[lttng-tools.git] / src / bin / lttng-relayd / health-relayd.c

/*
 * Copyright (C) 2013 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License, version 2 only,
 * as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License along
 * with this program; if not, write to the Free Software Foundation, Inc.,
 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
 */

#define _LGPL_SOURCE
#include <fcntl.h>
#include <getopt.h>
#include <grp.h>
#include <limits.h>
#include <pthread.h>
#include <signal.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/ipc.h>
#include <sys/resource.h>
#include <sys/shm.h>
#include <sys/socket.h>
#include <sys/stat.h>
#include <sys/types.h>
#include <urcu/list.h>
#include <poll.h>
#include <unistd.h>
#include <sys/mman.h>
#include <assert.h>
#include <urcu/compiler.h>
#include <inttypes.h>

#include <common/defaults.h>
#include <common/common.h>
#include <common/consumer/consumer.h>
#include <common/consumer/consumer-timer.h>
#include <common/compat/poll.h>
#include <common/sessiond-comm/sessiond-comm.h>
#include <common/utils.h>
#include <common/compat/getenv.h>
#include <common/fd-tracker/utils.h>

#include "lttng-relayd.h"
#include "health-relayd.h"

/* Global health check unix path */
static
char health_unix_sock_path[PATH_MAX];

int health_quit_pipe[2] = { -1, -1 };

/*
 * Check if the thread quit pipe was triggered.
 *
 * Return 1 if it was triggered else 0;
 */
static
int check_health_quit_pipe(int fd, uint32_t events)
{
	if (fd == health_quit_pipe[0] && (events & LPOLLIN)) {
		return 1;
	}

	return 0;
}

/*
 * Send data on a unix socket using the liblttsessiondcomm API.
 *
 * Return lttcomm error code.
 */
static int send_unix_sock(int sock, void *buf, size_t len)
{
	/* Check valid length */
	if (len == 0) {
		return -1;
	}

	return lttcomm_send_unix_sock(sock, buf, len);
}

static int create_lttng_rundir_with_perm(const char *rundir)
{
	int ret;

	DBG3("Creating LTTng run directory: %s", rundir);

	ret = mkdir(rundir, S_IRWXU);
	if (ret < 0) {
		if (errno != EEXIST) {
			ERR("Unable to create %s", rundir);
			goto error;
		} else {
			ret = 0;
		}
	} else if (ret == 0) {
		int is_root = !getuid();

		if (is_root) {
			gid_t gid;

			ret = utils_get_group_id(tracing_group_name, true, &gid);
			if (ret) {
				/* Default to root group. */
				gid = 0;
			}

			ret = chown(rundir, 0, gid);
			if (ret < 0) {
				ERR("Unable to set group on %s", rundir);
				PERROR("chown");
				ret = -1;
				goto error;
			}

			ret = chmod(rundir,
					S_IRUSR | S_IWUSR | S_IXUSR | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH);
			if (ret < 0) {
				ERR("Unable to set permissions on %s", health_unix_sock_path);
				PERROR("chmod");
				ret = -1;
				goto error;
			}
		}
	}

error:
	return ret;
}

static
int parse_health_env(void)
{
	const char *health_path;

	health_path = lttng_secure_getenv(LTTNG_RELAYD_HEALTH_ENV);
	if (health_path) {
		strncpy(health_unix_sock_path, health_path,
			PATH_MAX);
		health_unix_sock_path[PATH_MAX - 1] = '\0';
	}

	return 0;
}

static
int setup_health_path(void)
{
	int is_root, ret = 0;
	const char *home_path = NULL;
	char *rundir = NULL, *relayd_path = NULL;

	ret = parse_health_env();
	if (ret) {
		return ret;
	}

	is_root = !getuid();

