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                 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                         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, getpid());
 220         }
 221
 222 end:
 223         free(rundir);
 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                 ret = -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                         ret = -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                         ret = -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 & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
 347                                         ERR("Health socket poll error");
 348                                         goto error;
 349                                 }
 350                         }
 351                 }
 352
 353                 new_sock = lttcomm_accept_unix_sock(sock);
 354                 if (new_sock < 0) {
 355                         goto error;
 356                 }
 357
 358                 /*
 359                  * Set the CLOEXEC flag. Return code is useless because either way, the
 360                  * show must go on.
 361                  */
 362                 (void) utils_set_fd_cloexec(new_sock);
 363
 364                 DBG("Receiving data from client for health...");
 365                 ret = lttcomm_recv_unix_sock(new_sock, (void *)&msg, sizeof(msg));
 366                 if (ret <= 0) {
 367                         DBG("Nothing recv() from client... continuing");
 368                         ret = close(new_sock);
 369                         if (ret) {
 370                                 PERROR("close");
 371                         }
 372                         new_sock = -1;
 373                         continue;
 374                 }
 375
 376                 rcu_thread_online();
 377
 378                 assert(msg.cmd == HEALTH_CMD_CHECK);
 379
 380                 memset(&reply, 0, sizeof(reply));
 381                 for (i = 0; i < NR_HEALTH_RELAYD_TYPES; i++) {
 382                         /*
 383                          * health_check_state return 0 if thread is in
 384                          * error.
 385                          */
 386                         if (!health_check_state(health_relayd, i)) {
 387                                 reply.ret_code |= 1ULL << i;
 388                         }
 389                 }
 390
 391                 DBG2("Health check return value %" PRIx64, reply.ret_code);
 392
 393                 ret = send_unix_sock(new_sock, (void *) &reply, sizeof(reply));
 394                 if (ret < 0) {
 395                         ERR("Failed to send health data back to client");
 396                 }
 397
 398                 /* End of transmission */
 399                 ret = close(new_sock);
 400                 if (ret) {
 401                         PERROR("close");
 402                 }
 403                 new_sock = -1;
 404         }
 405
 406 exit:
 407 error:
 408         if (err) {
 409                 ERR("Health error occurred in %s", __func__);
 410         }
 411         DBG("Health check thread dying");
 412         unlink(health_unix_sock_path);
 413         if (sock >= 0) {
 414                 ret = close(sock);
 415                 if (ret) {
 416                         PERROR("close");
 417                 }
 418         }
 419
 420         /*
 421          * We do NOT rmdir rundir nor the relayd path because there are
 422          * other processes using them.
 423          */
 424
 425         lttng_poll_clean(&events);
 426
 427         rcu_unregister_thread();
 428         return NULL;
 429 }