bgpd.c revision 1.110 
1/*	$OpenBSD: bgpd.c,v 1.110 2004/10/19 12:02:49 henning Exp $ */
2
3/*
4 * Copyright (c) 2003, 2004 Henning Brauer <henning@openbsd.org>
5 *
6 * Permission to use, copy, modify, and distribute this software for any
7 * purpose with or without fee is hereby granted, provided that the above
8 * copyright notice and this permission notice appear in all copies.
9 *
10 * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
11 * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12 * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
13 * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
14 * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
15 * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
16 * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
17 */
18
19#include <sys/types.h>
20#include <sys/socket.h>
21#include <sys/wait.h>
22#include <netinet/in.h>
23#include <arpa/inet.h>
24#include <err.h>
25#include <errno.h>
26#include <fcntl.h>
27#include <poll.h>
28#include <pwd.h>
29#include <signal.h>
30#include <stdio.h>
31#include <stdlib.h>
32#include <string.h>
33#include <unistd.h>
34
35#include "mrt.h"
36#include "bgpd.h"
37#include "session.h"
38
39void	sighdlr(int);
40void	usage(void);
41int	main(int, char *[]);
42int	check_child(pid_t, const char *);
43int	reconfigure(char *, struct bgpd_config *, struct mrt_head *,
44	    struct peer **, struct filter_head *);
45int	dispatch_imsg(struct imsgbuf *, int);
46
47int			 rfd = -1;
48volatile sig_atomic_t	 mrtdump = 0;
49volatile sig_atomic_t	 quit = 0;
50volatile sig_atomic_t	 reconfig = 0;
51volatile sig_atomic_t	 sigchld = 0;
52struct imsgbuf		*ibuf_se;
53struct imsgbuf		*ibuf_rde;
54
55void
56sighdlr(int sig)
57{
58	switch (sig) {
59	case SIGTERM:
60	case SIGINT:
61		quit = 1;
62		break;
63	case SIGCHLD:
64		sigchld = 1;
65		break;
66	case SIGHUP:
67		reconfig = 1;
68		break;
69	case SIGALRM:
70	case SIGUSR1:
71		mrtdump = 1;
72		break;
73	}
74}
75
76void
77usage(void)
78{
79	extern char *__progname;
80
81	fprintf(stderr, "usage: %s [-dnv] ", __progname);
82	fprintf(stderr, "[-D macro=value] [-f file]\n");
83	exit(1);
84}
85
86#define PFD_PIPE_SESSION	0
87#define PFD_PIPE_ROUTE		1
88#define PFD_SOCK_ROUTE		2
89#define POLL_MAX		3
90#define MAX_TIMEOUT		3600
91
92int
93main(int argc, char *argv[])
94{
95	struct bgpd_config	 conf;
96	struct peer		*peer_l, *p;
97	struct mrt_head		 mrt_l;
98	struct network_head	 net_l;
99	struct filter_head	*rules_l;
100	struct network		*net;
101	struct filter_rule	*r;
102	struct mrt		*m;
103	struct listen_addr	*la;
104	struct pollfd		 pfd[POLL_MAX];
105	pid_t			 io_pid = 0, rde_pid = 0, pid;
106	char			*conffile;
107	int			 debug = 0;
108	int			 ch, nfds, timeout;
109	int			 pipe_m2s[2];
110	int			 pipe_m2r[2];
111	int			 pipe_s2r[2];
112
113	conffile = CONFFILE;
114	bgpd_process = PROC_MAIN;
115
116	log_init(1);		/* log to stderr until daemonized */
117
118	if ((rules_l = calloc(1, sizeof(struct filter_head))) == NULL)
119		err(1, NULL);
120
121	bzero(&conf, sizeof(conf));
122	LIST_INIT(&mrt_l);
123	TAILQ_INIT(&net_l);
124	TAILQ_INIT(rules_l);
125	peer_l = NULL;
126
