Commit | Line | Data |
---|---|---|
b8aa1682 | 1 | /* |
ab5be9fa MJ |
2 | * Copyright (C) 2012 Julien Desfossez <jdesfossez@efficios.com> |
3 | * Copyright (C) 2012 David Goulet <dgoulet@efficios.com> | |
4 | * Copyright (C) 2013 Jérémie Galarneau <jeremie.galarneau@efficios.com> | |
5 | * Copyright (C) 2015 Mathieu Desnoyers <mathieu.desnoyers@efficios.com> | |
b8aa1682 | 6 | * |
ab5be9fa | 7 | * SPDX-License-Identifier: GPL-2.0-only |
b8aa1682 | 8 | * |
b8aa1682 JD |
9 | */ |
10 | ||
6c1c0768 | 11 | #define _LGPL_SOURCE |
c9e313bc SM |
12 | #include "backward-compatibility-group-by.hpp" |
13 | #include "cmd.hpp" | |
14 | #include "connection.hpp" | |
15 | #include "ctf-trace.hpp" | |
16 | #include "health-relayd.hpp" | |
17 | #include "index.hpp" | |
18 | #include "live.hpp" | |
19 | #include "lttng-relayd.hpp" | |
20 | #include "session.hpp" | |
21 | #include "sessiond-trace-chunks.hpp" | |
22 | #include "stream.hpp" | |
23 | #include "tcp_keep_alive.hpp" | |
24 | #include "testpoint.hpp" | |
25 | #include "tracefile-array.hpp" | |
26 | #include "utils.hpp" | |
27 | #include "version.hpp" | |
98b82dfa | 28 | #include "viewer-session.hpp" |
c9e313bc | 29 | #include "viewer-stream.hpp" |
b8aa1682 | 30 | |
28ab034a JG |
31 | #include <common/align.hpp> |
32 | #include <common/buffer-view.hpp> | |
33 | #include <common/common.hpp> | |
34 | #include <common/compat/endian.hpp> | |
35 | #include <common/compat/getenv.hpp> | |
36 | #include <common/compat/poll.hpp> | |
37 | #include <common/compat/socket.hpp> | |
38 | #include <common/daemonize.hpp> | |
39 | #include <common/defaults.hpp> | |
40 | #include <common/dynamic-buffer.hpp> | |
41 | #include <common/fd-tracker/fd-tracker.hpp> | |
42 | #include <common/fd-tracker/utils.hpp> | |
43 | #include <common/futex.hpp> | |
44 | #include <common/ini-config/ini-config.hpp> | |
45 | #include <common/path.hpp> | |
c7b9f8c3 | 46 | #include <common/pthread-lock.hpp> |
28ab034a JG |
47 | #include <common/sessiond-comm/inet.hpp> |
48 | #include <common/sessiond-comm/relayd.hpp> | |
49 | #include <common/sessiond-comm/sessiond-comm.hpp> | |
50 | #include <common/string-utils/format.hpp> | |
56047f5a | 51 | #include <common/urcu.hpp> |
28ab034a JG |
52 | #include <common/uri.hpp> |
53 | #include <common/utils.hpp> | |
54 | ||
55 | #include <lttng/lttng.h> | |
56 | ||
57 | #include <algorithm> | |
58 | #include <ctype.h> | |
59 | #include <fcntl.h> | |
60 | #include <getopt.h> | |
61 | #include <grp.h> | |
62 | #include <inttypes.h> | |
63 | #include <limits.h> | |
64 | #include <pthread.h> | |
65 | #include <signal.h> | |
66 | #include <stdio.h> | |
67 | #include <stdlib.h> | |
68 | #include <string.h> | |
69 | #include <strings.h> | |
70 | #include <sys/mman.h> | |
71 | #include <sys/mount.h> | |
72 | #include <sys/resource.h> | |
73 | #include <sys/socket.h> | |
74 | #include <sys/stat.h> | |
75 | #include <sys/types.h> | |
76 | #include <sys/wait.h> | |
77 | #include <unistd.h> | |
78 | #include <urcu/futex.h> | |
79 | #include <urcu/rculist.h> | |
80 | #include <urcu/uatomic.h> | |
81 | ||
4fc83d94 PP |
82 | static const char *help_msg = |
83 | #ifdef LTTNG_EMBED_HELP | |
84 | #include <lttng-relayd.8.h> | |
85 | #else | |
cd9adb8b | 86 | nullptr |
4fc83d94 | 87 | #endif |
28ab034a | 88 | ; |
4fc83d94 | 89 | |
5569b118 JG |
90 | enum relay_connection_status { |
91 | RELAY_CONNECTION_STATUS_OK, | |
a9577b76 | 92 | /* An error occurred while processing an event on the connection. */ |
5569b118 JG |
93 | RELAY_CONNECTION_STATUS_ERROR, |
94 | /* Connection closed/shutdown cleanly. */ | |
95 | RELAY_CONNECTION_STATUS_CLOSED, | |
96 | }; | |
97 | ||
b8aa1682 | 98 | /* command line options */ |
ce9ee1fb | 99 | char *opt_output_path, *opt_working_directory; |
35ab25e5 | 100 | static int opt_daemon, opt_background, opt_print_version, opt_allow_clear = 1; |
a8b66566 | 101 | enum relay_group_output_by opt_group_output_by = RELAYD_GROUP_OUTPUT_BY_UNKNOWN; |
3fd27398 | 102 | |
e56e5792 | 103 | /* Argument variables */ |
28ab034a JG |
104 | int lttng_opt_quiet; /* not static in error.h */ |
105 | int lttng_opt_verbose; /* not static in error.h */ | |
106 | int lttng_opt_mi; /* not static in error.h */ | |
e56e5792 | 107 | |
3fd27398 MD |
108 | /* |
109 | * We need to wait for listener and live listener threads, as well as | |
110 | * health check thread, before being ready to signal readiness. | |
111 | */ | |
28ab034a | 112 | #define NR_LTTNG_RELAY_READY 3 |
3fd27398 | 113 | static int lttng_relay_ready = NR_LTTNG_RELAY_READY; |
0848dba7 MD |
114 | |
115 | /* Size of receive buffer. */ | |
28ab034a | 116 | #define RECV_DATA_BUFFER_SIZE 65536 |
0848dba7 | 117 | |
28ab034a JG |
118 | static int recv_child_signal; /* Set to 1 when a SIGUSR1 signal is received. */ |
119 | static pid_t child_ppid; /* Internal parent PID use with daemonize. */ | |
3fd27398 | 120 | |
095a4ae5 MD |
121 | static struct lttng_uri *control_uri; |
122 | static struct lttng_uri *data_uri; | |
d3e2ba59 | 123 | static struct lttng_uri *live_uri; |
b8aa1682 JD |
124 | |
125 | const char *progname; | |
b8aa1682 | 126 | |
65931c8b | 127 | const char *tracing_group_name = DEFAULT_TRACING_GROUP; |
cd60b05a JG |
128 | static int tracing_group_name_override; |
129 | ||
28ab034a | 130 | const char *const config_section_name = "relayd"; |
65931c8b | 131 | |
b8aa1682 JD |
132 | /* |
133 | * This pipe is used to inform the worker thread that a command is queued and | |
134 | * ready to be processed. | |
135 | */ | |
58eb9381 | 136 | static int relay_conn_pipe[2] = { -1, -1 }; |
b8aa1682 | 137 | |
26c9d55e | 138 | /* Shared between threads */ |
b8aa1682 JD |
139 | static int dispatch_thread_exit; |
140 | ||
141 | static pthread_t listener_thread; | |
142 | static pthread_t dispatcher_thread; | |
143 | static pthread_t worker_thread; | |
65931c8b | 144 | static pthread_t health_thread; |
b8aa1682 | 145 | |
7591bab1 MD |
146 | /* |
147 | * last_relay_stream_id_lock protects last_relay_stream_id increment | |
148 | * atomicity on 32-bit architectures. | |
149 | */ | |
150 | static pthread_mutex_t last_relay_stream_id_lock = PTHREAD_MUTEX_INITIALIZER; | |
095a4ae5 | 151 | static uint64_t last_relay_stream_id; |
b8aa1682 JD |
152 | |
153 | /* | |
154 | * Relay command queue. | |
155 | * | |
156 | * The relay_thread_listener and relay_thread_dispatcher communicate with this | |
157 | * queue. | |
158 | */ | |
58eb9381 | 159 | static struct relay_conn_queue relay_conn_queue; |
b8aa1682 | 160 | |
896010e3 | 161 | /* Cap of file desriptors to be in simultaneous use by the relay daemon. */ |
5c0551f9 | 162 | static unsigned int lttng_opt_fd_pool_size = -1; |
896010e3 | 163 | |
d3e2ba59 JD |
164 | /* Global relay stream hash table. */ |
165 | struct lttng_ht *relay_streams_ht; | |
166 | ||
92c6ca54 DG |
167 | /* Global relay viewer stream hash table. */ |
168 | struct lttng_ht *viewer_streams_ht; | |
169 | ||
7591bab1 MD |
170 | /* Global relay sessions hash table. */ |
171 | struct lttng_ht *sessions_ht; | |
0a6518b0 | 172 | |
98b82dfa KS |
173 | /* Global viewer sessions hash table. */ |
174 | struct lttng_ht *viewer_sessions_ht; | |
175 | ||
55706a7d | 176 | /* Relayd health monitoring */ |
eea7556c | 177 | struct health_app *health_relayd; |
55706a7d | 178 | |
23c8ff50 JG |
179 | struct sessiond_trace_chunk_registry *sessiond_trace_chunk_registry; |
180 | ||
00e3b7f1 JG |
181 | /* Global fd tracker. */ |
182 | struct fd_tracker *the_fd_tracker; | |
183 | ||
cd60b05a | 184 | static struct option long_options[] = { |
28ab034a JG |
185 | { |
186 | "control-port", | |
187 | 1, | |
cd9adb8b | 188 | nullptr, |
28ab034a JG |
189 | 'C', |
190 | }, | |
191 | { | |
192 | "data-port", | |
193 | 1, | |
cd9adb8b | 194 | nullptr, |
28ab034a JG |
195 | 'D', |
196 | }, | |
197 | { | |
198 | "live-port", | |
199 | 1, | |
cd9adb8b | 200 | nullptr, |
28ab034a JG |
201 | 'L', |
202 | }, | |
203 | { | |
204 | "daemonize", | |
205 | 0, | |
cd9adb8b | 206 | nullptr, |
28ab034a JG |
207 | 'd', |
208 | }, | |
209 | { | |
210 | "background", | |
211 | 0, | |
cd9adb8b | 212 | nullptr, |
28ab034a JG |
213 | 'b', |
214 | }, | |
215 | { | |
216 | "group", | |
217 | 1, | |
cd9adb8b | 218 | nullptr, |
28ab034a JG |
219 | 'g', |
220 | }, | |
221 | { | |
222 | "fd-pool-size", | |
223 | 1, | |
cd9adb8b | 224 | nullptr, |
28ab034a JG |
225 | '\0', |
226 | }, | |
227 | { | |
228 | "help", | |
229 | 0, | |
cd9adb8b | 230 | nullptr, |
28ab034a JG |
231 | 'h', |
232 | }, | |
233 | { | |
234 | "output", | |
235 | 1, | |
cd9adb8b | 236 | nullptr, |
28ab034a JG |
237 | 'o', |
238 | }, | |
239 | { | |
240 | "verbose", | |
241 | 0, | |
cd9adb8b | 242 | nullptr, |
28ab034a JG |
243 | 'v', |
244 | }, | |
cd9adb8b JG |
245 | { "config", 1, nullptr, 'f' }, |
246 | { "version", 0, nullptr, 'V' }, | |
28ab034a JG |
247 | { |
248 | "working-directory", | |
249 | 1, | |
cd9adb8b | 250 | nullptr, |
28ab034a JG |
251 | 'w', |
252 | }, | |
253 | { | |
254 | "group-output-by-session", | |
255 | 0, | |
cd9adb8b | 256 | nullptr, |
28ab034a JG |
257 | 's', |
258 | }, | |
259 | { | |
260 | "group-output-by-host", | |
261 | 0, | |
cd9adb8b | 262 | nullptr, |
28ab034a JG |
263 | 'p', |
264 | }, | |
cd9adb8b | 265 | { "disallow-clear", 0, nullptr, 'x' }, |
28ab034a | 266 | { |
cd9adb8b | 267 | nullptr, |
28ab034a | 268 | 0, |
cd9adb8b | 269 | nullptr, |
28ab034a JG |
270 | 0, |
271 | }, | |
cd60b05a JG |
272 | }; |
273 | ||
3a904098 | 274 | static const char *config_ignore_options[] = { "help", "config", "version" }; |
cd60b05a | 275 | |
cd9adb8b | 276 | static void print_version() |
28ab034a | 277 | { |
a3bc3918 JR |
278 | fprintf(stdout, "%s\n", VERSION); |
279 | } | |
280 | ||
cd9adb8b | 281 | static void relayd_config_log() |
a3bc3918 JR |
282 | { |
283 | DBG("LTTng-relayd " VERSION " - " VERSION_NAME "%s%s", | |
28ab034a JG |
284 | GIT_VERSION[0] == '\0' ? "" : " - " GIT_VERSION, |
285 | EXTRA_VERSION_NAME[0] == '\0' ? "" : " - " EXTRA_VERSION_NAME); | |
a3bc3918 JR |
286 | if (EXTRA_VERSION_DESCRIPTION[0] != '\0') { |
287 | DBG("LTTng-relayd extra version description:\n\t" EXTRA_VERSION_DESCRIPTION "\n"); | |
288 | } | |
7f5ed73a JR |
289 | if (EXTRA_VERSION_PATCHES[0] != '\0') { |
290 | DBG("LTTng-relayd extra patches:\n\t" EXTRA_VERSION_PATCHES "\n"); | |
291 | } | |
a3bc3918 JR |
292 | } |
293 | ||
cd60b05a JG |
294 | /* |
295 | * Take an option from the getopt output and set it in the right variable to be | |
296 | * used later. | |
297 | * | |
298 | * Return 0 on success else a negative value. | |
299 | */ | |
7591bab1 | 300 | static int set_option(int opt, const char *arg, const char *optname) |
b8aa1682 | 301 | { |
cd60b05a JG |
302 | int ret; |
303 | ||
304 | switch (opt) { | |
305 | case 0: | |
5c0551f9 | 306 | if (!strcmp(optname, "fd-pool-size")) { |
896010e3 JG |
307 | unsigned long v; |
308 | ||
309 | errno = 0; | |
cd9adb8b | 310 | v = strtoul(arg, nullptr, 0); |
60b7e1f8 | 311 | if (errno != 0 || !isdigit((unsigned char) arg[0])) { |
5c0551f9 | 312 | ERR("Wrong value in --fd-pool-size parameter: %s", arg); |
896010e3 JG |
313 | ret = -1; |
314 | goto end; | |
315 | } | |
896010e3 | 316 | if (v >= UINT_MAX) { |
28ab034a JG |
317 | ERR("File descriptor cap overflow in --fd-pool-size parameter: %s", |
318 | arg); | |
896010e3 JG |
319 | ret = -1; |
320 | goto end; | |
321 | } | |
5c0551f9 | 322 | lttng_opt_fd_pool_size = (unsigned int) v; |
896010e3 JG |
323 | } else { |
324 | fprintf(stderr, "unknown option %s", optname); | |
325 | if (arg) { | |
326 | fprintf(stderr, " with arg %s\n", arg); | |
327 | } | |
cd60b05a JG |
328 | } |
329 | break; | |
330 | case 'C': | |
e8fa9fb0 MD |
331 | if (lttng_is_setuid_setgid()) { |
332 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
28ab034a | 333 | "-C, --control-port"); |
e8fa9fb0 MD |
334 | } else { |
335 | ret = uri_parse(arg, &control_uri); | |
336 | if (ret < 0) { | |
337 | ERR("Invalid control URI specified"); | |
338 | goto end; | |
339 | } | |
340 | if (control_uri->port == 0) { | |
341 | control_uri->port = DEFAULT_NETWORK_CONTROL_PORT; | |
342 | } | |
cd60b05a JG |
343 | } |
344 | break; | |
345 | case 'D': | |
e8fa9fb0 MD |
346 | if (lttng_is_setuid_setgid()) { |
347 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
28ab034a | 348 | "-D, -data-port"); |
e8fa9fb0 MD |
349 | } else { |
350 | ret = uri_parse(arg, &data_uri); | |
351 | if (ret < 0) { | |
352 | ERR("Invalid data URI specified"); | |
353 | goto end; | |
354 | } | |
355 | if (data_uri->port == 0) { | |
356 | data_uri->port = DEFAULT_NETWORK_DATA_PORT; | |
357 | } | |
cd60b05a JG |
358 | } |
359 | break; | |
8d5c808e | 360 | case 'L': |
e8fa9fb0 MD |
361 | if (lttng_is_setuid_setgid()) { |
362 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
28ab034a | 363 | "-L, -live-port"); |
e8fa9fb0 MD |
364 | } else { |
365 | ret = uri_parse(arg, &live_uri); | |
366 | if (ret < 0) { | |
367 | ERR("Invalid live URI specified"); | |
368 | goto end; | |
369 | } | |
370 | if (live_uri->port == 0) { | |
371 | live_uri->port = DEFAULT_NETWORK_VIEWER_PORT; | |
372 | } | |
8d5c808e AM |
373 | } |
374 | break; | |
cd60b05a JG |
375 | case 'd': |
376 | opt_daemon = 1; | |
377 | break; | |
b5218ffb MD |
378 | case 'b': |
379 | opt_background = 1; | |
380 | break; | |
cd60b05a | 381 | case 'g': |
e8fa9fb0 MD |
382 | if (lttng_is_setuid_setgid()) { |
383 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
28ab034a | 384 | "-g, --group"); |
e8fa9fb0 MD |
385 | } else { |
386 | tracing_group_name = strdup(arg); | |
cd9adb8b | 387 | if (tracing_group_name == nullptr) { |
e8fa9fb0 MD |
388 | ret = -errno; |
389 | PERROR("strdup"); | |
390 | goto end; | |
391 | } | |
392 | tracing_group_name_override = 1; | |
330a40bb | 393 | } |
cd60b05a JG |
394 | break; |
395 | case 'h': | |
4fc83d94 | 396 | ret = utils_show_help(8, "lttng-relayd", help_msg); |
655b5cc1 | 397 | if (ret) { |
4fc83d94 | 398 | ERR("Cannot show --help for `lttng-relayd`"); |
655b5cc1 PP |
399 | perror("exec"); |
400 | } | |
cd60b05a | 401 | exit(EXIT_FAILURE); |
3a904098 | 402 | case 'V': |
a3bc3918 JR |
403 | opt_print_version = 1; |
404 | break; | |
cd60b05a | 405 | case 'o': |
e8fa9fb0 MD |
406 | if (lttng_is_setuid_setgid()) { |
407 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
28ab034a | 408 | "-o, --output"); |
e8fa9fb0 MD |
409 | } else { |
410 | ret = asprintf(&opt_output_path, "%s", arg); | |
411 | if (ret < 0) { | |
412 | ret = -errno; | |
413 | PERROR("asprintf opt_output_path"); | |
414 | goto end; | |
415 | } | |
cd60b05a JG |
416 | } |
417 | break; | |
ce9ee1fb JR |
418 | case 'w': |
419 | if (lttng_is_setuid_setgid()) { | |
420 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
28ab034a | 421 | "-w, --working-directory"); |
ce9ee1fb JR |
422 | } else { |
423 | ret = asprintf(&opt_working_directory, "%s", arg); | |
424 | if (ret < 0) { | |
425 | ret = -errno; | |
426 | PERROR("asprintf opt_working_directory"); | |
427 | goto end; | |
428 | } | |
429 | } | |
430 | break; | |
431 | ||
cd60b05a JG |
432 | case 'v': |
433 | /* Verbose level can increase using multiple -v */ | |
434 | if (arg) { | |
435 | lttng_opt_verbose = config_parse_value(arg); | |
436 | } else { | |
849e5b7b DG |
437 | /* Only 3 level of verbosity (-vvv). */ |
438 | if (lttng_opt_verbose < 3) { | |
439 | lttng_opt_verbose += 1; | |
440 | } | |
cd60b05a JG |
441 | } |
442 | break; | |
a8b66566 JR |
443 | case 's': |
444 | if (opt_group_output_by != RELAYD_GROUP_OUTPUT_BY_UNKNOWN) { | |
445 | ERR("Cannot set --group-output-by-session, another --group-output-by argument is present"); | |
446 | exit(EXIT_FAILURE); | |
447 | } | |
448 | opt_group_output_by = RELAYD_GROUP_OUTPUT_BY_SESSION; | |
449 | break; | |
450 | case 'p': | |
451 | if (opt_group_output_by != RELAYD_GROUP_OUTPUT_BY_UNKNOWN) { | |
452 | ERR("Cannot set --group-output-by-host, another --group-output-by argument is present"); | |
453 | exit(EXIT_FAILURE); | |
454 | } | |
455 | opt_group_output_by = RELAYD_GROUP_OUTPUT_BY_HOST; | |
456 | break; | |
35ab25e5 MD |
457 | case 'x': |
458 | /* Disallow clear */ | |
459 | opt_allow_clear = 0; | |
460 | break; | |
cd60b05a JG |
461 | default: |
462 | /* Unknown option or other error. | |
463 | * Error is printed by getopt, just return */ | |
464 | ret = -1; | |
465 | goto end; | |
466 | } | |
467 | ||
468 | /* All good. */ | |
469 | ret = 0; | |
470 | ||
471 | end: | |
472 | return ret; | |
473 | } | |
474 | ||
475 | /* | |
476 | * config_entry_handler_cb used to handle options read from a config file. | |
f40ef1d5 | 477 | * See config_entry_handler_cb comment in common/config/session-config.h for the |
cd60b05a JG |
478 | * return value conventions. |
479 | */ | |
f46376a1 | 480 | static int config_entry_handler(const struct config_entry *entry, |
28ab034a | 481 | void *unused __attribute__((unused))) |
cd60b05a JG |
482 | { |
483 | int ret = 0, i; | |
484 | ||
485 | if (!entry || !entry->name || !entry->value) { | |
486 | ret = -EINVAL; | |
487 | goto end; | |
488 | } | |
489 | ||
490 | /* Check if the option is to be ignored */ | |
491 | for (i = 0; i < sizeof(config_ignore_options) / sizeof(char *); i++) { | |
492 | if (!strcmp(entry->name, config_ignore_options[i])) { | |
493 | goto end; | |
494 | } | |
495 | } | |
496 | ||
497 | for (i = 0; i < (sizeof(long_options) / sizeof(struct option)) - 1; i++) { | |
498 | /* Ignore if entry name is not fully matched. */ | |
5c7248cd | 499 | if (strcmp(entry->name, long_options[i].name) != 0) { |
cd60b05a JG |
500 | continue; |
501 | } | |
502 | ||
503 | /* | |
7591bab1 MD |
504 | * If the option takes no argument on the command line, |
505 | * we have to check if the value is "true". We support | |
506 | * non-zero numeric values, true, on and yes. | |
cd60b05a JG |
507 | */ |
508 | if (!long_options[i].has_arg) { | |
509 | ret = config_parse_value(entry->value); | |
510 | if (ret <= 0) { | |
511 | if (ret) { | |
512 | WARN("Invalid configuration value \"%s\" for option %s", | |
28ab034a JG |
513 | entry->value, |
514 | entry->name); | |
cd60b05a JG |
515 | } |
516 | /* False, skip boolean config option. */ | |
517 | goto end; | |
518 | } | |
519 | } | |
520 | ||
521 | ret = set_option(long_options[i].val, entry->value, entry->name); | |
522 | goto end; | |
523 | } | |
524 | ||
28ab034a | 525 | WARN("Unrecognized option \"%s\" in daemon configuration file.", entry->name); |
cd60b05a JG |
526 | |
527 | end: | |
528 | return ret; | |
529 | } | |
530 | ||
cd9adb8b | 531 | static int parse_env_options() |
2a10de3b JR |
532 | { |
533 | int ret = 0; | |
cd9adb8b | 534 | char *value = nullptr; |
2a10de3b JR |
535 | |
536 | value = lttng_secure_getenv(DEFAULT_LTTNG_RELAYD_WORKING_DIRECTORY_ENV); | |
537 | if (value) { | |
538 | opt_working_directory = strdup(value); | |
539 | if (!opt_working_directory) { | |
28ab034a | 540 | ERR("Failed to allocate working directory string (\"%s\")", value); |
2a10de3b JR |
541 | ret = -1; |
542 | } | |
543 | } | |
544 | return ret; | |
545 | } | |
546 | ||
cd9adb8b | 547 | static int set_fd_pool_size() |
5c0551f9 JG |
548 | { |
549 | int ret = 0; | |
550 | struct rlimit rlimit; | |
551 | ||
552 | ret = getrlimit(RLIMIT_NOFILE, &rlimit); | |
553 | if (ret) { | |
554 | PERROR("Failed to get file descriptor limit"); | |
555 | ret = -1; | |
556 | goto end; | |
557 | } | |
558 | ||
559 | DBG("File descriptor count limits are %" PRIu64 " (soft) and %" PRIu64 " (hard)", | |
28ab034a JG |
560 | (uint64_t) rlimit.rlim_cur, |
561 | (uint64_t) rlimit.rlim_max); | |
5c0551f9 JG |
562 | if (lttng_opt_fd_pool_size == -1) { |
563 | /* Use default value (soft limit - reserve). */ | |
564 | if (rlimit.rlim_cur < DEFAULT_RELAYD_MIN_FD_POOL_SIZE) { | |
28ab034a JG |
565 | ERR("The process' file number limit is too low (%" PRIu64 |
566 | "). The process' file number limit must be set to at least %i.", | |
567 | (uint64_t) rlimit.rlim_cur, | |
568 | DEFAULT_RELAYD_MIN_FD_POOL_SIZE); | |
5c0551f9 JG |
569 | ret = -1; |
570 | goto end; | |
571 | } | |
28ab034a | 572 | lttng_opt_fd_pool_size = rlimit.rlim_cur - DEFAULT_RELAYD_FD_POOL_SIZE_RESERVE; |
5c0551f9 JG |
573 | goto end; |
574 | } | |
575 | ||
576 | if (lttng_opt_fd_pool_size < DEFAULT_RELAYD_MIN_FD_POOL_SIZE) { | |
577 | ERR("File descriptor pool size must be set to at least %d", | |
28ab034a | 578 | DEFAULT_RELAYD_MIN_FD_POOL_SIZE); |
5c0551f9 JG |
579 | ret = -1; |
580 | goto end; | |
581 | } | |
582 | ||
583 | if (lttng_opt_fd_pool_size > rlimit.rlim_cur) { | |
28ab034a JG |
584 | ERR("File descriptor pool size argument (%u) exceeds the process' soft limit (%" PRIu64 |
585 | ").", | |
586 | lttng_opt_fd_pool_size, | |
587 | (uint64_t) rlimit.rlim_cur); | |
5c0551f9 JG |
588 | ret = -1; |
589 | goto end; | |
590 | } | |
591 | ||
d49487dc | 592 | DBG("File descriptor pool size argument (%u) adjusted to %u to accommodates transient fd uses", |
28ab034a JG |
593 | lttng_opt_fd_pool_size, |
594 | lttng_opt_fd_pool_size - DEFAULT_RELAYD_FD_POOL_SIZE_RESERVE); | |
5c0551f9 JG |
595 | lttng_opt_fd_pool_size -= DEFAULT_RELAYD_FD_POOL_SIZE_RESERVE; |
596 | end: | |
597 | return ret; | |
598 | } | |
599 | ||
7591bab1 | 600 | static int set_options(int argc, char **argv) |
cd60b05a | 601 | { |
178a0557 | 602 | int c, ret = 0, option_index = 0, retval = 0; |
cd60b05a JG |
603 | int orig_optopt = optopt, orig_optind = optind; |
604 | char *default_address, *optstring; | |
cd9adb8b | 605 | char *config_path = nullptr; |
cd60b05a JG |
606 | |
607 | optstring = utils_generate_optstring(long_options, | |
28ab034a | 608 | sizeof(long_options) / sizeof(struct option)); |
cd60b05a | 609 | if (!optstring) { |
178a0557 | 610 | retval = -ENOMEM; |
cd60b05a JG |
611 | goto exit; |
612 | } | |
613 | ||
614 | /* Check for the --config option */ | |
615 | ||
28ab034a | 616 | while ((c = getopt_long(argc, argv, optstring, long_options, &option_index)) != -1) { |
cd60b05a | 617 | if (c == '?') { |
178a0557 | 618 | retval = -EINVAL; |
cd60b05a JG |
619 | goto exit; |
620 | } else if (c != 'f') { | |
621 | continue; | |
622 | } | |
623 | ||
e8fa9fb0 MD |
624 | if (lttng_is_setuid_setgid()) { |
625 | WARN("Getting '%s' argument from setuid/setgid binary refused for security reasons.", | |
28ab034a | 626 | "-f, --config"); |
e8fa9fb0 | 627 | } else { |
3a9e5d16 | 628 | free(config_path); |
e8fa9fb0 MD |
629 | config_path = utils_expand_path(optarg); |
630 | if (!config_path) { | |
631 | ERR("Failed to resolve path: %s", optarg); | |
632 | } | |
cd60b05a JG |
633 | } |
634 | } | |
635 | ||
28ab034a | 636 | ret = config_get_section_entries( |
cd9adb8b | 637 | config_path, config_section_name, config_entry_handler, nullptr); |
cd60b05a JG |
638 | if (ret) { |
639 | if (ret > 0) { | |
640 | ERR("Invalid configuration option at line %i", ret); | |
cd60b05a | 641 | } |
178a0557 | 642 | retval = -1; |
cd60b05a JG |
643 | goto exit; |
644 | } | |
b8aa1682 | 645 | |
cd60b05a JG |
646 | /* Reset getopt's global state */ |
647 | optopt = orig_optopt; | |
648 | optind = orig_optind; | |
cd9adb8b | 649 | while (true) { |
cd60b05a | 650 | c = getopt_long(argc, argv, optstring, long_options, &option_index); |
