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 _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 <ulimit.h>
  41 #include <inttypes.h>
  42
  43 #include <common/defaults.h>
  44 #include <common/common.h>
  45 #include <common/consumer/consumer.h>
  46 #include <common/consumer/consumer-timer.h>
  47 #include <common/compat/poll.h>
  48 #include <common/sessiond-comm/sessiond-comm.h>
  49 #include <common/utils.h>
  50
  51 #include "lttng-consumerd.h"
  52 #include "health-consumerd.h"
  53
  54 /* Global health check unix path */
  55 static char health_unix_sock_path[PATH_MAX];
  56
  57 int health_quit_pipe[2];
  58
  59 /*
  60  * Check if the thread quit pipe was triggered.
  61  *
  62  * Return 1 if it was triggered else 0;
  63  */
  64 static
  65 int check_health_quit_pipe(int fd, uint32_t events)
  66 {
  67         if (fd == health_quit_pipe[0] && (events & LPOLLIN)) {
  68                 return 1;
  69         }
  70
  71         return 0;
  72 }
  73
  74 /*
  75  * Send data on a unix socket using the liblttsessiondcomm API.
  76  *
  77  * Return lttcomm error code.
  78  */
  79 static int send_unix_sock(int sock, void *buf, size_t len)
  80 {
  81         /* Check valid length */
  82         if (len == 0) {
  83                 return -1;
  84         }
  85
  86         return lttcomm_send_unix_sock(sock, buf, len);
  87 }
  88
  89 static
  90 int setup_health_path(void)
  91 {
  92         int is_root, ret = 0;
  93         enum lttng_consumer_type type;
  94         const char *home_path;
  95
  96         type = lttng_consumer_get_type();
  97         is_root = !getuid();
  98
  99         if (is_root) {
 100                 if (strlen(health_unix_sock_path) != 0) {
 101                         goto end;
 102                 }
 103                 switch (type) {
 104                 case LTTNG_CONSUMER_KERNEL:
 105                         snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 106                                 DEFAULT_GLOBAL_KCONSUMER_HEALTH_UNIX_SOCK);
 107                         break;
 108                 case LTTNG_CONSUMER64_UST:
 109                         snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 110                                 DEFAULT_GLOBAL_USTCONSUMER64_HEALTH_UNIX_SOCK);
 111                         break;
 112                 case LTTNG_CONSUMER32_UST:
 113                         snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 114                                 DEFAULT_GLOBAL_USTCONSUMER32_HEALTH_UNIX_SOCK);
 115                         break;
 116                 default:
 117                         ret = -EINVAL;
 118                         goto end;
 119                 }
 120         } else {
 121                 home_path = utils_get_home_dir();
 122                 if (home_path == NULL) {
 123                         /* TODO: Add --socket PATH option */
 124                         ERR("Can't get HOME directory for sockets creation.");
 125                         ret = -EPERM;
 126                         goto end;
 127                 }
 128
 129                 /* Set health check Unix path */
 130                 if (strlen(health_unix_sock_path) != 0) {
 131                         goto end;
 132                 }
 133                 switch (type) {
 134                 case LTTNG_CONSUMER_KERNEL:
 135                         snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 136                                 DEFAULT_HOME_KCONSUMER_HEALTH_UNIX_SOCK, home_path);
 137                         break;
 138                 case LTTNG_CONSUMER64_UST:
 139                         snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 140                                 DEFAULT_HOME_USTCONSUMER64_HEALTH_UNIX_SOCK, home_path);
 141                         break;
 142                 case LTTNG_CONSUMER32_UST:
 143                         snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
 144                                 DEFAULT_HOME_USTCONSUMER32_HEALTH_UNIX_SOCK, home_path);
 145                         break;
 146                 default:
 147                         ret = -EINVAL;
 148                         goto end;
 149                 }
 150         }
 151 end:
 152         return ret;
 153 }
 154
 155 /*
 156  * Thread managing health check socket.
 157  */
 158 void *thread_manage_health(void *data)
 159 {
 160         int sock = -1, new_sock = -1, ret, i, pollfd, err = -1;
 161         uint32_t revents, nb_fd;
 162         struct lttng_poll_event events;
 163         struct health_comm_msg msg;
 164         struct health_comm_reply reply;
 165         int is_root;
 166
 167         DBG("[thread] Manage health check started");
 168
 169         setup_health_path();
 170
 171         rcu_register_thread();
 172
 173         /* We might hit an error path before this is created. */
 174         lttng_poll_init(&events);
 175
 176         /* Create unix socket */
 177         sock = lttcomm_create_unix_sock(health_unix_sock_path);
 178         if (sock < 0) {
 179                 ERR("Unable to create health check Unix socket");
 180                 err = -1;
 181                 goto error;
 182         }
 183
 184         is_root = !getuid();
 185         if (is_root) {
 186                 /* lttng health client socket path permissions */
 187                 ret = chown(health_unix_sock_path, 0,
 188                                 utils_get_group_id(tracing_group_name));
 189                 if (ret < 0) {
 190                         ERR("Unable to set group on %s", health_unix_sock_path);
 191                         PERROR("chown");
 192                         err = -1;
 193                         goto error;
 194                 }
 195
 196                 ret = chmod(health_unix_sock_path,
 197                                 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
 198                 if (ret < 0) {
 199                         ERR("Unable to set permissions on %s", health_unix_sock_path);
 200                         PERROR("chmod");
 201                         err = -1;
 202                         goto error;
 203                 }
 204         }
 205
 206         /*
 207          * Set the CLOEXEC flag. Return code is useless because either way, the
 208          * show must go on.
