bgpd.c revision 1.107 
1/*	$OpenBSD: bgpd.c,v 1.107 2004/09/16 00:25:12 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			if (check_child(rde_pid, "route decision engine"))
306				quit = 1;
307		}
308
309		if (mrtdump == 1) {
310			mrtdump = 0;
311			mrt_handler(&mrt_l);
312		}
313	}
314
315	signal(SIGCHLD, SIG_IGN);
316
317	if (io_pid)
318		kill(io_pid, SIGTERM);
319
320	if (rde_pid)
321		kill(rde_pid, SIGTERM);
322
323	while ((p = peer_l) != NULL) {
324		peer_l = p->next;
325		free(p);
326	}
327	while ((m = LIST_FIRST(&mrt_l)) != NULL) {
328		LIST_REMOVE(m, entry);
329		free(m);
330	}
331
332	free(rules_l);
333	control_cleanup();
334	kr_shutdown();
335	pftable_clear_all();
336	free(conf.listen_addrs);
337
338	do {
339		if ((pid = wait(NULL)) == -1 &&
340		    errno != EINTR && errno != ECHILD)
341			fatal("wait");
342	} while (pid != -1 || (pid == -1 && errno == EINTR));
343
344	msgbuf_clear(&ibuf_se->w);
345	free(ibuf_se);
346	msgbuf_clear(&ibuf_rde->w);
347	free(ibuf_rde);
348
349	log_info("Terminating");
350	return (0);
351}
352
353int
354check_child(pid_t pid, const char *pname)
355{
356	int	status;
357
358	if (waitpid(pid, &status, WNOHANG) > 0) {
359		if (WIFEXITED(status)) {
360			log_warnx("Lost child: %s exited", pname);
361			return (1);
362		}
363		if (WIFSIGNALED(status)) {
364			log_warnx("Lost child: %s terminated; signal %d",
365			    pname, WTERMSIG(status));
366			return (1);
367		}
368	}
369
370	return (0);
371}
372
373int
374reconfigure(char *conffile, struct bgpd_config *conf, struct mrt_head *mrt_l,
375    struct peer **peer_l, struct filter_head *rules_l)
376{
377	struct network_head	 net_l;
378	struct network		*n;
379	struct peer		*p;
380	struct filter_rule	*r;
381	struct listen_addr	*la;
382
383	if (parse_config(conffile, conf, mrt_l, peer_l, &net_l, rules_l)) {
384		log_warnx("config file %s has errors, not reloading",
385		    conffile);
386		return (-1);
387	}
388
389	prepare_listeners(conf);
390
391	if (imsg_compose(ibuf_se, IMSG_RECONF_CONF, 0,
392	    conf, sizeof(struct bgpd_config)) == -1)
393		return (-1);
394	if (imsg_compose(ibuf_rde, IMSG_RECONF_CONF, 0,
395	    conf, sizeof(struct bgpd_config)) == -1)
396		return (-1);
397	for (p = *peer_l; p != NULL; p = p->next)
398		if (imsg_compose(ibuf_se, IMSG_RECONF_PEER, p->conf.id,
399		    &p->conf, sizeof(struct peer_config)) == -1)
400			return (-1);
401	while ((n = TAILQ_FIRST(&net_l)) != NULL) {
402		if (imsg_compose(ibuf_rde, IMSG_NETWORK_ADD, 0,
403		    &n->net, sizeof(struct network_config)) == -1)
404			return (-1);
405		TAILQ_REMOVE(&net_l, n, entry);
406		free(n);
407	}
408	while ((r = TAILQ_FIRST(rules_l)) != NULL) {
409		if (imsg_compose(ibuf_rde, IMSG_RECONF_FILTER, 0,
410		    r, sizeof(struct filter_rule)) == -1)
411			return (-1);
412		TAILQ_REMOVE(rules_l, r, entry);
413		free(r);
414	}
415	while ((la = TAILQ_FIRST(conf->listen_addrs)) != NULL) {
416		if (imsg_compose_fdpass(ibuf_se, IMSG_RECONF_LISTENER, la->fd,
417		    la, sizeof(struct listen_addr)) == -1)
418			return (-1);
419		TAILQ_REMOVE(conf->listen_addrs, la, entry);
420		free(la);
421	}
422	free(conf->listen_addrs);
423	conf->listen_addrs = NULL;
424
425	if (imsg_compose(ibuf_se, IMSG_RECONF_DONE, 0, NULL, 0) == -1 ||
426	    imsg_compose(ibuf_rde, IMSG_RECONF_DONE, 0, NULL, 0) == -1)
427		return (-1);
428
429	/* mrt changes can be sent out of bound */
430	mrt_reconfigure(mrt_l);
431	return (0);
432}
433
434int
435dispatch_imsg(struct imsgbuf *ibuf, int idx)
436{
437	struct imsg		 imsg;
438	int			 n;
439
440	if ((n = imsg_read(ibuf)) == -1)
441		return (-1);
442
443	if (n == 0) {	/* connection closed */
444		log_warnx("dispatch_imsg in main: pipe closed");
