src/bin/lttng-consumerd/health-consumerd.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-consumerd.h"
  54 #include "health-consumerd.h"
  55
  56 /* Global health check unix path */
  57 static 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
  92 int setup_health_path(void)
  93 {
  94         int is_root, ret = 0;
  95         enum lttng_consumer_type type;
  96         const char *home_path;
  97
  98         type = lttng_consumer_get_type();
  99         is_root = !getuid();
 100
 101         if (is_root) {
 102                 if (strlen(health_unix_sock_path) != 0) {
 103                         goto end;
 104                 }
 105                 switch (type) {
 106                 case LTTNG_CONSUMER_KERNEL:
 107                         snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 108                                 DEFAULT_GLOBAL_KCONSUMER_HEALTH_UNIX_SOCK);
 109                         break;
 110                 case LTTNG_CONSUMER64_UST:
 111                         snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 112                                 DEFAULT_GLOBAL_USTCONSUMER64_HEALTH_UNIX_SOCK);
 113                         break;
 114                 case LTTNG_CONSUMER32_UST:
 115                         snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 116                                 DEFAULT_GLOBAL_USTCONSUMER32_HEALTH_UNIX_SOCK);
 117                         break;
 118                 default:
 119                         ret = -EINVAL;
 120                         goto end;
 121                 }
 122         } else {
 123                 home_path = utils_get_home_dir();
 124                 if (home_path == NULL) {
 125                         /* TODO: Add --socket PATH option */
 126                         ERR("Can't get HOME directory for sockets creation.");
 127                         ret = -EPERM;
 128                         goto end;
 129                 }
 130
 131                 /* Set health check Unix path */
 132                 if (strlen(health_unix_sock_path) != 0) {
 133                         goto end;
 134                 }
 135                 switch (type) {
 136                 case LTTNG_CONSUMER_KERNEL:
 137                         snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 138                                 DEFAULT_HOME_KCONSUMER_HEALTH_UNIX_SOCK, home_path);
 139                         break;
 140                 case LTTNG_CONSUMER64_UST:
 141                         snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 142                                 DEFAULT_HOME_USTCONSUMER64_HEALTH_UNIX_SOCK, home_path);
 143                         break;
 144                 case LTTNG_CONSUMER32_UST:
 145                         snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 146                                 DEFAULT_HOME_USTCONSUMER32_HEALTH_UNIX_SOCK, home_path);
 147                         break;
 148                 default:
 149                         ret = -EINVAL;
 150                         goto end;
 151                 }
 152         }
 153 end:
 154         return ret;
 155 }
 156
 157 /*
 158  * Thread managing health check socket.
 159  */
 160 void *thread_manage_health(void *data)
 161 {
 162         int sock = -1, new_sock = -1, ret, i, pollfd, err = -1;
 163         uint32_t revents, nb_fd;
 164         struct lttng_poll_event events;
 165         struct health_comm_msg msg;
 166         struct health_comm_reply reply;
 167         int is_root;
 168
 169         DBG("[thread] Manage health check started");
 170
 171         setup_health_path();
 172
 173         rcu_register_thread();
 174
 175         /* We might hit an error path before this is created. */
 176         lttng_poll_init(&events);
 177
 178         /* Create unix socket */
 179         sock = lttcomm_create_unix_sock(health_unix_sock_path);
 180         if (sock < 0) {
 181                 ERR("Unable to create health check Unix socket");
 182                 ret = -1;
 183                 goto error;
 184         }
 185
 186         is_root = !getuid();
 187         if (is_root) {
 188                 /* lttng health client socket path permissions */
 189                 ret = chown(health_unix_sock_path, 0,
 190                                 utils_get_group_id(tracing_group_name));
 191                 if (ret < 0) {
 192                         ERR("Unable to set group on %s", health_unix_sock_path);
 193                         PERROR("chown");
 194                         ret = -1;
 195                         goto error;
 196                 }
 197
 198                 ret = chmod(health_unix_sock_path,
 199                                 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
 200                 if (ret < 0) {
 201                         ERR("Unable to set permissions on %s", health_unix_sock_path);
 202                         PERROR("chmod");
 203                         ret = -1;
 204                         goto error;
 205                 }
 206         }
 207
 208         /*
 209          * Set the CLOEXEC flag. Return code is useless because either way, the
 210          * show must go on.
