bgpd.c revision 1.116 
1/*	$OpenBSD: bgpd.c,v 1.116 2005/03/30 11:23:15 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);
40__dead void	usage(void);
41int		main(int, char *[]);
42int		check_child(pid_t, const char *);
43int		send_filterset(struct imsgbuf *, struct filter_set_head *,
44		    int, int);
45int		reconfigure(char *, struct bgpd_config *, struct mrt_head *,
46		    struct peer **, struct filter_head *);
47int		dispatch_imsg(struct imsgbuf *, int);
48
49int			 rfd = -1;
50volatile sig_atomic_t	 mrtdump = 0;
51volatile sig_atomic_t	 quit = 0;
52volatile sig_atomic_t	 reconfig = 0;
53volatile sig_atomic_t	 sigchld = 0;
54struct imsgbuf		*ibuf_se;
55struct imsgbuf		*ibuf_rde;
56
57void
58sighdlr(int sig)
59{
60	switch (sig) {
61	case SIGTERM:
62	case SIGINT:
63		quit = 1;
64		break;
65	case SIGCHLD:
66		sigchld = 1;
67		break;
68	case SIGHUP:
69		reconfig = 1;
70		break;
71	case SIGALRM:
72	case SIGUSR1:
73		mrtdump = 1;
74		break;
75	}
76}
77
78__dead void
79usage(void)
80{
81	extern char *__progname;
82
83	fprintf(stderr, "usage: %s [-dnv] ", __progname);
84	fprintf(stderr, "[-D macro=value] [-f file]\n");
85	exit(1);
86}
87
88#define PFD_PIPE_SESSION	0
89#define PFD_PIPE_ROUTE		1
90#define PFD_SOCK_ROUTE		2
91#define POLL_MAX		3
92#define MAX_TIMEOUT		3600
93
94int
95main(int argc, char *argv[])
96{
97	struct bgpd_config	 conf;
98	struct peer		*peer_l, *p;
99	struct mrt_head		 mrt_l;
100	struct network_head	 net_l;
101	struct filter_head	*rules_l;
102	struct network		*net;
103	struct filter_rule	*r;
104	struct mrt		*m;
105	struct listen_addr	*la;
106	struct pollfd		 pfd[POLL_MAX];
107	pid_t			 io_pid = 0, rde_pid = 0, pid;
108	char			*conffile;
109	int			 debug = 0;
110	int			 ch, nfds, timeout;
111	int			 pipe_m2s[2];
112	int			 pipe_m2r[2];
113	int			 pipe_s2r[2];
114
115	conffile = CONFFILE;
116	bgpd_process = PROC_MAIN;
117
118	log_init(1);		/* log to stderr until daemonized */
119
120	if ((rules_l = calloc(1, sizeof(struct filter_head))) == NULL)
121		err(1, NULL);
122
123	bzero(&conf, sizeof(conf));
124	LIST_INIT(&mrt_l);
125	TAILQ_INIT(&net_l);
126	TAILQ_INIT(rules_l);
127	peer_l = NULL;
128
129	while ((ch = getopt(argc, argv, "dD:f:nv")) != -1) {
130		switch (ch) {
131		case 'd':
132			debug = 1;
133			break;
134		case 'D':
135			if (cmdline_symset(optarg) < 0)
136				log_warnx("could not parse macro definition %s",
137				    optarg);
138			break;
139		case 'f':
140			conffile = optarg;
141			break;
142		case 'n':
143			conf.opts |= BGPD_OPT_NOACTION;
144			break;
145		case 'v':
146			if (conf.opts & BGPD_OPT_VERBOSE)
147				conf.opts |= BGPD_OPT_VERBOSE2;
148			conf.opts |= BGPD_OPT_VERBOSE;
149			break;
150		default:
151			usage();
152			/* NOTREACHED */
153		}
154	}
155
156	if (parse_config(conffile, &conf, &mrt_l, &peer_l, &net_l, rules_l)) {
157		free(rules_l);
158		exit(1);
159	}
160
161	if (conf.opts & BGPD_OPT_NOACTION) {
162		if (conf.opts & BGPD_OPT_VERBOSE)
163			print_config(&conf, &net_l, peer_l, rules_l, &mrt_l);
164		else
165			fprintf(stderr, "configuration OK\n");
166		exit(0);
167	}
168
169	if (geteuid())
