bgpd.c revision 1.101 
1/*	$OpenBSD: bgpd.c,v 1.101 2004/07/05 02:13:43 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, struct mrt_head *);
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 (fcntl(pipe_m2s[0], F_SETFL, O_NONBLOCK) == -1 ||
182	    fcntl(pipe_m2s[1], F_SETFL, O_NONBLOCK) == -1)
183		fatal("fcntl");
184	if (socketpair(AF_UNIX, SOCK_STREAM, PF_UNSPEC, pipe_m2r) == -1)
185		fatal("socketpair");
186	if (fcntl(pipe_m2r[0], F_SETFL, O_NONBLOCK) == -1 ||
187	    fcntl(pipe_m2r[1], F_SETFL, O_NONBLOCK) == -1)
188		fatal("fcntl");
189	if (socketpair(AF_UNIX, SOCK_STREAM, PF_UNSPEC, pipe_s2r) == -1)
190		fatal("socketpair");
191	if (fcntl(pipe_s2r[0], F_SETFL, O_NONBLOCK) == -1 ||
192	    fcntl(pipe_s2r[1], F_SETFL, O_NONBLOCK) == -1)
193		fatal("fcntl");
194
195	prepare_listeners(&conf);
196
197	/* fork children */
198	rde_pid = rde_main(&conf, &net_l, rules_l, &mrt_l, pipe_m2r, pipe_s2r,
199	    pipe_m2s);
200	io_pid = session_main(&conf, peer_l, &net_l, rules_l, &mrt_l,
201	    pipe_m2s, pipe_s2r, pipe_m2r);
202
203	setproctitle("parent");
204
205	signal(SIGTERM, sighdlr);
206	signal(SIGINT, sighdlr);
207	signal(SIGCHLD, sighdlr);
208	signal(SIGHUP, sighdlr);
209	signal(SIGALRM, sighdlr);
210	signal(SIGUSR1, sighdlr);
211
212	close(pipe_m2s[1]);
213	close(pipe_m2r[1]);
214	close(pipe_s2r[0]);
215	close(pipe_s2r[1]);
216
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,
280			    &mrt_l) == -1)
281				quit = 1;
282		}
283
284		if (nfds > 0 && pfd[PFD_PIPE_ROUTE].revents & POLLIN) {
285			nfds--;
286			if (dispatch_imsg(&ibuf_rde, PFD_PIPE_ROUTE,
287			    &mrt_l) == -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			log_info("rereading config");
299			reconfigure(conffile, &conf, &mrt_l, &peer_l, rules_l);
300			reconfig = 0;
301		}
302
303		if (sigchld) {
304			if (check_child(io_pid, "session engine"))
305				quit = 1;
306			if (check_child(rde_pid, "route decision engine"))
307				quit = 1;
308			sigchld = 0;
309		}
310
311		if (mrtdump == 1) {
312			mrt_handler(&mrt_l);
313			mrtdump = 0;
314		}
315	}
316
317	signal(SIGCHLD, SIG_IGN);
318
319	if (io_pid)
320		kill(io_pid, SIGTERM);
321
322	if (rde_pid)
323		kill(rde_pid, SIGTERM);
324
325	while ((p = peer_l) != NULL) {
326		peer_l = p->next;
327		free(p);
328	}
329	while ((m = LIST_FIRST(&mrt_l)) != NULL) {
330		LIST_REMOVE(m, entry);
331		free(m);
332	}
333
334	free(rules_l);
335	control_cleanup();
336	kr_shutdown();
337	pftable_clear_all();
338
339	do {
340		if ((pid = wait(NULL)) == -1 &&
341		    errno != EINTR && errno != ECHILD)
342			fatal("wait");
343	} while (pid != -1 || (pid == -1 && errno == EINTR));
344
345	log_info("Terminating");
346	return (0);
347}
348
349int
350check_child(pid_t pid, const char *pname)
351{
352	int	status;
353
354	if (waitpid(pid, &status, WNOHANG) > 0) {
355		if (WIFEXITED(status)) {
356			log_warnx("Lost child: %s exited", pname);
357			return (1);
358		}
359		if (WIFSIGNALED(status)) {
360			log_warnx("Lost child: %s terminated; signal %d",
361			    pname, WTERMSIG(status));
362			return (1);
363		}
364	}
365
366	return (0);
367}
368
369int
