Commit | Line | Data |
---|---|---|
65931c8b MD |
1 | /* |
2 | * Copyright (C) 2013 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com> | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or modify | |
5 | * it under the terms of the GNU General Public License, version 2 only, | |
6 | * as published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it will be useful, | |
9 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
11 | * GNU General Public License for more details. | |
12 | * | |
13 | * You should have received a copy of the GNU General Public License along | |
14 | * with this program; if not, write to the Free Software Foundation, Inc., | |
15 | * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. | |
16 | */ | |
17 | ||
18 | #define _GNU_SOURCE | |
6c1c0768 | 19 | #define _LGPL_SOURCE |
65931c8b MD |
20 | #include <fcntl.h> |
21 | #include <getopt.h> | |
22 | #include <grp.h> | |
23 | #include <limits.h> | |
24 | #include <pthread.h> | |
25 | #include <signal.h> | |
26 | #include <stdio.h> | |
27 | #include <stdlib.h> | |
28 | #include <string.h> | |
29 | #include <sys/ipc.h> | |
30 | #include <sys/resource.h> | |
31 | #include <sys/shm.h> | |
32 | #include <sys/socket.h> | |
33 | #include <sys/stat.h> | |
34 | #include <sys/types.h> | |
35 | #include <urcu/list.h> | |
36 | #include <poll.h> | |
37 | #include <unistd.h> | |
38 | #include <sys/mman.h> | |
39 | #include <assert.h> | |
40 | #include <config.h> | |
41 | #include <urcu/compiler.h> | |
42 | #include <ulimit.h> | |
43 | #include <inttypes.h> | |
44 | ||
45 | #include <common/defaults.h> | |
46 | #include <common/common.h> | |
47 | #include <common/consumer.h> | |
48 | #include <common/consumer-timer.h> | |
49 | #include <common/compat/poll.h> | |
50 | #include <common/sessiond-comm/sessiond-comm.h> | |
51 | #include <common/utils.h> | |
e8fa9fb0 | 52 | #include <common/compat/getenv.h> |
65931c8b MD |
53 | |
54 | #include "lttng-relayd.h" | |
55 | #include "health-relayd.h" | |
56 | ||
57 | /* Global health check unix path */ | |
094fe907 MD |
58 | static |
59 | char health_unix_sock_path[PATH_MAX]; | |
65931c8b MD |
60 | |
61 | int health_quit_pipe[2]; | |
62 | ||
63 | /* | |
64 | * Check if the thread quit pipe was triggered. | |
65 | * | |
66 | * Return 1 if it was triggered else 0; | |
67 | */ | |
68 | static | |
69 | int check_health_quit_pipe(int fd, uint32_t events) | |
70 | { | |
71 | if (fd == health_quit_pipe[0] && (events & LPOLLIN)) { | |
72 | return 1; | |
73 | } | |
74 | ||
75 | return 0; | |
76 | } | |
77 | ||
78 | /* | |
79 | * Send data on a unix socket using the liblttsessiondcomm API. | |
80 | * | |
81 | * Return lttcomm error code. | |
82 | */ | |
83 | static int send_unix_sock(int sock, void *buf, size_t len) | |
84 | { | |
85 | /* Check valid length */ | |
86 | if (len == 0) { | |
87 | return -1; | |
88 | } | |
89 | ||
90 | return lttcomm_send_unix_sock(sock, buf, len); | |
91 | } | |
92 | ||
93 | static int create_lttng_rundir_with_perm(const char *rundir) | |
94 | { | |
95 | int ret; | |
96 | ||
97 | DBG3("Creating LTTng run directory: %s", rundir); | |
98 | ||
99 | ret = mkdir(rundir, S_IRWXU); | |
100 | if (ret < 0) { | |
