bgpd.c revision 1.78 
1/*	$OpenBSD: bgpd.c,v 1.78 2004/02/09 23:10:04 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 POLL_MAX		8
87#define PFD_PIPE_SESSION	0
88#define PFD_PIPE_ROUTE		1
89#define PFD_SOCK_ROUTE		2
90#define PFD_MRT_START		3
91
92int
93main(int argc, char *argv[])
94{
95	struct bgpd_config	 conf;
96	struct peer		*peer_l;
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		*(mrt[POLL_MAX]);
103	struct pollfd		 pfd[POLL_MAX];
104	pid_t			 io_pid = 0, rde_pid = 0, pid;
105	char			*conffile;
106	int			 debug = 0;
107	int			 ch, csock, i, j, n, nfds, timeout;
108	int			 pipe_m2s[2];
109	int			 pipe_m2r[2];
110	int			 pipe_s2r[2];
111
112	conffile = CONFFILE;
113	bgpd_process = PROC_MAIN;
114
115	log_init(1);		/* log to stderr until daemonized */
116
117	if ((rules_l = calloc(1, sizeof(struct filter_head))) == NULL)
118		err(1, NULL);
119
120	bzero(&conf, sizeof(conf));
121	LIST_INIT(&mrt_l);
122	TAILQ_INIT(&net_l);
123	TAILQ_INIT(rules_l);
124	peer_l = NULL;
125
126	while ((ch = getopt(argc, argv, "dD:f:nv")) != -1) {
127		switch (ch) {
128		case 'd':
129			debug = 1;
130			break;
131		case 'D':
132			if (cmdline_symset(optarg) < 0)
133				log_warnx("could not parse macro definition %s",
134				    optarg);
135			break;
136		case 'f':
137			conffile = optarg;
138			break;
139		case 'n':
140			conf.opts |= BGPD_OPT_NOACTION;
141			break;
142		case 'v':
143			if (conf.opts & BGPD_OPT_VERBOSE)
144				conf.opts |= BGPD_OPT_VERBOSE2;
145			conf.opts |= BGPD_OPT_VERBOSE;
146			break;
147		default:
148			usage();
149			/* NOTREACHED */
150		}
151	}
152
153	if (parse_config(conffile, &conf, &mrt_l, &peer_l, &net_l, rules_l))
154		exit(1);
155
156	if (conf.opts & BGPD_OPT_NOACTION) {
157		if (conf.opts & BGPD_OPT_VERBOSE)
158			print_config(&conf, peer_l, rules_l);
159		else
160			fprintf(stderr, "configuration OK\n");
161		exit(0);
162	}
163
164	if (geteuid())
165		errx(1, "need root privileges");
166
167	if (getpwnam(BGPD_USER) == NULL)
168		errx(1, "unknown user %s", BGPD_USER);
169	endpwent();
170
171	log_init(debug);
172
173	if (!debug)
174		daemon(1, 0);
175
176	log_info("startup");
177
178	if (pipe(pipe_m2s) == -1)
179		fatal("pipe");
180	if (fcntl(pipe_m2s[0], F_SETFL, O_NONBLOCK) == -1 ||
181	    fcntl(pipe_m2s[1], F_SETFL, O_NONBLOCK) == -1)
182		fatal("fcntl");
183	if (pipe(pipe_m2r) == -1)
184		fatal("pipe");
185	if (fcntl(pipe_m2r[0], F_SETFL, O_NONBLOCK) == -1 ||
186	    fcntl(pipe_m2r[1], F_SETFL, O_NONBLOCK) == -1)
187		fatal("fcntl");
188	if (pipe(pipe_s2r) == -1)
189		fatal("pipe");
190	if (fcntl(pipe_s2r[0], F_SETFL, O_NONBLOCK) == -1 ||
191	    fcntl(pipe_s2r[1], F_SETFL, O_NONBLOCK) == -1)
192		fatal("fcntl");
193
194	if ((csock = control_init()) == -1)
195		fatalx("control socket setup failed");
196
197	/* fork children */
