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 #define _LGPL_SOURCE
  20 #include <fcntl.h>
  21 #include <getopt.h>
  22 #include <grp.h>
  23 #include <limits.h>
  24 #include <pthread.h>
  25 #include <signal.h>
  26 #include <stdio.h>
  27 #include <stdlib.h>
  28 #include <string.h>
  29 #include <sys/ipc.h>
  30 #include <sys/resource.h>
  31 #include <sys/shm.h>
  32 #include <sys/socket.h>
  33 #include <sys/stat.h>
  34 #include <sys/types.h>
  35 #include <urcu/list.h>
  36 #include <poll.h>
  37 #include <unistd.h>
  38 #include <sys/mman.h>
  39 #include <assert.h>
  40 #include <config.h>
  41 #include <urcu/compiler.h>
  42 #include <ulimit.h>
  43 #include <inttypes.h>
  44
  45 #include <common/defaults.h>
  46 #include <common/common.h>
  47 #include <common/consumer.h>
  48 #include <common/consumer-timer.h>
  49 #include <common/compat/poll.h>
  50 #include <common/sessiond-comm/sessiond-comm.h>
  51 #include <common/utils.h>
  52
  53 #include "lttng-relayd.h"
  54 #include "health-relayd.h"
  55
  56 /* Global health check unix path */
  57 static
  58 char health_unix_sock_path[PATH_MAX];
  59
  60 int health_quit_pipe[2];
  61
  62 /*
  63  * Check if the thread quit pipe was triggered.
  64  *
  65  * Return 1 if it was triggered else 0;
  66  */
  67 static
  68 int check_health_quit_pipe(int fd, uint32_t events)
  69 {
  70         if (fd == health_quit_pipe[0] && (events & LPOLLIN)) {
  71                 return 1;
  72         }
  73
  74         return 0;
  75 }
  76
  77 /*
  78  * Send data on a unix socket using the liblttsessiondcomm API.
  79  *
  80  * Return lttcomm error code.
  81  */
  82 static int send_unix_sock(int sock, void *buf, size_t len)
  83 {
  84         /* Check valid length */
  85         if (len == 0) {
  86                 return -1;
  87         }
  88
  89         return lttcomm_send_unix_sock(sock, buf, len);
  90 }
  91
  92 static int create_lttng_rundir_with_perm(const char *rundir)
  93 {
  94         int ret;
  95
  96         DBG3("Creating LTTng run directory: %s", rundir);
  97
  98         ret = mkdir(rundir, S_IRWXU);
  99         if (ret < 0) {
 100                 if (errno != EEXIST) {
 101                         ERR("Unable to create %s", rundir);
 102                         goto error;
 103                 } else {
 104                         ret = 0;
 105                 }
 106         } else if (ret == 0) {
 107                 int is_root = !getuid();
 108
 109                 if (is_root) {
 110                         ret = chown(rundir, 0,
 111                                         utils_get_group_id(tracing_group_name));
 112                         if (ret < 0) {
 113                                 ERR("Unable to set group on %s", rundir);
 114                                 PERROR("chown");
 115                                 ret = -1;
 116                                 goto error;
 117                         }
 118
 119                         ret = chmod(rundir,
 120                                         S_IRUSR | S_IWUSR | S_IXUSR | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH);
 121                         if (ret < 0) {
 122                                 ERR("Unable to set permissions on %s", health_unix_sock_path);
 123                                 PERROR("chmod");
 124                                 ret = -1;
 125                                 goto error;
 126                         }
 127                 }
 128         }
 129
 130 error:
 131         return ret;
 132 }
 133
 134 static
 135 int parse_health_env(void)
 136 {
 137         const char *health_path;
 138
 139         health_path = getenv(LTTNG_RELAYD_HEALTH_ENV);
 140         if (health_path) {
 141                 strncpy(health_unix_sock_path, health_path,
 142                         PATH_MAX);
 143                 health_unix_sock_path[PATH_MAX - 1] = '\0';
 144         }
 145
 146         return 0;
 147 }
 148
 149 static
 150 int setup_health_path(void)
 151 {
 152         int is_root, ret = 0;
 153         char *home_path = NULL, *rundir = NULL, *relayd_path;
 154
 155         ret = parse_health_env();
 156         if (ret) {
 157                 return ret;
 158         }
 159
 160         is_root = !getuid();
 161
 162         if (is_root) {
