bgpd.c revision 1.71 
1/*	$OpenBSD: bgpd.c,v 1.71 2004/01/22 20:34:55 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 *);
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, *p, *next;
97	struct mrt_head		 mrt_l;
98	struct network_head	 net_l;
99	struct network		*net;
100	struct mrt		*(mrt[POLL_MAX]);
101	struct pollfd		 pfd[POLL_MAX];
102	pid_t			 io_pid = 0, rde_pid = 0, pid;
103	char			*conffile;
104	int			 debug = 0;
105	int			 ch, csock, i, j, n, nfds, timeout;
106	int			 pipe_m2s[2];
107	int			 pipe_m2r[2];
108	int			 pipe_s2r[2];
109
110	conffile = CONFFILE;
111	bgpd_process = PROC_MAIN;
112
113	log_init(1);		/* log to stderr until daemonized */
114
115	bzero(&conf, sizeof(conf));
116	LIST_INIT(&mrt_l);
117	TAILQ_INIT(&net_l);
118	peer_l = NULL;
119
120	while ((ch = getopt(argc, argv, "dD:f:nv")) != -1) {
121		switch (ch) {
122		case 'd':
123			debug = 1;
124			break;
125		case 'D':
126			if (cmdline_symset(optarg) < 0)
127				log_warnx("could not parse macro definition %s",
128				    optarg);
129			break;
130		case 'f':
131			conffile = optarg;
132			break;
133		case 'n':
134			conf.opts |= BGPD_OPT_NOACTION;
135			break;
136		case 'v':
137			if (conf.opts & BGPD_OPT_VERBOSE)
138				conf.opts |= BGPD_OPT_VERBOSE2;
139			conf.opts |= BGPD_OPT_VERBOSE;
140			break;
141		default:
142			usage();
143			/* NOTREACHED */
144		}
145	}
146
147	if (parse_config(conffile, &conf, &mrt_l, &peer_l, &net_l))
148		exit(1);
149
150	if (conf.opts & BGPD_OPT_NOACTION) {
151		fprintf(stderr, "configuration OK\n");
152		exit(0);
153	}
154
155	if (geteuid())
156		errx(1, "need root privileges");
157
158	if (getpwnam(BGPD_USER) == NULL)
159		errx(1, "unknown user %s", BGPD_USER);
160	endpwent();
161
162	log_init(debug);
163
164	if (!debug)
165		daemon(1, 0);
166
167	log_info("startup");
168
169	if (pipe(pipe_m2s) == -1)
170		fatal("pipe");
171	if (fcntl(pipe_m2s[0], F_SETFL, O_NONBLOCK) == -1 ||
172	    fcntl(pipe_m2s[1], F_SETFL, O_NONBLOCK) == -1)
173		fatal("fcntl");
174	if (pipe(pipe_m2r) == -1)
175		fatal("pipe");
176	if (fcntl(pipe_m2r[0], F_SETFL, O_NONBLOCK) == -1 ||
177	    fcntl(pipe_m2r[1], F_SETFL, O_NONBLOCK) == -1)
178		fatal("fcntl");
179	if (pipe(pipe_s2r) == -1)
180		fatal("pipe");
181	if (fcntl(pipe_s2r[0], F_SETFL, O_NONBLOCK) == -1 ||
182	    fcntl(pipe_s2r[1], F_SETFL, O_NONBLOCK) == -1)
183		fatal("fcntl");
184
185	if ((csock = control_init()) == -1)
186		fatalx("control socket setup failed");
187
188	/* fork children */
189	rde_pid = rde_main(&conf, peer_l, &net_l, pipe_m2r, pipe_s2r);
190	io_pid = session_main(&conf, peer_l, pipe_m2s, pipe_s2r);
191
192	setproctitle("parent");
193
194	signal(SIGTERM, sighdlr);
195	signal(SIGINT, sighdlr);
196	signal(SIGCHLD, sighdlr);
197	signal(SIGHUP, sighdlr);
