Fix: split index and data file rotation logic
[lttng-tools.git] / src / bin / lttng-relayd / main.c
CommitLineData
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>
d3ecc550 57#include <common/align.h>
f40ef1d5 58#include <common/config/session-config.h>
5312a3ed
JG
59#include <common/dynamic-buffer.h>
60#include <common/buffer-view.h>
7591bab1 61#include <urcu/rculist.h>
b8aa1682 62
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"
f056029c 76#include "tcp_keep_alive.h"
b8aa1682 77
4fc83d94
PP
78static const char *help_msg =
79#ifdef LTTNG_EMBED_HELP
80#include <lttng-relayd.8.h>
81#else
82NULL
83#endif
84;
85
5569b118
JG
86enum relay_connection_status {
87 RELAY_CONNECTION_STATUS_OK,
88 /* An error occured while processing an event on the connection. */
89 RELAY_CONNECTION_STATUS_ERROR,
90 /* Connection closed/shutdown cleanly. */
91 RELAY_CONNECTION_STATUS_CLOSED,
92};
93
b8aa1682 94/* command line options */
0f907de1 95char *opt_output_path;
b5218ffb 96static int opt_daemon, opt_background;
3fd27398
MD
97
98/*
99 * We need to wait for listener and live listener threads, as well as
100 * health check thread, before being ready to signal readiness.
101 */
102#define NR_LTTNG_RELAY_READY 3
103static int lttng_relay_ready = NR_LTTNG_RELAY_READY;
0848dba7
MD
104
105/* Size of receive buffer. */
106#define RECV_DATA_BUFFER_SIZE 65536
d3ecc550 107#define FILE_COPY_BUFFER_SIZE 65536
0848dba7 108
3fd27398
MD
109static int recv_child_signal; /* Set to 1 when a SIGUSR1 signal is received. */
110static pid_t child_ppid; /* Internal parent PID use with daemonize. */
111
095a4ae5
MD
112static struct lttng_uri *control_uri;
113static struct lttng_uri *data_uri;
d3e2ba59 114static struct lttng_uri *live_uri;
b8aa1682
JD
115
116const char *progname;
b8aa1682 117
65931c8b 118const char *tracing_group_name = DEFAULT_TRACING_GROUP;
cd60b05a
JG
119static int tracing_group_name_override;
120
121const char * const config_section_name = "relayd";
65931c8b 122
b8aa1682
JD
123/*
124 * Quit pipe for all threads. This permits a single cancellation point
125 * for all threads when receiving an event on the pipe.
126 */
0b242f62 127int thread_quit_pipe[2] = { -1, -1 };
b8aa1682
JD
128
129/*
130 * This pipe is used to inform the worker thread that a command is queued and
131 * ready to be processed.
132 */
58eb9381 133static int relay_conn_pipe[2] = { -1, -1 };
b8aa1682 134
26c9d55e 135/* Shared between threads */
b8aa1682
JD
136static int dispatch_thread_exit;
137
138static pthread_t listener_thread;
139static pthread_t dispatcher_thread;
140static pthread_t worker_thread;
65931c8b 141static pthread_t health_thread;
b8aa1682 142
7591bab1
MD
143/*
144 * last_relay_stream_id_lock protects last_relay_stream_id increment
145 * atomicity on 32-bit architectures.
146 */
147static pthread_mutex_t last_relay_stream_id_lock = PTHREAD_MUTEX_INITIALIZER;
095a4ae5 148static uint64_t last_relay_stream_id;
b8aa1682
JD
149
150/*
151 * Relay command queue.
152 *
153 * The relay_thread_listener and relay_thread_dispatcher communicate with this
154 * queue.
155 */
58eb9381 156static struct relay_conn_queue relay_conn_queue;
b8aa1682 157
d3e2ba59
JD
158/* Global relay stream hash table. */
159struct lttng_ht *relay_streams_ht;
160
92c6ca54
DG
161/* Global relay viewer stream hash table. */
162struct lttng_ht *viewer_streams_ht;
163
7591bab1
MD
164/* Global relay sessions hash table. */
165struct lttng_ht *sessions_ht;
0a6518b0 166
55706a7d 167/* Relayd health monitoring */
eea7556c 168struct health_app *health_relayd;
55706a7d 169
cd60b05a
JG
170static struct option long_options[] = {
171 { "control-port", 1, 0, 'C', },
172 { "data-port", 1, 0, 'D', },
8d5c808e 173 { "live-port", 1, 0, 'L', },
cd60b05a 174 { "daemonize", 0, 0, 'd', },
b5218ffb 175 { "background", 0, 0, 'b', },
cd60b05a
JG
176 { "group", 1, 0, 'g', },
177 { "help", 0, 0, 'h', },
178 { "output", 1, 0, 'o', },
179 { "verbose", 0, 0, 'v', },
180 { "config", 1, 0, 'f' },
3a904098 181 { "version", 0, 0, 'V' },
cd60b05a
JG
182 { NULL, 0, 0, 0, },
183};
184
3a904098 185static const char *config_ignore_options[] = { "help", "config", "version" };
cd60b05a 186
cd60b05a
JG
187/*
188 * Take an option from the getopt output and set it in the right variable to be
189 * used later.
190 *
191 * Return 0 on success else a negative value.
192 */
7591bab1 193static int set_option(int opt, const char *arg, const char *optname)
b8aa1682 194{
cd60b05a
JG
195 int ret;
196
197 switch (opt) {
198 case 0:
199 fprintf(stderr, "option %s", optname);
200 if (arg) {
201 fprintf(stderr, " with arg %s\n", arg);
202 }
203 break;
204 case 'C':
e8fa9fb0
MD
205 if (lttng_is_setuid_setgid()) {
206 WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.",
207 "-C, --control-port");
208 } else {
209 ret = uri_parse(arg, &control_uri);
210 if (ret < 0) {
211 ERR("Invalid control URI specified");
212 goto end;
213 }
214 if (control_uri->port == 0) {
215 control_uri->port = DEFAULT_NETWORK_CONTROL_PORT;
216 }
cd60b05a
JG
217 }
218 break;
219 case 'D':
e8fa9fb0
MD
220 if (lttng_is_setuid_setgid()) {
221 WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.",
222 "-D, -data-port");
223 } else {
224 ret = uri_parse(arg, &data_uri);
225 if (ret < 0) {
226 ERR("Invalid data URI specified");
227 goto end;
228 }
229 if (data_uri->port == 0) {
230 data_uri->port = DEFAULT_NETWORK_DATA_PORT;
231 }
cd60b05a
JG
232 }
233 break;
8d5c808e 234 case 'L':
e8fa9fb0
MD
235 if (lttng_is_setuid_setgid()) {
236 WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.",
237 "-L, -live-port");
238 } else {
239 ret = uri_parse(arg, &live_uri);
240 if (ret < 0) {
241 ERR("Invalid live URI specified");
242 goto end;
243 }
244 if (live_uri->port == 0) {
245 live_uri->port = DEFAULT_NETWORK_VIEWER_PORT;
246 }
8d5c808e
AM
247 }
248 break;
cd60b05a
JG
249 case 'd':
250 opt_daemon = 1;
251 break;
b5218ffb
MD
252 case 'b':
253 opt_background = 1;
254 break;
cd60b05a 255 case 'g':
e8fa9fb0
MD
256 if (lttng_is_setuid_setgid()) {
257 WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.",
258 "-g, --group");
259 } else {
260 tracing_group_name = strdup(arg);
261 if (tracing_group_name == NULL) {
262 ret = -errno;
263 PERROR("strdup");
264 goto end;
265 }
266 tracing_group_name_override = 1;
330a40bb 267 }
cd60b05a
JG
268 break;
269 case 'h':
4fc83d94 270 ret = utils_show_help(8, "lttng-relayd", help_msg);
655b5cc1 271 if (ret) {
4fc83d94 272 ERR("Cannot show --help for `lttng-relayd`");
655b5cc1
PP
273 perror("exec");
274 }
cd60b05a 275 exit(EXIT_FAILURE);
3a904098
AW
276 case 'V':
277 fprintf(stdout, "%s\n", VERSION);
278 exit(EXIT_SUCCESS);
cd60b05a 279 case 'o':
e8fa9fb0
MD
280 if (lttng_is_setuid_setgid()) {
281 WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.",
282 "-o, --output");
283 } else {
284 ret = asprintf(&opt_output_path, "%s", arg);
285 if (ret < 0) {
286 ret = -errno;
287 PERROR("asprintf opt_output_path");
288 goto end;
289 }
cd60b05a
JG
290 }
291 break;
292 case 'v':
293 /* Verbose level can increase using multiple -v */
294 if (arg) {
295 lttng_opt_verbose = config_parse_value(arg);
296 } else {
849e5b7b
DG
297 /* Only 3 level of verbosity (-vvv). */
298 if (lttng_opt_verbose < 3) {
299 lttng_opt_verbose += 1;
300 }
cd60b05a
JG
301 }
302 break;
303 default:
304 /* Unknown option or other error.
305 * Error is printed by getopt, just return */
306 ret = -1;
307 goto end;
308 }
309
310 /* All good. */
311 ret = 0;
312
313end:
314 return ret;
315}
316
317/*
318 * config_entry_handler_cb used to handle options read from a config file.
f40ef1d5 319 * See config_entry_handler_cb comment in common/config/session-config.h for the
cd60b05a
JG
320 * return value conventions.
321 */
7591bab1 322static int config_entry_handler(const struct config_entry *entry, void *unused)
cd60b05a
JG
323{
324 int ret = 0, i;
325
326 if (!entry || !entry->name || !entry->value) {
327 ret = -EINVAL;
328 goto end;
329 }
330
331 /* Check if the option is to be ignored */
332 for (i = 0; i < sizeof(config_ignore_options) / sizeof(char *); i++) {
333 if (!strcmp(entry->name, config_ignore_options[i])) {
334 goto end;
335 }
336 }
337
338 for (i = 0; i < (sizeof(long_options) / sizeof(struct option)) - 1; i++) {
339 /* Ignore if entry name is not fully matched. */
340 if (strcmp(entry->name, long_options[i].name)) {
341 continue;
342 }
343
344 /*
7591bab1
MD
345 * If the option takes no argument on the command line,
346 * we have to check if the value is "true". We support
347 * non-zero numeric values, true, on and yes.
cd60b05a
JG
348 */
349 if (!long_options[i].has_arg) {
350 ret = config_parse_value(entry->value);
351 if (ret <= 0) {
352 if (ret) {
353 WARN("Invalid configuration value \"%s\" for option %s",
354 entry->value, entry->name);
355 }
356 /* False, skip boolean config option. */
357 goto end;
358 }
359 }
360
361 ret = set_option(long_options[i].val, entry->value, entry->name);
362 goto end;
363 }
364
365 WARN("Unrecognized option \"%s\" in daemon configuration file.",
366 entry->name);
367
368end:
369 return ret;
370}
371
7591bab1 372static int set_options(int argc, char **argv)
cd60b05a 373{
178a0557 374 int c, ret = 0, option_index = 0, retval = 0;
cd60b05a
JG
375 int orig_optopt = optopt, orig_optind = optind;
376 char *default_address, *optstring;
377 const char *config_path = NULL;
378
379 optstring = utils_generate_optstring(long_options,
380 sizeof(long_options) / sizeof(struct option));
381 if (!optstring) {
178a0557 382 retval = -ENOMEM;
cd60b05a
JG
383 goto exit;
384 }
385
386 /* Check for the --config option */
387
388 while ((c = getopt_long(argc, argv, optstring, long_options,
389 &option_index)) != -1) {
390 if (c == '?') {
178a0557 391 retval = -EINVAL;
cd60b05a
JG
392 goto exit;
393 } else if (c != 'f') {
394 continue;
395 }
396
e8fa9fb0
MD
397 if (lttng_is_setuid_setgid()) {
398 WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.",
399 "-f, --config");
400 } else {
401 config_path = utils_expand_path(optarg);
402 if (!config_path) {
403 ERR("Failed to resolve path: %s", optarg);
404 }
cd60b05a
JG
405 }
406 }
407
408 ret = config_get_section_entries(config_path, config_section_name,
409 config_entry_handler, NULL);
410 if (ret) {
411 if (ret > 0) {
412 ERR("Invalid configuration option at line %i", ret);
cd60b05a 413 }
178a0557 414 retval = -1;
cd60b05a
JG
415 goto exit;
416 }
b8aa1682 417
cd60b05a
JG
418 /* Reset getopt's global state */
419 optopt = orig_optopt;
420 optind = orig_optind;
b8aa1682 421 while (1) {
cd60b05a 422 c = getopt_long(argc, argv, optstring, long_options, &option_index);
b8aa1682
JD
423 if (c == -1) {
424 break;
425 }
426
cd60b05a
JG
427 ret = set_option(c, optarg, long_options[option_index].name);
428 if (ret < 0) {
178a0557 429 retval = -1;
b8aa1682
JD
430 goto exit;
431 }
432 }
433
434 /* assign default values */
435 if (control_uri == NULL) {
fa91dc52
MD
436 ret = asprintf(&default_address,
437 "tcp://" DEFAULT_NETWORK_CONTROL_BIND_ADDRESS ":%d",
438 DEFAULT_NETWORK_CONTROL_PORT);
b8aa1682
JD
439 if (ret < 0) {
440 PERROR("asprintf default data address");
178a0557 441 retval = -1;
b8aa1682
JD
442 goto exit;
443 }
444
445 ret = uri_parse(default_address, &control_uri);
446 free(default_address);
447 if (ret < 0) {
448 ERR("Invalid control URI specified");
178a0557 449 retval = -1;
b8aa1682
JD
450 goto exit;
451 }
452 }
453 if (data_uri == NULL) {
fa91dc52
MD
454 ret = asprintf(&default_address,
455 "tcp://" DEFAULT_NETWORK_DATA_BIND_ADDRESS ":%d",
456 DEFAULT_NETWORK_DATA_PORT);
b8aa1682
JD
457 if (ret < 0) {
458 PERROR("asprintf default data address");
178a0557 459 retval = -1;
b8aa1682
JD
460 goto exit;
461 }
462
463 ret = uri_parse(default_address, &data_uri);
464 free(default_address);
465 if (ret < 0) {
466 ERR("Invalid data URI specified");
178a0557 467 retval = -1;
b8aa1682
JD
468 goto exit;
469 }
470 }
d3e2ba59 471 if (live_uri == NULL) {
fa91dc52
MD
472 ret = asprintf(&default_address,
473 "tcp://" DEFAULT_NETWORK_VIEWER_BIND_ADDRESS ":%d",
474 DEFAULT_NETWORK_VIEWER_PORT);
d3e2ba59
JD
475 if (ret < 0) {
476 PERROR("asprintf default viewer control address");
178a0557 477 retval = -1;
d3e2ba59
JD
478 goto exit;
479 }
480
481 ret = uri_parse(default_address, &live_uri);
482 free(default_address);
483 if (ret < 0) {
484 ERR("Invalid viewer control URI specified");
178a0557 485 retval = -1;
d3e2ba59
JD
486 goto exit;
487 }
488 }
b8aa1682
JD
489
490exit:
cd60b05a 491 free(optstring);
178a0557 492 return retval;
b8aa1682
JD
493}
494
7591bab1
MD
495static void print_global_objects(void)
496{
497 rcu_register_thread();
498
499 print_viewer_streams();
500 print_relay_streams();
501 print_sessions();
502
503 rcu_unregister_thread();
504}
505
b8aa1682
JD
506/*
507 * Cleanup the daemon
508 */
7591bab1 509static void relayd_cleanup(void)
b8aa1682 510{
7591bab1
MD
511 print_global_objects();
512
b8aa1682
JD
513 DBG("Cleaning up");
514
178a0557
MD
515 if (viewer_streams_ht)
516 lttng_ht_destroy(viewer_streams_ht);
517 if (relay_streams_ht)
518 lttng_ht_destroy(relay_streams_ht);
7591bab1
MD
519 if (sessions_ht)
520 lttng_ht_destroy(sessions_ht);
178a0557 521
095a4ae5
MD
522 /* free the dynamically allocated opt_output_path */
523 free(opt_output_path);
524
a02de639
CB
525 /* Close thread quit pipes */
526 utils_close_pipe(thread_quit_pipe);
527
710c1f73
DG
528 uri_free(control_uri);
529 uri_free(data_uri);
8d5c808e 530 /* Live URI is freed in the live thread. */
cd60b05a
JG
531
532 if (tracing_group_name_override) {
533 free((void *) tracing_group_name);
534 }
b8aa1682
JD
535}
536
537/*
538 * Write to writable pipe used to notify a thread.
