bgpd.c revision 1.197 
1/*	$OpenBSD: bgpd.c,v 1.197 2018/09/07 10:59:16 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 <syslog.h>
34#include <unistd.h>
35
36#include "bgpd.h"
37#include "mrt.h"
38#include "session.h"
39#include "log.h"
40
41void		sighdlr(int);
42__dead void	usage(void);
43int		main(int, char *[]);
44pid_t		start_child(enum bgpd_process, char *, int, int, int);
45int		send_filterset(struct imsgbuf *, struct filter_set_head *);
46int		reconfigure(char *, struct bgpd_config *, struct peer **);
47int		dispatch_imsg(struct imsgbuf *, int, struct bgpd_config *);
48int		control_setup(struct bgpd_config *);
49int		imsg_send_sockets(struct imsgbuf *, struct imsgbuf *);
50
51int			 cflags;
52volatile sig_atomic_t	 mrtdump;
53volatile sig_atomic_t	 quit;
54volatile sig_atomic_t	 reconfig;
55pid_t			 reconfpid;
56int			 reconfpending;
57struct imsgbuf		*ibuf_se;
58struct imsgbuf		*ibuf_rde;
59struct rib_names	 ribnames = SIMPLEQ_HEAD_INITIALIZER(ribnames);
60char			*cname;
61char			*rcname;
62
63void
64sighdlr(int sig)
65{
66	switch (sig) {
67	case SIGTERM:
68	case SIGINT:
69		quit = 1;
70		break;
71	case SIGHUP:
72		reconfig = 1;
73		break;
74	case SIGALRM:
75	case SIGUSR1:
76		mrtdump = 1;
77		break;
78	}
79}
80
81__dead void
82usage(void)
83{
84	extern char *__progname;
85
86	fprintf(stderr, "usage: %s [-cdnv] [-D macro=value] [-f file]\n",
87	    __progname);
88	exit(1);
89}
90
91#define PFD_PIPE_SESSION	0
92#define PFD_PIPE_ROUTE		1
93#define PFD_SOCK_ROUTE		2
94#define POLL_MAX		3
95#define MAX_TIMEOUT		3600
96
97int	 cmd_opts;
98
99int
100main(int argc, char *argv[])
101{
102	struct bgpd_config	*conf;
103	struct peer		*peer_l, *p;
104	struct pollfd		 pfd[POLL_MAX];
105	pid_t			 io_pid = 0, rde_pid = 0, pid;
106	char			*conffile;
107	char			*saved_argv0;
108	int			 debug = 0;
109	int			 rflag = 0, sflag = 0;
110	int			 rfd = -1;
111	int			 ch, timeout, status;
112	int			 pipe_m2s[2];
113	int			 pipe_m2r[2];
114
115	conffile = CONFFILE;
116	bgpd_process = PROC_MAIN;
117
118	log_init(1, LOG_DAEMON);	/* log to stderr until daemonized */
119	log_procinit(log_procnames[bgpd_process]);
120	log_setverbose(1);
121
122	saved_argv0 = argv[0];
123	if (saved_argv0 == NULL)
124		saved_argv0 = "bgpd";
125
126	conf = new_config();
127	peer_l = NULL;
128
129	while ((ch = getopt(argc, argv, "cdD:f:nRSv")) != -1) {
130		switch (ch) {
131		case 'c':
132			cmd_opts |= BGPD_OPT_FORCE_DEMOTE;
133			break;
134		case 'd':
135			debug = 1;
136			break;
137		case 'D':
138			if (cmdline_symset(optarg) < 0)
139				log_warnx("could not parse macro definition %s",
140				    optarg);
141			break;
142		case 'f':
143			conffile = optarg;
