bgpd.c revision 1.98 
1/*	$OpenBSD: bgpd.c,v 1.98 2004/07/03 17:19:59 claudio 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	io_pid = session_main(&conf, peer_l, &net_l, rules_l, &mrt_l,
200	    pipe_m2s, pipe_s2r);
201
202	setproctitle("parent");
203
204	signal(SIGTERM, sighdlr);
205	signal(SIGINT, sighdlr);
206	signal(SIGCHLD, sighdlr);
207	signal(SIGHUP, sighdlr);
208	signal(SIGALRM, sighdlr);
209	signal(SIGUSR1, sighdlr);
210
211	close(pipe_m2s[1]);
212	close(pipe_m2r[1]);
213	close(pipe_s2r[0]);
214	close(pipe_s2r[1]);
215
216	imsg_init(&ibuf_se, pipe_m2s[0]);
217	imsg_init(&ibuf_rde, pipe_m2r[0]);
218	mrt_init(&ibuf_rde, &ibuf_se);
219	if ((rfd = kr_init(!(conf.flags & BGPD_FLAG_NO_FIB_UPDATE))) == -1)
220		quit = 1;
221	if (pftable_clear_all() != 0)
222		quit = 1;
223
224	while ((net = TAILQ_FIRST(&net_l)) != NULL) {
225		TAILQ_REMOVE(&net_l, net, entry);
226		free(net);
227	}
228
229	while ((r = TAILQ_FIRST(rules_l)) != NULL) {
230		TAILQ_REMOVE(rules_l, r, entry);
231		free(r);
232	}
233
234	while ((la = TAILQ_FIRST(conf.listen_addrs)) != NULL) {
235		TAILQ_REMOVE(conf.listen_addrs, la, entry);
236		free(la);
237	}
238
239	mrt_reconfigure(&mrt_l);
240
241	while (quit == 0) {
242		pfd[PFD_PIPE_SESSION].fd = ibuf_se.fd;
243		pfd[PFD_PIPE_SESSION].events = POLLIN;
244		if (ibuf_se.w.queued)
245			pfd[PFD_PIPE_SESSION].events |= POLLOUT;
246		pfd[PFD_PIPE_ROUTE].fd = ibuf_rde.fd;
247		pfd[PFD_PIPE_ROUTE].events = POLLIN;
248		if (ibuf_rde.w.queued)
249			pfd[PFD_PIPE_ROUTE].events |= POLLOUT;
250		pfd[PFD_SOCK_ROUTE].fd = rfd;
251		pfd[PFD_SOCK_ROUTE].events = POLLIN;
252
253		timeout = mrt_timeout(&mrt_l);
254		if (timeout > MAX_TIMEOUT)
255			timeout = MAX_TIMEOUT;
256
257		if ((nfds = poll(pfd, POLL_MAX, timeout * 1000)) == -1)
258			if (errno != EINTR) {
259				log_warn("poll error");
260				quit = 1;
261			}
262
263		if (nfds > 0 && (pfd[PFD_PIPE_SESSION].revents & POLLOUT))
264			if (msgbuf_write(&ibuf_se.w) < 0) {
265				log_warn("pipe write error (to SE)");
266				quit = 1;
267			}
268
269		if (nfds > 0 && (pfd[PFD_PIPE_ROUTE].revents & POLLOUT))
270			if (msgbuf_write(&ibuf_rde.w) < 0) {
271				log_warn("pipe write error (to RDE)");
272				quit = 1;
273			}
274
275		if (nfds > 0 && pfd[PFD_PIPE_SESSION].revents & POLLIN) {
276			nfds--;
277			if (dispatch_imsg(&ibuf_se, PFD_PIPE_SESSION,
278			    &mrt_l) == -1)
279				quit = 1;
280		}
281
282		if (nfds > 0 && pfd[PFD_PIPE_ROUTE].revents & POLLIN) {
283			nfds--;
284			if (dispatch_imsg(&ibuf_rde, PFD_PIPE_ROUTE,
285			    &mrt_l) == -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			log_info("rereading config");
297			reconfigure(conffile, &conf, &mrt_l, &peer_l, rules_l);
298			reconfig = 0;
299		}
300
301		if (sigchld) {
302			if (check_child(io_pid, "session engine"))
303				quit = 1;
304			if (check_child(rde_pid, "route decision engine"))
305				quit = 1;
306			sigchld = 0;
307		}
308
309		if (mrtdump == 1) {
310			mrt_handler(&mrt_l);
311			mrtdump = 0;
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
337	do {
338		if ((pid = wait(NULL)) == -1 &&
339		    errno != EINTR && errno != ECHILD)
340			fatal("wait");
341	} while (pid != -1 || (pid == -1 && errno == EINTR));
342
343	log_info("Terminating");
344	return (0);
345}
346
347int
348check_child(pid_t pid, const char *pname)
349{
350	int	status;
351
352	if (waitpid(pid, &status, WNOHANG) > 0) {
353		if (WIFEXITED(status)) {
354			log_warnx("Lost child: %s exited", pname);
355			return (1);
356		}
357		if (WIFSIGNALED(status)) {
358			log_warnx("Lost child: %s terminated; signal %d",
359			    pname, WTERMSIG(status));
360			return (1);
361		}
