10a0e5fdffe1b88ab4ae144dcfff87f381b38b1a
[lttng-tools.git] / src / bin / lttng-relayd / health-relayd.c
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 _LGPL_SOURCE
19 #include <fcntl.h>
20 #include <getopt.h>
21 #include <grp.h>
22 #include <limits.h>
23 #include <pthread.h>
24 #include <signal.h>
25 #include <stdio.h>
26 #include <stdlib.h>
27 #include <string.h>
28 #include <sys/ipc.h>
29 #include <sys/resource.h>
30 #include <sys/shm.h>
31 #include <sys/socket.h>
32 #include <sys/stat.h>
33 #include <sys/types.h>
34 #include <urcu/list.h>
35 #include <poll.h>
36 #include <unistd.h>
37 #include <sys/mman.h>
38 #include <assert.h>
39 #include <urcu/compiler.h>
40 #include <ulimit.h>
41 #include <inttypes.h>
42
43 #include <common/defaults.h>
44 #include <common/common.h>
45 #include <common/consumer.h>
46 #include <common/consumer-timer.h>
47 #include <common/compat/poll.h>
48 #include <common/sessiond-comm/sessiond-comm.h>
49 #include <common/utils.h>
50 #include <common/compat/getenv.h>
51
52 #include "lttng-relayd.h"
53 #include "health-relayd.h"
54
55 /* Global health check unix path */
56 static
57 char health_unix_sock_path[PATH_MAX];
58
59 int health_quit_pipe[2];
60
61 /*
62 * Check if the thread quit pipe was triggered.
63 *
64 * Return 1 if it was triggered else 0;
65 */
66 static
67 int check_health_quit_pipe(int fd, uint32_t events)
68 {
69 if (fd == health_quit_pipe[0] && (events & LPOLLIN)) {
70 return 1;
71 }
72
73 return 0;
74 }
75
76 /*
77 * Send data on a unix socket using the liblttsessiondcomm API.
78 *
79 * Return lttcomm error code.
80 */
81 static int send_unix_sock(int sock, void *buf, size_t len)
82 {
83 /* Check valid length */
84 if (len == 0) {
85 return -1;
86 }
87
88 return lttcomm_send_unix_sock(sock, buf, len);
89 }
90
91 static int create_lttng_rundir_with_perm(const char *rundir)
92 {
93 int ret;
94
95 DBG3("Creating LTTng run directory: %s", rundir);
96
97 ret = mkdir(rundir, S_IRWXU);
98 if (ret < 0) {
99 if (errno != EEXIST) {
100 ERR("Unable to create %s", rundir);
101 goto error;
102 } else {
103 ret = 0;
104 }
105 } else if (ret == 0) {
106 int is_root = !getuid();
107
108 if (is_root) {
109 ret = chown(rundir, 0,
110 utils_get_group_id(tracing_group_name));
111 if (ret < 0) {
112 ERR("Unable to set group on %s", rundir);
113 PERROR("chown");
114 ret = -1;
115 goto error;
116 }
117
118 ret = chmod(rundir,
119 S_IRUSR | S_IWUSR | S_IXUSR | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH);
120 if (ret < 0) {
121 ERR("Unable to set permissions on %s", health_unix_sock_path);
122 PERROR("chmod");
123 ret = -1;
124 goto error;
125 }
126 }
127 }
128
129 error:
130 return ret;
131 }
132
133 static
134 int parse_health_env(void)
135 {
136 const char *health_path;
137
138 health_path = lttng_secure_getenv(LTTNG_RELAYD_HEALTH_ENV);
139 if (health_path) {
140 strncpy(health_unix_sock_path, health_path,
141 PATH_MAX);
142 health_unix_sock_path[PATH_MAX - 1] = '\0';
143 }
144
145 return 0;
146 }
147
148 static
149 int setup_health_path(void)