198	rde_pid = rde_main(&conf, peer_l, &net_l, rules_l, pipe_m2r, pipe_s2r);
199	io_pid = session_main(&conf, peer_l, pipe_m2s, pipe_s2r);
200
201	setproctitle("parent");
202
203	signal(SIGTERM, sighdlr);
204	signal(SIGINT, sighdlr);
205	signal(SIGCHLD, sighdlr);
206	signal(SIGHUP, sighdlr);
207	signal(SIGALRM, sighdlr);
208	signal(SIGUSR1, sighdlr);
209
210	close(pipe_m2s[1]);
211	close(pipe_m2r[1]);
212	close(pipe_s2r[0]);
213	close(pipe_s2r[1]);
214	close(csock);
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
222	while ((net = TAILQ_FIRST(&net_l)) != NULL) {
223		TAILQ_REMOVE(&net_l, net, network_l);
224		free(net);
225	}
226
227	while ((r = TAILQ_FIRST(rules_l)) != NULL) {
228		TAILQ_REMOVE(rules_l, r, entries);
229		free(r);
230	}
231
232	while (quit == 0) {
233		pfd[PFD_PIPE_SESSION].fd = ibuf_se.sock;
234		pfd[PFD_PIPE_SESSION].events = POLLIN;
235		if (ibuf_se.w.queued)
236			pfd[PFD_PIPE_SESSION].events |= POLLOUT;
237		pfd[PFD_PIPE_ROUTE].fd = ibuf_rde.sock;
238		pfd[PFD_PIPE_ROUTE].events = POLLIN;
239		if (ibuf_rde.w.queued)
240			pfd[PFD_PIPE_ROUTE].events |= POLLOUT;
241		pfd[PFD_SOCK_ROUTE].fd = rfd;
242		pfd[PFD_SOCK_ROUTE].events = POLLIN;
243		i = PFD_MRT_START;
244		i = mrt_select(&mrt_l, pfd, mrt, i, POLL_MAX, &timeout);
245
246		if ((nfds = poll(pfd, i, INFTIM)) == -1)
247			if (errno != EINTR) {
248				log_warn("poll error");
249				quit = 1;
250			}
251
252		if (nfds > 0 && (pfd[PFD_PIPE_SESSION].revents & POLLOUT))
253			if ((n = msgbuf_write(&ibuf_se.w)) < 0) {
254				log_warn("pipe write error (to SE)");
255				quit = 1;
256			}
257
258		if (nfds > 0 && (pfd[PFD_PIPE_ROUTE].revents & POLLOUT))
259			if ((n = msgbuf_write(&ibuf_rde.w)) < 0) {
260				log_warn("pipe write error (to RDE)");
261				quit = 1;
262			}
263
264		if (nfds > 0 && pfd[PFD_PIPE_SESSION].revents & POLLIN) {
265			nfds--;
266			if (dispatch_imsg(&ibuf_se, PFD_PIPE_SESSION,
267			    &mrt_l) == -1)
268				quit = 1;
269		}
270
271		if (nfds > 0 && pfd[PFD_PIPE_ROUTE].revents & POLLIN) {
272			nfds--;
273			if (dispatch_imsg(&ibuf_rde, PFD_PIPE_ROUTE,
274			    &mrt_l) == -1)
275				quit = 1;
276		}
277
278		if (nfds > 0 && pfd[PFD_SOCK_ROUTE].revents & POLLIN) {
279			nfds--;
280			if (kr_dispatch_msg() == -1)
281				quit = 1;
282		}
283
284		for (j = PFD_MRT_START; j < i && nfds > 0 ; j++) {
285			if (pfd[j].revents & POLLOUT) {
286				if ((n = mrt_write(mrt[j])) < 0) {
287					log_warn("mrt write error");
288				}
289			}
290		}
291
292		if (reconfig) {
293			log_info("rereading config");
294			reconfigure(conffile, &conf, &mrt_l, &peer_l, rules_l);
295			reconfig = 0;
296		}
297
298		if (sigchld) {
299			if (check_child(io_pid, "session engine"))
300				quit = 1;
301			if (check_child(rde_pid, "route decision engine"))
302				quit = 1;
303			sigchld = 0;
304		}
305
306		if (mrtdump == 1) {