127	while ((ch = getopt(argc, argv, "dD:f:nv")) != -1) {
128		switch (ch) {
129		case 'd':
130			debug = 1;
131			break;
132		case 'D':
133			if (cmdline_symset(optarg) < 0)
134				log_warnx("could not parse macro definition %s",
135				    optarg);
136			break;
137		case 'f':
138			conffile = optarg;
139			break;
140		case 'n':
141			conf.opts |= BGPD_OPT_NOACTION;
142			break;
143		case 'v':
144			if (conf.opts & BGPD_OPT_VERBOSE)
145				conf.opts |= BGPD_OPT_VERBOSE2;
146			conf.opts |= BGPD_OPT_VERBOSE;
147			break;
148		default:
149			usage();
150			/* NOTREACHED */
151		}
152	}
153
154	if (parse_config(conffile, &conf, &mrt_l, &peer_l, &net_l, rules_l))
155		exit(1);
156
157	if (conf.opts & BGPD_OPT_NOACTION) {
158		if (conf.opts & BGPD_OPT_VERBOSE)
159			print_config(&conf, &net_l, peer_l, rules_l, &mrt_l);
160		else
161			fprintf(stderr, "configuration OK\n");
162		exit(0);
163	}
164
165	if (geteuid())
166		errx(1, "need root privileges");
167
168	if (getpwnam(BGPD_USER) == NULL)
169		errx(1, "unknown user %s", BGPD_USER);
170	endpwent();
171
172	log_init(debug);
173
174	if (!debug)
175		daemon(1, 0);
176
177	log_info("startup");
178
179	if (socketpair(AF_UNIX, SOCK_STREAM, PF_UNSPEC, pipe_m2s) == -1)
180		fatal("socketpair");
181	if (socketpair(AF_UNIX, SOCK_STREAM, PF_UNSPEC, pipe_m2r) == -1)
182		fatal("socketpair");
183	if (socketpair(AF_UNIX, SOCK_STREAM, PF_UNSPEC, pipe_s2r) == -1)
184		fatal("socketpair");
185	session_socket_blockmode(pipe_m2s[0], BM_NONBLOCK);
186	session_socket_blockmode(pipe_m2s[1], BM_NONBLOCK);
187	session_socket_blockmode(pipe_m2r[0], BM_NONBLOCK);
188	session_socket_blockmode(pipe_m2r[1], BM_NONBLOCK);
189	session_socket_blockmode(pipe_s2r[0], BM_NONBLOCK);
190	session_socket_blockmode(pipe_s2r[1], BM_NONBLOCK);
191
192	prepare_listeners(&conf);
193
194	/* fork children */
195	rde_pid = rde_main(&conf, peer_l, &net_l, rules_l, &mrt_l,
196	    pipe_m2r, pipe_s2r, pipe_m2s);
197	io_pid = session_main(&conf, peer_l, &net_l, rules_l, &mrt_l,
198	    pipe_m2s, pipe_s2r, pipe_m2r);
199
200	setproctitle("parent");
201
202	signal(SIGTERM, sighdlr);
203	signal(SIGINT, sighdlr);
204	signal(SIGCHLD, sighdlr);
205	signal(SIGHUP, sighdlr);
206	signal(SIGALRM, sighdlr);
207	signal(SIGUSR1, sighdlr);
208
209	close(pipe_m2s[1]);
210	close(pipe_m2r[1]);
211	close(pipe_s2r[0]);
212	close(pipe_s2r[1]);
213
214	if ((ibuf_se = malloc(sizeof(struct imsgbuf))) == NULL ||
215	    (ibuf_rde = malloc(sizeof(struct imsgbuf))) == NULL)
216		fatal(NULL);
217	imsg_init(ibuf_se, pipe_m2s[0]);
218	imsg_init(ibuf_rde, pipe_m2r[0]);
219	mrt_init(ibuf_rde, ibuf_se);
220	if ((rfd = kr_init(!(conf.flags & BGPD_FLAG_NO_FIB_UPDATE))) == -1)
221		quit = 1;
222	if (pftable_clear_all() != 0)
223		quit = 1;
224
225	while ((net = TAILQ_FIRST(&net_l)) != NULL) {
226		TAILQ_REMOVE(&net_l, net, entry);
227		free(net);
228	}
229
230	while ((r = TAILQ_FIRST(rules_l)) != NULL) {
231		TAILQ_REMOVE(rules_l, r, entry);
232		free(r);
233	}
234
235	while ((la = TAILQ_FIRST(conf.listen_addrs)) != NULL) {
236		TAILQ_REMOVE(conf.listen_addrs, la, entry);
237		close(la->fd);
238		free(la);
239	}
240