 211          */
 212         (void) utils_set_fd_cloexec(sock);
 213
 214         ret = lttcomm_listen_unix_sock(sock);
 215         if (ret < 0) {
 216                 goto error;
 217         }
 218
 219         /* Size is set to 1 for the consumer_channel pipe */
 220         ret = lttng_poll_create(&events, 2, LTTNG_CLOEXEC);
 221         if (ret < 0) {
 222                 ERR("Poll set creation failed");
 223                 goto error;
 224         }
 225
 226         ret = lttng_poll_add(&events, health_quit_pipe[0], LPOLLIN);
 227         if (ret < 0) {
 228                 goto error;
 229         }
 230
 231         /* Add the application registration socket */
 232         ret = lttng_poll_add(&events, sock, LPOLLIN | LPOLLPRI);
 233         if (ret < 0) {
 234                 goto error;
 235         }
 236
 237         /* Perform prior memory accesses before decrementing ready */
 238         cmm_smp_mb__before_uatomic_dec();
 239         uatomic_dec(&lttng_consumer_ready);
 240
 241         while (1) {
 242                 DBG("Health check ready");
 243
 244                 /* Inifinite blocking call, waiting for transmission */
 245 restart:
 246                 ret = lttng_poll_wait(&events, -1);
 247                 if (ret < 0) {
 248                         /*
 249                          * Restart interrupted system call.
 250                          */
 251                         if (errno == EINTR) {
 252                                 goto restart;
 253                         }
 254                         goto error;
 255                 }
 256
 257                 nb_fd = ret;
 258
 259                 for (i = 0; i < nb_fd; i++) {
 260                         /* Fetch once the poll data */
 261                         revents = LTTNG_POLL_GETEV(&events, i);
 262                         pollfd = LTTNG_POLL_GETFD(&events, i);
 263
 264                         if (!revents) {
 265                                 /* No activity for this FD (poll implementation). */
 266                                 continue;
 267                         }
 268
 269                         /* Thread quit pipe has been closed. Killing thread. */
 270                         ret = check_health_quit_pipe(pollfd, revents);
 271                         if (ret) {
 272                                 err = 0;
 273                                 goto exit;
 274                         }
 275
 276                         /* Event on the registration socket */
 277                         if (pollfd == sock) {
 278                                 if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
 279                                         ERR("Health socket poll error");
 280                                         goto error;
 281                                 }
 282                         }
 283                 }
 284
 285                 new_sock = lttcomm_accept_unix_sock(sock);
 286                 if (new_sock < 0) {
 287                         goto error;
 288                 }
 289
 290                 /*
 291                  * Set the CLOEXEC flag. Return code is useless because either way, the
 292                  * show must go on.
 293                  */
 294                 (void) utils_set_fd_cloexec(new_sock);
 295
 296                 DBG("Receiving data from client for health...");
 297                 ret = lttcomm_recv_unix_sock(new_sock, (void *)&msg, sizeof(msg));
 298                 if (ret <= 0) {
 299                         DBG("Nothing recv() from client... continuing");
 300                         ret = close(new_sock);
 301                         if (ret) {
 302                                 PERROR("close");
 303                         }
 304                         new_sock = -1;
 305                         continue;
 306                 }
 307
 308                 rcu_thread_online();
 309
 310                 assert(msg.cmd == HEALTH_CMD_CHECK);
 311
 312                 memset(&reply, 0, sizeof(reply));
 313                 for (i = 0; i < NR_HEALTH_CONSUMERD_TYPES; i++) {
 314                         /*
 315                          * health_check_state return 0 if thread is in
 316                          * error.
 317                          */
 318                         if (!health_check_state(health_consumerd, i)) {
 319                                 reply.ret_code |= 1ULL << i;
 320                         }
 321                 }
 322
 323                 DBG("Health check return value %" PRIx64, reply.ret_code);
 324
 325                 ret = send_unix_sock(new_sock, (void *) &reply, sizeof(reply));
 326                 if (ret < 0) {
 327                         ERR("Failed to send health data back to client");
 328                 }
 329
 330                 /* End of transmission */
 331                 ret = close(new_sock);
 332                 if (ret) {
 333                         PERROR("close");
 334                 }
 335                 new_sock = -1;
 336         }
 337
 338 exit:
 339 error:
 340         if (err) {
 341                 ERR("Health error occurred in %s", __func__);
 342         }
 343         DBG("Health check thread dying");
 344         unlink(health_unix_sock_path);
 345         if (sock >= 0) {
 346                 ret = close(sock);
 347                 if (ret) {
 348                         PERROR("close");
 349                 }
 350         }
 351
 352         lttng_poll_clean(&events);
 353
 354         rcu_unregister_thread();
 355         return NULL;
 356 }