101 | if (errno != EEXIST) { | |
102 | ERR("Unable to create %s", rundir); | |
103 | goto error; | |
104 | } else { | |
105 | ret = 0; | |
106 | } | |
107 | } else if (ret == 0) { | |
108 | int is_root = !getuid(); | |
109 | ||
110 | if (is_root) { | |
111 | ret = chown(rundir, 0, | |
112 | utils_get_group_id(tracing_group_name)); | |
113 | if (ret < 0) { | |
114 | ERR("Unable to set group on %s", rundir); | |
115 | PERROR("chown"); | |
116 | ret = -1; | |
117 | goto error; | |
118 | } | |
119 | ||
120 | ret = chmod(rundir, | |
121 | S_IRUSR | S_IWUSR | S_IXUSR | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH); | |
122 | if (ret < 0) { | |
123 | ERR("Unable to set permissions on %s", health_unix_sock_path); | |
124 | PERROR("chmod"); | |
125 | ret = -1; | |
126 | goto error; | |
127 | } | |
128 | } | |
129 | } | |
130 | ||
131 | error: | |
132 | return ret; | |
133 | } | |
134 | ||
094fe907 MD |
135 | static |
136 | int parse_health_env(void) | |
137 | { | |
138 | const char *health_path; | |
139 | ||
e8fa9fb0 | 140 | health_path = lttng_secure_getenv(LTTNG_RELAYD_HEALTH_ENV); |
094fe907 MD |
141 | if (health_path) { |
142 | strncpy(health_unix_sock_path, health_path, | |
143 | PATH_MAX); | |
144 | health_unix_sock_path[PATH_MAX - 1] = '\0'; | |
145 | } | |
146 | ||
147 | return 0; | |
148 | } | |
149 | ||
65931c8b MD |
150 | static |
151 | int setup_health_path(void) | |
152 | { | |
153 | int is_root, ret = 0; | |
c3844e39 | 154 | char *home_path = NULL, *rundir = NULL, *relayd_path; |
65931c8b | 155 | |
094fe907 MD |
156 | ret = parse_health_env(); |
157 | if (ret) { | |
158 | return ret; | |
159 | } | |
160 | ||
65931c8b MD |
161 | is_root = !getuid(); |
162 | ||
163 | if (is_root) { | |
c3844e39 | 164 | rundir = strdup(DEFAULT_LTTNG_RUNDIR); |
b6ab01aa MD |
165 | if (!rundir) { |
166 | ret = -ENOMEM; | |
167 | goto end; | |
168 | } | |
65931c8b MD |
169 | } else { |
170 | /* | |
171 | * Create rundir from home path. This will create something like | |
172 | * $HOME/.lttng | |
173 | */ | |
174 | home_path = utils_get_home_dir(); | |
175 | ||
176 | if (home_path == NULL) { | |
177 | /* TODO: Add --socket PATH option */ | |
178 | ERR("Can't get HOME directory for sockets creation."); | |
179 | ret = -EPERM; | |
180 | goto end; | |
181 | } | |
182 | ||
c3844e39 | 183 | ret = asprintf(&rundir, DEFAULT_LTTNG_HOME_RUNDIR, home_path); |
65931c8b MD |
184 | if (ret < 0) { |
185 | ret = -ENOMEM; | |
186 | goto end; | |
187 | } | |
188 | } | |
189 | ||
c3844e39 | 190 | ret = asprintf(&relayd_path, DEFAULT_RELAYD_PATH, rundir); |
65931c8b MD |
191 | if (ret < 0) { |
192 | ret = -ENOMEM; | |
193 | goto end; | |
194 | } | |
195 | ||
c3844e39 | 196 | ret = create_lttng_rundir_with_perm(rundir); |
65931c8b MD |
197 | if (ret < 0) { |
198 | goto end; | |
199 | } | |
200 | ||
201 | ret = create_lttng_rundir_with_perm(relayd_path); | |
202 | if (ret < 0) { | |
203 | goto end; | |
204 | } | |
205 | ||
206 | if (is_root) { | |
207 | if (strlen(health_unix_sock_path) != 0) { | |
208 | goto end; | |
209 | } | |
210 | snprintf(health_unix_sock_path, sizeof(health_unix_sock_path), | |
211 | DEFAULT_GLOBAL_RELAY_HEALTH_UNIX_SOCK, | |
212 | getpid()); | |
213 | } else { | |