	if (is_root) {
		rundir = strdup(DEFAULT_LTTNG_RUNDIR);
		if (!rundir) {
			ret = -ENOMEM;
			goto end;
		}
	} else {
		/*
		 * Create rundir from home path. This will create something like
		 * $HOME/.lttng
		 */
		home_path = utils_get_home_dir();

		if (home_path == NULL) {
			/* TODO: Add --socket PATH option */
			ERR("Can't get HOME directory for sockets creation.");
			ret = -EPERM;
			goto end;
		}

		ret = asprintf(&rundir, DEFAULT_LTTNG_HOME_RUNDIR, home_path);
		if (ret < 0) {
			ret = -ENOMEM;
			goto end;
		}
	}

	ret = asprintf(&relayd_path, DEFAULT_RELAYD_PATH, rundir);
	if (ret < 0) {
		ret = -ENOMEM;
		goto end;
	}

	ret = create_lttng_rundir_with_perm(rundir);
	if (ret < 0) {
		goto end;
	}

	ret = create_lttng_rundir_with_perm(relayd_path);
	if (ret < 0) {
		goto end;
	}

	if (is_root) {
		if (strlen(health_unix_sock_path) != 0) {
			goto end;
		}
		snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
			DEFAULT_GLOBAL_RELAY_HEALTH_UNIX_SOCK,
			(int) getpid());
	} else {
		/* Set health check Unix path */
		if (strlen(health_unix_sock_path) != 0) {
			goto end;
		}

		snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
			DEFAULT_HOME_RELAY_HEALTH_UNIX_SOCK,
			home_path, (int) getpid());
	}

end:
	free(rundir);
	free(relayd_path);
	return ret;
}

static
int accept_unix_socket(void *data, int *out_fd)
{
	int ret;
	int accepting_sock = *((int *) data);

	ret = lttcomm_accept_unix_sock(accepting_sock);
	if (ret < 0) {
		goto end;
	}

	*out_fd = ret;
	ret = 0;
end:
	return ret;
}

static
int open_unix_socket(void *data, int *out_fd)
{
	int ret;
	const char *path = data;

	ret = lttcomm_create_unix_sock(path);
	if (ret < 0) {
		goto end;
	}

	*out_fd = ret;
	ret = 0;
end:
	return ret;
}

/*
 * Thread managing health check socket.
 */
void *thread_manage_health(void *data)
{
	int sock = -1, new_sock = -1, ret, i, pollfd, err = -1;
	uint32_t revents, nb_fd;
	struct lttng_poll_event events;
	struct health_comm_msg msg;
	struct health_comm_reply reply;
	int is_root;
	char *sock_name;

	DBG("[thread] Manage health check started");

	setup_health_path();

	rcu_register_thread();

	/* We might hit an error path before this is created. */
	lttng_poll_init(&events);

	/* Create unix socket */
	ret = asprintf(&sock_name, "Unix socket @ %s", health_unix_sock_path);
	if (ret == -1) {
		PERROR("Failed to allocate unix socket name");
		err = -1;
		goto error;
	}
	ret = fd_tracker_open_unsuspendable_fd(the_fd_tracker, &sock,
			(const char **) &sock_name, 1, open_unix_socket,
			health_unix_sock_path);
	free(sock_name);
	if (ret < 0) {
		ERR("Unable to create health check Unix socket");
		err = -1;
		goto error;
	}

	is_root = !getuid();
	if (is_root) {
		/* lttng health client socket path permissions */
		gid_t gid;

		ret = utils_get_group_id(tracing_group_name, true, &gid);
		if (ret) {
			/* Default to root group. */
			gid = 0;
		}

		ret = chown(health_unix_sock_path, 0, gid);
		if (ret < 0) {
			ERR("Unable to set group on %s", health_unix_sock_path);
			PERROR("chown");
			err = -1;
			goto error;
		}

		ret = chmod(health_unix_sock_path,
				S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
		if (ret < 0) {
			ERR("Unable to set permissions on %s", health_unix_sock_path);
			PERROR("chmod");
			err = -1;
			goto error;
		}
	}