170		errx(1, "need root privileges");
171
172	if (getpwnam(BGPD_USER) == NULL)
173		errx(1, "unknown user %s", BGPD_USER);
174	endpwent();
175
176	log_init(debug);
177
178	if (!debug)
179		daemon(1, 0);
180
181	log_info("startup");
182
183	if (socketpair(AF_UNIX, SOCK_STREAM, PF_UNSPEC, pipe_m2s) == -1)
184		fatal("socketpair");
185	if (socketpair(AF_UNIX, SOCK_STREAM, PF_UNSPEC, pipe_m2r) == -1)
186		fatal("socketpair");
187	if (socketpair(AF_UNIX, SOCK_STREAM, PF_UNSPEC, pipe_s2r) == -1)
188		fatal("socketpair");
189	session_socket_blockmode(pipe_m2s[0], BM_NONBLOCK);
190	session_socket_blockmode(pipe_m2s[1], BM_NONBLOCK);
191	session_socket_blockmode(pipe_m2r[0], BM_NONBLOCK);
192	session_socket_blockmode(pipe_m2r[1], BM_NONBLOCK);
193	session_socket_blockmode(pipe_s2r[0], BM_NONBLOCK);
194	session_socket_blockmode(pipe_s2r[1], BM_NONBLOCK);
195
196	prepare_listeners(&conf);
197
198	/* fork children */
199	rde_pid = rde_main(&conf, peer_l, &net_l, rules_l, &mrt_l,
200	    pipe_m2r, pipe_s2r, pipe_m2s);
201	io_pid = session_main(&conf, peer_l, &net_l, rules_l, &mrt_l,
202	    pipe_m2s, pipe_s2r, pipe_m2r);
203
204	setproctitle("parent");
205
206	signal(SIGTERM, sighdlr);
207	signal(SIGINT, sighdlr);
208	signal(SIGCHLD, sighdlr);
209	signal(SIGHUP, sighdlr);
210	signal(SIGALRM, sighdlr);
211	signal(SIGUSR1, sighdlr);
212
213	close(pipe_m2s[1]);
214	close(pipe_m2r[1]);
215	close(pipe_s2r[0]);
216	close(pipe_s2r[1]);
217
218	if ((ibuf_se = malloc(sizeof(struct imsgbuf))) == NULL ||
219	    (ibuf_rde = malloc(sizeof(struct imsgbuf))) == NULL)
220		fatal(NULL);
221	imsg_init(ibuf_se, pipe_m2s[0]);
222	imsg_init(ibuf_rde, pipe_m2r[0]);
223	mrt_init(ibuf_rde, ibuf_se);
224	if ((rfd = kr_init(!(conf.flags & BGPD_FLAG_NO_FIB_UPDATE))) == -1)
225		quit = 1;
226	if (pftable_clear_all() != 0)
227		quit = 1;
228
229	while ((net = TAILQ_FIRST(&net_l)) != NULL) {
230		TAILQ_REMOVE(&net_l, net, entry);
231		free(net);
232	}
233
234	while ((r = TAILQ_FIRST(rules_l)) != NULL) {
235		TAILQ_REMOVE(rules_l, r, entry);
236		free(r);
237	}
238	TAILQ_FOREACH(la, conf.listen_addrs, entry) {
239		close(la->fd);
240		la->fd = -1;
241	}
242
243	mrt_reconfigure(&mrt_l);
244
245	while (quit == 0) {
246		pfd[PFD_PIPE_SESSION].fd = ibuf_se->fd;
247		pfd[PFD_PIPE_SESSION].events = POLLIN;
248		if (ibuf_se->w.queued)
249			pfd[PFD_PIPE_SESSION].events |= POLLOUT;
250		pfd[PFD_PIPE_ROUTE].fd = ibuf_rde->fd;
251		pfd[PFD_PIPE_ROUTE].events = POLLIN;
252		if (ibuf_rde->w.queued)
253			pfd[PFD_PIPE_ROUTE].events |= POLLOUT;
254		pfd[PFD_SOCK_ROUTE].fd = rfd;
255		pfd[PFD_SOCK_ROUTE].events = POLLIN;
256
257		timeout = mrt_timeout(&mrt_l);
258		if (timeout > MAX_TIMEOUT)
259			timeout = MAX_TIMEOUT;
260
261		if ((nfds = poll(pfd, POLL_MAX, timeout * 1000)) == -1)
262			if (errno != EINTR) {
263				log_warn("poll error");
264				quit = 1;
265			}
266
267		if (nfds > 0 && (pfd[PFD_PIPE_SESSION].revents & POLLOUT))
268			if (msgbuf_write(&ibuf_se->w) < 0) {
269				log_warn("pipe write error (to SE)");
270				quit = 1;
271			}
272
273		if (nfds > 0 && (pfd[PFD_PIPE_ROUTE].revents & POLLOUT))
274			if (msgbuf_write(&ibuf_rde->w) < 0) {
