Backport: relayd: track the health unix socket with the fd-tracker

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

diff --git a/src/bin/lttng-relayd/health-relayd.c b/src/bin/lttng-relayd/health-relayd.c
index 01e54d2eae795a35e87d94dd6f06b51ef96be3ac..9b747dd67b7eefb40e2544430f9eef29404d326a 100644 (file)
--- a/src/bin/lttng-relayd/health-relayd.c
+++ b/src/bin/lttng-relayd/health-relayd.c
@@ -15,7 +15,7 @@
  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
  */
 
-#define _GNU_SOURCE
+#define _LGPL_SOURCE
 #include <fcntl.h>
 #include <getopt.h>
 #include <grp.h>
@@ -36,18 +36,18 @@
 #include <unistd.h>
 #include <sys/mman.h>
 #include <assert.h>
-#include <config.h>
 #include <urcu/compiler.h>
-#include <ulimit.h>
 #include <inttypes.h>
 
 #include <common/defaults.h>
 #include <common/common.h>
-#include <common/consumer.h>
-#include <common/consumer-timer.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"
@@ -135,7 +135,7 @@ int parse_health_env(void)
 {
        const char *health_path;
 
-       health_path = getenv(LTTNG_RELAYD_HEALTH_ENV);
+       health_path = lttng_secure_getenv(LTTNG_RELAYD_HEALTH_ENV);
        if (health_path) {
                strncpy(health_unix_sock_path, health_path,
                        PATH_MAX);
@@ -149,7 +149,7 @@ static
 int setup_health_path(void)
 {
        int is_root, ret = 0;
-       char *home_path = NULL, *rundir = NULL, *relayd_path;
+       char *home_path = NULL, *rundir = NULL, *relayd_path = NULL;
 
        ret = parse_health_env();
        if (ret) {
@@ -160,6 +160,10 @@ int setup_health_path(void)
 
        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
@@ -203,7 +207,7 @@ int setup_health_path(void)
                }
                snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
                        DEFAULT_GLOBAL_RELAY_HEALTH_UNIX_SOCK,
-                       getpid());
+                       (int) getpid());
        } else {
                /* Set health check Unix path */
                if (strlen(health_unix_sock_path) != 0) {
@@ -212,11 +216,29 @@ int setup_health_path(void)
 
                snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
                        DEFAULT_HOME_RELAY_HEALTH_UNIX_SOCK,
-                       home_path, getpid());
+                       home_path, (int) getpid());
        }
 
 end:
        free(rundir);
+       free(relayd_path);
+       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;
 }
 
@@ -231,6 +253,7 @@ void *thread_manage_health(void *data)
        struct health_comm_msg msg;
        struct health_comm_reply reply;
        int is_root;
+       char *sock_name;
 
        DBG("[thread] Manage health check started");
 
@@ -242,10 +265,19 @@ void *thread_manage_health(void *data)
        lttng_poll_init(&events);
 
        /* Create unix socket */
-       sock = lttcomm_create_unix_sock(health_unix_sock_path);
-       if (sock < 0) {
+       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");
-               ret = -1;
+               err = -1;
                goto error;
        }
 
@@ -257,7 +289,7 @@ void *thread_manage_health(void *data)
                if (ret < 0) {
                        ERR("Unable to set group on %s", health_unix_sock_path);
                        PERROR("chown");
-                       ret = -1;
+                       err = -1;
                        goto error;
                }
 
@@ -266,7 +298,7 @@ void *thread_manage_health(void *data)
                if (ret < 0) {
                        ERR("Unable to set permissions on %s", health_unix_sock_path);
                        PERROR("chmod");
-                       ret = -1;
+                       err = -1;
                        goto error;
                }
        }
@@ -300,6 +332,8 @@ void *thread_manage_health(void *data)
                goto error;
        }
 
+       lttng_relay_notify_ready();
+
        while (1) {
                DBG("Health check ready");
 
@@ -323,6 +357,11 @@ restart:
                        revents = LTTNG_POLL_GETEV(&events, i);
                        pollfd = LTTNG_POLL_GETFD(&events, i);
 
+                       if (!revents) {
+                               /* No activity for this FD (poll implementation). */
+                               continue;
+                       }
+
                        /* Thread quit pipe has been closed. Killing thread. */
                        ret = check_health_quit_pipe(pollfd, revents);
                        if (ret) {
@@ -332,9 +371,14 @@ restart:
 
                        /* Event on the registration socket */
                        if (pollfd == sock) {
-                               if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
+                               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;
                                }
                        }
                }
@@ -366,7 +410,7 @@ restart:
 
                assert(msg.cmd == HEALTH_CMD_CHECK);
 
-               reply.ret_code = 0;
+               memset(&reply, 0, sizeof(reply));
                for (i = 0; i < NR_HEALTH_RELAYD_TYPES; i++) {
                        /*
                         * health_check_state return 0 if thread is in
@@ -392,15 +436,17 @@ restart:
                new_sock = -1;
        }
 
-exit:
 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 = close(sock);
+               ret = fd_tracker_close_unsuspendable_fd(the_fd_tracker, &sock,
+                               1, fd_tracker_util_close_fd, NULL);
                if (ret) {
                        PERROR("close");
                }