 163                 rundir = strdup(DEFAULT_LTTNG_RUNDIR);
 164                 if (!rundir) {
 165                         ret = -ENOMEM;
 166                         goto end;
 167                 }
 168         } else {
 169                 /*
 170                  * Create rundir from home path. This will create something like
 171                  * $HOME/.lttng
 172                  */
 173                 home_path = utils_get_home_dir();
 174
 175                 if (home_path == NULL) {
 176                         /* TODO: Add --socket PATH option */
 177                         ERR("Can't get HOME directory for sockets creation.");
 178                         ret = -EPERM;
 179                         goto end;
 180                 }
 181
 182                 ret = asprintf(&rundir, DEFAULT_LTTNG_HOME_RUNDIR, home_path);
 183                 if (ret < 0) {
 184                         ret = -ENOMEM;
 185                         goto end;
 186                 }
 187         }
 188
 189         ret = asprintf(&relayd_path, DEFAULT_RELAYD_PATH, rundir);
 190         if (ret < 0) {
 191                 ret = -ENOMEM;
 192                 goto end;
 193         }
 194
 195         ret = create_lttng_rundir_with_perm(rundir);
 196         if (ret < 0) {
 197                 goto end;
 198         }
 199
 200         ret = create_lttng_rundir_with_perm(relayd_path);
 201         if (ret < 0) {
 202                 goto end;
 203         }
 204
 205         if (is_root) {
 206                 if (strlen(health_unix_sock_path) != 0) {
 207                         goto end;
 208                 }
 209                 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 210                         DEFAULT_GLOBAL_RELAY_HEALTH_UNIX_SOCK,
 211                         getpid());
 212         } else {
 213                 /* Set health check Unix path */
 214                 if (strlen(health_unix_sock_path) != 0) {
 215                         goto end;
 216                 }
 217
 218                 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 219                         DEFAULT_HOME_RELAY_HEALTH_UNIX_SOCK,
 220                         home_path, getpid());
 221         }
 222
 223 end:
 224         free(rundir);
 225         return ret;
 226 }
 227
 228 /*
 229  * Thread managing health check socket.
 230  */
 231 void *thread_manage_health(void *data)
 232 {
 233         int sock = -1, new_sock = -1, ret, i, pollfd, err = -1;
 234         uint32_t revents, nb_fd;
 235         struct lttng_poll_event events;
 236         struct health_comm_msg msg;
 237         struct health_comm_reply reply;
 238         int is_root;
 239
 240         DBG("[thread] Manage health check started");
 241
 242         setup_health_path();
 243
 244         rcu_register_thread();
 245
 246         /* We might hit an error path before this is created. */
 247         lttng_poll_init(&events);
 248
 249         /* Create unix socket */
 250         sock = lttcomm_create_unix_sock(health_unix_sock_path);
 251         if (sock < 0) {
 252                 ERR("Unable to create health check Unix socket");
 253                 ret = -1;
 254                 goto error;
 255         }
 256
 257         is_root = !getuid();
 258         if (is_root) {
 259                 /* lttng health client socket path permissions */
 260                 ret = chown(health_unix_sock_path, 0,
 261                                 utils_get_group_id(tracing_group_name));
 262                 if (ret < 0) {
 263                         ERR("Unable to set group on %s", health_unix_sock_path);
 264                         PERROR("chown");
 265                         ret = -1;
 266                         goto error;
 267                 }
 268
 269                 ret = chmod(health_unix_sock_path,
 270                                 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
 271                 if (ret < 0) {
 272                         ERR("Unable to set permissions on %s", health_unix_sock_path);
 273                         PERROR("chmod");
 274                         ret = -1;
 275                         goto error;
 276                 }
 277         }
 278
 279         /*
 280          * Set the CLOEXEC flag. Return code is useless because either way, the
 281          * show must go on.
