src/bin/lttng-sessiond/agent-thread.c

   1 /*
   2  * Copyright (C) 2013 - David Goulet <dgoulet@efficios.com>
   3  *
   4  * This program is free software; you can redistribute it and/or modify it
   5  * under the terms of the GNU General Public License, version 2 only, as
   6  * published by the Free Software Foundation.
   7  *
   8  * This program is distributed in the hope that it will be useful, but WITHOUT
   9  * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
  10  * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
  11  * more details.
  12  *
  13  * You should have received a copy of the GNU General Public License along with
  14  * this program; if not, write to the Free Software Foundation, Inc., 51
  15  * Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
  16  */
  17
  18 #define _LGPL_SOURCE
  19 #include <assert.h>
  20
  21 #include <common/common.h>
  22 #include <common/sessiond-comm/sessiond-comm.h>
  23 #include <common/uri.h>
  24 #include <common/utils.h>
  25
  26 #include <common/compat/endian.h>
  27
  28 #include "fd-limit.h"
  29 #include "agent-thread.h"
  30 #include "agent.h"
  31 #include "lttng-sessiond.h"
  32 #include "session.h"
  33 #include "utils.h"
  34
  35 /*
  36  * Note that there is not port here. It's set after this URI is parsed so we
  37  * can let the user define a custom one. However, localhost is ALWAYS the
  38  * default listening address.
  39  */
  40 static const char *default_reg_uri =
  41         "tcp://" DEFAULT_NETWORK_VIEWER_BIND_ADDRESS;
  42
  43 /*
  44  * Update agent application using the given socket. This is done just after
  45  * registration was successful.
  46  *
  47  * This is a quite heavy call in terms of locking since the session list lock
  48  * AND session lock are acquired.
  49  */
  50 static void update_agent_app(struct agent_app *app)
  51 {
  52         struct ltt_session *session, *stmp;
  53         struct ltt_session_list *list;
  54
  55         list = session_get_list();
  56         assert(list);
  57
  58         session_lock_list();
  59         cds_list_for_each_entry_safe(session, stmp, &list->head, list) {
  60                 session_lock(session);
  61                 if (session->ust_session) {
  62                         struct agent *agt;
  63
  64                         rcu_read_lock();
  65                         agt = trace_ust_find_agent(session->ust_session, app->domain);
  66                         if (agt) {
  67                                 agent_update(agt, app->sock->fd);
  68                         }
  69                         rcu_read_unlock();
  70                 }
  71                 session_unlock(session);
  72         }
  73         session_unlock_list();
  74 }
  75
  76 /*
  77  * Create and init socket from uri.
  78  */
  79 static struct lttcomm_sock *init_tcp_socket(void)
  80 {
  81         int ret;
  82         struct lttng_uri *uri = NULL;
  83         struct lttcomm_sock *sock = NULL;
  84
  85         /*
  86          * This should never fail since the URI is hardcoded and the port is set
  87          * before this thread is launched.
  88          */
  89         ret = uri_parse(default_reg_uri, &uri);
  90         assert(ret);
  91         assert(config.agent_tcp_port);
  92         uri->port = config.agent_tcp_port;
  93
  94         sock = lttcomm_alloc_sock_from_uri(uri);
  95         uri_free(uri);
  96         if (sock == NULL) {
  97                 ERR("[agent-thread] agent allocating TCP socket");
  98                 goto error;
  99         }
 100
 101         ret = lttcomm_create_sock(sock);
 102         if (ret < 0) {
 103                 goto error;
 104         }
 105
 106         ret = sock->ops->bind(sock);
 107         if (ret < 0) {
 108                 WARN("Another session daemon is using this agent port. Agent support "
 109                                 "will be deactivated to prevent interfering with the tracing.");
 110                 goto error;
 111         }
 112
 113         ret = sock->ops->listen(sock, -1);
 114         if (ret < 0) {
 115                 goto error;
 116         }
 117
 118         DBG("[agent-thread] Listening on TCP port %u and socket %d",
 119                         config.agent_tcp_port, sock->fd);
 120
 121         return sock;
 122
 123 error:
 124         if (sock) {
 125                 lttcomm_destroy_sock(sock);
 126         }
 127         return NULL;
 128 }
 129
 130 /*
 131  * Close and destroy the given TCP socket.