b8aa1682 JD |
651 | if (c == -1) { |
652 | break; | |
653 | } | |
654 | ||
cd60b05a JG |
655 | ret = set_option(c, optarg, long_options[option_index].name); |
656 | if (ret < 0) { | |
178a0557 | 657 | retval = -1; |
b8aa1682 JD |
658 | goto exit; |
659 | } | |
660 | } | |
661 | ||
662 | /* assign default values */ | |
cd9adb8b | 663 | if (control_uri == nullptr) { |
fa91dc52 | 664 | ret = asprintf(&default_address, |
28ab034a JG |
665 | "tcp://" DEFAULT_NETWORK_CONTROL_BIND_ADDRESS ":%d", |
666 | DEFAULT_NETWORK_CONTROL_PORT); | |
b8aa1682 JD |
667 | if (ret < 0) { |
668 | PERROR("asprintf default data address"); | |
178a0557 | 669 | retval = -1; |
b8aa1682 JD |
670 | goto exit; |
671 | } | |
672 | ||
673 | ret = uri_parse(default_address, &control_uri); | |
674 | free(default_address); | |
675 | if (ret < 0) { | |
676 | ERR("Invalid control URI specified"); | |
178a0557 | 677 | retval = -1; |
b8aa1682 JD |
678 | goto exit; |
679 | } | |
680 | } | |
cd9adb8b | 681 | if (data_uri == nullptr) { |
fa91dc52 | 682 | ret = asprintf(&default_address, |
28ab034a JG |
683 | "tcp://" DEFAULT_NETWORK_DATA_BIND_ADDRESS ":%d", |
684 | DEFAULT_NETWORK_DATA_PORT); | |
b8aa1682 JD |
685 | if (ret < 0) { |
686 | PERROR("asprintf default data address"); | |
178a0557 | 687 | retval = -1; |
b8aa1682 JD |
688 | goto exit; |
689 | } | |
690 | ||
691 | ret = uri_parse(default_address, &data_uri); | |
692 | free(default_address); | |
693 | if (ret < 0) { | |
694 | ERR("Invalid data URI specified"); | |
178a0557 | 695 | retval = -1; |
b8aa1682 JD |
696 | goto exit; |
697 | } | |
698 | } | |
cd9adb8b | 699 | if (live_uri == nullptr) { |
fa91dc52 | 700 | ret = asprintf(&default_address, |
28ab034a JG |
701 | "tcp://" DEFAULT_NETWORK_VIEWER_BIND_ADDRESS ":%d", |
702 | DEFAULT_NETWORK_VIEWER_PORT); | |
d3e2ba59 JD |
703 | if (ret < 0) { |
704 | PERROR("asprintf default viewer control address"); | |
178a0557 | 705 | retval = -1; |
d3e2ba59 JD |
706 | goto exit; |
707 | } | |
708 | ||
709 | ret = uri_parse(default_address, &live_uri); | |
710 | free(default_address); | |
711 | if (ret < 0) { | |
712 | ERR("Invalid viewer control URI specified"); | |
178a0557 | 713 | retval = -1; |
d3e2ba59 JD |
714 | goto exit; |
715 | } | |
716 | } | |
5c0551f9 JG |
717 | ret = set_fd_pool_size(); |
718 | if (ret) { | |
719 | retval = -1; | |
720 | goto exit; | |
896010e3 | 721 | } |
b8aa1682 | 722 | |
a8b66566 JR |
723 | if (opt_group_output_by == RELAYD_GROUP_OUTPUT_BY_UNKNOWN) { |
724 | opt_group_output_by = RELAYD_GROUP_OUTPUT_BY_HOST; | |
725 | } | |
35ab25e5 MD |
726 | if (opt_allow_clear) { |
727 | /* Check if env variable exists. */ | |
728 | const char *value = lttng_secure_getenv(DEFAULT_LTTNG_RELAYD_DISALLOW_CLEAR_ENV); | |
729 | if (value) { | |
730 | ret = config_parse_value(value); | |
731 | if (ret < 0) { | |
28ab034a JG |
732 | ERR("Invalid value for %s specified", |
733 | DEFAULT_LTTNG_RELAYD_DISALLOW_CLEAR_ENV); | |
35ab25e5 MD |
734 | retval = -1; |
735 | goto exit; | |
736 | } | |
737 | opt_allow_clear = !ret; | |
738 | } | |
739 | } | |
a8b66566 | 740 | |
b8aa1682 | 741 | exit: |
3a9e5d16 | 742 | free(config_path); |
cd60b05a | 743 | free(optstring); |
178a0557 | 744 | return retval; |
b8aa1682 JD |
745 | } |
746 | ||
cd9adb8b | 747 | static void print_global_objects() |
7591bab1 | 748 | { |
7591bab1 MD |
749 | print_viewer_streams(); |
750 | print_relay_streams(); | |
751 | print_sessions(); | |
7591bab1 MD |
752 | } |
753 | ||
28ab034a | 754 | static int noop_close(void *data __attribute__((unused)), int *fds __attribute__((unused))) |
5c0551f9 JG |
755 | { |
756 | return 0; | |
757 | } | |
758 | ||
cd9adb8b | 759 | static void untrack_stdio() |
5c0551f9 JG |
760 | { |
761 | int fds[] = { fileno(stdout), fileno(stderr) }; | |
762 | ||
763 | /* | |
764 | * noop_close is used since we don't really want to close | |
765 | * the stdio output fds; we merely want to stop tracking them. | |
766 | */ | |
cd9adb8b | 767 | (void) fd_tracker_close_unsuspendable_fd(the_fd_tracker, fds, 2, noop_close, nullptr); |
5c0551f9 JG |
768 | } |
769 | ||
b8aa1682 JD |
770 | /* |
771 | * Cleanup the daemon | |
772 | */ | |
cd9adb8b | 773 | static void relayd_cleanup() |
b8aa1682 | 774 | { |
7591bab1 MD |
775 | print_global_objects(); |
776 | ||
b8aa1682 JD |
777 | DBG("Cleaning up"); |
778 | ||
178a0557 MD |
779 | if (viewer_streams_ht) |
780 | lttng_ht_destroy(viewer_streams_ht); | |
98b82dfa KS |
781 | if (viewer_sessions_ht) { |
782 | lttng_ht_destroy(viewer_sessions_ht); | |
783 | } | |
178a0557 MD |
784 | if (relay_streams_ht) |
785 | lttng_ht_destroy(relay_streams_ht); | |
7591bab1 MD |
786 | if (sessions_ht) |
787 | lttng_ht_destroy(sessions_ht); | |
095a4ae5 | 788 | free(opt_output_path); |
ce9ee1fb | 789 | free(opt_working_directory); |
095a4ae5 | 790 | |
794e2e5f JG |
791 | if (health_relayd) { |
792 | health_app_destroy(health_relayd); | |
793 | } | |
a02de639 | 794 | /* Close thread quit pipes */ |
bcee2b96 | 795 | if (health_quit_pipe[0] != -1) { |
28ab034a | 796 | (void) fd_tracker_util_pipe_close(the_fd_tracker, health_quit_pipe); |
bcee2b96 | 797 | } |
8a00688e | 798 | relayd_close_thread_quit_pipe(); |
794e2e5f | 799 | if (sessiond_trace_chunk_registry) { |
28ab034a | 800 | sessiond_trace_chunk_registry_destroy(sessiond_trace_chunk_registry); |
794e2e5f | 801 | } |
00e3b7f1 | 802 | if (the_fd_tracker) { |
9c256b01 JG |
803 | untrack_stdio(); |
804 | /* | |
805 | * fd_tracker_destroy() will log the contents of the fd-tracker | |
806 | * if a leak is detected. | |
807 | */ | |
00e3b7f1 JG |
808 | fd_tracker_destroy(the_fd_tracker); |
809 | } | |
794e2e5f | 810 | |
710c1f73 DG |
811 | uri_free(control_uri); |
812 | uri_free(data_uri); | |
8d5c808e | 813 | /* Live URI is freed in the live thread. */ |
cd60b05a JG |
814 | |
815 | if (tracing_group_name_override) { | |
816 | free((void *) tracing_group_name); | |
817 | } | |
b8aa1682 JD |
818 | } |
819 | ||
7591bab1 | 820 | static int notify_health_quit_pipe(int *pipe) |
65931c8b | 821 | { |
6cd525e8 | 822 | ssize_t ret; |
65931c8b | 823 | |
6cd525e8 MD |
824 | ret = lttng_write(pipe[1], "4", 1); |
825 | if (ret < 1) { | |
65931c8b | 826 | PERROR("write relay health quit"); |
b4aacfdc | 827 | goto end; |
65931c8b | 828 | } |
b4aacfdc MD |
829 | ret = 0; |
830 | end: | |
831 | return ret; | |
65931c8b MD |
832 | } |
833 | ||
b8aa1682 | 834 | /* |
b4aacfdc | 835 | * Stop all relayd and relayd-live threads. |
b8aa1682 | 836 | */ |
cd9adb8b | 837 | int lttng_relay_stop_threads() |
b8aa1682 | 838 | { |
b4aacfdc | 839 | int retval = 0; |
b8aa1682 JD |
840 | |
841 | /* Stopping all threads */ | |
842 | DBG("Terminating all threads"); | |
8a00688e | 843 | if (relayd_notify_thread_quit_pipe()) { |
b8aa1682 | 844 | ERR("write error on thread quit pipe"); |
b4aacfdc | 845 | retval = -1; |
b8aa1682 JD |
846 | } |
847 | ||
b4aacfdc MD |
848 | if (notify_health_quit_pipe(health_quit_pipe)) { |
849 | ERR("write error on health quit pipe"); | |
850 | } | |
65931c8b | 851 | |
b8aa1682 | 852 | /* Dispatch thread */ |
26c9d55e | 853 | CMM_STORE_SHARED(dispatch_thread_exit, 1); |
58eb9381 | 854 | futex_nto1_wake(&relay_conn_queue.futex); |
178a0557 | 855 | |
b4aacfdc | 856 | if (relayd_live_stop()) { |
178a0557 | 857 | ERR("Error stopping live threads"); |
b4aacfdc | 858 | retval = -1; |
178a0557 | 859 | } |
b4aacfdc | 860 | return retval; |
b8aa1682 JD |
861 | } |
862 | ||
863 | /* | |
864 | * Signal handler for the daemon | |
865 | * | |
866 | * Simply stop all worker threads, leaving main() return gracefully after | |
867 | * joining all threads and calling cleanup(). | |
868 | */ | |
7591bab1 | 869 | static void sighandler(int sig) |
b8aa1682 JD |
870 | { |
871 | switch (sig) { | |
b8aa1682 JD |
872 | case SIGINT: |
873 | DBG("SIGINT caught"); | |
b4aacfdc MD |
874 | if (lttng_relay_stop_threads()) { |
875 | ERR("Error stopping threads"); | |
876 | } | |
b8aa1682 JD |
877 | break; |
878 | case SIGTERM: | |
879 | DBG("SIGTERM caught"); | |
b4aacfdc MD |
880 | if (lttng_relay_stop_threads()) { |
881 | ERR("Error stopping threads"); | |
882 | } | |
b8aa1682 | 883 | break; |
3fd27398 MD |
884 | case SIGUSR1: |
885 | CMM_STORE_SHARED(recv_child_signal, 1); | |
886 | break; | |
b8aa1682 JD |
887 | default: |
888 | break; | |
889 | } | |
890 | } | |
891 | ||
892 | /* | |
893 | * Setup signal handler for : | |
894 | * SIGINT, SIGTERM, SIGPIPE | |
895 | */ | |
cd9adb8b | 896 | static int set_signal_handler() |
b8aa1682 JD |
897 | { |
898 | int ret = 0; | |
899 | struct sigaction sa; | |
900 | sigset_t sigset; | |
901 | ||
902 | if ((ret = sigemptyset(&sigset)) < 0) { | |
903 | PERROR("sigemptyset"); | |
904 | return ret; | |
905 | } | |
906 | ||
b8aa1682 JD |
907 | sa.sa_mask = sigset; |
908 | sa.sa_flags = 0; | |
0072e5e2 MD |
909 | |
910 | sa.sa_handler = sighandler; | |
cd9adb8b | 911 | if ((ret = sigaction(SIGTERM, &sa, nullptr)) < 0) { |
b8aa1682 JD |
912 | PERROR("sigaction"); |
913 | return ret; | |
914 | } | |
915 | ||
cd9adb8b | 916 | if ((ret = sigaction(SIGINT, &sa, nullptr)) < 0) { |
b8aa1682 JD |
917 | PERROR("sigaction"); |
918 | return ret; | |
919 | } | |
920 | ||
cd9adb8b | 921 | if ((ret = sigaction(SIGUSR1, &sa, nullptr)) < 0) { |
b8aa1682 JD |
922 | PERROR("sigaction"); |
923 | return ret; | |
924 | } | |
925 | ||
0072e5e2 | 926 | sa.sa_handler = SIG_IGN; |
cd9adb8b | 927 | if ((ret = sigaction(SIGPIPE, &sa, nullptr)) < 0) { |
3fd27398 MD |
928 | PERROR("sigaction"); |
929 | return ret; | |
930 | } | |
931 | ||
932 | DBG("Signal handler set for SIGTERM, SIGUSR1, SIGPIPE and SIGINT"); | |
b8aa1682 JD |
933 | |
934 | return ret; | |
935 | } | |
936 | ||
cd9adb8b | 937 | void lttng_relay_notify_ready() |
3fd27398 MD |
938 | { |
939 | /* Notify the parent of the fork() process that we are ready. */ | |
940 | if (opt_daemon || opt_background) { | |
941 | if (uatomic_sub_return(<tng_relay_ready, 1) == 0) { | |
942 | kill(child_ppid, SIGUSR1); | |
943 | } | |
944 | } | |
945 | } | |
946 | ||
bcee2b96 JG |
947 | /* |
948 | * Init health quit pipe. | |
949 | * | |
950 | * Return -1 on error or 0 if all pipes are created. | |
951 | */ | |
cd9adb8b | 952 | static int init_health_quit_pipe() |
bcee2b96 | 953 | { |
28ab034a JG |
954 | return fd_tracker_util_pipe_open_cloexec( |
955 | the_fd_tracker, "Health quit pipe", health_quit_pipe); | |
bcee2b96 JG |
956 | } |
957 | ||
40212d87 JG |
958 | static int create_sock(void *data, int *out_fd) |
959 | { | |
960 | int ret; | |
ac497a37 | 961 | struct lttcomm_sock *sock = (lttcomm_sock *) data; |
40212d87 JG |
962 | |
963 | ret = lttcomm_create_sock(sock); | |
964 | if (ret < 0) { | |
965 | goto end; | |
966 | } | |
967 | ||
968 | *out_fd = sock->fd; | |
969 | end: | |
970 | return ret; | |
971 | } | |
972 | ||
f46376a1 | 973 | static int close_sock(void *data, int *in_fd __attribute__((unused))) |
40212d87 | 974 | { |
ac497a37 | 975 | struct lttcomm_sock *sock = (lttcomm_sock *) data; |
40212d87 JG |
976 | |
977 | return sock->ops->close(sock); | |
978 | } | |
979 | ||
f355467e JG |
980 | static int accept_sock(void *data, int *out_fd) |
981 | { | |
982 | int ret = 0; | |
983 | /* Socks is an array of in_sock, out_sock. */ | |
ac497a37 | 984 | struct lttcomm_sock **socks = (lttcomm_sock **) data; |
f355467e JG |
985 | struct lttcomm_sock *in_sock = socks[0]; |
986 | ||
f118099a | 987 | socks[1] = in_sock->ops->accept(in_sock); |
f355467e JG |
988 | if (!socks[1]) { |
989 | ret = -1; | |
990 | goto end; | |
991 | } | |
992 | *out_fd = socks[1]->fd; | |
993 | end: | |
994 | return ret; | |
995 | } | |
996 | ||
b8aa1682 JD |
997 | /* |
998 | * Create and init socket from uri. | |
999 | */ | |
28ab034a | 1000 | static struct lttcomm_sock *relay_socket_create(struct lttng_uri *uri, const char *name) |
b8aa1682 | 1001 | { |
40212d87 | 1002 | int ret, sock_fd; |
cd9adb8b | 1003 | struct lttcomm_sock *sock = nullptr; |
40212d87 | 1004 | char uri_str[PATH_MAX]; |
cd9adb8b | 1005 | char *formated_name = nullptr; |
b8aa1682 JD |
1006 | |
1007 | sock = lttcomm_alloc_sock_from_uri(uri); | |
cd9adb8b | 1008 | if (sock == nullptr) { |
b8aa1682 JD |
1009 | ERR("Allocating socket"); |
1010 | goto error; | |
1011 | } | |
1012 | ||
40212d87 JG |
1013 | /* |
1014 | * Don't fail to create the socket if the name can't be built as it is | |
1015 | * only used for debugging purposes. | |
1016 | */ | |
1017 | ret = uri_to_str_url(uri, uri_str, sizeof(uri_str)); | |
1018 | uri_str[sizeof(uri_str) - 1] = '\0'; | |
1019 | if (ret >= 0) { | |
28ab034a | 1020 | ret = asprintf(&formated_name, "%s socket @ %s", name, uri_str); |
40212d87 | 1021 | if (ret < 0) { |
cd9adb8b | 1022 | formated_name = nullptr; |
40212d87 | 1023 | } |
b8aa1682 | 1024 | } |
40212d87 | 1025 | |
28ab034a JG |
1026 | ret = fd_tracker_open_unsuspendable_fd(the_fd_tracker, |
1027 | &sock_fd, | |
1028 | (const char **) (formated_name ? &formated_name : | |
cd9adb8b | 1029 | nullptr), |
28ab034a JG |
1030 | 1, |
1031 | create_sock, | |
1032 | sock); | |
6016eb62 | 1033 | if (ret) { |
28ab034a | 1034 | PERROR("Failed to open \"%s\" relay socket", formated_name ?: "Unknown"); |
6016eb62 JG |
1035 | goto error; |
1036 | } | |
40212d87 | 1037 | DBG("Listening on %s socket %d", name, sock->fd); |
b8aa1682 JD |
1038 | |
1039 | ret = sock->ops->bind(sock); | |
1040 | if (ret < 0) { | |
2288467f | 1041 | PERROR("Failed to bind socket"); |
b8aa1682 JD |
1042 | goto error; |
1043 | } | |
1044 | ||
1045 | ret = sock->ops->listen(sock, -1); | |
1046 | if (ret < 0) { | |
1047 | goto error; | |
b8aa1682 JD |
1048 | } |
1049 | ||
6016eb62 | 1050 | free(formated_name); |
b8aa1682 JD |
1051 | return sock; |
1052 | ||
1053 | error: | |
1054 | if (sock) { | |
1055 | lttcomm_destroy_sock(sock); | |
1056 | } | |
6016eb62 | 1057 | free(formated_name); |
cd9adb8b | 1058 | return nullptr; |
b8aa1682 JD |
1059 | } |
1060 | ||
28ab034a JG |
1061 | static struct lttcomm_sock *accept_relayd_sock(struct lttcomm_sock *listening_sock, |
1062 | const char *name) | |
f355467e JG |
1063 | { |
1064 | int out_fd, ret; | |
cd9adb8b JG |
1065 | struct lttcomm_sock *socks[2] = { listening_sock, nullptr }; |
1066 | struct lttcomm_sock *new_sock = nullptr; | |
f355467e | 1067 | |
f118099a | 1068 | ret = fd_tracker_open_unsuspendable_fd( |
28ab034a | 1069 | the_fd_tracker, &out_fd, (const char **) &name, 1, accept_sock, &socks); |
f355467e JG |
1070 | if (ret) { |
1071 | goto end; | |
1072 | } | |
1073 | new_sock = socks[1]; | |
1074 | DBG("%s accepted, socket %d", name, new_sock->fd); | |
1075 | end: | |
1076 | return new_sock; | |
1077 | } | |
1078 | ||
b8aa1682 JD |
1079 | /* |
1080 | * This thread manages the listening for new connections on the network | |
1081 | */ | |
f46376a1 | 1082 | static void *relay_thread_listener(void *data __attribute__((unused))) |
b8aa1682 | 1083 | { |
8a00688e MJ |
1084 | int i, ret, err = -1; |
1085 | uint32_t nb_fd; | |
b8aa1682 JD |
1086 | struct lttng_poll_event events; |
1087 | struct lttcomm_sock *control_sock, *data_sock; | |
1088 | ||
b8aa1682 JD |
1089 | DBG("[thread] Relay listener started"); |
1090 | ||
8fba2b8d | 1091 | rcu_register_thread(); |
55706a7d MD |
1092 | health_register(health_relayd, HEALTH_RELAYD_TYPE_LISTENER); |
1093 | ||
f385ae0a MD |
1094 | health_code_update(); |
1095 | ||
40212d87 | 1096 | control_sock = relay_socket_create(control_uri, "Control listener"); |
b8aa1682 | 1097 | if (!control_sock) { |
095a4ae5 | 1098 | goto error_sock_control; |
b8aa1682 JD |
1099 | } |
1100 | ||
40212d87 | 1101 | data_sock = relay_socket_create(data_uri, "Data listener"); |
b8aa1682 | 1102 | if (!data_sock) { |
095a4ae5 | 1103 | goto error_sock_relay; |
b8aa1682 JD |
1104 | } |
1105 | ||
1106 | /* | |
7591bab1 MD |
1107 | * Pass 3 as size here for the thread quit pipe, control and |
1108 | * data socket. | |
b8aa1682 | 1109 | */ |
ba9cf8e1 | 1110 | ret = create_named_thread_poll_set(&events, 3, "Listener thread epoll"); |
b8aa1682 JD |
1111 | if (ret < 0) { |
1112 | goto error_create_poll; | |
1113 | } | |
1114 | ||
1115 | /* Add the control socket */ | |
1116 | ret = lttng_poll_add(&events, control_sock->fd, LPOLLIN | LPOLLRDHUP); | |
1117 | if (ret < 0) { | |
1118 | goto error_poll_add; | |
1119 | } | |
1120 | ||
1121 | /* Add the data socket */ | |
1122 | ret = lttng_poll_add(&events, data_sock->fd, LPOLLIN | LPOLLRDHUP); | |
1123 | if (ret < 0) { | |
1124 | goto error_poll_add; | |
1125 | } | |
1126 | ||
3fd27398 MD |
1127 | lttng_relay_notify_ready(); |
1128 | ||
9b5e0863 MD |
1129 | if (testpoint(relayd_thread_listener)) { |
1130 | goto error_testpoint; | |
1131 | } | |
1132 | ||
cd9adb8b | 1133 | while (true) { |
f385ae0a MD |
1134 | health_code_update(); |
1135 | ||
b8aa1682 JD |
1136 | DBG("Listener accepting connections"); |
1137 | ||
28ab034a | 1138 | restart: |
f385ae0a | 1139 | health_poll_entry(); |
b8aa1682 | 1140 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 1141 | health_poll_exit(); |
b8aa1682 JD |
1142 | if (ret < 0) { |
1143 | /* | |
1144 | * Restart interrupted system call. | |
1145 | */ | |
1146 | if (errno == EINTR) { | |
1147 | goto restart; | |
1148 | } | |
1149 | goto error; | |
1150 | } | |
1151 | ||
0d9c5d77 DG |
1152 | nb_fd = ret; |
1153 | ||
b8aa1682 JD |
1154 | DBG("Relay new connection received"); |
1155 | for (i = 0; i < nb_fd; i++) { | |
1156 | /* Fetch once the poll data */ | |
8a00688e MJ |
1157 | const auto revents = LTTNG_POLL_GETEV(&events, i); |
1158 | const auto pollfd = LTTNG_POLL_GETFD(&events, i); | |
b8aa1682 | 1159 | |
8a00688e MJ |
1160 | health_code_update(); |
1161 | ||
1162 | /* Activity on thread quit pipe, exiting. */ | |
1163 | if (relayd_is_thread_quit_pipe(pollfd)) { | |
1164 | DBG("Activity on thread quit pipe"); | |
095a4ae5 MD |
1165 | err = 0; |
1166 | goto exit; | |
b8aa1682 JD |
1167 | } |
1168 | ||
03e43155 | 1169 | if (revents & LPOLLIN) { |
4b7f17b2 | 1170 | /* |
7591bab1 MD |
1171 | * A new connection is requested, therefore a |
1172 | * sessiond/consumerd connection is allocated in | |
1173 | * this thread, enqueued to a global queue and | |
1174 | * dequeued (and freed) in the worker thread. | |
4b7f17b2 | 1175 | */ |
58eb9381 DG |
1176 | int val = 1; |
1177 | struct relay_connection *new_conn; | |
cd9adb8b | 1178 | struct lttcomm_sock *newsock = nullptr; |
7591bab1 | 1179 | enum connection_type type; |
b8aa1682 JD |
1180 | |
1181 | if (pollfd == data_sock->fd) { | |
7591bab1 | 1182 | type = RELAY_DATA; |
f355467e | 1183 | newsock = accept_relayd_sock(data_sock, |
28ab034a | 1184 | "Data socket to relayd"); |
4b7f17b2 | 1185 | } else { |
a0377dfe | 1186 | LTTNG_ASSERT(pollfd == control_sock->fd); |
7591bab1 | 1187 | type = RELAY_CONTROL; |
875e3164 | 1188 | newsock = accept_relayd_sock(control_sock, |
28ab034a | 1189 | "Control socket to relayd"); |
b8aa1682 | 1190 | } |
58eb9381 DG |
1191 | if (!newsock) { |
1192 | PERROR("accepting sock"); | |
58eb9381 DG |
1193 | goto error; |
1194 | } | |
1195 | ||
28ab034a JG |
1196 | ret = setsockopt( |
1197 | newsock->fd, SOL_SOCKET, SO_REUSEADDR, &val, sizeof(val)); | |
b8aa1682 JD |
1198 | if (ret < 0) { |
1199 | PERROR("setsockopt inet"); | |
4b7f17b2 | 1200 | lttcomm_destroy_sock(newsock); |
b8aa1682 JD |
1201 | goto error; |
1202 | } | |
f056029c JR |
1203 | |
1204 | ret = socket_apply_keep_alive_config(newsock->fd); | |
1205 | if (ret < 0) { | |
1206 | ERR("Failed to apply TCP keep-alive configuration on socket (%i)", | |
28ab034a | 1207 | newsock->fd); |
f056029c JR |
1208 | lttcomm_destroy_sock(newsock); |
1209 | goto error; | |
1210 | } | |
1211 | ||
7591bab1 MD |
1212 | new_conn = connection_create(newsock, type); |
1213 | if (!new_conn) { | |
1214 | lttcomm_destroy_sock(newsock); | |
1215 | goto error; | |
1216 | } | |
58eb9381 DG |
1217 | |
1218 | /* Enqueue request for the dispatcher thread. */ | |
ac497a37 SM |
1219 | cds_wfcq_head_ptr_t head; |
1220 | head.h = &relay_conn_queue.head; | |
28ab034a | 1221 | cds_wfcq_enqueue(head, &relay_conn_queue.tail, &new_conn->qnode); |
b8aa1682 JD |
1222 | |
1223 | /* | |
7591bab1 MD |
1224 | * Wake the dispatch queue futex. |
1225 | * Implicit memory barrier with the | |
1226 | * exchange in cds_wfcq_enqueue. | |
b8aa1682 | 1227 | */ |
58eb9381 | 1228 | futex_nto1_wake(&relay_conn_queue.futex); |
03e43155 MD |
1229 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
1230 | ERR("socket poll error"); | |
1231 | goto error; | |
1232 | } else { | |
1233 | ERR("Unexpected poll events %u for sock %d", revents, pollfd); | |
1234 | goto error; | |
b8aa1682 JD |
1235 | } |
1236 | } | |
1237 | } | |
1238 | ||
095a4ae5 | 1239 | exit: |
b8aa1682 JD |
1240 | error: |
1241 | error_poll_add: | |
9b5e0863 | 1242 | error_testpoint: |
ba9cf8e1 | 1243 | (void) fd_tracker_util_poll_clean(the_fd_tracker, &events); |
b8aa1682 | 1244 | error_create_poll: |
095a4ae5 | 1245 | if (data_sock->fd >= 0) { |
40212d87 JG |
1246 | int data_sock_fd = data_sock->fd; |
1247 | ||
28ab034a JG |
1248 | ret = fd_tracker_close_unsuspendable_fd( |
1249 | the_fd_tracker, &data_sock_fd, 1, close_sock, data_sock); | |
b8aa1682 | 1250 | if (ret) { |
40212d87 | 1251 | PERROR("Failed to close the data listener socket file descriptor"); |
b8aa1682 | 1252 | } |
40212d87 | 1253 | data_sock->fd = -1; |
b8aa1682 | 1254 | } |
095a4ae5 MD |
1255 | lttcomm_destroy_sock(data_sock); |
1256 | error_sock_relay: | |
1257 | if (control_sock->fd >= 0) { | |
40212d87 JG |
1258 | int control_sock_fd = control_sock->fd; |
1259 | ||
28ab034a JG |
1260 | ret = fd_tracker_close_unsuspendable_fd( |
1261 | the_fd_tracker, &control_sock_fd, 1, close_sock, control_sock); | |
b8aa1682 | 1262 | if (ret) { |
40212d87 | 1263 | PERROR("Failed to close the control listener socket file descriptor"); |
b8aa1682 | 1264 | } |
40212d87 | 1265 | control_sock->fd = -1; |
b8aa1682 | 1266 | } |
095a4ae5 MD |
1267 | lttcomm_destroy_sock(control_sock); |
1268 | error_sock_control: | |
1269 | if (err) { | |
f385ae0a MD |
1270 | health_error(); |
1271 | ERR("Health error occurred in %s", __func__); | |
095a4ae5 | 1272 | } |
55706a7d | 1273 | health_unregister(health_relayd); |
8fba2b8d | 1274 | rcu_unregister_thread(); |
b8aa1682 | 1275 | DBG("Relay listener thread cleanup complete"); |
b4aacfdc | 1276 | lttng_relay_stop_threads(); |
cd9adb8b | 1277 | return nullptr; |
b8aa1682 JD |
1278 | } |
1279 | ||
1280 | /* | |
1281 | * This thread manages the dispatching of the requests to worker threads | |
1282 | */ | |
f46376a1 | 1283 | static void *relay_thread_dispatcher(void *data __attribute__((unused))) |
b8aa1682 | 1284 | { |
6cd525e8 MD |
1285 | int err = -1; |
1286 | ssize_t ret; | |
8bdee6e2 | 1287 | struct cds_wfcq_node *node; |
cd9adb8b | 1288 | struct relay_connection *new_conn = nullptr; |
b8aa1682 JD |
1289 | |
1290 | DBG("[thread] Relay dispatcher started"); | |
1291 | ||
55706a7d MD |
1292 | health_register(health_relayd, HEALTH_RELAYD_TYPE_DISPATCHER); |
1293 | ||
9b5e0863 MD |
1294 | if (testpoint(relayd_thread_dispatcher)) { |
1295 | goto error_testpoint; | |
1296 | } | |
1297 | ||
f385ae0a MD |
1298 | health_code_update(); |
1299 | ||
0ed3b1a8 | 1300 | for (;;) { |