	/*
	 * Set the CLOEXEC flag. Return code is useless because either way, the
	 * show must go on.
	 */
	(void) utils_set_fd_cloexec(sock);

	ret = lttcomm_listen_unix_sock(sock);
	if (ret < 0) {
		goto error;
	}

	/* Size is set to 2 for the unix socket and quit pipe. */
	ret = fd_tracker_util_poll_create(the_fd_tracker,
			"Health management thread epoll", &events, 2,
			LTTNG_CLOEXEC);
	if (ret < 0) {
		ERR("Poll set creation failed");
		goto error;
	}

	ret = lttng_poll_add(&events, health_quit_pipe[0], LPOLLIN);
	if (ret < 0) {
		goto error;
	}

	/* Add the application registration socket */
	ret = lttng_poll_add(&events, sock, LPOLLIN | LPOLLPRI);
	if (ret < 0) {
		goto error;
	}

	lttng_relay_notify_ready();

	while (1) {
		char *accepted_socket_name;

		DBG("Health check ready");

		/* Inifinite blocking call, waiting for transmission */
restart:
		ret = lttng_poll_wait(&events, -1);
		if (ret < 0) {
			/*
			 * Restart interrupted system call.
			 */
			if (errno == EINTR) {
				goto restart;
			}
			goto error;
		}

		nb_fd = ret;

		for (i = 0; i < nb_fd; i++) {
			/* Fetch once the poll data */
			revents = LTTNG_POLL_GETEV(&events, i);
			pollfd = LTTNG_POLL_GETFD(&events, i);

			/* Thread quit pipe has been closed. Killing thread. */
			ret = check_health_quit_pipe(pollfd, revents);
			if (ret) {
				err = 0;
				goto exit;
			}

			/* Event on the registration socket */
			if (pollfd == sock) {
				if (revents & LPOLLIN) {
					continue;
				} else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
					ERR("Health socket poll error");
					goto error;
				} else {
					ERR("Unexpected poll events %u for sock %d", revents, pollfd);
					goto error;
				}
			}
		}

		ret = asprintf(&accepted_socket_name, "Socket accepted from unix socket @ %s",
				health_unix_sock_path);
		if (ret == -1) {
			PERROR("Failed to allocate name of accepted socket from unix socket @ %s",
					health_unix_sock_path);
			goto error;
		}
		ret = fd_tracker_open_unsuspendable_fd(the_fd_tracker, &new_sock,
				(const char **) &accepted_socket_name, 1,
				accept_unix_socket, &sock);
		free(accepted_socket_name);
		if (ret < 0) {
			goto error;
		}

		/*
		 * Set the CLOEXEC flag. Return code is useless because either way, the
		 * show must go on.
		 */
		(void) utils_set_fd_cloexec(new_sock);

		DBG("Receiving data from client for health...");
		ret = lttcomm_recv_unix_sock(new_sock, (void *)&msg, sizeof(msg));
		if (ret <= 0) {
			DBG("Nothing recv() from client... continuing");
			ret = fd_tracker_close_unsuspendable_fd(the_fd_tracker,
					&new_sock, 1, fd_tracker_util_close_fd,
					NULL);
			if (ret) {
				PERROR("close");
			}
			new_sock = -1;
			continue;
		}

		rcu_thread_online();

		assert(msg.cmd == HEALTH_CMD_CHECK);

		memset(&reply, 0, sizeof(reply));
		for (i = 0; i < NR_HEALTH_RELAYD_TYPES; i++) {
			/*
			 * health_check_state return 0 if thread is in
			 * error.
			 */
			if (!health_check_state(health_relayd, i)) {
				reply.ret_code |= 1ULL << i;
			}
		}

		DBG2("Health check return value %" PRIx64, reply.ret_code);

		ret = send_unix_sock(new_sock, (void *) &reply, sizeof(reply));
		if (ret < 0) {
			ERR("Failed to send health data back to client");
		}