241	mrt_reconfigure(&mrt_l);
242
243	while (quit == 0) {
244		pfd[PFD_PIPE_SESSION].fd = ibuf_se->fd;
245		pfd[PFD_PIPE_SESSION].events = POLLIN;
246		if (ibuf_se->w.queued)
247			pfd[PFD_PIPE_SESSION].events |= POLLOUT;
248		pfd[PFD_PIPE_ROUTE].fd = ibuf_rde->fd;
249		pfd[PFD_PIPE_ROUTE].events = POLLIN;
250		if (ibuf_rde->w.queued)
251			pfd[PFD_PIPE_ROUTE].events |= POLLOUT;
252		pfd[PFD_SOCK_ROUTE].fd = rfd;
253		pfd[PFD_SOCK_ROUTE].events = POLLIN;
254
255		timeout = mrt_timeout(&mrt_l);
256		if (timeout > MAX_TIMEOUT)
257			timeout = MAX_TIMEOUT;
258
259		if ((nfds = poll(pfd, POLL_MAX, timeout * 1000)) == -1)
260			if (errno != EINTR) {
261				log_warn("poll error");
262				quit = 1;
263			}
264
265		if (nfds > 0 && (pfd[PFD_PIPE_SESSION].revents & POLLOUT))
266			if (msgbuf_write(&ibuf_se->w) < 0) {
267				log_warn("pipe write error (to SE)");
268				quit = 1;
269			}
270
271		if (nfds > 0 && (pfd[PFD_PIPE_ROUTE].revents & POLLOUT))
272			if (msgbuf_write(&ibuf_rde->w) < 0) {
273				log_warn("pipe write error (to RDE)");
274				quit = 1;
275			}
276
277		if (nfds > 0 && pfd[PFD_PIPE_SESSION].revents & POLLIN) {
278			nfds--;
279			if (dispatch_imsg(ibuf_se, PFD_PIPE_SESSION) == -1)
280				quit = 1;
281		}
282
283		if (nfds > 0 && pfd[PFD_PIPE_ROUTE].revents & POLLIN) {
284			nfds--;
285			if (dispatch_imsg(ibuf_rde, PFD_PIPE_ROUTE) == -1)
286				quit = 1;
287		}
288
289		if (nfds > 0 && pfd[PFD_SOCK_ROUTE].revents & POLLIN) {
290			nfds--;
291			if (kr_dispatch_msg() == -1)
292				quit = 1;
293		}
294
295		if (reconfig) {
296			reconfig = 0;
297			log_info("rereading config");
298			reconfigure(conffile, &conf, &mrt_l, &peer_l, rules_l);
299		}
300
301		if (sigchld) {
302			sigchld = 0;
303			if (check_child(io_pid, "session engine")) {
304				quit = 1;
305				io_pid = 0;
306			}
307			if (check_child(rde_pid, "route decision engine")) {
308				quit = 1;
309				rde_pid = 0;
310			}
311		}
312
313		if (mrtdump == 1) {
314			mrtdump = 0;
315			mrt_handler(&mrt_l);
316		}
317	}
318
319	signal(SIGCHLD, SIG_IGN);
320
321	if (io_pid)
322		kill(io_pid, SIGTERM);
323
324	if (rde_pid)
325		kill(rde_pid, SIGTERM);
326
327	while ((p = peer_l) != NULL) {
328		peer_l = p->next;
329		free(p);
330	}
331	while ((m = LIST_FIRST(&mrt_l)) != NULL) {
332		LIST_REMOVE(m, entry);
333		free(m);
334	}
335
336	free(rules_l);
337	control_cleanup();
338	kr_shutdown();
339	pftable_clear_all();
340	free(conf.listen_addrs);
341
342	do {
343		if ((pid = wait(NULL)) == -1 &&
344		    errno != EINTR && errno != ECHILD)
345			fatal("wait");
346	} while (pid != -1 || (pid == -1 && errno == EINTR));
347
348	msgbuf_clear(&ibuf_se->w);
349	free(ibuf_se);
350	msgbuf_clear(&ibuf_rde->w);
351	free(ibuf_rde);
352
353	log_info("Terminating");
354	return (0);
355}
356
357int
358check_child(pid_t pid, const char *pname)
359{
360	int	status;
361
362	if (waitpid(pid, &status, WNOHANG) > 0) {
363		if (WIFEXITED(status)) {
364			log_warnx("Lost child: %s exited", pname);
365			return (1);
366		}
367		if (WIFSIGNALED(status)) {
368			log_warnx("Lost child: %s terminated; signal %d",
369			    pname, WTERMSIG(status));
370			return (1);
371		}