539 */
7591bab1 540static int notify_thread_pipe(int wpipe)
b8aa1682 541{
6cd525e8 542 ssize_t ret;
b8aa1682 543
6cd525e8
MD
544 ret = lttng_write(wpipe, "!", 1);
545 if (ret < 1) {
b8aa1682 546 PERROR("write poll pipe");
b4aacfdc 547 goto end;
b8aa1682 548 }
b4aacfdc
MD
549 ret = 0;
550end:
b8aa1682
JD
551 return ret;
552}
553
7591bab1 554static int notify_health_quit_pipe(int *pipe)
65931c8b 555{
6cd525e8 556 ssize_t ret;
65931c8b 557
6cd525e8
MD
558 ret = lttng_write(pipe[1], "4", 1);
559 if (ret < 1) {
65931c8b 560 PERROR("write relay health quit");
b4aacfdc 561 goto end;
65931c8b 562 }
b4aacfdc
MD
563 ret = 0;
564end:
565 return ret;
65931c8b
MD
566}
567
b8aa1682 568/*
b4aacfdc 569 * Stop all relayd and relayd-live threads.
b8aa1682 570 */
b4aacfdc 571int lttng_relay_stop_threads(void)
b8aa1682 572{
b4aacfdc 573 int retval = 0;
b8aa1682
JD
574
575 /* Stopping all threads */
576 DBG("Terminating all threads");
b4aacfdc 577 if (notify_thread_pipe(thread_quit_pipe[1])) {
b8aa1682 578 ERR("write error on thread quit pipe");
b4aacfdc 579 retval = -1;
b8aa1682
JD
580 }
581
b4aacfdc
MD
582 if (notify_health_quit_pipe(health_quit_pipe)) {
583 ERR("write error on health quit pipe");
584 }
65931c8b 585
b8aa1682 586 /* Dispatch thread */
26c9d55e 587 CMM_STORE_SHARED(dispatch_thread_exit, 1);
58eb9381 588 futex_nto1_wake(&relay_conn_queue.futex);
178a0557 589
b4aacfdc 590 if (relayd_live_stop()) {
178a0557 591 ERR("Error stopping live threads");
b4aacfdc 592 retval = -1;
178a0557 593 }
b4aacfdc 594 return retval;
b8aa1682
JD
595}
596
597/*
598 * Signal handler for the daemon
599 *
600 * Simply stop all worker threads, leaving main() return gracefully after
601 * joining all threads and calling cleanup().
602 */
7591bab1 603static void sighandler(int sig)
b8aa1682
JD
604{
605 switch (sig) {
b8aa1682
JD
606 case SIGINT:
607 DBG("SIGINT caught");
b4aacfdc
MD
608 if (lttng_relay_stop_threads()) {
609 ERR("Error stopping threads");
610 }
b8aa1682
JD
611 break;
612 case SIGTERM:
613 DBG("SIGTERM caught");
b4aacfdc
MD
614 if (lttng_relay_stop_threads()) {
615 ERR("Error stopping threads");
616 }
b8aa1682 617 break;
3fd27398
MD
618 case SIGUSR1:
619 CMM_STORE_SHARED(recv_child_signal, 1);
620 break;
b8aa1682
JD
621 default:
622 break;
623 }
624}
625
626/*
627 * Setup signal handler for :
628 * SIGINT, SIGTERM, SIGPIPE
629 */
7591bab1 630static int set_signal_handler(void)
b8aa1682
JD
631{
632 int ret = 0;
633 struct sigaction sa;
634 sigset_t sigset;
635
636 if ((ret = sigemptyset(&sigset)) < 0) {
637 PERROR("sigemptyset");
638 return ret;
639 }
640
b8aa1682
JD
641 sa.sa_mask = sigset;
642 sa.sa_flags = 0;
0072e5e2
MD
643
644 sa.sa_handler = sighandler;
b8aa1682
JD
645 if ((ret = sigaction(SIGTERM, &sa, NULL)) < 0) {
646 PERROR("sigaction");
647 return ret;
648 }
649
650 if ((ret = sigaction(SIGINT, &sa, NULL)) < 0) {
651 PERROR("sigaction");
652 return ret;
653 }
654
0072e5e2 655 if ((ret = sigaction(SIGUSR1, &sa, NULL)) < 0) {
b8aa1682
JD
656 PERROR("sigaction");
657 return ret;
658 }
659
0072e5e2
MD
660 sa.sa_handler = SIG_IGN;
661 if ((ret = sigaction(SIGPIPE, &sa, NULL)) < 0) {
3fd27398
MD
662 PERROR("sigaction");
663 return ret;
664 }
665
666 DBG("Signal handler set for SIGTERM, SIGUSR1, SIGPIPE and SIGINT");
b8aa1682
JD
667
668 return ret;
669}
670
3fd27398
MD
671void lttng_relay_notify_ready(void)
672{
673 /* Notify the parent of the fork() process that we are ready. */
674 if (opt_daemon || opt_background) {
675 if (uatomic_sub_return(&lttng_relay_ready, 1) == 0) {
676 kill(child_ppid, SIGUSR1);
677 }
678 }
679}
680
b8aa1682
JD
681/*
682 * Init thread quit pipe.
683 *
684 * Return -1 on error or 0 if all pipes are created.
685 */
7591bab1 686static int init_thread_quit_pipe(void)
b8aa1682 687{
a02de639 688 int ret;
b8aa1682 689
a02de639 690 ret = utils_create_pipe_cloexec(thread_quit_pipe);
b8aa1682 691
b8aa1682
JD
692 return ret;
693}
694
695/*
696 * Create a poll set with O_CLOEXEC and add the thread quit pipe to the set.
697 */
7591bab1 698static int create_thread_poll_set(struct lttng_poll_event *events, int size)
b8aa1682
JD
699{
700 int ret;
701
702 if (events == NULL || size == 0) {
703 ret = -1;
704 goto error;
705 }
706
707 ret = lttng_poll_create(events, size, LTTNG_CLOEXEC);
708 if (ret < 0) {
709 goto error;
710 }
711
712 /* Add quit pipe */
c7759e6a 713 ret = lttng_poll_add(events, thread_quit_pipe[0], LPOLLIN | LPOLLERR);
b8aa1682
JD
714 if (ret < 0) {
715 goto error;
716 }
717
718 return 0;
719
720error:
721 return ret;
722}
723
724/*
725 * Check if the thread quit pipe was triggered.
726 *
727 * Return 1 if it was triggered else 0;
728 */
7591bab1 729static int check_thread_quit_pipe(int fd, uint32_t events)
b8aa1682
JD
730{
731 if (fd == thread_quit_pipe[0] && (events & LPOLLIN)) {
732 return 1;
733 }
734
735 return 0;
736}
737
738/*
739 * Create and init socket from uri.
740 */
7591bab1 741static struct lttcomm_sock *relay_socket_create(struct lttng_uri *uri)
b8aa1682
JD
742{
743 int ret;
744 struct lttcomm_sock *sock = NULL;
745
746 sock = lttcomm_alloc_sock_from_uri(uri);
747 if (sock == NULL) {
748 ERR("Allocating socket");
749 goto error;
750 }
751
752 ret = lttcomm_create_sock(sock);
753 if (ret < 0) {
754 goto error;
755 }
756 DBG("Listening on sock %d", sock->fd);
757
758 ret = sock->ops->bind(sock);
759 if (ret < 0) {
2288467f 760 PERROR("Failed to bind socket");
b8aa1682
JD
761 goto error;
762 }
763
764 ret = sock->ops->listen(sock, -1);
765 if (ret < 0) {
766 goto error;
767
768 }
769
770 return sock;
771
772error:
773 if (sock) {
774 lttcomm_destroy_sock(sock);
775 }
776 return NULL;
777}
778
779/*
780 * This thread manages the listening for new connections on the network
781 */
7591bab1 782static void *relay_thread_listener(void *data)
b8aa1682 783{
095a4ae5 784 int i, ret, pollfd, err = -1;
b8aa1682
JD
785 uint32_t revents, nb_fd;
786 struct lttng_poll_event events;
787 struct lttcomm_sock *control_sock, *data_sock;
788
b8aa1682
JD
789 DBG("[thread] Relay listener started");
790
55706a7d
MD
791 health_register(health_relayd, HEALTH_RELAYD_TYPE_LISTENER);
792
f385ae0a
MD
793 health_code_update();
794
7591bab1 795 control_sock = relay_socket_create(control_uri);
b8aa1682 796 if (!control_sock) {
095a4ae5 797 goto error_sock_control;
b8aa1682
JD
798 }
799
7591bab1 800 data_sock = relay_socket_create(data_uri);
b8aa1682 801 if (!data_sock) {
095a4ae5 802 goto error_sock_relay;
b8aa1682
JD
803 }
804
805 /*
7591bab1
MD
806 * Pass 3 as size here for the thread quit pipe, control and
807 * data socket.
b8aa1682
JD
808 */
809 ret = create_thread_poll_set(&events, 3);
810 if (ret < 0) {
811 goto error_create_poll;
812 }
813
814 /* Add the control socket */
815 ret = lttng_poll_add(&events, control_sock->fd, LPOLLIN | LPOLLRDHUP);
816 if (ret < 0) {
817 goto error_poll_add;
818 }
819
820 /* Add the data socket */
821 ret = lttng_poll_add(&events, data_sock->fd, LPOLLIN | LPOLLRDHUP);
822 if (ret < 0) {
823 goto error_poll_add;
824 }
825
3fd27398
MD
826 lttng_relay_notify_ready();
827
9b5e0863
MD
828 if (testpoint(relayd_thread_listener)) {
829 goto error_testpoint;
830 }
831
b8aa1682 832 while (1) {
f385ae0a
MD
833 health_code_update();
834
b8aa1682
JD
835 DBG("Listener accepting connections");
836
b8aa1682 837restart:
f385ae0a 838 health_poll_entry();
b8aa1682 839 ret = lttng_poll_wait(&events, -1);
f385ae0a 840 health_poll_exit();
b8aa1682
JD
841 if (ret < 0) {
842 /*
843 * Restart interrupted system call.
844 */
845 if (errno == EINTR) {
846 goto restart;
847 }
848 goto error;
849 }
850
0d9c5d77
DG
851 nb_fd = ret;
852
b8aa1682
JD
853 DBG("Relay new connection received");
854 for (i = 0; i < nb_fd; i++) {
f385ae0a
MD
855 health_code_update();
856
b8aa1682
JD
857 /* Fetch once the poll data */
858 revents = LTTNG_POLL_GETEV(&events, i);
859 pollfd = LTTNG_POLL_GETFD(&events, i);
860
fd20dac9 861 if (!revents) {
7591bab1
MD
862 /*
863 * No activity for this FD (poll
864 * implementation).
865 */
fd20dac9
MD
866 continue;
867 }
868
b8aa1682
JD
869 /* Thread quit pipe has been closed. Killing thread. */
870 ret = check_thread_quit_pipe(pollfd, revents);
871 if (ret) {
095a4ae5
MD
872 err = 0;
873 goto exit;
b8aa1682
JD
874 }
875
03e43155 876 if (revents & LPOLLIN) {
4b7f17b2 877 /*
7591bab1
MD
878 * A new connection is requested, therefore a
879 * sessiond/consumerd connection is allocated in
880 * this thread, enqueued to a global queue and
881 * dequeued (and freed) in the worker thread.
4b7f17b2 882 */
58eb9381
DG
883 int val = 1;
884 struct relay_connection *new_conn;
4b7f17b2 885 struct lttcomm_sock *newsock;
7591bab1 886 enum connection_type type;
b8aa1682
JD
887
888 if (pollfd == data_sock->fd) {
7591bab1 889 type = RELAY_DATA;
b8aa1682 890 newsock = data_sock->ops->accept(data_sock);
58eb9381
DG
891 DBG("Relay data connection accepted, socket %d",
892 newsock->fd);
4b7f17b2
MD
893 } else {
894 assert(pollfd == control_sock->fd);
7591bab1 895 type = RELAY_CONTROL;
b8aa1682 896 newsock = control_sock->ops->accept(control_sock);
58eb9381
DG
897 DBG("Relay control connection accepted, socket %d",
898 newsock->fd);
b8aa1682 899 }
58eb9381
DG
900 if (!newsock) {
901 PERROR("accepting sock");
58eb9381
DG
902 goto error;
903 }
904
905 ret = setsockopt(newsock->fd, SOL_SOCKET, SO_REUSEADDR, &val,
906 sizeof(val));
b8aa1682
JD
907 if (ret < 0) {
908 PERROR("setsockopt inet");
4b7f17b2 909 lttcomm_destroy_sock(newsock);
b8aa1682
JD
910 goto error;
911 }
f056029c
JR
912
913 ret = socket_apply_keep_alive_config(newsock->fd);
914 if (ret < 0) {
915 ERR("Failed to apply TCP keep-alive configuration on socket (%i)",
916 newsock->fd);
917 lttcomm_destroy_sock(newsock);
918 goto error;
919 }
920
7591bab1
MD
921 new_conn = connection_create(newsock, type);
922 if (!new_conn) {
923 lttcomm_destroy_sock(newsock);
924 goto error;
925 }
58eb9381
DG
926
927 /* Enqueue request for the dispatcher thread. */
8bdee6e2
SM
928 cds_wfcq_enqueue(&relay_conn_queue.head, &relay_conn_queue.tail,
929 &new_conn->qnode);
b8aa1682
JD
930
931 /*
7591bab1
MD
932 * Wake the dispatch queue futex.