 132  */
 133 static void destroy_tcp_socket(struct lttcomm_sock *sock)
 134 {
 135         assert(sock);
 136
 137         DBG3("[agent-thread] Destroy TCP socket on port %u", config.agent_tcp_port);
 138
 139         /* This will return gracefully if fd is invalid. */
 140         sock->ops->close(sock);
 141         lttcomm_destroy_sock(sock);
 142 }
 143
 144 /*
 145  * Handle a new agent registration using the reg socket. After that, a new
 146  * agent application is added to the global hash table and attach to an UST app
 147  * object. If r_app is not NULL, the created app is set to the pointer.
 148  *
 149  * Return the new FD created upon accept() on success or else a negative errno
 150  * value.
 151  */
 152 static int handle_registration(struct lttcomm_sock *reg_sock,
 153                 struct agent_app **r_app)
 154 {
 155         int ret;
 156         pid_t pid;
 157         uint32_t major_version, minor_version;
 158         ssize_t size;
 159         enum lttng_domain_type domain;
 160         struct agent_app *app;
 161         struct agent_register_msg msg;
 162         struct lttcomm_sock *new_sock;
 163
 164         assert(reg_sock);
 165
 166         new_sock = reg_sock->ops->accept(reg_sock);
 167         if (!new_sock) {
 168                 ret = -ENOTCONN;
 169                 goto error;
 170         }
 171
 172         size = new_sock->ops->recvmsg(new_sock, &msg, sizeof(msg), 0);
 173         if (size < sizeof(msg)) {
 174                 ret = -EINVAL;
 175                 goto error_socket;
 176         }
 177         domain = be32toh(msg.domain);
 178         pid = be32toh(msg.pid);
 179         major_version = be32toh(msg.major_version);
 180         minor_version = be32toh(msg.minor_version);
 181
 182         /* Test communication protocol version of the registring agent. */
 183         if (major_version != AGENT_MAJOR_VERSION) {
 184                 ret = -EINVAL;
 185                 goto error_socket;
 186         }
 187         if (minor_version != AGENT_MINOR_VERSION) {
 188                 ret = -EINVAL;
 189                 goto error_socket;
 190         }
 191
 192         DBG2("[agent-thread] New registration for pid %d domain %d on socket %d",
 193                         pid, domain, new_sock->fd);
 194
 195         app = agent_create_app(pid, domain, new_sock);
 196         if (!app) {
 197                 ret = -ENOMEM;
 198                 goto error_socket;
 199         }
 200
 201         /*
 202          * Add before assigning the socket value to the UST app so it can be found
 203          * concurrently.
 204          */
 205         agent_add_app(app);
 206
 207         /*
 208          * We don't need to attach the agent app to the app. If we ever do so, we
 209          * should consider both registration order of agent before app and app
 210          * before agent.
 211          */
 212
 213         if (r_app) {
 214                 *r_app = app;
 215         }
 216
 217         return new_sock->fd;
 218
 219 error_socket:
 220         new_sock->ops->close(new_sock);
 221         lttcomm_destroy_sock(new_sock);
 222 error:
 223         return ret;
 224 }
 225
 226 /*
 227  * This thread manage application notify communication.