144			break;
145		case 'n':
146			cmd_opts |= BGPD_OPT_NOACTION;
147			break;
148		case 'v':
149			if (cmd_opts & BGPD_OPT_VERBOSE)
150				cmd_opts |= BGPD_OPT_VERBOSE2;
151			cmd_opts |= BGPD_OPT_VERBOSE;
152			break;
153		case 'R':
154			rflag = 1;
155			break;
156		case 'S':
157			sflag = 1;
158			break;
159		default:
160			usage();
161			/* NOTREACHED */
162		}
163	}
164
165	argc -= optind;
166	argv += optind;
167	if (argc > 0 || (sflag && rflag))
168		usage();
169
170	if (cmd_opts & BGPD_OPT_NOACTION) {
171		if (parse_config(conffile, conf, &peer_l))
172			exit(1);
173
174		if (cmd_opts & BGPD_OPT_VERBOSE)
175			print_config(conf, &ribnames, &conf->networks, peer_l,
176			    conf->filters, conf->mrt, &conf->rdomains);
177		else
178			fprintf(stderr, "configuration OK\n");
179		exit(0);
180	}
181
182	if (rflag)
183		rde_main(debug, cmd_opts & BGPD_OPT_VERBOSE);
184	else if (sflag)
185		session_main(debug, cmd_opts & BGPD_OPT_VERBOSE);
186
187	if (geteuid())
188		errx(1, "need root privileges");
189
190	if (getpwnam(BGPD_USER) == NULL)
191		errx(1, "unknown user %s", BGPD_USER);
192
193	log_init(debug, LOG_DAEMON);
194	log_setverbose(cmd_opts & BGPD_OPT_VERBOSE);
195
196	if (!debug)
197		daemon(1, 0);
198
199	log_info("startup");
200
201	if (socketpair(AF_UNIX, SOCK_STREAM | SOCK_CLOEXEC | SOCK_NONBLOCK,
202	    PF_UNSPEC, pipe_m2s) == -1)
203		fatal("socketpair");
204	if (socketpair(AF_UNIX, SOCK_STREAM | SOCK_CLOEXEC | SOCK_NONBLOCK,
205	    PF_UNSPEC, pipe_m2r) == -1)
206		fatal("socketpair");
207
208	/* fork children */
209	rde_pid = start_child(PROC_RDE, saved_argv0, pipe_m2r[1], debug,
210	    cmd_opts & BGPD_OPT_VERBOSE);
211	io_pid = start_child(PROC_SE, saved_argv0, pipe_m2s[1], debug,
212	    cmd_opts & BGPD_OPT_VERBOSE);
213
214	signal(SIGTERM, sighdlr);
215	signal(SIGINT, sighdlr);
216	signal(SIGHUP, sighdlr);
217	signal(SIGALRM, sighdlr);
218	signal(SIGUSR1, sighdlr);
219	signal(SIGPIPE, SIG_IGN);
220
221	if ((ibuf_se = malloc(sizeof(struct imsgbuf))) == NULL ||
222	    (ibuf_rde = malloc(sizeof(struct imsgbuf))) == NULL)
223		fatal(NULL);
224	imsg_init(ibuf_se, pipe_m2s[0]);
225	imsg_init(ibuf_rde, pipe_m2r[0]);
226	mrt_init(ibuf_rde, ibuf_se);
227	if ((rfd = kr_init()) == -1)
228		quit = 1;
229
230	/*
231	 * rpath, read config file
232	 * cpath, unlink control socket
233	 * fattr, chmod on control socket
234	 * wpath, needed if we are doing mrt dumps
235	 *
236	 * pledge placed here because kr_init() does a setsockopt on the
237	 * routing socket thats not allowed at all.
238	 */
239#if 0
240	/*
241	 * disabled because we do ioctls on /dev/pf and SIOCSIFGATTR
242	 * this needs some redesign of bgpd to be fixed.