		/* End of transmission */
		ret = fd_tracker_close_unsuspendable_fd(the_fd_tracker,
				&new_sock, 1, fd_tracker_util_close_fd,
				NULL);
		if (ret) {
			PERROR("close");
		}
		new_sock = -1;
	}

error:
	lttng_relay_stop_threads();
exit:
	if (err) {
		ERR("Health error occurred in %s", __func__);
	}
	DBG("Health check thread dying");
	unlink(health_unix_sock_path);
	if (sock >= 0) {
		ret = fd_tracker_close_unsuspendable_fd(the_fd_tracker, &sock,
				1, fd_tracker_util_close_fd, NULL);
		if (ret) {
			PERROR("close");
		}
	}

	/*
	 * We do NOT rmdir rundir nor the relayd path because there are
	 * other processes using them.
	 */

	(void) fd_tracker_util_poll_clean(the_fd_tracker, &events);

	rcu_unregister_thread();
	return NULL;
}
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
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>
c8fea79c JR	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>
1c9bd75b	50	#include <common/fd-tracker/utils.h>
65931c8b MD	51
	52	#include "lttng-relayd.h"
	53	#include "health-relayd.h"
	54
	55	/* Global health check unix path */
094fe907 MD	56	static
094fe907 MD	57	char health_unix_sock_path[PATH_MAX];
65931c8b	58
794e2e5f	59	int health_quit_pipe[2] = { -1, -1 };
65931c8b MD	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) {
28ab59d0 JR	109	gid_t gid;
	110
	111	ret = utils_get_group_id(tracing_group_name, true, &gid);
	112	if (ret) {
	113	/* Default to root group. */
	114	gid = 0;
	115	}
	116
	117	ret = chown(rundir, 0, gid);
65931c8b MD	118	if (ret < 0) {
	119	ERR("Unable to set group on %s", rundir);
	120	PERROR("chown");
	121	ret = -1;
	122	goto error;
	123	}
	124
	125	ret = chmod(rundir,
	126	S_IRUSR \| S_IWUSR \| S_IXUSR \| S_IRGRP \| S_IXGRP \| S_IROTH \| S_IXOTH);
	127	if (ret < 0) {
	128	ERR("Unable to set permissions on %s", health_unix_sock_path);
	129	PERROR("chmod");
	130	ret = -1;
	131	goto error;
	132	}
	133	}
	134	}
	135
	136	error:
	137	return ret;
	138	}
	139
094fe907 MD	140	static
	141	int parse_health_env(void)
	142	{
	143	const char *health_path;
	144
e8fa9fb0	145	health_path = lttng_secure_getenv(LTTNG_RELAYD_HEALTH_ENV);
094fe907 MD	146	if (health_path) {
	147	strncpy(health_unix_sock_path, health_path,
	148	PATH_MAX);
	149	health_unix_sock_path[PATH_MAX - 1] = '\0';
	150	}
	151
	152	return 0;
	153	}
	154
65931c8b MD	155	static
	156	int setup_health_path(void)
	157	{
	158	int is_root, ret = 0;
4f00620d JG	159	const char *home_path = NULL;
4f00620d JG	160	char rundir = NULL, relayd_path = NULL;
65931c8b	161
094fe907 MD	162	ret = parse_health_env();
	163	if (ret) {
	164	return ret;
	165	}
	166
65931c8b MD	167	is_root = !getuid();
	168
	169	if (is_root) {
c3844e39	170	rundir = strdup(DEFAULT_LTTNG_RUNDIR);
b6ab01aa MD	171	if (!rundir) {
	172	ret = -ENOMEM;
	173	goto end;
	174	}
65931c8b MD	175	} else {
	176	/*
	177	* Create rundir from home path. This will create something like