 228  */
 229 void *agent_thread_manage_registration(void *data)
 230 {
 231         int i, ret, pollfd;
 232         uint32_t revents, nb_fd;
 233         struct lttng_poll_event events;
 234         struct lttcomm_sock *reg_sock;
 235
 236         DBG("[agent-thread] Manage agent application registration.");
 237
 238         rcu_register_thread();
 239         rcu_thread_online();
 240
 241         /* Agent initialization call MUST be called before starting the thread. */
 242         assert(agent_apps_ht_by_sock);
 243
 244         /* Create pollset with size 2, quit pipe and socket. */
 245         ret = sessiond_set_thread_pollset(&events, 2);
 246         if (ret < 0) {
 247                 goto error_poll_create;
 248         }
 249
 250         reg_sock = init_tcp_socket();
 251         sessiond_notify_ready();
 252         if (!reg_sock) {
 253                 goto error_tcp_socket;
 254         }
 255
 256         /* Add TCP socket to poll set. */
 257         ret = lttng_poll_add(&events, reg_sock->fd,
 258                         LPOLLIN | LPOLLERR | LPOLLHUP | LPOLLRDHUP);
 259         if (ret < 0) {
 260                 goto error;
 261         }
 262
 263         while (1) {
 264                 DBG3("[agent-thread] Manage agent polling");
 265
 266                 /* Inifinite blocking call, waiting for transmission */
 267 restart:
 268                 ret = lttng_poll_wait(&events, -1);
 269                 DBG3("[agent-thread] Manage agent return from poll on %d fds",
 270                                 LTTNG_POLL_GETNB(&events));
 271                 if (ret < 0) {
 272                         /*
 273                          * Restart interrupted system call.
 274                          */
 275                         if (errno == EINTR) {
 276                                 goto restart;
 277                         }
 278                         goto error;
 279                 }
 280                 nb_fd = ret;
 281                 DBG3("[agent-thread] %d fd ready", nb_fd);
 282
 283                 for (i = 0; i < nb_fd; i++) {
 284                         /* Fetch once the poll data */
 285                         revents = LTTNG_POLL_GETEV(&events, i);
 286                         pollfd = LTTNG_POLL_GETFD(&events, i);
 287
 288                         if (!revents) {
 289                                 /* No activity for this FD (poll implementation). */
 290                                 continue;
 291                         }
 292
 293                         /* Thread quit pipe has been closed. Killing thread. */
 294                         ret = sessiond_check_thread_quit_pipe(pollfd, revents);
 295                         if (ret) {
 296                                 goto exit;
 297                         }
 298
 299                         if (revents & LPOLLIN) {
 300                                 int new_fd;
 301                                 struct agent_app *app = NULL;
 302
 303                                 assert(pollfd == reg_sock->fd);
 304                                 new_fd = handle_registration(reg_sock, &app);
 305                                 if (new_fd < 0) {
 306                                         continue;
 307                                 }
 308                                 /* Should not have a NULL app on success. */
 309                                 assert(app);
 310
 311                                 /*
 312                                  * Since this is a command socket (write then read),
 313                                  * only add poll error event to only detect shutdown.
 314                                  */
 315                                 ret = lttng_poll_add(&events, new_fd,
 316                                                 LPOLLERR | LPOLLHUP | LPOLLRDHUP);
 317                                 if (ret < 0) {
 318                                         agent_destroy_app_by_sock(new_fd);
 319                                         continue;
 320                                 }
 321
 322                                 /* Update newly registered app. */
 323                                 update_agent_app(app);
 324
 325                                 /* On failure, the poll will detect it and clean it up. */
 326                                 ret = agent_send_registration_done(app);
 327                                 if (ret < 0) {
 328                                         /* Removing from the poll set */
 329                                         ret = lttng_poll_del(&events, new_fd);
 330                                         if (ret < 0) {
 331                                                 goto error;
 332                                         }
 333                                         agent_destroy_app_by_sock(new_fd);
 334                                         continue;
 335                                 }
 336                         } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
 337                                 /* Removing from the poll set */
 338                                 ret = lttng_poll_del(&events, pollfd);
 339                                 if (ret < 0) {
 340                                         goto error;
 341                                 }
 342                                 agent_destroy_app_by_sock(pollfd);
 343                         } else {
 344                                 ERR("Unexpected poll events %u for sock %d", revents, pollfd);
 345                                 goto error;
 346                         }
 347                 }
 348         }
 349
 350 exit:
 351         /* Whatever happens, try to delete it and exit. */
 352         (void) lttng_poll_del(&events, reg_sock->fd);
 353 error:
 354         destroy_tcp_socket(reg_sock);
 355 error_tcp_socket:
 356         lttng_poll_clean(&events);
 357 error_poll_create:
 358         DBG("[agent-thread] is cleaning up and stopping.");
 359
 360         rcu_thread_offline();
 361         rcu_unregister_thread();
 362         return NULL;
 363 }