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