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 static char *relayd_path;
  58 static char *lttng_rundir;
  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 setup_health_path(void)
 136 {
 137         int is_root, ret = 0;
 138         char *home_path = NULL;
 139
 140         is_root = !getuid();
 141
 142         if (is_root) {
 143                 lttng_rundir = strdup(DEFAULT_LTTNG_RUNDIR);
 144         } else {
 145                 /*
 146                  * Create rundir from home path. This will create something like
 147                  * $HOME/.lttng
 148                  */
 149                 home_path = utils_get_home_dir();
 150
 151                 if (home_path == NULL) {
 152                         /* TODO: Add --socket PATH option */
 153                         ERR("Can't get HOME directory for sockets creation.");
 154                         ret = -EPERM;
 155                         goto end;
 156                 }
 157
 158                 ret = asprintf(&lttng_rundir, DEFAULT_LTTNG_HOME_RUNDIR, home_path);
 159                 if (ret < 0) {
 160                         ret = -ENOMEM;
 161                         goto end;
 162                 }
 163         }
 164
 165         ret = asprintf(&relayd_path, DEFAULT_RELAYD_PATH, lttng_rundir);
 166         if (ret < 0) {
 167                 ret = -ENOMEM;
 168                 goto end;
 169         }
 170
 171         ret = create_lttng_rundir_with_perm(lttng_rundir);
 172         if (ret < 0) {
 173                 goto end;
 174         }
 175
 176         ret = create_lttng_rundir_with_perm(relayd_path);
 177         if (ret < 0) {
 178                 goto end;
 179         }
 180
 181         if (is_root) {
 182                 if (strlen(health_unix_sock_path) != 0) {
 183                         goto end;
 184                 }
 185                 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 186                         DEFAULT_GLOBAL_RELAY_HEALTH_UNIX_SOCK,
 187                         getpid());
 188         } else {
 189                 /* Set health check Unix path */
 190                 if (strlen(health_unix_sock_path) != 0) {
 191                         goto end;
 192                 }
 193
 194                 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 195                         DEFAULT_HOME_RELAY_HEALTH_UNIX_SOCK,
 196                         home_path, getpid());
 197         }
 198
 199 end:
 200         return ret;
 201 }
 202
 203 /*
 204  * Thread managing health check socket.
 205  */
 206 void *thread_manage_health(void *data)
 207 {
 208         int sock = -1, new_sock = -1, ret, i, pollfd, err = -1;
 209         uint32_t revents, nb_fd;
 210         struct lttng_poll_event events;
 211         struct health_comm_msg msg;
 212         struct health_comm_reply reply;
 213         int is_root;
 214
 215         DBG("[thread] Manage health check started");
 216
 217         setup_health_path();
 218
 219         rcu_register_thread();
 220
 221         /* We might hit an error path before this is created. */
 222         lttng_poll_init(&events);
 223
 224         /* Create unix socket */
 225         sock = lttcomm_create_unix_sock(health_unix_sock_path);
 226         if (sock < 0) {
 227                 ERR("Unable to create health check Unix socket");
 228                 ret = -1;
 229                 goto error;
 230         }
 231
 232         is_root = !getuid();
 233         if (is_root) {
 234                 /* lttng health client socket path permissions */
 235                 ret = chown(health_unix_sock_path, 0,
 236                                 utils_get_group_id(tracing_group_name));
 237                 if (ret < 0) {
 238                         ERR("Unable to set group on %s", health_unix_sock_path);
 239                         PERROR("chown");
 240                         ret = -1;
 241                         goto error;
 242                 }
 243
 244                 ret = chmod(health_unix_sock_path,
 245                                 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
 246                 if (ret < 0) {
 247                         ERR("Unable to set permissions on %s", health_unix_sock_path);
 248                         PERROR("chmod");
 249                         ret = -1;
 250                         goto error;
 251                 }
 252         }
 253
 254         /*
 255          * Set the CLOEXEC flag. Return code is useless because either way, the
 256          * show must go on.
