src/bin/lttng-sessiond/health.c

   1 /*
   2  * Copyright (C) 2012 - David Goulet <dgoulet@efficios.com>
   3  * Copyright (C) 2018 - Jérémie Galarneau <jeremie.galarneau@efficios.com>
   4  *
   5  * This program is free software; you can redistribute it and/or modify it
   6  * under the terms of the GNU General Public License, version 2 only, as
   7  * published by the Free Software Foundation.
   8  *
   9  * This program is distributed in the hope that it will be useful, but WITHOUT
  10  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  11  * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
  12  * more details.
  13  *
  14  * You should have received a copy of the GNU General Public License along with
  15  * this program; if not, write to the Free Software Foundation, Inc., 51
  16  * Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
  17  */
  18
  19 #include "lttng-sessiond.h"
  20 #include "health-sessiond.h"
  21 #include <common/macros.h>
  22 #include <common/error.h>
  23 #include <common/utils.h>
  24 #include <common/pipe.h>
  25 #include <inttypes.h>
  26 #include <sys/stat.h>
  27 #include "utils.h"
  28 #include "thread.h"
  29
  30 struct thread_notifiers {
  31         struct lttng_pipe *quit_pipe;
  32         sem_t ready;
  33 };
  34
  35 static
  36 void mark_thread_as_ready(struct thread_notifiers *notifiers)
  37 {
  38         DBG("Marking health management thread as ready");
  39         sem_post(&notifiers->ready);
  40 }
  41
  42 static
  43 void wait_until_thread_is_ready(struct thread_notifiers *notifiers)
  44 {
  45         DBG("Waiting for health management thread to be ready");
  46         sem_wait(&notifiers->ready);
  47         DBG("Health management thread is ready");
  48 }
  49
  50 static void cleanup_health_management_thread(void *data)
  51 {
  52         struct thread_notifiers *notifiers = data;
  53
  54         lttng_pipe_destroy(notifiers->quit_pipe);
  55         sem_destroy(&notifiers->ready);
  56         free(notifiers);
  57 }
  58
  59 /*
  60  * Thread managing health check socket.
  61  */
  62 static void *thread_manage_health(void *data)
  63 {
  64         const bool is_root = (getuid() == 0);
  65         int sock = -1, new_sock = -1, ret, i, pollfd, err = -1;
  66         uint32_t revents, nb_fd;
  67         struct lttng_poll_event events;
  68         struct health_comm_msg msg;
  69         struct health_comm_reply reply;
  70         /* Thread-specific quit pipe. */
  71         struct thread_notifiers *notifiers = data;
  72         const int quit_pipe_read_fd = lttng_pipe_get_readfd(
  73                         notifiers->quit_pipe);
  74
  75         DBG("[thread] Manage health check started");
  76
  77         rcu_register_thread();
  78
  79         /*
  80          * Created with a size of two for:
  81          *   - client socket
  82          *   - thread quit pipe
  83          */
  84         ret = lttng_poll_create(&events, 2, LTTNG_CLOEXEC);
  85         if (ret < 0) {
  86                 goto error;
  87         }
  88
  89         /* Create unix socket */
  90         sock = lttcomm_create_unix_sock(config.health_unix_sock_path.value);
  91         if (sock < 0) {
  92                 ERR("Unable to create health check Unix socket");
  93                 goto error;
  94         }
  95
  96         if (is_root) {
  97                 /* lttng health client socket path permissions */
  98                 ret = chown(config.health_unix_sock_path.value, 0,
  99                                 utils_get_group_id(config.tracing_group_name.value));
 100                 if (ret < 0) {
 101                         ERR("Unable to set group on %s", config.health_unix_sock_path.value);
 102                         PERROR("chown");
 103                         goto error;
 104                 }
 105
 106                 ret = chmod(config.health_unix_sock_path.value,
 107                                 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
 108                 if (ret < 0) {
 109                         ERR("Unable to set permissions on %s", config.health_unix_sock_path.value);
 110                         PERROR("chmod");
 111                         goto error;
 112                 }
 113         }
 114
 115         /*
 116          * Set the CLOEXEC flag. Return code is useless because either way, the
 117          * show must go on.
 118          */
 119         (void) utils_set_fd_cloexec(sock);
 120
 121         ret = lttcomm_listen_unix_sock(sock);
 122         if (ret < 0) {
 123                 goto error;
 124         }
 125
 126         ret = lttng_poll_add(&events, quit_pipe_read_fd, LPOLLIN | LPOLLERR);
 127         if (ret < 0) {
 128                 goto error;
 129         }
 130
 131         /* Add the application registration socket */
 132         ret = lttng_poll_add(&events, sock, LPOLLIN | LPOLLPRI);
 133         if (ret < 0) {
 134                 goto error;
 135         }
 136
 137         mark_thread_as_ready(notifiers);
 138         while (1) {
 139                 DBG("Health check ready");
 140
 141                 /* Infinite blocking call, waiting for transmission */
 142 restart:
 143                 ret = lttng_poll_wait(&events, -1);
 144                 if (ret < 0) {
 145                         /*
 146                          * Restart interrupted system call.
