bgpd.c revision 1.106 
1/*	$OpenBSD: bgpd.c,v 1.106 2004/09/15 18:30:42 otto 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	imsg_init(&ibuf_se, pipe_m2s[0]);
215	imsg_init(&ibuf_rde, pipe_m2r[0]);
216	mrt_init(&ibuf_rde, &ibuf_se);
217	if ((rfd = kr_init(!(conf.flags & BGPD_FLAG_NO_FIB_UPDATE))) == -1)
218		quit = 1;
219	if (pftable_clear_all() != 0)
220		quit = 1;
221
222	while ((net = TAILQ_FIRST(&net_l)) != NULL) {
223		TAILQ_REMOVE(&net_l, net, entry);
224		free(net);
225	}
226
227	while ((r = TAILQ_FIRST(rules_l)) != NULL) {
228		TAILQ_REMOVE(rules_l, r, entry);
229		free(r);
230	}
231
232	while ((la = TAILQ_FIRST(conf.listen_addrs)) != NULL) {
233		TAILQ_REMOVE(conf.listen_addrs, la, entry);
234		close(la->fd);
235		free(la);
236	}
237
238	mrt_reconfigure(&mrt_l);
239
240	while (quit == 0) {
241		pfd[PFD_PIPE_SESSION].fd = ibuf_se.fd;
242		pfd[PFD_PIPE_SESSION].events = POLLIN;
243		if (ibuf_se.w.queued)
244			pfd[PFD_PIPE_SESSION].events |= POLLOUT;
245		pfd[PFD_PIPE_ROUTE].fd = ibuf_rde.fd;
246		pfd[PFD_PIPE_ROUTE].events = POLLIN;
247		if (ibuf_rde.w.queued)
248			pfd[PFD_PIPE_ROUTE].events |= POLLOUT;
249		pfd[PFD_SOCK_ROUTE].fd = rfd;
250		pfd[PFD_SOCK_ROUTE].events = POLLIN;
251
252		timeout = mrt_timeout(&mrt_l);
253		if (timeout > MAX_TIMEOUT)
254			timeout = MAX_TIMEOUT;
255
256		if ((nfds = poll(pfd, POLL_MAX, timeout * 1000)) == -1)
257			if (errno != EINTR) {
258				log_warn("poll error");
259				quit = 1;
260			}
261
262		if (nfds > 0 && (pfd[PFD_PIPE_SESSION].revents & POLLOUT))
263			if (msgbuf_write(&ibuf_se.w) < 0) {
264				log_warn("pipe write error (to SE)");
265				quit = 1;
266			}
267
268		if (nfds > 0 && (pfd[PFD_PIPE_ROUTE].revents & POLLOUT))
269			if (msgbuf_write(&ibuf_rde.w) < 0) {
270				log_warn("pipe write error (to RDE)");
271				quit = 1;
272			}
273
274		if (nfds > 0 && pfd[PFD_PIPE_SESSION].revents & POLLIN) {
275			nfds--;
276			if (dispatch_imsg(&ibuf_se, PFD_PIPE_SESSION) == -1)
277				quit = 1;
278		}
279
280		if (nfds > 0 && pfd[PFD_PIPE_ROUTE].revents & POLLIN) {
281			nfds--;
282			if (dispatch_imsg(&ibuf_rde, PFD_PIPE_ROUTE) == -1)
283				quit = 1;
284		}
285
286		if (nfds > 0 && pfd[PFD_SOCK_ROUTE].revents & POLLIN) {
287			nfds--;
288			if (kr_dispatch_msg() == -1)
289				quit = 1;
290		}
291
292		if (reconfig) {
293			reconfig = 0;
294			log_info("rereading config");
295			reconfigure(conffile, &conf, &mrt_l, &peer_l, rules_l);
296		}
297
298		if (sigchld) {
299			sigchld = 0;
300			if (check_child(io_pid, "session engine"))
301				quit = 1;
302			if (check_child(rde_pid, "route decision engine"))
303				quit = 1;
304		}
305
306		if (mrtdump == 1) {
307			mrtdump = 0;
308			mrt_handler(&mrt_l);
309		}
310	}
311
312	signal(SIGCHLD, SIG_IGN);
313
314	if (io_pid)
315		kill(io_pid, SIGTERM);
316
317	if (rde_pid)
318		kill(rde_pid, SIGTERM);
319
320	while ((p = peer_l) != NULL) {
321		peer_l = p->next;
322		free(p);
323	}
324	while ((m = LIST_FIRST(&mrt_l)) != NULL) {
325		LIST_REMOVE(m, entry);
326		free(m);
327	}
328
329	free(rules_l);
330	control_cleanup();
331	kr_shutdown();
332	pftable_clear_all();
333	free(conf.listen_addrs);
334
335	do {
336		if ((pid = wait(NULL)) == -1 &&
337		    errno != EINTR && errno != ECHILD)
