Fix: getgrnam is not MT-Safe, use getgrnam_r
[lttng-tools.git] / src / bin / lttng-relayd / health-relayd.c
CommitLineData
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
6c1c0768 18#define _LGPL_SOURCE
65931c8b
MD
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>
65931c8b 39#include <urcu/compiler.h>
65931c8b
MD
40#include <inttypes.h>
41
42#include <common/defaults.h>
43#include <common/common.h>
c8fea79c
JR
44#include <common/consumer/consumer.h>
45#include <common/consumer/consumer-timer.h>
65931c8b
MD
46#include <common/compat/poll.h>
47#include <common/sessiond-comm/sessiond-comm.h>
48#include <common/utils.h>
e8fa9fb0 49#include <common/compat/getenv.h>
65931c8b
MD
50
51#include "lttng-relayd.h"
52#include "health-relayd.h"
53
54/* Global health check unix path */
094fe907
MD
55static
56char health_unix_sock_path[PATH_MAX];
65931c8b
MD
57
58int health_quit_pipe[2];
59
60/*
61 * Check if the thread quit pipe was triggered.
62 *
63 * Return 1 if it was triggered else 0;
64 */
65static
66int check_health_quit_pipe(int fd, uint32_t events)
67{
68 if (fd == health_quit_pipe[0] && (events & LPOLLIN)) {
69 return 1;
70 }
71
72 return 0;
73}
74
75/*
76 * Send data on a unix socket using the liblttsessiondcomm API.
77 *
78 * Return lttcomm error code.
79 */
80static int send_unix_sock(int sock, void *buf, size_t len)
81{
82 /* Check valid length */
83 if (len == 0) {
84 return -1;
85 }
86
87 return lttcomm_send_unix_sock(sock, buf, len);
88}
89
90static int create_lttng_rundir_with_perm(const char *rundir)
91{
92 int ret;
93
94 DBG3("Creating LTTng run directory: %s", rundir);
95
96 ret = mkdir(rundir, S_IRWXU);
97 if (ret < 0) {
98 if (errno != EEXIST) {
99 ERR("Unable to create %s", rundir);
100 goto error;
101 } else {
102 ret = 0;
103 }
104 } else if (ret == 0) {
105 int is_root = !getuid();
106
107 if (is_root) {
28ab59d0
JR
108 gid_t gid;
109
110 ret = utils_get_group_id(tracing_group_name, true, &gid);
111 if (ret) {
112 /* Default to root group. */
113 gid = 0;
114 }
115
116 ret = chown(rundir, 0, gid);
65931c8b
MD
117 if (ret < 0) {
118 ERR("Unable to set group on %s", rundir);
119 PERROR("chown");
120 ret = -1;
121 goto error;
122 }
123
124 ret = chmod(rundir,
125 S_IRUSR | S_IWUSR | S_IXUSR | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH);
126 if (ret < 0) {
127 ERR("Unable to set permissions on %s", health_unix_sock_path);
128 PERROR("chmod");
129 ret = -1;
130 goto error;
131 }
132 }
133 }
134
135error:
136 return ret;
137}
138
094fe907
MD
139static
140int parse_health_env(void)
141{
142 const char *health_path;
143
e8fa9fb0 144 health_path = lttng_secure_getenv(LTTNG_RELAYD_HEALTH_ENV);
094fe907
MD
145 if (health_path) {
146 strncpy(health_unix_sock_path, health_path,
147 PATH_MAX);
148 health_unix_sock_path[PATH_MAX - 1] = '\0';
149 }
150
151 return 0;
152}
153
65931c8b
MD
154static
155int setup_health_path(void)
156{
157 int is_root, ret = 0;
edd94901 158 char *home_path = NULL, *rundir = NULL, *relayd_path = NULL;
65931c8b 159
094fe907
MD
160 ret = parse_health_env();
161 if (ret) {
162 return ret;
163 }
164
65931c8b
MD
165 is_root = !getuid();
166
167 if (is_root) {