243	 */
244BROKEN	if (pledge("stdio rpath wpath cpath fattr unix route recvfd sendfd",
245	    NULL) == -1)
246		fatal("pledge");
247#endif
248
249	if (imsg_send_sockets(ibuf_se, ibuf_rde))
250		fatal("could not establish imsg links");
251	quit = reconfigure(conffile, conf, &peer_l);
252	if (pftable_clear_all() != 0)
253		quit = 1;
254
255	while (quit == 0) {
256		bzero(pfd, sizeof(pfd));
257
258		set_pollfd(&pfd[PFD_PIPE_SESSION], ibuf_se);
259		set_pollfd(&pfd[PFD_PIPE_ROUTE], ibuf_rde);
260
261		pfd[PFD_SOCK_ROUTE].fd = rfd;
262		pfd[PFD_SOCK_ROUTE].events = POLLIN;
263
264		timeout = mrt_timeout(conf->mrt);
265		if (timeout > MAX_TIMEOUT)
266			timeout = MAX_TIMEOUT;
267
268		if (poll(pfd, POLL_MAX, timeout * 1000) == -1)
269			if (errno != EINTR) {
270				log_warn("poll error");
271				quit = 1;
272			}
273
274		if (handle_pollfd(&pfd[PFD_PIPE_SESSION], ibuf_se) == -1) {
275			log_warnx("main: Lost connection to SE");
276			msgbuf_clear(&ibuf_se->w);
277			free(ibuf_se);
278			ibuf_se = NULL;
279			quit = 1;
280		} else {
281			if (dispatch_imsg(ibuf_se, PFD_PIPE_SESSION, conf) ==
282			    -1)
283				quit = 1;
284		}
285
286		if (handle_pollfd(&pfd[PFD_PIPE_ROUTE], ibuf_rde) == -1) {
287			log_warnx("main: Lost connection to RDE");
288			msgbuf_clear(&ibuf_rde->w);
289			free(ibuf_rde);
290			ibuf_rde = NULL;
291			quit = 1;
292		} else {
293			if (dispatch_imsg(ibuf_rde, PFD_PIPE_ROUTE, conf) ==
294			    -1)
295				quit = 1;
296		}
297
298		if (pfd[PFD_SOCK_ROUTE].revents & POLLIN) {
299			if (kr_dispatch_msg(conf->default_tableid) == -1)
300				quit = 1;
301		}
302
303		if (reconfig) {
304			u_int	error;
305
306			reconfig = 0;
307			switch (reconfigure(conffile, conf, &peer_l)) {
308			case -1:	/* fatal error */
309				quit = 1;
310				break;
311			case 0:		/* all OK */
312				error = 0;
313				break;
314			case 2:
315				error = CTL_RES_PENDING;
316				break;
317			default:	/* parse error */
318				error = CTL_RES_PARSE_ERROR;
319				break;
320			}
321			if (reconfpid != 0) {
322				send_imsg_session(IMSG_CTL_RESULT, reconfpid,
323				    &error, sizeof(error));
324				reconfpid = 0;
325			}
326		}
327
328		if (mrtdump) {
329			mrtdump = 0;
330			mrt_handler(conf->mrt);
331		}
332	}
333
334	/* close pipes */
335	if (ibuf_se) {
336		msgbuf_clear(&ibuf_se->w);
337		close(ibuf_se->fd);
338		free(ibuf_se);
339	}
340	if (ibuf_rde) {
341		msgbuf_clear(&ibuf_rde->w);
342		close(ibuf_rde->fd);
343		free(ibuf_rde);
344	}
345
346	while ((p = peer_l) != NULL) {
347		peer_l = p->next;
348		free(p);
349	}
350
351	control_cleanup(conf->csock);
352	control_cleanup(conf->rcsock);
353	carp_demote_shutdown();
354	kr_shutdown(conf->fib_priority, conf->default_tableid);
355	pftable_clear_all();
356
357	free_config(conf);
358
359	log_debug("waiting for children to terminate");
360	do {
361		pid = wait(&status);
362		if (pid == -1) {
363			if (errno != EINTR && errno != ECHILD)
364				fatal("wait");
365		} else if (WIFSIGNALED(status))
366			log_warnx("%s terminated; signal %d",
367			    (pid == rde_pid) ? "route decision engine" :
368			    "session engine", WTERMSIG(status));
369	} while (pid != -1 || (pid == -1 && errno == EINTR));
370
371	free(rcname);
372	free(cname);
373