933 * Implicit memory barrier with the
934 * exchange in cds_wfcq_enqueue.
b8aa1682 935 */
58eb9381 936 futex_nto1_wake(&relay_conn_queue.futex);
03e43155
MD
937 } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
938 ERR("socket poll error");
939 goto error;
940 } else {
941 ERR("Unexpected poll events %u for sock %d", revents, pollfd);
942 goto error;
b8aa1682
JD
943 }
944 }
945 }
946
095a4ae5 947exit:
b8aa1682
JD
948error:
949error_poll_add:
9b5e0863 950error_testpoint:
b8aa1682
JD
951 lttng_poll_clean(&events);
952error_create_poll:
095a4ae5
MD
953 if (data_sock->fd >= 0) {
954 ret = data_sock->ops->close(data_sock);
b8aa1682
JD
955 if (ret) {
956 PERROR("close");
957 }
b8aa1682 958 }
095a4ae5
MD
959 lttcomm_destroy_sock(data_sock);
960error_sock_relay:
961 if (control_sock->fd >= 0) {
962 ret = control_sock->ops->close(control_sock);
b8aa1682
JD
963 if (ret) {
964 PERROR("close");
965 }
b8aa1682 966 }
095a4ae5
MD
967 lttcomm_destroy_sock(control_sock);
968error_sock_control:
969 if (err) {
f385ae0a
MD
970 health_error();
971 ERR("Health error occurred in %s", __func__);
095a4ae5 972 }
55706a7d 973 health_unregister(health_relayd);
b8aa1682 974 DBG("Relay listener thread cleanup complete");
b4aacfdc 975 lttng_relay_stop_threads();
b8aa1682
JD
976 return NULL;
977}
978
979/*
980 * This thread manages the dispatching of the requests to worker threads
981 */
7591bab1 982static void *relay_thread_dispatcher(void *data)
b8aa1682 983{
6cd525e8
MD
984 int err = -1;
985 ssize_t ret;
8bdee6e2 986 struct cds_wfcq_node *node;
58eb9381 987 struct relay_connection *new_conn = NULL;
b8aa1682
JD
988
989 DBG("[thread] Relay dispatcher started");
990
55706a7d
MD
991 health_register(health_relayd, HEALTH_RELAYD_TYPE_DISPATCHER);
992
9b5e0863
MD
993 if (testpoint(relayd_thread_dispatcher)) {
994 goto error_testpoint;
995 }
996
f385ae0a
MD
997 health_code_update();
998
0ed3b1a8 999 for (;;) {
f385ae0a
MD
1000 health_code_update();
1001
b8aa1682 1002 /* Atomically prepare the queue futex */
58eb9381 1003 futex_nto1_prepare(&relay_conn_queue.futex);
b8aa1682 1004
0ed3b1a8
MD
1005 if (CMM_LOAD_SHARED(dispatch_thread_exit)) {
1006 break;
1007 }
1008
b8aa1682 1009 do {
f385ae0a
MD
1010 health_code_update();
1011
b8aa1682 1012 /* Dequeue commands */
8bdee6e2
SM
1013 node = cds_wfcq_dequeue_blocking(&relay_conn_queue.head,
1014 &relay_conn_queue.tail);
b8aa1682
JD
1015 if (node == NULL) {
1016 DBG("Woken up but nothing in the relay command queue");
1017 /* Continue thread execution */
1018 break;
1019 }
58eb9381 1020 new_conn = caa_container_of(node, struct relay_connection, qnode);
b8aa1682 1021
58eb9381 1022 DBG("Dispatching request waiting on sock %d", new_conn->sock->fd);
b8aa1682
JD
1023
1024 /*
7591bab1
MD
1025 * Inform worker thread of the new request. This
1026 * call is blocking so we can be assured that
1027 * the data will be read at some point in time
1028 * or wait to the end of the world :)
b8aa1682 1029 */
58eb9381
DG
1030 ret = lttng_write(relay_conn_pipe[1], &new_conn, sizeof(new_conn));
1031 if (ret < 0) {
1032 PERROR("write connection pipe");
7591bab1 1033 connection_put(new_conn);
b8aa1682
JD
1034 goto error;
1035 }
1036 } while (node != NULL);
1037
1038 /* Futex wait on queue. Blocking call on futex() */
f385ae0a 1039 health_poll_entry();
58eb9381 1040 futex_nto1_wait(&relay_conn_queue.futex);
f385ae0a 1041 health_poll_exit();
b8aa1682
JD
1042 }
1043
f385ae0a
MD
1044 /* Normal exit, no error */
1045 err = 0;
1046
b8aa1682 1047error:
9b5e0863 1048error_testpoint:
f385ae0a
MD
1049 if (err) {
1050 health_error();
1051 ERR("Health error occurred in %s", __func__);
1052 }
55706a7d 1053 health_unregister(health_relayd);
b8aa1682 1054 DBG("Dispatch thread dying");
b4aacfdc 1055 lttng_relay_stop_threads();
b8aa1682
JD
1056 return NULL;
1057}
1058
b8aa1682 1059/*
7591bab1 1060 * Set index data from the control port to a given index object.
b8aa1682 1061 */
7591bab1 1062static int set_index_control_data(struct relay_index *index,
234cd636
JD
1063 struct lttcomm_relayd_index *data,
1064 struct relay_connection *conn)
1c20f0e2 1065{
7591bab1 1066 struct ctf_packet_index index_data;
1c20f0e2
JD
1067
1068 /*
5312a3ed 1069 * The index on disk is encoded in big endian.
1c20f0e2 1070 */
5312a3ed
JG
1071 index_data.packet_size = htobe64(data->packet_size);
1072 index_data.content_size = htobe64(data->content_size);
1073 index_data.timestamp_begin = htobe64(data->timestamp_begin);
1074 index_data.timestamp_end = htobe64(data->timestamp_end);
1075 index_data.events_discarded = htobe64(data->events_discarded);
1076 index_data.stream_id = htobe64(data->stream_id);
234cd636
JD
1077
1078 if (conn->minor >= 8) {
5312a3ed
JG
1079 index->index_data.stream_instance_id = htobe64(data->stream_instance_id);
1080 index->index_data.packet_seq_num = htobe64(data->packet_seq_num);
234cd636
JD
1081 }
1082
7591bab1 1083 return relay_index_set_data(index, &index_data);
1c20f0e2
JD
1084}
1085
298a25ca
JG
1086static bool session_streams_have_index(const struct relay_session *session)
1087{
1088 return session->minor >= 4 && !session->snapshot;
1089}
1090
c5b6f4f0
DG
1091/*
1092 * Handle the RELAYD_CREATE_SESSION command.
1093 *
1094 * On success, send back the session id or else return a negative value.
1095 */
5312a3ed
JG
1096static int relay_create_session(const struct lttcomm_relayd_hdr *recv_hdr,
1097 struct relay_connection *conn,
1098 const struct lttng_buffer_view *payload)
c5b6f4f0 1099{
5312a3ed
JG
1100 int ret = 0;
1101 ssize_t send_ret;
c5b6f4f0
DG
1102 struct relay_session *session;
1103 struct lttcomm_relayd_status_session reply;
36d2e35d 1104 char session_name[LTTNG_NAME_MAX];
9cf3eb20 1105 char hostname[LTTNG_HOST_NAME_MAX];
7591bab1
MD
1106 uint32_t live_timer = 0;
1107 bool snapshot = false;
c5b6f4f0 1108
36d2e35d 1109 memset(session_name, 0, LTTNG_NAME_MAX);
9cf3eb20 1110 memset(hostname, 0, LTTNG_HOST_NAME_MAX);
c5b6f4f0
DG
1111
1112 memset(&reply, 0, sizeof(reply));
1113
f86f6389
JR
1114 if (conn->minor < 4) {
1115 /* From 2.1 to 2.3 */
1116 ret = 0;
1117 } else if (conn->minor >= 4 && conn->minor < 11) {
1118 /* From 2.4 to 2.10 */
5312a3ed 1119 ret = cmd_create_session_2_4(payload, session_name,
7591bab1 1120 hostname, &live_timer, &snapshot);
f86f6389
JR
1121 } else {
1122 /* From 2.11 to ... */
1123 ret = cmd_create_session_2_11(payload, session_name,
1124 hostname, &live_timer, &snapshot);
7591bab1 1125 }
f86f6389 1126
7591bab1
MD
1127 if (ret < 0) {
1128 goto send_reply;
d3e2ba59
JD
1129 }
1130
7591bab1
MD
1131 session = session_create(session_name, hostname, live_timer,
1132 snapshot, conn->major, conn->minor);
1133 if (!session) {
1134 ret = -1;
1135 goto send_reply;
1136 }
1137 assert(!conn->session);
1138 conn->session = session;
c5b6f4f0
DG
1139 DBG("Created session %" PRIu64, session->id);
1140
7591bab1
MD
1141 reply.session_id = htobe64(session->id);
1142
1143send_reply:
c5b6f4f0
DG
1144 if (ret < 0) {
1145 reply.ret_code = htobe32(LTTNG_ERR_FATAL);
1146 } else {
1147 reply.ret_code = htobe32(LTTNG_OK);
1148 }
1149
58eb9381 1150 send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0);
5312a3ed
JG
1151 if (send_ret < (ssize_t) sizeof(reply)) {
1152 ERR("Failed to send \"create session\" command reply (ret = %zd)",
1153 send_ret);
1154 ret = -1;
c5b6f4f0
DG
1155 }
1156
1157 return ret;
1158}
1159
a4baae1b
JD
1160/*
1161 * When we have received all the streams and the metadata for a channel,
1162 * we make them visible to the viewer threads.
1163 */
7591bab1 1164static void publish_connection_local_streams(struct relay_connection *conn)
a4baae1b 1165{
7591bab1
MD
1166 struct relay_stream *stream;
1167 struct relay_session *session = conn->session;
a4baae1b 1168
7591bab1
MD
1169 /*
1170 * We publish all streams belonging to a session atomically wrt
1171 * session lock.
1172 */
1173 pthread_mutex_lock(&session->lock);
1174 rcu_read_lock();
1175 cds_list_for_each_entry_rcu(stream, &session->recv_list,
1176 recv_node) {
1177 stream_publish(stream);
a4baae1b 1178 }
7591bab1 1179 rcu_read_unlock();
a4baae1b 1180
7591bab1
MD
1181 /*
1182 * Inform the viewer that there are new streams in the session.
1183 */
1184 if (session->viewer_attached) {
1185 uatomic_set(&session->new_streams, 1);
1186 }
1187 pthread_mutex_unlock(&session->lock);
a4baae1b
JD
1188}
1189
b8aa1682
JD
1190/*
1191 * relay_add_stream: allocate a new stream for a session
1192 */
5312a3ed
JG
1193static int relay_add_stream(const struct lttcomm_relayd_hdr *recv_hdr,
1194 struct relay_connection *conn,
1195 const struct lttng_buffer_view *payload)
b8aa1682 1196{
7591bab1
MD
1197 int ret;
1198 ssize_t send_ret;
58eb9381 1199 struct relay_session *session = conn->session;
b8aa1682
JD
1200 struct relay_stream *stream = NULL;
1201 struct lttcomm_relayd_status_stream reply;
4030a636 1202 struct ctf_trace *trace = NULL;
7591bab1
MD
1203 uint64_t stream_handle = -1ULL;
1204 char *path_name = NULL, *channel_name = NULL;
1205 uint64_t tracefile_size = 0, tracefile_count = 0;
81164b6b 1206 struct relay_stream_chunk_id stream_chunk_id = { 0 };
b8aa1682 1207
5312a3ed 1208 if (!session || !conn->version_check_done) {
b8aa1682
JD
1209 ERR("Trying to add a stream before version check");
1210 ret = -1;
1211 goto end_no_session;
1212 }
1213
2f21a469
JR
1214 if (session->minor == 1) {
1215 /* For 2.1 */
5312a3ed 1216 ret = cmd_recv_stream_2_1(payload, &path_name,
7591bab1 1217 &channel_name);
2f21a469
JR
1218 } else if (session->minor > 1 && session->minor < 11) {
1219 /* From 2.2 to 2.10 */
5312a3ed 1220 ret = cmd_recv_stream_2_2(payload, &path_name,
7591bab1 1221 &channel_name, &tracefile_size, &tracefile_count);
2f21a469
JR
1222 } else {
1223 /* From 2.11 to ... */
1224 ret = cmd_recv_stream_2_11(payload, &path_name,
0b50e4b3
JG
1225 &channel_name, &tracefile_size, &tracefile_count,
1226 &stream_chunk_id.value);
1227 stream_chunk_id.is_set = true;
0f907de1 1228 }
2f21a469 1229
0f907de1 1230 if (ret < 0) {
7591bab1 1231 goto send_reply;
b8aa1682
JD
1232 }
1233
7591bab1 1234 trace = ctf_trace_get_by_path_or_create(session, path_name);
2a174661 1235 if (!trace) {
7591bab1 1236 goto send_reply;
2a174661 1237 }
7591bab1 1238 /* This stream here has one reference on the trace. */
2a174661 1239
7591bab1
MD
1240 pthread_mutex_lock(&last_relay_stream_id_lock);
1241 stream_handle = ++last_relay_stream_id;
1242 pthread_mutex_unlock(&last_relay_stream_id_lock);
d3e2ba59 1243
7591bab1
MD
1244 /* We pass ownership of path_name and channel_name. */
1245 stream = stream_create(trace, stream_handle, path_name,
81164b6b
JG
1246 channel_name, tracefile_size, tracefile_count,
1247 &stream_chunk_id);
7591bab1
MD
1248 path_name = NULL;
1249 channel_name = NULL;
a4baae1b 1250
2a174661 1251 /*
7591bab1
MD
1252 * Streams are the owners of their trace. Reference to trace is
1253 * kept within stream_create().
2a174661 1254 */
7591bab1 1255 ctf_trace_put(trace);
d3e2ba59 1256
7591bab1 1257send_reply:
53efb85a 1258 memset(&reply, 0, sizeof(reply));
7591bab1
MD
1259 reply.handle = htobe64(stream_handle);
1260 if (!stream) {
f73fabfd 1261 reply.ret_code = htobe32(LTTNG_ERR_UNK);
b8aa1682 1262 } else {
f73fabfd 1263 reply.ret_code = htobe32(LTTNG_OK);
b8aa1682 1264 }
5af40280 1265
58eb9381 1266 send_ret = conn->sock->ops->sendmsg(conn->sock, &reply,
b8aa1682 1267 sizeof(struct lttcomm_relayd_status_stream), 0);
5312a3ed
JG
1268 if (send_ret < (ssize_t) sizeof(reply)) {
1269 ERR("Failed to send \"add stream\" command reply (ret = %zd)",
1270 send_ret);
1271 ret = -1;
b8aa1682
JD
1272 }
1273
1274end_no_session:
7591bab1
MD
1275 free(path_name);
1276 free(channel_name);
0f907de1 1277 return ret;
b8aa1682
JD
1278}
1279
173af62f
DG
1280/*
1281 * relay_close_stream: close a specific stream
1282 */
5312a3ed
JG
1283static int relay_close_stream(const struct lttcomm_relayd_hdr *recv_hdr,
1284 struct relay_connection *conn,
1285 const struct lttng_buffer_view *payload)
173af62f 1286{
5312a3ed
JG
1287 int ret;
1288 ssize_t send_ret;
58eb9381 1289 struct relay_session *session = conn->session;
173af62f
DG
1290 struct lttcomm_relayd_close_stream stream_info;
1291 struct lttcomm_relayd_generic_reply reply;
1292 struct relay_stream *stream;
173af62f
DG
1293
1294 DBG("Close stream received");
1295
5312a3ed 1296 if (!session || !conn->version_check_done) {
173af62f
DG
1297 ERR("Trying to close a stream before version check");
1298 ret = -1;
1299 goto end_no_session;
1300 }
1301
5312a3ed
JG
1302 if (payload->size < sizeof(stream_info)) {
1303 ERR("Unexpected payload size in \"relay_close_stream\": expected >= %zu bytes, got %zu bytes",
1304 sizeof(stream_info), payload->size);
173af62f
DG
1305 ret = -1;
1306 goto end_no_session;
1307 }
5312a3ed
JG
1308 memcpy(&stream_info, payload->data, sizeof(stream_info));
1309 stream_info.stream_id = be64toh(stream_info.stream_id);
1310 stream_info.last_net_seq_num = be64toh(stream_info.last_net_seq_num);
173af62f 1311
5312a3ed 1312 stream = stream_get_by_id(stream_info.stream_id);
173af62f
DG
1313 if (!stream) {
1314 ret = -1;
7591bab1 1315 goto end;
173af62f 1316 }
77f7bd85
MD
1317
1318 /*
1319 * Set last_net_seq_num before the close flag. Required by data
1320 * pending check.
1321 */
7591bab1 1322 pthread_mutex_lock(&stream->lock);
5312a3ed 1323 stream->last_net_seq_num = stream_info.last_net_seq_num;
77f7bd85
MD
1324 pthread_mutex_unlock(&stream->lock);
1325
bda7c7b9
JG
1326 /*
1327 * This is one of the conditions which may trigger a stream close
1328 * with the others being:
1329 * 1) A close command is received for a stream
1330 * 2) The control connection owning the stream is closed
1331 * 3) We have received all of the stream's data _after_ a close
1332 * request.
1333 */
1334 try_stream_close(stream);
7591bab1
MD
1335 if (stream->is_metadata) {
1336 struct relay_viewer_stream *vstream;
173af62f 1337
7591bab1
MD
1338 vstream = viewer_stream_get_by_id(stream->stream_handle);
1339 if (vstream) {
1340 if (vstream->metadata_sent == stream->metadata_received) {
1341 /*
1342 * Since all the metadata has been sent to the
1343 * viewer and that we have a request to close
1344 * its stream, we can safely teardown the
1345 * corresponding metadata viewer stream.