c3844e39 168 rundir = strdup(DEFAULT_LTTNG_RUNDIR);
b6ab01aa
MD
169 if (!rundir) {
170 ret = -ENOMEM;
171 goto end;
172 }
65931c8b
MD
173 } else {
174 /*
175 * Create rundir from home path. This will create something like
176 * $HOME/.lttng
177 */
178 home_path = utils_get_home_dir();
179
180 if (home_path == NULL) {
181 /* TODO: Add --socket PATH option */
182 ERR("Can't get HOME directory for sockets creation.");
183 ret = -EPERM;
184 goto end;
185 }
186
c3844e39 187 ret = asprintf(&rundir, DEFAULT_LTTNG_HOME_RUNDIR, home_path);
65931c8b
MD
188 if (ret < 0) {
189 ret = -ENOMEM;
190 goto end;
191 }
192 }
193
c3844e39 194 ret = asprintf(&relayd_path, DEFAULT_RELAYD_PATH, rundir);
65931c8b
MD
195 if (ret < 0) {
196 ret = -ENOMEM;
197 goto end;
198 }
199
c3844e39 200 ret = create_lttng_rundir_with_perm(rundir);
65931c8b
MD
201 if (ret < 0) {
202 goto end;
203 }
204
205 ret = create_lttng_rundir_with_perm(relayd_path);
206 if (ret < 0) {
207 goto end;
208 }
209
210 if (is_root) {
211 if (strlen(health_unix_sock_path) != 0) {
212 goto end;
213 }
214 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
215 DEFAULT_GLOBAL_RELAY_HEALTH_UNIX_SOCK,
d1f721c5 216 (int) getpid());
65931c8b
MD
217 } else {
218 /* Set health check Unix path */
219 if (strlen(health_unix_sock_path) != 0) {
220 goto end;
221 }
222
223 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
224 DEFAULT_HOME_RELAY_HEALTH_UNIX_SOCK,
d1f721c5 225 home_path, (int) getpid());
65931c8b
MD
226 }
227
228end:
c3844e39 229 free(rundir);
edd94901 230 free(relayd_path);
65931c8b
MD
231 return ret;
232}
233
234/*
235 * Thread managing health check socket.
236 */
237void *thread_manage_health(void *data)
238{
239 int sock = -1, new_sock = -1, ret, i, pollfd, err = -1;
240 uint32_t revents, nb_fd;
241 struct lttng_poll_event events;
242 struct health_comm_msg msg;
243 struct health_comm_reply reply;
244 int is_root;
245
246 DBG("[thread] Manage health check started");
247
248 setup_health_path();
249
250 rcu_register_thread();
251
252 /* We might hit an error path before this is created. */
253 lttng_poll_init(&events);
254
255 /* Create unix socket */
256 sock = lttcomm_create_unix_sock(health_unix_sock_path);
257 if (sock < 0) {
258 ERR("Unable to create health check Unix socket");
7568ddbf 259 err = -1;
65931c8b
MD
260 goto error;
261 }
262
263 is_root = !getuid();
264 if (is_root) {
265 /* lttng health client socket path permissions */
28ab59d0
JR
266 gid_t gid;
267
268 ret = utils_get_group_id(tracing_group_name, true, &gid);
269 if (ret) {
270 /* Default to root group. */
271 gid = 0;
272 }
273
274 ret = chown(health_unix_sock_path, 0, gid);
65931c8b
MD
275 if (ret < 0) {
276 ERR("Unable to set group on %s", health_unix_sock_path);
277 PERROR("chown");
7568ddbf 278 err = -1;
65931c8b
MD
279 goto error;
280 }
281
282 ret = chmod(health_unix_sock_path,
283 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
284 if (ret < 0) {
285 ERR("Unable to set permissions on %s", health_unix_sock_path);
286 PERROR("chmod");
7568ddbf 287 err = -1;
65931c8b
MD
288 goto error;
289 }
290 }
291
292 /*
293 * Set the CLOEXEC flag. Return code is useless because either way, the
294 * show must go on.