	178	* $HOME/.lttng
	179	*/
	180	home_path = utils_get_home_dir();
	181
	182	if (home_path == NULL) {
	183	/* TODO: Add --socket PATH option */
	184	ERR("Can't get HOME directory for sockets creation.");
	185	ret = -EPERM;
	186	goto end;
	187	}
	188
c3844e39	189	ret = asprintf(&rundir, DEFAULT_LTTNG_HOME_RUNDIR, home_path);
65931c8b MD	190	if (ret < 0) {
	191	ret = -ENOMEM;
	192	goto end;
	193	}
	194	}
	195
c3844e39	196	ret = asprintf(&relayd_path, DEFAULT_RELAYD_PATH, rundir);
65931c8b MD	197	if (ret < 0) {
	198	ret = -ENOMEM;
	199	goto end;
	200	}
	201
c3844e39	202	ret = create_lttng_rundir_with_perm(rundir);
65931c8b MD	203	if (ret < 0) {
	204	goto end;
	205	}
	206
	207	ret = create_lttng_rundir_with_perm(relayd_path);
	208	if (ret < 0) {
	209	goto end;
	210	}
	211
	212	if (is_root) {
	213	if (strlen(health_unix_sock_path) != 0) {
	214	goto end;
	215	}
	216	snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
	217	DEFAULT_GLOBAL_RELAY_HEALTH_UNIX_SOCK,
d1f721c5	218	(int) getpid());
65931c8b MD	219	} else {
	220	/* Set health check Unix path */
	221	if (strlen(health_unix_sock_path) != 0) {
	222	goto end;
	223	}
	224
	225	snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
	226	DEFAULT_HOME_RELAY_HEALTH_UNIX_SOCK,
d1f721c5	227	home_path, (int) getpid());
65931c8b MD	228	}
	229
	230	end:
c3844e39	231	free(rundir);
edd94901	232	free(relayd_path);
65931c8b MD	233	return ret;
	234	}
	235
1c9bd75b JG	236	static
	237	int accept_unix_socket(void data, int out_fd)
	238	{
	239	int ret;
	240	int accepting_sock = ((int ) data);
	241
	242	ret = lttcomm_accept_unix_sock(accepting_sock);
	243	if (ret < 0) {
	244	goto end;
	245	}
	246
	247	*out_fd = ret;
	248	ret = 0;
	249	end:
	250	return ret;
	251	}
	252
d7eddab9 JG	253	static
	254	int open_unix_socket(void data, int out_fd)
	255	{
	256	int ret;
	257	const char *path = data;
	258
	259	ret = lttcomm_create_unix_sock(path);
	260	if (ret < 0) {
	261	goto end;
	262	}
	263
	264	*out_fd = ret;
	265	ret = 0;
	266	end:
	267	return ret;
	268	}
	269
65931c8b MD	270	/*
	271	* Thread managing health check socket.
	272	*/
	273	void thread_manage_health(void data)
	274	{
	275	int sock = -1, new_sock = -1, ret, i, pollfd, err = -1;
	276	uint32_t revents, nb_fd;
	277	struct lttng_poll_event events;
	278	struct health_comm_msg msg;
	279	struct health_comm_reply reply;
	280	int is_root;
d7eddab9	281	char *sock_name;
65931c8b MD	282
	283	DBG("[thread] Manage health check started");
	284
	285	setup_health_path();
	286
	287	rcu_register_thread();
	288
	289	/* We might hit an error path before this is created. */
	290	lttng_poll_init(&events);
	291
	292	/* Create unix socket */
d7eddab9 JG	293	ret = asprintf(&sock_name, "Unix socket @ %s", health_unix_sock_path);
	294	if (ret == -1) {
	295	PERROR("Failed to allocate unix socket name");
	296	err = -1;
	297	goto error;
	298	}
	299	ret = fd_tracker_open_unsuspendable_fd(the_fd_tracker, &sock,
