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
  51 #include "lttng-relayd.h"
  52 #include "health-relayd.h"
  53
  54 /* Global health check unix path */
  55 static
  56 char health_unix_sock_path[PATH_MAX];
  57
  58 int health_quit_pipe[2];
  59
  60 /*
  61  * Check if the thread quit pipe was triggered.
  62  *
  63  * Return 1 if it was triggered else 0;
  64  */
  65 static
  66 int check_health_quit_pipe(int fd, uint32_t events)
  67 {
  68         if (fd == health_quit_pipe[0] && (events & LPOLLIN)) {
  69                 return 1;
  70         }
  71
  72         return 0;
  73 }
  74
  75 /*
  76  * Send data on a unix socket using the liblttsessiondcomm API.
  77  *
  78  * Return lttcomm error code.
  79  */
  80 static int send_unix_sock(int sock, void *buf, size_t len)
  81 {
  82         /* Check valid length */
  83         if (len == 0) {
  84                 return -1;
  85         }
  86
  87         return lttcomm_send_unix_sock(sock, buf, len);
  88 }
  89
  90 static int create_lttng_rundir_with_perm(const char *rundir)
  91 {
  92         int ret;
  93
  94         DBG3("Creating LTTng run directory: %s", rundir);
  95
  96         ret = mkdir(rundir, S_IRWXU);
  97         if (ret < 0) {
  98                 if (errno != EEXIST) {
  99                         ERR("Unable to create %s", rundir);
 100                         goto error;
 101                 } else {
 102                         ret = 0;
 103                 }
 104         } else if (ret == 0) {
 105                 int is_root = !getuid();
 106
 107                 if (is_root) {
 108                         gid_t gid;
 109
 110                         ret = utils_get_group_id(tracing_group_name, true, &gid);
 111                         if (ret) {
 112                                 /* Default to root group. */
 113                                 gid = 0;
 114                         }
 115
 116                         ret = chown(rundir, 0, gid);
 117                         if (ret < 0) {
 118                                 ERR("Unable to set group on %s", rundir);
 119                                 PERROR("chown");
 120                                 ret = -1;
 121                                 goto error;
 122                         }
 123
 124                         ret = chmod(rundir,
 125                                         S_IRUSR | S_IWUSR | S_IXUSR | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH);
 126                         if (ret < 0) {
 127                                 ERR("Unable to set permissions on %s", health_unix_sock_path);
 128                                 PERROR("chmod");
 129                                 ret = -1;
 130                                 goto error;
 131                         }
 132                 }
 133         }
 134
 135 error:
 136         return ret;
 137 }
 138
 139 static
 140 int parse_health_env(void)
 141 {
 142         const char *health_path;
 143
 144         health_path = lttng_secure_getenv(LTTNG_RELAYD_HEALTH_ENV);
 145         if (health_path) {
 146                 strncpy(health_unix_sock_path, health_path,
 147                         PATH_MAX);
 148                 health_unix_sock_path[PATH_MAX - 1] = '\0';
 149         }
 150
 151         return 0;
 152 }
 153
 154 static
 155 int setup_health_path(void)
 156 {
 157         int is_root, ret = 0;
 158         const char *home_path = NULL;
 159         char *rundir = NULL, *relayd_path = NULL;
 160
 161         ret = parse_health_env();
 162         if (ret) {
 163                 return ret;
 164         }
 165
 166         is_root = !getuid();
 167
 168         if (is_root) {
 169                 rundir = strdup(DEFAULT_LTTNG_RUNDIR);
 170                 if (!rundir) {
 171                         ret = -ENOMEM;
 172                         goto end;
 173                 }
 174         } else {
 175                 /*
 176                  * Create rundir from home path. This will create something like
 177                  * $HOME/.lttng
 178                  */
 179                 home_path = utils_get_home_dir();
 180
 181                 if (home_path == NULL) {
 182                         /* TODO: Add --socket PATH option */
 183                         ERR("Can't get HOME directory for sockets creation.");
 184                         ret = -EPERM;
 185                         goto end;
 186                 }
 187
 188                 ret = asprintf(&rundir, DEFAULT_LTTNG_HOME_RUNDIR, home_path);
 189                 if (ret < 0) {
 190                         ret = -ENOMEM;
 191                         goto end;
 192                 }
 193         }
 194
 195         ret = asprintf(&relayd_path, DEFAULT_RELAYD_PATH, rundir);