f385ae0a MD |
1301 | health_code_update(); |
1302 | ||
b8aa1682 | 1303 | /* Atomically prepare the queue futex */ |
58eb9381 | 1304 | futex_nto1_prepare(&relay_conn_queue.futex); |
b8aa1682 | 1305 | |
0ed3b1a8 MD |
1306 | if (CMM_LOAD_SHARED(dispatch_thread_exit)) { |
1307 | break; | |
1308 | } | |
1309 | ||
b8aa1682 | 1310 | do { |
f385ae0a MD |
1311 | health_code_update(); |
1312 | ||
b8aa1682 | 1313 | /* Dequeue commands */ |
8bdee6e2 SM |
1314 | node = cds_wfcq_dequeue_blocking(&relay_conn_queue.head, |
1315 | &relay_conn_queue.tail); | |
cd9adb8b | 1316 | if (node == nullptr) { |
b8aa1682 JD |
1317 | DBG("Woken up but nothing in the relay command queue"); |
1318 | /* Continue thread execution */ | |
1319 | break; | |
1320 | } | |
0114db0e | 1321 | new_conn = lttng::utils::container_of(node, &relay_connection::qnode); |
b8aa1682 | 1322 | |
58eb9381 | 1323 | DBG("Dispatching request waiting on sock %d", new_conn->sock->fd); |
b8aa1682 JD |
1324 | |
1325 | /* | |
7591bab1 MD |
1326 | * Inform worker thread of the new request. This |
1327 | * call is blocking so we can be assured that | |
1328 | * the data will be read at some point in time | |
1329 | * or wait to the end of the world :) | |
b8aa1682 | 1330 | */ |
28f23191 JG |
1331 | ret = lttng_write( |
1332 | relay_conn_pipe[1], &new_conn, sizeof(new_conn)); /* NOLINT | |
1333 | sizeof | |
1334 | used | |
1335 | on a | |
1336 | pointer. | |
1337 | */ | |
58eb9381 DG |
1338 | if (ret < 0) { |
1339 | PERROR("write connection pipe"); | |
7591bab1 | 1340 | connection_put(new_conn); |
b8aa1682 JD |
1341 | goto error; |
1342 | } | |
cd9adb8b | 1343 | } while (node != nullptr); |
b8aa1682 JD |
1344 | |
1345 | /* Futex wait on queue. Blocking call on futex() */ | |
f385ae0a | 1346 | health_poll_entry(); |
58eb9381 | 1347 | futex_nto1_wait(&relay_conn_queue.futex); |
f385ae0a | 1348 | health_poll_exit(); |
b8aa1682 JD |
1349 | } |
1350 | ||
f385ae0a MD |
1351 | /* Normal exit, no error */ |
1352 | err = 0; | |
1353 | ||
b8aa1682 | 1354 | error: |
9b5e0863 | 1355 | error_testpoint: |
f385ae0a MD |
1356 | if (err) { |
1357 | health_error(); | |
1358 | ERR("Health error occurred in %s", __func__); | |
1359 | } | |
55706a7d | 1360 | health_unregister(health_relayd); |
b8aa1682 | 1361 | DBG("Dispatch thread dying"); |
b4aacfdc | 1362 | lttng_relay_stop_threads(); |
cd9adb8b | 1363 | return nullptr; |
b8aa1682 JD |
1364 | } |
1365 | ||
298a25ca JG |
1366 | static bool session_streams_have_index(const struct relay_session *session) |
1367 | { | |
1368 | return session->minor >= 4 && !session->snapshot; | |
1369 | } | |
1370 | ||
c5b6f4f0 DG |
1371 | /* |
1372 | * Handle the RELAYD_CREATE_SESSION command. | |
1373 | * | |
1374 | * On success, send back the session id or else return a negative value. | |
1375 | */ | |
28ab034a JG |
1376 | static int relay_create_session(const struct lttcomm_relayd_hdr *recv_hdr __attribute__((unused)), |
1377 | struct relay_connection *conn, | |
1378 | const struct lttng_buffer_view *payload) | |
c5b6f4f0 | 1379 | { |
5312a3ed JG |
1380 | int ret = 0; |
1381 | ssize_t send_ret; | |
cd9adb8b | 1382 | struct relay_session *session = nullptr; |
ecd1a12f | 1383 | struct lttcomm_relayd_create_session_reply_2_11 reply = {}; |
1e791a74 JG |
1384 | char session_name[LTTNG_NAME_MAX] = {}; |
1385 | char hostname[LTTNG_HOST_NAME_MAX] = {}; | |
7591bab1 MD |
1386 | uint32_t live_timer = 0; |
1387 | bool snapshot = false; | |
46ef2188 | 1388 | bool session_name_contains_creation_timestamp = false; |
23c8ff50 | 1389 | /* Left nil for peers < 2.11. */ |
6fa5fe7c | 1390 | char base_path[LTTNG_PATH_MAX] = {}; |
23c8ff50 | 1391 | lttng_uuid sessiond_uuid = {}; |
1e791a74 JG |
1392 | LTTNG_OPTIONAL(uint64_t) id_sessiond = {}; |
1393 | LTTNG_OPTIONAL(uint64_t) current_chunk_id = {}; | |
db1da059 | 1394 | LTTNG_OPTIONAL(time_t) creation_time = {}; |
ecd1a12f MD |
1395 | struct lttng_dynamic_buffer reply_payload; |
1396 | ||
1397 | lttng_dynamic_buffer_init(&reply_payload); | |
c5b6f4f0 | 1398 | |
f86f6389 JR |
1399 | if (conn->minor < 4) { |
1400 | /* From 2.1 to 2.3 */ | |
1401 | ret = 0; | |
1402 | } else if (conn->minor >= 4 && conn->minor < 11) { | |
1403 | /* From 2.4 to 2.10 */ | |
28ab034a JG |
1404 | ret = cmd_create_session_2_4( |
1405 | payload, session_name, hostname, &live_timer, &snapshot); | |
f86f6389 | 1406 | } else { |
84fa4db5 | 1407 | bool has_current_chunk; |
db1da059 JG |
1408 | uint64_t current_chunk_id_value; |
1409 | time_t creation_time_value; | |
1410 | uint64_t id_sessiond_value; | |
84fa4db5 | 1411 | |
f86f6389 | 1412 | /* From 2.11 to ... */ |
28ab034a JG |
1413 | ret = cmd_create_session_2_11(payload, |
1414 | session_name, | |
1415 | hostname, | |
1416 | base_path, | |
1417 | &live_timer, | |
1418 | &snapshot, | |
1419 | &id_sessiond_value, | |
1420 | sessiond_uuid, | |
1421 | &has_current_chunk, | |
1422 | ¤t_chunk_id_value, | |
1423 | &creation_time_value, | |
1424 | &session_name_contains_creation_timestamp); | |
23c8ff50 JG |
1425 | if (lttng_uuid_is_nil(sessiond_uuid)) { |
1426 | /* The nil UUID is reserved for pre-2.11 clients. */ | |
1427 | ERR("Illegal nil UUID announced by peer in create session command"); | |
1428 | ret = -1; | |
1429 | goto send_reply; | |
1430 | } | |
db1da059 JG |
1431 | LTTNG_OPTIONAL_SET(&id_sessiond, id_sessiond_value); |
1432 | LTTNG_OPTIONAL_SET(&creation_time, creation_time_value); | |
1433 | if (has_current_chunk) { | |
28ab034a | 1434 | LTTNG_OPTIONAL_SET(¤t_chunk_id, current_chunk_id_value); |
db1da059 | 1435 | } |
7591bab1 | 1436 | } |
f86f6389 | 1437 | |
7591bab1 MD |
1438 | if (ret < 0) { |
1439 | goto send_reply; | |
d3e2ba59 JD |
1440 | } |
1441 | ||
28ab034a JG |
1442 | session = session_create(session_name, |
1443 | hostname, | |
1444 | base_path, | |
1445 | live_timer, | |
1446 | snapshot, | |
1447 | sessiond_uuid, | |
cd9adb8b JG |
1448 | id_sessiond.is_set ? &id_sessiond.value : nullptr, |
1449 | current_chunk_id.is_set ? ¤t_chunk_id.value : nullptr, | |
1450 | creation_time.is_set ? &creation_time.value : nullptr, | |
28ab034a JG |
1451 | conn->major, |
1452 | conn->minor, | |
1453 | session_name_contains_creation_timestamp); | |
7591bab1 MD |
1454 | if (!session) { |
1455 | ret = -1; | |
1456 | goto send_reply; | |
1457 | } | |
a0377dfe | 1458 | LTTNG_ASSERT(!conn->session); |
7591bab1 | 1459 | conn->session = session; |
c5b6f4f0 DG |
1460 | DBG("Created session %" PRIu64, session->id); |
1461 | ||
ecd1a12f | 1462 | reply.generic.session_id = htobe64(session->id); |
7591bab1 MD |
1463 | |
1464 | send_reply: | |
c5b6f4f0 | 1465 | if (ret < 0) { |
ecd1a12f | 1466 | reply.generic.ret_code = htobe32(LTTNG_ERR_FATAL); |
c5b6f4f0 | 1467 | } else { |
ecd1a12f | 1468 | reply.generic.ret_code = htobe32(LTTNG_OK); |
c5b6f4f0 DG |
1469 | } |
1470 | ||
ecd1a12f MD |
1471 | if (conn->minor < 11) { |
1472 | /* From 2.1 to 2.10 */ | |
28ab034a JG |
1473 | ret = lttng_dynamic_buffer_append( |
1474 | &reply_payload, &reply.generic, sizeof(reply.generic)); | |
ecd1a12f MD |
1475 | if (ret) { |
1476 | ERR("Failed to append \"create session\" command reply header to payload buffer"); | |
1477 | ret = -1; | |
1478 | goto end; | |
1479 | } | |
1480 | } else { | |
28ab034a | 1481 | const uint32_t output_path_length = session ? strlen(session->output_path) + 1 : 0; |
ecd1a12f MD |
1482 | |
1483 | reply.output_path_length = htobe32(output_path_length); | |
28ab034a | 1484 | ret = lttng_dynamic_buffer_append(&reply_payload, &reply, sizeof(reply)); |
ecd1a12f MD |
1485 | if (ret) { |
1486 | ERR("Failed to append \"create session\" command reply header to payload buffer"); | |
1487 | goto end; | |
1488 | } | |
1489 | ||
8d382dd4 | 1490 | if (output_path_length) { |
28ab034a JG |
1491 | ret = lttng_dynamic_buffer_append( |
1492 | &reply_payload, session->output_path, output_path_length); | |
8d382dd4 JG |
1493 | if (ret) { |
1494 | ERR("Failed to append \"create session\" command reply path to payload buffer"); | |
1495 | goto end; | |
1496 | } | |
ecd1a12f MD |
1497 | } |
1498 | } | |
1499 | ||
28ab034a | 1500 | send_ret = conn->sock->ops->sendmsg(conn->sock, reply_payload.data, reply_payload.size, 0); |
ecd1a12f MD |
1501 | if (send_ret < (ssize_t) reply_payload.size) { |
1502 | ERR("Failed to send \"create session\" command reply of %zu bytes (ret = %zd)", | |
28ab034a JG |
1503 | reply_payload.size, |
1504 | send_ret); | |
5312a3ed | 1505 | ret = -1; |
c5b6f4f0 | 1506 | } |
ecd1a12f | 1507 | end: |
4c6885d2 JG |
1508 | if (ret < 0 && session) { |
1509 | session_put(session); | |
1510 | } | |
ecd1a12f | 1511 | lttng_dynamic_buffer_reset(&reply_payload); |
c5b6f4f0 DG |
1512 | return ret; |
1513 | } | |
1514 | ||
a4baae1b JD |
1515 | /* |
1516 | * When we have received all the streams and the metadata for a channel, | |
1517 | * we make them visible to the viewer threads. | |
1518 | */ | |
7591bab1 | 1519 | static void publish_connection_local_streams(struct relay_connection *conn) |
a4baae1b | 1520 | { |
7591bab1 | 1521 | struct relay_session *session = conn->session; |
98b82dfa KS |
1522 | unsigned int created = 0; |
1523 | bool closed = false; | |
1524 | ||
1525 | LTTNG_ASSERT(viewer_sessions_ht); | |
a4baae1b | 1526 | |
7591bab1 MD |
1527 | /* |
1528 | * We publish all streams belonging to a session atomically wrt | |
1529 | * session lock. | |
1530 | */ | |
c7b9f8c3 JG |
1531 | const lttng::pthread::lock_guard session_lock(session->lock); |
1532 | ||
1533 | for (auto *stream : | |
1534 | lttng::urcu::rcu_list_iteration_adapter<relay_stream, &relay_stream::recv_node>( | |
1535 | session->recv_list)) { | |
7591bab1 | 1536 | stream_publish(stream); |
a4baae1b | 1537 | } |
a4baae1b | 1538 | |
7591bab1 MD |
1539 | /* |
1540 | * Inform the viewer that there are new streams in the session. | |
1541 | */ | |
98b82dfa KS |
1542 | if (!session->viewer_attached) { |
1543 | goto unlock; | |
1544 | } | |
1545 | ||
1546 | /* | |
1547 | * Create viewer_streams for all the newly published streams for this relay session. | |
1548 | * This searches through all known viewer sessions and finds those that are | |
1549 | * attached to this connection's relay session. This is done so that the newer | |
1550 | * viewer streams will hold a reference on any relay streams that already exist, | |
1551 | * but may be unpublished between now and the next GET_NEW_STREAMS from the | |
1552 | * attached live viewer. | |
1553 | */ | |
7c8d0f41 JG |
1554 | for (auto *viewer_session : |
1555 | lttng::urcu::lfht_iteration_adapter<relay_viewer_session, | |
1556 | decltype(relay_viewer_session::viewer_session_n), | |
1557 | &relay_viewer_session::viewer_session_n>( | |
1558 | *viewer_sessions_ht->ht)) { | |
1559 | for (auto *session_iter : | |
1560 | lttng::urcu::rcu_list_iteration_adapter<relay_session, | |
1561 | &relay_session::viewer_session_node>( | |
1562 | viewer_session->session_list)) { | |
98b82dfa KS |
1563 | if (session != session_iter) { |
1564 | continue; | |
1565 | } | |
1566 | const int ret = make_viewer_streams(session, | |
1567 | viewer_session, | |
1568 | LTTNG_VIEWER_SEEK_BEGINNING, | |
1569 | nullptr, | |
1570 | nullptr, | |
1571 | &created, | |
1572 | &closed); | |
1573 | if (ret == 0) { | |
1574 | DBG("Created %d new viewer streams during publication of relay streams for relay session %" PRIu64, | |
1575 | created, | |
1576 | session->id); | |
1577 | } else if (ret < 0) { | |
1578 | /* | |
1579 | * Warning, since the creation of the | |
1580 | * streams will be retried when the viewer | |
1581 | * next sends the GET_NEW_STREAMS again. | |
1582 | */ | |
1583 | WARN("Failed to create new viewer streams during publication of relay streams for relay session %" PRIu64 | |
1584 | ", ret=%d, created=%d, closed=%d", | |
1585 | session->id, | |
1586 | ret, | |
1587 | created, | |
1588 | closed); | |
1589 | } | |
1590 | } | |
7591bab1 | 1591 | } |
98b82dfa KS |
1592 | unlock: |
1593 | uatomic_set(&session->new_streams, 1); | |
1594 | pthread_mutex_unlock(&session->lock); | |
a4baae1b JD |
1595 | } |
1596 | ||
348a81dc JG |
1597 | static int conform_channel_path(char *channel_path) |
1598 | { | |
1599 | int ret = 0; | |
1600 | ||
1601 | if (strstr("../", channel_path)) { | |
1602 | ERR("Refusing channel path as it walks up the path hierarchy: \"%s\"", | |
28ab034a | 1603 | channel_path); |
348a81dc JG |
1604 | ret = -1; |
1605 | goto end; | |
1606 | } | |
1607 | ||
1608 | if (*channel_path == '/') { | |
1609 | const size_t len = strlen(channel_path); | |
1610 | ||
1611 | /* | |
1612 | * Channel paths from peers prior to 2.11 are expressed as an | |
1613 | * absolute path that is, in reality, relative to the relay | |
1614 | * daemon's output directory. Remove the leading slash so it | |
1615 | * is correctly interpreted as a relative path later on. | |
1616 | * | |
1617 | * len (and not len - 1) is used to copy the trailing NULL. | |
1618 | */ | |
1619 | bcopy(channel_path + 1, channel_path, len); | |
1620 | } | |
1621 | end: | |
1622 | return ret; | |
1623 | } | |
1624 | ||
b8aa1682 JD |
1625 | /* |
1626 | * relay_add_stream: allocate a new stream for a session | |
1627 | */ | |
28ab034a JG |
1628 | static int relay_add_stream(const struct lttcomm_relayd_hdr *recv_hdr __attribute__((unused)), |
1629 | struct relay_connection *conn, | |
1630 | const struct lttng_buffer_view *payload) | |
b8aa1682 | 1631 | { |
7591bab1 MD |
1632 | int ret; |
1633 | ssize_t send_ret; | |
58eb9381 | 1634 | struct relay_session *session = conn->session; |
cd9adb8b | 1635 | struct relay_stream *stream = nullptr; |
b8aa1682 | 1636 | struct lttcomm_relayd_status_stream reply; |
cd9adb8b | 1637 | struct ctf_trace *trace = nullptr; |
7591bab1 | 1638 | uint64_t stream_handle = -1ULL; |
cd9adb8b | 1639 | char *path_name = nullptr, *channel_name = nullptr; |
7591bab1 | 1640 | uint64_t tracefile_size = 0, tracefile_count = 0; |
348a81dc | 1641 | LTTNG_OPTIONAL(uint64_t) stream_chunk_id = {}; |
b8aa1682 | 1642 | |
5312a3ed | 1643 | if (!session || !conn->version_check_done) { |
b8aa1682 JD |
1644 | ERR("Trying to add a stream before version check"); |
1645 | ret = -1; | |
1646 | goto end_no_session; | |
1647 | } | |
1648 | ||
2f21a469 JR |
1649 | if (session->minor == 1) { |
1650 | /* For 2.1 */ | |
28ab034a | 1651 | ret = cmd_recv_stream_2_1(payload, &path_name, &channel_name); |
2f21a469 JR |
1652 | } else if (session->minor > 1 && session->minor < 11) { |
1653 | /* From 2.2 to 2.10 */ | |
28ab034a JG |
1654 | ret = cmd_recv_stream_2_2( |
1655 | payload, &path_name, &channel_name, &tracefile_size, &tracefile_count); | |
2f21a469 JR |
1656 | } else { |
1657 | /* From 2.11 to ... */ | |
28ab034a JG |
1658 | ret = cmd_recv_stream_2_11(payload, |
1659 | &path_name, | |
1660 | &channel_name, | |
1661 | &tracefile_size, | |
1662 | &tracefile_count, | |
1663 | &stream_chunk_id.value); | |
0b50e4b3 | 1664 | stream_chunk_id.is_set = true; |
0f907de1 | 1665 | } |
2f21a469 | 1666 | |
0f907de1 | 1667 | if (ret < 0) { |
7591bab1 | 1668 | goto send_reply; |
b8aa1682 JD |
1669 | } |
1670 | ||
348a81dc JG |
1671 | if (conform_channel_path(path_name)) { |
1672 | goto send_reply; | |
1673 | } | |
1674 | ||
2a635488 JR |
1675 | /* |
1676 | * Backward compatibility for --group-output-by-session. | |
1677 | * Prior to lttng 2.11, the complete path is passed by the stream. | |
1678 | * Starting at 2.11, lttng-relayd uses chunk. When dealing with producer | |
1679 | * >=2.11 the chunk is responsible for the output path. When dealing | |
1680 | * with producer < 2.11 the chunk output_path is the root output path | |
1681 | * and the stream carries the complete path (path_name). | |
1682 | * To support --group-output-by-session with older producer (<2.11), we | |
1683 | * need to craft the path based on the stream path. | |
1684 | */ | |
1685 | if (opt_group_output_by == RELAYD_GROUP_OUTPUT_BY_SESSION) { | |
1686 | if (conn->minor < 4) { | |
1687 | /* | |
1688 | * From 2.1 to 2.3, the session_name is not passed on | |
1689 | * the RELAYD_CREATE_SESSION command. The session name | |
1690 | * is necessary to detect the presence of a base_path | |
1691 | * inside the stream path. Without it we cannot perform | |
1692 | * a valid group-output-by-session transformation. | |
1693 | */ | |
1694 | WARN("Unable to perform a --group-by-session transformation for session %" PRIu64 | |
1695 | " for stream with path \"%s\" as it is produced by a peer using a protocol older than v2.4", | |
28ab034a JG |
1696 | session->id, |
1697 | path_name); | |
2a635488 JR |
1698 | } else if (conn->minor >= 4 && conn->minor < 11) { |
1699 | char *group_by_session_path_name; | |
1700 | ||
a0377dfe | 1701 | LTTNG_ASSERT(session->session_name[0] != '\0'); |
2a635488 | 1702 | |
28ab034a JG |
1703 | group_by_session_path_name = backward_compat_group_by_session( |
1704 | path_name, session->session_name, session->creation_time.value); | |
2a635488 JR |
1705 | if (!group_by_session_path_name) { |
1706 | ERR("Failed to apply group by session to stream of session %" PRIu64, | |
28ab034a | 1707 | session->id); |
2a635488 JR |
1708 | goto send_reply; |
1709 | } | |
1710 | ||
1711 | DBG("Transformed session path from \"%s\" to \"%s\" to honor per-session name grouping", | |
28ab034a JG |
1712 | path_name, |
1713 | group_by_session_path_name); | |
2a635488 JR |
1714 | |
1715 | free(path_name); | |
1716 | path_name = group_by_session_path_name; | |
1717 | } | |
1718 | } | |
1719 | ||
7591bab1 | 1720 | trace = ctf_trace_get_by_path_or_create(session, path_name); |
2a174661 | 1721 | if (!trace) { |
7591bab1 | 1722 | goto send_reply; |
2a174661 | 1723 | } |
2a174661 | 1724 | |
2a635488 | 1725 | /* This stream here has one reference on the trace. */ |
7591bab1 MD |
1726 | pthread_mutex_lock(&last_relay_stream_id_lock); |
1727 | stream_handle = ++last_relay_stream_id; | |
1728 | pthread_mutex_unlock(&last_relay_stream_id_lock); | |
d3e2ba59 | 1729 | |
7591bab1 | 1730 | /* We pass ownership of path_name and channel_name. */ |
28ab034a JG |
1731 | stream = stream_create( |
1732 | trace, stream_handle, path_name, channel_name, tracefile_size, tracefile_count); | |
cd9adb8b JG |
1733 | path_name = nullptr; |
1734 | channel_name = nullptr; | |
a4baae1b | 1735 | |
2a174661 | 1736 | /* |
7591bab1 MD |
1737 | * Streams are the owners of their trace. Reference to trace is |
1738 | * kept within stream_create(). | |
2a174661 | 1739 | */ |
7591bab1 | 1740 | ctf_trace_put(trace); |
d3e2ba59 | 1741 | |
7591bab1 | 1742 | send_reply: |
53efb85a | 1743 | memset(&reply, 0, sizeof(reply)); |
7591bab1 MD |
1744 | reply.handle = htobe64(stream_handle); |
1745 | if (!stream) { | |
f73fabfd | 1746 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
b8aa1682 | 1747 | } else { |
f73fabfd | 1748 | reply.ret_code = htobe32(LTTNG_OK); |
b8aa1682 | 1749 | } |
5af40280 | 1750 | |
28ab034a JG |
1751 | send_ret = conn->sock->ops->sendmsg( |
1752 | conn->sock, &reply, sizeof(struct lttcomm_relayd_status_stream), 0); | |
5312a3ed | 1753 | if (send_ret < (ssize_t) sizeof(reply)) { |
28ab034a | 1754 | ERR("Failed to send \"add stream\" command reply (ret = %zd)", send_ret); |
5312a3ed | 1755 | ret = -1; |
b8aa1682 JD |
1756 | } |
1757 | ||
1758 | end_no_session: | |
7591bab1 MD |
1759 | free(path_name); |
1760 | free(channel_name); | |
0f907de1 | 1761 | return ret; |
b8aa1682 JD |
1762 | } |
1763 | ||
173af62f DG |
1764 | /* |
1765 | * relay_close_stream: close a specific stream | |
1766 | */ | |
28ab034a JG |
1767 | static int relay_close_stream(const struct lttcomm_relayd_hdr *recv_hdr __attribute__((unused)), |
1768 | struct relay_connection *conn, | |
1769 | const struct lttng_buffer_view *payload) | |
173af62f | 1770 | { |
5312a3ed JG |
1771 | int ret; |
1772 | ssize_t send_ret; | |
58eb9381 | 1773 | struct relay_session *session = conn->session; |
173af62f DG |
1774 | struct lttcomm_relayd_close_stream stream_info; |
1775 | struct lttcomm_relayd_generic_reply reply; | |
1776 | struct relay_stream *stream; | |
173af62f DG |
1777 | |
1778 | DBG("Close stream received"); | |
1779 | ||
5312a3ed | 1780 | if (!session || !conn->version_check_done) { |
173af62f DG |
1781 | ERR("Trying to close a stream before version check"); |
1782 | ret = -1; | |
1783 | goto end_no_session; | |
1784 | } | |
1785 | ||
5312a3ed JG |
1786 | if (payload->size < sizeof(stream_info)) { |
1787 | ERR("Unexpected payload size in \"relay_close_stream\": expected >= %zu bytes, got %zu bytes", | |
28ab034a JG |
1788 | sizeof(stream_info), |
1789 | payload->size); | |
173af62f DG |
1790 | ret = -1; |
1791 | goto end_no_session; | |
1792 | } | |
5312a3ed JG |
1793 | memcpy(&stream_info, payload->data, sizeof(stream_info)); |
1794 | stream_info.stream_id = be64toh(stream_info.stream_id); | |
1795 | stream_info.last_net_seq_num = be64toh(stream_info.last_net_seq_num); | |
173af62f | 1796 | |
5312a3ed | 1797 | stream = stream_get_by_id(stream_info.stream_id); |
173af62f DG |
1798 | if (!stream) { |
1799 | ret = -1; | |
7591bab1 | 1800 | goto end; |
173af62f | 1801 | } |
77f7bd85 MD |
1802 | |
1803 | /* | |
1804 | * Set last_net_seq_num before the close flag. Required by data | |
1805 | * pending check. | |
1806 | */ | |
7591bab1 | 1807 | pthread_mutex_lock(&stream->lock); |
5312a3ed | 1808 | stream->last_net_seq_num = stream_info.last_net_seq_num; |
77f7bd85 MD |
1809 | pthread_mutex_unlock(&stream->lock); |
1810 | ||
bda7c7b9 JG |
1811 | /* |
1812 | * This is one of the conditions which may trigger a stream close | |
1813 | * with the others being: | |
1814 | * 1) A close command is received for a stream | |
1815 | * 2) The control connection owning the stream is closed | |
1816 | * 3) We have received all of the stream's data _after_ a close | |
1817 | * request. | |
1818 | */ | |
1819 | try_stream_close(stream); | |
7591bab1 | 1820 | stream_put(stream); |
5312a3ed | 1821 | ret = 0; |
173af62f | 1822 | |
7591bab1 | 1823 | end: |
53efb85a | 1824 | memset(&reply, 0, sizeof(reply)); |
173af62f | 1825 | if (ret < 0) { |
f73fabfd | 1826 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
173af62f | 1827 | } else { |
f73fabfd | 1828 | reply.ret_code = htobe32(LTTNG_OK); |
173af62f | 1829 | } |
28ab034a JG |
1830 | send_ret = conn->sock->ops->sendmsg( |
1831 | conn->sock, &reply, sizeof(struct lttcomm_relayd_generic_reply), 0); | |
5312a3ed | 1832 | if (send_ret < (ssize_t) sizeof(reply)) { |
28ab034a | 1833 | ERR("Failed to send \"close stream\" command reply (ret = %zd)", send_ret); |
5312a3ed | 1834 | ret = -1; |
173af62f DG |
1835 | } |
1836 | ||
1837 | end_no_session: | |
1838 | return ret; | |
1839 | } | |
1840 | ||
93ec662e JD |
1841 | /* |
1842 | * relay_reset_metadata: reset a metadata stream | |
1843 | */ | |
28ab034a JG |
1844 | static int relay_reset_metadata(const struct lttcomm_relayd_hdr *recv_hdr __attribute__((unused)), |
1845 | struct relay_connection *conn, | |
1846 | const struct lttng_buffer_view *payload) | |
93ec662e | 1847 | { |
5312a3ed JG |
1848 | int ret; |
1849 | ssize_t send_ret; | |
93ec662e JD |
1850 | struct relay_session *session = conn->session; |
1851 | struct lttcomm_relayd_reset_metadata stream_info; | |