198	signal(SIGALRM, sighdlr);
199	signal(SIGUSR1, sighdlr);
200
201	close(pipe_m2s[1]);
202	close(pipe_m2r[1]);
203	close(pipe_s2r[0]);
204	close(pipe_s2r[1]);
205	close(csock);
206
207	imsg_init(&ibuf_se, pipe_m2s[0]);
208	imsg_init(&ibuf_rde, pipe_m2r[0]);
209	mrt_init(&ibuf_rde, &ibuf_se);
210	if ((rfd = kr_init(!(conf.flags & BGPD_FLAG_NO_FIB_UPDATE))) == -1)
211		quit = 1;
212
213	for (p = peer_l; p != NULL; p = next) {
214		next = p->next;
215		free(p);
216	}
217	for (net = TAILQ_FIRST(&net_l); net != TAILQ_END(&net_l);
218	    net = TAILQ_FIRST(&net_l)) {
219		TAILQ_REMOVE(&net_l, net, network_l);
220		free(net);
221	}
222
223	while (quit == 0) {
224		pfd[PFD_PIPE_SESSION].fd = ibuf_se.sock;
225		pfd[PFD_PIPE_SESSION].events = POLLIN;
226		if (ibuf_se.w.queued)
227			pfd[PFD_PIPE_SESSION].events |= POLLOUT;
228		pfd[PFD_PIPE_ROUTE].fd = ibuf_rde.sock;
229		pfd[PFD_PIPE_ROUTE].events = POLLIN;
230		if (ibuf_rde.w.queued)
231			pfd[PFD_PIPE_ROUTE].events |= POLLOUT;
232		pfd[PFD_SOCK_ROUTE].fd = rfd;
233		pfd[PFD_SOCK_ROUTE].events = POLLIN;
234		i = PFD_MRT_START;
235		i = mrt_select(&mrt_l, pfd, mrt, i, POLL_MAX, &timeout);
236
237		if ((nfds = poll(pfd, i, INFTIM)) == -1)
238			if (errno != EINTR) {
239				log_warn("poll error");
240				quit = 1;
241			}
242
243		if (nfds > 0 && (pfd[PFD_PIPE_SESSION].revents & POLLOUT))
244			if ((n = msgbuf_write(&ibuf_se.w)) < 0) {
245				log_warn("pipe write error (to SE)");
246				quit = 1;
247			}
248
249		if (nfds > 0 && (pfd[PFD_PIPE_ROUTE].revents & POLLOUT))
250			if ((n = msgbuf_write(&ibuf_rde.w)) < 0) {
251				log_warn("pipe write error (to RDE)");
252				quit = 1;
253			}
254
255		if (nfds > 0 && pfd[PFD_PIPE_SESSION].revents & POLLIN) {
256			nfds--;
257			if (dispatch_imsg(&ibuf_se, PFD_PIPE_SESSION,
258			    &mrt_l) == -1)
259				quit = 1;
260		}
261
262		if (nfds > 0 && pfd[PFD_PIPE_ROUTE].revents & POLLIN) {
263			nfds--;
264			if (dispatch_imsg(&ibuf_rde, PFD_PIPE_ROUTE,
265			    &mrt_l) == -1)
266				quit = 1;
267		}
268
269		if (nfds > 0 && pfd[PFD_SOCK_ROUTE].revents & POLLIN) {
270			nfds--;
271			if (kr_dispatch_msg() == -1)
272				quit = 1;
273		}
274
275		for (j = PFD_MRT_START; j < i && nfds > 0 ; j++) {
276			if (pfd[j].revents & POLLOUT) {
277				if ((n = mrt_write(mrt[j])) < 0) {
278					log_warn("mrt write error");
279				}
280			}
281		}
282
283		if (reconfig) {
284			log_info("rereading config");
285			reconfigure(conffile, &conf, &mrt_l, peer_l);
286			reconfig = 0;
287		}
288
289		if (sigchld) {
290			if (check_child(io_pid, "session engine"))
291				quit = 1;
292			if (check_child(rde_pid, "route decision engine"))
293				quit = 1;
294			sigchld = 0;
295		}
296
297		if (mrtdump == 1) {
298			mrt_handler(&mrt_l);
