relayd: track clients of the health unix socket with the fd-tracker
[lttng-tools.git] / src / bin / lttng-relayd / health-relayd.c
CommitLineData
65931c8b
MD
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
6c1c0768 18#define _LGPL_SOURCE
65931c8b
MD
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>
65931c8b 39#include <urcu/compiler.h>
65931c8b
MD
40#include <inttypes.h>
41
42#include <common/defaults.h>
43#include <common/common.h>
c8fea79c
JR
44#include <common/consumer/consumer.h>
45#include <common/consumer/consumer-timer.h>
65931c8b
MD
46#include <common/compat/poll.h>
47#include <common/sessiond-comm/sessiond-comm.h>
48#include <common/utils.h>
e8fa9fb0 49#include <common/compat/getenv.h>
1c9bd75b 50#include <common/fd-tracker/utils.h>
65931c8b
MD
51
52#include "lttng-relayd.h"
53#include "health-relayd.h"
54
55/* Global health check unix path */
094fe907
MD
56static
57char health_unix_sock_path[PATH_MAX];
65931c8b 58
794e2e5f 59int health_quit_pipe[2] = { -1, -1 };
65931c8b
MD
60
61/*
62 * Check if the thread quit pipe was triggered.
63 *
64 * Return 1 if it was triggered else 0;
65 */
66static
67int 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 */
81static 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
91static int create_lttng_rundir_with_perm(const char *rundir)
92{
93 int ret;
94
95 DBG3("Creating LTTng run directory: %s", rundir);
96
97 ret = mkdir(rundir, S_IRWXU);
98 if (ret < 0) {
99 if (errno != EEXIST) {
100 ERR("Unable to create %s", rundir);
101 goto error;
102 } else {
103 ret = 0;
104 }
105 } else if (ret == 0) {
106 int is_root = !getuid();
107
108 if (is_root) {
28ab59d0
JR
109 gid_t gid;
110
111 ret = utils_get_group_id(tracing_group_name, true, &gid);
112 if (ret) {
113 /* Default to root group. */
114 gid = 0;
115 }
116
117 ret = chown(rundir, 0, gid);
65931c8b
MD
118 if (ret < 0) {
119 ERR("Unable to set group on %s", rundir);
120 PERROR("chown");
121 ret = -1;
122 goto error;
123 }
124
125 ret = chmod(rundir,
126 S_IRUSR | S_IWUSR | S_IXUSR | S_IRGRP | S_IXGRP | S_IROTH | S_IXOTH);
127 if (ret < 0) {
128 ERR("Unable to set permissions on %s", health_unix_sock_path);
129 PERROR("chmod");
130 ret = -1;
131 goto error;
132 }
133 }
134 }
135
136error:
137 return ret;
138}
139
094fe907
MD
140static
141int parse_health_env(void)
142{
143 const char *health_path;
144
e8fa9fb0 145 health_path = lttng_secure_getenv(LTTNG_RELAYD_HEALTH_ENV);
094fe907
MD
146 if (health_path) {
147 strncpy(health_unix_sock_path, health_path,
148 PATH_MAX);
149 health_unix_sock_path[PATH_MAX - 1] = '\0';
150 }
151
152 return 0;
153}
154
65931c8b
MD
155static
156int setup_health_path(void)
157{
158 int is_root, ret = 0;
4f00620d
JG
159 const char *home_path = NULL;
160 char *rundir = NULL, *relayd_path = NULL;
65931c8b 161
094fe907
MD
162 ret = parse_health_env();
163 if (ret) {
164 return ret;
165 }
166
65931c8b
MD
167 is_root = !getuid();
168
169 if (is_root) {
c3844e39 170 rundir = strdup(DEFAULT_LTTNG_RUNDIR);
b6ab01aa
MD
171 if (!rundir) {
172 ret = -ENOMEM;
173 goto end;
174 }
65931c8b
MD
175 } else {
176 /*
177 * Create rundir from home path. This will create something like
178 * $HOME/.lttng
179 */
180 home_path = utils_get_home_dir();
181
182 if (home_path == NULL) {
183 /* TODO: Add --socket PATH option */
184 ERR("Can't get HOME directory for sockets creation.");
185 ret = -EPERM;
186 goto end;
187 }
188
c3844e39 189 ret = asprintf(&rundir, DEFAULT_LTTNG_HOME_RUNDIR, home_path);
65931c8b
MD
190 if (ret < 0) {
191 ret = -ENOMEM;
192 goto end;
193 }
194 }
195
c3844e39 196 ret = asprintf(&relayd_path, DEFAULT_RELAYD_PATH, rundir);
65931c8b
MD
197 if (ret < 0) {
198 ret = -ENOMEM;
199 goto end;
200 }