1852 | struct lttcomm_relayd_generic_reply reply; | |
1853 | struct relay_stream *stream; | |
1854 | ||
1855 | DBG("Reset metadata received"); | |
1856 | ||
5312a3ed | 1857 | if (!session || !conn->version_check_done) { |
93ec662e JD |
1858 | ERR("Trying to reset a metadata stream before version check"); |
1859 | ret = -1; | |
1860 | goto end_no_session; | |
1861 | } | |
1862 | ||
5312a3ed JG |
1863 | if (payload->size < sizeof(stream_info)) { |
1864 | ERR("Unexpected payload size in \"relay_reset_metadata\": expected >= %zu bytes, got %zu bytes", | |
28ab034a JG |
1865 | sizeof(stream_info), |
1866 | payload->size); | |
93ec662e JD |
1867 | ret = -1; |
1868 | goto end_no_session; | |
1869 | } | |
5312a3ed JG |
1870 | memcpy(&stream_info, payload->data, sizeof(stream_info)); |
1871 | stream_info.stream_id = be64toh(stream_info.stream_id); | |
1872 | stream_info.version = be64toh(stream_info.version); | |
1873 | ||
1874 | DBG("Update metadata to version %" PRIu64, stream_info.version); | |
93ec662e JD |
1875 | |
1876 | /* Unsupported for live sessions for now. */ | |
1877 | if (session->live_timer != 0) { | |
1878 | ret = -1; | |
1879 | goto end; | |
1880 | } | |
1881 | ||
5312a3ed | 1882 | stream = stream_get_by_id(stream_info.stream_id); |
93ec662e JD |
1883 | if (!stream) { |
1884 | ret = -1; | |
1885 | goto end; | |
1886 | } | |
1887 | pthread_mutex_lock(&stream->lock); | |
1888 | if (!stream->is_metadata) { | |
1889 | ret = -1; | |
1890 | goto end_unlock; | |
1891 | } | |
1892 | ||
c35f9726 | 1893 | ret = stream_reset_file(stream); |
93ec662e | 1894 | if (ret < 0) { |
28ab034a JG |
1895 | ERR("Failed to reset metadata stream %" PRIu64 ": stream_path = %s, channel = %s", |
1896 | stream->stream_handle, | |
1897 | stream->path_name, | |
1898 | stream->channel_name); | |
93ec662e JD |
1899 | goto end_unlock; |
1900 | } | |
93ec662e JD |
1901 | end_unlock: |
1902 | pthread_mutex_unlock(&stream->lock); | |
1903 | stream_put(stream); | |
1904 | ||
1905 | end: | |
1906 | memset(&reply, 0, sizeof(reply)); | |
1907 | if (ret < 0) { | |
1908 | reply.ret_code = htobe32(LTTNG_ERR_UNK); | |
1909 | } else { | |
1910 | reply.ret_code = htobe32(LTTNG_OK); | |
1911 | } | |
28ab034a JG |
1912 | send_ret = conn->sock->ops->sendmsg( |
1913 | conn->sock, &reply, sizeof(struct lttcomm_relayd_generic_reply), 0); | |
5312a3ed | 1914 | if (send_ret < (ssize_t) sizeof(reply)) { |
28ab034a | 1915 | ERR("Failed to send \"reset metadata\" command reply (ret = %zd)", send_ret); |
5312a3ed | 1916 | ret = -1; |
93ec662e JD |
1917 | } |
1918 | ||
1919 | end_no_session: | |
1920 | return ret; | |
1921 | } | |
1922 | ||
b8aa1682 JD |
1923 | /* |
1924 | * relay_unknown_command: send -1 if received unknown command | |
1925 | */ | |
7591bab1 | 1926 | static void relay_unknown_command(struct relay_connection *conn) |
b8aa1682 JD |
1927 | { |
1928 | struct lttcomm_relayd_generic_reply reply; | |
5312a3ed | 1929 | ssize_t send_ret; |
b8aa1682 | 1930 | |
53efb85a | 1931 | memset(&reply, 0, sizeof(reply)); |
f73fabfd | 1932 | reply.ret_code = htobe32(LTTNG_ERR_UNK); |
5312a3ed JG |
1933 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
1934 | if (send_ret < sizeof(reply)) { | |
1935 | ERR("Failed to send \"unknown command\" command reply (ret = %zd)", send_ret); | |
b8aa1682 JD |
1936 | } |
1937 | } | |
1938 | ||
1939 | /* | |
1940 | * relay_start: send an acknowledgment to the client to tell if we are | |
1941 | * ready to receive data. We are ready if a session is established. | |
1942 | */ | |
28ab034a JG |
1943 | static int relay_start(const struct lttcomm_relayd_hdr *recv_hdr __attribute__((unused)), |
1944 | struct relay_connection *conn, | |
1945 | const struct lttng_buffer_view *payload __attribute__((unused))) | |
b8aa1682 | 1946 | { |
5312a3ed JG |
1947 | int ret = 0; |
1948 | ssize_t send_ret; | |
b8aa1682 | 1949 | struct lttcomm_relayd_generic_reply reply; |
58eb9381 | 1950 | struct relay_session *session = conn->session; |
b8aa1682 JD |
1951 | |
1952 | if (!session) { | |
1953 | DBG("Trying to start the streaming without a session established"); | |
f73fabfd | 1954 | ret = htobe32(LTTNG_ERR_UNK); |
b8aa1682 JD |
1955 | } |
1956 | ||
53efb85a | 1957 | memset(&reply, 0, sizeof(reply)); |
5312a3ed | 1958 | reply.ret_code = htobe32(LTTNG_OK); |
28ab034a | 1959 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
5312a3ed | 1960 | if (send_ret < (ssize_t) sizeof(reply)) { |
28ab034a | 1961 | ERR("Failed to send \"relay_start\" command reply (ret = %zd)", send_ret); |
5312a3ed | 1962 | ret = -1; |
b8aa1682 JD |
1963 | } |
1964 | ||
1965 | return ret; | |
1966 | } | |
1967 | ||
b8aa1682 | 1968 | /* |
7591bab1 | 1969 | * relay_recv_metadata: receive the metadata for the session. |
b8aa1682 | 1970 | */ |
5312a3ed | 1971 | static int relay_recv_metadata(const struct lttcomm_relayd_hdr *recv_hdr, |
28ab034a JG |
1972 | struct relay_connection *conn, |
1973 | const struct lttng_buffer_view *payload) | |
b8aa1682 | 1974 | { |
32d1569c | 1975 | int ret = 0; |
58eb9381 | 1976 | struct relay_session *session = conn->session; |
5312a3ed | 1977 | struct lttcomm_relayd_metadata_payload metadata_payload_header; |
b8aa1682 | 1978 | struct relay_stream *metadata_stream; |
5312a3ed | 1979 | uint64_t metadata_payload_size; |
c35f9726 | 1980 | struct lttng_buffer_view packet_view; |
b8aa1682 JD |
1981 | |
1982 | if (!session) { | |
1983 | ERR("Metadata sent before version check"); | |
1984 | ret = -1; | |
1985 | goto end; | |
1986 | } | |
1987 | ||
5312a3ed | 1988 | if (recv_hdr->data_size < sizeof(struct lttcomm_relayd_metadata_payload)) { |
f6416125 MD |
1989 | ERR("Incorrect data size"); |
1990 | ret = -1; | |
1991 | goto end; | |
1992 | } | |
28ab034a JG |
1993 | metadata_payload_size = |
1994 | recv_hdr->data_size - sizeof(struct lttcomm_relayd_metadata_payload); | |
f6416125 | 1995 | |
28ab034a JG |
1996 | memcpy(&metadata_payload_header, payload->data, sizeof(metadata_payload_header)); |
1997 | metadata_payload_header.stream_id = be64toh(metadata_payload_header.stream_id); | |
1998 | metadata_payload_header.padding_size = be32toh(metadata_payload_header.padding_size); | |
9d1bbf21 | 1999 | |
5312a3ed | 2000 | metadata_stream = stream_get_by_id(metadata_payload_header.stream_id); |
b8aa1682 JD |
2001 | if (!metadata_stream) { |
2002 | ret = -1; | |
7591bab1 | 2003 | goto end; |
b8aa1682 JD |
2004 | } |
2005 | ||
28ab034a JG |
2006 | packet_view = lttng_buffer_view_from_view( |
2007 | payload, sizeof(metadata_payload_header), metadata_payload_size); | |
3e6e0df2 | 2008 | if (!lttng_buffer_view_is_valid(&packet_view)) { |
c35f9726 | 2009 | ERR("Invalid metadata packet length announced by header"); |
b8aa1682 | 2010 | ret = -1; |
7591bab1 | 2011 | goto end_put; |
b8aa1682 | 2012 | } |
1d4dfdef | 2013 | |
c35f9726 | 2014 | pthread_mutex_lock(&metadata_stream->lock); |
28ab034a | 2015 | ret = stream_write(metadata_stream, &packet_view, metadata_payload_header.padding_size); |
c35f9726 | 2016 | pthread_mutex_unlock(&metadata_stream->lock); |
28ab034a | 2017 | if (ret) { |
5312a3ed | 2018 | ret = -1; |
7591bab1 | 2019 | goto end_put; |
1d4dfdef | 2020 | } |
7591bab1 | 2021 | end_put: |
7591bab1 | 2022 | stream_put(metadata_stream); |
b8aa1682 JD |
2023 | end: |
2024 | return ret; | |
2025 | } | |
2026 | ||
2027 | /* | |
2028 | * relay_send_version: send relayd version number | |
2029 | */ | |
28ab034a JG |
2030 | static int relay_send_version(const struct lttcomm_relayd_hdr *recv_hdr __attribute__((unused)), |
2031 | struct relay_connection *conn, | |
2032 | const struct lttng_buffer_view *payload) | |
b8aa1682 | 2033 | { |
7f51dcba | 2034 | int ret; |
5312a3ed | 2035 | ssize_t send_ret; |
092b6259 | 2036 | struct lttcomm_relayd_version reply, msg; |
87cb6359 | 2037 | bool compatible = true; |
b8aa1682 | 2038 | |
5312a3ed | 2039 | conn->version_check_done = true; |
b8aa1682 | 2040 | |
092b6259 | 2041 | /* Get version from the other side. */ |
5312a3ed JG |
2042 | if (payload->size < sizeof(msg)) { |
2043 | ERR("Unexpected payload size in \"relay_send_version\": expected >= %zu bytes, got %zu bytes", | |
28ab034a JG |
2044 | sizeof(msg), |
2045 | payload->size); | |
092b6259 | 2046 | ret = -1; |
092b6259 DG |
2047 | goto end; |
2048 | } | |
2049 | ||
5312a3ed JG |
2050 | memcpy(&msg, payload->data, sizeof(msg)); |
2051 | msg.major = be32toh(msg.major); | |
2052 | msg.minor = be32toh(msg.minor); | |
2053 | ||
53efb85a | 2054 | memset(&reply, 0, sizeof(reply)); |
d83a952c MD |
2055 | reply.major = RELAYD_VERSION_COMM_MAJOR; |
2056 | reply.minor = RELAYD_VERSION_COMM_MINOR; | |
d4519fa3 JD |
2057 | |
2058 | /* Major versions must be the same */ | |
5312a3ed | 2059 | if (reply.major != msg.major) { |
6151a90f | 2060 | DBG("Incompatible major versions (%u vs %u), deleting session", |
28ab034a JG |
2061 | reply.major, |
2062 | msg.major); | |
87cb6359 | 2063 | compatible = false; |
d4519fa3 JD |
2064 | } |
2065 | ||
58eb9381 | 2066 | conn->major = reply.major; |
0f907de1 | 2067 | /* We adapt to the lowest compatible version */ |
5312a3ed | 2068 | if (reply.minor <= msg.minor) { |
58eb9381 | 2069 | conn->minor = reply.minor; |
0f907de1 | 2070 | } else { |
5312a3ed | 2071 | conn->minor = msg.minor; |
0f907de1 JD |
2072 | } |
2073 | ||
6151a90f JD |
2074 | reply.major = htobe32(reply.major); |
2075 | reply.minor = htobe32(reply.minor); | |
28ab034a | 2076 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
5312a3ed | 2077 | if (send_ret < (ssize_t) sizeof(reply)) { |
28ab034a | 2078 | ERR("Failed to send \"send version\" command reply (ret = %zd)", send_ret); |
5312a3ed JG |
2079 | ret = -1; |
2080 | goto end; | |
2081 | } else { | |
2082 | ret = 0; | |
6151a90f JD |
2083 | } |
2084 | ||
87cb6359 JD |
2085 | if (!compatible) { |
2086 | ret = -1; | |
2087 | goto end; | |
2088 | } | |
2089 | ||
28ab034a | 2090 | DBG("Version check done using protocol %u.%u", conn->major, conn->minor); |
b8aa1682 JD |
2091 | |
2092 | end: | |
2093 | return ret; | |
2094 | } | |
2095 | ||
c8f59ee5 | 2096 | /* |
6d805429 | 2097 | * Check for data pending for a given stream id from the session daemon. |
c8f59ee5 | 2098 | */ |
28ab034a JG |
2099 | static int relay_data_pending(const struct lttcomm_relayd_hdr *recv_hdr __attribute__((unused)), |
2100 | struct relay_connection *conn, | |
2101 | const struct lttng_buffer_view *payload) | |
c8f59ee5 | 2102 | { |
58eb9381 | 2103 | struct relay_session *session = conn->session; |
6d805429 | 2104 | struct lttcomm_relayd_data_pending msg; |
c8f59ee5 DG |
2105 | struct lttcomm_relayd_generic_reply reply; |
2106 | struct relay_stream *stream; | |
5312a3ed | 2107 | ssize_t send_ret; |
c8f59ee5 | 2108 | int ret; |
298a25ca | 2109 | uint64_t stream_seq; |
c8f59ee5 | 2110 | |
6d805429 | 2111 | DBG("Data pending command received"); |
c8f59ee5 | 2112 | |
5312a3ed | 2113 | if (!session || !conn->version_check_done) { |
c8f59ee5 DG |
2114 | ERR("Trying to check for data before version check"); |
2115 | ret = -1; | |
2116 | goto end_no_session; | |
2117 | } | |
2118 | ||
5312a3ed JG |
2119 | if (payload->size < sizeof(msg)) { |
2120 | ERR("Unexpected payload size in \"relay_data_pending\": expected >= %zu bytes, got %zu bytes", | |
28ab034a JG |
2121 | sizeof(msg), |
2122 | payload->size); | |
c8f59ee5 DG |
2123 | ret = -1; |
2124 | goto end_no_session; | |
2125 | } | |
5312a3ed JG |
2126 | memcpy(&msg, payload->data, sizeof(msg)); |
2127 | msg.stream_id = be64toh(msg.stream_id); | |
2128 | msg.last_net_seq_num = be64toh(msg.last_net_seq_num); | |
c8f59ee5 | 2129 | |
5312a3ed | 2130 | stream = stream_get_by_id(msg.stream_id); |
cd9adb8b | 2131 | if (stream == nullptr) { |
c8f59ee5 | 2132 | ret = -1; |
7591bab1 | 2133 | goto end; |
c8f59ee5 DG |
2134 | } |
2135 | ||
7591bab1 MD |
2136 | pthread_mutex_lock(&stream->lock); |
2137 | ||
298a25ca JG |
2138 | if (session_streams_have_index(session)) { |
2139 | /* | |
2140 | * Ensure that both the index and stream data have been | |
2141 | * flushed up to the requested point. | |
2142 | */ | |
ac497a37 | 2143 | stream_seq = std::min(stream->prev_data_seq, stream->prev_index_seq); |
298a25ca | 2144 | } else { |
a8f9f353 | 2145 | stream_seq = stream->prev_data_seq; |
298a25ca | 2146 | } |
a8f9f353 | 2147 | DBG("Data pending for stream id %" PRIu64 ": prev_data_seq %" PRIu64 |
28ab034a JG |
2148 | ", prev_index_seq %" PRIu64 ", and last_seq %" PRIu64, |
2149 | msg.stream_id, | |
2150 | stream->prev_data_seq, | |
2151 | stream->prev_index_seq, | |
2152 | msg.last_net_seq_num); | |
c8f59ee5 | 2153 | |
33832e64 | 2154 | /* Avoid wrapping issue */ |
298a25ca | 2155 | if (((int64_t) (stream_seq - msg.last_net_seq_num)) >= 0) { |
6d805429 | 2156 | /* Data has in fact been written and is NOT pending */ |
c8f59ee5 | 2157 | ret = 0; |
6d805429 DG |
2158 | } else { |
2159 | /* Data still being streamed thus pending */ | |
2160 | ret = 1; | |
c8f59ee5 DG |
2161 | } |
2162 | ||
7591bab1 MD |
2163 | stream->data_pending_check_done = true; |
2164 | pthread_mutex_unlock(&stream->lock); | |
f7079f67 | 2165 | |
7591bab1 MD |
2166 | stream_put(stream); |
2167 | end: | |
c8f59ee5 | 2168 | |
53efb85a | 2169 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 2170 | reply.ret_code = htobe32(ret); |
5312a3ed JG |
2171 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
2172 | if (send_ret < (ssize_t) sizeof(reply)) { | |
28ab034a | 2173 | ERR("Failed to send \"data pending\" command reply (ret = %zd)", send_ret); |
5312a3ed | 2174 | ret = -1; |
c8f59ee5 DG |
2175 | } |
2176 | ||
2177 | end_no_session: | |
2178 | return ret; | |
2179 | } | |
2180 | ||
2181 | /* | |
2182 | * Wait for the control socket to reach a quiescent state. | |
2183 | * | |
7591bab1 MD |
2184 | * Note that for now, when receiving this command from the session |
2185 | * daemon, this means that every subsequent commands or data received on | |
2186 | * the control socket has been handled. So, this is why we simply return | |
2187 | * OK here. | |
c8f59ee5 | 2188 | */ |
28ab034a JG |
2189 | static int relay_quiescent_control(const struct lttcomm_relayd_hdr *recv_hdr |
2190 | __attribute__((unused)), | |
2191 | struct relay_connection *conn, | |
2192 | const struct lttng_buffer_view *payload) | |
c8f59ee5 DG |
2193 | { |
2194 | int ret; | |
5312a3ed | 2195 | ssize_t send_ret; |
ad7051c0 | 2196 | struct relay_stream *stream; |
ad7051c0 | 2197 | struct lttcomm_relayd_quiescent_control msg; |
c8f59ee5 DG |
2198 | struct lttcomm_relayd_generic_reply reply; |
2199 | ||
2200 | DBG("Checking quiescent state on control socket"); | |
2201 | ||
5312a3ed | 2202 | if (!conn->session || !conn->version_check_done) { |
ad7051c0 DG |
2203 | ERR("Trying to check for data before version check"); |
2204 | ret = -1; | |
2205 | goto end_no_session; | |
2206 | } | |
2207 | ||
5312a3ed JG |
2208 | if (payload->size < sizeof(msg)) { |
2209 | ERR("Unexpected payload size in \"relay_quiescent_control\": expected >= %zu bytes, got %zu bytes", | |
28ab034a JG |
2210 | sizeof(msg), |
2211 | payload->size); | |
ad7051c0 DG |
2212 | ret = -1; |
2213 | goto end_no_session; | |
2214 | } | |
5312a3ed JG |
2215 | memcpy(&msg, payload->data, sizeof(msg)); |
2216 | msg.stream_id = be64toh(msg.stream_id); | |
ad7051c0 | 2217 | |
5312a3ed | 2218 | stream = stream_get_by_id(msg.stream_id); |
7591bab1 MD |
2219 | if (!stream) { |
2220 | goto reply; | |
2221 | } | |
2222 | pthread_mutex_lock(&stream->lock); | |
2223 | stream->data_pending_check_done = true; | |
2224 | pthread_mutex_unlock(&stream->lock); | |
5312a3ed JG |
2225 | |
2226 | DBG("Relay quiescent control pending flag set to %" PRIu64, msg.stream_id); | |
7591bab1 MD |
2227 | stream_put(stream); |
2228 | reply: | |
53efb85a | 2229 | memset(&reply, 0, sizeof(reply)); |
c8f59ee5 | 2230 | reply.ret_code = htobe32(LTTNG_OK); |
5312a3ed JG |
2231 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
2232 | if (send_ret < (ssize_t) sizeof(reply)) { | |
28ab034a | 2233 | ERR("Failed to send \"quiescent control\" command reply (ret = %zd)", send_ret); |
5312a3ed JG |
2234 | ret = -1; |
2235 | } else { | |
2236 | ret = 0; | |
c8f59ee5 DG |
2237 | } |
2238 | ||
ad7051c0 | 2239 | end_no_session: |
c8f59ee5 DG |
2240 | return ret; |
2241 | } | |
2242 | ||
f7079f67 | 2243 | /* |
7591bab1 MD |
2244 | * Initialize a data pending command. This means that a consumer is about |
2245 | * to ask for data pending for each stream it holds. Simply iterate over | |
2246 | * all streams of a session and set the data_pending_check_done flag. | |
f7079f67 DG |
2247 | * |
2248 | * This command returns to the client a LTTNG_OK code. | |
2249 | */ | |
5312a3ed | 2250 | static int relay_begin_data_pending(const struct lttcomm_relayd_hdr *recv_hdr, |
28ab034a JG |
2251 | struct relay_connection *conn, |
2252 | const struct lttng_buffer_view *payload) | |
f7079f67 DG |
2253 | { |
2254 | int ret; | |
5312a3ed | 2255 | ssize_t send_ret; |
f7079f67 DG |
2256 | struct lttcomm_relayd_begin_data_pending msg; |
2257 | struct lttcomm_relayd_generic_reply reply; | |
f7079f67 | 2258 | |
a0377dfe FD |
2259 | LTTNG_ASSERT(recv_hdr); |
2260 | LTTNG_ASSERT(conn); | |
f7079f67 DG |
2261 | |
2262 | DBG("Init streams for data pending"); | |
2263 | ||
5312a3ed | 2264 | if (!conn->session || !conn->version_check_done) { |
f7079f67 DG |
2265 | ERR("Trying to check for data before version check"); |
2266 | ret = -1; | |
2267 | goto end_no_session; | |
2268 | } | |
2269 | ||
5312a3ed JG |
2270 | if (payload->size < sizeof(msg)) { |
2271 | ERR("Unexpected payload size in \"relay_begin_data_pending\": expected >= %zu bytes, got %zu bytes", | |
28ab034a JG |
2272 | sizeof(msg), |
2273 | payload->size); | |
f7079f67 DG |
2274 | ret = -1; |
2275 | goto end_no_session; | |
2276 | } | |
5312a3ed JG |
2277 | memcpy(&msg, payload->data, sizeof(msg)); |
2278 | msg.session_id = be64toh(msg.session_id); | |
f7079f67 DG |
2279 | |
2280 | /* | |
7591bab1 MD |
2281 | * Iterate over all streams to set the begin data pending flag. |
2282 | * For now, the streams are indexed by stream handle so we have | |
2283 | * to iterate over all streams to find the one associated with | |
2284 | * the right session_id. | |
f7079f67 | 2285 | */ |
85c34f4c JG |
2286 | for (auto *stream : |
2287 | lttng::urcu::lfht_iteration_adapter<relay_stream, | |
2288 | decltype(relay_stream::node), | |
2289 | &relay_stream::node>(*relay_streams_ht->ht)) { | |
2290 | if (!stream_get(stream)) { | |
2291 | continue; | |
2292 | } | |
56047f5a | 2293 | |
85c34f4c JG |
2294 | if (stream->trace->session->id == msg.session_id) { |
2295 | pthread_mutex_lock(&stream->lock); | |
2296 | stream->data_pending_check_done = false; | |
2297 | pthread_mutex_unlock(&stream->lock); | |
2298 | DBG("Set begin data pending flag to stream %" PRIu64, | |
2299 | stream->stream_handle); | |
f7079f67 | 2300 | } |
85c34f4c JG |
2301 | |
2302 | stream_put(stream); | |
f7079f67 | 2303 | } |
f7079f67 | 2304 | |
53efb85a | 2305 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
2306 | /* All good, send back reply. */ |
2307 | reply.ret_code = htobe32(LTTNG_OK); | |
2308 | ||
5312a3ed JG |
2309 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
2310 | if (send_ret < (ssize_t) sizeof(reply)) { | |
28ab034a | 2311 | ERR("Failed to send \"begin data pending\" command reply (ret = %zd)", send_ret); |
5312a3ed JG |
2312 | ret = -1; |
2313 | } else { | |
2314 | ret = 0; | |
f7079f67 DG |
2315 | } |
2316 | ||
2317 | end_no_session: | |
2318 | return ret; | |
2319 | } | |
2320 | ||
2321 | /* | |
7591bab1 MD |
2322 | * End data pending command. This will check, for a given session id, if |
2323 | * each stream associated with it has its data_pending_check_done flag | |
2324 | * set. If not, this means that the client lost track of the stream but | |
2325 | * the data is still being streamed on our side. In this case, we inform | |
2326 | * the client that data is in flight. | |
f7079f67 DG |
2327 | * |
2328 | * Return to the client if there is data in flight or not with a ret_code. | |
2329 | */ | |
28ab034a JG |
2330 | static int relay_end_data_pending(const struct lttcomm_relayd_hdr *recv_hdr __attribute__((unused)), |
2331 | struct relay_connection *conn, | |
2332 | const struct lttng_buffer_view *payload) | |
f7079f67 DG |
2333 | { |
2334 | int ret; | |
5312a3ed | 2335 | ssize_t send_ret; |
f7079f67 DG |
2336 | struct lttcomm_relayd_end_data_pending msg; |
2337 | struct lttcomm_relayd_generic_reply reply; | |
f7079f67 DG |
2338 | uint32_t is_data_inflight = 0; |
2339 | ||
f7079f67 DG |
2340 | DBG("End data pending command"); |
2341 | ||
5312a3ed | 2342 | if (!conn->session || !conn->version_check_done) { |
f7079f67 DG |
2343 | ERR("Trying to check for data before version check"); |
2344 | ret = -1; | |
2345 | goto end_no_session; | |
2346 | } | |
2347 | ||
5312a3ed JG |
2348 | if (payload->size < sizeof(msg)) { |
2349 | ERR("Unexpected payload size in \"relay_end_data_pending\": expected >= %zu bytes, got %zu bytes", | |
28ab034a JG |
2350 | sizeof(msg), |
2351 | payload->size); | |
f7079f67 DG |
2352 | ret = -1; |
2353 | goto end_no_session; | |
2354 | } | |
5312a3ed JG |
2355 | memcpy(&msg, payload->data, sizeof(msg)); |
2356 | msg.session_id = be64toh(msg.session_id); | |
f7079f67 | 2357 | |
7591bab1 MD |
2358 | /* |
2359 | * Iterate over all streams to see if the begin data pending | |
2360 | * flag is set. | |
2361 | */ | |
85c34f4c JG |
2362 | for (auto *stream : |
2363 | lttng::urcu::lfht_iteration_adapter<relay_stream, | |
2364 | decltype(relay_stream::node), | |
2365 | &relay_stream::node>(*relay_streams_ht->ht)) { | |
2366 | if (!stream_get(stream)) { | |
2367 | continue; | |
2368 | } | |
56047f5a | 2369 | |
85c34f4c JG |
2370 | if (stream->trace->session->id != msg.session_id) { |
2371 | stream_put(stream); | |
2372 | continue; | |
2373 | } | |
56047f5a | 2374 | |
85c34f4c JG |
2375 | pthread_mutex_lock(&stream->lock); |
2376 | if (!stream->data_pending_check_done) { | |
2377 | uint64_t stream_seq; | |
56047f5a | 2378 | |
85c34f4c JG |
2379 | if (session_streams_have_index(conn->session)) { |
2380 | /* | |
2381 | * Ensure that both the index and stream data have been | |
2382 | * flushed up to the requested point. | |
2383 | */ | |
2384 | stream_seq = | |
2385 | std::min(stream->prev_data_seq, stream->prev_index_seq); | |
2386 | } else { | |
2387 | stream_seq = stream->prev_data_seq; | |
7591bab1 | 2388 | } |
56047f5a | 2389 | |
85c34f4c JG |
2390 | if (!stream->closed || |
2391 | !(((int64_t) (stream_seq - stream->last_net_seq_num)) >= 0)) { | |
2392 | is_data_inflight = 1; | |
2393 | DBG("Data is still in flight for stream %" PRIu64, | |
2394 | stream->stream_handle); | |
2395 | pthread_mutex_unlock(&stream->lock); | |
2396 | stream_put(stream); | |
2397 | break; | |
2398 | } | |
f7079f67 | 2399 | } |
85c34f4c JG |
2400 | |
2401 | pthread_mutex_unlock(&stream->lock); | |
2402 | stream_put(stream); | |
f7079f67 | 2403 | } |
f7079f67 | 2404 | |
53efb85a | 2405 | memset(&reply, 0, sizeof(reply)); |
f7079f67 DG |
2406 | /* All good, send back reply. */ |
2407 | reply.ret_code = htobe32(is_data_inflight); | |
