Fix: double RCU unlock on event_agent_disable_all
[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
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
58static
59char health_unix_sock_path[PATH_MAX];
65931c8b
MD
60
61int 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 */
68static
69int 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 */
83static 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
93static 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
131error:
132 return ret;
133}
134
094fe907
MD
135static
136int 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
150static
151int setup_health_path(void)
152{
153 int is_root, ret = 0;
edd94901 154 char *home_path = NULL, *rundir = NULL, *relayd_path = NULL;
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
224end:
c3844e39 225 free(rundir);
edd94901 226 free(relayd_path);
65931c8b
MD
227 return ret;
228}
229
230/*
231 * Thread managing health check socket.
232 */
233void *thread_manage_health(void *data)
234{
235 int sock = -1, new_sock = -1, ret, i, pollfd, err = -1;
236 uint32_t revents, nb_fd;
237 struct lttng_poll_event events;
238 struct health_comm_msg msg;
239 struct health_comm_reply reply;
240 int is_root;
241
242 DBG("[thread] Manage health check started");
243
244 setup_health_path();
245
246 rcu_register_thread();
247
248 /* We might hit an error path before this is created. */
249 lttng_poll_init(&events);
250
251 /* Create unix socket */
252 sock = lttcomm_create_unix_sock(health_unix_sock_path);
253 if (sock < 0) {
254 ERR("Unable to create health check Unix socket");
255 ret = -1;
256 goto error;
257 }
258
259 is_root = !getuid();
260 if (is_root) {
261 /* lttng health client socket path permissions */
262 ret = chown(health_unix_sock_path, 0,
263 utils_get_group_id(tracing_group_name));
264 if (ret < 0) {
265 ERR("Unable to set group on %s", health_unix_sock_path);
266 PERROR("chown");
267 ret = -1;
268 goto error;
269 }
270
271 ret = chmod(health_unix_sock_path,
272 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
273 if (ret < 0) {
274 ERR("Unable to set permissions on %s", health_unix_sock_path);
275 PERROR("chmod");
276 ret = -1;
277 goto error;
278 }
279 }
280
281 /*
282 * Set the CLOEXEC flag. Return code is useless because either way, the
283 * show must go on.
284 */
285 (void) utils_set_fd_cloexec(sock);
286
287 ret = lttcomm_listen_unix_sock(sock);
288 if (ret < 0) {
289 goto error;
290 }
291
292 /* Size is set to 1 for the consumer_channel pipe */
293 ret = lttng_poll_create(&events, 2, LTTNG_CLOEXEC);
294 if (ret < 0) {
295 ERR("Poll set creation failed");
296 goto error;
297 }
298
299 ret = lttng_poll_add(&events, health_quit_pipe[0], LPOLLIN);
300 if (ret < 0) {
301 goto error;
302 }
303
304 /* Add the application registration socket */
305 ret = lttng_poll_add(&events, sock, LPOLLIN | LPOLLPRI);
306 if (ret < 0) {
307 goto error;
308 }
309
3fd27398
MD
310 lttng_relay_notify_ready();
311
65931c8b
MD
312 while (1) {
313 DBG("Health check ready");
314
315 /* Inifinite blocking call, waiting for transmission */
316restart:
317 ret = lttng_poll_wait(&events, -1);
318 if (ret < 0) {
319 /*
320 * Restart interrupted system call.
321 */
322 if (errno == EINTR) {
323 goto restart;
324 }
325 goto error;
326 }
327
328 nb_fd = ret;
329
330 for (i = 0; i < nb_fd; i++) {
331 /* Fetch once the poll data */
332 revents = LTTNG_POLL_GETEV(&events, i);
333 pollfd = LTTNG_POLL_GETFD(&events, i);
334
fd20dac9
MD
335 if (!revents) {
336 /* No activity for this FD (poll implementation). */
337 continue;
338 }
339
65931c8b
MD
340 /* Thread quit pipe has been closed. Killing thread. */
341 ret = check_health_quit_pipe(pollfd, revents);
342 if (ret) {
343 err = 0;
344 goto exit;
345 }
346
347 /* Event on the registration socket */
348 if (pollfd == sock) {
349 if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
350 ERR("Health socket poll error");
351 goto error;
352 }
353 }
354 }
355
356 new_sock = lttcomm_accept_unix_sock(sock);
357 if (new_sock < 0) {
358 goto error;
359 }
360
361 /*
362 * Set the CLOEXEC flag. Return code is useless because either way, the
363 * show must go on.
364 */
365 (void) utils_set_fd_cloexec(new_sock);
366
367 DBG("Receiving data from client for health...");
368 ret = lttcomm_recv_unix_sock(new_sock, (void *)&msg, sizeof(msg));
369 if (ret <= 0) {
370 DBG("Nothing recv() from client... continuing");
371 ret = close(new_sock);
372 if (ret) {
373 PERROR("close");
374 }
375 new_sock = -1;
376 continue;
377 }
378
379 rcu_thread_online();
380
381 assert(msg.cmd == HEALTH_CMD_CHECK);
382
53efb85a 383 memset(&reply, 0, sizeof(reply));
65931c8b
MD
384 for (i = 0; i < NR_HEALTH_RELAYD_TYPES; i++) {
385 /*
386 * health_check_state return 0 if thread is in
387 * error.
388 */
389 if (!health_check_state(health_relayd, i)) {
390 reply.ret_code |= 1ULL << i;
391 }
392 }
393
394 DBG2("Health check return value %" PRIx64, reply.ret_code);
395
396 ret = send_unix_sock(new_sock, (void *) &reply, sizeof(reply));
397 if (ret < 0) {
398 ERR("Failed to send health data back to client");
399 }
400
401 /* End of transmission */
402 ret = close(new_sock);
403 if (ret) {
404 PERROR("close");
405 }
406 new_sock = -1;
407 }
408
409exit:
410error:
411 if (err) {
412 ERR("Health error occurred in %s", __func__);
413 }
414 DBG("Health check thread dying");
415 unlink(health_unix_sock_path);
416 if (sock >= 0) {
417 ret = close(sock);
418 if (ret) {
419 PERROR("close");
420 }
421 }
422
dcbcae3e
MD
423 /*
424 * We do NOT rmdir rundir nor the relayd path because there are
425 * other processes using them.
426 */
427
65931c8b
MD
428 lttng_poll_clean(&events);
429
430 rcu_unregister_thread();
431 return NULL;
432}
This page took 0.04476 seconds and 4 git commands to generate.