214 | /* Set health check Unix path */ | |
215 | if (strlen(health_unix_sock_path) != 0) { | |
216 | goto end; | |
217 | } | |
218 | ||
219 | snprintf(health_unix_sock_path, sizeof(health_unix_sock_path), | |
220 | DEFAULT_HOME_RELAY_HEALTH_UNIX_SOCK, | |
221 | home_path, getpid()); | |
222 | } | |
223 | ||
224 | end: | |
c3844e39 | 225 | free(rundir); |
65931c8b MD |
226 | return ret; |
227 | } | |
228 | ||
229 | /* | |
230 | * Thread managing health check socket. | |
231 | */ | |
232 | void *thread_manage_health(void *data) | |
233 | { | |
234 | int sock = -1, new_sock = -1, ret, i, pollfd, err = -1; | |
235 | uint32_t revents, nb_fd; | |
236 | struct lttng_poll_event events; | |
237 | struct health_comm_msg msg; | |
238 | struct health_comm_reply reply; | |
239 | int is_root; | |
240 | ||
241 | DBG("[thread] Manage health check started"); | |
242 | ||
243 | setup_health_path(); | |
244 | ||
245 | rcu_register_thread(); | |
246 | ||
247 | /* We might hit an error path before this is created. */ | |
248 | lttng_poll_init(&events); | |
249 | ||
250 | /* Create unix socket */ | |
251 | sock = lttcomm_create_unix_sock(health_unix_sock_path); | |
252 | if (sock < 0) { | |
253 | ERR("Unable to create health check Unix socket"); | |
254 | ret = -1; | |
255 | goto error; | |
256 | } | |
257 | ||
258 | is_root = !getuid(); | |
259 | if (is_root) { | |
260 | /* lttng health client socket path permissions */ | |
261 | ret = chown(health_unix_sock_path, 0, | |
262 | utils_get_group_id(tracing_group_name)); | |
263 | if (ret < 0) { | |
264 | ERR("Unable to set group on %s", health_unix_sock_path); | |
265 | PERROR("chown"); | |
266 | ret = -1; | |
267 | goto error; | |
268 | } | |
269 | ||
270 | ret = chmod(health_unix_sock_path, | |
271 | S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP); | |
272 | if (ret < 0) { | |
273 | ERR("Unable to set permissions on %s", health_unix_sock_path); | |
274 | PERROR("chmod"); | |
275 | ret = -1; | |
276 | goto error; | |
277 | } | |
278 | } | |
279 | ||
280 | /* | |
281 | * Set the CLOEXEC flag. Return code is useless because either way, the | |
282 | * show must go on. | |
283 | */ | |
284 | (void) utils_set_fd_cloexec(sock); | |
285 | ||
286 | ret = lttcomm_listen_unix_sock(sock); | |
287 | if (ret < 0) { | |
288 | goto error; | |
289 | } | |
290 | ||
291 | /* Size is set to 1 for the consumer_channel pipe */ | |
292 | ret = lttng_poll_create(&events, 2, LTTNG_CLOEXEC); | |
293 | if (ret < 0) { | |
294 | ERR("Poll set creation failed"); | |
295 | goto error; | |
296 | } | |
297 | ||
298 | ret = lttng_poll_add(&events, health_quit_pipe[0], LPOLLIN); | |
299 | if (ret < 0) { | |
300 | goto error; | |
301 | } | |
302 | ||
303 | /* Add the application registration socket */ | |
304 | ret = lttng_poll_add(&events, sock, LPOLLIN | LPOLLPRI); | |
305 | if (ret < 0) { | |
306 | goto error; | |
307 | } | |
308 | ||
3fd27398 MD |
309 | lttng_relay_notify_ready(); |
310 | ||
65931c8b MD |
311 | while (1) { |
312 | DBG("Health check ready"); | |
313 | ||
314 | /* Inifinite blocking call, waiting for transmission */ | |
315 | restart: | |
316 | ret = lttng_poll_wait(&events, -1); | |
317 | if (ret < 0) { | |
318 | /* | |