338			fatal("wait");
339	} while (pid != -1 || (pid == -1 && errno == EINTR));
340
341	log_info("Terminating");
342	return (0);
343}
344
345int
346check_child(pid_t pid, const char *pname)
347{
348	int	status;
349
350	if (waitpid(pid, &status, WNOHANG) > 0) {
351		if (WIFEXITED(status)) {
352			log_warnx("Lost child: %s exited", pname);
353			return (1);
354		}
355		if (WIFSIGNALED(status)) {
356			log_warnx("Lost child: %s terminated; signal %d",
357			    pname, WTERMSIG(status));
358			return (1);
359		}
360	}
361
362	return (0);
363}
364
365int
366reconfigure(char *conffile, struct bgpd_config *conf, struct mrt_head *mrt_l,
367    struct peer **peer_l, struct filter_head *rules_l)
368{
369	struct network_head	 net_l;
370	struct network		*n;
371	struct peer		*p;
372	struct filter_rule	*r;
373	struct listen_addr	*la;
374
375	if (parse_config(conffile, conf, mrt_l, peer_l, &net_l, rules_l)) {
376		log_warnx("config file %s has errors, not reloading",
377		    conffile);
378		return (-1);
379	}
380
381	prepare_listeners(conf);
382
383	if (imsg_compose(&ibuf_se, IMSG_RECONF_CONF, 0,
384	    conf, sizeof(struct bgpd_config)) == -1)
385		return (-1);
386	if (imsg_compose(&ibuf_rde, IMSG_RECONF_CONF, 0,
387	    conf, sizeof(struct bgpd_config)) == -1)
388		return (-1);
389	for (p = *peer_l; p != NULL; p = p->next)
390		if (imsg_compose(&ibuf_se, IMSG_RECONF_PEER, p->conf.id,
391		    &p->conf, sizeof(struct peer_config)) == -1)
392			return (-1);
393	while ((n = TAILQ_FIRST(&net_l)) != NULL) {
394		if (imsg_compose(&ibuf_rde, IMSG_NETWORK_ADD, 0,
395		    &n->net, sizeof(struct network_config)) == -1)
396			return (-1);
397		TAILQ_REMOVE(&net_l, n, entry);
398		free(n);
399	}
400	while ((r = TAILQ_FIRST(rules_l)) != NULL) {
401		if (imsg_compose(&ibuf_rde, IMSG_RECONF_FILTER, 0,
402		    r, sizeof(struct filter_rule)) == -1)
403			return (-1);
404		TAILQ_REMOVE(rules_l, r, entry);
405		free(r);
406	}
407	while ((la = TAILQ_FIRST(conf->listen_addrs)) != NULL) {
408		if (imsg_compose_fdpass(&ibuf_se, IMSG_RECONF_LISTENER, la->fd,
409		    la, sizeof(struct listen_addr)) == -1)
410			return (-1);
411		TAILQ_REMOVE(conf->listen_addrs, la, entry);
412		free(la);
413	}
414	free(conf->listen_addrs);
415	conf->listen_addrs = NULL;
416
417	if (imsg_compose(&ibuf_se, IMSG_RECONF_DONE, 0, NULL, 0) == -1 ||
418	    imsg_compose(&ibuf_rde, IMSG_RECONF_DONE, 0, NULL, 0) == -1)
419		return (-1);
420
421	/* mrt changes can be sent out of bound */
422	mrt_reconfigure(mrt_l);
423	return (0);
424}
425
426int
427dispatch_imsg(struct imsgbuf *ibuf, int idx)
428{
429	struct imsg		 imsg;
430	int			 n;
431
432	if ((n = imsg_read(ibuf)) == -1)
433		return (-1);
434
435	if (n == 0) {	/* connection closed */
436		log_warnx("dispatch_imsg in main: pipe closed");
437		return (-1);
438	}
439
440	for (;;) {
441		if ((n = imsg_get(ibuf, &imsg)) == -1)
442			return (-1);
443
444		if (n == 0)
445			break;
446
447		switch (imsg.hdr.type) {
448		case IMSG_KROUTE_CHANGE:
449			if (idx != PFD_PIPE_ROUTE)
450				log_warnx("route request not from RDE");
451			else if (kr_change(imsg.data))
452				return (-1);
453			break;
454		case IMSG_KROUTE_DELETE:
455			if (idx != PFD_PIPE_ROUTE)
456				log_warnx("route request not from RDE");
