2 * Copyright (C) 2013 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
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.
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.
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.
29 #include <sys/resource.h>
31 #include <sys/socket.h>
33 #include <sys/types.h>
34 #include <urcu/list.h>
39 #include <urcu/compiler.h>
42 #include <common/defaults.h>
43 #include <common/common.h>
44 #include <common/consumer/consumer.h>
45 #include <common/consumer/consumer-timer.h>
46 #include <common/compat/poll.h>
47 #include <common/sessiond-comm/sessiond-comm.h>
48 #include <common/utils.h>
49 #include <common/compat/getenv.h>
50 #include <common/fd-tracker/utils.h>
52 #include "lttng-relayd.h"
53 #include "health-relayd.h"
55 /* Global health check unix path */
57 char health_unix_sock_path
[PATH_MAX
];
59 int health_quit_pipe
[2];
62 * Check if the thread quit pipe was triggered.
64 * Return 1 if it was triggered else 0;
67 int check_health_quit_pipe(int fd
, uint32_t events
)
69 if (fd
== health_quit_pipe
[0] && (events
& LPOLLIN
)) {
77 * Send data on a unix socket using the liblttsessiondcomm API.
79 * Return lttcomm error code.
81 static int send_unix_sock(int sock
, void *buf
, size_t len
)
83 /* Check valid length */
88 return lttcomm_send_unix_sock(sock
, buf
, len
);
91 static int create_lttng_rundir_with_perm(const char *rundir
)
95 DBG3("Creating LTTng run directory: %s", rundir
);
97 ret
= mkdir(rundir
, S_IRWXU
);
99 if (errno
!= EEXIST
) {
100 ERR("Unable to create %s", rundir
);
105 } else if (ret
== 0) {
106 int is_root
= !getuid();
109 ret
= chown(rundir
, 0,
110 utils_get_group_id(tracing_group_name
));
112 ERR("Unable to set group on %s", rundir
);
119 S_IRUSR
| S_IWUSR
| S_IXUSR
| S_IRGRP
| S_IXGRP
| S_IROTH
| S_IXOTH
);
121 ERR("Unable to set permissions on %s", health_unix_sock_path
);
134 int parse_health_env(void)
136 const char *health_path
;
138 health_path
= lttng_secure_getenv(LTTNG_RELAYD_HEALTH_ENV
);
140 strncpy(health_unix_sock_path
, health_path
,
142 health_unix_sock_path
[PATH_MAX
- 1] = '\0';
149 int setup_health_path(void)
151 int is_root
, ret
= 0;
152 char *home_path
= NULL
, *rundir
= NULL
, *relayd_path
= NULL
;
154 ret
= parse_health_env();
162 rundir
= strdup(DEFAULT_LTTNG_RUNDIR
);
169 * Create rundir from home path. This will create something like
172 home_path
= utils_get_home_dir();
174 if (home_path
== NULL
) {
175 /* TODO: Add --socket PATH option */
176 ERR("Can't get HOME directory for sockets creation.");
181 ret
= asprintf(&rundir
, DEFAULT_LTTNG_HOME_RUNDIR
, home_path
);
188 ret
= asprintf(&relayd_path
, DEFAULT_RELAYD_PATH
, rundir
);
194 ret
= create_lttng_rundir_with_perm(rundir
);
199 ret
= create_lttng_rundir_with_perm(relayd_path
);
205 if (strlen(health_unix_sock_path
) != 0) {
208 snprintf(health_unix_sock_path
, sizeof(health_unix_sock_path
),
209 DEFAULT_GLOBAL_RELAY_HEALTH_UNIX_SOCK
,
212 /* Set health check Unix path */
213 if (strlen(health_unix_sock_path
) != 0) {
217 snprintf(health_unix_sock_path
, sizeof(health_unix_sock_path
),
218 DEFAULT_HOME_RELAY_HEALTH_UNIX_SOCK
,
219 home_path
, (int) getpid());
229 int open_unix_socket(void *data
, int *out_fd
)
232 const char *path
= data
;
234 ret
= lttcomm_create_unix_sock(path
);
246 * Thread managing health check socket.