370reconfigure(char *conffile, struct bgpd_config *conf, struct mrt_head *mrt_l,
371    struct peer **peer_l, struct filter_head *rules_l)
372{
373	struct network_head	 net_l;
374	struct network		*n;
375	struct peer		*p;
376	struct filter_rule	*r;
377	struct listen_addr	*la;
378
379	if (parse_config(conffile, conf, mrt_l, peer_l, &net_l, rules_l)) {
380		log_warnx("config file %s has errors, not reloading",
381		    conffile);
382		return (-1);
383	}
384
385	prepare_listeners(conf);
386
387	if (imsg_compose(&ibuf_se, IMSG_RECONF_CONF, 0,
388	    conf, sizeof(struct bgpd_config)) == -1)
389		return (-1);
390	if (imsg_compose(&ibuf_rde, IMSG_RECONF_CONF, 0,
391	    conf, sizeof(struct bgpd_config)) == -1)
392		return (-1);
393	for (p = *peer_l; p != NULL; p = p->next)
394		if (imsg_compose(&ibuf_se, IMSG_RECONF_PEER, p->conf.id,
395		    &p->conf, sizeof(struct peer_config)) == -1)
396			return (-1);
397	while ((n = TAILQ_FIRST(&net_l)) != NULL) {
398		if (imsg_compose(&ibuf_rde, IMSG_NETWORK_ADD, 0,
399		    &n->net, sizeof(struct network_config)) == -1)
400			return (-1);
401		TAILQ_REMOVE(&net_l, n, entry);
402		free(n);
403	}
404	while ((r = TAILQ_FIRST(rules_l)) != NULL) {
405		if (imsg_compose(&ibuf_rde, IMSG_RECONF_FILTER, 0,
406		    r, sizeof(struct filter_rule)) == -1)
407			return (-1);
408		TAILQ_REMOVE(rules_l, r, entry);
409		free(r);
410	}
411	while ((la = TAILQ_FIRST(conf->listen_addrs)) != NULL) {
412		if (imsg_compose_fdpass(&ibuf_se, IMSG_RECONF_LISTENER, la->fd,
413		    la, sizeof(struct listen_addr)) == -1)
414			return (-1);
415		TAILQ_REMOVE(conf->listen_addrs, la, entry);
416		free(la);
417	}
418	free(conf->listen_addrs);
419	conf->listen_addrs = NULL;
420
421	if (imsg_compose(&ibuf_se, IMSG_RECONF_DONE, 0, NULL, 0) == -1 ||
422	    imsg_compose(&ibuf_rde, IMSG_RECONF_DONE, 0, NULL, 0) == -1)
423		return (-1);
424
425	/* mrt changes can be sent out of bound */
426	mrt_reconfigure(mrt_l);
427	return (0);
428}
429
430int
431dispatch_imsg(struct imsgbuf *ibuf, int idx, struct mrt_head *mrt_l)
432{
433	struct imsg		 imsg;
434	struct mrt		 mrt;
435	int			 n;
436
437	if ((n = imsg_read(ibuf)) == -1)
438		return (-1);
439
440	if (n == 0) {	/* connection closed */
441		log_warnx("dispatch_imsg in main: pipe closed");
442		return (-1);
443	}
444
445	for (;;) {
446		if ((n = imsg_get(ibuf, &imsg)) == -1)
447			return (-1);
448
449		if (n == 0)
450			break;
451
452		switch (imsg.hdr.type) {
453		case IMSG_MRT_CLOSE:
454			if (imsg.hdr.len != IMSG_HEADER_SIZE +
455			    sizeof(struct mrt)) {
456				log_warnx("wrong imsg len");
457				break;
458			}
459			memcpy(&mrt, imsg.data, sizeof(struct mrt));
460			mrt_close(mrt_get(mrt_l, &mrt));
461			break;
462		case IMSG_KROUTE_CHANGE:
463			if (idx != PFD_PIPE_ROUTE)
464				log_warnx("route request not from RDE");
465			else if (kr_change(imsg.data))
466				return (-1);
467			break;
468		case IMSG_KROUTE_DELETE:
469			if (idx != PFD_PIPE_ROUTE)
470				log_warnx("route request not from RDE");
471			else if (kr_delete(imsg.data))
472				return (-1);
473			break;
474		case IMSG_NEXTHOP_ADD:
475			if (idx != PFD_PIPE_ROUTE)