 196         if (ret < 0) {
 197                 ret = -ENOMEM;
 198                 goto end;
 199         }
 200
 201         ret = create_lttng_rundir_with_perm(rundir);
 202         if (ret < 0) {
 203                 goto end;
 204         }
 205
 206         ret = create_lttng_rundir_with_perm(relayd_path);
 207         if (ret < 0) {
 208                 goto end;
 209         }
 210
 211         if (is_root) {
 212                 if (strlen(health_unix_sock_path) != 0) {
 213                         goto end;
 214                 }
 215                 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 216                         DEFAULT_GLOBAL_RELAY_HEALTH_UNIX_SOCK,
 217                         (int) getpid());
 218         } else {
 219                 /* Set health check Unix path */
 220                 if (strlen(health_unix_sock_path) != 0) {
 221                         goto end;
 222                 }
 223
 224                 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 225                         DEFAULT_HOME_RELAY_HEALTH_UNIX_SOCK,
 226                         home_path, (int) getpid());
 227         }
 228
 229 end:
 230         free(rundir);
 231         free(relayd_path);
 232         return ret;
 233 }
 234
 235 /*
 236  * Thread managing health check socket.
 237  */
 238 void *thread_manage_health(void *data)
 239 {
 240         int sock = -1, new_sock = -1, ret, i, pollfd, err = -1;
 241         uint32_t revents, nb_fd;
 242         struct lttng_poll_event events;
 243         struct health_comm_msg msg;
 244         struct health_comm_reply reply;
 245         int is_root;
 246
 247         DBG("[thread] Manage health check started");
 248
 249         setup_health_path();
 250
 251         rcu_register_thread();
 252
 253         /* We might hit an error path before this is created. */
 254         lttng_poll_init(&events);
 255
 256         /* Create unix socket */
 257         sock = lttcomm_create_unix_sock(health_unix_sock_path);
 258         if (sock < 0) {
 259                 ERR("Unable to create health check Unix socket");
 260                 err = -1;
 261                 goto error;
 262         }
 263
 264         is_root = !getuid();
 265         if (is_root) {
 266                 /* lttng health client socket path permissions */
 267                 gid_t gid;
 268
 269                 ret = utils_get_group_id(tracing_group_name, true, &gid);
 270                 if (ret) {
 271                         /* Default to root group. */
 272                         gid = 0;
 273                 }
 274
 275                 ret = chown(health_unix_sock_path, 0, gid);
 276                 if (ret < 0) {
 277                         ERR("Unable to set group on %s", health_unix_sock_path);
 278                         PERROR("chown");
 279                         err = -1;
 280                         goto error;
 281                 }
 282
 283                 ret = chmod(health_unix_sock_path,
 284                                 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
 285                 if (ret < 0) {
 286                         ERR("Unable to set permissions on %s", health_unix_sock_path);
 287                         PERROR("chmod");
 288                         err = -1;
 289                         goto error;
 290                 }
 291         }
 292
 293         /*
 294          * Set the CLOEXEC flag. Return code is useless because either way, the
 295          * show must go on.
 296          */
 297         (void) utils_set_fd_cloexec(sock);
 298
 299         ret = lttcomm_listen_unix_sock(sock);
 300         if (ret < 0) {
 301                 goto error;
 302         }
 303
 304         /* Size is set to 1 for the consumer_channel pipe */
 305         ret = lttng_poll_create(&events, 2, LTTNG_CLOEXEC);
 306         if (ret < 0) {
 307                 ERR("Poll set creation failed");
 308                 goto error;
 309         }
 310
 311         ret = lttng_poll_add(&events, health_quit_pipe[0], LPOLLIN);
 312         if (ret < 0) {
 313                 goto error;
 314         }
 315
 316         /* Add the application registration socket */
 317         ret = lttng_poll_add(&events, sock, LPOLLIN | LPOLLPRI);
 318         if (ret < 0) {
 319                 goto error;
 320         }
 321
 322         lttng_relay_notify_ready();
 323
 324         while (1) {
 325                 DBG("Health check ready");
 326
 327                 /* Inifinite blocking call, waiting for transmission */
 328 restart:
 329                 ret = lttng_poll_wait(&events, -1);
 330                 if (ret < 0) {
 331                         /*
 332                          * Restart interrupted system call.