2408 | ||
5312a3ed JG |
2409 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
2410 | if (send_ret < (ssize_t) sizeof(reply)) { | |
28ab034a | 2411 | ERR("Failed to send \"end data pending\" command reply (ret = %zd)", send_ret); |
5312a3ed JG |
2412 | ret = -1; |
2413 | } else { | |
2414 | ret = 0; | |
f7079f67 DG |
2415 | } |
2416 | ||
2417 | end_no_session: | |
2418 | return ret; | |
2419 | } | |
2420 | ||
1c20f0e2 JD |
2421 | /* |
2422 | * Receive an index for a specific stream. | |
2423 | * | |
2424 | * Return 0 on success else a negative value. | |
2425 | */ | |
28ab034a JG |
2426 | static int relay_recv_index(const struct lttcomm_relayd_hdr *recv_hdr __attribute__((unused)), |
2427 | struct relay_connection *conn, | |
2428 | const struct lttng_buffer_view *payload) | |
1c20f0e2 | 2429 | { |
5312a3ed JG |
2430 | int ret; |
2431 | ssize_t send_ret; | |
58eb9381 | 2432 | struct relay_session *session = conn->session; |
1c20f0e2 | 2433 | struct lttcomm_relayd_index index_info; |
1c20f0e2 JD |
2434 | struct lttcomm_relayd_generic_reply reply; |
2435 | struct relay_stream *stream; | |
f8f3885c | 2436 | size_t msg_len; |
1c20f0e2 | 2437 | |
a0377dfe | 2438 | LTTNG_ASSERT(conn); |
1c20f0e2 JD |
2439 | |
2440 | DBG("Relay receiving index"); | |
2441 | ||
5312a3ed | 2442 | if (!session || !conn->version_check_done) { |
1c20f0e2 JD |
2443 | ERR("Trying to close a stream before version check"); |
2444 | ret = -1; | |
2445 | goto end_no_session; | |
2446 | } | |
2447 | ||
28ab034a JG |
2448 | msg_len = lttcomm_relayd_index_len(lttng_to_index_major(conn->major, conn->minor), |
2449 | lttng_to_index_minor(conn->major, conn->minor)); | |
5312a3ed JG |
2450 | if (payload->size < msg_len) { |
2451 | ERR("Unexpected payload size in \"relay_recv_index\": expected >= %zu bytes, got %zu bytes", | |
28ab034a JG |
2452 | msg_len, |
2453 | payload->size); | |
1c20f0e2 JD |
2454 | ret = -1; |
2455 | goto end_no_session; | |
2456 | } | |
5312a3ed JG |
2457 | memcpy(&index_info, payload->data, msg_len); |
2458 | index_info.relay_stream_id = be64toh(index_info.relay_stream_id); | |
2459 | index_info.net_seq_num = be64toh(index_info.net_seq_num); | |
2460 | index_info.packet_size = be64toh(index_info.packet_size); | |
2461 | index_info.content_size = be64toh(index_info.content_size); | |
2462 | index_info.timestamp_begin = be64toh(index_info.timestamp_begin); | |
2463 | index_info.timestamp_end = be64toh(index_info.timestamp_end); | |
2464 | index_info.events_discarded = be64toh(index_info.events_discarded); | |
2465 | index_info.stream_id = be64toh(index_info.stream_id); | |
81df238b JR |
2466 | |
2467 | if (conn->minor >= 8) { | |
28ab034a | 2468 | index_info.stream_instance_id = be64toh(index_info.stream_instance_id); |
81df238b | 2469 | index_info.packet_seq_num = be64toh(index_info.packet_seq_num); |
0f83d1cc MD |
2470 | } else { |
2471 | index_info.stream_instance_id = -1ULL; | |
2472 | index_info.packet_seq_num = -1ULL; | |
81df238b | 2473 | } |
5312a3ed JG |
2474 | |
2475 | stream = stream_get_by_id(index_info.relay_stream_id); | |
1c20f0e2 | 2476 | if (!stream) { |
7591bab1 | 2477 | ERR("stream_get_by_id not found"); |
1c20f0e2 | 2478 | ret = -1; |
7591bab1 | 2479 | goto end; |
1c20f0e2 | 2480 | } |
d3e2ba59 | 2481 | |
c35f9726 JG |
2482 | pthread_mutex_lock(&stream->lock); |
2483 | ret = stream_add_index(stream, &index_info); | |
2484 | pthread_mutex_unlock(&stream->lock); | |
2485 | if (ret) { | |
7591bab1 MD |
2486 | goto end_stream_put; |
2487 | } | |
1c20f0e2 | 2488 | |
7591bab1 | 2489 | end_stream_put: |
7591bab1 | 2490 | stream_put(stream); |
7591bab1 | 2491 | end: |
53efb85a | 2492 | memset(&reply, 0, sizeof(reply)); |
1c20f0e2 JD |
2493 | if (ret < 0) { |
2494 | reply.ret_code = htobe32(LTTNG_ERR_UNK); | |
2495 | } else { | |
2496 | reply.ret_code = htobe32(LTTNG_OK); | |
2497 | } | |
58eb9381 | 2498 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
5312a3ed JG |
2499 | if (send_ret < (ssize_t) sizeof(reply)) { |
2500 | ERR("Failed to send \"recv index\" command reply (ret = %zd)", send_ret); | |
2501 | ret = -1; | |
1c20f0e2 JD |
2502 | } |
2503 | ||
2504 | end_no_session: | |
2505 | return ret; | |
2506 | } | |
2507 | ||
a4baae1b JD |
2508 | /* |
2509 | * Receive the streams_sent message. | |
2510 | * | |
2511 | * Return 0 on success else a negative value. | |
2512 | */ | |
28ab034a JG |
2513 | static int relay_streams_sent(const struct lttcomm_relayd_hdr *recv_hdr __attribute__((unused)), |
2514 | struct relay_connection *conn, | |
2515 | const struct lttng_buffer_view *payload __attribute__((unused))) | |
a4baae1b | 2516 | { |
5312a3ed JG |
2517 | int ret; |
2518 | ssize_t send_ret; | |
a4baae1b JD |
2519 | struct lttcomm_relayd_generic_reply reply; |
2520 | ||
a0377dfe | 2521 | LTTNG_ASSERT(conn); |
a4baae1b JD |
2522 | |
2523 | DBG("Relay receiving streams_sent"); | |
2524 | ||
5312a3ed | 2525 | if (!conn->session || !conn->version_check_done) { |
a4baae1b JD |
2526 | ERR("Trying to close a stream before version check"); |
2527 | ret = -1; | |
2528 | goto end_no_session; | |
2529 | } | |
2530 | ||
2531 | /* | |
7591bab1 MD |
2532 | * Publish every pending stream in the connection recv list which are |
2533 | * now ready to be used by the viewer. | |
4a9daf17 | 2534 | */ |
7591bab1 | 2535 | publish_connection_local_streams(conn); |
4a9daf17 | 2536 | |
53efb85a | 2537 | memset(&reply, 0, sizeof(reply)); |
a4baae1b | 2538 | reply.ret_code = htobe32(LTTNG_OK); |
58eb9381 | 2539 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
5312a3ed | 2540 | if (send_ret < (ssize_t) sizeof(reply)) { |
28ab034a | 2541 | ERR("Failed to send \"streams sent\" command reply (ret = %zd)", send_ret); |
5312a3ed | 2542 | ret = -1; |
a4baae1b JD |
2543 | } else { |
2544 | /* Success. */ | |
2545 | ret = 0; | |
2546 | } | |
2547 | ||
2548 | end_no_session: | |
2549 | return ret; | |
2550 | } | |
2551 | ||
28ab034a JG |
2552 | static ssize_t |
2553 | relay_unpack_rotate_streams_header(const struct lttng_buffer_view *payload, | |
2554 | struct lttcomm_relayd_rotate_streams *_rotate_streams) | |
ce9dbd47 JG |
2555 | { |
2556 | struct lttcomm_relayd_rotate_streams rotate_streams; | |
2557 | /* | |
2558 | * Set to the smallest version (packed) of `lttcomm_relayd_rotate_streams`. | |
2559 | * This is the smallest version of this structure, but it can be larger; | |
2560 | * this variable is updated once the proper size of the structure is known. | |
2561 | * | |
2562 | * See comment at the declaration of this structure for more information. | |
2563 | */ | |
2564 | ssize_t header_len = sizeof(struct lttcomm_relayd_rotate_streams_packed); | |
28ab034a | 2565 | size_t expected_payload_size_no_padding, expected_payload_size_3_bytes_padding, |
ce9dbd47 JG |
2566 | expected_payload_size_7_bytes_padding; |
2567 | ||
2568 | if (payload->size < header_len) { | |
2569 | ERR("Unexpected payload size in \"relay_rotate_session_stream\": expected >= %zu bytes, got %zu bytes", | |
28ab034a JG |
2570 | header_len, |
2571 | payload->size); | |
ce9dbd47 JG |
2572 | goto error; |
2573 | } | |
2574 | ||
2575 | /* | |
2576 | * Some versions incorrectly omitted the LTTNG_PACKED annotation on the | |
2577 | * `new_chunk_id` optional field of struct lttcomm_relayd_rotate_streams. | |
2578 | * | |
2579 | * We start by "unpacking" `stream_count` to figure out the padding length | |
2580 | * emited by our peer. | |
2581 | */ | |
11bcbf89 JG |
2582 | { |
2583 | decltype(rotate_streams.stream_count) stream_count; | |
2584 | ||
2585 | memcpy(&stream_count, payload->data, sizeof(stream_count)); | |
2586 | rotate_streams.stream_count = be32toh(stream_count); | |
2587 | } | |
2588 | ||
2589 | rotate_streams.new_chunk_id = LTTNG_OPTIONAL_INIT_UNSET; | |
ce9dbd47 JG |
2590 | |
2591 | /* | |
2592 | * Payload size expected given the possible padding lengths in | |
2593 | * `struct lttcomm_relayd_rotate_streams`. | |
2594 | */ | |
28ab034a JG |
2595 | expected_payload_size_no_padding = |
2596 | (rotate_streams.stream_count * sizeof(*rotate_streams.rotation_positions)) + | |
ce9dbd47 | 2597 | sizeof(lttcomm_relayd_rotate_streams_packed); |
28ab034a JG |
2598 | expected_payload_size_3_bytes_padding = |
2599 | (rotate_streams.stream_count * sizeof(*rotate_streams.rotation_positions)) + | |
ce9dbd47 | 2600 | sizeof(lttcomm_relayd_rotate_streams_3_bytes_padding); |
28ab034a JG |
2601 | expected_payload_size_7_bytes_padding = |
2602 | (rotate_streams.stream_count * sizeof(*rotate_streams.rotation_positions)) + | |
ce9dbd47 JG |
2603 | sizeof(lttcomm_relayd_rotate_streams_7_bytes_padding); |
2604 | ||
2605 | if (payload->size == expected_payload_size_no_padding) { | |
2606 | struct lttcomm_relayd_rotate_streams_packed packed_rotate_streams; | |
2607 | ||
2608 | /* | |
2609 | * This handles cases where someone might build with | |
2610 | * -fpack-struct or any other toolchain that wouldn't produce | |
2611 | * padding to align `value`. | |
2612 | */ | |
2613 | DBG("Received `struct lttcomm_relayd_rotate_streams` with no padding"); | |
2614 | ||
2615 | header_len = sizeof(packed_rotate_streams); | |
2616 | memcpy(&packed_rotate_streams, payload->data, header_len); | |
2617 | ||
2618 | /* Unpack the packed structure to the natively-packed version. */ | |
11bcbf89 JG |
2619 | _rotate_streams->new_chunk_id = (typeof(_rotate_streams->new_chunk_id)){ |
2620 | .is_set = !!packed_rotate_streams.new_chunk_id.is_set, | |
2621 | .value = be64toh(packed_rotate_streams.new_chunk_id.value), | |
ce9dbd47 | 2622 | }; |
11bcbf89 | 2623 | _rotate_streams->stream_count = be32toh(packed_rotate_streams.stream_count); |
ce9dbd47 JG |
2624 | } else if (payload->size == expected_payload_size_3_bytes_padding) { |
2625 | struct lttcomm_relayd_rotate_streams_3_bytes_padding padded_rotate_streams; | |
2626 | ||
2627 | DBG("Received `struct lttcomm_relayd_rotate_streams` with 3 bytes of padding (4-byte aligned peer)"); | |
2628 | ||
2629 | header_len = sizeof(padded_rotate_streams); | |
2630 | memcpy(&padded_rotate_streams, payload->data, header_len); | |
2631 | ||
2632 | /* Unpack the 3-byte padded structure to the natively-packed version. */ | |
11bcbf89 JG |
2633 | _rotate_streams->new_chunk_id = (typeof(_rotate_streams->new_chunk_id)){ |
2634 | .is_set = !!padded_rotate_streams.new_chunk_id.is_set, | |
2635 | .value = be64toh(padded_rotate_streams.new_chunk_id.value), | |
ce9dbd47 | 2636 | }; |
11bcbf89 | 2637 | _rotate_streams->stream_count = be32toh(padded_rotate_streams.stream_count); |
ce9dbd47 JG |
2638 | } else if (payload->size == expected_payload_size_7_bytes_padding) { |
2639 | struct lttcomm_relayd_rotate_streams_7_bytes_padding padded_rotate_streams; | |
2640 | ||
2641 | DBG("Received `struct lttcomm_relayd_rotate_streams` with 7 bytes of padding (8-byte aligned peer)"); | |
2642 | ||
2643 | header_len = sizeof(padded_rotate_streams); | |
2644 | memcpy(&padded_rotate_streams, payload->data, header_len); | |
2645 | ||
2646 | /* Unpack the 7-byte padded structure to the natively-packed version. */ | |
11bcbf89 JG |
2647 | _rotate_streams->new_chunk_id = (typeof(_rotate_streams->new_chunk_id)){ |
2648 | .is_set = !!padded_rotate_streams.new_chunk_id.is_set, | |
2649 | .value = be64toh(padded_rotate_streams.new_chunk_id.value), | |
ce9dbd47 | 2650 | }; |
11bcbf89 | 2651 | _rotate_streams->stream_count = be32toh(padded_rotate_streams.stream_count); |
ce9dbd47 JG |
2652 | |
2653 | header_len = sizeof(padded_rotate_streams); | |
2654 | } else { | |
2655 | ERR("Unexpected payload size in \"relay_rotate_session_stream\": expected %zu, %zu or %zu bytes, got %zu bytes", | |
28ab034a JG |
2656 | expected_payload_size_no_padding, |
2657 | expected_payload_size_3_bytes_padding, | |
2658 | expected_payload_size_7_bytes_padding, | |
2659 | payload->size); | |
ce9dbd47 JG |
2660 | goto error; |
2661 | } | |
2662 | ||
2663 | return header_len; | |
2664 | error: | |
2665 | return -1; | |
2666 | } | |
2667 | ||
d3ecc550 | 2668 | /* |
c35f9726 JG |
2669 | * relay_rotate_session_stream: rotate a stream to a new tracefile for the |
2670 | * session rotation feature (not the tracefile rotation feature). | |
d3ecc550 | 2671 | */ |
28ab034a JG |
2672 | static int relay_rotate_session_streams(const struct lttcomm_relayd_hdr *recv_hdr |
2673 | __attribute__((unused)), | |
2674 | struct relay_connection *conn, | |
2675 | const struct lttng_buffer_view *payload) | |
d3ecc550 | 2676 | { |
30b9d5ab | 2677 | int ret = 0; |
c35f9726 | 2678 | uint32_t i; |
5312a3ed | 2679 | ssize_t send_ret; |
c35f9726 | 2680 | enum lttng_error_code reply_code = LTTNG_ERR_UNK; |
d3ecc550 | 2681 | struct relay_session *session = conn->session; |
c35f9726 JG |
2682 | struct lttcomm_relayd_rotate_streams rotate_streams; |
2683 | struct lttcomm_relayd_generic_reply reply = {}; | |
cd9adb8b JG |
2684 | struct relay_stream *stream = nullptr; |
2685 | struct lttng_trace_chunk *next_trace_chunk = nullptr; | |
c35f9726 | 2686 | struct lttng_buffer_view stream_positions; |
70626904 JG |
2687 | char chunk_id_buf[MAX_INT_DEC_LEN(uint64_t)]; |
2688 | const char *chunk_id_str = "none"; | |
ce9dbd47 | 2689 | ssize_t header_len; |
d3ecc550 | 2690 | |
d3ecc550 JD |
2691 | if (!session || !conn->version_check_done) { |
2692 | ERR("Trying to rotate a stream before version check"); | |
2693 | ret = -1; | |
2694 | goto end_no_reply; | |
2695 | } | |
2696 | ||
2697 | if (session->major == 2 && session->minor < 11) { | |
2698 | ERR("Unsupported feature before 2.11"); | |
2699 | ret = -1; | |
2700 | goto end_no_reply; | |
2701 | } | |
2702 | ||
ce9dbd47 JG |
2703 | header_len = relay_unpack_rotate_streams_header(payload, &rotate_streams); |
2704 | if (header_len < 0) { | |
d3ecc550 JD |
2705 | ret = -1; |
2706 | goto end_no_reply; | |
2707 | } | |
2708 | ||
c35f9726 JG |
2709 | if (rotate_streams.new_chunk_id.is_set) { |
2710 | /* | |
2711 | * Retrieve the trace chunk the stream must transition to. As | |
2712 | * per the protocol, this chunk should have been created | |
2713 | * before this command is received. | |
2714 | */ | |
2715 | next_trace_chunk = sessiond_trace_chunk_registry_get_chunk( | |
28ab034a JG |
2716 | sessiond_trace_chunk_registry, |
2717 | session->sessiond_uuid, | |
2718 | conn->session->id_sessiond.is_set ? conn->session->id_sessiond.value : | |
2719 | conn->session->id, | |
2720 | rotate_streams.new_chunk_id.value); | |
c35f9726 | 2721 | if (!next_trace_chunk) { |
c70636a7 | 2722 | char uuid_str[LTTNG_UUID_STR_LEN]; |
c35f9726 JG |
2723 | |
2724 | lttng_uuid_to_str(session->sessiond_uuid, uuid_str); | |
2725 | ERR("Unknown next trace chunk in ROTATE_STREAMS command: sessiond_uuid = {%s}, session_id = %" PRIu64 | |
28ab034a JG |
2726 | ", trace_chunk_id = %" PRIu64, |
2727 | uuid_str, | |
2728 | session->id, | |
2729 | rotate_streams.new_chunk_id.value); | |
c35f9726 JG |
2730 | reply_code = LTTNG_ERR_INVALID_PROTOCOL; |
2731 | ret = -1; | |
2732 | goto end; | |
2733 | } | |
70626904 | 2734 | |
28ab034a JG |
2735 | ret = snprintf(chunk_id_buf, |
2736 | sizeof(chunk_id_buf), | |
2737 | "%" PRIu64, | |
2738 | rotate_streams.new_chunk_id.value); | |
70626904 JG |
2739 | if (ret < 0 || ret >= sizeof(chunk_id_buf)) { |
2740 | chunk_id_str = "formatting error"; | |
2741 | } else { | |
2742 | chunk_id_str = chunk_id_buf; | |
2743 | } | |
d3ecc550 JD |
2744 | } |
2745 | ||
70626904 | 2746 | DBG("Rotate %" PRIu32 " streams of session \"%s\" to chunk \"%s\"", |
28ab034a JG |
2747 | rotate_streams.stream_count, |
2748 | session->session_name, | |
2749 | chunk_id_str); | |
70626904 | 2750 | |
28ab034a | 2751 | stream_positions = lttng_buffer_view_from_view(payload, header_len, -1); |
c35f9726 | 2752 | if (!stream_positions.data || |
28ab034a JG |
2753 | stream_positions.size < (rotate_streams.stream_count * |
2754 | sizeof(struct lttcomm_relayd_stream_rotation_position))) { | |
c35f9726 | 2755 | reply_code = LTTNG_ERR_INVALID_PROTOCOL; |
d3ecc550 | 2756 | ret = -1; |
5312a3ed | 2757 | goto end; |
d3ecc550 JD |
2758 | } |
2759 | ||
c35f9726 JG |
2760 | for (i = 0; i < rotate_streams.stream_count; i++) { |
2761 | struct lttcomm_relayd_stream_rotation_position *position_comm = | |
28ab034a | 2762 | &((typeof(position_comm)) stream_positions.data)[i]; |
c35f9726 JG |
2763 | const struct lttcomm_relayd_stream_rotation_position pos = { |
2764 | .stream_id = be64toh(position_comm->stream_id), | |
28ab034a | 2765 | .rotate_at_seq_num = be64toh(position_comm->rotate_at_seq_num), |
c35f9726 | 2766 | }; |
5312a3ed | 2767 | |
c35f9726 JG |
2768 | stream = stream_get_by_id(pos.stream_id); |
2769 | if (!stream) { | |
2770 | reply_code = LTTNG_ERR_INVALID; | |
2771 | ret = -1; | |
2772 | goto end; | |
c6db3843 JG |
2773 | } |
2774 | ||
c35f9726 | 2775 | pthread_mutex_lock(&stream->lock); |
28ab034a | 2776 | ret = stream_set_pending_rotation(stream, next_trace_chunk, pos.rotate_at_seq_num); |
c35f9726 JG |
2777 | pthread_mutex_unlock(&stream->lock); |
2778 | if (ret) { | |
2779 | reply_code = LTTNG_ERR_FILE_CREATION_ERROR; | |
2780 | goto end; | |
c6db3843 | 2781 | } |
c35f9726 JG |
2782 | |
2783 | stream_put(stream); | |
cd9adb8b | 2784 | stream = nullptr; |
d3ecc550 JD |
2785 | } |
2786 | ||
c35f9726 | 2787 | reply_code = LTTNG_OK; |
eaeb64a9 | 2788 | ret = 0; |
d3ecc550 | 2789 | end: |
c35f9726 JG |
2790 | if (stream) { |
2791 | stream_put(stream); | |
d3ecc550 | 2792 | } |
c35f9726 JG |
2793 | |
2794 | reply.ret_code = htobe32((uint32_t) reply_code); | |
28ab034a JG |
2795 | send_ret = conn->sock->ops->sendmsg( |
2796 | conn->sock, &reply, sizeof(struct lttcomm_relayd_generic_reply), 0); | |
5312a3ed | 2797 | if (send_ret < (ssize_t) sizeof(reply)) { |
28ab034a | 2798 | ERR("Failed to send \"rotate session stream\" command reply (ret = %zd)", send_ret); |
5312a3ed | 2799 | ret = -1; |
d3ecc550 | 2800 | } |
d3ecc550 | 2801 | end_no_reply: |
c35f9726 | 2802 | lttng_trace_chunk_put(next_trace_chunk); |
d3ecc550 JD |
2803 | return ret; |
2804 | } | |
2805 | ||
e5add6d0 JG |
2806 | /* |
2807 | * relay_create_trace_chunk: create a new trace chunk | |
2808 | */ | |
28ab034a JG |
2809 | static int relay_create_trace_chunk(const struct lttcomm_relayd_hdr *recv_hdr |
2810 | __attribute__((unused)), | |
2811 | struct relay_connection *conn, | |
2812 | const struct lttng_buffer_view *payload) | |
e5add6d0 JG |
2813 | { |
2814 | int ret = 0; | |
2815 | ssize_t send_ret; | |
2816 | struct relay_session *session = conn->session; | |
2817 | struct lttcomm_relayd_create_trace_chunk *msg; | |
2818 | struct lttcomm_relayd_generic_reply reply = {}; | |
2819 | struct lttng_buffer_view header_view; | |
cd9adb8b | 2820 | struct lttng_trace_chunk *chunk = nullptr, *published_chunk = nullptr; |
e5add6d0 JG |
2821 | enum lttng_error_code reply_code = LTTNG_OK; |
2822 | enum lttng_trace_chunk_status chunk_status; | |
a7ceb342 | 2823 | const char *new_path; |
e5add6d0 JG |
2824 | |
2825 | if (!session || !conn->version_check_done) { | |
2826 | ERR("Trying to create a trace chunk before version check"); | |
2827 | ret = -1; | |
2828 | goto end_no_reply; | |
2829 | } | |
2830 | ||
2831 | if (session->major == 2 && session->minor < 11) { | |
2832 | ERR("Chunk creation command is unsupported before 2.11"); | |
2833 | ret = -1; | |
2834 | goto end_no_reply; | |
2835 | } | |
2836 | ||
2837 | header_view = lttng_buffer_view_from_view(payload, 0, sizeof(*msg)); | |
3e6e0df2 | 2838 | if (!lttng_buffer_view_is_valid(&header_view)) { |
e5add6d0 JG |
2839 | ERR("Failed to receive payload of chunk creation command"); |
2840 | ret = -1; | |
2841 | goto end_no_reply; | |
2842 | } | |
2843 | ||
2844 | /* Convert to host endianness. */ | |
2845 | msg = (typeof(msg)) header_view.data; | |
2846 | msg->chunk_id = be64toh(msg->chunk_id); | |
2847 | msg->creation_timestamp = be64toh(msg->creation_timestamp); | |
2848 | msg->override_name_length = be32toh(msg->override_name_length); | |
2849 | ||
8cd15f6a MD |
2850 | pthread_mutex_lock(&conn->session->lock); |
2851 | session->ongoing_rotation = true; | |
a7ceb342 | 2852 | if (session->current_trace_chunk && |
28ab034a | 2853 | !lttng_trace_chunk_get_name_overridden(session->current_trace_chunk)) { |
a7ceb342 | 2854 | chunk_status = lttng_trace_chunk_rename_path(session->current_trace_chunk, |
28ab034a | 2855 | DEFAULT_CHUNK_TMP_OLD_DIRECTORY); |
a7ceb342 MD |
2856 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { |
2857 | ERR("Failed to rename old chunk"); | |
2858 | ret = -1; | |
2859 | reply_code = LTTNG_ERR_UNK; | |
2860 | goto end; | |
2861 | } | |
2862 | } | |
a7ceb342 MD |
2863 | if (!session->current_trace_chunk) { |
2864 | if (!session->has_rotated) { | |
2865 | new_path = ""; | |
2866 | } else { | |
cd9adb8b | 2867 | new_path = nullptr; |
a7ceb342 MD |
2868 | } |
2869 | } else { | |
2870 | new_path = DEFAULT_CHUNK_TMP_NEW_DIRECTORY; | |
2871 | } | |
28ab034a | 2872 | chunk = lttng_trace_chunk_create(msg->chunk_id, msg->creation_timestamp, new_path); |
e5add6d0 JG |
2873 | if (!chunk) { |
2874 | ERR("Failed to create trace chunk in trace chunk creation command"); | |
2875 | ret = -1; | |
2876 | reply_code = LTTNG_ERR_NOMEM; | |
2877 | goto end; | |
2878 | } | |
7145f5e9 | 2879 | lttng_trace_chunk_set_fd_tracker(chunk, the_fd_tracker); |
e5add6d0 JG |
2880 | |
2881 | if (msg->override_name_length) { | |
2882 | const char *name; | |
28ab034a JG |
2883 | const struct lttng_buffer_view chunk_name_view = lttng_buffer_view_from_view( |
2884 | payload, sizeof(*msg), msg->override_name_length); | |
3e6e0df2 JG |
2885 | |
2886 | if (!lttng_buffer_view_is_valid(&chunk_name_view)) { | |
2887 | ERR("Invalid payload of chunk creation command (protocol error): buffer too short for expected name length"); | |
2888 | ret = -1; | |
2889 | reply_code = LTTNG_ERR_INVALID; | |
2890 | goto end; | |
2891 | } | |
e5add6d0 | 2892 | |
e5add6d0 | 2893 | name = chunk_name_view.data; |
3e6e0df2 JG |
2894 | if (name[msg->override_name_length - 1]) { |
2895 | ERR("Invalid payload of chunk creation command (protocol error): name is not null-terminated"); | |
e5add6d0 JG |
2896 | ret = -1; |
2897 | reply_code = LTTNG_ERR_INVALID; | |
2898 | goto end; | |
2899 | } | |
2900 | ||
28ab034a | 2901 | chunk_status = lttng_trace_chunk_override_name(chunk, chunk_name_view.data); |
e5add6d0 JG |
2902 | switch (chunk_status) { |
2903 | case LTTNG_TRACE_CHUNK_STATUS_OK: | |
2904 | break; | |
2905 | case LTTNG_TRACE_CHUNK_STATUS_INVALID_ARGUMENT: | |
2906 | ERR("Failed to set the name of new trace chunk in trace chunk creation command (invalid name)"); | |
2907 | reply_code = LTTNG_ERR_INVALID; | |
2908 | ret = -1; | |
2909 | goto end; | |
2910 | default: | |
2911 | ERR("Failed to set the name of new trace chunk in trace chunk creation command (unknown error)"); | |
2912 | reply_code = LTTNG_ERR_UNK; | |
2913 | ret = -1; | |
2914 | goto end; | |
2915 | } | |
2916 | } | |
2917 | ||
e5add6d0 JG |
2918 | chunk_status = lttng_trace_chunk_set_credentials_current_user(chunk); |
2919 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { | |
2920 | reply_code = LTTNG_ERR_UNK; | |
2921 | ret = -1; | |
2922 | goto end; | |
