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         char *home_path = NULL, *rundir = NULL, *relayd_path = NULL;
 159
 160         ret = parse_health_env();
 161         if (ret) {
 162                 return ret;
 163         }
 164
 165         is_root = !getuid();
 166
 167         if (is_root) {
 168                 rundir = strdup(DEFAULT_LTTNG_RUNDIR);
 169                 if (!rundir) {
 170                         ret = -ENOMEM;
 171                         goto end;
 172                 }
 173         } else {
 174                 /*
 175                  * Create rundir from home path. This will create something like
 176                  * $HOME/.lttng
 177                  */
 178                 home_path = utils_get_home_dir();
 179
 180                 if (home_path == NULL) {
 181                         /* TODO: Add --socket PATH option */
 182                         ERR("Can't get HOME directory for sockets creation.");
 183                         ret = -EPERM;
 184                         goto end;
 185                 }
 186
 187                 ret = asprintf(&rundir, DEFAULT_LTTNG_HOME_RUNDIR, home_path);
 188                 if (ret < 0) {
 189                         ret = -ENOMEM;
 190                         goto end;
 191                 }
 192         }
 193
 194         ret = asprintf(&relayd_path, DEFAULT_RELAYD_PATH, rundir);
 195         if (ret < 0) {
 196                 ret = -ENOMEM;
 197                 goto end;
 198         }
 199
 200         ret = create_lttng_rundir_with_perm(rundir);
 201         if (ret < 0) {
 202                 goto end;
 203         }
 204
 205         ret = create_lttng_rundir_with_perm(relayd_path);
 206         if (ret < 0) {
 207                 goto end;
 208         }
 209
 210         if (is_root) {
 211                 if (strlen(health_unix_sock_path) != 0) {
 212                         goto end;
 213                 }
 214                 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 215                         DEFAULT_GLOBAL_RELAY_HEALTH_UNIX_SOCK,
 216                         (int) getpid());
 217         } else {
 218                 /* Set health check Unix path */
 219                 if (strlen(health_unix_sock_path) != 0) {
 220                         goto end;
 221                 }
 222
 223                 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 224                         DEFAULT_HOME_RELAY_HEALTH_UNIX_SOCK,
 225                         home_path, (int) getpid());
 226         }
 227
 228 end:
 229         free(rundir);
 230         free(relayd_path);
 231         return ret;
 232 }
 233
 234 /*
 235  * Thread managing health check socket.
 236  */
 237 void *thread_manage_health(void *data)
 238 {
 239         int sock = -1, new_sock = -1, ret, i, pollfd, err = -1;
 240         uint32_t revents, nb_fd;
 241         struct lttng_poll_event events;
 242         struct health_comm_msg msg;
 243         struct health_comm_reply reply;
 244         int is_root;
 245
 246         DBG("[thread] Manage health check started");
 247
 248         setup_health_path();
 249
 250         rcu_register_thread();
 251
 252         /* We might hit an error path before this is created. */
 253         lttng_poll_init(&events);
 254
 255         /* Create unix socket */
 256         sock = lttcomm_create_unix_sock(health_unix_sock_path);
 257         if (sock < 0) {
 258                 ERR("Unable to create health check Unix socket");
 259                 err = -1;
 260                 goto error;
 261         }
 262
 263         is_root = !getuid();
 264         if (is_root) {
 265                 /* lttng health client socket path permissions */
 266                 gid_t gid;
 267
 268                 ret = utils_get_group_id(tracing_group_name, true, &gid);
 269                 if (ret) {
 270                         /* Default to root group. */
 271                         gid = 0;
 272                 }
 273
 274                 ret = chown(health_unix_sock_path, 0, gid);
 275                 if (ret < 0) {
 276                         ERR("Unable to set group on %s", health_unix_sock_path);
 277                         PERROR("chown");
 278                         err = -1;
 279                         goto error;
 280                 }
 281
 282                 ret = chmod(health_unix_sock_path,
 283                                 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
 284                 if (ret < 0) {
 285                         ERR("Unable to set permissions on %s", health_unix_sock_path);
 286                         PERROR("chmod");
 287                         err = -1;
 288                         goto error;
 289                 }
 290         }
 291
 292         /*
 293          * Set the CLOEXEC flag. Return code is useless because either way, the
 294          * show must go on.
