src/bin/lttng-relayd/health-relayd.c

   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
  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>
  39 #include <config.h>
  40 #include <urcu/compiler.h>
  41 #include <ulimit.h>
  42 #include <inttypes.h>
  43
  44 #include <common/defaults.h>
  45 #include <common/common.h>
  46 #include <common/consumer.h>
  47 #include <common/consumer-timer.h>
  48 #include <common/compat/poll.h>
  49 #include <common/sessiond-comm/sessiond-comm.h>
  50 #include <common/utils.h>
  51
  52 #include "lttng-relayd.h"
  53 #include "health-relayd.h"
  54
  55 /* Global health check unix path */
  56 static
  57 char health_unix_sock_path[PATH_MAX];
  58
  59 int health_quit_pipe[2];
  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) {
 109                         ret = chown(rundir, 0,
 110                                         utils_get_group_id(tracing_group_name));
 111                         if (ret < 0) {
 112                                 ERR("Unable to set group on %s", rundir);
 113                                 PERROR("chown");
 114                                 ret = -1;
 115                                 goto error;
 116                         }
 117
 118                         ret = chmod(rundir,
 119                                         S_IRUSR | S_IWUSR | S_IXUSR | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH);
 120                         if (ret < 0) {
 121                                 ERR("Unable to set permissions on %s", health_unix_sock_path);
 122                                 PERROR("chmod");
 123                                 ret = -1;
 124                                 goto error;
 125                         }
 126                 }
 127         }
 128
 129 error:
 130         return ret;
 131 }
 132
 133 static
 134 int parse_health_env(void)
 135 {
 136         const char *health_path;
 137
 138         health_path = getenv(LTTNG_RELAYD_HEALTH_ENV);
 139         if (health_path) {
 140                 strncpy(health_unix_sock_path, health_path,
 141                         PATH_MAX);
 142                 health_unix_sock_path[PATH_MAX - 1] = '\0';
 143         }
 144
 145         return 0;
 146 }
 147
 148 static
 149 int setup_health_path(void)
 150 {
 151         int is_root, ret = 0;
 152         char *home_path = NULL, *rundir = NULL, *relayd_path;
 153
 154         ret = parse_health_env();
 155         if (ret) {
 156                 return ret;
 157         }
 158
 159         is_root = !getuid();
 160
 161         if (is_root) {
 162                 rundir = strdup(DEFAULT_LTTNG_RUNDIR);
 163         } else {
 164                 /*
 165                  * Create rundir from home path. This will create something like
 166                  * $HOME/.lttng
 167                  */
 168                 home_path = utils_get_home_dir();
 169
 170                 if (home_path == NULL) {
 171                         /* TODO: Add --socket PATH option */
 172                         ERR("Can't get HOME directory for sockets creation.");
 173                         ret = -EPERM;
 174                         goto end;
 175                 }
 176
 177                 ret = asprintf(&rundir, DEFAULT_LTTNG_HOME_RUNDIR, home_path);
 178                 if (ret < 0) {
 179                         ret = -ENOMEM;
 180                         goto end;
 181                 }
 182         }
 183
 184         ret = asprintf(&relayd_path, DEFAULT_RELAYD_PATH, rundir);
 185         if (ret < 0) {
 186                 ret = -ENOMEM;
 187                 goto end;
 188         }
 189
 190         ret = create_lttng_rundir_with_perm(rundir);
 191         if (ret < 0) {
 192                 goto end;
 193         }
 194
 195         ret = create_lttng_rundir_with_perm(relayd_path);
 196         if (ret < 0) {
 197                 goto end;
 198         }
 199
 200         if (is_root) {
 201                 if (strlen(health_unix_sock_path) != 0) {
 202                         goto end;
 203                 }
 204                 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 205                         DEFAULT_GLOBAL_RELAY_HEALTH_UNIX_SOCK,
 206                         getpid());
 207         } else {
 208                 /* Set health check Unix path */
 209                 if (strlen(health_unix_sock_path) != 0) {
 210                         goto end;
 211                 }
 212
 213                 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 214                         DEFAULT_HOME_RELAY_HEALTH_UNIX_SOCK,
 215                         home_path, getpid());
 216         }
 217
 218 end:
 219         free(rundir);
 220         return ret;
 221 }
 222
 223 /*
 224  * Thread managing health check socket.
 225  */
 226 void *thread_manage_health(void *data)
 227 {
 228         int sock = -1, new_sock = -1, ret, i, pollfd, err = -1;
 229         uint32_t revents, nb_fd;
 230         struct lttng_poll_event events;
 231         struct health_comm_msg msg;
 232         struct health_comm_reply reply;
 233         int is_root;
 234
 235         DBG("[thread] Manage health check started");
 236
 237         setup_health_path();
 238
 239         rcu_register_thread();
 240
 241         /* We might hit an error path before this is created. */
 242         lttng_poll_init(&events);
 243
 244         /* Create unix socket */
 245         sock = lttcomm_create_unix_sock(health_unix_sock_path);
 246         if (sock < 0) {
 247                 ERR("Unable to create health check Unix socket");
 248                 ret = -1;
 249                 goto error;
 250         }
 251
 252         is_root = !getuid();
 253         if (is_root) {
 254                 /* lttng health client socket path permissions */
 255                 ret = chown(health_unix_sock_path, 0,
 256                                 utils_get_group_id(tracing_group_name));
 257                 if (ret < 0) {
 258                         ERR("Unable to set group on %s", health_unix_sock_path);
 259                         PERROR("chown");
 260                         ret = -1;
 261                         goto error;
 262                 }
 263
 264                 ret = chmod(health_unix_sock_path,
 265                                 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
 266                 if (ret < 0) {
 267                         ERR("Unable to set permissions on %s", health_unix_sock_path);
 268                         PERROR("chmod");
 269                         ret = -1;
 270                         goto error;
 271                 }
 272         }
 273
 274         /*
 275          * Set the CLOEXEC flag. Return code is useless because either way, the
 276          * show must go on.
