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 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 setup_health_path(void)
 134 {
 135         int is_root, ret = 0;
 136         char *home_path = NULL, *rundir = NULL, *relayd_path;
 137
 138         is_root = !getuid();
 139
 140         if (is_root) {
 141                 rundir = strdup(DEFAULT_LTTNG_RUNDIR);
 142         } else {
 143                 /*
 144                  * Create rundir from home path. This will create something like
 145                  * $HOME/.lttng
 146                  */
 147                 home_path = utils_get_home_dir();
 148
 149                 if (home_path == NULL) {
 150                         /* TODO: Add --socket PATH option */
 151                         ERR("Can't get HOME directory for sockets creation.");
 152                         ret = -EPERM;
 153                         goto end;
 154                 }
 155
 156                 ret = asprintf(&rundir, DEFAULT_LTTNG_HOME_RUNDIR, home_path);
 157                 if (ret < 0) {
 158                         ret = -ENOMEM;
 159                         goto end;
 160                 }
 161         }
 162
 163         ret = asprintf(&relayd_path, DEFAULT_RELAYD_PATH, rundir);
 164         if (ret < 0) {
 165                 ret = -ENOMEM;
 166                 goto end;
 167         }
 168
 169         ret = create_lttng_rundir_with_perm(rundir);
 170         if (ret < 0) {
 171                 goto end;
 172         }
 173
 174         ret = create_lttng_rundir_with_perm(relayd_path);
 175         if (ret < 0) {
 176                 goto end;
 177         }
 178
 179         if (is_root) {
 180                 if (strlen(health_unix_sock_path) != 0) {
 181                         goto end;
 182                 }
 183                 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 184                         DEFAULT_GLOBAL_RELAY_HEALTH_UNIX_SOCK,
 185                         getpid());
 186         } else {
 187                 /* Set health check Unix path */
 188                 if (strlen(health_unix_sock_path) != 0) {
 189                         goto end;
 190                 }
 191
 192                 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 193                         DEFAULT_HOME_RELAY_HEALTH_UNIX_SOCK,
 194                         home_path, getpid());
 195         }
 196
 197 end:
 198         free(rundir);
 199         return ret;
 200 }
 201
 202 /*
 203  * Thread managing health check socket.
 204  */
 205 void *thread_manage_health(void *data)
 206 {
 207         int sock = -1, new_sock = -1, ret, i, pollfd, err = -1;
 208         uint32_t revents, nb_fd;
 209         struct lttng_poll_event events;
 210         struct health_comm_msg msg;
 211         struct health_comm_reply reply;
 212         int is_root;
 213
 214         DBG("[thread] Manage health check started");
 215
 216         setup_health_path();
 217
 218         rcu_register_thread();
 219
 220         /* We might hit an error path before this is created. */
 221         lttng_poll_init(&events);
 222
 223         /* Create unix socket */
 224         sock = lttcomm_create_unix_sock(health_unix_sock_path);
 225         if (sock < 0) {
 226                 ERR("Unable to create health check Unix socket");
 227                 ret = -1;
 228                 goto error;
 229         }
 230
 231         is_root = !getuid();
 232         if (is_root) {
 233                 /* lttng health client socket path permissions */
 234                 ret = chown(health_unix_sock_path, 0,
 235                                 utils_get_group_id(tracing_group_name));
 236                 if (ret < 0) {
 237                         ERR("Unable to set group on %s", health_unix_sock_path);
 238                         PERROR("chown");
 239                         ret = -1;
 240                         goto error;
 241                 }
 242
 243                 ret = chmod(health_unix_sock_path,
 244                                 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
 245                 if (ret < 0) {
 246                         ERR("Unable to set permissions on %s", health_unix_sock_path);
 247                         PERROR("chmod");
 248                         ret = -1;
 249                         goto error;
 250                 }
 251         }
 252
 253         /*
 254          * Set the CLOEXEC flag. Return code is useless because either way, the
 255          * show must go on.