307			mrt_handler(&mrt_l);
308			mrtdump = 0;
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	do {
321		pid = waitpid(-1, NULL, WNOHANG);
322	} while (pid > 0 || (pid == -1 && errno == EINTR));
323
324	free(rules_l);
325	control_cleanup();
326	kr_shutdown();
327
328	log_info("Terminating");
329	return (0);
330}
331
332int
333check_child(pid_t pid, const char *pname)
334{
335	int	status;
336
337	if (waitpid(pid, &status, WNOHANG) > 0) {
338		if (WIFEXITED(status)) {
339			log_warnx("Lost child: %s exited", pname);
340			return (1);
341		}
342		if (WIFSIGNALED(status)) {
343			log_warnx("Lost child: %s terminated; signal %d",
344			    pname, WTERMSIG(status));
345			return (1);
346		}
347	}
348
349	return (0);
350}
351
352int
353reconfigure(char *conffile, struct bgpd_config *conf, struct mrt_head *mrt_l,
354    struct peer **peer_l, struct filter_head *rules_l)
355{
356	struct network_head	 net_l;
357	struct network		*n;
358	struct peer		*p;
359	struct filter_rule	*r;
360
361	if (parse_config(conffile, conf, mrt_l, peer_l, &net_l, rules_l)) {
362		log_warnx("config file %s has errors, not reloading",
363		    conffile);
364		return (-1);
365	}
366
367	if (imsg_compose(&ibuf_se, IMSG_RECONF_CONF, 0,
368	    conf, sizeof(struct bgpd_config)) == -1)
369		return (-1);
370	if (imsg_compose(&ibuf_rde, IMSG_RECONF_CONF, 0,
371	    conf, sizeof(struct bgpd_config)) == -1)
372		return (-1);
373	for (p = *peer_l; p != NULL; p = p->next) {
374		if (imsg_compose(&ibuf_se, IMSG_RECONF_PEER, p->conf.id,
375		    &p->conf, sizeof(struct peer_config)) == -1)
376			return (-1);
377		if (imsg_compose(&ibuf_rde, IMSG_RECONF_PEER, p->conf.id,
378		    &p->conf, sizeof(struct peer_config)) == -1)
379			return (-1);
380	}
381	for (n = TAILQ_FIRST(&net_l); n != TAILQ_END(&net_l);
382	    n = TAILQ_FIRST(&net_l)) {
383		if (imsg_compose(&ibuf_rde, IMSG_RECONF_NETWORK, 0,
384		    &n->net, sizeof(struct network_config)) == -1)
385			return (-1);
386		TAILQ_REMOVE(&net_l, n, network_l);
387		free(n);
388	}
389	for (r = TAILQ_FIRST(rules_l); r != NULL; r = TAILQ_FIRST(rules_l)) {
390		if (imsg_compose(&ibuf_rde, IMSG_RECONF_FILTER, 0,
391		    r, sizeof(struct filter_rule)) == -1)
392			return (-1);
393		TAILQ_REMOVE(rules_l, r, entries);
394		free(r);
395	}
396	if (imsg_compose(&ibuf_se, IMSG_RECONF_DONE, 0, NULL, 0) == -1 ||
397	    imsg_compose(&ibuf_rde, IMSG_RECONF_DONE, 0, NULL, 0) == -1)
398		return (-1);
399
400	return (0);
401}
402
403int
404dispatch_imsg(struct imsgbuf *ibuf, int idx, struct mrt_head *mrt_l)
405{
406	struct imsg		 imsg;
407	int			 n;
408
409	if ((n = imsg_read(ibuf)) == -1)
410		return (-1);
411
412	if (n == 0) {	/* connection closed */
413		log_warnx("dispatch_imsg in main: pipe closed");
414		return (-1);
415	}
416
417	for (;;) {
418		if ((n = imsg_get(ibuf, &imsg)) == -1)