2923 | } | |
2924 | ||
a0377dfe | 2925 | LTTNG_ASSERT(conn->session->output_directory); |
28ab034a | 2926 | chunk_status = lttng_trace_chunk_set_as_owner(chunk, conn->session->output_directory); |
e5add6d0 JG |
2927 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { |
2928 | reply_code = LTTNG_ERR_UNK; | |
2929 | ret = -1; | |
2930 | goto end; | |
2931 | } | |
2932 | ||
2933 | published_chunk = sessiond_trace_chunk_registry_publish_chunk( | |
28ab034a JG |
2934 | sessiond_trace_chunk_registry, |
2935 | conn->session->sessiond_uuid, | |
2936 | conn->session->id_sessiond.is_set ? conn->session->id_sessiond.value : | |
2937 | conn->session->id, | |
2938 | chunk); | |
e5add6d0 | 2939 | if (!published_chunk) { |
c70636a7 | 2940 | char uuid_str[LTTNG_UUID_STR_LEN]; |
e5add6d0 JG |
2941 | |
2942 | lttng_uuid_to_str(conn->session->sessiond_uuid, uuid_str); | |
28ab034a JG |
2943 | ERR("Failed to publish chunk: sessiond_uuid = %s, session_id = %" PRIu64 |
2944 | ", chunk_id = %" PRIu64, | |
2945 | uuid_str, | |
2946 | conn->session->id, | |
2947 | msg->chunk_id); | |
e5add6d0 JG |
2948 | ret = -1; |
2949 | reply_code = LTTNG_ERR_NOMEM; | |
2950 | goto end; | |
2951 | } | |
2952 | ||
62bad3bf JG |
2953 | if (conn->session->pending_closure_trace_chunk) { |
2954 | /* | |
2955 | * Invalid; this means a second create_trace_chunk command was | |
2956 | * received before a close_trace_chunk. | |
2957 | */ | |
2958 | ERR("Invalid trace chunk close command received; a trace chunk is already waiting for a trace chunk close command"); | |
2959 | reply_code = LTTNG_ERR_INVALID_PROTOCOL; | |
2960 | ret = -1; | |
8cd15f6a | 2961 | goto end; |
62bad3bf | 2962 | } |
28ab034a | 2963 | conn->session->pending_closure_trace_chunk = conn->session->current_trace_chunk; |
e5add6d0 | 2964 | conn->session->current_trace_chunk = published_chunk; |
cd9adb8b | 2965 | published_chunk = nullptr; |
a7ceb342 MD |
2966 | if (!conn->session->pending_closure_trace_chunk) { |
2967 | session->ongoing_rotation = false; | |
2968 | } | |
e5add6d0 | 2969 | end: |
8cd15f6a | 2970 | pthread_mutex_unlock(&conn->session->lock); |
e5add6d0 | 2971 | reply.ret_code = htobe32((uint32_t) reply_code); |
28ab034a JG |
2972 | send_ret = conn->sock->ops->sendmsg( |
2973 | conn->sock, &reply, sizeof(struct lttcomm_relayd_generic_reply), 0); | |
e5add6d0 | 2974 | if (send_ret < (ssize_t) sizeof(reply)) { |
28ab034a | 2975 | ERR("Failed to send \"create trace chunk\" command reply (ret = %zd)", send_ret); |
e5add6d0 JG |
2976 | ret = -1; |
2977 | } | |
2978 | end_no_reply: | |
2979 | lttng_trace_chunk_put(chunk); | |
2980 | lttng_trace_chunk_put(published_chunk); | |
e5add6d0 JG |
2981 | return ret; |
2982 | } | |
2983 | ||
bbc4768c JG |
2984 | /* |
2985 | * relay_close_trace_chunk: close a trace chunk | |
2986 | */ | |
28ab034a JG |
2987 | static int relay_close_trace_chunk(const struct lttcomm_relayd_hdr *recv_hdr |
2988 | __attribute__((unused)), | |
2989 | struct relay_connection *conn, | |
2990 | const struct lttng_buffer_view *payload) | |
bbc4768c | 2991 | { |
9898f786 | 2992 | int ret = 0, buf_ret; |
bbc4768c JG |
2993 | ssize_t send_ret; |
2994 | struct relay_session *session = conn->session; | |
2995 | struct lttcomm_relayd_close_trace_chunk *msg; | |
ecd1a12f | 2996 | struct lttcomm_relayd_close_trace_chunk_reply reply = {}; |
bbc4768c | 2997 | struct lttng_buffer_view header_view; |
cd9adb8b | 2998 | struct lttng_trace_chunk *chunk = nullptr; |
bbc4768c JG |
2999 | enum lttng_error_code reply_code = LTTNG_OK; |
3000 | enum lttng_trace_chunk_status chunk_status; | |
3001 | uint64_t chunk_id; | |
c35f9726 | 3002 | LTTNG_OPTIONAL(enum lttng_trace_chunk_command_type) close_command = {}; |
bbc4768c | 3003 | time_t close_timestamp; |
ecd1a12f MD |
3004 | char closed_trace_chunk_path[LTTNG_PATH_MAX]; |
3005 | size_t path_length = 0; | |
cd9adb8b | 3006 | const char *chunk_name = nullptr; |
ecd1a12f | 3007 | struct lttng_dynamic_buffer reply_payload; |
a7ceb342 | 3008 | const char *new_path; |
ecd1a12f MD |
3009 | |
3010 | lttng_dynamic_buffer_init(&reply_payload); | |
bbc4768c JG |
3011 | |
3012 | if (!session || !conn->version_check_done) { | |
3013 | ERR("Trying to close a trace chunk before version check"); | |
3014 | ret = -1; | |
3015 | goto end_no_reply; | |
3016 | } | |
3017 | ||
3018 | if (session->major == 2 && session->minor < 11) { | |
3019 | ERR("Chunk close command is unsupported before 2.11"); | |
3020 | ret = -1; | |
3021 | goto end_no_reply; | |
3022 | } | |
3023 | ||
3024 | header_view = lttng_buffer_view_from_view(payload, 0, sizeof(*msg)); | |
3e6e0df2 | 3025 | if (!lttng_buffer_view_is_valid(&header_view)) { |
bbc4768c JG |
3026 | ERR("Failed to receive payload of chunk close command"); |
3027 | ret = -1; | |
3028 | goto end_no_reply; | |
3029 | } | |
3030 | ||
3031 | /* Convert to host endianness. */ | |
3032 | msg = (typeof(msg)) header_view.data; | |
3033 | chunk_id = be64toh(msg->chunk_id); | |
3034 | close_timestamp = (time_t) be64toh(msg->close_timestamp); | |
ac497a37 SM |
3035 | close_command.value = (lttng_trace_chunk_command_type) be32toh(msg->close_command.value); |
3036 | close_command.is_set = msg->close_command.is_set; | |
bbc4768c | 3037 | |
28ab034a JG |
3038 | chunk = sessiond_trace_chunk_registry_get_chunk(sessiond_trace_chunk_registry, |
3039 | conn->session->sessiond_uuid, | |
3040 | conn->session->id_sessiond.is_set ? | |
3041 | conn->session->id_sessiond.value : | |
3042 | conn->session->id, | |
3043 | chunk_id); | |
bbc4768c | 3044 | if (!chunk) { |
c70636a7 | 3045 | char uuid_str[LTTNG_UUID_STR_LEN]; |
bbc4768c JG |
3046 | |
3047 | lttng_uuid_to_str(conn->session->sessiond_uuid, uuid_str); | |
28ab034a JG |
3048 | ERR("Failed to find chunk to close: sessiond_uuid = %s, session_id = %" PRIu64 |
3049 | ", chunk_id = %" PRIu64, | |
3050 | uuid_str, | |
3051 | conn->session->id, | |
3052 | msg->chunk_id); | |
bbc4768c JG |
3053 | ret = -1; |
3054 | reply_code = LTTNG_ERR_NOMEM; | |
3055 | goto end; | |
3056 | } | |
3057 | ||
62bad3bf | 3058 | pthread_mutex_lock(&session->lock); |
28ab034a | 3059 | if (close_command.is_set && close_command.value == LTTNG_TRACE_CHUNK_COMMAND_TYPE_DELETE) { |
f77a6ec2 MD |
3060 | /* |
3061 | * Clear command. It is a protocol error to ask for a | |
3062 | * clear on a relay which does not allow it. Querying | |
3063 | * the configuration allows figuring out whether | |
3064 | * clearing is allowed before doing the clear. | |
3065 | */ | |
3066 | if (!opt_allow_clear) { | |
3067 | ret = -1; | |
3068 | reply_code = LTTNG_ERR_INVALID_PROTOCOL; | |
3069 | goto end_unlock_session; | |
3070 | } | |
3071 | } | |
28ab034a | 3072 | if (session->pending_closure_trace_chunk && session->pending_closure_trace_chunk != chunk) { |
62bad3bf | 3073 | ERR("Trace chunk close command for session \"%s\" does not target the trace chunk pending closure", |
28ab034a | 3074 | session->session_name); |
62bad3bf JG |
3075 | reply_code = LTTNG_ERR_INVALID_PROTOCOL; |
3076 | ret = -1; | |
3077 | goto end_unlock_session; | |
3078 | } | |
3079 | ||
a7ceb342 | 3080 | if (session->current_trace_chunk && session->current_trace_chunk != chunk && |
28ab034a | 3081 | !lttng_trace_chunk_get_name_overridden(session->current_trace_chunk)) { |
a7ceb342 | 3082 | if (close_command.is_set && |
28ab034a JG |
3083 | close_command.value == LTTNG_TRACE_CHUNK_COMMAND_TYPE_DELETE && |
3084 | !session->has_rotated) { | |
a7ceb342 MD |
3085 | /* New chunk stays in session output directory. */ |
3086 | new_path = ""; | |
3087 | } else { | |
3088 | /* Use chunk name for new chunk. */ | |
cd9adb8b | 3089 | new_path = nullptr; |
a7ceb342 MD |
3090 | } |
3091 | /* Rename new chunk path. */ | |
28ab034a JG |
3092 | chunk_status = |
3093 | lttng_trace_chunk_rename_path(session->current_trace_chunk, new_path); | |
a7ceb342 MD |
3094 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { |
3095 | ret = -1; | |
88188199 | 3096 | goto end_unlock_session; |
a7ceb342 MD |
3097 | } |
3098 | session->ongoing_rotation = false; | |
3099 | } | |
3100 | if ((!close_command.is_set || | |
28ab034a JG |
3101 | close_command.value == LTTNG_TRACE_CHUNK_COMMAND_TYPE_NO_OPERATION) && |
3102 | !lttng_trace_chunk_get_name_overridden(chunk)) { | |
a7ceb342 MD |
3103 | const char *old_path; |
3104 | ||
3105 | if (!session->has_rotated) { | |
3106 | old_path = ""; | |
3107 | } else { | |
cd9adb8b | 3108 | old_path = nullptr; |
a7ceb342 MD |
3109 | } |
3110 | /* We need to move back the .tmp_old_chunk to its rightful place. */ | |
3111 | chunk_status = lttng_trace_chunk_rename_path(chunk, old_path); | |
3112 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { | |
3113 | ret = -1; | |
88188199 | 3114 | goto end_unlock_session; |
a7ceb342 MD |
3115 | } |
3116 | } | |
28ab034a | 3117 | chunk_status = lttng_trace_chunk_set_close_timestamp(chunk, close_timestamp); |
bbc4768c JG |
3118 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { |
3119 | ERR("Failed to set trace chunk close timestamp"); | |
3120 | ret = -1; | |
3121 | reply_code = LTTNG_ERR_UNK; | |
62bad3bf | 3122 | goto end_unlock_session; |
bbc4768c JG |
3123 | } |
3124 | ||
3125 | if (close_command.is_set) { | |
28ab034a | 3126 | chunk_status = lttng_trace_chunk_set_close_command(chunk, close_command.value); |
bbc4768c JG |
3127 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { |
3128 | ret = -1; | |
3129 | reply_code = LTTNG_ERR_INVALID; | |
62bad3bf | 3130 | goto end_unlock_session; |
bbc4768c JG |
3131 | } |
3132 | } | |
cd9adb8b | 3133 | chunk_status = lttng_trace_chunk_get_name(chunk, &chunk_name, nullptr); |
ecd1a12f MD |
3134 | if (chunk_status != LTTNG_TRACE_CHUNK_STATUS_OK) { |
3135 | ERR("Failed to get chunk name"); | |
3136 | ret = -1; | |
3137 | reply_code = LTTNG_ERR_UNK; | |
3138 | goto end_unlock_session; | |
3139 | } | |
3140 | if (!session->has_rotated && !session->snapshot) { | |
3141 | ret = lttng_strncpy(closed_trace_chunk_path, | |
28ab034a JG |
3142 | session->output_path, |
3143 | sizeof(closed_trace_chunk_path)); | |
ecd1a12f MD |
3144 | if (ret) { |
3145 | ERR("Failed to send trace chunk path: path length of %zu bytes exceeds the maximal allowed length of %zu bytes", | |
28ab034a JG |
3146 | strlen(session->output_path), |
3147 | sizeof(closed_trace_chunk_path)); | |
ecd1a12f MD |
3148 | reply_code = LTTNG_ERR_NOMEM; |
3149 | ret = -1; | |
3150 | goto end_unlock_session; | |
3151 | } | |
3152 | } else { | |
3153 | if (session->snapshot) { | |
3154 | ret = snprintf(closed_trace_chunk_path, | |
28ab034a JG |
3155 | sizeof(closed_trace_chunk_path), |
3156 | "%s/%s", | |
3157 | session->output_path, | |
3158 | chunk_name); | |
ecd1a12f MD |
3159 | } else { |
3160 | ret = snprintf(closed_trace_chunk_path, | |
28ab034a JG |
3161 | sizeof(closed_trace_chunk_path), |
3162 | "%s/" DEFAULT_ARCHIVED_TRACE_CHUNKS_DIRECTORY "/%s", | |
3163 | session->output_path, | |
3164 | chunk_name); | |
ecd1a12f MD |
3165 | } |
3166 | if (ret < 0 || ret == sizeof(closed_trace_chunk_path)) { | |
3167 | ERR("Failed to format closed trace chunk resulting path"); | |
3168 | reply_code = ret < 0 ? LTTNG_ERR_UNK : LTTNG_ERR_NOMEM; | |
3169 | ret = -1; | |
3170 | goto end_unlock_session; | |
3171 | } | |
3172 | } | |
489ea154 | 3173 | if (close_command.is_set && |
28ab034a | 3174 | close_command.value == LTTNG_TRACE_CHUNK_COMMAND_TYPE_MOVE_TO_COMPLETED) { |
489ea154 MD |
3175 | session->has_rotated = true; |
3176 | } | |
ecd1a12f MD |
3177 | DBG("Reply chunk path on close: %s", closed_trace_chunk_path); |
3178 | path_length = strlen(closed_trace_chunk_path) + 1; | |
3179 | if (path_length > UINT32_MAX) { | |
3180 | ERR("Closed trace chunk path exceeds the maximal length allowed by the protocol"); | |
3181 | ret = -1; | |
3182 | reply_code = LTTNG_ERR_INVALID_PROTOCOL; | |
3183 | goto end_unlock_session; | |
3184 | } | |
bbc4768c | 3185 | |
c35f9726 JG |
3186 | if (session->current_trace_chunk == chunk) { |
3187 | /* | |
3188 | * After a trace chunk close command, no new streams | |
3189 | * referencing the chunk may be created. Hence, on the | |
3190 | * event that no new trace chunk have been created for | |
3191 | * the session, the reference to the current trace chunk | |
3192 | * is released in order to allow it to be reclaimed when | |
3193 | * the last stream releases its reference to it. | |
3194 | */ | |
3195 | lttng_trace_chunk_put(session->current_trace_chunk); | |
cd9adb8b | 3196 | session->current_trace_chunk = nullptr; |
c35f9726 | 3197 | } |
62bad3bf | 3198 | lttng_trace_chunk_put(session->pending_closure_trace_chunk); |
cd9adb8b | 3199 | session->pending_closure_trace_chunk = nullptr; |
62bad3bf | 3200 | end_unlock_session: |
c35f9726 JG |
3201 | pthread_mutex_unlock(&session->lock); |
3202 | ||
bbc4768c | 3203 | end: |
ecd1a12f MD |
3204 | reply.generic.ret_code = htobe32((uint32_t) reply_code); |
3205 | reply.path_length = htobe32((uint32_t) path_length); | |
28ab034a | 3206 | buf_ret = lttng_dynamic_buffer_append(&reply_payload, &reply, sizeof(reply)); |
9898f786 | 3207 | if (buf_ret) { |
ecd1a12f MD |
3208 | ERR("Failed to append \"close trace chunk\" command reply header to payload buffer"); |
3209 | goto end_no_reply; | |
3210 | } | |
3211 | ||
3212 | if (reply_code == LTTNG_OK) { | |
28ab034a JG |
3213 | buf_ret = lttng_dynamic_buffer_append( |
3214 | &reply_payload, closed_trace_chunk_path, path_length); | |
9898f786 | 3215 | if (buf_ret) { |
ecd1a12f MD |
3216 | ERR("Failed to append \"close trace chunk\" command reply path to payload buffer"); |
3217 | goto end_no_reply; | |
3218 | } | |
3219 | } | |
3220 | ||
28ab034a | 3221 | send_ret = conn->sock->ops->sendmsg(conn->sock, reply_payload.data, reply_payload.size, 0); |
ecd1a12f MD |
3222 | if (send_ret < reply_payload.size) { |
3223 | ERR("Failed to send \"close trace chunk\" command reply of %zu bytes (ret = %zd)", | |
28ab034a JG |
3224 | reply_payload.size, |
3225 | send_ret); | |
bbc4768c | 3226 | ret = -1; |
ecd1a12f | 3227 | goto end_no_reply; |
bbc4768c JG |
3228 | } |
3229 | end_no_reply: | |
3230 | lttng_trace_chunk_put(chunk); | |
ecd1a12f | 3231 | lttng_dynamic_buffer_reset(&reply_payload); |
bbc4768c JG |
3232 | return ret; |
3233 | } | |
3234 | ||
c35f9726 JG |
3235 | /* |
3236 | * relay_trace_chunk_exists: check if a trace chunk exists | |
3237 | */ | |
28ab034a JG |
3238 | static int relay_trace_chunk_exists(const struct lttcomm_relayd_hdr *recv_hdr |
3239 | __attribute__((unused)), | |
3240 | struct relay_connection *conn, | |
3241 | const struct lttng_buffer_view *payload) | |
c35f9726 JG |
3242 | { |
3243 | int ret = 0; | |
3244 | ssize_t send_ret; | |
3245 | struct relay_session *session = conn->session; | |
3246 | struct lttcomm_relayd_trace_chunk_exists *msg; | |
3247 | struct lttcomm_relayd_trace_chunk_exists_reply reply = {}; | |
3248 | struct lttng_buffer_view header_view; | |
c35f9726 | 3249 | uint64_t chunk_id; |
6b584c2e | 3250 | bool chunk_exists; |
c35f9726 JG |
3251 | |
3252 | if (!session || !conn->version_check_done) { | |
0f1b1d25 | 3253 | ERR("Trying to check for the presence of a trace chunk before version check"); |
c35f9726 JG |
3254 | ret = -1; |
3255 | goto end_no_reply; | |
3256 | } | |
3257 | ||
3258 | if (session->major == 2 && session->minor < 11) { | |
8a82be4c | 3259 | ERR("Chunk exists command is unsupported before 2.11"); |
c35f9726 JG |
3260 | ret = -1; |
3261 | goto end_no_reply; | |
3262 | } | |
3263 | ||
3264 | header_view = lttng_buffer_view_from_view(payload, 0, sizeof(*msg)); | |
3e6e0df2 | 3265 | if (!lttng_buffer_view_is_valid(&header_view)) { |
8a82be4c | 3266 | ERR("Failed to receive payload of chunk exists command"); |
c35f9726 JG |
3267 | ret = -1; |
3268 | goto end_no_reply; | |
3269 | } | |
3270 | ||
3271 | /* Convert to host endianness. */ | |
3272 | msg = (typeof(msg)) header_view.data; | |
3273 | chunk_id = be64toh(msg->chunk_id); | |
3274 | ||
28ab034a JG |
3275 | ret = sessiond_trace_chunk_registry_chunk_exists(sessiond_trace_chunk_registry, |
3276 | conn->session->sessiond_uuid, | |
3277 | conn->session->id, | |
3278 | chunk_id, | |
3279 | &chunk_exists); | |
6b584c2e JG |
3280 | /* |
3281 | * If ret is not 0, send the reply and report the error to the caller. | |
3282 | * It is a protocol (or internal) error and the session/connection | |
3283 | * should be torn down. | |
3284 | */ | |
28ab034a JG |
3285 | reply.generic.ret_code = |
3286 | htobe32((uint32_t) (ret == 0 ? LTTNG_OK : LTTNG_ERR_INVALID_PROTOCOL)); | |
ac497a37 SM |
3287 | reply.trace_chunk_exists = ret == 0 ? chunk_exists : 0; |
3288 | ||
28ab034a | 3289 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
c35f9726 | 3290 | if (send_ret < (ssize_t) sizeof(reply)) { |
28ab034a | 3291 | ERR("Failed to send \"create trace chunk\" command reply (ret = %zd)", send_ret); |
c35f9726 JG |
3292 | ret = -1; |
3293 | } | |
3294 | end_no_reply: | |
c35f9726 JG |
3295 | return ret; |
3296 | } | |
3297 | ||
8614e600 MD |
3298 | /* |
3299 | * relay_get_configuration: query whether feature is available | |
3300 | */ | |
28ab034a JG |
3301 | static int relay_get_configuration(const struct lttcomm_relayd_hdr *recv_hdr |
3302 | __attribute__((unused)), | |
3303 | struct relay_connection *conn, | |
3304 | const struct lttng_buffer_view *payload) | |
8614e600 MD |
3305 | { |
3306 | int ret = 0; | |
3307 | ssize_t send_ret; | |
3308 | struct lttcomm_relayd_get_configuration *msg; | |
3309 | struct lttcomm_relayd_get_configuration_reply reply = {}; | |
3310 | struct lttng_buffer_view header_view; | |
3311 | uint64_t query_flags = 0; | |
3312 | uint64_t result_flags = 0; | |
3313 | ||
3314 | header_view = lttng_buffer_view_from_view(payload, 0, sizeof(*msg)); | |
3e6e0df2 | 3315 | if (!lttng_buffer_view_is_valid(&header_view)) { |
8614e600 MD |
3316 | ERR("Failed to receive payload of chunk close command"); |
3317 | ret = -1; | |
3318 | goto end_no_reply; | |
3319 | } | |
3320 | ||
3321 | /* Convert to host endianness. */ | |
3322 | msg = (typeof(msg)) header_view.data; | |
3323 | query_flags = be64toh(msg->query_flags); | |
3324 | ||
3325 | if (query_flags) { | |
3326 | ret = LTTNG_ERR_INVALID_PROTOCOL; | |
3327 | goto reply; | |
3328 | } | |
3329 | if (opt_allow_clear) { | |
3330 | result_flags |= LTTCOMM_RELAYD_CONFIGURATION_FLAG_CLEAR_ALLOWED; | |
3331 | } | |
3332 | ret = 0; | |
3333 | reply: | |
28ab034a JG |
3334 | reply.generic.ret_code = |
3335 | htobe32((uint32_t) (ret == 0 ? LTTNG_OK : LTTNG_ERR_INVALID_PROTOCOL)); | |
ac497a37 SM |
3336 | reply.relayd_configuration_flags = htobe64(result_flags); |
3337 | ||
28ab034a | 3338 | send_ret = conn->sock->ops->sendmsg(conn->sock, &reply, sizeof(reply), 0); |
8614e600 | 3339 | if (send_ret < (ssize_t) sizeof(reply)) { |
28ab034a | 3340 | ERR("Failed to send \"get configuration\" command reply (ret = %zd)", send_ret); |
8614e600 MD |
3341 | ret = -1; |
3342 | } | |
3343 | end_no_reply: | |
3344 | return ret; | |
3345 | } | |
3346 | ||
5312a3ed | 3347 | static int relay_process_control_command(struct relay_connection *conn, |
28ab034a JG |
3348 | const struct lttcomm_relayd_hdr *header, |
3349 | const struct lttng_buffer_view *payload) | |
b8aa1682 JD |
3350 | { |
3351 | int ret = 0; | |
3352 | ||
00e71031 | 3353 | DBG3("Processing \"%s\" command for socket %i", |
28ab034a JG |
3354 | lttcomm_relayd_command_str((lttcomm_relayd_command) header->cmd), |
3355 | conn->sock->fd); | |
5312a3ed | 3356 | switch (header->cmd) { |
b8aa1682 | 3357 | case RELAYD_CREATE_SESSION: |
5312a3ed | 3358 | ret = relay_create_session(header, conn, payload); |
b8aa1682 | 3359 | break; |
b8aa1682 | 3360 | case RELAYD_ADD_STREAM: |
5312a3ed | 3361 | ret = relay_add_stream(header, conn, payload); |
b8aa1682 JD |
3362 | break; |
3363 | case RELAYD_START_DATA: | |
5312a3ed | 3364 | ret = relay_start(header, conn, payload); |
b8aa1682 JD |
3365 | break; |
3366 | case RELAYD_SEND_METADATA: | |
5312a3ed | 3367 | ret = relay_recv_metadata(header, conn, payload); |
b8aa1682 JD |
3368 | break; |
3369 | case RELAYD_VERSION: | |
5312a3ed | 3370 | ret = relay_send_version(header, conn, payload); |
b8aa1682 | 3371 | break; |
173af62f | 3372 | case RELAYD_CLOSE_STREAM: |
5312a3ed | 3373 | ret = relay_close_stream(header, conn, payload); |
173af62f | 3374 | break; |
6d805429 | 3375 | case RELAYD_DATA_PENDING: |
5312a3ed | 3376 | ret = relay_data_pending(header, conn, payload); |
c8f59ee5 DG |
3377 | break; |
3378 | case RELAYD_QUIESCENT_CONTROL: | |
5312a3ed | 3379 | ret = relay_quiescent_control(header, conn, payload); |
c8f59ee5 | 3380 | break; |
f7079f67 | 3381 | case RELAYD_BEGIN_DATA_PENDING: |
5312a3ed | 3382 | ret = relay_begin_data_pending(header, conn, payload); |
f7079f67 DG |
3383 | break; |
3384 | case RELAYD_END_DATA_PENDING: | |
5312a3ed | 3385 | ret = relay_end_data_pending(header, conn, payload); |
f7079f67 | 3386 | break; |
1c20f0e2 | 3387 | case RELAYD_SEND_INDEX: |
5312a3ed | 3388 | ret = relay_recv_index(header, conn, payload); |
1c20f0e2 | 3389 | break; |
a4baae1b | 3390 | case RELAYD_STREAMS_SENT: |
5312a3ed | 3391 | ret = relay_streams_sent(header, conn, payload); |
a4baae1b | 3392 | break; |
93ec662e | 3393 | case RELAYD_RESET_METADATA: |
5312a3ed | 3394 | ret = relay_reset_metadata(header, conn, payload); |
93ec662e | 3395 | break; |
c35f9726 | 3396 | case RELAYD_ROTATE_STREAMS: |
c35f9726 | 3397 | ret = relay_rotate_session_streams(header, conn, payload); |
d3ecc550 | 3398 | break; |
e5add6d0 | 3399 | case RELAYD_CREATE_TRACE_CHUNK: |
e5add6d0 JG |
3400 | ret = relay_create_trace_chunk(header, conn, payload); |
3401 | break; | |
bbc4768c | 3402 | case RELAYD_CLOSE_TRACE_CHUNK: |
bbc4768c JG |
3403 | ret = relay_close_trace_chunk(header, conn, payload); |
3404 | break; | |
c35f9726 | 3405 | case RELAYD_TRACE_CHUNK_EXISTS: |
c35f9726 JG |
3406 | ret = relay_trace_chunk_exists(header, conn, payload); |
3407 | break; | |
8614e600 | 3408 | case RELAYD_GET_CONFIGURATION: |
8614e600 MD |
3409 | ret = relay_get_configuration(header, conn, payload); |
3410 | break; | |
b8aa1682 JD |
3411 | case RELAYD_UPDATE_SYNC_INFO: |
3412 | default: | |
5312a3ed | 3413 | ERR("Received unknown command (%u)", header->cmd); |
58eb9381 | 3414 | relay_unknown_command(conn); |
b8aa1682 JD |
3415 | ret = -1; |
3416 | goto end; | |
3417 | } | |
3418 | ||
3419 | end: | |
3420 | return ret; | |
3421 | } | |
3422 | ||
28ab034a JG |
3423 | static enum relay_connection_status |
3424 | relay_process_control_receive_payload(struct relay_connection *conn) | |
5312a3ed JG |
3425 | { |
3426 | int ret = 0; | |
5569b118 | 3427 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; |
28ab034a | 3428 | struct lttng_dynamic_buffer *reception_buffer = &conn->protocol.ctrl.reception_buffer; |
5312a3ed | 3429 | struct ctrl_connection_state_receive_payload *state = |
28ab034a | 3430 | &conn->protocol.ctrl.state.receive_payload; |
5312a3ed JG |
3431 | struct lttng_buffer_view payload_view; |
3432 | ||
3433 | if (state->left_to_receive == 0) { | |
3434 | /* Short-circuit for payload-less commands. */ | |
3435 | goto reception_complete; | |
3436 | } | |
3437 | ||
