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 #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-consumerd.h"
  53 #include "health-consumerd.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
  91 int setup_health_path(void)
  92 {
  93         int is_root, ret = 0;
  94         enum lttng_consumer_type type;
  95         const char *home_path;
  96
  97         type = lttng_consumer_get_type();
  98         is_root = !getuid();
  99
 100         if (is_root) {
 101                 if (strlen(health_unix_sock_path) != 0) {
 102                         goto end;
 103                 }
 104                 switch (type) {
 105                 case LTTNG_CONSUMER_KERNEL:
 106                         snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 107                                 DEFAULT_GLOBAL_KCONSUMER_HEALTH_UNIX_SOCK);
 108                         break;
 109                 case LTTNG_CONSUMER64_UST:
 110                         snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 111                                 DEFAULT_GLOBAL_USTCONSUMER64_HEALTH_UNIX_SOCK);
 112                         break;
 113                 case LTTNG_CONSUMER32_UST:
 114                         snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 115                                 DEFAULT_GLOBAL_USTCONSUMER32_HEALTH_UNIX_SOCK);
 116                         break;
 117                 default:
 118                         ret = -EINVAL;
 119                         goto end;
 120                 }
 121         } else {
 122                 home_path = utils_get_home_dir();
 123                 if (home_path == NULL) {
 124                         /* TODO: Add --socket PATH option */
 125                         ERR("Can't get HOME directory for sockets creation.");
 126                         ret = -EPERM;
 127                         goto end;
 128                 }
 129
 130                 /* Set health check Unix path */
 131                 if (strlen(health_unix_sock_path) != 0) {
 132                         goto end;
 133                 }
 134                 switch (type) {
 135                 case LTTNG_CONSUMER_KERNEL:
 136                         snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 137                                 DEFAULT_HOME_KCONSUMER_HEALTH_UNIX_SOCK, home_path);
 138                         break;
 139                 case LTTNG_CONSUMER64_UST:
 140                         snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 141                                 DEFAULT_HOME_USTCONSUMER64_HEALTH_UNIX_SOCK, home_path);
 142                         break;
 143                 case LTTNG_CONSUMER32_UST:
 144                         snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 145                                 DEFAULT_HOME_USTCONSUMER32_HEALTH_UNIX_SOCK, home_path);
 146                         break;
 147                 default:
 148                         ret = -EINVAL;
 149                         goto end;
 150                 }
 151         }
 152 end:
 153         return ret;
 154 }
 155
 156 /*
 157  * Thread managing health check socket.
 158  */
 159 void *thread_manage_health(void *data)
 160 {
 161         int sock = -1, new_sock = -1, ret, i, pollfd, err = -1;
 162         uint32_t revents, nb_fd;
 163         struct lttng_poll_event events;
 164         struct health_comm_msg msg;
 165         struct health_comm_reply reply;
 166         int is_root;
 167
 168         DBG("[thread] Manage health check started");
 169
 170         setup_health_path();
 171
 172         rcu_register_thread();
 173
 174         /* We might hit an error path before this is created. */
 175         lttng_poll_init(&events);
 176
 177         /* Create unix socket */
 178         sock = lttcomm_create_unix_sock(health_unix_sock_path);
 179         if (sock < 0) {
 180                 ERR("Unable to create health check Unix socket");
 181                 ret = -1;
 182                 goto error;
 183         }
 184
 185         is_root = !getuid();
 186         if (is_root) {
 187                 /* lttng health client socket path permissions */
 188                 ret = chown(health_unix_sock_path, 0,
 189                                 utils_get_group_id(tracing_group_name));
 190                 if (ret < 0) {
 191                         ERR("Unable to set group on %s", health_unix_sock_path);
 192                         PERROR("chown");
 193                         ret = -1;
 194                         goto error;
 195                 }
 196
 197                 ret = chmod(health_unix_sock_path,
 198                                 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
 199                 if (ret < 0) {
 200                         ERR("Unable to set permissions on %s", health_unix_sock_path);
 201                         PERROR("chmod");
 202                         ret = -1;
 203                         goto error;
 204                 }
 205         }
 206
 207         /*
 208          * Set the CLOEXEC flag. Return code is useless because either way, the
 209          * show must go on.