 277          */
 278         (void) utils_set_fd_cloexec(sock);
 279
 280         ret = lttcomm_listen_unix_sock(sock);
 281         if (ret < 0) {
 282                 goto error;
 283         }
 284
 285         /* Size is set to 1 for the consumer_channel pipe */
 286         ret = lttng_poll_create(&events, 2, LTTNG_CLOEXEC);
 287         if (ret < 0) {
 288                 ERR("Poll set creation failed");
 289                 goto error;
 290         }
 291
 292         ret = lttng_poll_add(&events, health_quit_pipe[0], LPOLLIN);
 293         if (ret < 0) {
 294                 goto error;
 295         }
 296
 297         /* Add the application registration socket */
 298         ret = lttng_poll_add(&events, sock, LPOLLIN | LPOLLPRI);
 299         if (ret < 0) {
 300                 goto error;
 301         }
 302
 303         lttng_relay_notify_ready();
 304
 305         while (1) {
 306                 DBG("Health check ready");
 307
 308                 /* Inifinite blocking call, waiting for transmission */
 309 restart:
 310                 ret = lttng_poll_wait(&events, -1);
 311                 if (ret < 0) {
 312                         /*
 313                          * Restart interrupted system call.
 314                          */
 315                         if (errno == EINTR) {
 316                                 goto restart;
 317                         }
 318                         goto error;
 319                 }
 320
 321                 nb_fd = ret;
 322
 323                 for (i = 0; i < nb_fd; i++) {
 324                         /* Fetch once the poll data */
 325                         revents = LTTNG_POLL_GETEV(&events, i);
 326                         pollfd = LTTNG_POLL_GETFD(&events, i);
 327
 328                         /* Thread quit pipe has been closed. Killing thread. */
 329                         ret = check_health_quit_pipe(pollfd, revents);
 330                         if (ret) {
 331                                 err = 0;
 332                                 goto exit;
 333                         }
 334
 335                         /* Event on the registration socket */
 336                         if (pollfd == sock) {
 337                                 if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
 338                                         ERR("Health socket poll error");
 339                                         goto error;
 340                                 }
 341                         }
 342                 }
 343
 344                 new_sock = lttcomm_accept_unix_sock(sock);
 345                 if (new_sock < 0) {
 346                         goto error;
 347                 }
 348
 349                 /*
 350                  * Set the CLOEXEC flag. Return code is useless because either way, the
 351                  * show must go on.
 352                  */
 353                 (void) utils_set_fd_cloexec(new_sock);
 354
 355                 DBG("Receiving data from client for health...");
 356                 ret = lttcomm_recv_unix_sock(new_sock, (void *)&msg, sizeof(msg));
 357                 if (ret <= 0) {
 358                         DBG("Nothing recv() from client... continuing");
 359                         ret = close(new_sock);
 360                         if (ret) {
 361                                 PERROR("close");
 362                         }
 363                         new_sock = -1;
 364                         continue;
 365                 }
 366
 367                 rcu_thread_online();
 368
 369                 assert(msg.cmd == HEALTH_CMD_CHECK);
 370
 371                 memset(&reply, 0, sizeof(reply));
 372                 for (i = 0; i < NR_HEALTH_RELAYD_TYPES; i++) {
 373                         /*
 374                          * health_check_state return 0 if thread is in
 375                          * error.
 376                          */
 377                         if (!health_check_state(health_relayd, i)) {
 378                                 reply.ret_code |= 1ULL << i;
 379                         }
 380                 }
 381
 382                 DBG2("Health check return value %" PRIx64, reply.ret_code);
 383
 384                 ret = send_unix_sock(new_sock, (void *) &reply, sizeof(reply));
 385                 if (ret < 0) {
 386                         ERR("Failed to send health data back to client");
 387                 }
 388
 389                 /* End of transmission */
 390                 ret = close(new_sock);
 391                 if (ret) {
 392                         PERROR("close");
 393                 }
 394                 new_sock = -1;
 395         }
 396
 397 exit:
 398 error:
 399         if (err) {
 400                 ERR("Health error occurred in %s", __func__);
 401         }
 402         DBG("Health check thread dying");
 403         unlink(health_unix_sock_path);
 404         if (sock >= 0) {
 405                 ret = close(sock);
 406                 if (ret) {
 407                         PERROR("close");
 408                 }
 409         }
 410
 411         /*
 412          * We do NOT rmdir rundir nor the relayd path because there are
 413          * other processes using them.
 414          */
 415
 416         lttng_poll_clean(&events);
 417
 418         rcu_unregister_thread();
 419         return NULL;
 420 }