Fix: failure to launch agent thread is not reported
[lttng-tools.git] / 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 static int agent_tracing_enabled = -1;
36
37 /*
38 * Note that there is not port here. It's set after this URI is parsed so we
39 * can let the user define a custom one. However, localhost is ALWAYS the
40 * default listening address.
41 */
42 static const char *default_reg_uri =
43 "tcp://" DEFAULT_NETWORK_VIEWER_BIND_ADDRESS;
44
45 /*
46 * Update agent application using the given socket. This is done just after
47 * registration was successful.
48 *
49 * This is a quite heavy call in terms of locking since the session list lock
50 * AND session lock are acquired.
51 */
52 static void update_agent_app(struct agent_app *app)
53 {
54 struct ltt_session *session, *stmp;
55 struct ltt_session_list *list;
56
57 list = session_get_list();
58 assert(list);
59
60 session_lock_list();
61 cds_list_for_each_entry_safe(session, stmp, &list->head, list) {
62 session_lock(session);
63 if (session->ust_session) {
64 struct agent *agt;
65
66 rcu_read_lock();
67 agt = trace_ust_find_agent(session->ust_session, app->domain);
68 if (agt) {
69 agent_update(agt, app->sock->fd);
70 }
71 rcu_read_unlock();
72 }
73 session_unlock(session);
74 }
75 session_unlock_list();
76 }
77
78 /*
79 * Create and init socket from uri.
80 */
81 static struct lttcomm_sock *init_tcp_socket(void)
82 {
83 int ret;
84 struct lttng_uri *uri = NULL;
85 struct lttcomm_sock *sock = NULL;
86
87 /*
88 * This should never fail since the URI is hardcoded and the port is set
89 * before this thread is launched.
90 */
91 ret = uri_parse(default_reg_uri, &uri);
92 assert(ret);
93 assert(config.agent_tcp_port);
94 uri->port = config.agent_tcp_port;
95
96 sock = lttcomm_alloc_sock_from_uri(uri);
97 uri_free(uri);
98 if (sock == NULL) {
99 ERR("[agent-thread] agent allocating TCP socket");
100 goto error;
101 }
102
103 ret = lttcomm_create_sock(sock);
104 if (ret < 0) {
105 goto error;
106 }
107
108 ret = sock->ops->bind(sock);
109 if (ret < 0) {
110 WARN("Another session daemon is using this agent port. Agent support "
111 "will be deactivated to prevent interfering with the tracing.");
112 goto error;
113 }
114
115 ret = sock->ops->listen(sock, -1);
116 if (ret < 0) {
117 goto error;
118 }
119
120 DBG("[agent-thread] Listening on TCP port %u and socket %d",
121 config.agent_tcp_port, sock->fd);
122
123 return sock;
124
125 error:
126 if (sock) {
127 lttcomm_destroy_sock(sock);
128 }
129 return NULL;
130 }
131
132 /*
133 * Close and destroy the given TCP socket.
134 */
135 static void destroy_tcp_socket(struct lttcomm_sock *sock)
136 {
137 assert(sock);
138
139 DBG3("[agent-thread] Destroy TCP socket on port %u", config.agent_tcp_port);
140
141 /* This will return gracefully if fd is invalid. */
142 sock->ops->close(sock);
143 lttcomm_destroy_sock(sock);
144 }
145
146 /*
147 * Handle a new agent registration using the reg socket. After that, a new
148 * agent application is added to the global hash table and attach to an UST app
149 * object. If r_app is not NULL, the created app is set to the pointer.
150 *
151 * Return the new FD created upon accept() on success or else a negative errno
152 * value.
153 */
154 static int handle_registration(struct lttcomm_sock *reg_sock,
155 struct agent_app **r_app)
156 {
157 int ret;
158 pid_t pid;
159 uint32_t major_version, minor_version;
160 ssize_t size;
161 enum lttng_domain_type domain;
162 struct agent_app *app;
163 struct agent_register_msg msg;
164 struct lttcomm_sock *new_sock;
165
166 assert(reg_sock);
167
168 new_sock = reg_sock->ops->accept(reg_sock);
169 if (!new_sock) {
170 ret = -ENOTCONN;
171 goto error;
172 }
173
174 size = new_sock->ops->recvmsg(new_sock, &msg, sizeof(msg), 0);
175 if (size < sizeof(msg)) {
176 ret = -EINVAL;
177 goto error_socket;
178 }
179 domain = be32toh(msg.domain);
180 pid = be32toh(msg.pid);
181 major_version = be32toh(msg.major_version);
182 minor_version = be32toh(msg.minor_version);
183
184 /* Test communication protocol version of the registring agent. */
185 if (major_version != AGENT_MAJOR_VERSION) {
186 ret = -EINVAL;
187 goto error_socket;
188 }
189 if (minor_version != AGENT_MINOR_VERSION) {
190 ret = -EINVAL;
191 goto error_socket;
192 }
193
194 DBG2("[agent-thread] New registration for pid %d domain %d on socket %d",
195 pid, domain, new_sock->fd);
196
197 app = agent_create_app(pid, domain, new_sock);
198 if (!app) {
199 ret = -ENOMEM;
200 goto error_socket;
201 }
202
203 /*
204 * Add before assigning the socket value to the UST app so it can be found
205 * concurrently.