1346 */
1347 viewer_stream_put(vstream);
1348 }
1349 /* Put local reference. */
1350 viewer_stream_put(vstream);
1351 }
1352 }
7591bab1 1353 stream_put(stream);
5312a3ed 1354 ret = 0;
173af62f 1355
7591bab1 1356end:
53efb85a 1357 memset(&reply, 0, sizeof(reply));
173af62f 1358 if (ret < 0) {
f73fabfd 1359 reply.ret_code = htobe32(LTTNG_ERR_UNK);
173af62f 1360 } else {
f73fabfd 1361 reply.ret_code = htobe32(LTTNG_OK);
173af62f 1362 }
58eb9381 1363 send_ret = conn->sock->ops->sendmsg(conn->sock, &reply,
173af62f 1364 sizeof(struct lttcomm_relayd_generic_reply), 0);
5312a3ed
JG
1365 if (send_ret < (ssize_t) sizeof(reply)) {
1366 ERR("Failed to send \"close stream\" command reply (ret = %zd)",
1367 send_ret);
1368 ret = -1;
173af62f
DG
1369 }
1370
1371end_no_session:
1372 return ret;
1373}
1374
93ec662e
JD
1375/*
1376 * relay_reset_metadata: reset a metadata stream
1377 */
1378static
5312a3ed
JG
1379int relay_reset_metadata(const struct lttcomm_relayd_hdr *recv_hdr,
1380 struct relay_connection *conn,
1381 const struct lttng_buffer_view *payload)
93ec662e 1382{
5312a3ed
JG
1383 int ret;
1384 ssize_t send_ret;
93ec662e
JD
1385 struct relay_session *session = conn->session;
1386 struct lttcomm_relayd_reset_metadata stream_info;
1387 struct lttcomm_relayd_generic_reply reply;
1388 struct relay_stream *stream;
1389
1390 DBG("Reset metadata received");
1391
5312a3ed 1392 if (!session || !conn->version_check_done) {
93ec662e
JD
1393 ERR("Trying to reset a metadata stream before version check");
1394 ret = -1;
1395 goto end_no_session;
1396 }
1397
5312a3ed
JG
1398 if (payload->size < sizeof(stream_info)) {
1399 ERR("Unexpected payload size in \"relay_reset_metadata\": expected >= %zu bytes, got %zu bytes",
1400 sizeof(stream_info), payload->size);
93ec662e
JD
1401 ret = -1;
1402 goto end_no_session;
1403 }
5312a3ed
JG
1404 memcpy(&stream_info, payload->data, sizeof(stream_info));
1405 stream_info.stream_id = be64toh(stream_info.stream_id);
1406 stream_info.version = be64toh(stream_info.version);
1407
1408 DBG("Update metadata to version %" PRIu64, stream_info.version);
93ec662e
JD
1409
1410 /* Unsupported for live sessions for now. */
1411 if (session->live_timer != 0) {
1412 ret = -1;
1413 goto end;
1414 }
1415
5312a3ed 1416 stream = stream_get_by_id(stream_info.stream_id);
93ec662e
JD
1417 if (!stream) {
1418 ret = -1;
1419 goto end;
1420 }
1421 pthread_mutex_lock(&stream->lock);
1422 if (!stream->is_metadata) {
1423 ret = -1;
1424 goto end_unlock;
1425 }
1426
1427 ret = utils_rotate_stream_file(stream->path_name, stream->channel_name,
1428 0, 0, -1, -1, stream->stream_fd->fd, NULL,
1429 &stream->stream_fd->fd);
1430 if (ret < 0) {
1431 ERR("Failed to rotate metadata file %s of channel %s",
1432 stream->path_name, stream->channel_name);
1433 goto end_unlock;
1434 }
1435
1436end_unlock:
1437 pthread_mutex_unlock(&stream->lock);
1438 stream_put(stream);
1439
1440end:
1441 memset(&reply, 0, sizeof(reply));
1442 if (ret < 0) {
1443 reply.ret_code = htobe32(LTTNG_ERR_UNK);
1444 } else {
1445 reply.ret_code = htobe32(LTTNG_OK);
1446 }
1447 send_ret = conn->sock->ops->sendmsg(conn->sock, &reply,
1448 sizeof(struct lttcomm_relayd_generic_reply), 0);
5312a3ed
JG
1449 if (send_ret < (ssize_t) sizeof(reply)) {
1450 ERR("Failed to send \"reset metadata\" command reply (ret = %zd)",
1451 send_ret);
1452 ret = -1;
93ec662e
JD
1453 }
1454
1455end_no_session:
1456 return ret;
1457}
1458
b8aa1682
JD
1459/*
1460 * relay_unknown_command: send -1 if received unknown command
1461 */
7591bab1 1462static void relay_unknown_command(struct relay_connection *conn)
b8aa1682
JD
1463{
1464 struct lttcomm_relayd_generic_reply reply;
5312a3ed 1465 ssize_t send_ret;
b8aa1682 1466
53efb85a 1467 memset(&reply, 0, sizeof(reply));
f73fabfd 1468 reply.ret_code = htobe32(LTTNG_ERR_UNK);
5312a3ed
JG
1469 send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0);
1470 if (send_ret < sizeof(reply)) {
1471 ERR("Failed to send \"unknown command\" command reply (ret = %zd)", send_ret);
b8aa1682
JD
1472 }
1473}
1474
1475/*
1476 * relay_start: send an acknowledgment to the client to tell if we are
1477 * ready to receive data. We are ready if a session is established.
1478 */
5312a3ed
JG
1479static int relay_start(const struct lttcomm_relayd_hdr *recv_hdr,
1480 struct relay_connection *conn,
1481 const struct lttng_buffer_view *payload)
b8aa1682 1482{
5312a3ed
JG
1483 int ret = 0;
1484 ssize_t send_ret;
b8aa1682 1485 struct lttcomm_relayd_generic_reply reply;
58eb9381 1486 struct relay_session *session = conn->session;
b8aa1682
JD
1487
1488 if (!session) {
1489 DBG("Trying to start the streaming without a session established");
f73fabfd 1490 ret = htobe32(LTTNG_ERR_UNK);
b8aa1682
JD
1491 }
1492
53efb85a 1493 memset(&reply, 0, sizeof(reply));
5312a3ed
JG
1494 reply.ret_code = htobe32(LTTNG_OK);
1495 send_ret = conn->sock->ops->sendmsg(conn->sock, &reply,
1496 sizeof(reply), 0);
1497 if (send_ret < (ssize_t) sizeof(reply)) {
1498 ERR("Failed to send \"relay_start\" command reply (ret = %zd)",
1499 send_ret);
1500 ret = -1;
b8aa1682
JD
1501 }
1502
1503 return ret;
1504}
1505
1d4dfdef
DG
1506/*
1507 * Append padding to the file pointed by the file descriptor fd.
1508 */
1509static int write_padding_to_file(int fd, uint32_t size)
1510{
6cd525e8 1511 ssize_t ret = 0;
1d4dfdef
DG
1512 char *zeros;
1513
1514 if (size == 0) {
1515 goto end;
1516 }
1517
1518 zeros = zmalloc(size);
1519 if (zeros == NULL) {
1520 PERROR("zmalloc zeros for padding");
1521 ret = -1;
1522 goto end;
1523 }
1524
6cd525e8
MD
1525 ret = lttng_write(fd, zeros, size);
1526 if (ret < size) {
1d4dfdef
DG
1527 PERROR("write padding to file");
1528 }
1529
e986c7a1
DG
1530 free(zeros);
1531
1d4dfdef
DG
1532end:
1533 return ret;
1534}
1535
d3ecc550
JD
1536/*
1537 * Close the current index file if it is open, and create a new one.
1538 *
1539 * Return 0 on success, -1 on error.
1540 */
1541static
cb523e02
JG
1542int create_rotate_index_file(struct relay_stream *stream,
1543 const char *stream_path)
d3ecc550
JD
1544{
1545 int ret;
1546 uint32_t major, minor;
1547
1548 /* Put ref on previous index_file. */
1549 if (stream->index_file) {
1550 lttng_index_file_put(stream->index_file);
1551 stream->index_file = NULL;
1552 }
1553 major = stream->trace->session->major;
1554 minor = stream->trace->session->minor;
cb523e02 1555 stream->index_file = lttng_index_file_create(stream_path,
d3ecc550
JD
1556 stream->channel_name,
1557 -1, -1, stream->tracefile_size,
1558 tracefile_array_get_file_index_head(stream->tfa),
1559 lttng_to_index_major(major, minor),
1560 lttng_to_index_minor(major, minor));
1561 if (!stream->index_file) {
1562 ret = -1;
1563 goto end;
1564 }
1565
1566 ret = 0;
1567
1568end:
1569 return ret;
1570}
1571
1572static
c6db3843 1573int do_rotate_stream_data(struct relay_stream *stream)
d3ecc550
JD
1574{
1575 int ret;
1576
c6db3843
JG
1577 DBG("Rotating stream %" PRIu64 " data file",
1578 stream->stream_handle);
d3ecc550
JD
1579 /* Perform the stream rotation. */
1580 ret = utils_rotate_stream_file(stream->path_name,
1581 stream->channel_name, stream->tracefile_size,
1582 stream->tracefile_count, -1,
1583 -1, stream->stream_fd->fd,
1584 NULL, &stream->stream_fd->fd);
1585 if (ret < 0) {
1586 ERR("Rotating stream output file");
1587 goto end;
1588 }
1589 stream->tracefile_size_current = 0;
d3ecc550 1590 stream->pos_after_last_complete_data_index = 0;
c6db3843 1591 stream->data_rotated = true;
d3ecc550 1592
c6db3843
JG
1593 if (stream->data_rotated && stream->index_rotated) {
1594 /* Rotation completed; reset its state. */
1595 DBG("Rotation completed for stream %" PRIu64,
1596 stream->stream_handle);
1597 stream->rotate_at_seq_num = -1ULL;
1598 stream->data_rotated = false;
1599 stream->index_rotated = false;
1600 }
d3ecc550
JD
1601end:
1602 return ret;
1603}
1604
1605/*
1606 * If too much data has been written in a tracefile before we received the
1607 * rotation command, we have to move the excess data to the new tracefile and
1608 * perform the rotation. This can happen because the control and data
1609 * connections are separate, the indexes as well as the commands arrive from
1610 * the control connection and we have no control over the order so we could be
1611 * in a situation where too much data has been received on the data connection
c6db3843 1612 * before the rotation command on the control connection arrives.
d3ecc550
JD
1613 */
1614static
1615int rotate_truncate_stream(struct relay_stream *stream)
1616{
1617 int ret, new_fd;
a5df8828 1618 off_t lseek_ret;
d3ecc550
JD
1619 uint64_t diff, pos = 0;
1620 char buf[FILE_COPY_BUFFER_SIZE];
1621
1622 assert(!stream->is_metadata);
1623
1624 assert(stream->tracefile_size_current >
1625 stream->pos_after_last_complete_data_index);
1626 diff = stream->tracefile_size_current -
1627 stream->pos_after_last_complete_data_index;
1628
1629 /* Create the new tracefile. */
1630 new_fd = utils_create_stream_file(stream->path_name,
1631 stream->channel_name,
1632 stream->tracefile_size, stream->tracefile_count,
1633 /* uid */ -1, /* gid */ -1, /* suffix */ NULL);
1634 if (new_fd < 0) {
1635 ERR("Failed to create new stream file at path %s for channel %s",
1636 stream->path_name, stream->channel_name);
1637 ret = -1;
1638 goto end;
1639 }
1640
1641 /*
1642 * Rewind the current tracefile to the position at which the rotation
1643 * should have occured.
1644 */
a5df8828 1645 lseek_ret = lseek(stream->stream_fd->fd,
d3ecc550 1646 stream->pos_after_last_complete_data_index, SEEK_SET);
a5df8828 1647 if (lseek_ret < 0) {
d3ecc550 1648 PERROR("seek truncate stream");
a5df8828 1649 ret = -1;
d3ecc550
JD
1650 goto end;
1651 }
1652
1653 /* Move data from the old file to the new file. */
1654 while (pos < diff) {
1655 uint64_t count, bytes_left;
1656 ssize_t io_ret;
1657
1658 bytes_left = diff - pos;
1659 count = bytes_left > sizeof(buf) ? sizeof(buf) : bytes_left;
1660 assert(count <= SIZE_MAX);
1661
1662 io_ret = lttng_read(stream->stream_fd->fd, buf, count);
1663 if (io_ret < (ssize_t) count) {
1664 char error_string[256];
1665
1666 snprintf(error_string, sizeof(error_string),
1667 "Failed to read %" PRIu64 " bytes from fd %i in rotate_truncate_stream(), returned %zi",
1668 count, stream->stream_fd->fd, io_ret);
1669 if (io_ret == -1) {
1670 PERROR("%s", error_string);
1671 } else {
1672 ERR("%s", error_string);
1673 }
1674 ret = -1;
1675 goto end;
1676 }
1677
1678 io_ret = lttng_write(new_fd, buf, count);
1679 if (io_ret < (ssize_t) count) {
1680 char error_string[256];
1681
1682 snprintf(error_string, sizeof(error_string),
1683 "Failed to write %" PRIu64 " bytes from fd %i in rotate_truncate_stream(), returned %zi",
1684 count, new_fd, io_ret);
1685 if (io_ret == -1) {
1686 PERROR("%s", error_string);
1687 } else {
1688 ERR("%s", error_string);
1689 }
1690 ret = -1;
1691 goto end;
1692 }
1693
1694 pos += count;
1695 }
1696
1697 /* Truncate the file to get rid of the excess data. */
1698 ret = ftruncate(stream->stream_fd->fd,
1699 stream->pos_after_last_complete_data_index);
1700 if (ret) {
1701 PERROR("ftruncate");
1702 goto end;
1703 }
1704
1705 ret = close(stream->stream_fd->fd);
1706 if (ret < 0) {
1707 PERROR("Closing tracefile");
1708 goto end;
1709 }
1710
d3ecc550
JD
1711 /*
1712 * Update the offset and FD of all the eventual indexes created by the
1713 * data connection before the rotation command arrived.
1714 */
1715 ret = relay_index_switch_all_files(stream);
1716 if (ret < 0) {
1717 ERR("Failed to rotate index file");
1718 goto end;
1719 }
1720
1721 stream->stream_fd->fd = new_fd;
1722 stream->tracefile_size_current = diff;
1723 stream->pos_after_last_complete_data_index = 0;
1724 stream->rotate_at_seq_num = -1ULL;
1725
1726 ret = 0;
1727
1728end:
1729 return ret;
1730}
1731
1732/*
c6db3843 1733 * Check if a stream's index file should be rotated (for session rotation).
d3ecc550
JD
1734 * Must be called with the stream lock held.
1735 *
1736 * Return 0 on success, a negative value on error.
1737 */
1738static
c6db3843 1739int try_rotate_stream_index(struct relay_stream *stream)
d3ecc550
JD
1740{
1741 int ret = 0;
1742
d3ecc550 1743 if (stream->rotate_at_seq_num == -1ULL) {
c6db3843 1744 /* No rotation expected. */
d3ecc550
JD
1745 goto end;
1746 }
1747
c6db3843
JG
1748 if (stream->index_rotated) {
1749 /* Rotation of the index has already occurred. */
1750 goto end;
1751 }
1752
1753 if (stream->prev_index_seq == -1ULL ||
1754 stream->prev_index_seq < stream->rotate_at_seq_num) {
1755 DBG("Stream %" PRIu64 " index not yet ready for rotation (rotate_at_seq_num = %" PRIu64 ", prev_index_seq = %" PRIu64 ")",
1756 stream->stream_handle,
1757 stream->rotate_at_seq_num,
1758 stream->prev_index_seq);
1759 goto end;
1760 } else if (stream->prev_index_seq != stream->rotate_at_seq_num) {
1761 /*
1762 * Unexpected, protocol error/bug.
1763 * It could mean that we received a rotation position
1764 * that is in the past.
1765 */
1766 ERR("Stream %" PRIu64 " index is in an inconsistent state (rotate_at_seq_num = %" PRIu64 ", prev_data_seq = %" PRIu64 ", prev_index_seq = %" PRIu64 ")",
c64c70f5
JG
1767 stream->stream_handle,
1768 stream->rotate_at_seq_num,
a8f9f353 1769 stream->prev_data_seq,
c64c70f5 1770 stream->prev_index_seq);
c6db3843
JG
1771 ret = -1;
1772 goto end;
1773 } else {
1774 DBG("Rotating stream %" PRIu64 " index file",
1775 stream->stream_handle);
1776 ret = create_rotate_index_file(stream, stream->path_name);
1777 stream->index_rotated = true;
1778
1779 if (stream->data_rotated && stream->index_rotated) {
1780 /* Rotation completed; reset its state. */
1781 DBG("Rotation completed for stream %" PRIu64,
1782 stream->stream_handle);
1783 stream->rotate_at_seq_num = -1ULL;
1784 stream->data_rotated = false;
1785 stream->index_rotated = false;
1786 }
1787 }
1788
1789end:
1790 return ret;
1791}
1792
1793/*
1794 * Check if a stream's data file (as opposed to index) should be rotated
1795 * (for session rotation).