3438 | ret = conn->sock->ops->recvmsg(conn->sock, | |
28ab034a JG |
3439 | reception_buffer->data + state->received, |
3440 | state->left_to_receive, | |
3441 | MSG_DONTWAIT); | |
5312a3ed | 3442 | if (ret < 0) { |
942003e5 MJ |
3443 | DIAGNOSTIC_PUSH |
3444 | DIAGNOSTIC_IGNORE_LOGICAL_OP | |
5569b118 | 3445 | if (errno != EAGAIN && errno != EWOULDBLOCK) { |
28ab034a JG |
3446 | DIAGNOSTIC_POP |
3447 | PERROR("Unable to receive command payload on sock %d", conn->sock->fd); | |
5569b118 JG |
3448 | status = RELAY_CONNECTION_STATUS_ERROR; |
3449 | } | |
5312a3ed JG |
3450 | goto end; |
3451 | } else if (ret == 0) { | |
3452 | DBG("Socket %d performed an orderly shutdown (received EOF)", conn->sock->fd); | |
5569b118 | 3453 | status = RELAY_CONNECTION_STATUS_CLOSED; |
5312a3ed JG |
3454 | goto end; |
3455 | } | |
3456 | ||
a0377dfe FD |
3457 | LTTNG_ASSERT(ret > 0); |
3458 | LTTNG_ASSERT(ret <= state->left_to_receive); | |
5312a3ed JG |
3459 | |
3460 | state->left_to_receive -= ret; | |
3461 | state->received += ret; | |
3462 | ||
3463 | if (state->left_to_receive > 0) { | |
3464 | /* | |
3465 | * Can't transition to the protocol's next state, wait to | |
3466 | * receive the rest of the header. | |
3467 | */ | |
28ab034a JG |
3468 | DBG3("Partial reception of control connection protocol payload (received %" PRIu64 |
3469 | " bytes, %" PRIu64 " bytes left to receive, fd = %i)", | |
3470 | state->received, | |
3471 | state->left_to_receive, | |
3472 | conn->sock->fd); | |
5312a3ed JG |
3473 | goto end; |
3474 | } | |
3475 | ||
3476 | reception_complete: | |
3477 | DBG("Done receiving control command payload: fd = %i, payload size = %" PRIu64 " bytes", | |
28ab034a JG |
3478 | conn->sock->fd, |
3479 | state->received); | |
5312a3ed JG |
3480 | /* |
3481 | * The payload required to process the command has been received. | |
3482 | * A view to the reception buffer is forwarded to the various | |
3483 | * commands and the state of the control is reset on success. | |
3484 | * | |
3485 | * Commands are responsible for sending their reply to the peer. | |
3486 | */ | |
28ab034a JG |
3487 | payload_view = lttng_buffer_view_from_dynamic_buffer(reception_buffer, 0, -1); |
3488 | ret = relay_process_control_command(conn, &state->header, &payload_view); | |
5312a3ed | 3489 | if (ret < 0) { |
5569b118 | 3490 | status = RELAY_CONNECTION_STATUS_ERROR; |
5312a3ed JG |
3491 | goto end; |
3492 | } | |
3493 | ||
3494 | ret = connection_reset_protocol_state(conn); | |
5569b118 JG |
3495 | if (ret) { |
3496 | status = RELAY_CONNECTION_STATUS_ERROR; | |
3497 | } | |
5312a3ed | 3498 | end: |
5569b118 | 3499 | return status; |
5312a3ed JG |
3500 | } |
3501 | ||
28ab034a JG |
3502 | static enum relay_connection_status |
3503 | relay_process_control_receive_header(struct relay_connection *conn) | |
5312a3ed JG |
3504 | { |
3505 | int ret = 0; | |
5569b118 | 3506 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; |
5312a3ed | 3507 | struct lttcomm_relayd_hdr header; |
28ab034a | 3508 | struct lttng_dynamic_buffer *reception_buffer = &conn->protocol.ctrl.reception_buffer; |
5312a3ed | 3509 | struct ctrl_connection_state_receive_header *state = |
28ab034a | 3510 | &conn->protocol.ctrl.state.receive_header; |
5312a3ed | 3511 | |
a0377dfe | 3512 | LTTNG_ASSERT(state->left_to_receive != 0); |
5312a3ed JG |
3513 | |
3514 | ret = conn->sock->ops->recvmsg(conn->sock, | |
28ab034a JG |
3515 | reception_buffer->data + state->received, |
3516 | state->left_to_receive, | |
3517 | MSG_DONTWAIT); | |
5312a3ed | 3518 | if (ret < 0) { |
942003e5 MJ |
3519 | DIAGNOSTIC_PUSH |
3520 | DIAGNOSTIC_IGNORE_LOGICAL_OP | |
5569b118 | 3521 | if (errno != EAGAIN && errno != EWOULDBLOCK) { |
28ab034a | 3522 | DIAGNOSTIC_POP |
5569b118 | 3523 | PERROR("Unable to receive control command header on sock %d", |
28ab034a | 3524 | conn->sock->fd); |
5569b118 JG |
3525 | status = RELAY_CONNECTION_STATUS_ERROR; |
3526 | } | |
5312a3ed JG |
3527 | goto end; |
3528 | } else if (ret == 0) { | |
3529 | DBG("Socket %d performed an orderly shutdown (received EOF)", conn->sock->fd); | |
5569b118 | 3530 | status = RELAY_CONNECTION_STATUS_CLOSED; |
5312a3ed JG |
3531 | goto end; |
3532 | } | |
3533 | ||
a0377dfe FD |
3534 | LTTNG_ASSERT(ret > 0); |
3535 | LTTNG_ASSERT(ret <= state->left_to_receive); | |
5312a3ed JG |
3536 | |
3537 | state->left_to_receive -= ret; | |
3538 | state->received += ret; | |
3539 | ||
3540 | if (state->left_to_receive > 0) { | |
3541 | /* | |
3542 | * Can't transition to the protocol's next state, wait to | |
3543 | * receive the rest of the header. | |
3544 | */ | |
28ab034a JG |
3545 | DBG3("Partial reception of control connection protocol header (received %" PRIu64 |
3546 | " bytes, %" PRIu64 " bytes left to receive, fd = %i)", | |
3547 | state->received, | |
3548 | state->left_to_receive, | |
3549 | conn->sock->fd); | |
5312a3ed JG |
3550 | goto end; |
3551 | } | |
3552 | ||
3553 | /* Transition to next state: receiving the command's payload. */ | |
28ab034a | 3554 | conn->protocol.ctrl.state_id = CTRL_CONNECTION_STATE_RECEIVE_PAYLOAD; |
5312a3ed JG |
3555 | memcpy(&header, reception_buffer->data, sizeof(header)); |
3556 | header.circuit_id = be64toh(header.circuit_id); | |
3557 | header.data_size = be64toh(header.data_size); | |
3558 | header.cmd = be32toh(header.cmd); | |
3559 | header.cmd_version = be32toh(header.cmd_version); | |
28ab034a | 3560 | memcpy(&conn->protocol.ctrl.state.receive_payload.header, &header, sizeof(header)); |
5312a3ed | 3561 | |
28ab034a JG |
3562 | DBG("Done receiving control command header: fd = %i, cmd = %s, cmd_version = %" PRIu32 |
3563 | ", payload size = %" PRIu64 " bytes", | |
3564 | conn->sock->fd, | |
3565 | lttcomm_relayd_command_str((enum lttcomm_relayd_command) header.cmd), | |
3566 | header.cmd_version, | |
3567 | header.data_size); | |
5312a3ed | 3568 | |
715e6fb1 | 3569 | if (header.data_size > DEFAULT_NETWORK_RELAYD_CTRL_MAX_PAYLOAD_SIZE) { |
28ab034a JG |
3570 | ERR("Command header indicates a payload (%" PRIu64 |
3571 | " bytes) that exceeds the maximal payload size allowed on a control connection.", | |
3572 | header.data_size); | |
5569b118 | 3573 | status = RELAY_CONNECTION_STATUS_ERROR; |
5312a3ed JG |
3574 | goto end; |
3575 | } | |
3576 | ||
28ab034a | 3577 | conn->protocol.ctrl.state.receive_payload.left_to_receive = header.data_size; |
5312a3ed | 3578 | conn->protocol.ctrl.state.receive_payload.received = 0; |
28ab034a | 3579 | ret = lttng_dynamic_buffer_set_size(reception_buffer, header.data_size); |
5312a3ed | 3580 | if (ret) { |
5569b118 | 3581 | status = RELAY_CONNECTION_STATUS_ERROR; |
5312a3ed JG |
3582 | goto end; |
3583 | } | |
3584 | ||
3585 | if (header.data_size == 0) { | |
3586 | /* | |
3587 | * Manually invoke the next state as the poll loop | |
3588 | * will not wake-up to allow us to proceed further. | |
3589 | */ | |
5569b118 | 3590 | status = relay_process_control_receive_payload(conn); |
5312a3ed JG |
3591 | } |
3592 | end: | |
5569b118 | 3593 | return status; |
5312a3ed JG |
3594 | } |
3595 | ||
3596 | /* | |
3597 | * Process the commands received on the control socket | |
3598 | */ | |
28ab034a | 3599 | static enum relay_connection_status relay_process_control(struct relay_connection *conn) |
5312a3ed | 3600 | { |
5569b118 | 3601 | enum relay_connection_status status; |
5312a3ed JG |
3602 | |
3603 | switch (conn->protocol.ctrl.state_id) { | |
3604 | case CTRL_CONNECTION_STATE_RECEIVE_HEADER: | |
5569b118 | 3605 | status = relay_process_control_receive_header(conn); |
5312a3ed JG |
3606 | break; |
3607 | case CTRL_CONNECTION_STATE_RECEIVE_PAYLOAD: | |
5569b118 | 3608 | status = relay_process_control_receive_payload(conn); |
5312a3ed JG |
3609 | break; |
3610 | default: | |
3611 | ERR("Unknown control connection protocol state encountered."); | |
3612 | abort(); | |
3613 | } | |
3614 | ||
5569b118 | 3615 | return status; |
5312a3ed JG |
3616 | } |
3617 | ||
28ab034a | 3618 | static enum relay_connection_status relay_process_data_receive_header(struct relay_connection *conn) |
b8aa1682 | 3619 | { |
5312a3ed | 3620 | int ret; |
5569b118 | 3621 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; |
5312a3ed | 3622 | struct data_connection_state_receive_header *state = |
28ab034a | 3623 | &conn->protocol.data.state.receive_header; |
5312a3ed | 3624 | struct lttcomm_relayd_data_hdr header; |
b8aa1682 | 3625 | struct relay_stream *stream; |
5312a3ed | 3626 | |
a0377dfe | 3627 | LTTNG_ASSERT(state->left_to_receive != 0); |
5312a3ed JG |
3628 | |
3629 | ret = conn->sock->ops->recvmsg(conn->sock, | |
28ab034a JG |
3630 | state->header_reception_buffer + state->received, |
3631 | state->left_to_receive, | |
3632 | MSG_DONTWAIT); | |
5312a3ed | 3633 | if (ret < 0) { |
942003e5 MJ |
3634 | DIAGNOSTIC_PUSH |
3635 | DIAGNOSTIC_IGNORE_LOGICAL_OP | |
5569b118 | 3636 | if (errno != EAGAIN && errno != EWOULDBLOCK) { |
28ab034a | 3637 | DIAGNOSTIC_POP |
5569b118 JG |
3638 | PERROR("Unable to receive data header on sock %d", conn->sock->fd); |
3639 | status = RELAY_CONNECTION_STATUS_ERROR; | |
3640 | } | |
5312a3ed JG |
3641 | goto end; |
3642 | } else if (ret == 0) { | |
3643 | /* Orderly shutdown. Not necessary to print an error. */ | |
3644 | DBG("Socket %d performed an orderly shutdown (received EOF)", conn->sock->fd); | |
5569b118 | 3645 | status = RELAY_CONNECTION_STATUS_CLOSED; |
b8aa1682 JD |
3646 | goto end; |
3647 | } | |
3648 | ||
a0377dfe FD |
3649 | LTTNG_ASSERT(ret > 0); |
3650 | LTTNG_ASSERT(ret <= state->left_to_receive); | |
5312a3ed JG |
3651 | |
3652 | state->left_to_receive -= ret; | |
3653 | state->received += ret; | |
3654 | ||
3655 | if (state->left_to_receive > 0) { | |
3656 | /* | |
3657 | * Can't transition to the protocol's next state, wait to | |
3658 | * receive the rest of the header. | |
3659 | */ | |
28ab034a JG |
3660 | DBG3("Partial reception of data connection header (received %" PRIu64 |
3661 | " bytes, %" PRIu64 " bytes left to receive, fd = %i)", | |
3662 | state->received, | |
3663 | state->left_to_receive, | |
3664 | conn->sock->fd); | |
7591bab1 | 3665 | goto end; |
b8aa1682 | 3666 | } |
b8aa1682 | 3667 | |
5312a3ed JG |
3668 | /* Transition to next state: receiving the payload. */ |
3669 | conn->protocol.data.state_id = DATA_CONNECTION_STATE_RECEIVE_PAYLOAD; | |
173af62f | 3670 | |
5312a3ed JG |
3671 | memcpy(&header, state->header_reception_buffer, sizeof(header)); |
3672 | header.circuit_id = be64toh(header.circuit_id); | |
3673 | header.stream_id = be64toh(header.stream_id); | |
3674 | header.data_size = be32toh(header.data_size); | |
3675 | header.net_seq_num = be64toh(header.net_seq_num); | |
3676 | header.padding_size = be32toh(header.padding_size); | |
3677 | memcpy(&conn->protocol.data.state.receive_payload.header, &header, sizeof(header)); | |
3678 | ||
28ab034a | 3679 | conn->protocol.data.state.receive_payload.left_to_receive = header.data_size; |
5312a3ed JG |
3680 | conn->protocol.data.state.receive_payload.received = 0; |
3681 | conn->protocol.data.state.receive_payload.rotate_index = false; | |
3682 | ||
28ab034a JG |
3683 | DBG("Received data connection header on fd %i: circuit_id = %" PRIu64 |
3684 | ", stream_id = %" PRIu64 ", data_size = %" PRIu32 ", net_seq_num = %" PRIu64 | |
3685 | ", padding_size = %" PRIu32, | |
3686 | conn->sock->fd, | |
3687 | header.circuit_id, | |
3688 | header.stream_id, | |
3689 | header.data_size, | |
3690 | header.net_seq_num, | |
3691 | header.padding_size); | |
5312a3ed JG |
3692 | |
3693 | stream = stream_get_by_id(header.stream_id); | |
3694 | if (!stream) { | |
3695 | DBG("relay_process_data_receive_payload: Cannot find stream %" PRIu64, | |
28ab034a | 3696 | header.stream_id); |
5569b118 JG |
3697 | /* Protocol error. */ |
3698 | status = RELAY_CONNECTION_STATUS_ERROR; | |
5312a3ed JG |
3699 | goto end; |
3700 | } | |
b8aa1682 | 3701 | |
7591bab1 | 3702 | pthread_mutex_lock(&stream->lock); |
c35f9726 | 3703 | /* Prepare stream for the reception of a new packet. */ |
28ab034a JG |
3704 | ret = stream_init_packet( |
3705 | stream, header.data_size, &conn->protocol.data.state.receive_payload.rotate_index); | |
c35f9726 JG |
3706 | pthread_mutex_unlock(&stream->lock); |
3707 | if (ret) { | |
3708 | ERR("Failed to rotate stream output file"); | |
3709 | status = RELAY_CONNECTION_STATUS_ERROR; | |
3710 | goto end_stream_unlock; | |
1c20f0e2 JD |
3711 | } |
3712 | ||
5312a3ed | 3713 | end_stream_unlock: |
5312a3ed JG |
3714 | stream_put(stream); |
3715 | end: | |
5569b118 | 3716 | return status; |
5312a3ed JG |
3717 | } |
3718 | ||
28ab034a JG |
3719 | static enum relay_connection_status |
3720 | relay_process_data_receive_payload(struct relay_connection *conn) | |
5312a3ed JG |
3721 | { |
3722 | int ret; | |
5569b118 | 3723 | enum relay_connection_status status = RELAY_CONNECTION_STATUS_OK; |
5312a3ed JG |
3724 | struct relay_stream *stream; |
3725 | struct data_connection_state_receive_payload *state = | |
28ab034a | 3726 | &conn->protocol.data.state.receive_payload; |
5312a3ed JG |
3727 | const size_t chunk_size = RECV_DATA_BUFFER_SIZE; |
3728 | char data_buffer[chunk_size]; | |
3729 | bool partial_recv = false; | |
3730 | bool new_stream = false, close_requested = false, index_flushed = false; | |
3731 | uint64_t left_to_receive = state->left_to_receive; | |
3732 | struct relay_session *session; | |
3733 | ||
28ab034a JG |
3734 | DBG3("Receiving data for stream id %" PRIu64 " seqnum %" PRIu64 ", %" PRIu64 |
3735 | " bytes received, %" PRIu64 " bytes left to receive", | |
3736 | state->header.stream_id, | |
3737 | state->header.net_seq_num, | |
3738 | state->received, | |
3739 | left_to_receive); | |
fd0f1e3e | 3740 | |
5312a3ed JG |
3741 | stream = stream_get_by_id(state->header.stream_id); |
3742 | if (!stream) { | |
5569b118 | 3743 | /* Protocol error. */ |
fd0f1e3e | 3744 | ERR("relay_process_data_receive_payload: cannot find stream %" PRIu64, |
28ab034a | 3745 | state->header.stream_id); |
5569b118 | 3746 | status = RELAY_CONNECTION_STATUS_ERROR; |
5312a3ed | 3747 | goto end; |
1c20f0e2 JD |
3748 | } |
3749 | ||
5312a3ed JG |
3750 | pthread_mutex_lock(&stream->lock); |
3751 | session = stream->trace->session; | |
fd0f1e3e JR |
3752 | if (!conn->session) { |
3753 | ret = connection_set_session(conn, session); | |
3754 | if (ret) { | |
3755 | status = RELAY_CONNECTION_STATUS_ERROR; | |
3756 | goto end_stream_unlock; | |
3757 | } | |
3758 | } | |
5312a3ed JG |
3759 | |
3760 | /* | |
3761 | * The size of the "chunk" received on any iteration is bounded by: | |
3762 | * - the data left to receive, | |
3763 | * - the data immediately available on the socket, | |
3764 | * - the on-stack data buffer | |
3765 | */ | |
3766 | while (left_to_receive > 0 && !partial_recv) { | |
ff2aa8f0 | 3767 | size_t recv_size = std::min<uint64_t>(left_to_receive, chunk_size); |
c35f9726 | 3768 | struct lttng_buffer_view packet_chunk; |
5312a3ed | 3769 | |
28ab034a | 3770 | ret = conn->sock->ops->recvmsg(conn->sock, data_buffer, recv_size, MSG_DONTWAIT); |
5312a3ed | 3771 | if (ret < 0) { |
942003e5 MJ |
3772 | DIAGNOSTIC_PUSH |
3773 | DIAGNOSTIC_IGNORE_LOGICAL_OP | |
5569b118 | 3774 | if (errno != EAGAIN && errno != EWOULDBLOCK) { |
28ab034a | 3775 | DIAGNOSTIC_POP |
5569b118 JG |
3776 | PERROR("Socket %d error", conn->sock->fd); |
3777 | status = RELAY_CONNECTION_STATUS_ERROR; | |
3778 | } | |
0848dba7 | 3779 | goto end_stream_unlock; |
5312a3ed JG |
3780 | } else if (ret == 0) { |
3781 | /* No more data ready to be consumed on socket. */ | |
3782 | DBG3("No more data ready for consumption on data socket of stream id %" PRIu64, | |
28ab034a | 3783 | state->header.stream_id); |
5569b118 | 3784 | status = RELAY_CONNECTION_STATUS_CLOSED; |
5312a3ed JG |
3785 | break; |
3786 | } else if (ret < (int) recv_size) { | |
3787 | /* | |
3788 | * All the data available on the socket has been | |
3789 | * consumed. | |
3790 | */ | |
3791 | partial_recv = true; | |
c35f9726 | 3792 | recv_size = ret; |
0848dba7 MD |
3793 | } |
3794 | ||
28ab034a | 3795 | packet_chunk = lttng_buffer_view_init(data_buffer, 0, recv_size); |
a0377dfe | 3796 | LTTNG_ASSERT(packet_chunk.data); |
5312a3ed | 3797 | |
c35f9726 JG |
3798 | ret = stream_write(stream, &packet_chunk, 0); |
3799 | if (ret) { | |
0848dba7 | 3800 | ERR("Relay error writing data to file"); |
5569b118 | 3801 | status = RELAY_CONNECTION_STATUS_ERROR; |
0848dba7 MD |
3802 | goto end_stream_unlock; |
3803 | } | |
3804 | ||
5312a3ed JG |
3805 | left_to_receive -= recv_size; |
3806 | state->received += recv_size; | |
3807 | state->left_to_receive = left_to_receive; | |
5312a3ed JG |
3808 | } |
3809 | ||
3810 | if (state->left_to_receive > 0) { | |
3811 | /* | |
3812 | * Did not receive all the data expected, wait for more data to | |
3813 | * become available on the socket. | |
3814 | */ | |
28ab034a JG |
3815 | DBG3("Partial receive on data connection of stream id %" PRIu64 ", %" PRIu64 |
3816 | " bytes received, %" PRIu64 " bytes left to receive", | |
3817 | state->header.stream_id, | |
3818 | state->received, | |
3819 | state->left_to_receive); | |
5312a3ed | 3820 | goto end_stream_unlock; |
0848dba7 | 3821 | } |
5ab7344e | 3822 | |
cd9adb8b | 3823 | ret = stream_write(stream, nullptr, state->header.padding_size); |
c35f9726 | 3824 | if (ret) { |
5569b118 | 3825 | status = RELAY_CONNECTION_STATUS_ERROR; |
7591bab1 | 3826 | goto end_stream_unlock; |
1d4dfdef | 3827 | } |
5312a3ed | 3828 | |
298a25ca | 3829 | if (session_streams_have_index(session)) { |
28ab034a JG |
3830 | ret = stream_update_index(stream, |
3831 | state->header.net_seq_num, | |
3832 | state->rotate_index, | |
3833 | &index_flushed, | |
3834 | state->header.data_size + state->header.padding_size); | |
5312a3ed | 3835 | if (ret < 0) { |
28ab034a JG |
3836 | ERR("Failed to update index: stream %" PRIu64 " net_seq_num %" PRIu64 |
3837 | " ret %d", | |
3838 | stream->stream_handle, | |
3839 | state->header.net_seq_num, | |
3840 | ret); | |
5569b118 | 3841 | status = RELAY_CONNECTION_STATUS_ERROR; |
5312a3ed JG |
3842 | goto end_stream_unlock; |
3843 | } | |
3844 | } | |
3845 | ||
a8f9f353 | 3846 | if (stream->prev_data_seq == -1ULL) { |
c0bae11d MD |
3847 | new_stream = true; |
3848 | } | |
3849 | ||
28ab034a JG |
3850 | ret = stream_complete_packet(stream, |
3851 | state->header.data_size + state->header.padding_size, | |
3852 | state->header.net_seq_num, | |
3853 | index_flushed); | |
c35f9726 JG |
3854 | if (ret) { |
3855 | status = RELAY_CONNECTION_STATUS_ERROR; | |
3856 | goto end_stream_unlock; | |
3857 | } | |
5312a3ed JG |
3858 | |
3859 | /* | |
3860 | * Resetting the protocol state (to RECEIVE_HEADER) will trash the | |
3861 | * contents of *state which are aliased (union) to the same location as | |
3862 | * the new state. Don't use it beyond this point. | |
3863 | */ | |
3864 | connection_reset_protocol_state(conn); | |
cd9adb8b | 3865 | state = nullptr; |
173af62f | 3866 | |
7591bab1 | 3867 | end_stream_unlock: |
bda7c7b9 | 3868 | close_requested = stream->close_requested; |
7591bab1 | 3869 | pthread_mutex_unlock(&stream->lock); |
5312a3ed | 3870 | if (close_requested && left_to_receive == 0) { |
bda7c7b9 JG |
3871 | try_stream_close(stream); |
3872 | } | |
3873 | ||
c0bae11d MD |
3874 | if (new_stream) { |
3875 | pthread_mutex_lock(&session->lock); | |
3876 | uatomic_set(&session->new_streams, 1); | |
3877 | pthread_mutex_unlock(&session->lock); | |
3878 | } | |
5312a3ed | 3879 | |
7591bab1 | 3880 | stream_put(stream); |
b8aa1682 | 3881 | end: |
5569b118 | 3882 | return status; |
b8aa1682 JD |
3883 | } |
3884 | ||
5312a3ed JG |
3885 | /* |
3886 | * relay_process_data: Process the data received on the data socket | |
3887 | */ | |
28ab034a | 3888 | static enum relay_connection_status relay_process_data(struct relay_connection *conn) |
5312a3ed | 3889 | { |
5569b118 | 3890 | enum relay_connection_status status; |
5312a3ed JG |
3891 | |
3892 | switch (conn->protocol.data.state_id) { | |
3893 | case DATA_CONNECTION_STATE_RECEIVE_HEADER: | |
5569b118 | 3894 | status = relay_process_data_receive_header(conn); |
5312a3ed JG |
3895 | break; |
3896 | case DATA_CONNECTION_STATE_RECEIVE_PAYLOAD: | |
5569b118 | 3897 | status = relay_process_data_receive_payload(conn); |
5312a3ed JG |
3898 | break; |
3899 | default: | |
3900 | ERR("Unexpected data connection communication state."); | |
3901 | abort(); | |
3902 | } | |
3903 | ||
5569b118 | 3904 | return status; |
5312a3ed JG |
3905 | } |
3906 | ||
7591bab1 | 3907 | static void cleanup_connection_pollfd(struct lttng_poll_event *events, int pollfd) |
b8aa1682 JD |
3908 | { |
3909 | int ret; | |
3910 | ||
58eb9381 | 3911 | (void) lttng_poll_del(events, pollfd); |
b8aa1682 | 3912 | |
28ab034a | 3913 | ret = fd_tracker_close_unsuspendable_fd( |
cd9adb8b | 3914 | the_fd_tracker, &pollfd, 1, fd_tracker_util_close_fd, nullptr); |
b8aa1682 JD |
3915 | if (ret < 0) { |
3916 | ERR("Closing pollfd %d", pollfd); | |
3917 | } | |
3918 | } | |
3919 | ||
7591bab1 | 3920 | static void relay_thread_close_connection(struct lttng_poll_event *events, |
28ab034a JG |
3921 | int pollfd, |
3922 | struct relay_connection *conn) | |
9d1bbf21 | 3923 | { |
7591bab1 | 3924 | const char *type_str; |
2a174661 | 3925 | |
7591bab1 MD |
3926 | switch (conn->type) { |
3927 | case RELAY_DATA: | |
3928 | type_str = "Data"; | |
3929 | break; | |
3930 | case RELAY_CONTROL: | |
3931 | type_str = "Control"; | |
3932 | break; | |
3933 | case RELAY_VIEWER_COMMAND: | |
3934 | type_str = "Viewer Command"; | |
3935 | break; | |
3936 | case RELAY_VIEWER_NOTIFICATION: | |
3937 | type_str = "Viewer Notification"; | |
3938 | break; | |
3939 | default: | |
3940 | type_str = "Unknown"; | |
9d1bbf21 | 3941 | } |
7591bab1 MD |
3942 | cleanup_connection_pollfd(events, pollfd); |
3943 | connection_put(conn); | |
3944 | DBG("%s connection closed with %d", type_str, pollfd); | |
b8aa1682 JD |
3945 | } |
3946 | ||
3947 | /* | |
3948 | * This thread does the actual work | |
3949 | */ | |
f46376a1 | 3950 | static void *relay_thread_worker(void *data __attribute__((unused))) |
b8aa1682 | 3951 | { |
beaad64c DG |
3952 | int ret, err = -1, last_seen_data_fd = -1; |
3953 | uint32_t nb_fd; | |
b8aa1682 JD |
3954 | struct lttng_poll_event events; |
3955 | struct lttng_ht *relay_connections_ht; | |
b8aa1682 JD |
3956 | |
3957 | DBG("[thread] Relay worker started"); | |
3958 | ||
9d1bbf21 MD |
3959 | rcu_register_thread(); |
3960 | ||
55706a7d MD |
3961 | health_register(health_relayd, HEALTH_RELAYD_TYPE_WORKER); |
3962 | ||
9b5e0863 MD |
3963 | if (testpoint(relayd_thread_worker)) { |
3964 | goto error_testpoint; | |
3965 | } | |
3966 | ||
f385ae0a MD |
3967 | health_code_update(); |
3968 | ||
b8aa1682 JD |
3969 | /* table of connections indexed on socket */ |
3970 | relay_connections_ht = lttng_ht_new(0, LTTNG_HT_TYPE_ULONG); | |
095a4ae5 MD |
3971 | if (!relay_connections_ht) { |
3972 | goto relay_connections_ht_error; | |
3973 | } | |
b8aa1682 | 3974 | |
e32a0864 | 3975 | ret = create_named_thread_poll_set(&events, 2, "Worker thread epoll"); |
b8aa1682 JD |
3976 | if (ret < 0) { |
3977 | goto error_poll_create; | |