 256          */
 257         (void) utils_set_fd_cloexec(sock);
 258
 259         ret = lttcomm_listen_unix_sock(sock);
 260         if (ret < 0) {
 261                 goto error;
 262         }
 263
 264         /* Size is set to 1 for the consumer_channel pipe */
 265         ret = lttng_poll_create(&events, 2, LTTNG_CLOEXEC);
 266         if (ret < 0) {
 267                 ERR("Poll set creation failed");
 268                 goto error;
 269         }
 270
 271         ret = lttng_poll_add(&events, health_quit_pipe[0], LPOLLIN);
 272         if (ret < 0) {
 273                 goto error;
 274         }
 275
 276         /* Add the application registration socket */
 277         ret = lttng_poll_add(&events, sock, LPOLLIN | LPOLLPRI);
 278         if (ret < 0) {
 279                 goto error;
 280         }
 281
 282         while (1) {
 283                 DBG("Health check ready");
 284
 285                 /* Inifinite blocking call, waiting for transmission */
 286 restart:
 287                 ret = lttng_poll_wait(&events, -1);
 288                 if (ret < 0) {
 289                         /*
 290                          * Restart interrupted system call.
 291                          */
 292                         if (errno == EINTR) {
 293                                 goto restart;
 294                         }
 295                         goto error;
 296                 }
 297
 298                 nb_fd = ret;
 299
 300                 for (i = 0; i < nb_fd; i++) {
 301                         /* Fetch once the poll data */
 302                         revents = LTTNG_POLL_GETEV(&events, i);
 303                         pollfd = LTTNG_POLL_GETFD(&events, i);
 304
 305                         /* Thread quit pipe has been closed. Killing thread. */
 306                         ret = check_health_quit_pipe(pollfd, revents);
 307                         if (ret) {
 308                                 err = 0;
 309                                 goto exit;
 310                         }
 311
 312                         /* Event on the registration socket */
 313                         if (pollfd == sock) {
 314                                 if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
 315                                         ERR("Health socket poll error");
 316                                         goto error;
 317                                 }
 318                         }
 319                 }
 320
 321                 new_sock = lttcomm_accept_unix_sock(sock);
 322                 if (new_sock < 0) {
 323                         goto error;
 324                 }
 325
 326                 /*
 327                  * Set the CLOEXEC flag. Return code is useless because either way, the
 328                  * show must go on.
 329                  */
 330                 (void) utils_set_fd_cloexec(new_sock);
 331
 332                 DBG("Receiving data from client for health...");
 333                 ret = lttcomm_recv_unix_sock(new_sock, (void *)&msg, sizeof(msg));
 334                 if (ret <= 0) {
 335                         DBG("Nothing recv() from client... continuing");
 336                         ret = close(new_sock);
 337                         if (ret) {
 338                                 PERROR("close");
 339                         }
 340                         new_sock = -1;
 341                         continue;
 342                 }
 343
 344                 rcu_thread_online();
 345
 346                 assert(msg.cmd == HEALTH_CMD_CHECK);
 347
 348                 reply.ret_code = 0;
 349                 for (i = 0; i < NR_HEALTH_RELAYD_TYPES; i++) {
 350                         /*
 351                          * health_check_state return 0 if thread is in
 352                          * error.
 353                          */
 354                         if (!health_check_state(health_relayd, i)) {
 355                                 reply.ret_code |= 1ULL << i;
 356                         }
 357                 }
 358
 359                 DBG2("Health check return value %" PRIx64, reply.ret_code);
 360
 361                 ret = send_unix_sock(new_sock, (void *) &reply, sizeof(reply));
 362                 if (ret < 0) {
 363                         ERR("Failed to send health data back to client");
 364                 }
 365
 366                 /* End of transmission */
 367                 ret = close(new_sock);
 368                 if (ret) {
 369                         PERROR("close");
 370                 }
 371                 new_sock = -1;
 372         }
 373
 374 exit:
 375 error:
 376         if (err) {
 377                 ERR("Health error occurred in %s", __func__);
 378         }
 379         DBG("Health check thread dying");
 380         unlink(health_unix_sock_path);
 381         if (sock >= 0) {
 382                 ret = close(sock);
 383                 if (ret) {
 384                         PERROR("close");
 385                 }
 386         }
 387
 388         /*
 389          * We do NOT rmdir rundir nor the relayd path because there are
 390          * other processes using them.
 391          */
 392
 393         lttng_poll_clean(&events);
 394
 395         rcu_unregister_thread();
 396         return NULL;
 397 }