1796 * Must be called with the stream lock held.
1797 *
1798 * Return 0 on success, a negative value on error.
1799 */
1800static
1801int try_rotate_stream_data(struct relay_stream *stream)
1802{
1803 int ret = 0;
1804
1805 if (stream->rotate_at_seq_num == -1ULL) {
1806 /* No rotation expected. */
1807 goto end;
1808 }
1809
1810 if (stream->data_rotated) {
1811 /* Rotation of the data file has already occurred. */
1812 goto end;
1813 }
1814
1815 if (stream->prev_data_seq == -1ULL ||
1816 stream->prev_data_seq < stream->rotate_at_seq_num) {
1817 DBG("Stream %" PRIu64 " not yet ready for rotation (rotate_at_seq_num = %" PRIu64 ", prev_data_seq = %" PRIu64 ")",
1818 stream->stream_handle,
1819 stream->rotate_at_seq_num,
1820 stream->prev_data_seq);
d3ecc550 1821 goto end;
a8f9f353 1822 } else if (stream->prev_data_seq > stream->rotate_at_seq_num) {
c64c70f5 1823 /*
a8f9f353 1824 * prev_data_seq is checked here since indexes and rotation
c64c70f5
JG
1825 * commands are serialized with respect to each other.
1826 */
d3ecc550
JD
1827 DBG("Rotation after too much data has been written in tracefile "
1828 "for stream %" PRIu64 ", need to truncate before "
1829 "rotating", stream->stream_handle);
1830 ret = rotate_truncate_stream(stream);
1831 if (ret) {
1832 ERR("Failed to truncate stream");
1833 goto end;
1834 }
c6db3843
JG
1835 } else if (stream->prev_data_seq != stream->rotate_at_seq_num) {
1836 /*
1837 * Unexpected, protocol error/bug.
1838 * It could mean that we received a rotation position
1839 * that is in the past.
1840 */
1841 ERR("Stream %" PRIu64 " data is in an inconsistent state (rotate_at_seq_num = %" PRIu64 ", prev_data_seq = %" PRIu64 ")",
c64c70f5
JG
1842 stream->stream_handle,
1843 stream->rotate_at_seq_num,
c6db3843
JG
1844 stream->prev_data_seq);
1845 ret = -1;
1846 goto end;
1847 } else {
1848 ret = do_rotate_stream_data(stream);
d3ecc550
JD
1849 }
1850
1851end:
1852 return ret;
1853}
1854
b8aa1682 1855/*
7591bab1 1856 * relay_recv_metadata: receive the metadata for the session.
b8aa1682 1857 */
5312a3ed
JG
1858static int relay_recv_metadata(const struct lttcomm_relayd_hdr *recv_hdr,
1859 struct relay_connection *conn,
1860 const struct lttng_buffer_view *payload)
b8aa1682 1861{
32d1569c 1862 int ret = 0;
6cd525e8 1863 ssize_t size_ret;
58eb9381 1864 struct relay_session *session = conn->session;
5312a3ed 1865 struct lttcomm_relayd_metadata_payload metadata_payload_header;
b8aa1682 1866 struct relay_stream *metadata_stream;
5312a3ed 1867 uint64_t metadata_payload_size;
b8aa1682
JD
1868
1869 if (!session) {
1870 ERR("Metadata sent before version check");
1871 ret = -1;
1872 goto end;
1873 }
1874
5312a3ed 1875 if (recv_hdr->data_size < sizeof(struct lttcomm_relayd_metadata_payload)) {
f6416125
MD
1876 ERR("Incorrect data size");
1877 ret = -1;
1878 goto end;
1879 }
5312a3ed
JG
1880 metadata_payload_size = recv_hdr->data_size -
1881 sizeof(struct lttcomm_relayd_metadata_payload);
f6416125 1882
5312a3ed
JG
1883 memcpy(&metadata_payload_header, payload->data,
1884 sizeof(metadata_payload_header));
1885 metadata_payload_header.stream_id = be64toh(
1886 metadata_payload_header.stream_id);
1887 metadata_payload_header.padding_size = be32toh(
1888 metadata_payload_header.padding_size);
9d1bbf21 1889
5312a3ed 1890 metadata_stream = stream_get_by_id(metadata_payload_header.stream_id);
b8aa1682
JD
1891 if (!metadata_stream) {
1892 ret = -1;
7591bab1 1893 goto end;
b8aa1682
JD
1894 }
1895
7591bab1
MD
1896 pthread_mutex_lock(&metadata_stream->lock);
1897
5312a3ed
JG
1898 size_ret = lttng_write(metadata_stream->stream_fd->fd,
1899 payload->data + sizeof(metadata_payload_header),
1900 metadata_payload_size);
1901 if (size_ret < metadata_payload_size) {
b8aa1682
JD
1902 ERR("Relay error writing metadata on file");
1903 ret = -1;
7591bab1 1904 goto end_put;
b8aa1682 1905 }
1d4dfdef 1906
32d1569c 1907 size_ret = write_padding_to_file(metadata_stream->stream_fd->fd,
5312a3ed 1908 metadata_payload_header.padding_size);
715e6fb1 1909 if (size_ret < (int64_t) metadata_payload_header.padding_size) {
5312a3ed 1910 ret = -1;
7591bab1 1911 goto end_put;
1d4dfdef 1912 }
2a174661 1913
7591bab1 1914 metadata_stream->metadata_received +=
5312a3ed 1915 metadata_payload_size + metadata_payload_header.padding_size;
7591bab1
MD
1916 DBG2("Relay metadata written. Updated metadata_received %" PRIu64,
1917 metadata_stream->metadata_received);
1d4dfdef 1918
c6db3843 1919 ret = try_rotate_stream_data(metadata_stream);
d3ecc550
JD
1920 if (ret < 0) {
1921 goto end_put;
1922 }
1923
7591bab1
MD
1924end_put:
1925 pthread_mutex_unlock(&metadata_stream->lock);
1926 stream_put(metadata_stream);
b8aa1682
JD
1927end:
1928 return ret;
1929}
1930
1931/*
1932 * relay_send_version: send relayd version number
1933 */
5312a3ed
JG
1934static int relay_send_version(const struct lttcomm_relayd_hdr *recv_hdr,
1935 struct relay_connection *conn,
1936 const struct lttng_buffer_view *payload)
b8aa1682 1937{
7f51dcba 1938 int ret;
5312a3ed 1939 ssize_t send_ret;
092b6259 1940 struct lttcomm_relayd_version reply, msg;
87cb6359 1941 bool compatible = true;
b8aa1682 1942
5312a3ed 1943 conn->version_check_done = true;
b8aa1682 1944
092b6259 1945 /* Get version from the other side. */
5312a3ed
JG
1946 if (payload->size < sizeof(msg)) {
1947 ERR("Unexpected payload size in \"relay_send_version\": expected >= %zu bytes, got %zu bytes",
1948 sizeof(msg), payload->size);
092b6259 1949 ret = -1;
092b6259
DG
1950 goto end;
1951 }
1952
5312a3ed
JG
1953 memcpy(&msg, payload->data, sizeof(msg));
1954 msg.major = be32toh(msg.major);
1955 msg.minor = be32toh(msg.minor);
1956
53efb85a 1957 memset(&reply, 0, sizeof(reply));
d83a952c
MD
1958 reply.major = RELAYD_VERSION_COMM_MAJOR;
1959 reply.minor = RELAYD_VERSION_COMM_MINOR;
d4519fa3
JD
1960
1961 /* Major versions must be the same */
5312a3ed 1962 if (reply.major != msg.major) {
6151a90f 1963 DBG("Incompatible major versions (%u vs %u), deleting session",
5312a3ed 1964 reply.major, msg.major);
87cb6359 1965 compatible = false;
d4519fa3
JD
1966 }
1967
58eb9381 1968 conn->major = reply.major;
0f907de1 1969 /* We adapt to the lowest compatible version */
5312a3ed 1970 if (reply.minor <= msg.minor) {
58eb9381 1971 conn->minor = reply.minor;
0f907de1 1972 } else {
5312a3ed 1973 conn->minor = msg.minor;
0f907de1
JD
1974 }
1975
6151a90f
JD
1976 reply.major = htobe32(reply.major);
1977 reply.minor = htobe32(reply.minor);
5312a3ed
JG
1978 send_ret = conn->sock->ops->sendmsg(conn->sock, &reply,
1979 sizeof(reply), 0);
1980 if (send_ret < (ssize_t) sizeof(reply)) {
1981 ERR("Failed to send \"send version\" command reply (ret = %zd)",
1982 send_ret);
1983 ret = -1;
1984 goto end;
1985 } else {
1986 ret = 0;
6151a90f
JD
1987 }
1988
87cb6359
JD
1989 if (!compatible) {
1990 ret = -1;
1991 goto end;
1992 }
1993
58eb9381
DG
1994 DBG("Version check done using protocol %u.%u", conn->major,
1995 conn->minor);
b8aa1682
JD
1996
1997end:
1998 return ret;
1999}
2000
c8f59ee5 2001/*
6d805429 2002 * Check for data pending for a given stream id from the session daemon.
c8f59ee5 2003 */
5312a3ed
JG
2004static int relay_data_pending(const struct lttcomm_relayd_hdr *recv_hdr,
2005 struct relay_connection *conn,
2006 const struct lttng_buffer_view *payload)
c8f59ee5 2007{
58eb9381 2008 struct relay_session *session = conn->session;
6d805429 2009 struct lttcomm_relayd_data_pending msg;
c8f59ee5
DG
2010 struct lttcomm_relayd_generic_reply reply;
2011 struct relay_stream *stream;
5312a3ed 2012 ssize_t send_ret;
c8f59ee5 2013 int ret;
298a25ca 2014 uint64_t stream_seq;
c8f59ee5 2015
6d805429 2016 DBG("Data pending command received");
c8f59ee5 2017
5312a3ed 2018 if (!session || !conn->version_check_done) {
c8f59ee5
DG
2019 ERR("Trying to check for data before version check");
2020 ret = -1;
2021 goto end_no_session;
2022 }
2023
5312a3ed
JG
2024 if (payload->size < sizeof(msg)) {
2025 ERR("Unexpected payload size in \"relay_data_pending\": expected >= %zu bytes, got %zu bytes",
2026 sizeof(msg), payload->size);
c8f59ee5
DG
2027 ret = -1;
2028 goto end_no_session;
2029 }
5312a3ed
JG
2030 memcpy(&msg, payload->data, sizeof(msg));
2031 msg.stream_id = be64toh(msg.stream_id);
2032 msg.last_net_seq_num = be64toh(msg.last_net_seq_num);
c8f59ee5 2033
5312a3ed 2034 stream = stream_get_by_id(msg.stream_id);
de91f48a 2035 if (stream == NULL) {
c8f59ee5 2036 ret = -1;
7591bab1 2037 goto end;
c8f59ee5
DG
2038 }
2039
7591bab1
MD
2040 pthread_mutex_lock(&stream->lock);
2041
298a25ca
JG
2042 if (session_streams_have_index(session)) {
2043 /*
2044 * Ensure that both the index and stream data have been
2045 * flushed up to the requested point.
2046 */
a8f9f353 2047 stream_seq = min(stream->prev_data_seq, stream->prev_index_seq);
298a25ca 2048 } else {
a8f9f353 2049 stream_seq = stream->prev_data_seq;
298a25ca 2050 }
a8f9f353 2051 DBG("Data pending for stream id %" PRIu64 ": prev_data_seq %" PRIu64
298a25ca
JG
2052 ", prev_index_seq %" PRIu64
2053 ", and last_seq %" PRIu64, msg.stream_id,
a8f9f353 2054 stream->prev_data_seq, stream->prev_index_seq,
298a25ca 2055 msg.last_net_seq_num);
c8f59ee5 2056
33832e64 2057 /* Avoid wrapping issue */
298a25ca 2058 if (((int64_t) (stream_seq - msg.last_net_seq_num)) >= 0) {
6d805429 2059 /* Data has in fact been written and is NOT pending */
c8f59ee5 2060 ret = 0;
6d805429
DG
2061 } else {
2062 /* Data still being streamed thus pending */
2063 ret = 1;
c8f59ee5
DG
2064 }
2065
7591bab1
MD
2066 stream->data_pending_check_done = true;
2067 pthread_mutex_unlock(&stream->lock);
f7079f67 2068
7591bab1
MD
2069 stream_put(stream);
2070end:
c8f59ee5 2071
53efb85a 2072 memset(&reply, 0, sizeof(reply));
c8f59ee5 2073 reply.ret_code = htobe32(ret);
5312a3ed
JG
2074 send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0);
2075 if (send_ret < (ssize_t) sizeof(reply)) {
2076 ERR("Failed to send \"data pending\" command reply (ret = %zd)",
2077 send_ret);
2078 ret = -1;
c8f59ee5
DG
2079 }
2080
2081end_no_session:
2082 return ret;
2083}
2084
2085/*
2086 * Wait for the control socket to reach a quiescent state.
2087 *
7591bab1
MD
2088 * Note that for now, when receiving this command from the session
2089 * daemon, this means that every subsequent commands or data received on
2090 * the control socket has been handled. So, this is why we simply return
2091 * OK here.
c8f59ee5 2092 */
5312a3ed
JG
2093static int relay_quiescent_control(const struct lttcomm_relayd_hdr *recv_hdr,
2094 struct relay_connection *conn,
2095 const struct lttng_buffer_view *payload)
c8f59ee5
DG
2096{
2097 int ret;
5312a3ed 2098 ssize_t send_ret;
ad7051c0 2099 struct relay_stream *stream;
ad7051c0 2100 struct lttcomm_relayd_quiescent_control msg;
c8f59ee5
DG
2101 struct lttcomm_relayd_generic_reply reply;
2102
2103 DBG("Checking quiescent state on control socket");
2104
5312a3ed 2105 if (!conn->session || !conn->version_check_done) {
ad7051c0
DG
2106 ERR("Trying to check for data before version check");
2107 ret = -1;
2108 goto end_no_session;
2109 }
2110
5312a3ed
JG
2111 if (payload->size < sizeof(msg)) {
2112 ERR("Unexpected payload size in \"relay_quiescent_control\": expected >= %zu bytes, got %zu bytes",
2113 sizeof(msg), payload->size);
ad7051c0
DG
2114 ret = -1;
2115 goto end_no_session;
2116 }
5312a3ed
JG
2117 memcpy(&msg, payload->data, sizeof(msg));
2118 msg.stream_id = be64toh(msg.stream_id);
ad7051c0 2119
5312a3ed 2120 stream = stream_get_by_id(msg.stream_id);
7591bab1
MD
2121 if (!stream) {
2122 goto reply;
2123 }
2124 pthread_mutex_lock(&stream->lock);
2125 stream->data_pending_check_done = true;
2126 pthread_mutex_unlock(&stream->lock);
5312a3ed
JG
2127
2128 DBG("Relay quiescent control pending flag set to %" PRIu64, msg.stream_id);
7591bab1
MD
2129 stream_put(stream);
2130reply:
53efb85a 2131 memset(&reply, 0, sizeof(reply));
c8f59ee5 2132 reply.ret_code = htobe32(LTTNG_OK);
5312a3ed
JG
2133 send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0);
2134 if (send_ret < (ssize_t) sizeof(reply)) {
2135 ERR("Failed to send \"quiescent control\" command reply (ret = %zd)",
2136 send_ret);
2137 ret = -1;
2138 } else {
2139 ret = 0;
c8f59ee5
DG
2140 }
2141
ad7051c0 2142end_no_session:
c8f59ee5
DG
2143 return ret;
2144}
2145
f7079f67 2146/*
7591bab1
MD
2147 * Initialize a data pending command. This means that a consumer is about
2148 * to ask for data pending for each stream it holds. Simply iterate over
2149 * all streams of a session and set the data_pending_check_done flag.
f7079f67
DG
2150 *
2151 * This command returns to the client a LTTNG_OK code.