3978 | } | |
3979 | ||
58eb9381 | 3980 | ret = lttng_poll_add(&events, relay_conn_pipe[0], LPOLLIN | LPOLLRDHUP); |
b8aa1682 JD |
3981 | if (ret < 0) { |
3982 | goto error; | |
3983 | } | |
3984 | ||
beaad64c | 3985 | restart: |
cd9adb8b | 3986 | while (true) { |
beaad64c DG |
3987 | int idx = -1, i, seen_control = 0, last_notdel_data_fd = -1; |
3988 | ||
f385ae0a MD |
3989 | health_code_update(); |
3990 | ||
b8aa1682 | 3991 | /* Infinite blocking call, waiting for transmission */ |
87c1611d | 3992 | DBG3("Relayd worker thread polling..."); |
f385ae0a | 3993 | health_poll_entry(); |
b8aa1682 | 3994 | ret = lttng_poll_wait(&events, -1); |
f385ae0a | 3995 | health_poll_exit(); |
b8aa1682 JD |
3996 | if (ret < 0) { |
3997 | /* | |
3998 | * Restart interrupted system call. | |
3999 | */ | |
4000 | if (errno == EINTR) { | |
4001 | goto restart; | |
4002 | } | |
4003 | goto error; | |
4004 | } | |
4005 | ||
0d9c5d77 DG |
4006 | nb_fd = ret; |
4007 | ||
beaad64c | 4008 | /* |
7591bab1 MD |
4009 | * Process control. The control connection is |
4010 | * prioritized so we don't starve it with high | |
4011 | * throughput tracing data on the data connection. | |
beaad64c | 4012 | */ |
b8aa1682 JD |
4013 | for (i = 0; i < nb_fd; i++) { |
4014 | /* Fetch once the poll data */ | |
8a00688e MJ |
4015 | const auto revents = LTTNG_POLL_GETEV(&events, i); |
4016 | const auto pollfd = LTTNG_POLL_GETFD(&events, i); | |
b8aa1682 | 4017 | |
f385ae0a MD |
4018 | health_code_update(); |
4019 | ||
8a00688e MJ |
4020 | /* Activity on thread quit pipe, exiting. */ |
4021 | if (relayd_is_thread_quit_pipe(pollfd)) { | |
4022 | DBG("Activity on thread quit pipe"); | |
095a4ae5 MD |
4023 | err = 0; |
4024 | goto exit; | |
b8aa1682 JD |
4025 | } |
4026 | ||
58eb9381 DG |
4027 | /* Inspect the relay conn pipe for new connection */ |
4028 | if (pollfd == relay_conn_pipe[0]) { | |
03e43155 | 4029 | if (revents & LPOLLIN) { |
90e7d72f JG |
4030 | struct relay_connection *conn; |
4031 | ||
5c7248cd JG |
4032 | ret = lttng_read(relay_conn_pipe[0], |
4033 | &conn, | |
4034 | sizeof(conn)); /* NOLINT sizeof used on a | |
4035 | pointer. */ | |
b8aa1682 JD |
4036 | if (ret < 0) { |
4037 | goto error; | |
4038 | } | |
28ab034a JG |
4039 | ret = lttng_poll_add( |
4040 | &events, conn->sock->fd, LPOLLIN | LPOLLRDHUP); | |
73039936 FD |
4041 | if (ret) { |
4042 | ERR("Failed to add new connection file descriptor to poll set"); | |
4043 | goto error; | |
4044 | } | |
7591bab1 | 4045 | connection_ht_add(relay_connections_ht, conn); |
58eb9381 | 4046 | DBG("Connection socket %d added", conn->sock->fd); |
03e43155 MD |
4047 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
4048 | ERR("Relay connection pipe error"); | |
4049 | goto error; | |
4050 | } else { | |
28ab034a JG |
4051 | ERR("Unexpected poll events %u for sock %d", |
4052 | revents, | |
4053 | pollfd); | |
03e43155 | 4054 | goto error; |
b8aa1682 | 4055 | } |
58eb9381 | 4056 | } else { |
90e7d72f JG |
4057 | struct relay_connection *ctrl_conn; |
4058 | ||
7591bab1 | 4059 | ctrl_conn = connection_get_by_sock(relay_connections_ht, pollfd); |
58eb9381 | 4060 | /* If not found, there is a synchronization issue. */ |
a0377dfe | 4061 | LTTNG_ASSERT(ctrl_conn); |
58eb9381 | 4062 | |
03e43155 MD |
4063 | if (ctrl_conn->type == RELAY_DATA) { |
4064 | if (revents & LPOLLIN) { | |
beaad64c | 4065 | /* |
28ab034a JG |
4066 | * Flag the last seen data fd not deleted. It will |
4067 | * be used as the last seen fd if any fd gets | |
4068 | * deleted in this first loop. | |
beaad64c DG |
4069 | */ |
4070 | last_notdel_data_fd = pollfd; | |
4071 | } | |
03e43155 MD |
4072 | goto put_ctrl_connection; |
4073 | } | |
a0377dfe | 4074 | LTTNG_ASSERT(ctrl_conn->type == RELAY_CONTROL); |
03e43155 MD |
4075 | |
4076 | if (revents & LPOLLIN) { | |
5569b118 JG |
4077 | enum relay_connection_status status; |
4078 | ||
4079 | status = relay_process_control(ctrl_conn); | |
4080 | if (status != RELAY_CONNECTION_STATUS_OK) { | |
fd0f1e3e | 4081 | /* |
28ab034a JG |
4082 | * On socket error flag the session as aborted to |
4083 | * force the cleanup of its stream otherwise it can | |
4084 | * leak during the lifetime of the relayd. | |
fd0f1e3e JR |
4085 | * |
4086 | * This prevents situations in which streams can be | |
4087 | * left opened because an index was received, the | |
4088 | * control connection is closed, and the data | |
28ab034a JG |
4089 | * connection is closed (uncleanly) before the |
4090 | * packet's data provided. | |
fd0f1e3e | 4091 | * |
28ab034a JG |
4092 | * Since the control connection encountered an |
4093 | * error, it is okay to be conservative and close | |
4094 | * the session right now as we can't rely on the | |
4095 | * protocol being respected anymore. | |
fd0f1e3e JR |
4096 | */ |
4097 | if (status == RELAY_CONNECTION_STATUS_ERROR) { | |
4098 | session_abort(ctrl_conn->session); | |
4099 | } | |
4100 | ||
5569b118 | 4101 | /* Clear the connection on error or close. */ |
28ab034a JG |
4102 | relay_thread_close_connection( |
4103 | &events, pollfd, ctrl_conn); | |
03e43155 | 4104 | } |
5312a3ed | 4105 | seen_control = 1; |
03e43155 | 4106 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
28ab034a | 4107 | relay_thread_close_connection(&events, pollfd, ctrl_conn); |
03e43155 MD |
4108 | if (last_seen_data_fd == pollfd) { |
4109 | last_seen_data_fd = last_notdel_data_fd; | |
4110 | } | |
58eb9381 | 4111 | } else { |
03e43155 | 4112 | ERR("Unexpected poll events %u for control sock %d", |
28ab034a JG |
4113 | revents, |
4114 | pollfd); | |
03e43155 MD |
4115 | connection_put(ctrl_conn); |
4116 | goto error; | |
beaad64c | 4117 | } |
03e43155 | 4118 | put_ctrl_connection: |
7591bab1 | 4119 | connection_put(ctrl_conn); |
beaad64c DG |
4120 | } |
4121 | } | |
4122 | ||
4123 | /* | |
4124 | * The last loop handled a control request, go back to poll to make | |
4125 | * sure we prioritise the control socket. | |
4126 | */ | |
4127 | if (seen_control) { | |
4128 | continue; | |
4129 | } | |
4130 | ||
4131 | if (last_seen_data_fd >= 0) { | |
4132 | for (i = 0; i < nb_fd; i++) { | |
07c4863f | 4133 | const int pollfd = LTTNG_POLL_GETFD(&events, i); |
f385ae0a MD |
4134 | |
4135 | health_code_update(); | |
4136 | ||
beaad64c DG |
4137 | if (last_seen_data_fd == pollfd) { |
4138 | idx = i; | |
4139 | break; | |
4140 | } | |
4141 | } | |
4142 | } | |
4143 | ||
4144 | /* Process data connection. */ | |
4145 | for (i = idx + 1; i < nb_fd; i++) { | |
4146 | /* Fetch the poll data. */ | |
07c4863f JG |
4147 | const uint32_t revents = LTTNG_POLL_GETEV(&events, i); |
4148 | const int pollfd = LTTNG_POLL_GETFD(&events, i); | |
90e7d72f | 4149 | struct relay_connection *data_conn; |
beaad64c | 4150 | |
f385ae0a MD |
4151 | health_code_update(); |
4152 | ||
fd20dac9 MD |
4153 | if (!revents) { |
4154 | /* No activity for this FD (poll implementation). */ | |
4155 | continue; | |
4156 | } | |
4157 | ||
beaad64c | 4158 | /* Skip the command pipe. It's handled in the first loop. */ |
58eb9381 | 4159 | if (pollfd == relay_conn_pipe[0]) { |
beaad64c DG |
4160 | continue; |
4161 | } | |
4162 | ||
7591bab1 | 4163 | data_conn = connection_get_by_sock(relay_connections_ht, pollfd); |
90e7d72f | 4164 | if (!data_conn) { |
fd20dac9 | 4165 | /* Skip it. Might be removed before. */ |
fd20dac9 MD |
4166 | continue; |
4167 | } | |
03e43155 MD |
4168 | if (data_conn->type == RELAY_CONTROL) { |
4169 | goto put_data_connection; | |
4170 | } | |
a0377dfe | 4171 | LTTNG_ASSERT(data_conn->type == RELAY_DATA); |
fd20dac9 MD |
4172 | |
4173 | if (revents & LPOLLIN) { | |
5569b118 JG |
4174 | enum relay_connection_status status; |
4175 | ||
4176 | status = relay_process_data(data_conn); | |
4177 | /* Connection closed or error. */ | |
4178 | if (status != RELAY_CONNECTION_STATUS_OK) { | |
fd0f1e3e JR |
4179 | /* |
4180 | * On socket error flag the session as aborted to force | |
4181 | * the cleanup of its stream otherwise it can leak | |
4182 | * during the lifetime of the relayd. | |
4183 | * | |
4184 | * This prevents situations in which streams can be | |
4185 | * left opened because an index was received, the | |
4186 | * control connection is closed, and the data | |
4187 | * connection is closed (uncleanly) before the packet's | |
4188 | * data provided. | |
4189 | * | |
4190 | * Since the data connection encountered an error, | |
4191 | * it is okay to be conservative and close the | |
4192 | * session right now as we can't rely on the protocol | |
4193 | * being respected anymore. | |
4194 | */ | |
4195 | if (status == RELAY_CONNECTION_STATUS_ERROR) { | |
4196 | session_abort(data_conn->session); | |
4197 | } | |
28ab034a | 4198 | relay_thread_close_connection(&events, pollfd, data_conn); |
fd20dac9 MD |
4199 | /* |
4200 | * Every goto restart call sets the last seen fd where | |
4201 | * here we don't really care since we gracefully | |
4202 | * continue the loop after the connection is deleted. | |
4203 | */ | |
4204 | } else { | |
4205 | /* Keep last seen port. */ | |
4206 | last_seen_data_fd = pollfd; | |
7591bab1 | 4207 | connection_put(data_conn); |
fd20dac9 | 4208 | goto restart; |
b8aa1682 | 4209 | } |
03e43155 | 4210 | } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { |
28ab034a | 4211 | relay_thread_close_connection(&events, pollfd, data_conn); |
03e43155 | 4212 | } else { |
28ab034a | 4213 | ERR("Unknown poll events %u for data sock %d", revents, pollfd); |
b8aa1682 | 4214 | } |
03e43155 | 4215 | put_data_connection: |
7591bab1 | 4216 | connection_put(data_conn); |
b8aa1682 | 4217 | } |
beaad64c | 4218 | last_seen_data_fd = -1; |
b8aa1682 JD |
4219 | } |
4220 | ||
f385ae0a MD |
4221 | /* Normal exit, no error */ |
4222 | ret = 0; | |
4223 | ||
095a4ae5 | 4224 | exit: |
b8aa1682 | 4225 | error: |
71efa8ef | 4226 | /* Cleanup remaining connection object. */ |
85c34f4c JG |
4227 | for (auto *destroy_conn : |
4228 | lttng::urcu::lfht_iteration_adapter<relay_connection, | |
4229 | decltype(relay_connection::sock_n), | |
4230 | &relay_connection::sock_n>( | |
4231 | *relay_connections_ht->ht)) { | |
4232 | health_code_update(); | |
98ba050e | 4233 | |
85c34f4c | 4234 | session_abort(destroy_conn->session); |
56047f5a | 4235 | |
85c34f4c JG |
4236 | /* |
4237 | * No need to grab another ref, because we own | |
4238 | * destroy_conn. | |
4239 | */ | |
4240 | relay_thread_close_connection(&events, destroy_conn->sock->fd, destroy_conn); | |
b8aa1682 | 4241 | } |
7591bab1 | 4242 | |
e32a0864 | 4243 | (void) fd_tracker_util_poll_clean(the_fd_tracker, &events); |
7d2f7452 | 4244 | error_poll_create: |
b8aa1682 | 4245 | lttng_ht_destroy(relay_connections_ht); |
095a4ae5 | 4246 | relay_connections_ht_error: |
58eb9381 | 4247 | /* Close relay conn pipes */ |
28ab034a | 4248 | (void) fd_tracker_util_pipe_close(the_fd_tracker, relay_conn_pipe); |
095a4ae5 MD |
4249 | if (err) { |
4250 | DBG("Thread exited with error"); | |
4251 | } | |
b8aa1682 | 4252 | DBG("Worker thread cleanup complete"); |
9b5e0863 | 4253 | error_testpoint: |
f385ae0a MD |
4254 | if (err) { |
4255 | health_error(); | |
4256 | ERR("Health error occurred in %s", __func__); | |
4257 | } | |
4258 | health_unregister(health_relayd); | |
9d1bbf21 | 4259 | rcu_unregister_thread(); |
b4aacfdc | 4260 | lttng_relay_stop_threads(); |
cd9adb8b | 4261 | return nullptr; |
b8aa1682 JD |
4262 | } |
4263 | ||
4264 | /* | |
4265 | * Create the relay command pipe to wake thread_manage_apps. | |
4266 | * Closed in cleanup(). | |
4267 | */ | |
cd9adb8b | 4268 | static int create_relay_conn_pipe() |
b8aa1682 | 4269 | { |
28ab034a JG |
4270 | return fd_tracker_util_pipe_open_cloexec( |
4271 | the_fd_tracker, "Relayd connection pipe", relay_conn_pipe); | |
b8aa1682 JD |
4272 | } |
4273 | ||
f46376a1 | 4274 | static int stdio_open(void *data __attribute__((unused)), int *fds) |
9c256b01 JG |
4275 | { |
4276 | fds[0] = fileno(stdout); | |
4277 | fds[1] = fileno(stderr); | |
4278 | return 0; | |
4279 | } | |
4280 | ||
cd9adb8b | 4281 | static int track_stdio() |
9c256b01 JG |
4282 | { |
4283 | int fds[2]; | |
4284 | const char *names[] = { "stdout", "stderr" }; | |
4285 | ||
cd9adb8b | 4286 | return fd_tracker_open_unsuspendable_fd(the_fd_tracker, fds, names, 2, stdio_open, nullptr); |
9c256b01 JG |
4287 | } |
4288 | ||
b8aa1682 JD |
4289 | /* |
4290 | * main | |
4291 | */ | |
4292 | int main(int argc, char **argv) | |
4293 | { | |
00e3b7f1 | 4294 | bool thread_is_rcu_registered = false; |
178a0557 | 4295 | int ret = 0, retval = 0; |
b8aa1682 | 4296 | void *status; |
cd9adb8b | 4297 | char *unlinked_file_directory_path = nullptr, *output_path = nullptr; |
b8aa1682 | 4298 | |
2a10de3b JR |
4299 | /* Parse environment variables */ |
4300 | ret = parse_env_options(); | |
4301 | if (ret) { | |
4302 | retval = -1; | |
4303 | goto exit_options; | |
4304 | } | |
4305 | ||
4306 | /* | |
4307 | * Parse arguments. | |
4308 | * Command line arguments overwrite environment. | |
4309 | */ | |
b8aa1682 | 4310 | progname = argv[0]; |
178a0557 MD |
4311 | if (set_options(argc, argv)) { |
4312 | retval = -1; | |
4313 | goto exit_options; | |
b8aa1682 JD |
4314 | } |
4315 | ||
178a0557 MD |
4316 | if (set_signal_handler()) { |
4317 | retval = -1; | |
4318 | goto exit_options; | |
b8aa1682 JD |
4319 | } |
4320 | ||
a3bc3918 JR |
4321 | relayd_config_log(); |
4322 | ||
4323 | if (opt_print_version) { | |
4324 | print_version(); | |
4325 | retval = 0; | |
4326 | goto exit_options; | |
4327 | } | |
4328 | ||
c0407718 JG |
4329 | ret = fclose(stdin); |
4330 | if (ret) { | |
4331 | PERROR("Failed to close stdin"); | |
4332 | goto exit_options; | |
4333 | } | |
4334 | ||
35ab25e5 MD |
4335 | DBG("Clear command %s", opt_allow_clear ? "allowed" : "disallowed"); |
4336 | ||
4d513a50 DG |
4337 | /* Try to create directory if -o, --output is specified. */ |
4338 | if (opt_output_path) { | |
994fa64f DG |
4339 | if (*opt_output_path != '/') { |
4340 | ERR("Please specify an absolute path for -o, --output PATH"); | |
178a0557 MD |
4341 | retval = -1; |
4342 | goto exit_options; | |
994fa64f DG |
4343 | } |
4344 | ||
28ab034a | 4345 | ret = utils_mkdir_recursive(opt_output_path, S_IRWXU | S_IRWXG, -1, -1); |
4d513a50 DG |
4346 | if (ret < 0) { |
4347 | ERR("Unable to create %s", opt_output_path); | |
178a0557 MD |
4348 | retval = -1; |
4349 | goto exit_options; | |
4d513a50 DG |
4350 | } |
4351 | } | |
4352 | ||
b8aa1682 | 4353 | /* Daemonize */ |
b5218ffb | 4354 | if (opt_daemon || opt_background) { |
28ab034a | 4355 | ret = lttng_daemonize(&child_ppid, &recv_child_signal, !opt_background); |
b8aa1682 | 4356 | if (ret < 0) { |
178a0557 MD |
4357 | retval = -1; |
4358 | goto exit_options; | |
b8aa1682 | 4359 | } |
3fd27398 MD |
4360 | } |
4361 | ||
ce9ee1fb JR |
4362 | if (opt_working_directory) { |
4363 | ret = utils_change_working_directory(opt_working_directory); | |
4364 | if (ret) { | |
4365 | /* All errors are already logged. */ | |
4366 | goto exit_options; | |
4367 | } | |
4368 | } | |
4369 | ||
23c8ff50 JG |
4370 | sessiond_trace_chunk_registry = sessiond_trace_chunk_registry_create(); |
4371 | if (!sessiond_trace_chunk_registry) { | |
4372 | ERR("Failed to initialize session daemon trace chunk registry"); | |
4373 | retval = -1; | |
794e2e5f | 4374 | goto exit_options; |
23c8ff50 JG |
4375 | } |
4376 | ||
00e3b7f1 JG |
4377 | /* |
4378 | * The RCU thread registration (and use, through the fd-tracker's | |
4379 | * creation) is done after the daemonization to allow us to not | |
4380 | * deal with liburcu's fork() management as the call RCU needs to | |
4381 | * be restored. | |
4382 | */ | |
4383 | rcu_register_thread(); | |
4384 | thread_is_rcu_registered = true; | |
4385 | ||
f7c3ffd7 JG |
4386 | output_path = create_output_path(""); |
4387 | if (!output_path) { | |
4388 | ERR("Failed to get output path"); | |
4389 | retval = -1; | |
4390 | goto exit_options; | |
4391 | } | |
28ab034a JG |
4392 | ret = asprintf(&unlinked_file_directory_path, |
4393 | "%s/%s", | |
4394 | output_path, | |
4395 | DEFAULT_UNLINKED_FILES_DIRECTORY); | |
f7c3ffd7 JG |
4396 | free(output_path); |
4397 | if (ret < 0) { | |
4398 | ERR("Failed to format unlinked file directory path"); | |
4399 | retval = -1; | |
4400 | goto exit_options; | |
4401 | } | |
28ab034a | 4402 | the_fd_tracker = fd_tracker_create(unlinked_file_directory_path, lttng_opt_fd_pool_size); |
f7c3ffd7 | 4403 | free(unlinked_file_directory_path); |
00e3b7f1 JG |
4404 | if (!the_fd_tracker) { |
4405 | retval = -1; | |
4406 | goto exit_options; | |
4407 | } | |
4408 | ||
9c256b01 JG |
4409 | ret = track_stdio(); |
4410 | if (ret) { | |
4411 | retval = -1; | |
4412 | goto exit_options; | |
4413 | } | |
4414 | ||
178a0557 MD |
4415 | /* Initialize thread health monitoring */ |
4416 | health_relayd = health_app_create(NR_HEALTH_RELAYD_TYPES); | |
4417 | if (!health_relayd) { | |
4418 | PERROR("health_app_create error"); | |
4419 | retval = -1; | |
794e2e5f | 4420 | goto exit_options; |
178a0557 MD |
4421 | } |
4422 | ||
3fd27398 | 4423 | /* Create thread quit pipe */ |
8a00688e | 4424 | if (relayd_init_thread_quit_pipe()) { |
178a0557 | 4425 | retval = -1; |
794e2e5f | 4426 | goto exit_options; |
b8aa1682 JD |
4427 | } |
4428 | ||
b8aa1682 | 4429 | /* Setup the thread apps communication pipe. */ |
178a0557 MD |
4430 | if (create_relay_conn_pipe()) { |
4431 | retval = -1; | |
794e2e5f | 4432 | goto exit_options; |
b8aa1682 JD |
4433 | } |
4434 | ||
4435 | /* Init relay command queue. */ | |
8bdee6e2 | 4436 | cds_wfcq_init(&relay_conn_queue.head, &relay_conn_queue.tail); |
b8aa1682 | 4437 | |
554831e7 MD |
4438 | /* Initialize communication library */ |
4439 | lttcomm_init(); | |
87e45c13 | 4440 | lttcomm_inet_init(); |
554831e7 | 4441 | |
d3e2ba59 | 4442 | /* tables of sessions indexed by session ID */ |
7591bab1 MD |
4443 | sessions_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
4444 | if (!sessions_ht) { | |
178a0557 | 4445 | retval = -1; |
794e2e5f | 4446 | goto exit_options; |
d3e2ba59 JD |
4447 | } |
4448 | ||
4449 | /* tables of streams indexed by stream ID */ | |
2a174661 | 4450 | relay_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
d3e2ba59 | 4451 | if (!relay_streams_ht) { |
178a0557 | 4452 | retval = -1; |
794e2e5f | 4453 | goto exit_options; |
d3e2ba59 JD |
4454 | } |
4455 | ||
98b82dfa KS |
4456 | /* tables of viewer sessions indexed by session ID */ |
4457 | viewer_sessions_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); | |
4458 | if (!viewer_sessions_ht) { | |
4459 | retval = -1; | |
4460 | goto exit_options; | |
4461 | } | |
4462 | ||
d3e2ba59 | 4463 | /* tables of streams indexed by stream ID */ |
92c6ca54 DG |
4464 | viewer_streams_ht = lttng_ht_new(0, LTTNG_HT_TYPE_U64); |
4465 | if (!viewer_streams_ht) { | |
178a0557 | 4466 | retval = -1; |
794e2e5f | 4467 | goto exit_options; |
55706a7d MD |
4468 | } |
4469 | ||
bcee2b96 | 4470 | ret = init_health_quit_pipe(); |
178a0557 MD |
4471 | if (ret) { |
4472 | retval = -1; | |
794e2e5f | 4473 | goto exit_options; |
65931c8b MD |
4474 | } |
4475 | ||
4476 | /* Create thread to manage the client socket */ | |
cd9adb8b JG |
4477 | ret = pthread_create(&health_thread, |
4478 | default_pthread_attr(), | |
4479 | thread_manage_health_relayd, | |
4480 | (void *) nullptr); | |
178a0557 MD |
4481 | if (ret) { |
4482 | errno = ret; | |
65931c8b | 4483 | PERROR("pthread_create health"); |
178a0557 | 4484 | retval = -1; |
794e2e5f | 4485 | goto exit_options; |
65931c8b MD |
4486 | } |
4487 | ||
b8aa1682 | 4488 | /* Setup the dispatcher thread */ |
cd9adb8b JG |
4489 | ret = pthread_create(&dispatcher_thread, |
4490 | default_pthread_attr(), | |
4491 | relay_thread_dispatcher, | |
4492 | (void *) nullptr); | |
178a0557 MD |
4493 | if (ret) { |
4494 | errno = ret; | |
b8aa1682 | 4495 | PERROR("pthread_create dispatcher"); |
178a0557 MD |
4496 | retval = -1; |
4497 | goto exit_dispatcher_thread; | |
b8aa1682 JD |
4498 | } |
4499 | ||
4500 | /* Setup the worker thread */ | |
cd9adb8b | 4501 | ret = pthread_create(&worker_thread, default_pthread_attr(), relay_thread_worker, nullptr); |
178a0557 MD |
4502 | if (ret) { |
4503 | errno = ret; | |
b8aa1682 | 4504 | PERROR("pthread_create worker"); |
178a0557 MD |
4505 | retval = -1; |
4506 | goto exit_worker_thread; | |
b8aa1682 JD |
4507 | } |
4508 | ||
4509 | /* Setup the listener thread */ | |
28ab034a | 4510 | ret = pthread_create( |
cd9adb8b | 4511 | &listener_thread, default_pthread_attr(), relay_thread_listener, (void *) nullptr); |
178a0557 MD |
4512 | if (ret) { |
4513 | errno = ret; | |
b8aa1682 | 4514 | PERROR("pthread_create listener"); |
178a0557 MD |
4515 | retval = -1; |
4516 | goto exit_listener_thread; | |
b8aa1682 JD |
4517 | } |
4518 | ||
7591bab1 | 4519 | ret = relayd_live_create(live_uri); |
178a0557 | 4520 | if (ret) { |
d3e2ba59 | 4521 | ERR("Starting live viewer threads"); |
178a0557 | 4522 | retval = -1; |
50138f51 | 4523 | goto exit_live; |
d3e2ba59 JD |
4524 | } |
4525 | ||
178a0557 MD |
4526 | /* |
4527 | * This is where we start awaiting program completion (e.g. through | |
4528 | * signal that asks threads to teardown). | |
4529 | */ | |
4530 | ||
4531 | ret = relayd_live_join(); | |
4532 | if (ret) { | |
4533 | retval = -1; | |
4534 | } | |
50138f51 | 4535 | exit_live: |
178a0557 | 4536 | |
b8aa1682 | 4537 | ret = pthread_join(listener_thread, &status); |
178a0557 MD |
4538 | if (ret) { |
4539 | errno = ret; | |
4540 | PERROR("pthread_join listener_thread"); | |
4541 | retval = -1; | |
b8aa1682 JD |
4542 | } |
4543 | ||
178a0557 | 4544 | exit_listener_thread: |
b8aa1682 | 4545 | ret = pthread_join(worker_thread, &status); |
178a0557 MD |
4546 | if (ret) { |
4547 | errno = ret; | |
4548 | PERROR("pthread_join worker_thread"); | |
4549 | retval = -1; | |
b8aa1682 JD |
4550 | } |
4551 | ||
178a0557 | 4552 | exit_worker_thread: |
b8aa1682 | 4553 | ret = pthread_join(dispatcher_thread, &status); |
178a0557 MD |
4554 | if (ret) { |
4555 | errno = ret; | |
4556 | PERROR("pthread_join dispatcher_thread"); | |
4557 | retval = -1; | |
b8aa1682 | 4558 | } |
178a0557 | 4559 | exit_dispatcher_thread: |
42415026 | 4560 | |
65931c8b | 4561 | ret = pthread_join(health_thread, &status); |
178a0557 MD |
4562 | if (ret) { |
4563 | errno = ret; | |
4564 | PERROR("pthread_join health_thread"); | |
4565 | retval = -1; | |
65931c8b | 4566 | } |
178a0557 | 4567 | exit_options: |
4d62fbf8 MD |
4568 | /* |
4569 | * Wait for all pending call_rcu work to complete before tearing | |
4570 | * down data structures. call_rcu worker may be trying to | |
4571 | * perform lookups in those structures. | |
4572 | */ | |
4573 | rcu_barrier(); | |
7591bab1 MD |
4574 | relayd_cleanup(); |
4575 | ||
4576 | /* Ensure all prior call_rcu are done. */ | |
4577 | rcu_barrier(); | |
d3e2ba59 | 4578 | |
00e3b7f1 JG |
4579 | if (thread_is_rcu_registered) { |
4580 | rcu_unregister_thread(); | |
4581 | } | |
4582 | ||
178a0557 | 4583 | if (!retval) { |
b8aa1682 | 4584 | exit(EXIT_SUCCESS); |
178a0557 MD |
4585 | } else { |
4586 | exit(EXIT_FAILURE); | |
b8aa1682 | 4587 | } |
b8aa1682 | 4588 | } |