419			return (-1);
420
421		if (n == 0)
422			break;
423
424		switch (imsg.hdr.type) {
425		case IMSG_MRT_MSG:
426		case IMSG_MRT_END:
427			if (mrt_queue(mrt_l, &imsg) == -1)
428				log_warnx("mrt_queue failed.");
429			break;
430		case IMSG_KROUTE_CHANGE:
431			if (idx != PFD_PIPE_ROUTE)
432				log_warnx("route request not from RDE");
433			else if (kr_change(imsg.data))
434				return (-1);
435			break;
436		case IMSG_KROUTE_DELETE:
437			if (idx != PFD_PIPE_ROUTE)
438				log_warnx("route request not from RDE");
439			else if (kr_delete(imsg.data))
440				return (-1);
441			break;
442		case IMSG_NEXTHOP_ADD:
443			if (idx != PFD_PIPE_ROUTE)
444				log_warnx("nexthop request not from RDE");
445			else
446				if (imsg.hdr.len != IMSG_HEADER_SIZE +
447				    sizeof(struct bgpd_addr))
448					log_warnx("wrong imsg len");
449				else if (kr_nexthop_add(imsg.data) == -1)
450					return (-1);
451			break;
452		case IMSG_NEXTHOP_REMOVE:
453			if (idx != PFD_PIPE_ROUTE)
454				log_warnx("nexthop request not from RDE");
455			else
456				if (imsg.hdr.len != IMSG_HEADER_SIZE +
457				    sizeof(struct bgpd_addr))
458					log_warnx("wrong imsg len");
459				else kr_nexthop_delete(imsg.data);
460			break;
461		case IMSG_CTL_RELOAD:
462			if (idx != PFD_PIPE_SESSION)
463				log_warnx("reload request not from SE");
464			else
465				reconfig = 1;
466			break;
467		case IMSG_CTL_FIB_COUPLE:
468			if (idx != PFD_PIPE_SESSION)
469				log_warnx("couple request not from SE");
470			else
471				kr_fib_couple();
472			break;
473		case IMSG_CTL_FIB_DECOUPLE:
474			if (idx != PFD_PIPE_SESSION)
475				log_warnx("decouple request not from SE");
476			else
477				kr_fib_decouple();
478			break;
479		case IMSG_CTL_KROUTE:
480		case IMSG_CTL_KROUTE_ADDR:
481		case IMSG_CTL_SHOW_NEXTHOP:
482		case IMSG_CTL_SHOW_INTERFACE:
483			if (idx != PFD_PIPE_SESSION)
484				log_warnx("kroute request not from SE");
485			else
486				kr_show_route(&imsg);
487			break;
488		default:
489			break;
490		}
491		imsg_free(&imsg);
492	}
493	return (0);
494}
495
496void
497send_nexthop_update(struct kroute_nexthop *msg)
498{
499	char	*gw = NULL;
500
501	if (msg->gateway.af)
502		if (asprintf(&gw, ": via %s",
503		    log_addr(&msg->gateway)) == -1) {
504			log_warn("send_nexthop_update");
505			quit = 1;
506		}
507
508	log_info("nexthop %s now %s%s%s", log_addr(&msg->nexthop),
509	    msg->valid ? "valid" : "invalid",
510	    msg->connected ? ": directly connected" : "",
511	    msg->gateway.af ? gw : "");
512
513	free(gw);
514
515	if (imsg_compose(&ibuf_rde, IMSG_NEXTHOP_UPDATE, 0,
516	    msg, sizeof(struct kroute_nexthop)) == -1)
517		quit = 1;
518}
519
520void
521send_imsg_session(int type, pid_t pid, void *data, u_int16_t datalen)
522{
523	imsg_compose_pid(&ibuf_se, type, pid, data, datalen);
524}
525
526
527