2152 */
5312a3ed
JG
2153static int relay_begin_data_pending(const struct lttcomm_relayd_hdr *recv_hdr,
2154 struct relay_connection *conn,
2155 const struct lttng_buffer_view *payload)
f7079f67
DG
2156{
2157 int ret;
5312a3ed 2158 ssize_t send_ret;
f7079f67
DG
2159 struct lttng_ht_iter iter;
2160 struct lttcomm_relayd_begin_data_pending msg;
2161 struct lttcomm_relayd_generic_reply reply;
2162 struct relay_stream *stream;
f7079f67
DG
2163
2164 assert(recv_hdr);
58eb9381 2165 assert(conn);
f7079f67
DG
2166
2167 DBG("Init streams for data pending");
2168
5312a3ed 2169 if (!conn->session || !conn->version_check_done) {
f7079f67
DG
2170 ERR("Trying to check for data before version check");
2171 ret = -1;
2172 goto end_no_session;
2173 }
2174
5312a3ed
JG
2175 if (payload->size < sizeof(msg)) {
2176 ERR("Unexpected payload size in \"relay_begin_data_pending\": expected >= %zu bytes, got %zu bytes",
2177 sizeof(msg), payload->size);
f7079f67
DG
2178 ret = -1;
2179 goto end_no_session;
2180 }
5312a3ed
JG
2181 memcpy(&msg, payload->data, sizeof(msg));
2182 msg.session_id = be64toh(msg.session_id);
f7079f67
DG
2183
2184 /*
7591bab1
MD
2185 * Iterate over all streams to set the begin data pending flag.
2186 * For now, the streams are indexed by stream handle so we have
2187 * to iterate over all streams to find the one associated with
2188 * the right session_id.
f7079f67
DG
2189 */
2190 rcu_read_lock();
d3e2ba59 2191 cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream,
2a174661 2192 node.node) {
7591bab1
MD
2193 if (!stream_get(stream)) {
2194 continue;
2195 }
5312a3ed 2196 if (stream->trace->session->id == msg.session_id) {
7591bab1
MD
2197 pthread_mutex_lock(&stream->lock);
2198 stream->data_pending_check_done = false;
2199 pthread_mutex_unlock(&stream->lock);
f7079f67
DG
2200 DBG("Set begin data pending flag to stream %" PRIu64,
2201 stream->stream_handle);
2202 }
7591bab1 2203 stream_put(stream);
f7079f67
DG
2204 }
2205 rcu_read_unlock();
2206
53efb85a 2207 memset(&reply, 0, sizeof(reply));
f7079f67
DG
2208 /* All good, send back reply. */
2209 reply.ret_code = htobe32(LTTNG_OK);
2210
5312a3ed
JG
2211 send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0);
2212 if (send_ret < (ssize_t) sizeof(reply)) {
2213 ERR("Failed to send \"begin data pending\" command reply (ret = %zd)",
2214 send_ret);
2215 ret = -1;
2216 } else {
2217 ret = 0;
f7079f67
DG
2218 }
2219
2220end_no_session:
2221 return ret;
2222}
2223
2224/*
7591bab1
MD
2225 * End data pending command. This will check, for a given session id, if
2226 * each stream associated with it has its data_pending_check_done flag
2227 * set. If not, this means that the client lost track of the stream but
2228 * the data is still being streamed on our side. In this case, we inform
2229 * the client that data is in flight.
f7079f67
DG
2230 *
2231 * Return to the client if there is data in flight or not with a ret_code.
2232 */
5312a3ed
JG
2233static int relay_end_data_pending(const struct lttcomm_relayd_hdr *recv_hdr,
2234 struct relay_connection *conn,
2235 const struct lttng_buffer_view *payload)
f7079f67
DG
2236{
2237 int ret;
5312a3ed 2238 ssize_t send_ret;
f7079f67
DG
2239 struct lttng_ht_iter iter;
2240 struct lttcomm_relayd_end_data_pending msg;
2241 struct lttcomm_relayd_generic_reply reply;
2242 struct relay_stream *stream;
f7079f67
DG
2243 uint32_t is_data_inflight = 0;
2244
f7079f67
DG
2245 DBG("End data pending command");
2246
5312a3ed 2247 if (!conn->session || !conn->version_check_done) {
f7079f67
DG
2248 ERR("Trying to check for data before version check");
2249 ret = -1;
2250 goto end_no_session;
2251 }
2252
5312a3ed
JG
2253 if (payload->size < sizeof(msg)) {
2254 ERR("Unexpected payload size in \"relay_end_data_pending\": expected >= %zu bytes, got %zu bytes",
2255 sizeof(msg), payload->size);
f7079f67
DG
2256 ret = -1;
2257 goto end_no_session;
2258 }
5312a3ed
JG
2259 memcpy(&msg, payload->data, sizeof(msg));
2260 msg.session_id = be64toh(msg.session_id);
f7079f67 2261
7591bab1
MD
2262 /*
2263 * Iterate over all streams to see if the begin data pending
2264 * flag is set.
2265 */
f7079f67 2266 rcu_read_lock();
d3e2ba59 2267 cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream,
2a174661 2268 node.node) {
7591bab1
MD
2269 if (!stream_get(stream)) {
2270 continue;
2271 }
5312a3ed 2272 if (stream->trace->session->id != msg.session_id) {
7591bab1
MD
2273 stream_put(stream);
2274 continue;
2275 }
2276 pthread_mutex_lock(&stream->lock);
2277 if (!stream->data_pending_check_done) {
298a25ca
JG
2278 uint64_t stream_seq;
2279
2280 if (session_streams_have_index(conn->session)) {
2281 /*
2282 * Ensure that both the index and stream data have been
2283 * flushed up to the requested point.
2284 */
a8f9f353 2285 stream_seq = min(stream->prev_data_seq, stream->prev_index_seq);
298a25ca 2286 } else {
a8f9f353 2287 stream_seq = stream->prev_data_seq;
298a25ca
JG
2288 }
2289 if (!stream->closed || !(((int64_t) (stream_seq - stream->last_net_seq_num)) >= 0)) {
7591bab1
MD
2290 is_data_inflight = 1;
2291 DBG("Data is still in flight for stream %" PRIu64,
2292 stream->stream_handle);
2293 pthread_mutex_unlock(&stream->lock);
2294 stream_put(stream);
2295 break;
2296 }
f7079f67 2297 }
7591bab1
MD
2298 pthread_mutex_unlock(&stream->lock);
2299 stream_put(stream);
f7079f67
DG
2300 }
2301 rcu_read_unlock();
2302
53efb85a 2303 memset(&reply, 0, sizeof(reply));
f7079f67
DG
2304 /* All good, send back reply. */
2305 reply.ret_code = htobe32(is_data_inflight);
2306
5312a3ed
JG
2307 send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0);
2308 if (send_ret < (ssize_t) sizeof(reply)) {
2309 ERR("Failed to send \"end data pending\" command reply (ret = %zd)",
2310 send_ret);
2311 ret = -1;
2312 } else {
2313 ret = 0;
f7079f67
DG
2314 }
2315
2316end_no_session:
2317 return ret;
2318}
2319
1c20f0e2
JD
2320/*
2321 * Receive an index for a specific stream.
2322 *
2323 * Return 0 on success else a negative value.
2324 */
5312a3ed
JG
2325static int relay_recv_index(const struct lttcomm_relayd_hdr *recv_hdr,
2326 struct relay_connection *conn,
2327 const struct lttng_buffer_view *payload)
1c20f0e2 2328{
5312a3ed
JG
2329 int ret;
2330 ssize_t send_ret;
58eb9381 2331 struct relay_session *session = conn->session;
1c20f0e2 2332 struct lttcomm_relayd_index index_info;
7591bab1 2333 struct relay_index *index;
1c20f0e2
JD
2334 struct lttcomm_relayd_generic_reply reply;
2335 struct relay_stream *stream;
f8f3885c 2336 size_t msg_len;
1c20f0e2 2337
58eb9381 2338 assert(conn);
1c20f0e2
JD
2339
2340 DBG("Relay receiving index");
2341
5312a3ed 2342 if (!session || !conn->version_check_done) {
1c20f0e2
JD
2343 ERR("Trying to close a stream before version check");
2344 ret = -1;
2345 goto end_no_session;
2346 }
2347
f8f3885c
MD
2348 msg_len = lttcomm_relayd_index_len(
2349 lttng_to_index_major(conn->major, conn->minor),
2350 lttng_to_index_minor(conn->major, conn->minor));
5312a3ed
JG
2351 if (payload->size < msg_len) {
2352 ERR("Unexpected payload size in \"relay_recv_index\": expected >= %zu bytes, got %zu bytes",
2353 msg_len, payload->size);
1c20f0e2
JD
2354 ret = -1;
2355 goto end_no_session;
2356 }
5312a3ed
JG
2357 memcpy(&index_info, payload->data, msg_len);
2358 index_info.relay_stream_id = be64toh(index_info.relay_stream_id);
2359 index_info.net_seq_num = be64toh(index_info.net_seq_num);
2360 index_info.packet_size = be64toh(index_info.packet_size);
2361 index_info.content_size = be64toh(index_info.content_size);
2362 index_info.timestamp_begin = be64toh(index_info.timestamp_begin);
2363 index_info.timestamp_end = be64toh(index_info.timestamp_end);
2364 index_info.events_discarded = be64toh(index_info.events_discarded);
2365 index_info.stream_id = be64toh(index_info.stream_id);
81df238b
JR
2366
2367 if (conn->minor >= 8) {
2368 index_info.stream_instance_id =
2369 be64toh(index_info.stream_instance_id);
2370 index_info.packet_seq_num = be64toh(index_info.packet_seq_num);
2371 }
5312a3ed
JG
2372
2373 stream = stream_get_by_id(index_info.relay_stream_id);
1c20f0e2 2374 if (!stream) {
7591bab1 2375 ERR("stream_get_by_id not found");
1c20f0e2 2376 ret = -1;
7591bab1 2377 goto end;
1c20f0e2 2378 }
7591bab1 2379 pthread_mutex_lock(&stream->lock);
1c20f0e2 2380
d3e2ba59
JD
2381 /* Live beacon handling */
2382 if (index_info.packet_size == 0) {
7591bab1
MD
2383 DBG("Received live beacon for stream %" PRIu64,
2384 stream->stream_handle);
d3e2ba59
JD
2385
2386 /*
7591bab1
MD
2387 * Only flag a stream inactive when it has already
2388 * received data and no indexes are in flight.
d3e2ba59 2389 */
a44ca2ca 2390 if (stream->index_received_seqcount > 0
7591bab1 2391 && stream->indexes_in_flight == 0) {
5312a3ed 2392 stream->beacon_ts_end = index_info.timestamp_end;
d3e2ba59
JD
2393 }
2394 ret = 0;
7591bab1 2395 goto end_stream_put;
d3e2ba59
JD
2396 } else {
2397 stream->beacon_ts_end = -1ULL;
2398 }
2399
528f2ffa 2400 if (stream->ctf_stream_id == -1ULL) {
5312a3ed 2401 stream->ctf_stream_id = index_info.stream_id;
528f2ffa 2402 }
5312a3ed 2403 index = relay_index_get_by_id_or_create(stream, index_info.net_seq_num);
7591bab1
MD
2404 if (!index) {
2405 ret = -1;
2406 ERR("relay_index_get_by_id_or_create index NULL");
2407 goto end_stream_put;
1c20f0e2 2408 }
234cd636 2409 if (set_index_control_data(index, &index_info, conn)) {
7591bab1
MD
2410 ERR("set_index_control_data error");
2411 relay_index_put(index);
2412 ret = -1;
2413 goto end_stream_put;
2414 }
2415 ret = relay_index_try_flush(index);
2416 if (ret == 0) {
a44ca2ca
MD
2417 tracefile_array_commit_seq(stream->tfa);
2418 stream->index_received_seqcount++;
d3ecc550 2419 stream->pos_after_last_complete_data_index += index->total_size;
7a45c7e6 2420 stream->prev_index_seq = index_info.net_seq_num;
c6db3843
JG
2421
2422 ret = try_rotate_stream_index(stream);
2423 if (ret < 0) {
2424 goto end_stream_put;
2425 }
7591bab1
MD
2426 } else if (ret > 0) {
2427 /* no flush. */
2428 ret = 0;
2429 } else {
245f8bec
JR
2430 /*
2431 * ret < 0
2432 *
2433 * relay_index_try_flush is responsible for the self-reference
2434 * put of the index object on error.
2435 */
7591bab1 2436 ERR("relay_index_try_flush error %d", ret);
7591bab1 2437 ret = -1;
1c20f0e2
JD
2438 }
2439
7591bab1
MD
2440end_stream_put:
2441 pthread_mutex_unlock(&stream->lock);
2442 stream_put(stream);
2443
2444end:
1c20f0e2 2445
53efb85a 2446 memset(&reply, 0, sizeof(reply));
1c20f0e2
JD
2447 if (ret < 0) {
2448 reply.ret_code = htobe32(LTTNG_ERR_UNK);
2449 } else {
2450 reply.ret_code = htobe32(LTTNG_OK);
2451 }
58eb9381 2452 send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0);
5312a3ed
JG
2453 if (send_ret < (ssize_t) sizeof(reply)) {
2454 ERR("Failed to send \"recv index\" command reply (ret = %zd)", send_ret);
2455 ret = -1;
1c20f0e2
JD
2456 }
2457
2458end_no_session:
2459 return ret;
2460}
2461
a4baae1b
JD
2462/*
2463 * Receive the streams_sent message.
2464 *
2465 * Return 0 on success else a negative value.
2466 */
5312a3ed
JG
2467static int relay_streams_sent(const struct lttcomm_relayd_hdr *recv_hdr,
2468 struct relay_connection *conn,
2469 const struct lttng_buffer_view *payload)
a4baae1b 2470{
5312a3ed
JG
2471 int ret;
2472 ssize_t send_ret;
a4baae1b
JD
2473 struct lttcomm_relayd_generic_reply reply;
2474
58eb9381 2475 assert(conn);
a4baae1b
JD
2476
2477 DBG("Relay receiving streams_sent");
2478
5312a3ed 2479 if (!conn->session || !conn->version_check_done) {
a4baae1b
JD
2480 ERR("Trying to close a stream before version check");
2481 ret = -1;
2482 goto end_no_session;
2483 }
2484
2485 /*
7591bab1
MD
2486 * Publish every pending stream in the connection recv list which are
2487 * now ready to be used by the viewer.
4a9daf17 2488 */
7591bab1 2489 publish_connection_local_streams(conn);
4a9daf17 2490
53efb85a 2491 memset(&reply, 0, sizeof(reply));
a4baae1b 2492 reply.ret_code = htobe32(LTTNG_OK);
58eb9381 2493 send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0);
5312a3ed
JG
2494 if (send_ret < (ssize_t) sizeof(reply)) {
2495 ERR("Failed to send \"streams sent\" command reply (ret = %zd)",
2496 send_ret);
2497 ret = -1;
a4baae1b
JD
2498 } else {
2499 /* Success. */
2500 ret = 0;
2501 }
2502
2503end_no_session:
2504 return ret;
2505}
2506
d3ecc550 2507/*
5312a3ed 2508 * relay_rotate_session_stream: rotate a stream to a new tracefile for the session
d3ecc550
JD
2509 * rotation feature (not the tracefile rotation feature).