201
c3844e39 202 ret = create_lttng_rundir_with_perm(rundir);
65931c8b
MD
203 if (ret < 0) {
204 goto end;
205 }
206
207 ret = create_lttng_rundir_with_perm(relayd_path);
208 if (ret < 0) {
209 goto end;
210 }
211
212 if (is_root) {
213 if (strlen(health_unix_sock_path) != 0) {
214 goto end;
215 }
216 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
217 DEFAULT_GLOBAL_RELAY_HEALTH_UNIX_SOCK,
d1f721c5 218 (int) getpid());
65931c8b
MD
219 } else {
220 /* Set health check Unix path */
221 if (strlen(health_unix_sock_path) != 0) {
222 goto end;
223 }
224
225 snprintf(health_unix_sock_path, sizeof(health_unix_sock_path),
226 DEFAULT_HOME_RELAY_HEALTH_UNIX_SOCK,
d1f721c5 227 home_path, (int) getpid());
65931c8b
MD
228 }
229
230end:
c3844e39 231 free(rundir);
edd94901 232 free(relayd_path);
65931c8b
MD
233 return ret;
234}
235
1c9bd75b
JG
236static
237int accept_unix_socket(void *data, int *out_fd)
238{
239 int ret;
240 int accepting_sock = *((int *) data);
241
242 ret = lttcomm_accept_unix_sock(accepting_sock);
243 if (ret < 0) {
244 goto end;
245 }
246
247 *out_fd = ret;
248 ret = 0;
249end:
250 return ret;
251}
252
65931c8b
MD
253/*
254 * Thread managing health check socket.
255 */
256void *thread_manage_health(void *data)
257{
258 int sock = -1, new_sock = -1, ret, i, pollfd, err = -1;
259 uint32_t revents, nb_fd;
260 struct lttng_poll_event events;
261 struct health_comm_msg msg;
262 struct health_comm_reply reply;
263 int is_root;
264
265 DBG("[thread] Manage health check started");
266
267 setup_health_path();
268
269 rcu_register_thread();
270
271 /* We might hit an error path before this is created. */
272 lttng_poll_init(&events);
273
274 /* Create unix socket */
275 sock = lttcomm_create_unix_sock(health_unix_sock_path);
276 if (sock < 0) {
277 ERR("Unable to create health check Unix socket");
7568ddbf 278 err = -1;
65931c8b
MD
279 goto error;
280 }
281
282 is_root = !getuid();
283 if (is_root) {
284 /* lttng health client socket path permissions */
28ab59d0
JR
285 gid_t gid;
286
287 ret = utils_get_group_id(tracing_group_name, true, &gid);
288 if (ret) {
289 /* Default to root group. */
290 gid = 0;
291 }
292
293 ret = chown(health_unix_sock_path, 0, gid);
65931c8b
MD
294 if (ret < 0) {
295 ERR("Unable to set group on %s", health_unix_sock_path);
296 PERROR("chown");
7568ddbf 297 err = -1;
65931c8b
MD
298 goto error;
299 }
300
301 ret = chmod(health_unix_sock_path,
302 S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP);
303 if (ret < 0) {
304 ERR("Unable to set permissions on %s", health_unix_sock_path);
305 PERROR("chmod");
7568ddbf 306 err = -1;
65931c8b
MD
307 goto error;
308 }
309 }
310
311 /*
312 * Set the CLOEXEC flag. Return code is useless because either way, the
313 * show must go on.
314 */
315 (void) utils_set_fd_cloexec(sock);
316
317 ret = lttcomm_listen_unix_sock(sock);
318 if (ret < 0) {
319 goto error;
320 }
321
322 /* Size is set to 1 for the consumer_channel pipe */
323 ret = lttng_poll_create(&events, 2, LTTNG_CLOEXEC);
324 if (ret < 0) {
325 ERR("Poll set creation failed");
326 goto error;
327 }
328
329 ret = lttng_poll_add(&events, health_quit_pipe[0], LPOLLIN);
330 if (ret < 0) {
331 goto error;
332 }
333
334 /* Add the application registration socket */
335 ret = lttng_poll_add(&events, sock, LPOLLIN | LPOLLPRI);
336 if (ret < 0) {
337 goto error;
338 }
339
3fd27398
MD
340 lttng_relay_notify_ready();
341
65931c8b 342 while (1) {
1c9bd75b
JG
343 char *accepted_socket_name;
344
65931c8b
MD
345 DBG("Health check ready");
346
347 /* Inifinite blocking call, waiting for transmission */
348restart:
349 ret = lttng_poll_wait(&events, -1);
350 if (ret < 0) {
351 /*
352 * Restart interrupted system call.