275				log_warn("pipe write error (to RDE)");
276				quit = 1;
277			}
278
279		if (nfds > 0 && pfd[PFD_PIPE_SESSION].revents & POLLIN) {
280			nfds--;
281			if (dispatch_imsg(ibuf_se, PFD_PIPE_SESSION) == -1)
282				quit = 1;
283		}
284
285		if (nfds > 0 && pfd[PFD_PIPE_ROUTE].revents & POLLIN) {
286			nfds--;
287			if (dispatch_imsg(ibuf_rde, PFD_PIPE_ROUTE) == -1)
288				quit = 1;
289		}
290
291		if (nfds > 0 && pfd[PFD_SOCK_ROUTE].revents & POLLIN) {
292			nfds--;
293			if (kr_dispatch_msg() == -1)
294				quit = 1;
295		}
296
297		if (reconfig) {
298			reconfig = 0;
299			log_info("rereading config");
300			reconfigure(conffile, &conf, &mrt_l, &peer_l, rules_l);
301		}
302
303		if (sigchld) {
304			sigchld = 0;
305			if (check_child(io_pid, "session engine")) {
306				quit = 1;
307				io_pid = 0;
308			}
309			if (check_child(rde_pid, "route decision engine")) {
310				quit = 1;
311				rde_pid = 0;
312			}
313		}
314
315		if (mrtdump == 1) {
316			mrtdump = 0;
317			mrt_handler(&mrt_l);
318		}
319	}
320
321	signal(SIGCHLD, SIG_IGN);
322
323	if (io_pid)
324		kill(io_pid, SIGTERM);
325
326	if (rde_pid)
327		kill(rde_pid, SIGTERM);
328
329	while ((p = peer_l) != NULL) {
330		peer_l = p->next;
331		free(p);
332	}
333	while ((m = LIST_FIRST(&mrt_l)) != NULL) {
334		LIST_REMOVE(m, entry);
335		free(m);
336	}
337	while ((la = TAILQ_FIRST(conf.listen_addrs)) != NULL) {
338		TAILQ_REMOVE(conf.listen_addrs, la, entry);
339		close(la->fd);
340		free(la);
341	}
342
343	free(rules_l);
344	control_cleanup();
345	kr_shutdown();
346	pftable_clear_all();
347	free(conf.listen_addrs);
348
349	do {
350		if ((pid = wait(NULL)) == -1 &&
351		    errno != EINTR && errno != ECHILD)
352			fatal("wait");
353	} while (pid != -1 || (pid == -1 && errno == EINTR));
354
355	msgbuf_clear(&ibuf_se->w);
356	free(ibuf_se);
357	msgbuf_clear(&ibuf_rde->w);
358	free(ibuf_rde);
359
360	log_info("Terminating");
361	return (0);
362}
363
364int
365check_child(pid_t pid, const char *pname)
366{
367	int	status;
368
369	if (waitpid(pid, &status, WNOHANG) > 0) {
370		if (WIFEXITED(status)) {
371			log_warnx("Lost child: %s exited", pname);
372			return (1);
373		}
374		if (WIFSIGNALED(status)) {
375			log_warnx("Lost child: %s terminated; signal %d",
376			    pname, WTERMSIG(status));
377			return (1);
378		}
379	}
380
381	return (0);
382}
383
384int
385send_filterset(struct imsgbuf *i, struct filter_set_head *set, int id, int f)
386{
387	struct filter_set	*s;
388
389	for (s = SIMPLEQ_FIRST(set); s != NULL; ) {
390		if (imsg_compose(i, IMSG_FILTER_SET, id, 0, -1, s,
391		    sizeof(struct filter_set)) == -1)
392			return (-1);
393		if (f) {
394			SIMPLEQ_REMOVE_HEAD(set, entry);
395			free(s);
396			s = SIMPLEQ_FIRST(set);
397		} else
398			s = SIMPLEQ_NEXT(s, entry);
399	}
400	return (0);
401}
402
403int
404reconfigure(char *conffile, struct bgpd_config *conf, struct mrt_head *mrt_l,
405    struct peer **peer_l, struct filter_head *rules_l)
406{
407	struct network_head	 net_l;
408	struct network		*n;
409	struct peer		*p;
410	struct filter_rule	*r;
411	struct listen_addr	*la;
412