295 */
296 (void) utils_set_fd_cloexec(sock);
297
298 ret = lttcomm_listen_unix_sock(sock);
299 if (ret < 0) {
300 goto error;
301 }
302
303 /* Size is set to 1 for the consumer_channel pipe */
304 ret = lttng_poll_create(&events, 2, LTTNG_CLOEXEC);
305 if (ret < 0) {
306 ERR("Poll set creation failed");
307 goto error;
308 }
309
310 ret = lttng_poll_add(&events, health_quit_pipe[0], LPOLLIN);
311 if (ret < 0) {
312 goto error;
313 }
314
315 /* Add the application registration socket */
316 ret = lttng_poll_add(&events, sock, LPOLLIN | LPOLLPRI);
317 if (ret < 0) {
318 goto error;
319 }
320
3fd27398
MD
321 lttng_relay_notify_ready();
322
65931c8b
MD
323 while (1) {
324 DBG("Health check ready");
325
326 /* Inifinite blocking call, waiting for transmission */
327restart:
328 ret = lttng_poll_wait(&events, -1);
329 if (ret < 0) {
330 /*
331 * Restart interrupted system call.
332 */
333 if (errno == EINTR) {
334 goto restart;
335 }
336 goto error;
337 }
338
339 nb_fd = ret;
340
341 for (i = 0; i < nb_fd; i++) {
342 /* Fetch once the poll data */
343 revents = LTTNG_POLL_GETEV(&events, i);
344 pollfd = LTTNG_POLL_GETFD(&events, i);
345
fd20dac9
MD
346 if (!revents) {
347 /* No activity for this FD (poll implementation). */
348 continue;
349 }
350
65931c8b
MD
351 /* Thread quit pipe has been closed. Killing thread. */
352 ret = check_health_quit_pipe(pollfd, revents);
353 if (ret) {
354 err = 0;
355 goto exit;
356 }
357
358 /* Event on the registration socket */
359 if (pollfd == sock) {
03e43155
MD
360 if (revents & LPOLLIN) {
361 continue;
362 } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
65931c8b
MD
363 ERR("Health socket poll error");
364 goto error;
03e43155
MD
365 } else {
366 ERR("Unexpected poll events %u for sock %d", revents, pollfd);
367 goto error;
65931c8b
MD
368 }
369 }
370 }
371
372 new_sock = lttcomm_accept_unix_sock(sock);
373 if (new_sock < 0) {
374 goto error;
375 }
376
377 /*
378 * Set the CLOEXEC flag. Return code is useless because either way, the
379 * show must go on.
380 */
381 (void) utils_set_fd_cloexec(new_sock);
382
383 DBG("Receiving data from client for health...");
384 ret = lttcomm_recv_unix_sock(new_sock, (void *)&msg, sizeof(msg));
385 if (ret <= 0) {
386 DBG("Nothing recv() from client... continuing");
387 ret = close(new_sock);
388 if (ret) {
389 PERROR("close");
390 }
391 new_sock = -1;
392 continue;
393 }
394
395 rcu_thread_online();
396
397 assert(msg.cmd == HEALTH_CMD_CHECK);
398
53efb85a 399 memset(&reply, 0, sizeof(reply));
65931c8b
MD
400 for (i = 0; i < NR_HEALTH_RELAYD_TYPES; i++) {
401 /*
402 * health_check_state return 0 if thread is in
403 * error.
404 */
405 if (!health_check_state(health_relayd, i)) {
406 reply.ret_code |= 1ULL << i;
407 }
408 }
409
410 DBG2("Health check return value %" PRIx64, reply.ret_code);
411
412 ret = send_unix_sock(new_sock, (void *) &reply, sizeof(reply));
413 if (ret < 0) {
414 ERR("Failed to send health data back to client");
415 }
416
417 /* End of transmission */
418 ret = close(new_sock);
419 if (ret) {
420 PERROR("close");
421 }
422 new_sock = -1;
423 }
424
65931c8b 425error:
81714439
JG
426 lttng_relay_stop_threads();
427exit:
65931c8b
MD
428 if (err) {
429 ERR("Health error occurred in %s", __func__);
430 }
431 DBG("Health check thread dying");
432 unlink(health_unix_sock_path);
433 if (sock >= 0) {
434 ret = close(sock);
435 if (ret) {
436 PERROR("close");
437 }
438 }
439
dcbcae3e
MD
440 /*
441 * We do NOT rmdir rundir nor the relayd path because there are
442 * other processes using them.
443 */
444
65931c8b
MD
445 lttng_poll_clean(&events);
446
447 rcu_unregister_thread();
448 return NULL;
449}
This page took 0.054394 seconds and 4 git commands to generate.