	300	(const char **) &sock_name, 1, open_unix_socket,
	301	health_unix_sock_path);
	302	free(sock_name);
	303	if (ret < 0) {
65931c8b	304	ERR("Unable to create health check Unix socket");
7568ddbf	305	err = -1;
65931c8b MD	306	goto error;
	307	}
	308
	309	is_root = !getuid();
	310	if (is_root) {
	311	/* lttng health client socket path permissions */
28ab59d0 JR	312	gid_t gid;
	313
	314	ret = utils_get_group_id(tracing_group_name, true, &gid);
	315	if (ret) {
	316	/* Default to root group. */
	317	gid = 0;
	318	}
	319
	320	ret = chown(health_unix_sock_path, 0, gid);
65931c8b MD	321	if (ret < 0) {
	322	ERR("Unable to set group on %s", health_unix_sock_path);
	323	PERROR("chown");
7568ddbf	324	err = -1;
65931c8b MD	325	goto error;
	326	}
	327
	328	ret = chmod(health_unix_sock_path,
	329	S_IRUSR \| S_IWUSR \| S_IRGRP \| S_IWGRP);
	330	if (ret < 0) {
	331	ERR("Unable to set permissions on %s", health_unix_sock_path);
	332	PERROR("chmod");
7568ddbf	333	err = -1;
65931c8b MD	334	goto error;
	335	}
	336	}
	337
	338	/*
	339	* Set the CLOEXEC flag. Return code is useless because either way, the
	340	* show must go on.
	341	*/
	342	(void) utils_set_fd_cloexec(sock);
	343
	344	ret = lttcomm_listen_unix_sock(sock);
	345	if (ret < 0) {
	346	goto error;
	347	}
	348
aa91fbc5 JG	349	/* Size is set to 2 for the unix socket and quit pipe. */
	350	ret = fd_tracker_util_poll_create(the_fd_tracker,
	351	"Health management thread epoll", &events, 2,
	352	LTTNG_CLOEXEC);
65931c8b MD	353	if (ret < 0) {
	354	ERR("Poll set creation failed");
	355	goto error;
	356	}
	357
	358	ret = lttng_poll_add(&events, health_quit_pipe[0], LPOLLIN);
	359	if (ret < 0) {
	360	goto error;
	361	}
	362
	363	/* Add the application registration socket */
	364	ret = lttng_poll_add(&events, sock, LPOLLIN \| LPOLLPRI);
	365	if (ret < 0) {
	366	goto error;
	367	}
	368
3fd27398 MD	369	lttng_relay_notify_ready();
3fd27398 MD	370
65931c8b	371	while (1) {
1c9bd75b JG	372	char *accepted_socket_name;
1c9bd75b JG	373
65931c8b MD	374	DBG("Health check ready");
	375
	376	/* Inifinite blocking call, waiting for transmission */
	377	restart:
	378	ret = lttng_poll_wait(&events, -1);
	379	if (ret < 0) {
	380	/*
	381	* Restart interrupted system call.
	382	*/
	383	if (errno == EINTR) {
	384	goto restart;
	385	}
	386	goto error;
	387	}
	388
	389	nb_fd = ret;
	390
	391	for (i = 0; i < nb_fd; i++) {
	392	/* Fetch once the poll data */
	393	revents = LTTNG_POLL_GETEV(&events, i);
	394	pollfd = LTTNG_POLL_GETFD(&events, i);
	395
	396	/* Thread quit pipe has been closed. Killing thread. */
	397	ret = check_health_quit_pipe(pollfd, revents);
	398	if (ret) {
	399	err = 0;
	400	goto exit;
	401	}
	402
	403	/* Event on the registration socket */
	404	if (pollfd == sock) {
03e43155 MD	405	if (revents & LPOLLIN) {
	406	continue;
	407	} else if (revents & (LPOLLERR \| LPOLLHUP \| LPOLLRDHUP)) {
65931c8b MD	408	ERR("Health socket poll error");