150 {
151 int is_root, ret = 0;
152 char *home_path = NULL, *rundir = NULL, *relayd_path = NULL;
153
154 ret = parse_health_env();
155 if (ret) {
156 return ret;
157 }
158
159 is_root = !getuid();
160
161 if (is_root) {
162 rundir = strdup(DEFAULT_LTTNG_RUNDIR);
163 if (!rundir) {
164 ret = -ENOMEM;
165 goto end;
166 }
167 } else {
168 /*
169 * Create rundir from home path. This will create something like
170 * $HOME/.lttng
171 */
172 home_path = utils_get_home_dir();
173
174 if (home_path == NULL) {
175 /* TODO: Add --socket PATH option */
176 ERR("Can't get HOME directory for sockets creation.");
177 ret = -EPERM;
178 goto end;
179 }
180
181 ret = asprintf(&rundir, DEFAULT_LTTNG_HOME_RUNDIR, home_path);
182 if (ret < 0) {
183 ret = -ENOMEM;
184 goto end;
185 }
186 }
187
188 ret = asprintf(&relayd_path, DEFAULT_RELAYD_PATH, rundir);
189 if (ret < 0) {
190 ret = -ENOMEM;
191 goto end;
192 }
193
194 ret = create_lttng_rundir_with_perm(rundir);
195 if (ret < 0) {
196 goto end;
197 }
198
199 ret = create_lttng_rundir_with_perm(relayd_path);
200 if (ret < 0) {
201 goto end;
202 }
203
204 if (is_root) {
205 if (strlen(health_unix_sock_path) != 0) {
206 goto end;
207 }
208 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
209 DEFAULT_GLOBAL_RELAY_HEALTH_UNIX_SOCK,
210 getpid());
211 } else {
212 /* Set health check Unix path */
213 if (strlen(health_unix_sock_path) != 0) {
214 goto end;
215 }
216
217 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
218 DEFAULT_HOME_RELAY_HEALTH_UNIX_SOCK,
219 home_path, getpid());
220 }
221
222 end:
223 free(rundir);
224 free(relayd_path);
225 return ret;
226 }
227
228 /*
229 * Thread managing health check socket.
230 */
231 void *thread_manage_health(void *data)
232 {
233 int sock = -1, new_sock = -1, ret, i, pollfd, err = -1;
234 uint32_t revents, nb_fd;
235 struct lttng_poll_event events;
236 struct health_comm_msg msg;
237 struct health_comm_reply reply;
238 int is_root;
239
240 DBG("[thread] Manage health check started");
241
242 setup_health_path();
243
244 rcu_register_thread();
245
246 /* We might hit an error path before this is created. */
247 lttng_poll_init(&events);
248
249 /* Create unix socket */
250 sock = lttcomm_create_unix_sock(health_unix_sock_path);
251 if (sock < 0) {
252 ERR("Unable to create health check Unix socket");
253 ret = -1;
254 goto error;
255 }
256
257 is_root = !getuid();
258 if (is_root) {
259 /* lttng health client socket path permissions */
260 ret = chown(health_unix_sock_path, 0,
261 utils_get_group_id(tracing_group_name));
262 if (ret < 0) {
263 ERR("Unable to set group on %s", health_unix_sock_path);
264 PERROR("chown");
265 ret = -1;
266 goto error;
267 }
268
269 ret = chmod(health_unix_sock_path,
270 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
271 if (ret < 0) {
272 ERR("Unable to set permissions on %s", health_unix_sock_path);
273 PERROR("chmod");
274 ret = -1;
275 goto error;
276 }
277 }
278
279 /*
280 * Set the CLOEXEC flag. Return code is useless because either way, the
281 * show must go on.