2510 */
5312a3ed
JG
2511static int relay_rotate_session_stream(const struct lttcomm_relayd_hdr *recv_hdr,
2512 struct relay_connection *conn,
2513 const struct lttng_buffer_view *payload)
d3ecc550 2514{
5312a3ed
JG
2515 int ret;
2516 ssize_t send_ret;
d3ecc550
JD
2517 struct relay_session *session = conn->session;
2518 struct lttcomm_relayd_rotate_stream stream_info;
2519 struct lttcomm_relayd_generic_reply reply;
2520 struct relay_stream *stream;
5312a3ed
JG
2521 size_t header_len;
2522 size_t path_len;
2523 struct lttng_buffer_view new_path_view;
d3ecc550
JD
2524
2525 DBG("Rotate stream received");
2526
2527 if (!session || !conn->version_check_done) {
2528 ERR("Trying to rotate a stream before version check");
2529 ret = -1;
2530 goto end_no_reply;
2531 }
2532
2533 if (session->major == 2 && session->minor < 11) {
2534 ERR("Unsupported feature before 2.11");
2535 ret = -1;
2536 goto end_no_reply;
2537 }
2538
5312a3ed 2539 header_len = sizeof(struct lttcomm_relayd_rotate_stream);
d3ecc550 2540
5312a3ed
JG
2541 if (payload->size < header_len) {
2542 ERR("Unexpected payload size in \"relay_rotate_session_stream\": expected >= %zu bytes, got %zu bytes",
2543 header_len, payload->size);
d3ecc550
JD
2544 ret = -1;
2545 goto end_no_reply;
2546 }
2547
5312a3ed
JG
2548 memcpy(&stream_info, payload->data, header_len);
2549
2550 /* Convert to host */
2551 stream_info.pathname_length = be32toh(stream_info.pathname_length);
2552 stream_info.stream_id = be64toh(stream_info.stream_id);
2553 stream_info.new_chunk_id = be64toh(stream_info.new_chunk_id);
2554 stream_info.rotate_at_seq_num = be64toh(stream_info.rotate_at_seq_num);
d3ecc550 2555
5312a3ed
JG
2556 path_len = stream_info.pathname_length;
2557 if (payload->size < header_len + path_len) {
2558 ERR("Unexpected payload size in \"relay_rotate_session_stream\" including path: expected >= %zu bytes, got %zu bytes",
2559 header_len + path_len, payload->size);
2560 ret = -1;
2561 goto end_no_reply;
2562 }
2563
d3ecc550 2564 /* Ensure it fits in local filename length. */
5312a3ed 2565 if (path_len >= LTTNG_PATH_MAX) {
d3ecc550
JD
2566 ret = -ENAMETOOLONG;
2567 ERR("Length of relay_rotate_session_stream command's path name (%zu bytes) exceeds the maximal allowed length of %i bytes",
5312a3ed 2568 path_len, LTTNG_PATH_MAX);
d3ecc550
JD
2569 goto end;
2570 }
2571
5312a3ed
JG
2572 new_path_view = lttng_buffer_view_from_view(payload, header_len,
2573 stream_info.pathname_length);
d3ecc550 2574
5312a3ed
JG
2575 stream = stream_get_by_id(stream_info.stream_id);
2576 if (!stream) {
d3ecc550 2577 ret = -1;
5312a3ed 2578 goto end;
d3ecc550
JD
2579 }
2580
2581 pthread_mutex_lock(&stream->lock);
2582
2583 /*
2584 * Update the trace path (just the folder, the stream name does not
2585 * change).
2586 */
cb523e02
JG
2587 free(stream->prev_path_name);
2588 stream->prev_path_name = stream->path_name;
5312a3ed 2589 stream->path_name = create_output_path(new_path_view.data);
d3ecc550
JD
2590 if (!stream->path_name) {
2591 ERR("Failed to create a new output path");
5312a3ed 2592 ret = -1;
d3ecc550
JD
2593 goto end_stream_unlock;
2594 }
2595 ret = utils_mkdir_recursive(stream->path_name, S_IRWXU | S_IRWXG,
2596 -1, -1);
2597 if (ret < 0) {
2598 ERR("relay creating output directory");
5312a3ed 2599 ret = -1;
d3ecc550
JD
2600 goto end_stream_unlock;
2601 }
5312a3ed 2602
81164b6b
JG
2603 assert(stream->current_chunk_id.is_set);
2604 stream->current_chunk_id.value = stream_info.new_chunk_id;
d3ecc550
JD
2605
2606 if (stream->is_metadata) {
c6db3843
JG
2607 /*
2608 * Metadata streams have no index; consider its rotation
2609 * complete.
2610 */
2611 stream->index_rotated = true;
d3ecc550
JD
2612 /*
2613 * The metadata stream is sent only over the control connection
2614 * so we know we have all the data to perform the stream
2615 * rotation.
2616 */
c6db3843 2617 ret = do_rotate_stream_data(stream);
d3ecc550 2618 } else {
5312a3ed 2619 stream->rotate_at_seq_num = stream_info.rotate_at_seq_num;
c6db3843
JG
2620 ret = try_rotate_stream_data(stream);
2621 if (ret < 0) {
2622 goto end_stream_unlock;
2623 }
2624
2625 ret = try_rotate_stream_index(stream);
2626 if (ret < 0) {
2627 goto end_stream_unlock;
2628 }
d3ecc550
JD
2629 }
2630
2631end_stream_unlock:
2632 pthread_mutex_unlock(&stream->lock);
2633 stream_put(stream);
2634end:
2635 memset(&reply, 0, sizeof(reply));
2636 if (ret < 0) {
2637 reply.ret_code = htobe32(LTTNG_ERR_UNK);
2638 } else {
2639 reply.ret_code = htobe32(LTTNG_OK);
2640 }
2641 send_ret = conn->sock->ops->sendmsg(conn->sock, &reply,
2642 sizeof(struct lttcomm_relayd_generic_reply), 0);
5312a3ed
JG
2643 if (send_ret < (ssize_t) sizeof(reply)) {
2644 ERR("Failed to send \"rotate session stream\" command reply (ret = %zd)",
2645 send_ret);
2646 ret = -1;
d3ecc550
JD
2647 }
2648
2649end_no_reply:
d3ecc550
JD
2650 return ret;
2651}
2652
a1ae2ea5
JD
2653/*
2654 * relay_mkdir: Create a folder on the disk.
2655 */
5312a3ed
JG
2656static int relay_mkdir(const struct lttcomm_relayd_hdr *recv_hdr,
2657 struct relay_connection *conn,
2658 const struct lttng_buffer_view *payload)
a1ae2ea5
JD
2659{
2660 int ret;
a1ae2ea5
JD
2661 struct relay_session *session = conn->session;
2662 struct lttcomm_relayd_mkdir path_info_header;
a1ae2ea5
JD
2663 struct lttcomm_relayd_generic_reply reply;
2664 char *path = NULL;
5312a3ed
JG
2665 size_t header_len;
2666 ssize_t send_ret;
2667 struct lttng_buffer_view path_view;
a1ae2ea5
JD
2668
2669 if (!session || !conn->version_check_done) {
00fb02ac 2670 ERR("Trying to create a directory before version check");
a1ae2ea5
JD
2671 ret = -1;
2672 goto end_no_session;
2673 }
2674
2675 if (session->major == 2 && session->minor < 11) {
2676 /*
2677 * This client is not supposed to use this command since
2678 * it predates its introduction.
2679 */
2680 ERR("relay_mkdir command is unsupported before LTTng 2.11");
2681 ret = -1;
2682 goto end_no_session;
2683 }
2684
5312a3ed
JG
2685 header_len = sizeof(path_info_header);
2686 if (payload->size < header_len) {
2687 ERR("Unexpected payload size in \"relay_mkdir\": expected >= %zu bytes, got %zu bytes",
2688 header_len, payload->size);
a1ae2ea5
JD
2689 ret = -1;
2690 goto end_no_session;
2691 }
2692
5312a3ed
JG
2693 memcpy(&path_info_header, payload->data, header_len);
2694
a1ae2ea5
JD
2695 path_info_header.length = be32toh(path_info_header.length);
2696
5312a3ed
JG
2697 if (payload->size < header_len + path_info_header.length) {
2698 ERR("Unexpected payload size in \"relay_mkdir\" including path: expected >= %zu bytes, got %zu bytes",
2699 header_len + path_info_header.length, payload->size);
2700 ret = -1;
2701 goto end_no_session;
2702 }
2703
a1ae2ea5
JD
2704 /* Ensure that it fits in local path length. */
2705 if (path_info_header.length >= LTTNG_PATH_MAX) {
2706 ret = -ENAMETOOLONG;
2707 ERR("Path name argument of mkdir command (%" PRIu32 " bytes) exceeds the maximal length allowed (%d bytes)",
2708 path_info_header.length, LTTNG_PATH_MAX);
2709 goto end;
2710 }
2711
5312a3ed
JG
2712 path_view = lttng_buffer_view_from_view(payload, header_len,
2713 path_info_header.length);
a1ae2ea5 2714
5312a3ed 2715 path = create_output_path(path_view.data);
a1ae2ea5
JD
2716 if (!path) {
2717 ERR("Failed to create output path");
2718 ret = -1;
2719 goto end;
2720 }
2721
116ffe56 2722 DBG("MKDIR command has path \"%s\", changed to \"%s\"", path_view.data, path);
a1ae2ea5
JD
2723 ret = utils_mkdir_recursive(path, S_IRWXU | S_IRWXG, -1, -1);
2724 if (ret < 0) {
2725 ERR("relay creating output directory");
2726 goto end;
2727 }
2728
2729 ret = 0;
2730
2731end:
2732 memset(&reply, 0, sizeof(reply));
2733 if (ret < 0) {
2734 reply.ret_code = htobe32(LTTNG_ERR_UNK);
2735 } else {
2736 reply.ret_code = htobe32(LTTNG_OK);
2737 }
5312a3ed
JG
2738 send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0);
2739 if (send_ret < (ssize_t) sizeof(reply)) {
2740 ERR("Failed to send \"mkdir\" command reply (ret = %zd)", send_ret);
a1ae2ea5
JD
2741 ret = -1;
2742 }
2743
2744end_no_session:
2745 free(path);
a1ae2ea5
JD
2746 return ret;
2747}
2748
00fb02ac
JD
2749static int validate_rotate_rename_path_length(const char *path_type,
2750 uint32_t path_length)
2751{
2752 int ret = 0;
2753
2754 if (path_length > LTTNG_PATH_MAX) {
2755 ret = -ENAMETOOLONG;
2756 ERR("rotate rename \"%s\" path name length (%" PRIu32 " bytes) exceeds the allowed size of %i bytes",
2757 path_type, path_length, LTTNG_PATH_MAX);
2758 } else if (path_length == 0) {
2759 ret = -EINVAL;
2760 ERR("rotate rename \"%s\" path name has an illegal length of 0", path_type);
2761 }
2762 return ret;
2763}
2764
2765/*
2766 * relay_rotate_rename: rename the trace folder after a rotation is
2767 * completed. We are not closing any fd here, just moving the folder, so it
2768 * works even if data is still in-flight.
2769 */
5312a3ed
JG
2770static int relay_rotate_rename(const struct lttcomm_relayd_hdr *recv_hdr,
2771 struct relay_connection *conn,
2772 const struct lttng_buffer_view *payload)
00fb02ac
JD
2773{
2774 int ret;
5312a3ed 2775 ssize_t send_ret;
00fb02ac
JD
2776 struct relay_session *session = conn->session;
2777 struct lttcomm_relayd_generic_reply reply;
2778 struct lttcomm_relayd_rotate_rename header;
5312a3ed 2779 size_t header_len;
00fb02ac 2780 size_t received_paths_size;
00fb02ac 2781 char *complete_old_path = NULL, *complete_new_path = NULL;
5312a3ed
JG
2782 struct lttng_buffer_view old_path_view;
2783 struct lttng_buffer_view new_path_view;
00fb02ac
JD
2784
2785 if (!session || !conn->version_check_done) {
2786 ERR("Trying to rename a trace folder before version check");
2787 ret = -1;
2788 goto end_no_reply;
2789 }
2790
2791 if (session->major == 2 && session->minor < 11) {
2792 ERR("relay_rotate_rename command is unsupported before LTTng 2.11");
2793 ret = -1;
2794 goto end_no_reply;
2795 }
2796
5312a3ed
JG
2797 header_len = sizeof(header);
2798 if (payload->size < header_len) {
2799 ERR("Unexpected payload size in \"relay_rotate_rename\": expected >= %zu bytes, got %zu bytes",
2800 header_len, payload->size);
00fb02ac
JD
2801 ret = -1;
2802 goto end_no_reply;
2803 }
2804
5312a3ed
JG
2805 memcpy(&header, payload->data, header_len);
2806
00fb02ac
JD
2807 header.old_path_length = be32toh(header.old_path_length);
2808 header.new_path_length = be32toh(header.new_path_length);
2809 received_paths_size = header.old_path_length + header.new_path_length;
2810
5312a3ed
JG
2811 if (payload->size < header_len + received_paths_size) {
2812 ERR("Unexpected payload size in \"relay_rotate_rename\" including paths: expected >= %zu bytes, got %zu bytes",
2813 header_len, payload->size);
2814 ret = -1;
2815 goto end_no_reply;
2816 }
2817
00fb02ac
JD
2818 /* Ensure the paths don't exceed their allowed size. */
2819 ret = validate_rotate_rename_path_length("old", header.old_path_length);
2820 if (ret) {
2821 goto end;
2822 }
2823 ret = validate_rotate_rename_path_length("new", header.new_path_length);
2824 if (ret) {
2825 goto end;
2826 }
2827
5312a3ed
JG
2828 old_path_view = lttng_buffer_view_from_view(payload, header_len,
2829 header.old_path_length);
2830 new_path_view = lttng_buffer_view_from_view(payload,
2831 header_len + header.old_path_length,
2832 header.new_path_length);
00fb02ac
JD
2833
2834 /* Validate that both paths received are NULL terminated. */
5312a3ed 2835 if (old_path_view.data[old_path_view.size - 1] != '\0') {
00fb02ac
JD
2836 ERR("relay_rotate_rename command's \"old\" path is invalid (not NULL terminated)");
2837 ret = -1;
2838 goto end;
2839 }
5312a3ed 2840 if (new_path_view.data[new_path_view.size - 1] != '\0') {
00fb02ac
JD
2841 ERR("relay_rotate_rename command's \"new\" path is invalid (not NULL terminated)");
2842 ret = -1;
2843 goto end;
2844 }
2845
116ffe56
JG
2846 DBG("ROTATE_RENAME command has argument old path = \"%s\", new_path = \"%s\"",
2847 old_path_view.data, new_path_view.data);
5312a3ed 2848 complete_old_path = create_output_path(old_path_view.data);
00fb02ac
JD
2849 if (!complete_old_path) {
2850 ERR("Failed to build old output path in rotate_rename command");
2851 ret = -1;
2852 goto end;
2853 }
2854
5312a3ed 2855 complete_new_path = create_output_path(new_path_view.data);
00fb02ac
JD
2856 if (!complete_new_path) {
2857 ERR("Failed to build new output path in rotate_rename command");
2858 ret = -1;
2859 goto end;
2860 }
116ffe56
JG
2861 DBG("Expanded ROTATE_RENAME arguments to old path = \"%s\", new_path = \"%s\"",
2862 complete_old_path, complete_new_path);
00fb02ac
JD
2863
2864 ret = utils_mkdir_recursive(complete_new_path, S_IRWXU | S_IRWXG,
2865 -1, -1);
2866 if (ret < 0) {
2867 ERR("Failed to mkdir() rotate_rename's \"new\" output directory at \"%s\"",
2868 complete_new_path);
2869 goto end;
2870 }
2871
2872 /*
2873 * If a domain has not yet created its channel, the domain-specific
2874 * folder might not exist, but this is not an error.
2875 */
2876 ret = rename(complete_old_path, complete_new_path);
2877 if (ret < 0 && errno != ENOENT) {
2878 PERROR("Renaming chunk in rotate_rename command from \"%s\" to \"%s\"",
2879 complete_old_path, complete_new_path);
2880 goto end;
2881 }
2882 ret = 0;
2883
2884end:
2885 memset(&reply, 0, sizeof(reply));
2886 if (ret < 0) {
2887 reply.ret_code = htobe32(LTTNG_ERR_UNK);
2888 } else {
2889 reply.ret_code = htobe32(LTTNG_OK);
2890 }
5312a3ed
JG
2891 send_ret = conn->sock->ops->sendmsg(conn->sock, &reply,
2892 sizeof(reply), 0);
2893 if (send_ret < sizeof(reply)) {
2894 ERR("Failed to send \"rotate rename\" command reply (ret = %zd)",
2895 send_ret);
00fb02ac
JD
2896 ret = -1;
2897 }
2898
2899end_no_reply:
00fb02ac
JD
2900 free(complete_old_path);
2901 free(complete_new_path);
2902 return ret;
2903}
2904
e2acc8d2
JD
2905/*
2906 * Check if all the streams in the session have completed the last rotation.