248 void *thread_manage_health(void *data
)
250 int sock
= -1, new_sock
= -1, ret
, i
, pollfd
, err
= -1;
251 uint32_t revents
, nb_fd
;
252 struct lttng_poll_event events
;
253 struct health_comm_msg msg
;
254 struct health_comm_reply reply
;
258 DBG("[thread] Manage health check started");
262 rcu_register_thread();
264 /* We might hit an error path before this is created. */
265 lttng_poll_init(&events
);
267 /* Create unix socket */
268 ret
= asprintf(&sock_name
, "Unix socket @ %s", health_unix_sock_path
);
270 PERROR("Failed to allocate unix socket name");
274 ret
= fd_tracker_open_unsuspendable_fd(the_fd_tracker
, &sock
,
275 (const char **) &sock_name
, 1, open_unix_socket
,
276 health_unix_sock_path
);
279 ERR("Unable to create health check Unix socket");
286 /* lttng health client socket path permissions */
287 ret
= chown(health_unix_sock_path
, 0,
288 utils_get_group_id(tracing_group_name
));
290 ERR("Unable to set group on %s", health_unix_sock_path
);
296 ret
= chmod(health_unix_sock_path
,
297 S_IRUSR
| S_IWUSR
| S_IRGRP
| S_IWGRP
);
299 ERR("Unable to set permissions on %s", health_unix_sock_path
);
307 * Set the CLOEXEC flag. Return code is useless because either way, the
310 (void) utils_set_fd_cloexec(sock
);
312 ret
= lttcomm_listen_unix_sock(sock
);
317 /* Size is set to 1 for the consumer_channel pipe */
318 ret
= lttng_poll_create(&events
, 2, LTTNG_CLOEXEC
);
320 ERR("Poll set creation failed");
324 ret
= lttng_poll_add(&events
, health_quit_pipe
[0], LPOLLIN
);
329 /* Add the application registration socket */
330 ret
= lttng_poll_add(&events
, sock
, LPOLLIN
| LPOLLPRI
);
335 lttng_relay_notify_ready();
338 DBG("Health check ready");
340 /* Inifinite blocking call, waiting for transmission */
342 ret
= lttng_poll_wait(&events
, -1);
345 * Restart interrupted system call.
347 if (errno
== EINTR
) {
355 for (i
= 0; i
< nb_fd
; i
++) {
356 /* Fetch once the poll data */
357 revents
= LTTNG_POLL_GETEV(&events
, i
);
358 pollfd
= LTTNG_POLL_GETFD(&events
, i
);
361 /* No activity for this FD (poll implementation). */
365 /* Thread quit pipe has been closed. Killing thread. */
366 ret
= check_health_quit_pipe(pollfd
, revents
);
372 /* Event on the registration socket */
373 if (pollfd
== sock
) {
374 if (revents
& LPOLLIN
) {
376 } else if (revents
& (LPOLLERR
| LPOLLHUP
| LPOLLRDHUP
)) {
377 ERR("Health socket poll error");
380 ERR("Unexpected poll events %u for sock %d", revents
, pollfd
);
386 new_sock
= lttcomm_accept_unix_sock(sock
);
392 * Set the CLOEXEC flag. Return code is useless because either way, the
395 (void) utils_set_fd_cloexec(new_sock
);
397 DBG("Receiving data from client for health...");
398 ret
= lttcomm_recv_unix_sock(new_sock
, (void *)&msg
, sizeof(msg
));
400 DBG("Nothing recv() from client... continuing");
401 ret
= close(new_sock
);
411 assert(msg
.cmd
== HEALTH_CMD_CHECK
);
413 memset(&reply
, 0, sizeof(reply
));
414 for (i
= 0; i
< NR_HEALTH_RELAYD_TYPES
; i
++) {
416 * health_check_state return 0 if thread is in
419 if (!health_check_state(health_relayd
, i
)) {
420 reply
.ret_code
|= 1ULL << i
;
424 DBG2("Health check return value %" PRIx64
, reply
.ret_code
);
426 ret
= send_unix_sock(new_sock
, (void *) &reply
, sizeof(reply
));
428 ERR("Failed to send health data back to client");
431 /* End of transmission */
432 ret
= close(new_sock
);
440 lttng_relay_stop_threads();
443 ERR("Health error occurred in %s", __func__
);
445 DBG("Health check thread dying");
446 unlink(health_unix_sock_path
);
448 ret
= fd_tracker_close_unsuspendable_fd(the_fd_tracker
, &sock
,
449 1, fd_tracker_util_close_fd
, NULL
);
456 * We do NOT rmdir rundir nor the relayd path because there are
457 * other processes using them.
460 lttng_poll_clean(&events
);
462 rcu_unregister_thread();