372	}
373
374	return (0);
375}
376
377int
378reconfigure(char *conffile, struct bgpd_config *conf, struct mrt_head *mrt_l,
379    struct peer **peer_l, struct filter_head *rules_l)
380{
381	struct network_head	 net_l;
382	struct network		*n;
383	struct peer		*p;
384	struct filter_rule	*r;
385	struct listen_addr	*la;
386
387	if (parse_config(conffile, conf, mrt_l, peer_l, &net_l, rules_l)) {
388		log_warnx("config file %s has errors, not reloading",
389		    conffile);
390		return (-1);
391	}
392
393	prepare_listeners(conf);
394
395	if (imsg_compose(ibuf_se, IMSG_RECONF_CONF, 0, 0, -1,
396	    conf, sizeof(struct bgpd_config)) == -1)
397		return (-1);
398	if (imsg_compose(ibuf_rde, IMSG_RECONF_CONF, 0, 0, -1,
399	    conf, sizeof(struct bgpd_config)) == -1)
400		return (-1);
401	for (p = *peer_l; p != NULL; p = p->next)
402		if (imsg_compose(ibuf_se, IMSG_RECONF_PEER, p->conf.id, 0, -1,
403		    &p->conf, sizeof(struct peer_config)) == -1)
404			return (-1);
405	while ((n = TAILQ_FIRST(&net_l)) != NULL) {
406		if (imsg_compose(ibuf_rde, IMSG_NETWORK_ADD, 0, 0, -1,
407		    &n->net, sizeof(struct network_config)) == -1)
408			return (-1);
409		TAILQ_REMOVE(&net_l, n, entry);
410		free(n);
411	}
412	while ((r = TAILQ_FIRST(rules_l)) != NULL) {
413		if (imsg_compose(ibuf_rde, IMSG_RECONF_FILTER, 0, 0, -1,
414		    r, sizeof(struct filter_rule)) == -1)
415			return (-1);
416		TAILQ_REMOVE(rules_l, r, entry);
417		free(r);
418	}
419	while ((la = TAILQ_FIRST(conf->listen_addrs)) != NULL) {
420		if (imsg_compose(ibuf_se, IMSG_RECONF_LISTENER, 0, 0, la->fd,
421		    la, sizeof(struct listen_addr)) == -1)
422			return (-1);
423		TAILQ_REMOVE(conf->listen_addrs, la, entry);
424		free(la);
425	}
426	free(conf->listen_addrs);
427	conf->listen_addrs = NULL;
428
429	if (imsg_compose(ibuf_se, IMSG_RECONF_DONE, 0, 0, -1, NULL, 0) == -1 ||
430	    imsg_compose(ibuf_rde, IMSG_RECONF_DONE, 0, 0, -1, NULL, 0) == -1)
431		return (-1);
432
433	/* mrt changes can be sent out of bound */
434	mrt_reconfigure(mrt_l);
435	return (0);
436}
437
438int
439dispatch_imsg(struct imsgbuf *ibuf, int idx)
440{
441	struct imsg		 imsg;
442	int			 n;
443
444	if ((n = imsg_read(ibuf)) == -1)
445		return (-1);
446
447	if (n == 0) {	/* connection closed */
448		log_warnx("dispatch_imsg in main: pipe closed");
449		return (-1);
450	}
451
452	for (;;) {
453		if ((n = imsg_get(ibuf, &imsg)) == -1)
454			return (-1);
455
456		if (n == 0)
457			break;
458
459		switch (imsg.hdr.type) {
460		case IMSG_KROUTE_CHANGE:
461			if (idx != PFD_PIPE_ROUTE)
462				log_warnx("route request not from RDE");
463			else if (kr_change(imsg.data))
464				return (-1);
465			break;
466		case IMSG_KROUTE_DELETE:
467			if (idx != PFD_PIPE_ROUTE)
468				log_warnx("route request not from RDE");
469			else if (kr_delete(imsg.data))
470				return (-1);
471			break;
472		case IMSG_NEXTHOP_ADD:
473			if (idx != PFD_PIPE_ROUTE)
474				log_warnx("nexthop request not from RDE");
475			else
476				if (imsg.hdr.len != IMSG_HEADER_SIZE +
477				    sizeof(struct bgpd_addr))
478					log_warnx("wrong imsg len");
479				else if (kr_nexthop_add(imsg.data) == -1)