 333                          */
 334                         if (errno == EINTR) {
 335                                 goto restart;
 336                         }
 337                         goto error;
 338                 }
 339
 340                 nb_fd = ret;
 341
 342                 for (i = 0; i < nb_fd; i++) {
 343                         /* Fetch once the poll data */
 344                         revents = LTTNG_POLL_GETEV(&events, i);
 345                         pollfd = LTTNG_POLL_GETFD(&events, i);
 346
 347                         /* Thread quit pipe has been closed. Killing thread. */
 348                         ret = check_health_quit_pipe(pollfd, revents);
 349                         if (ret) {
 350                                 err = 0;
 351                                 goto exit;
 352                         }
 353
 354                         /* Event on the registration socket */
 355                         if (pollfd == sock) {
 356                                 if (revents & LPOLLIN) {
 357                                         continue;
 358                                 } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
 359                                         ERR("Health socket poll error");
 360                                         goto error;
 361                                 } else {
 362                                         ERR("Unexpected poll events %u for sock %d", revents, pollfd);
 363                                         goto error;
 364                                 }
 365                         }
 366                 }
 367
 368                 new_sock = lttcomm_accept_unix_sock(sock);
 369                 if (new_sock < 0) {
 370                         goto error;
 371                 }
 372
 373                 /*
 374                  * Set the CLOEXEC flag. Return code is useless because either way, the
 375                  * show must go on.
 376                  */
 377                 (void) utils_set_fd_cloexec(new_sock);
 378
 379                 DBG("Receiving data from client for health...");
 380                 ret = lttcomm_recv_unix_sock(new_sock, (void *)&msg, sizeof(msg));
 381                 if (ret <= 0) {
 382                         DBG("Nothing recv() from client... continuing");
 383                         ret = close(new_sock);
 384                         if (ret) {
 385                                 PERROR("close");
 386                         }
 387                         new_sock = -1;
 388                         continue;
 389                 }
 390
 391                 rcu_thread_online();
 392
 393                 assert(msg.cmd == HEALTH_CMD_CHECK);
 394
 395                 memset(&reply, 0, sizeof(reply));
 396                 for (i = 0; i < NR_HEALTH_RELAYD_TYPES; i++) {
 397                         /*
 398                          * health_check_state return 0 if thread is in
 399                          * error.
 400                          */
 401                         if (!health_check_state(health_relayd, i)) {
 402                                 reply.ret_code |= 1ULL << i;
 403                         }
 404                 }
 405
 406                 DBG2("Health check return value %" PRIx64, reply.ret_code);
 407
 408                 ret = send_unix_sock(new_sock, (void *) &reply, sizeof(reply));
 409                 if (ret < 0) {
 410                         ERR("Failed to send health data back to client");
 411                 }
 412
 413                 /* End of transmission */
 414                 ret = close(new_sock);
 415                 if (ret) {
 416                         PERROR("close");
 417                 }
 418                 new_sock = -1;
 419         }
 420
 421 error:
 422         lttng_relay_stop_threads();
 423 exit:
 424         if (err) {
 425                 ERR("Health error occurred in %s", __func__);
 426         }
 427         DBG("Health check thread dying");
 428         unlink(health_unix_sock_path);
 429         if (sock >= 0) {
 430                 ret = close(sock);
 431                 if (ret) {
 432                         PERROR("close");
 433                 }
 434         }
 435
 436         /*
 437          * We do NOT rmdir rundir nor the relayd path because there are
 438          * other processes using them.
 439          */
 440
 441         lttng_poll_clean(&events);
 442
 443         rcu_unregister_thread();
 444         return NULL;
 445 }