457			else if (kr_delete(imsg.data))
458				return (-1);
459			break;
460		case IMSG_NEXTHOP_ADD:
461			if (idx != PFD_PIPE_ROUTE)
462				log_warnx("nexthop request not from RDE");
463			else
464				if (imsg.hdr.len != IMSG_HEADER_SIZE +
465				    sizeof(struct bgpd_addr))
466					log_warnx("wrong imsg len");
467				else if (kr_nexthop_add(imsg.data) == -1)
468					return (-1);
469			break;
470		case IMSG_NEXTHOP_REMOVE:
471			if (idx != PFD_PIPE_ROUTE)
472				log_warnx("nexthop request not from RDE");
473			else
474				if (imsg.hdr.len != IMSG_HEADER_SIZE +
475				    sizeof(struct bgpd_addr))
476					log_warnx("wrong imsg len");
477				else
478					kr_nexthop_delete(imsg.data);
479			break;
480		case IMSG_PFTABLE_ADD:
481			if (idx != PFD_PIPE_ROUTE)
482				log_warnx("pftable request not from RDE");
483			else
484				if (imsg.hdr.len != IMSG_HEADER_SIZE +
485				    sizeof(struct pftable_msg))
486					log_warnx("wrong imsg len");
487				else if (pftable_addr_add(imsg.data) != 0)
488					return (-1);
489			break;
490		case IMSG_PFTABLE_REMOVE:
491			if (idx != PFD_PIPE_ROUTE)
492				log_warnx("pftable request not from RDE");
493			else
494				if (imsg.hdr.len != IMSG_HEADER_SIZE +
495				    sizeof(struct pftable_msg))
496					log_warnx("wrong imsg len");
497				else if (pftable_addr_remove(imsg.data) != 0)
498					return (-1);
499			break;
500		case IMSG_PFTABLE_COMMIT:
501			if (idx != PFD_PIPE_ROUTE)
502				log_warnx("pftable request not from RDE");
503			else
504				if (imsg.hdr.len != IMSG_HEADER_SIZE)
505					log_warnx("wrong imsg len");
506				else if (pftable_commit() != 0)
507					return (-1);
508			break;
509		case IMSG_CTL_RELOAD:
510			if (idx != PFD_PIPE_SESSION)
511				log_warnx("reload request not from SE");
512			else
513				reconfig = 1;
514			break;
515		case IMSG_CTL_FIB_COUPLE:
516			if (idx != PFD_PIPE_SESSION)
517				log_warnx("couple request not from SE");
518			else
519				kr_fib_couple();
520			break;
521		case IMSG_CTL_FIB_DECOUPLE:
522			if (idx != PFD_PIPE_SESSION)
523				log_warnx("decouple request not from SE");
524			else
525				kr_fib_decouple();
526			break;
527		case IMSG_CTL_KROUTE:
528		case IMSG_CTL_KROUTE_ADDR:
529		case IMSG_CTL_SHOW_NEXTHOP:
530		case IMSG_CTL_SHOW_INTERFACE:
531			if (idx != PFD_PIPE_SESSION)
532				log_warnx("kroute request not from SE");
533			else
534				kr_show_route(&imsg);
535			break;
536		default:
537			break;
538		}
539		imsg_free(&imsg);
540	}
541	return (0);
542}
543
544void
545send_nexthop_update(struct kroute_nexthop *msg)
546{
547	char	*gw = NULL;
548
549	if (msg->gateway.af)
550		if (asprintf(&gw, ": via %s",
551		    log_addr(&msg->gateway)) == -1) {
552			log_warn("send_nexthop_update");
553			quit = 1;
554		}
555
556	log_info("nexthop %s now %s%s%s", log_addr(&msg->nexthop),
557	    msg->valid ? "valid" : "invalid",
558	    msg->connected ? ": directly connected" : "",
559	    msg->gateway.af ? gw : "");
560
561	free(gw);
562
563	if (imsg_compose(&ibuf_rde, IMSG_NEXTHOP_UPDATE, 0,
564	    msg, sizeof(struct kroute_nexthop)) == -1)
565		quit = 1;
566}
567
568void
569send_imsg_session(int type, pid_t pid, void *data, u_int16_t datalen)
570{
571	imsg_compose_pid(&ibuf_se, type, pid, data, datalen);
572}
573