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