319 | * Restart interrupted system call. | |
320 | */ | |
321 | if (errno == EINTR) { | |
322 | goto restart; | |
323 | } | |
324 | goto error; | |
325 | } | |
326 | ||
327 | nb_fd = ret; | |
328 | ||
329 | for (i = 0; i < nb_fd; i++) { | |
330 | /* Fetch once the poll data */ | |
331 | revents = LTTNG_POLL_GETEV(&events, i); | |
332 | pollfd = LTTNG_POLL_GETFD(&events, i); | |
333 | ||
fd20dac9 MD |
334 | if (!revents) { |
335 | /* No activity for this FD (poll implementation). */ | |
336 | continue; | |
337 | } | |
338 | ||
65931c8b MD |
339 | /* Thread quit pipe has been closed. Killing thread. */ |
340 | ret = check_health_quit_pipe(pollfd, revents); | |
341 | if (ret) { | |
342 | err = 0; | |
343 | goto exit; | |
344 | } | |
345 | ||
346 | /* Event on the registration socket */ | |
347 | if (pollfd == sock) { | |
348 | if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) { | |
349 | ERR("Health socket poll error"); | |
350 | goto error; | |
351 | } | |
352 | } | |
353 | } | |
354 | ||
355 | new_sock = lttcomm_accept_unix_sock(sock); | |
356 | if (new_sock < 0) { | |
357 | goto error; | |
358 | } | |
359 | ||
360 | /* | |
361 | * Set the CLOEXEC flag. Return code is useless because either way, the | |
362 | * show must go on. | |
363 | */ | |
364 | (void) utils_set_fd_cloexec(new_sock); | |
365 | ||
366 | DBG("Receiving data from client for health..."); | |
367 | ret = lttcomm_recv_unix_sock(new_sock, (void *)&msg, sizeof(msg)); | |
368 | if (ret <= 0) { | |
369 | DBG("Nothing recv() from client... continuing"); | |
370 | ret = close(new_sock); | |
371 | if (ret) { | |
372 | PERROR("close"); | |
373 | } | |
374 | new_sock = -1; | |
375 | continue; | |
376 | } | |
377 | ||
378 | rcu_thread_online(); | |
379 | ||
380 | assert(msg.cmd == HEALTH_CMD_CHECK); | |
381 | ||
53efb85a | 382 | memset(&reply, 0, sizeof(reply)); |
65931c8b MD |
383 | for (i = 0; i < NR_HEALTH_RELAYD_TYPES; i++) { |
384 | /* | |
385 | * health_check_state return 0 if thread is in | |
386 | * error. | |
387 | */ | |
388 | if (!health_check_state(health_relayd, i)) { | |
389 | reply.ret_code |= 1ULL << i; | |
390 | } | |
391 | } | |
392 | ||
393 | DBG2("Health check return value %" PRIx64, reply.ret_code); | |
394 | ||
395 | ret = send_unix_sock(new_sock, (void *) &reply, sizeof(reply)); | |
396 | if (ret < 0) { | |
397 | ERR("Failed to send health data back to client"); | |
398 | } | |
399 | ||
400 | /* End of transmission */ | |
401 | ret = close(new_sock); | |
402 | if (ret) { | |
403 | PERROR("close"); | |
404 | } | |
405 | new_sock = -1; | |
406 | } | |
407 | ||
408 | exit: | |
409 | error: | |
410 | if (err) { | |
411 | ERR("Health error occurred in %s", __func__); | |
412 | } | |
413 | DBG("Health check thread dying"); | |
414 | unlink(health_unix_sock_path); | |
415 | if (sock >= 0) { | |
416 | ret = close(sock); | |
417 | if (ret) { | |
418 | PERROR("close"); | |
419 | } | |
420 | } | |
421 | ||
dcbcae3e MD |
422 | /* |
423 | * We do NOT rmdir rundir nor the relayd path because there are | |
424 | * other processes using them. | |
425 | */ | |
426 | ||
65931c8b MD |
427 | lttng_poll_clean(&events); |
428 | ||
429 | rcu_unregister_thread(); | |
430 | return NULL; | |
431 | } |