 282          */
 283         (void) utils_set_fd_cloexec(sock);
 284
 285         ret = lttcomm_listen_unix_sock(sock);
 286         if (ret < 0) {
 287                 goto error;
 288         }
 289
 290         /* Size is set to 1 for the consumer_channel pipe */
 291         ret = lttng_poll_create(&events, 2, LTTNG_CLOEXEC);
 292         if (ret < 0) {
 293                 ERR("Poll set creation failed");
 294                 goto error;
 295         }
 296
 297         ret = lttng_poll_add(&events, health_quit_pipe[0], LPOLLIN);
 298         if (ret < 0) {
 299                 goto error;
 300         }
 301
 302         /* Add the application registration socket */
 303         ret = lttng_poll_add(&events, sock, LPOLLIN | LPOLLPRI);
 304         if (ret < 0) {
 305                 goto error;
 306         }
 307
 308         lttng_relay_notify_ready();
 309
 310         while (1) {
 311                 DBG("Health check ready");
 312
 313                 /* Inifinite blocking call, waiting for transmission */
 314 restart:
 315                 ret = lttng_poll_wait(&events, -1);
 316                 if (ret < 0) {
 317                         /*
 318                          * Restart interrupted system call.
 319                          */
 320                         if (errno == EINTR) {
 321                                 goto restart;
 322                         }
 323                         goto error;
 324                 }
 325
 326                 nb_fd = ret;
 327
 328                 for (i = 0; i < nb_fd; i++) {
 329                         /* Fetch once the poll data */
 330                         revents = LTTNG_POLL_GETEV(&events, i);
 331                         pollfd = LTTNG_POLL_GETFD(&events, i);
 332
 333                         if (!revents) {
 334                                 /* No activity for this FD (poll implementation). */
 335                                 continue;
 336                         }
 337
 338                         /* Thread quit pipe has been closed. Killing thread. */
 339                         ret = check_health_quit_pipe(pollfd, revents);
 340                         if (ret) {
 341                                 err = 0;
 342                                 goto exit;
 343                         }
 344
 345                         /* Event on the registration socket */
 346                         if (pollfd == sock) {
 347                                 if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
 348                                         ERR("Health socket poll error");
 349                                         goto error;
 350                                 }
 351                         }
 352                 }
 353
 354                 new_sock = lttcomm_accept_unix_sock(sock);
 355                 if (new_sock < 0) {
 356                         goto error;
 357                 }
 358
 359                 /*
 360                  * Set the CLOEXEC flag. Return code is useless because either way, the
 361                  * show must go on.
 362                  */
 363                 (void) utils_set_fd_cloexec(new_sock);
 364
 365                 DBG("Receiving data from client for health...");
 366                 ret = lttcomm_recv_unix_sock(new_sock, (void *)&msg, sizeof(msg));
 367                 if (ret <= 0) {
 368                         DBG("Nothing recv() from client... continuing");
 369                         ret = close(new_sock);
 370                         if (ret) {
 371                                 PERROR("close");
 372                         }
 373                         new_sock = -1;
 374                         continue;
 375                 }
 376
 377                 rcu_thread_online();
 378
 379                 assert(msg.cmd == HEALTH_CMD_CHECK);
 380
 381                 memset(&reply, 0, sizeof(reply));
 382                 for (i = 0; i < NR_HEALTH_RELAYD_TYPES; i++) {
 383                         /*
 384                          * health_check_state return 0 if thread is in
 385                          * error.
 386                          */
 387                         if (!health_check_state(health_relayd, i)) {
 388                                 reply.ret_code |= 1ULL << i;
 389                         }
 390                 }
 391
 392                 DBG2("Health check return value %" PRIx64, reply.ret_code);
 393
 394                 ret = send_unix_sock(new_sock, (void *) &reply, sizeof(reply));
 395                 if (ret < 0) {
 396                         ERR("Failed to send health data back to client");
 397                 }
 398
 399                 /* End of transmission */
 400                 ret = close(new_sock);
 401                 if (ret) {
 402                         PERROR("close");
 403                 }
 404                 new_sock = -1;
 405         }
 406
 407 exit:
 408 error:
 409         if (err) {
 410                 ERR("Health error occurred in %s", __func__);
 411         }
 412         DBG("Health check thread dying");
 413         unlink(health_unix_sock_path);
 414         if (sock >= 0) {
 415                 ret = close(sock);
 416                 if (ret) {
 417                         PERROR("close");
 418                 }
 419         }
 420
 421         /*
 422          * We do NOT rmdir rundir nor the relayd path because there are
 423          * other processes using them.
 424          */
 425
 426         lttng_poll_clean(&events);
 427
 428         rcu_unregister_thread();
 429         return NULL;
 430 }