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 _LGPL_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 <urcu/compiler.h>
  40 #include <inttypes.h>
  41
  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>
  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] = { -1, -1 };
  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                         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);
 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
 140 static
 141 int parse_health_env(void)
 142 {
 143         const char *health_path;
 144
 145         health_path = lttng_secure_getenv(LTTNG_RELAYD_HEALTH_ENV);
 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
 155 static
 156 int setup_health_path(void)
 157 {
 158         int is_root, ret = 0;
 159         const char *home_path = NULL;
 160         char *rundir = NULL, *relayd_path = NULL;
 161
 162         ret = parse_health_env();
 163         if (ret) {
 164                 return ret;
 165         }
 166
 167         is_root = !getuid();
 168
 169         if (is_root) {
 170                 rundir = strdup(DEFAULT_LTTNG_RUNDIR);
 171                 if (!rundir) {
 172                         ret = -ENOMEM;
 173                         goto end;
 174                 }
 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
 189                 ret = asprintf(&rundir, DEFAULT_LTTNG_HOME_RUNDIR, home_path);
 190                 if (ret < 0) {
 191                         ret = -ENOMEM;
 192                         goto end;
 193                 }
 194         }
 195
 196         ret = asprintf(&relayd_path, DEFAULT_RELAYD_PATH, rundir);
 197         if (ret < 0) {
 198                 ret = -ENOMEM;
 199                 goto end;
 200         }
 201
 202         ret = create_lttng_rundir_with_perm(rundir);
 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,
 218                         (int) getpid());
 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,
 227                         home_path, (int) getpid());
 228         }
 229
 230 end:
 231         free(rundir);
 232         free(relayd_path);
 233         return ret;
 234 }
 235
 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
 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
 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;
 281         char *sock_name;
 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 */
 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) {
 304                 ERR("Unable to create health check Unix socket");
 305                 err = -1;
 306                 goto error;
 307         }
 308
 309         is_root = !getuid();
 310         if (is_root) {
 311                 /* lttng health client socket path permissions */
 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);
 321                 if (ret < 0) {
 322                         ERR("Unable to set group on %s", health_unix_sock_path);
 323                         PERROR("chown");
 324                         err = -1;
 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");
 333                         err = -1;
 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
 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);
 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
 369         lttng_relay_notify_ready();
 370
 371         while (1) {
 372                 char *accepted_socket_name;
 373
 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) {
 405                                 if (revents & LPOLLIN) {
 406                                         continue;
 407                                 } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
 408                                         ERR("Health socket poll error");
 409                                         goto error;
 410                                 } else {
 411                                         ERR("Unexpected poll events %u for sock %d", revents, pollfd);
 412                                         goto error;
 413                                 }
 414                         }
 415                 }
 416
 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) {
 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");
 442                         ret = fd_tracker_close_unsuspendable_fd(the_fd_tracker,
 443                                         &new_sock, 1, fd_tracker_util_close_fd,
 444                                         NULL);
 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
 456                 memset(&reply, 0, sizeof(reply));
 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 */
 475                 ret = fd_tracker_close_unsuspendable_fd(the_fd_tracker,
 476                                 &new_sock, 1, fd_tracker_util_close_fd,
 477                                 NULL);
 478                 if (ret) {
 479                         PERROR("close");
 480                 }
 481                 new_sock = -1;
 482         }
 483
 484 error:
 485         lttng_relay_stop_threads();
 486 exit:
 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) {
 493                 ret = fd_tracker_close_unsuspendable_fd(the_fd_tracker, &sock,
 494                                 1, fd_tracker_util_close_fd, NULL);
 495                 if (ret) {
 496                         PERROR("close");
 497                 }
 498         }
 499
 500         /*
 501          * We do NOT rmdir rundir nor the relayd path because there are
 502          * other processes using them.
 503          */
 504
 505         (void) fd_tracker_util_poll_clean(the_fd_tracker, &events);
 506
 507         rcu_unregister_thread();
 508         return NULL;
 509 }