 295          */
 296         (void) utils_set_fd_cloexec(sock);
 297
 298         ret = lttcomm_listen_unix_sock(sock);
 299         if (ret < 0) {
 300                 goto error;
 301         }
 302
 303         /* Size is set to 1 for the consumer_channel pipe */
 304         ret = lttng_poll_create(&events, 2, LTTNG_CLOEXEC);
 305         if (ret < 0) {
 306                 ERR("Poll set creation failed");
 307                 goto error;
 308         }
 309
 310         ret = lttng_poll_add(&events, health_quit_pipe[0], LPOLLIN);
 311         if (ret < 0) {
 312                 goto error;
 313         }
 314
 315         /* Add the application registration socket */
 316         ret = lttng_poll_add(&events, sock, LPOLLIN | LPOLLPRI);
 317         if (ret < 0) {
 318                 goto error;
 319         }
 320
 321         lttng_relay_notify_ready();
 322
 323         while (1) {
 324                 DBG("Health check ready");
 325
 326                 /* Inifinite blocking call, waiting for transmission */
 327 restart:
 328                 ret = lttng_poll_wait(&events, -1);
 329                 if (ret < 0) {
 330                         /*
 331                          * Restart interrupted system call.
 332                          */
 333                         if (errno == EINTR) {
 334                                 goto restart;
 335                         }
 336                         goto error;
 337                 }
 338
 339                 nb_fd = ret;
 340
 341                 for (i = 0; i < nb_fd; i++) {
 342                         /* Fetch once the poll data */
 343                         revents = LTTNG_POLL_GETEV(&events, i);
 344                         pollfd = LTTNG_POLL_GETFD(&events, i);
 345
 346                         /* Thread quit pipe has been closed. Killing thread. */
 347                         ret = check_health_quit_pipe(pollfd, revents);
 348                         if (ret) {
 349                                 err = 0;
 350                                 goto exit;
 351                         }
 352
 353                         /* Event on the registration socket */
 354                         if (pollfd == sock) {
 355                                 if (revents & LPOLLIN) {
 356                                         continue;
 357                                 } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
 358                                         ERR("Health socket poll error");
 359                                         goto error;
 360                                 } else {
 361                                         ERR("Unexpected poll events %u for sock %d", revents, pollfd);
 362                                         goto error;
 363                                 }
 364                         }
 365                 }
 366
 367                 new_sock = lttcomm_accept_unix_sock(sock);
 368                 if (new_sock < 0) {
 369                         goto error;
 370                 }
 371
 372                 /*
 373                  * Set the CLOEXEC flag. Return code is useless because either way, the
 374                  * show must go on.
 375                  */
 376                 (void) utils_set_fd_cloexec(new_sock);
 377
 378                 DBG("Receiving data from client for health...");
 379                 ret = lttcomm_recv_unix_sock(new_sock, (void *)&msg, sizeof(msg));
 380                 if (ret <= 0) {
 381                         DBG("Nothing recv() from client... continuing");
 382                         ret = close(new_sock);
 383                         if (ret) {
 384                                 PERROR("close");
 385                         }
 386                         new_sock = -1;
 387                         continue;
 388                 }
 389
 390                 rcu_thread_online();
 391
 392                 assert(msg.cmd == HEALTH_CMD_CHECK);
 393
 394                 memset(&reply, 0, sizeof(reply));
 395                 for (i = 0; i < NR_HEALTH_RELAYD_TYPES; i++) {
 396                         /*
 397                          * health_check_state return 0 if thread is in
 398                          * error.
 399                          */
 400                         if (!health_check_state(health_relayd, i)) {
 401                                 reply.ret_code |= 1ULL << i;
 402                         }
 403                 }
 404
 405                 DBG2("Health check return value %" PRIx64, reply.ret_code);
 406
 407                 ret = send_unix_sock(new_sock, (void *) &reply, sizeof(reply));
 408                 if (ret < 0) {
 409                         ERR("Failed to send health data back to client");
 410                 }
 411
 412                 /* End of transmission */
 413                 ret = close(new_sock);
 414                 if (ret) {
 415                         PERROR("close");
 416                 }
 417                 new_sock = -1;
 418         }
 419
 420 error:
 421         lttng_relay_stop_threads();
 422 exit:
 423         if (err) {
 424                 ERR("Health error occurred in %s", __func__);
 425         }
 426         DBG("Health check thread dying");
 427         unlink(health_unix_sock_path);
 428         if (sock >= 0) {
 429                 ret = close(sock);
 430                 if (ret) {
 431                         PERROR("close");
 432                 }
 433         }
 434
 435         /*
 436          * We do NOT rmdir rundir nor the relayd path because there are
 437          * other processes using them.
 438          */
 439
 440         lttng_poll_clean(&events);
 441
 442         rcu_unregister_thread();
 443         return NULL;
 444 }