282 */
283 (void) utils_set_fd_cloexec(sock);
284
285 ret = lttcomm_listen_unix_sock(sock);
286 if (ret < 0) {
287 goto error;
288 }
289
290 /* Size is set to 1 for the consumer_channel pipe */
291 ret = lttng_poll_create(&events, 2, LTTNG_CLOEXEC);
292 if (ret < 0) {
293 ERR("Poll set creation failed");
294 goto error;
295 }
296
297 ret = lttng_poll_add(&events, health_quit_pipe[0], LPOLLIN);
298 if (ret < 0) {
299 goto error;
300 }
301
302 /* Add the application registration socket */
303 ret = lttng_poll_add(&events, sock, LPOLLIN | LPOLLPRI);
304 if (ret < 0) {
305 goto error;
306 }
307
308 lttng_relay_notify_ready();
309
310 while (1) {
311 DBG("Health check ready");
312
313 /* Inifinite blocking call, waiting for transmission */
314 restart:
315 ret = lttng_poll_wait(&events, -1);
316 if (ret < 0) {
317 /*
318 * Restart interrupted system call.
319 */
320 if (errno == EINTR) {
321 goto restart;
322 }
323 goto error;
324 }
325
326 nb_fd = ret;
327
328 for (i = 0; i < nb_fd; i++) {
329 /* Fetch once the poll data */
330 revents = LTTNG_POLL_GETEV(&events, i);
331 pollfd = LTTNG_POLL_GETFD(&events, i);
332
333 if (!revents) {
334 /* No activity for this FD (poll implementation). */
335 continue;
336 }
337
338 /* Thread quit pipe has been closed. Killing thread. */
339 ret = check_health_quit_pipe(pollfd, revents);
340 if (ret) {
341 err = 0;
342 goto exit;
343 }
344
345 /* Event on the registration socket */
346 if (pollfd == sock) {
347 if (revents & LPOLLIN) {
348 continue;
349 } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
350 ERR("Health socket poll error");
351 goto error;
352 } else {
353 ERR("Unexpected poll events %u for sock %d", revents, pollfd);
354 goto error;
355 }
356 }
357 }
358
359 new_sock = lttcomm_accept_unix_sock(sock);
360 if (new_sock < 0) {
361 goto error;
362 }
363
364 /*
365 * Set the CLOEXEC flag. Return code is useless because either way, the
366 * show must go on.
367 */
368 (void) utils_set_fd_cloexec(new_sock);
369
370 DBG("Receiving data from client for health...");
371 ret = lttcomm_recv_unix_sock(new_sock, (void *)&msg, sizeof(msg));
372 if (ret <= 0) {
373 DBG("Nothing recv() from client... continuing");
374 ret = close(new_sock);
375 if (ret) {
376 PERROR("close");
377 }
378 new_sock = -1;
379 continue;
380 }
381
382 rcu_thread_online();
383
384 assert(msg.cmd == HEALTH_CMD_CHECK);
385
386 memset(&reply, 0, sizeof(reply));
387 for (i = 0; i < NR_HEALTH_RELAYD_TYPES; i++) {
388 /*
389 * health_check_state return 0 if thread is in
390 * error.
391 */
392 if (!health_check_state(health_relayd, i)) {
393 reply.ret_code |= 1ULL << i;
394 }
395 }
396
397 DBG2("Health check return value %" PRIx64, reply.ret_code);
398
399 ret = send_unix_sock(new_sock, (void *) &reply, sizeof(reply));
400 if (ret < 0) {
401 ERR("Failed to send health data back to client");
402 }
403
404 /* End of transmission */
405 ret = close(new_sock);
406 if (ret) {
407 PERROR("close");
408 }
409 new_sock = -1;
410 }
411
412 exit:
413 error:
414 if (err) {
415 ERR("Health error occurred in %s", __func__);
416 }
417 DBG("Health check thread dying");
418 unlink(health_unix_sock_path);
419 if (sock >= 0) {
420 ret = close(sock);
421 if (ret) {
422 PERROR("close");
423 }
424 }
425
426 /*
427 * We do NOT rmdir rundir nor the relayd path because there are
428 * other processes using them.
429 */
430
431 lttng_poll_clean(&events);
432
433 rcu_unregister_thread();
434 return NULL;
435 }
This page took 0.05736 seconds and 3 git commands to generate.