65931c8b MD	409	goto error;
03e43155 MD	410	} else {
	411	ERR("Unexpected poll events %u for sock %d", revents, pollfd);
	412	goto error;
65931c8b MD	413	}
	414	}
	415	}
	416
1c9bd75b JG	417	ret = asprintf(&accepted_socket_name, "Socket accepted from unix socket @ %s",
	418	health_unix_sock_path);
	419	if (ret == -1) {
	420	PERROR("Failed to allocate name of accepted socket from unix socket @ %s",
	421	health_unix_sock_path);
	422	goto error;
	423	}
	424	ret = fd_tracker_open_unsuspendable_fd(the_fd_tracker, &new_sock,
	425	(const char **) &accepted_socket_name, 1,
	426	accept_unix_socket, &sock);
	427	free(accepted_socket_name);
	428	if (ret < 0) {
65931c8b MD	429	goto error;
	430	}
	431
	432	/*
	433	* Set the CLOEXEC flag. Return code is useless because either way, the
	434	* show must go on.
	435	*/
	436	(void) utils_set_fd_cloexec(new_sock);
	437
	438	DBG("Receiving data from client for health...");
	439	ret = lttcomm_recv_unix_sock(new_sock, (void *)&msg, sizeof(msg));
	440	if (ret <= 0) {
	441	DBG("Nothing recv() from client... continuing");
1c9bd75b JG	442	ret = fd_tracker_close_unsuspendable_fd(the_fd_tracker,
	443	&new_sock, 1, fd_tracker_util_close_fd,
	444	NULL);
65931c8b MD	445	if (ret) {
	446	PERROR("close");
	447	}
	448	new_sock = -1;
	449	continue;
	450	}
	451
	452	rcu_thread_online();
	453
	454	assert(msg.cmd == HEALTH_CMD_CHECK);
	455
53efb85a	456	memset(&reply, 0, sizeof(reply));
65931c8b MD	457	for (i = 0; i < NR_HEALTH_RELAYD_TYPES; i++) {
	458	/*
	459	* health_check_state return 0 if thread is in
	460	* error.
	461	*/
	462	if (!health_check_state(health_relayd, i)) {
	463	reply.ret_code \|= 1ULL << i;
	464	}
	465	}
	466
	467	DBG2("Health check return value %" PRIx64, reply.ret_code);
	468
	469	ret = send_unix_sock(new_sock, (void *) &reply, sizeof(reply));
	470	if (ret < 0) {
	471	ERR("Failed to send health data back to client");
	472	}
	473
	474	/* End of transmission */
1c9bd75b JG	475	ret = fd_tracker_close_unsuspendable_fd(the_fd_tracker,
	476	&new_sock, 1, fd_tracker_util_close_fd,
	477	NULL);
65931c8b MD	478	if (ret) {
	479	PERROR("close");
	480	}
	481	new_sock = -1;
	482	}
	483
65931c8b	484	error:
81714439 JG	485	lttng_relay_stop_threads();
81714439 JG	486	exit:
65931c8b MD	487	if (err) {
	488	ERR("Health error occurred in %s", __func__);
	489	}
	490	DBG("Health check thread dying");
	491	unlink(health_unix_sock_path);
	492	if (sock >= 0) {
d7eddab9 JG	493	ret = fd_tracker_close_unsuspendable_fd(the_fd_tracker, &sock,
d7eddab9 JG	494	1, fd_tracker_util_close_fd, NULL);
65931c8b MD	495	if (ret) {
	496	PERROR("close");
	497	}
	498	}
	499
dcbcae3e MD	500	/*
	501	* We do NOT rmdir rundir nor the relayd path because there are
	502	* other processes using them.
	503	*/
	504
aa91fbc5	505	(void) fd_tracker_util_poll_clean(the_fd_tracker, &events);
65931c8b MD	506
	507	rcu_unregister_thread();
	508	return NULL;
	509	}