362	}
363
364	return (0);
365}
366
367int
368reconfigure(char *conffile, struct bgpd_config *conf, struct mrt_head *mrt_l,
369    struct peer **peer_l, struct filter_head *rules_l)
370{
371	struct network_head	 net_l;
372	struct network		*n;
373	struct peer		*p;
374	struct filter_rule	*r;
375	struct listen_addr	*la;
376
377	if (parse_config(conffile, conf, mrt_l, peer_l, &net_l, rules_l)) {
378		log_warnx("config file %s has errors, not reloading",
379		    conffile);
380		return (-1);
381	}
382
383	prepare_listeners(conf);
384
385	if (imsg_compose(&ibuf_se, IMSG_RECONF_CONF, 0,
386	    conf, sizeof(struct bgpd_config)) == -1)
387		return (-1);
388	if (imsg_compose(&ibuf_rde, IMSG_RECONF_CONF, 0,
389	    conf, sizeof(struct bgpd_config)) == -1)
390		return (-1);
391	for (p = *peer_l; p != NULL; p = p->next)
392		if (imsg_compose(&ibuf_se, IMSG_RECONF_PEER, p->conf.id,
393		    &p->conf, sizeof(struct peer_config)) == -1)
394			return (-1);
395	while ((n = TAILQ_FIRST(&net_l)) != NULL) {
396		if (imsg_compose(&ibuf_rde, IMSG_NETWORK_ADD, 0,
397		    &n->net, sizeof(struct network_config)) == -1)
398			return (-1);
399		TAILQ_REMOVE(&net_l, n, entry);
400		free(n);
401	}
402	while ((r = TAILQ_FIRST(rules_l)) != NULL) {
403		if (imsg_compose(&ibuf_rde, IMSG_RECONF_FILTER, 0,
404		    r, sizeof(struct filter_rule)) == -1)
405			return (-1);
406		TAILQ_REMOVE(rules_l, r, entry);
407		free(r);
408	}
409	while ((la = TAILQ_FIRST(conf->listen_addrs)) != NULL) {
410		if (imsg_compose_fdpass(&ibuf_se, IMSG_RECONF_LISTENER, la->fd,
411		    la, sizeof(struct listen_addr)) == -1)
412			return (-1);
413		TAILQ_REMOVE(conf->listen_addrs, la, entry);
414		free(la);
415	}
416	free(conf->listen_addrs);
417	conf->listen_addrs = NULL;
418
419	if (imsg_compose(&ibuf_se, IMSG_RECONF_DONE, 0, NULL, 0) == -1 ||
420	    imsg_compose(&ibuf_rde, IMSG_RECONF_DONE, 0, NULL, 0) == -1)
421		return (-1);
422
423	/* mrt changes can be sent out of bound */
424	mrt_reconfigure(mrt_l);
425	return (0);
426}
427
428int
429dispatch_imsg(struct imsgbuf *ibuf, int idx, struct mrt_head *mrt_l)
430{
431	struct imsg		 imsg;
432	struct mrt		 mrt;
433	int			 n;
434
435	if ((n = imsg_read(ibuf)) == -1)
436		return (-1);
437
438	if (n == 0) {	/* connection closed */
439		log_warnx("dispatch_imsg in main: pipe closed");
440		return (-1);
441	}
442
443	for (;;) {
444		if ((n = imsg_get(ibuf, &imsg)) == -1)
445			return (-1);
446
447		if (n == 0)
448			break;
449
450		switch (imsg.hdr.type) {
451		case IMSG_MRT_CLOSE:
452			if (imsg.hdr.len != IMSG_HEADER_SIZE +
453			    sizeof(struct mrt)) {
454				log_warnx("wrong imsg len");
455				break;
456			}
457			memcpy(&mrt, imsg.data, sizeof(struct mrt));
458			mrt_close(mrt_get(mrt_l, &mrt));
459			break;
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		default:
549			break;
550		}
551		imsg_free(&imsg);
552	}
553	return (0);
554}
555
556void
557send_nexthop_update(struct kroute_nexthop *msg)
558{
559	char	*gw = NULL;
560
561	if (msg->gateway.af)
562		if (asprintf(&gw, ": via %s",
563		    log_addr(&msg->gateway)) == -1) {
564			log_warn("send_nexthop_update");
565			quit = 1;
566		}
567
568	log_info("nexthop %s now %s%s%s", log_addr(&msg->nexthop),
569	    msg->valid ? "valid" : "invalid",
570	    msg->connected ? ": directly connected" : "",
571	    msg->gateway.af ? gw : "");
572
573	free(gw);
574
575	if (imsg_compose(&ibuf_rde, IMSG_NEXTHOP_UPDATE, 0,
576	    msg, sizeof(struct kroute_nexthop)) == -1)
577		quit = 1;
578}
579
580void
581send_imsg_session(int type, pid_t pid, void *data, u_int16_t datalen)
582{
583	imsg_compose_pid(&ibuf_se, type, pid, data, datalen);
584}
585