206 */
207 agent_add_app(app);
208
209 /*
210 * We don't need to attach the agent app to the app. If we ever do so, we
211 * should consider both registration order of agent before app and app
212 * before agent.
213 */
214
215 if (r_app) {
216 *r_app = app;
217 }
218
219 return new_sock->fd;
220
221 error_socket:
222 new_sock->ops->close(new_sock);
223 lttcomm_destroy_sock(new_sock);
224 error:
225 return ret;
226 }
227
228 bool agent_tracing_is_enabled(void)
229 {
230 int enabled;
231
232 enabled = uatomic_read(&agent_tracing_enabled);
233 assert(enabled != -1);
234 return enabled == 1;
235 }
236
237 /*
238 * This thread manage application notify communication.
239 */
240 void *agent_thread_manage_registration(void *data)
241 {
242 int i, ret, pollfd;
243 uint32_t revents, nb_fd;
244 struct lttng_poll_event events;
245 struct lttcomm_sock *reg_sock;
246
247 DBG("[agent-thread] Manage agent application registration.");
248
249 rcu_register_thread();
250 rcu_thread_online();
251
252 /* Agent initialization call MUST be called before starting the thread. */
253 assert(agent_apps_ht_by_sock);
254
255 /* Create pollset with size 2, quit pipe and socket. */
256 ret = sessiond_set_thread_pollset(&events, 2);
257 if (ret < 0) {
258 goto error_poll_create;
259 }
260
261 reg_sock = init_tcp_socket();
262 uatomic_set(&agent_tracing_enabled, !!reg_sock);
263
264 /*
265 * Signal that the agent thread is ready. The command thread
266 * may start to query whether or not agent tracing is enabled.
267 */
268 sessiond_notify_ready();
269 if (!reg_sock) {
270 goto error_tcp_socket;
271 }
272
273 /* Add TCP socket to poll set. */
274 ret = lttng_poll_add(&events, reg_sock->fd,
275 LPOLLIN | LPOLLERR | LPOLLHUP | LPOLLRDHUP);
276 if (ret < 0) {
277 goto error;
278 }
279
280 while (1) {
281 DBG3("[agent-thread] Manage agent polling");
282
283 /* Inifinite blocking call, waiting for transmission */
284 restart:
285 ret = lttng_poll_wait(&events, -1);
286 DBG3("[agent-thread] Manage agent return from poll on %d fds",
287 LTTNG_POLL_GETNB(&events));
288 if (ret < 0) {
289 /*
290 * Restart interrupted system call.
291 */
292 if (errno == EINTR) {
293 goto restart;
294 }
295 goto error;
296 }
297 nb_fd = ret;
298 DBG3("[agent-thread] %d fd ready", nb_fd);
299
300 for (i = 0; i < nb_fd; i++) {
301 /* Fetch once the poll data */
302 revents = LTTNG_POLL_GETEV(&events, i);
303 pollfd = LTTNG_POLL_GETFD(&events, i);
304
305 if (!revents) {
306 /* No activity for this FD (poll implementation). */
307 continue;
308 }
309
310 /* Thread quit pipe has been closed. Killing thread. */
311 ret = sessiond_check_thread_quit_pipe(pollfd, revents);
312 if (ret) {
313 goto exit;
314 }
315
316 if (revents & LPOLLIN) {
317 int new_fd;
318 struct agent_app *app = NULL;
319
320 assert(pollfd == reg_sock->fd);
321 new_fd = handle_registration(reg_sock, &app);
322 if (new_fd < 0) {
323 continue;
324 }
325 /* Should not have a NULL app on success. */
326 assert(app);
327
328 /*
329 * Since this is a command socket (write then read),
330 * only add poll error event to only detect shutdown.
331 */
332 ret = lttng_poll_add(&events, new_fd,
333 LPOLLERR | LPOLLHUP | LPOLLRDHUP);
334 if (ret < 0) {
335 agent_destroy_app_by_sock(new_fd);
336 continue;
337 }
338
339 /* Update newly registered app. */
340 update_agent_app(app);
341
342 /* On failure, the poll will detect it and clean it up. */
343 ret = agent_send_registration_done(app);
344 if (ret < 0) {
345 /* Removing from the poll set */
346 ret = lttng_poll_del(&events, new_fd);
347 if (ret < 0) {
348 goto error;
349 }
350 agent_destroy_app_by_sock(new_fd);
351 continue;
352 }
353 } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
354 /* Removing from the poll set */
355 ret = lttng_poll_del(&events, pollfd);
356 if (ret < 0) {
357 goto error;
358 }
359 agent_destroy_app_by_sock(pollfd);
360 } else {
361 ERR("Unexpected poll events %u for sock %d", revents, pollfd);
362 goto error;
363 }
364 }
365 }
366
367 exit:
368 uatomic_set(&agent_tracing_enabled, 0);
369 /* Whatever happens, try to delete it and exit. */
370 (void) lttng_poll_del(&events, reg_sock->fd);
371 error:
372 destroy_tcp_socket(reg_sock);
373 error_tcp_socket:
374 lttng_poll_clean(&events);
375 error_poll_create:
376 DBG("[agent-thread] is cleaning up and stopping.");
377
378 rcu_thread_offline();
379 rcu_unregister_thread();
380 return NULL;
381 }
This page took 0.038632 seconds and 4 git commands to generate.