2907 * The chunk_id value is used to distinguish the cases where a stream was
2908 * closed on the consumerd before the rotation started but it still active on
2909 * the relayd, and the case where a stream appeared on the consumerd/relayd
2910 * after the last rotation started (in that case, it is already writing in the
2911 * new chunk folder).
2912 */
2913static
5312a3ed
JG
2914int relay_rotate_pending(const struct lttcomm_relayd_hdr *recv_hdr,
2915 struct relay_connection *conn,
2916 const struct lttng_buffer_view *payload)
e2acc8d2
JD
2917{
2918 struct relay_session *session = conn->session;
2919 struct lttcomm_relayd_rotate_pending msg;
d88744a4 2920 struct lttcomm_relayd_rotate_pending_reply reply;
e2acc8d2
JD
2921 struct lttng_ht_iter iter;
2922 struct relay_stream *stream;
4f5fb4c3 2923 int ret = 0;
5312a3ed 2924 ssize_t send_ret;
e2acc8d2
JD
2925 uint64_t chunk_id;
2926 bool rotate_pending = false;
2927
2928 DBG("Rotate pending command received");
2929
2930 if (!session || !conn->version_check_done) {
2931 ERR("Trying to check for data before version check");
2932 ret = -1;
2933 goto end_no_reply;
2934 }
2935
2936 if (session->major == 2 && session->minor < 11) {
2937 ERR("Unsupported feature before 2.11");
2938 ret = -1;
2939 goto end_no_reply;
2940 }
2941
5312a3ed
JG
2942 if (payload->size < sizeof(msg)) {
2943 ERR("Unexpected payload size in \"relay_rotate_pending\": expected >= %zu bytes, got %zu bytes",
2944 sizeof(msg), payload->size);
e2acc8d2
JD
2945 ret = -1;
2946 goto end_no_reply;
2947 }
2948
5312a3ed
JG
2949 memcpy(&msg, payload->data, sizeof(msg));
2950
e2acc8d2 2951 chunk_id = be64toh(msg.chunk_id);
5312a3ed 2952
ce6877c9
JG
2953 DBG("Evaluating rotate pending for session \"%s\" and chunk id %" PRIu64,
2954 session->session_name, chunk_id);
e2acc8d2
JD
2955
2956 /*
2957 * Iterate over all the streams in the session and check if they are
2958 * still waiting for data to perform their rotation.
2959 */
2960 rcu_read_lock();
2961 cds_lfht_for_each_entry(relay_streams_ht->ht, &iter.iter, stream,
2962 node.node) {
2963 if (!stream_get(stream)) {
2964 continue;
2965 }
2966 if (stream->trace->session != session) {
2967 stream_put(stream);
2968 continue;
2969 }
2970 pthread_mutex_lock(&stream->lock);
2971 if (stream->rotate_at_seq_num != -1ULL) {
2972 /* We have not yet performed the rotation. */
2973 rotate_pending = true;
2974 DBG("Stream %" PRIu64 " is still rotating",
2975 stream->stream_handle);
81164b6b 2976 } else if (stream->current_chunk_id.value < chunk_id) {
e2acc8d2
JD
2977 /*
2978 * Stream closed on the consumer but still active on the
2979 * relay.
2980 */
2981 rotate_pending = true;
2982 DBG("Stream %" PRIu64 " did not exist on the consumer "
2983 "when the last rotation started, but is"
2984 "still waiting for data before getting"
2985 "closed",
2986 stream->stream_handle);
2987 }
2988 pthread_mutex_unlock(&stream->lock);
2989 stream_put(stream);
2990 if (rotate_pending) {
2991 goto send_reply;
2992 }
2993 }
2994
2995send_reply:
2996 rcu_read_unlock();
2997 memset(&reply, 0, sizeof(reply));
d88744a4
JD
2998 reply.generic.ret_code = htobe32((uint32_t) LTTNG_OK);
2999 reply.is_pending = (uint8_t) !!rotate_pending;
5312a3ed 3000 send_ret = conn->sock->ops->sendmsg(conn->sock, &reply,
e2acc8d2 3001 sizeof(reply), 0);
5312a3ed
JG
3002 if (send_ret < (ssize_t) sizeof(reply)) {
3003 ERR("Failed to send \"rotate pending\" command reply (ret = %zd)",
3004 send_ret);
e2acc8d2
JD
3005 ret = -1;
3006 }
3007
3008end_no_reply:
3009 return ret;
3010}
3011
5312a3ed
JG
3012#define DBG_CMD(cmd_name, conn) \
3013 DBG3("Processing \"%s\" command for socket %i", cmd_name, conn->sock->fd);
3014
3015static int relay_process_control_command(struct relay_connection *conn,
3016 const struct lttcomm_relayd_hdr *header,
3017 const struct lttng_buffer_view *payload)
b8aa1682
JD
3018{
3019 int ret = 0;
3020
5312a3ed 3021 switch (header->cmd) {
b8aa1682 3022 case RELAYD_CREATE_SESSION:
5312a3ed
JG
3023 DBG_CMD("RELAYD_CREATE_SESSION", conn);
3024 ret = relay_create_session(header, conn, payload);
b8aa1682 3025 break;
b8aa1682 3026 case RELAYD_ADD_STREAM:
5312a3ed
JG
3027 DBG_CMD("RELAYD_ADD_STREAM", conn);
3028 ret = relay_add_stream(header, conn, payload);
b8aa1682
JD
3029 break;
3030 case RELAYD_START_DATA:
5312a3ed
JG
3031 DBG_CMD("RELAYD_START_DATA", conn);
3032 ret = relay_start(header, conn, payload);
b8aa1682
JD
3033 break;
3034 case RELAYD_SEND_METADATA:
5312a3ed
JG
3035 DBG_CMD("RELAYD_SEND_METADATA", conn);
3036 ret = relay_recv_metadata(header, conn, payload);
b8aa1682
JD
3037 break;
3038 case RELAYD_VERSION:
5312a3ed
JG
3039 DBG_CMD("RELAYD_VERSION", conn);
3040 ret = relay_send_version(header, conn, payload);
b8aa1682 3041 break;
173af62f 3042 case RELAYD_CLOSE_STREAM:
5312a3ed
JG
3043 DBG_CMD("RELAYD_CLOSE_STREAM", conn);
3044 ret = relay_close_stream(header, conn, payload);
173af62f 3045 break;
6d805429 3046 case RELAYD_DATA_PENDING:
5312a3ed
JG
3047 DBG_CMD("RELAYD_DATA_PENDING", conn);
3048 ret = relay_data_pending(header, conn, payload);
c8f59ee5
DG
3049 break;
3050 case RELAYD_QUIESCENT_CONTROL:
5312a3ed
JG
3051 DBG_CMD("RELAYD_QUIESCENT_CONTROL", conn);
3052 ret = relay_quiescent_control(header, conn, payload);
c8f59ee5 3053 break;
f7079f67 3054 case RELAYD_BEGIN_DATA_PENDING:
5312a3ed
JG
3055 DBG_CMD("RELAYD_BEGIN_DATA_PENDING", conn);
3056 ret = relay_begin_data_pending(header, conn, payload);
f7079f67
DG
3057 break;
3058 case RELAYD_END_DATA_PENDING:
5312a3ed
JG
3059 DBG_CMD("RELAYD_END_DATA_PENDING", conn);
3060 ret = relay_end_data_pending(header, conn, payload);
f7079f67 3061 break;
1c20f0e2 3062 case RELAYD_SEND_INDEX:
5312a3ed
JG
3063 DBG_CMD("RELAYD_SEND_INDEX", conn);
3064 ret = relay_recv_index(header, conn, payload);
1c20f0e2 3065 break;
a4baae1b 3066 case RELAYD_STREAMS_SENT:
5312a3ed
JG
3067 DBG_CMD("RELAYD_STREAMS_SENT", conn);
3068 ret = relay_streams_sent(header, conn, payload);
a4baae1b 3069 break;
93ec662e 3070 case RELAYD_RESET_METADATA:
5312a3ed
JG
3071 DBG_CMD("RELAYD_RESET_METADATA", conn);
3072 ret = relay_reset_metadata(header, conn, payload);
93ec662e 3073 break;
d3ecc550 3074 case RELAYD_ROTATE_STREAM:
5312a3ed
JG
3075 DBG_CMD("RELAYD_ROTATE_STREAM", conn);
3076 ret = relay_rotate_session_stream(header, conn, payload);
d3ecc550 3077 break;
00fb02ac 3078 case RELAYD_ROTATE_RENAME:
5312a3ed
JG
3079 DBG_CMD("RELAYD_ROTATE_RENAME", conn);
3080 ret = relay_rotate_rename(header, conn, payload);
00fb02ac 3081 break;
e2acc8d2 3082 case RELAYD_ROTATE_PENDING:
5312a3ed
JG
3083 DBG_CMD("RELAYD_ROTATE_PENDING", conn);
3084 ret = relay_rotate_pending(header, conn, payload);
e2acc8d2 3085 break;
a1ae2ea5 3086 case RELAYD_MKDIR:
5312a3ed
JG
3087 DBG_CMD("RELAYD_MKDIR", conn);
3088 ret = relay_mkdir(header, conn, payload);
a1ae2ea5 3089 break;
b8aa1682
JD
3090 case RELAYD_UPDATE_SYNC_INFO:
3091 default:
5312a3ed 3092 ERR("Received unknown command (%u)", header->cmd);
58eb9381 3093 relay_unknown_command(conn);
b8aa1682
JD
3094 ret = -1;
3095 goto end;
3096 }
3097
3098end:
3099 return ret;
3100}
3101
5569b118
JG
3102static enum relay_connection_status relay_process_control_receive_payload(
3103 struct relay_connection *conn)
5312a3ed
JG
3104{
3105 int ret = 0;
5569b118 3106 enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK;
5312a3ed
JG
3107 struct lttng_dynamic_buffer *reception_buffer =
3108 &conn->protocol.ctrl.reception_buffer;
3109 struct ctrl_connection_state_receive_payload *state =
3110 &conn->protocol.ctrl.state.receive_payload;
3111 struct lttng_buffer_view payload_view;
3112
3113 if (state->left_to_receive == 0) {
3114 /* Short-circuit for payload-less commands. */
3115 goto reception_complete;
3116 }
3117
3118 ret = conn->sock->ops->recvmsg(conn->sock,
3119 reception_buffer->data + state->received,
3120 state->left_to_receive, MSG_DONTWAIT);
3121 if (ret < 0) {
5569b118
JG
3122 if (errno != EAGAIN && errno != EWOULDBLOCK) {
3123 PERROR("Unable to receive command payload on sock %d",
3124 conn->sock->fd);
3125 status = RELAY_CONNECTION_STATUS_ERROR;
3126 }
5312a3ed
JG
3127 goto end;
3128 } else if (ret == 0) {
3129 DBG("Socket %d performed an orderly shutdown (received EOF)", conn->sock->fd);
5569b118 3130 status = RELAY_CONNECTION_STATUS_CLOSED;
5312a3ed
JG
3131 goto end;
3132 }
3133
3134 assert(ret > 0);
3135 assert(ret <= state->left_to_receive);
3136
3137 state->left_to_receive -= ret;
3138 state->received += ret;
3139
3140 if (state->left_to_receive > 0) {
3141 /*
3142 * Can't transition to the protocol's next state, wait to
3143 * receive the rest of the header.
3144 */
3145 DBG3("Partial reception of control connection protocol payload (received %" PRIu64 " bytes, %" PRIu64 " bytes left to receive, fd = %i)",
3146 state->received, state->left_to_receive,
3147 conn->sock->fd);
5312a3ed
JG
3148 goto end;
3149 }
3150
3151reception_complete:
3152 DBG("Done receiving control command payload: fd = %i, payload size = %" PRIu64 " bytes",
3153 conn->sock->fd, state->received);
3154 /*
3155 * The payload required to process the command has been received.
3156 * A view to the reception buffer is forwarded to the various
3157 * commands and the state of the control is reset on success.
3158 *
3159 * Commands are responsible for sending their reply to the peer.
3160 */
3161 payload_view = lttng_buffer_view_from_dynamic_buffer(reception_buffer,
3162 0, -1);
3163 ret = relay_process_control_command(conn,
3164 &state->header, &payload_view);
3165 if (ret < 0) {
5569b118 3166 status = RELAY_CONNECTION_STATUS_ERROR;
5312a3ed
JG
3167 goto end;
3168 }
3169
3170 ret = connection_reset_protocol_state(conn);
5569b118
JG
3171 if (ret) {
3172 status = RELAY_CONNECTION_STATUS_ERROR;
3173 }
5312a3ed 3174end:
5569b118 3175 return status;
5312a3ed
JG
3176}
3177
5569b118
JG
3178static enum relay_connection_status relay_process_control_receive_header(
3179 struct relay_connection *conn)
5312a3ed
JG
3180{
3181 int ret = 0;
5569b118 3182 enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK;
5312a3ed
JG
3183 struct lttcomm_relayd_hdr header;
3184 struct lttng_dynamic_buffer *reception_buffer =
3185 &conn->protocol.ctrl.reception_buffer;
3186 struct ctrl_connection_state_receive_header *state =
3187 &conn->protocol.ctrl.state.receive_header;
3188
3189 assert(state->left_to_receive != 0);
3190
3191 ret = conn->sock->ops->recvmsg(conn->sock,
3192 reception_buffer->data + state->received,
3193 state->left_to_receive, MSG_DONTWAIT);
3194 if (ret < 0) {
5569b118
JG
3195 if (errno != EAGAIN && errno != EWOULDBLOCK) {
3196 PERROR("Unable to receive control command header on sock %d",
3197 conn->sock->fd);
3198 status = RELAY_CONNECTION_STATUS_ERROR;
3199 }
5312a3ed
JG
3200 goto end;
3201 } else if (ret == 0) {
3202 DBG("Socket %d performed an orderly shutdown (received EOF)", conn->sock->fd);
5569b118 3203 status = RELAY_CONNECTION_STATUS_CLOSED;
5312a3ed
JG
3204 goto end;
3205 }
3206
3207 assert(ret > 0);
3208 assert(ret <= state->left_to_receive);
3209
3210 state->left_to_receive -= ret;
3211 state->received += ret;
3212
3213 if (state->left_to_receive > 0) {
3214 /*
3215 * Can't transition to the protocol's next state, wait to
3216 * receive the rest of the header.
3217 */
3218 DBG3("Partial reception of control connection protocol header (received %" PRIu64 " bytes, %" PRIu64 " bytes left to receive, fd = %i)",
3219 state->received, state->left_to_receive,
3220 conn->sock->fd);
5312a3ed
JG
3221 goto end;
3222 }
3223
3224 /* Transition to next state: receiving the command's payload. */
3225 conn->protocol.ctrl.state_id =
3226 CTRL_CONNECTION_STATE_RECEIVE_PAYLOAD;
3227 memcpy(&header, reception_buffer->data, sizeof(header));
3228 header.circuit_id = be64toh(header.circuit_id);
3229 header.data_size = be64toh(header.data_size);
3230 header.cmd = be32toh(header.cmd);
3231 header.cmd_version = be32toh(header.cmd_version);
3232 memcpy(&conn->protocol.ctrl.state.receive_payload.header,
3233 &header, sizeof(header));
3234
3235 DBG("Done receiving control command header: fd = %i, cmd = %" PRIu32 ", cmd_version = %" PRIu32 ", payload size = %" PRIu64 " bytes",
3236 conn->sock->fd, header.cmd, header.cmd_version,
3237 header.data_size);
3238
715e6fb1 3239 if (header.data_size > DEFAULT_NETWORK_RELAYD_CTRL_MAX_PAYLOAD_SIZE) {
5312a3ed
JG
3240 ERR("Command header indicates a payload (%" PRIu64 " bytes) that exceeds the maximal payload size allowed on a control connection.",
3241 header.data_size);
5569b118 3242 status = RELAY_CONNECTION_STATUS_ERROR;
5312a3ed
JG
3243 goto end;
3244 }
3245