480					return (-1);
481			break;
482		case IMSG_NEXTHOP_REMOVE:
483			if (idx != PFD_PIPE_ROUTE)
484				log_warnx("nexthop request not from RDE");
485			else
486				if (imsg.hdr.len != IMSG_HEADER_SIZE +
487				    sizeof(struct bgpd_addr))
488					log_warnx("wrong imsg len");
489				else
490					kr_nexthop_delete(imsg.data);
491			break;
492		case IMSG_PFTABLE_ADD:
493			if (idx != PFD_PIPE_ROUTE)
494				log_warnx("pftable request not from RDE");
495			else
496				if (imsg.hdr.len != IMSG_HEADER_SIZE +
497				    sizeof(struct pftable_msg))
498					log_warnx("wrong imsg len");
499				else if (pftable_addr_add(imsg.data) != 0)
500					return (-1);
501			break;
502		case IMSG_PFTABLE_REMOVE:
503			if (idx != PFD_PIPE_ROUTE)
504				log_warnx("pftable request not from RDE");
505			else
506				if (imsg.hdr.len != IMSG_HEADER_SIZE +
507				    sizeof(struct pftable_msg))
508					log_warnx("wrong imsg len");
509				else if (pftable_addr_remove(imsg.data) != 0)
510					return (-1);
511			break;
512		case IMSG_PFTABLE_COMMIT:
513			if (idx != PFD_PIPE_ROUTE)
514				log_warnx("pftable request not from RDE");
515			else
516				if (imsg.hdr.len != IMSG_HEADER_SIZE)
517					log_warnx("wrong imsg len");
518				else if (pftable_commit() != 0)
519					return (-1);
520			break;
521		case IMSG_CTL_RELOAD:
522			if (idx != PFD_PIPE_SESSION)
523				log_warnx("reload request not from SE");
524			else
525				reconfig = 1;
526			break;
527		case IMSG_CTL_FIB_COUPLE:
528			if (idx != PFD_PIPE_SESSION)
529				log_warnx("couple request not from SE");
530			else
531				kr_fib_couple();
532			break;
533		case IMSG_CTL_FIB_DECOUPLE:
534			if (idx != PFD_PIPE_SESSION)
535				log_warnx("decouple request not from SE");
536			else
537				kr_fib_decouple();
538			break;
539		case IMSG_CTL_KROUTE:
540		case IMSG_CTL_KROUTE_ADDR:
541		case IMSG_CTL_SHOW_NEXTHOP:
542		case IMSG_CTL_SHOW_INTERFACE:
543			if (idx != PFD_PIPE_SESSION)
544				log_warnx("kroute request not from SE");
545			else
546				kr_show_route(&imsg);
547			break;
548		case IMSG_IFINFO:
549			if (idx != PFD_PIPE_SESSION)
550				log_warnx("IFINFO request not from SE");
551			else if (imsg.hdr.len != IMSG_HEADER_SIZE + IFNAMSIZ)
552				log_warnx("IFINFO request with wrong len");
553			else
554				kr_ifinfo(imsg.data);
555			break;
556		default:
557			break;
558		}
559		imsg_free(&imsg);
560	}
561	return (0);
562}
563
564void
565send_nexthop_update(struct kroute_nexthop *msg)
566{
567	char	*gw = NULL;
568
569	if (msg->gateway.af)
570		if (asprintf(&gw, ": via %s",
571		    log_addr(&msg->gateway)) == -1) {
572			log_warn("send_nexthop_update");
573			quit = 1;
574		}
575
576	log_info("nexthop %s now %s%s%s", log_addr(&msg->nexthop),
577	    msg->valid ? "valid" : "invalid",
578	    msg->connected ? ": directly connected" : "",
579	    msg->gateway.af ? gw : "");
580
581	free(gw);
582
583	if (imsg_compose(ibuf_rde, IMSG_NEXTHOP_UPDATE, 0, 0, -1,
584	    msg, sizeof(struct kroute_nexthop)) == -1)
585		quit = 1;
586}
587
588void
589send_imsg_session(int type, pid_t pid, void *data, u_int16_t datalen)
590{
591	imsg_compose(ibuf_se, type, 0, pid, -1, data, datalen);
592}
593