413	if (parse_config(conffile, conf, mrt_l, peer_l, &net_l, rules_l)) {
414		log_warnx("config file %s has errors, not reloading",
415		    conffile);
416		return (-1);
417	}
418
419	prepare_listeners(conf);
420
421	if (imsg_compose(ibuf_se, IMSG_RECONF_CONF, 0, 0, -1,
422	    conf, sizeof(struct bgpd_config)) == -1)
423		return (-1);
424	if (imsg_compose(ibuf_rde, IMSG_RECONF_CONF, 0, 0, -1,
425	    conf, sizeof(struct bgpd_config)) == -1)
426		return (-1);
427	for (p = *peer_l; p != NULL; p = p->next) {
428		if (imsg_compose(ibuf_se, IMSG_RECONF_PEER, p->conf.id, 0, -1,
429		    &p->conf, sizeof(struct peer_config)) == -1)
430			return (-1);
431		if (send_filterset(ibuf_se, &p->conf.attrset,
432		    p->conf.id, 0) == -1)
433			return (-1);
434	}
435	while ((n = TAILQ_FIRST(&net_l)) != NULL) {
436		if (imsg_compose(ibuf_rde, IMSG_NETWORK_ADD, 0, 0, -1,
437		    &n->net, sizeof(struct network_config)) == -1)
438			return (-1);
439		if (send_filterset(ibuf_rde, &n->net.attrset, 0, 1) == -1)
440			return (-1);
441		if (imsg_compose(ibuf_rde, IMSG_NETWORK_DONE, 0, 0, -1,
442		    NULL, 0) == -1)
443			return (-1);
444		TAILQ_REMOVE(&net_l, n, entry);
445		free(n);
446	}
447	while ((r = TAILQ_FIRST(rules_l)) != NULL) {
448		if (imsg_compose(ibuf_rde, IMSG_RECONF_FILTER, 0, 0, -1,
449		    r, sizeof(struct filter_rule)) == -1)
450			return (-1);
451		if (send_filterset(ibuf_rde, &r->set, 0, 1) == -1)
452			return (-1);
453		TAILQ_REMOVE(rules_l, r, entry);
454		free(r);
455	}
456	TAILQ_FOREACH(la, conf->listen_addrs, entry) {
457		if (imsg_compose(ibuf_se, IMSG_RECONF_LISTENER, 0, 0, la->fd,
458		    la, sizeof(struct listen_addr)) == -1)
459			return (-1);
460		la->fd = -1;
461	}
462
463	if (imsg_compose(ibuf_se, IMSG_RECONF_DONE, 0, 0, -1, NULL, 0) == -1 ||
464	    imsg_compose(ibuf_rde, IMSG_RECONF_DONE, 0, 0, -1, NULL, 0) == -1)
465		return (-1);
466
467	/* mrt changes can be sent out of bound */
468	mrt_reconfigure(mrt_l);
469	return (0);
470}
471
472int
473dispatch_imsg(struct imsgbuf *ibuf, int idx)
474{
475	struct imsg		 imsg;
476	int			 n;
477	int			 rv;
478
479	if ((n = imsg_read(ibuf)) == -1)
480		return (-1);
481
482	if (n == 0) {	/* connection closed */
483		log_warnx("dispatch_imsg in main: pipe closed");
484		return (-1);
485	}
486
487	rv = 0;
488	for (;;) {
489		if ((n = imsg_get(ibuf, &imsg)) == -1)
490			return (-1);
491
492		if (n == 0)
493			break;
494
495		switch (imsg.hdr.type) {
496		case IMSG_KROUTE_CHANGE:
497			if (idx != PFD_PIPE_ROUTE)
498				log_warnx("route request not from RDE");
499			else if (kr_change(imsg.data))
500				rv = -1;
501			break;
502		case IMSG_KROUTE_DELETE:
503			if (idx != PFD_PIPE_ROUTE)
504				log_warnx("route request not from RDE");
505			else if (kr_delete(imsg.data))
506				rv = -1;
507			break;
508		case IMSG_NEXTHOP_ADD:
509			if (idx != PFD_PIPE_ROUTE)
510				log_warnx("nexthop request not from RDE");
511			else
512				if (imsg.hdr.len != IMSG_HEADER_SIZE +
513				    sizeof(struct bgpd_addr))
514					log_warnx("wrong imsg len");
515				else if (kr_nexthop_add(imsg.data) == -1)
516					rv = -1;
517			break;
518		case IMSG_NEXTHOP_REMOVE:
519			if (idx != PFD_PIPE_ROUTE)
520				log_warnx("nexthop request not from RDE");
521			else
522				if (imsg.hdr.len != IMSG_HEADER_SIZE +
523				    sizeof(struct bgpd_addr))
524					log_warnx("wrong imsg len");
525				else
526					kr_nexthop_delete(imsg.data);
527			break;
528		case IMSG_PFTABLE_ADD:
529			if (idx != PFD_PIPE_ROUTE)
530				log_warnx("pftable request not from RDE");
531			else
532				if (imsg.hdr.len != IMSG_HEADER_SIZE +
533				    sizeof(struct pftable_msg))
534					log_warnx("wrong imsg len");
535				else if (pftable_addr_add(imsg.data) != 0)
536					rv = -1;
537			break;
538		case IMSG_PFTABLE_REMOVE:
539			if (idx != PFD_PIPE_ROUTE)
540				log_warnx("pftable request not from RDE");
541			else
542				if (imsg.hdr.len != IMSG_HEADER_SIZE +
543				    sizeof(struct pftable_msg))
544					log_warnx("wrong imsg len");
545				else if (pftable_addr_remove(imsg.data) != 0)
546					rv = -1;
547			break;
548		case IMSG_PFTABLE_COMMIT:
549			if (idx != PFD_PIPE_ROUTE)
550				log_warnx("pftable request not from RDE");
551			else
552				if (imsg.hdr.len != IMSG_HEADER_SIZE)
553					log_warnx("wrong imsg len");
554				else if (pftable_commit() != 0)
555					rv = -1;
556			break;
557		case IMSG_CTL_RELOAD:
558			if (idx != PFD_PIPE_SESSION)
559				log_warnx("reload request not from SE");
560			else
561				reconfig = 1;
562			break;
563		case IMSG_CTL_FIB_COUPLE:
564			if (idx != PFD_PIPE_SESSION)
565				log_warnx("couple request not from SE");
566			else
567				kr_fib_couple();
568			break;
569		case IMSG_CTL_FIB_DECOUPLE:
570			if (idx != PFD_PIPE_SESSION)
571				log_warnx("decouple request not from SE");
572			else
573				kr_fib_decouple();
574			break;
575		case IMSG_CTL_KROUTE:
576		case IMSG_CTL_KROUTE_ADDR:
577		case IMSG_CTL_SHOW_NEXTHOP:
578		case IMSG_CTL_SHOW_INTERFACE:
579			if (idx != PFD_PIPE_SESSION)
580				log_warnx("kroute request not from SE");
581			else
582				kr_show_route(&imsg);
583			break;
584		case IMSG_IFINFO:
585			if (idx != PFD_PIPE_SESSION)
586				log_warnx("IFINFO request not from SE");
587			else if (imsg.hdr.len != IMSG_HEADER_SIZE + IFNAMSIZ)
588				log_warnx("IFINFO request with wrong len");
589			else
590				kr_ifinfo(imsg.data);
591			break;
592		default:
593			break;
594		}
595		imsg_free(&imsg);
596		if (rv != 0)
597			return (rv);
598	}
599	return (0);
600}
601
602void
603send_nexthop_update(struct kroute_nexthop *msg)
604{
605	char	*gw = NULL;
606
607	if (msg->gateway.af)
608		if (asprintf(&gw, ": via %s",
609		    log_addr(&msg->gateway)) == -1) {
610			log_warn("send_nexthop_update");
611			quit = 1;
612		}
613
614	log_info("nexthop %s now %s%s%s", log_addr(&msg->nexthop),
615	    msg->valid ? "valid" : "invalid",
616	    msg->connected ? ": directly connected" : "",
617	    msg->gateway.af ? gw : "");
618
619	free(gw);
620
621	if (imsg_compose(ibuf_rde, IMSG_NEXTHOP_UPDATE, 0, 0, -1,
622	    msg, sizeof(struct kroute_nexthop)) == -1)
623		quit = 1;
624}
625
626void
627send_imsg_session(int type, pid_t pid, void *data, u_int16_t datalen)
628{
629	imsg_compose(ibuf_se, type, 0, pid, -1, data, datalen);
630}
631