 210          */
 211         (void) utils_set_fd_cloexec(sock);
 212
 213         ret = lttcomm_listen_unix_sock(sock);
 214         if (ret < 0) {
 215                 goto error;
 216         }
 217
 218         /* Size is set to 1 for the consumer_channel pipe */
 219         ret = lttng_poll_create(&events, 2, LTTNG_CLOEXEC);
 220         if (ret < 0) {
 221                 ERR("Poll set creation failed");
 222                 goto error;
 223         }
 224
 225         ret = lttng_poll_add(&events, health_quit_pipe[0], LPOLLIN);
 226         if (ret < 0) {
 227                 goto error;
 228         }
 229
 230         /* Add the application registration socket */
 231         ret = lttng_poll_add(&events, sock, LPOLLIN | LPOLLPRI);
 232         if (ret < 0) {
 233                 goto error;
 234         }
 235
 236         /* Perform prior memory accesses before decrementing ready */
 237         cmm_smp_mb__before_uatomic_dec();
 238         uatomic_dec(&lttng_consumer_ready);
 239
 240         while (1) {
 241                 DBG("Health check ready");
 242
 243                 /* Inifinite blocking call, waiting for transmission */
 244 restart:
 245                 ret = lttng_poll_wait(&events, -1);
 246                 if (ret < 0) {
 247                         /*
 248                          * Restart interrupted system call.
 249                          */
 250                         if (errno == EINTR) {
 251                                 goto restart;
 252                         }
 253                         goto error;
 254                 }
 255
 256                 nb_fd = ret;
 257
 258                 for (i = 0; i < nb_fd; i++) {
 259                         /* Fetch once the poll data */
 260                         revents = LTTNG_POLL_GETEV(&events, i);
 261                         pollfd = LTTNG_POLL_GETFD(&events, i);
 262
 263                         /* Thread quit pipe has been closed. Killing thread. */
 264                         ret = check_health_quit_pipe(pollfd, revents);
 265                         if (ret) {
 266                                 err = 0;
 267                                 goto exit;
 268                         }
 269
 270                         /* Event on the registration socket */
 271                         if (pollfd == sock) {
 272                                 if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
 273                                         ERR("Health socket poll error");
 274                                         goto error;
 275                                 }
 276                         }
 277                 }
 278
 279                 new_sock = lttcomm_accept_unix_sock(sock);
 280                 if (new_sock < 0) {
 281                         goto error;
 282                 }
 283
 284                 /*
 285                  * Set the CLOEXEC flag. Return code is useless because either way, the
 286                  * show must go on.
 287                  */
 288                 (void) utils_set_fd_cloexec(new_sock);
 289
 290                 DBG("Receiving data from client for health...");
 291                 ret = lttcomm_recv_unix_sock(new_sock, (void *)&msg, sizeof(msg));
 292                 if (ret <= 0) {
 293                         DBG("Nothing recv() from client... continuing");
 294                         ret = close(new_sock);
 295                         if (ret) {
 296                                 PERROR("close");
 297                         }
 298                         new_sock = -1;
 299                         continue;
 300                 }
 301
 302                 rcu_thread_online();
 303
 304                 assert(msg.cmd == HEALTH_CMD_CHECK);
 305
 306                 reply.ret_code = 0;
 307                 for (i = 0; i < NR_HEALTH_CONSUMERD_TYPES; i++) {
 308                         /*
 309                          * health_check_state return 0 if thread is in
 310                          * error.
 311                          */
 312                         if (!health_check_state(health_consumerd, i)) {
 313                                 reply.ret_code |= 1ULL << i;
 314                         }
 315                 }
 316
 317                 DBG2("Health check return value %" PRIx64, reply.ret_code);
 318
 319                 ret = send_unix_sock(new_sock, (void *) &reply, sizeof(reply));
 320                 if (ret < 0) {
 321                         ERR("Failed to send health data back to client");
 322                 }
 323
 324                 /* End of transmission */
 325                 ret = close(new_sock);
 326                 if (ret) {
 327                         PERROR("close");
 328                 }
 329                 new_sock = -1;
 330         }
 331
 332 exit:
 333 error:
 334         if (err) {
 335                 ERR("Health error occurred in %s", __func__);
 336         }
 337         DBG("Health check thread dying");
 338         unlink(health_unix_sock_path);
 339         if (sock >= 0) {
 340                 ret = close(sock);
 341                 if (ret) {
 342                         PERROR("close");
 343                 }
 344         }
 345
 346         lttng_poll_clean(&events);
 347
 348         rcu_unregister_thread();
 349         return NULL;
 350 }