299			mrtdump = 0;
300		}
301	}
302
303	signal(SIGCHLD, SIG_IGN);
304
305	if (io_pid)
306		kill(io_pid, SIGTERM);
307
308	if (rde_pid)
309		kill(rde_pid, SIGTERM);
310
311	do {
312		pid = waitpid(-1, NULL, WNOHANG);
313	} while (pid > 0 || (pid == -1 && errno == EINTR));
314
315	control_cleanup();
316	kr_shutdown();
317
318	log_info("Terminating");
319	return (0);
320}
321
322int
323check_child(pid_t pid, const char *pname)
324{
325	int	status;
326
327	if (waitpid(pid, &status, WNOHANG) > 0) {
328		if (WIFEXITED(status)) {
329			log_warnx("Lost child: %s exited", pname);
330			return (1);
331		}
332		if (WIFSIGNALED(status)) {
333			log_warnx("Lost child: %s terminated; signal %d",
334			    pname, WTERMSIG(status));
335			return (1);
336		}
337	}
338
339	return (0);
340}
341
342int
343reconfigure(char *conffile, struct bgpd_config *conf, struct mrt_head *mrt_l,
344    struct peer *peer_l)
345{
346	struct network_head	 net_l;
347	struct network		*n;
348	struct peer		*p, *next;
349
350	if (parse_config(conffile, conf, mrt_l, &peer_l, &net_l)) {
351		log_warnx("config file %s has errors, not reloading",
352		    conffile);
353		return (-1);
354	}
355
356	if (imsg_compose(&ibuf_se, IMSG_RECONF_CONF, 0,
357	    conf, sizeof(struct bgpd_config)) == -1)
358		return (-1);
359	if (imsg_compose(&ibuf_rde, IMSG_RECONF_CONF, 0,
360	    conf, sizeof(struct bgpd_config)) == -1)
361		return (-1);
362	for (p = peer_l; p != NULL; p = next) {
363		next = p->next;
364		if (imsg_compose(&ibuf_se, IMSG_RECONF_PEER, p->conf.id,
365		    &p->conf, sizeof(struct peer_config)) == -1)
366			return (-1);
367		if (imsg_compose(&ibuf_rde, IMSG_RECONF_PEER, p->conf.id,
368		    &p->conf, sizeof(struct peer_config)) == -1)
369			return (-1);
370		free(p);
371	}
372	for (n = TAILQ_FIRST(&net_l); n != TAILQ_END(&net_l);
373	    n = TAILQ_FIRST(&net_l)) {
374		if (imsg_compose(&ibuf_rde, IMSG_RECONF_NETWORK, 0,
375		    &n->net, sizeof(struct network_config)) == -1)
376			return (-1);
377		TAILQ_REMOVE(&net_l, n, network_l);
378		free(n);
379	}
380	if (imsg_compose(&ibuf_se, IMSG_RECONF_DONE, 0, NULL, 0) == -1 ||
381	    imsg_compose(&ibuf_rde, IMSG_RECONF_DONE, 0, NULL, 0) == -1)
382		return (-1);
383
384	return (0);
385}
386
387int
388dispatch_imsg(struct imsgbuf *ibuf, int idx, struct mrt_head *mrt_l)
389{
390	struct imsg		 imsg;
391	int			 n;
392
393	if ((n = imsg_read(ibuf)) == -1)
394		return (-1);
395
396	if (n == 0) {	/* connection closed */
397		log_warnx("dispatch_imsg in main: pipe closed");
398		return (-1);
399	}
400
401	for (;;) {
402		if ((n = imsg_get(ibuf, &imsg)) == -1)
403			return (-1);
404
405		if (n == 0)
406			break;
407
408		switch (imsg.hdr.type) {
409		case IMSG_MRT_MSG:
410		case IMSG_MRT_END:
411			if (mrt_queue(mrt_l, &imsg) == -1)