445		return (-1);
446	}
447
448	for (;;) {
449		if ((n = imsg_get(ibuf, &imsg)) == -1)
450			return (-1);
451
452		if (n == 0)
453			break;
454
455		switch (imsg.hdr.type) {
456		case IMSG_KROUTE_CHANGE:
457			if (idx != PFD_PIPE_ROUTE)
458				log_warnx("route request not from RDE");
459			else if (kr_change(imsg.data))
460				return (-1);
461			break;
462		case IMSG_KROUTE_DELETE:
463			if (idx != PFD_PIPE_ROUTE)
464				log_warnx("route request not from RDE");
465			else if (kr_delete(imsg.data))
466				return (-1);
467			break;
468		case IMSG_NEXTHOP_ADD:
469			if (idx != PFD_PIPE_ROUTE)
470				log_warnx("nexthop request not from RDE");
471			else
472				if (imsg.hdr.len != IMSG_HEADER_SIZE +
473				    sizeof(struct bgpd_addr))
474					log_warnx("wrong imsg len");
475				else if (kr_nexthop_add(imsg.data) == -1)
476					return (-1);
477			break;
478		case IMSG_NEXTHOP_REMOVE:
479			if (idx != PFD_PIPE_ROUTE)
480				log_warnx("nexthop request not from RDE");
481			else
482				if (imsg.hdr.len != IMSG_HEADER_SIZE +
483				    sizeof(struct bgpd_addr))
484					log_warnx("wrong imsg len");
485				else
486					kr_nexthop_delete(imsg.data);
487			break;
488		case IMSG_PFTABLE_ADD:
489			if (idx != PFD_PIPE_ROUTE)
490				log_warnx("pftable request not from RDE");
491			else
492				if (imsg.hdr.len != IMSG_HEADER_SIZE +
493				    sizeof(struct pftable_msg))
494					log_warnx("wrong imsg len");
495				else if (pftable_addr_add(imsg.data) != 0)
496					return (-1);
497			break;
498		case IMSG_PFTABLE_REMOVE:
499			if (idx != PFD_PIPE_ROUTE)
500				log_warnx("pftable request not from RDE");
501			else
502				if (imsg.hdr.len != IMSG_HEADER_SIZE +
503				    sizeof(struct pftable_msg))
504					log_warnx("wrong imsg len");
505				else if (pftable_addr_remove(imsg.data) != 0)
506					return (-1);
507			break;
508		case IMSG_PFTABLE_COMMIT:
509			if (idx != PFD_PIPE_ROUTE)
510				log_warnx("pftable request not from RDE");
511			else
512				if (imsg.hdr.len != IMSG_HEADER_SIZE)
513					log_warnx("wrong imsg len");
514				else if (pftable_commit() != 0)
515					return (-1);
516			break;
517		case IMSG_CTL_RELOAD:
518			if (idx != PFD_PIPE_SESSION)
519				log_warnx("reload request not from SE");
520			else
521				reconfig = 1;
522			break;
523		case IMSG_CTL_FIB_COUPLE:
524			if (idx != PFD_PIPE_SESSION)
525				log_warnx("couple request not from SE");
526			else
527				kr_fib_couple();
528			break;
529		case IMSG_CTL_FIB_DECOUPLE:
530			if (idx != PFD_PIPE_SESSION)
531				log_warnx("decouple request not from SE");
532			else
533				kr_fib_decouple();
534			break;
535		case IMSG_CTL_KROUTE:
536		case IMSG_CTL_KROUTE_ADDR:
537		case IMSG_CTL_SHOW_NEXTHOP:
538		case IMSG_CTL_SHOW_INTERFACE:
539			if (idx != PFD_PIPE_SESSION)
540				log_warnx("kroute request not from SE");
541			else
542				kr_show_route(&imsg);
543			break;
544		default:
545			break;
546		}
547		imsg_free(&imsg);
548	}
549	return (0);
550}
551
552void
553send_nexthop_update(struct kroute_nexthop *msg)
554{
555	char	*gw = NULL;
556
557	if (msg->gateway.af)
558		if (asprintf(&gw, ": via %s",
559		    log_addr(&msg->gateway)) == -1) {
560			log_warn("send_nexthop_update");
561			quit = 1;
562		}
563
564	log_info("nexthop %s now %s%s%s", log_addr(&msg->nexthop),
565	    msg->valid ? "valid" : "invalid",
566	    msg->connected ? ": directly connected" : "",
567	    msg->gateway.af ? gw : "");
568
569	free(gw);
570
571	if (imsg_compose(ibuf_rde, IMSG_NEXTHOP_UPDATE, 0,
572	    msg, sizeof(struct kroute_nexthop)) == -1)
573		quit = 1;
574}
575
576void
577send_imsg_session(int type, pid_t pid, void *data, u_int16_t datalen)
578{
579	imsg_compose_pid(ibuf_se, type, pid, data, datalen);
580}
581