 147                          */
 148                         if (errno == EINTR) {
 149                                 goto restart;
 150                         }
 151                         goto error;
 152                 }
 153
 154                 nb_fd = ret;
 155
 156                 for (i = 0; i < nb_fd; i++) {
 157                         /* Fetch once the poll data */
 158                         revents = LTTNG_POLL_GETEV(&events, i);
 159                         pollfd = LTTNG_POLL_GETFD(&events, i);
 160
 161                         if (!revents) {
 162                                 /* No activity for this FD (poll implementation). */
 163                                 continue;
 164                         }
 165
 166                         /* Event on the registration socket */
 167                         if (pollfd == sock) {
 168                                 if (revents & LPOLLIN) {
 169                                         continue;
 170                                 } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
 171                                         ERR("Health socket poll error");
 172                                         goto error;
 173                                 } else {
 174                                         ERR("Unexpected poll events %u for sock %d", revents, pollfd);
 175                                         goto error;
 176                                 }
 177                         } else {
 178                                 /* Event on the thread's quit pipe. */
 179                                 err = 0;
 180                                 goto exit;
 181                         }
 182                 }
 183
 184                 new_sock = lttcomm_accept_unix_sock(sock);
 185                 if (new_sock < 0) {
 186                         goto error;
 187                 }
 188
 189                 /*
 190                  * Set the CLOEXEC flag. Return code is useless because either way, the
 191                  * show must go on.
 192                  */
 193                 (void) utils_set_fd_cloexec(new_sock);
 194
 195                 DBG("Receiving data from client for health...");
 196                 ret = lttcomm_recv_unix_sock(new_sock, (void *)&msg, sizeof(msg));
 197                 if (ret <= 0) {
 198                         DBG("Nothing recv() from client... continuing");
 199                         ret = close(new_sock);
 200                         if (ret) {
 201                                 PERROR("close");
 202                         }
 203                         continue;
 204                 }
 205
 206                 rcu_thread_online();
 207
 208                 memset(&reply, 0, sizeof(reply));
 209                 for (i = 0; i < NR_HEALTH_SESSIOND_TYPES; i++) {
 210                         /*
 211                          * health_check_state returns 0 if health is
 212                          * bad.
 213                          */
 214                         if (!health_check_state(health_sessiond, i)) {
 215                                 reply.ret_code |= 1ULL << i;
 216                         }
 217                 }
 218
 219                 DBG2("Health check return value %" PRIx64, reply.ret_code);
 220
 221                 ret = lttcomm_send_unix_sock(new_sock, (void *) &reply,
 222                                 sizeof(reply));
 223                 if (ret < 0) {
 224                         ERR("Failed to send health data back to client");
 225                 }
 226
 227                 /* End of transmission */
 228                 ret = close(new_sock);
 229                 if (ret) {
 230                         PERROR("close");
 231                 }
 232         }
 233
 234 exit:
 235 error:
 236         if (err) {
 237                 ERR("Health error occurred in %s", __func__);
 238         }
 239         DBG("Health check thread dying");
 240         unlink(config.health_unix_sock_path.value);
 241         if (sock >= 0) {
 242                 ret = close(sock);
 243                 if (ret) {
 244                         PERROR("close");
 245                 }
 246         }
 247
 248         lttng_poll_clean(&events);
 249         rcu_unregister_thread();
 250         return NULL;
 251 }
 252
 253 static bool shutdown_health_management_thread(void *data)
 254 {
 255         struct thread_notifiers *notifiers = data;
 256         const int write_fd = lttng_pipe_get_writefd(notifiers->quit_pipe);
 257
 258         return notify_thread_pipe(write_fd) == 1;
 259 }
 260
 261 bool launch_health_management_thread(void)
 262 {
 263         struct thread_notifiers *notifiers;
 264         struct lttng_thread *thread;
 265
 266         notifiers = zmalloc(sizeof(*notifiers));
 267         if (!notifiers) {
 268                 goto error;
 269         }
 270
 271         sem_init(&notifiers->ready, 0, 0);
 272         notifiers->quit_pipe = lttng_pipe_open(FD_CLOEXEC);
 273         if (!notifiers->quit_pipe) {
 274                 goto error;
 275         }
 276         thread = lttng_thread_create("Health management",
 277                         thread_manage_health,
 278                         shutdown_health_management_thread,
 279                         cleanup_health_management_thread,
 280                         notifiers);
 281         if (!thread) {
 282                 goto error;
 283         }
 284
 285         wait_until_thread_is_ready(notifiers);
 286         lttng_thread_put(thread);
 287         return true;
 288 error:
 289         cleanup_health_management_thread(notifiers);
 290         return false;
 291 }