412				log_warnx("mrt_queue failed.");
413			break;
414		case IMSG_KROUTE_CHANGE:
415			if (idx != PFD_PIPE_ROUTE)
416				log_warnx("route request not from RDE");
417			else if (kr_change(imsg.data))
418				return (-1);
419			break;
420		case IMSG_KROUTE_DELETE:
421			if (idx != PFD_PIPE_ROUTE)
422				log_warnx("route request not from RDE");
423			else if (kr_delete(imsg.data))
424				return (-1);
425			break;
426		case IMSG_NEXTHOP_ADD:
427			if (idx != PFD_PIPE_ROUTE)
428				log_warnx("nexthop request not from RDE");
429			else
430				if (imsg.hdr.len != IMSG_HEADER_SIZE +
431				    sizeof(struct bgpd_addr))
432					log_warnx("wrong imsg len");
433				else if (kr_nexthop_add(imsg.data) == -1)
434					return (-1);
435			break;
436		case IMSG_NEXTHOP_REMOVE:
437			if (idx != PFD_PIPE_ROUTE)
438				log_warnx("nexthop request not from RDE");
439			else
440				if (imsg.hdr.len != IMSG_HEADER_SIZE +
441				    sizeof(struct bgpd_addr))
442					log_warnx("wrong imsg len");
443				else kr_nexthop_delete(imsg.data);
444			break;
445		case IMSG_CTL_RELOAD:
446			if (idx != PFD_PIPE_SESSION)
447				log_warnx("reload request not from SE");
448			else
449				reconfig = 1;
450			break;
451		case IMSG_CTL_FIB_COUPLE:
452			if (idx != PFD_PIPE_SESSION)
453				log_warnx("couple request not from SE");
454			else
455				kr_fib_couple();
456			break;
457		case IMSG_CTL_FIB_DECOUPLE:
458			if (idx != PFD_PIPE_SESSION)
459				log_warnx("decouple request not from SE");
460			else
461				kr_fib_decouple();
462			break;
463		case IMSG_CTL_KROUTE:
464		case IMSG_CTL_KROUTE_ADDR:
465		case IMSG_CTL_SHOW_NEXTHOP:
466		case IMSG_CTL_SHOW_INTERFACE:
467			if (idx != PFD_PIPE_SESSION)
468				log_warnx("kroute request not from SE");
469			else
470				kr_show_route(&imsg);
471			break;
472		default:
473			break;
474		}
475		imsg_free(&imsg);
476	}
477	return (0);
478}
479
480void
481send_nexthop_update(struct kroute_nexthop *msg)
482{
483	char	*gw = NULL, *nh = NULL;
484
485	if (msg->gateway.af == AF_INET)
486		if (asprintf(&gw, ": via %s",
487		    log_ntoa(msg->gateway.v4.s_addr)) == -1) {
488			log_warn("send_nexthop_update");
489			quit = 1;
490		}
491
492	if (msg->nexthop.af == AF_INET)
493		nh = log_ntoa(msg->nexthop.v4.s_addr);
494
495	log_info("nexthop %s now %s%s%s", nh,
496	    msg->valid ? "valid" : "invalid",
497	    msg->connected ? ": directly connected" : "",
498	    msg->gateway.af ? gw : "");
499
500	free(gw);
501
502	if (imsg_compose(&ibuf_rde, IMSG_NEXTHOP_UPDATE, 0,
503	    msg, sizeof(struct kroute_nexthop)) == -1)
504		quit = 1;
505}
506
507void
508send_imsg_session(int type, pid_t pid, void *data, u_int16_t datalen)
509{
510	imsg_compose_pid(&ibuf_se, type, pid, data, datalen);
511}
512
513
514