 209          */
 210         (void) utils_set_fd_cloexec(sock);
 211
 212         ret = lttcomm_listen_unix_sock(sock);
 213         if (ret < 0) {
 214                 goto error;
 215         }
 216
 217         /* Size is set to 1 for the consumer_channel pipe */
 218         ret = lttng_poll_create(&events, 2, LTTNG_CLOEXEC);
 219         if (ret < 0) {
 220                 ERR("Poll set creation failed");
 221                 goto error;
 222         }
 223
 224         ret = lttng_poll_add(&events, health_quit_pipe[0], LPOLLIN);
 225         if (ret < 0) {
 226                 goto error;
 227         }
 228
 229         /* Add the application registration socket */
 230         ret = lttng_poll_add(&events, sock, LPOLLIN | LPOLLPRI);
 231         if (ret < 0) {
 232                 goto error;
 233         }
 234
 235         /* Perform prior memory accesses before decrementing ready */
 236         cmm_smp_mb__before_uatomic_dec();
 237         uatomic_dec(&lttng_consumer_ready);
 238
 239         while (1) {
 240                 DBG("Health check ready");
 241
 242                 /* Inifinite blocking call, waiting for transmission */
 243 restart:
 244                 ret = lttng_poll_wait(&events, -1);
 245                 if (ret < 0) {
 246                         /*
 247                          * Restart interrupted system call.
 248                          */
 249                         if (errno == EINTR) {
 250                                 goto restart;
 251                         }
 252                         goto error;
 253                 }
 254
 255                 nb_fd = ret;
 256
 257                 for (i = 0; i < nb_fd; i++) {
 258                         /* Fetch once the poll data */
 259                         revents = LTTNG_POLL_GETEV(&events, i);
 260                         pollfd = LTTNG_POLL_GETFD(&events, i);
 261
 262                         if (!revents) {
 263                                 /* No activity for this FD (poll implementation). */
 264                                 continue;
 265                         }
 266
 267                         /* Thread quit pipe has been closed. Killing thread. */
 268                         ret = check_health_quit_pipe(pollfd, revents);
 269                         if (ret) {
 270                                 err = 0;
 271                                 goto exit;
 272                         }
 273
 274                         /* Event on the registration socket */
 275                         if (pollfd == sock) {
 276                                 if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)
 277                                                 && !(revents & LPOLLIN)) {
 278                                         ERR("Health socket poll error");
 279                                         goto error;
 280                                 }
 281                         }
 282                 }
 283
 284                 new_sock = lttcomm_accept_unix_sock(sock);
 285                 if (new_sock < 0) {
 286                         goto error;
 287                 }
 288
 289                 /*
 290                  * Set the CLOEXEC flag. Return code is useless because either way, the
 291                  * show must go on.
 292                  */
 293                 (void) utils_set_fd_cloexec(new_sock);
 294
 295                 DBG("Receiving data from client for health...");
 296                 ret = lttcomm_recv_unix_sock(new_sock, (void *)&msg, sizeof(msg));
 297                 if (ret <= 0) {
 298                         DBG("Nothing recv() from client... continuing");
 299                         ret = close(new_sock);
 300                         if (ret) {
 301                                 PERROR("close");
 302                         }
 303                         new_sock = -1;
 304                         continue;
 305                 }
 306
 307                 rcu_thread_online();
 308
 309                 assert(msg.cmd == HEALTH_CMD_CHECK);
 310
 311                 memset(&reply, 0, sizeof(reply));
 312                 for (i = 0; i < NR_HEALTH_CONSUMERD_TYPES; i++) {
 313                         /*
 314                          * health_check_state return 0 if thread is in
 315                          * error.
 316                          */
 317                         if (!health_check_state(health_consumerd, i)) {
 318                                 reply.ret_code |= 1ULL << i;
 319                         }
 320                 }
 321
 322                 DBG("Health check return value %" PRIx64, reply.ret_code);
 323
 324                 ret = send_unix_sock(new_sock, (void *) &reply, sizeof(reply));
 325                 if (ret < 0) {
 326                         ERR("Failed to send health data back to client");
 327                 }
 328
 329                 /* End of transmission */
 330                 ret = close(new_sock);
 331                 if (ret) {
 332                         PERROR("close");
 333                 }
 334                 new_sock = -1;
 335         }
 336
 337 exit:
 338 error:
 339         if (err) {
 340                 ERR("Health error occurred in %s", __func__);
 341         }
 342         DBG("Health check thread dying");
 343         unlink(health_unix_sock_path);
 344         if (sock >= 0) {
 345                 ret = close(sock);
 346                 if (ret) {
 347                         PERROR("close");
 348                 }
 349         }
 350
 351         lttng_poll_clean(&events);
 352
 353         rcu_unregister_thread();
 354         return NULL;
 355 }