353 */
354 if (errno == EINTR) {
355 goto restart;
356 }
357 goto error;
358 }
359
360 nb_fd = ret;
361
362 for (i = 0; i < nb_fd; i++) {
363 /* Fetch once the poll data */
364 revents = LTTNG_POLL_GETEV(&events, i);
365 pollfd = LTTNG_POLL_GETFD(&events, i);
366
367 /* Thread quit pipe has been closed. Killing thread. */
368 ret = check_health_quit_pipe(pollfd, revents);
369 if (ret) {
370 err = 0;
371 goto exit;
372 }
373
374 /* Event on the registration socket */
375 if (pollfd == sock) {
03e43155
MD
376 if (revents & LPOLLIN) {
377 continue;
378 } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
65931c8b
MD
379 ERR("Health socket poll error");
380 goto error;
03e43155
MD
381 } else {
382 ERR("Unexpected poll events %u for sock %d", revents, pollfd);
383 goto error;
65931c8b
MD
384 }
385 }
386 }
387
1c9bd75b
JG
388 ret = asprintf(&accepted_socket_name, "Socket accepted from unix socket @ %s",
389 health_unix_sock_path);
390 if (ret == -1) {
391 PERROR("Failed to allocate name of accepted socket from unix socket @ %s",
392 health_unix_sock_path);
393 goto error;
394 }
395 ret = fd_tracker_open_unsuspendable_fd(the_fd_tracker, &new_sock,
396 (const char **) &accepted_socket_name, 1,
397 accept_unix_socket, &sock);
398 free(accepted_socket_name);
399 if (ret < 0) {
65931c8b
MD
400 goto error;
401 }
402
403 /*
404 * Set the CLOEXEC flag. Return code is useless because either way, the
405 * show must go on.
406 */
407 (void) utils_set_fd_cloexec(new_sock);
408
409 DBG("Receiving data from client for health...");
410 ret = lttcomm_recv_unix_sock(new_sock, (void *)&msg, sizeof(msg));
411 if (ret <= 0) {
412 DBG("Nothing recv() from client... continuing");
1c9bd75b
JG
413 ret = fd_tracker_close_unsuspendable_fd(the_fd_tracker,
414 &new_sock, 1, fd_tracker_util_close_fd,
415 NULL);
65931c8b
MD
416 if (ret) {
417 PERROR("close");
418 }
419 new_sock = -1;
420 continue;
421 }
422
423 rcu_thread_online();
424
425 assert(msg.cmd == HEALTH_CMD_CHECK);
426
53efb85a 427 memset(&reply, 0, sizeof(reply));
65931c8b
MD
428 for (i = 0; i < NR_HEALTH_RELAYD_TYPES; i++) {
429 /*
430 * health_check_state return 0 if thread is in
431 * error.
432 */
433 if (!health_check_state(health_relayd, i)) {
434 reply.ret_code |= 1ULL << i;
435 }
436 }
437
438 DBG2("Health check return value %" PRIx64, reply.ret_code);
439
440 ret = send_unix_sock(new_sock, (void *) &reply, sizeof(reply));
441 if (ret < 0) {
442 ERR("Failed to send health data back to client");
443 }
444
445 /* End of transmission */
1c9bd75b
JG
446 ret = fd_tracker_close_unsuspendable_fd(the_fd_tracker,
447 &new_sock, 1, fd_tracker_util_close_fd,
448 NULL);
65931c8b
MD
449 if (ret) {
450 PERROR("close");
451 }
452 new_sock = -1;
453 }
454
65931c8b 455error:
81714439
JG
456 lttng_relay_stop_threads();
457exit:
65931c8b
MD
458 if (err) {
459 ERR("Health error occurred in %s", __func__);
460 }
461 DBG("Health check thread dying");
462 unlink(health_unix_sock_path);
463 if (sock >= 0) {
464 ret = close(sock);
465 if (ret) {
466 PERROR("close");
467 }
468 }
469
dcbcae3e
MD
470 /*
471 * We do NOT rmdir rundir nor the relayd path because there are
472 * other processes using them.
473 */
474
65931c8b
MD
475 lttng_poll_clean(&events);
476
477 rcu_unregister_thread();
478 return NULL;
479}
This page took 0.141014 seconds and 5 git commands to generate.