 257          */
 258         (void) utils_set_fd_cloexec(sock);
 259
 260         ret = lttcomm_listen_unix_sock(sock);
 261         if (ret < 0) {
 262                 goto error;
 263         }
 264
 265         /* Size is set to 1 for the consumer_channel pipe */
 266         ret = lttng_poll_create(&events, 2, LTTNG_CLOEXEC);
 267         if (ret < 0) {
 268                 ERR("Poll set creation failed");
 269                 goto error;
 270         }
 271
 272         ret = lttng_poll_add(&events, health_quit_pipe[0], LPOLLIN);
 273         if (ret < 0) {
 274                 goto error;
 275         }
 276
 277         /* Add the application registration socket */
 278         ret = lttng_poll_add(&events, sock, LPOLLIN | LPOLLPRI);
 279         if (ret < 0) {
 280                 goto error;
 281         }
 282
 283         while (1) {
 284                 DBG("Health check ready");
 285
 286                 /* Inifinite blocking call, waiting for transmission */
 287 restart:
 288                 ret = lttng_poll_wait(&events, -1);
 289                 if (ret < 0) {
 290                         /*
 291                          * Restart interrupted system call.
 292                          */
 293                         if (errno == EINTR) {
 294                                 goto restart;
 295                         }
 296                         goto error;
 297                 }
 298
 299                 nb_fd = ret;
 300
 301                 for (i = 0; i < nb_fd; i++) {
 302                         /* Fetch once the poll data */
 303                         revents = LTTNG_POLL_GETEV(&events, i);
 304                         pollfd = LTTNG_POLL_GETFD(&events, i);
 305
 306                         /* Thread quit pipe has been closed. Killing thread. */
 307                         ret = check_health_quit_pipe(pollfd, revents);
 308                         if (ret) {
 309                                 err = 0;
 310                                 goto exit;
 311                         }
 312
 313                         /* Event on the registration socket */
 314                         if (pollfd == sock) {
 315                                 if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
 316                                         ERR("Health socket poll error");
 317                                         goto error;
 318                                 }
 319                         }
 320                 }
 321
 322                 new_sock = lttcomm_accept_unix_sock(sock);
 323                 if (new_sock < 0) {
 324                         goto error;
 325                 }
 326
 327                 /*
 328                  * Set the CLOEXEC flag. Return code is useless because either way, the
 329                  * show must go on.
 330                  */
 331                 (void) utils_set_fd_cloexec(new_sock);
 332
 333                 DBG("Receiving data from client for health...");
 334                 ret = lttcomm_recv_unix_sock(new_sock, (void *)&msg, sizeof(msg));
 335                 if (ret <= 0) {
 336                         DBG("Nothing recv() from client... continuing");
 337                         ret = close(new_sock);
 338                         if (ret) {
 339                                 PERROR("close");
 340                         }
 341                         new_sock = -1;
 342                         continue;
 343                 }
 344
 345                 rcu_thread_online();
 346
 347                 assert(msg.cmd == HEALTH_CMD_CHECK);
 348
 349                 reply.ret_code = 0;
 350                 for (i = 0; i < NR_HEALTH_RELAYD_TYPES; i++) {
 351                         /*
 352                          * health_check_state return 0 if thread is in
 353                          * error.
 354                          */
 355                         if (!health_check_state(health_relayd, i)) {
 356                                 reply.ret_code |= 1ULL << i;
 357                         }
 358                 }
 359
 360                 DBG2("Health check return value %" PRIx64, reply.ret_code);
 361
 362                 ret = send_unix_sock(new_sock, (void *) &reply, sizeof(reply));
 363                 if (ret < 0) {
 364                         ERR("Failed to send health data back to client");
 365                 }
 366
 367                 /* End of transmission */
 368                 ret = close(new_sock);
 369                 if (ret) {
 370                         PERROR("close");
 371                 }
 372                 new_sock = -1;
 373         }
 374
 375 exit:
 376 error:
 377         if (err) {
 378                 ERR("Health error occurred in %s", __func__);
 379         }
 380         DBG("Health check thread dying");
 381         unlink(health_unix_sock_path);
 382         (void) rmdir(relayd_path);
 383         free(relayd_path);
 384         (void) rmdir(lttng_rundir);
 385         free(lttng_rundir);
 386
 387         if (sock >= 0) {
 388                 ret = close(sock);
 389                 if (ret) {
 390                         PERROR("close");
 391                 }
 392         }
 393
 394         lttng_poll_clean(&events);
 395
 396         rcu_unregister_thread();
 397         return NULL;
 398 }