476				log_warnx("nexthop request not from RDE");
477			else
478				if (imsg.hdr.len != IMSG_HEADER_SIZE +
479				    sizeof(struct bgpd_addr))
480					log_warnx("wrong imsg len");
481				else if (kr_nexthop_add(imsg.data) == -1)
482					return (-1);
483			break;
484		case IMSG_NEXTHOP_REMOVE:
485			if (idx != PFD_PIPE_ROUTE)
486				log_warnx("nexthop request not from RDE");
487			else
488				if (imsg.hdr.len != IMSG_HEADER_SIZE +
489				    sizeof(struct bgpd_addr))
490					log_warnx("wrong imsg len");
491				else
492					kr_nexthop_delete(imsg.data);
493			break;
494		case IMSG_PFTABLE_ADD:
495			if (idx != PFD_PIPE_ROUTE)
496				log_warnx("pftable request not from RDE");
497			else
498				if (imsg.hdr.len != IMSG_HEADER_SIZE +
499				    sizeof(struct pftable_msg))
500					log_warnx("wrong imsg len");
501				else if (pftable_addr_add(imsg.data) != 0)
502					return (-1);
503			break;
504		case IMSG_PFTABLE_REMOVE:
505			if (idx != PFD_PIPE_ROUTE)
506				log_warnx("pftable request not from RDE");
507			else
508				if (imsg.hdr.len != IMSG_HEADER_SIZE +
509				    sizeof(struct pftable_msg))
510					log_warnx("wrong imsg len");
511				else if (pftable_addr_remove(imsg.data) != 0)
512					return (-1);
513			break;
514		case IMSG_PFTABLE_COMMIT:
515			if (idx != PFD_PIPE_ROUTE)
516				log_warnx("pftable request not from RDE");
517			else
518				if (imsg.hdr.len != IMSG_HEADER_SIZE)
519					log_warnx("wrong imsg len");
520				else if (pftable_commit() != 0)
521					return (-1);
522			break;
523		case IMSG_CTL_RELOAD:
524			if (idx != PFD_PIPE_SESSION)
525				log_warnx("reload request not from SE");
526			else
527				reconfig = 1;
528			break;
529		case IMSG_CTL_FIB_COUPLE:
530			if (idx != PFD_PIPE_SESSION)
531				log_warnx("couple request not from SE");
532			else
533				kr_fib_couple();
534			break;
535		case IMSG_CTL_FIB_DECOUPLE:
536			if (idx != PFD_PIPE_SESSION)
537				log_warnx("decouple request not from SE");
538			else
539				kr_fib_decouple();
540			break;
541		case IMSG_CTL_KROUTE:
542		case IMSG_CTL_KROUTE_ADDR:
543		case IMSG_CTL_SHOW_NEXTHOP:
544		case IMSG_CTL_SHOW_INTERFACE:
545			if (idx != PFD_PIPE_SESSION)
546				log_warnx("kroute request not from SE");
547			else
548				kr_show_route(&imsg);
549			break;
550		default:
551			break;
552		}
553		imsg_free(&imsg);
554	}
555	return (0);
556}
557
558void
559send_nexthop_update(struct kroute_nexthop *msg)
560{
561	char	*gw = NULL;
562
563	if (msg->gateway.af)
564		if (asprintf(&gw, ": via %s",
565		    log_addr(&msg->gateway)) == -1) {
566			log_warn("send_nexthop_update");
567			quit = 1;
568		}
569
570	log_info("nexthop %s now %s%s%s", log_addr(&msg->nexthop),
571	    msg->valid ? "valid" : "invalid",
572	    msg->connected ? ": directly connected" : "",
573	    msg->gateway.af ? gw : "");
574
575	free(gw);
576
577	if (imsg_compose(&ibuf_rde, IMSG_NEXTHOP_UPDATE, 0,
578	    msg, sizeof(struct kroute_nexthop)) == -1)
579		quit = 1;
580}
581
582void
583send_imsg_session(int type, pid_t pid, void *data, u_int16_t datalen)
584{
585	imsg_compose_pid(&ibuf_se, type, pid, data, datalen);
586}
587