374	log_info("terminating");
375	return (0);
376}
377
378pid_t
379start_child(enum bgpd_process p, char *argv0, int fd, int debug, int verbose)
380{
381	char *argv[5];
382	int argc = 0;
383	pid_t pid;
384
385	switch (pid = fork()) {
386	case -1:
387		fatal("cannot fork");
388	case 0:
389		break;
390	default:
391		close(fd);
392		return (pid);
393	}
394
395	if (dup2(fd, 3) == -1)
396		fatal("cannot setup imsg fd");
397
398	argv[argc++] = argv0;
399	switch (p) {
400	case PROC_MAIN:
401		fatalx("Can not start main process");
402	case PROC_RDE:
403		argv[argc++] = "-R";
404		break;
405	case PROC_SE:
406		argv[argc++] = "-S";
407		break;
408	}
409	if (debug)
410		argv[argc++] = "-d";
411	if (verbose)
412		argv[argc++] = "-v";
413	argv[argc++] = NULL;
414
415	execvp(argv0, argv);
416	fatal("execvp");
417}
418
419int
420send_filterset(struct imsgbuf *i, struct filter_set_head *set)
421{
422	struct filter_set	*s;
423
424	TAILQ_FOREACH(s, set, entry)
425		if (imsg_compose(i, IMSG_FILTER_SET, 0, 0, -1, s,
426		    sizeof(struct filter_set)) == -1)
427			return (-1);
428	return (0);
429}
430
431int
432reconfigure(char *conffile, struct bgpd_config *conf, struct peer **peer_l)
433{
434	struct peer		*p;
435	struct filter_rule	*r;
436	struct listen_addr	*la;
437	struct rde_rib		*rr;
438	struct rdomain		*rd;
439	struct prefixset	*ps;
440	struct prefixset_item	*psi;
441
442	if (reconfpending) {
443		log_info("previous reload still running");
444		return (2);
445	}
446	reconfpending = 2;	/* one per child */
447
448	log_info("rereading config");
449	if (parse_config(conffile, conf, peer_l)) {
450		log_warnx("config file %s has errors, not reloading",
451		    conffile);
452		reconfpending = 0;
453		return (1);
454	}
455
456	cflags = conf->flags;
457	prepare_listeners(conf);
458
459	/* start reconfiguration */
460	if (imsg_compose(ibuf_se, IMSG_RECONF_CONF, 0, 0, -1,
461	    conf, sizeof(struct bgpd_config)) == -1)
462		return (-1);
463	if (imsg_compose(ibuf_rde, IMSG_RECONF_CONF, 0, 0, -1,
464	    conf, sizeof(struct bgpd_config)) == -1)
465		return (-1);
466
467	TAILQ_FOREACH(la, conf->listen_addrs, entry) {
468		if (imsg_compose(ibuf_se, IMSG_RECONF_LISTENER, 0, 0, la->fd,
469		    la, sizeof(struct listen_addr)) == -1)
470			return (-1);
471		la->fd = -1;
472	}
473
474	if (control_setup(conf) == -1)
475		return (-1);
476
477	/* adjust fib syncing on reload */
478	ktable_preload();
479
480	/* RIBs for the RDE */
481	while ((rr = SIMPLEQ_FIRST(&ribnames))) {
482		SIMPLEQ_REMOVE_HEAD(&ribnames, entry);
483		if (ktable_update(rr->rtableid, rr->name, NULL,
484		    rr->flags, conf->fib_priority) == -1) {
485			log_warnx("failed to load rdomain %d",
486			    rr->rtableid);
487			return (-1);
488		}
489		if (imsg_compose(ibuf_rde, IMSG_RECONF_RIB, 0, 0, -1,
490		    rr, sizeof(struct rde_rib)) == -1)
491			return (-1);
492		free(rr);
493	}
494
495	/* send peer list to the SE */
496	for (p = *peer_l; p != NULL; p = p->next) {
497		if (imsg_compose(ibuf_se, IMSG_RECONF_PEER, p->conf.id, 0, -1,
498		    &p->conf, sizeof(struct peer_config)) == -1)
499			return (-1);
500	}
501
502	/* networks go via kroute to the RDE */
503	if (kr_net_reload(conf->default_tableid, &conf->networks))
504		return (-1);
505
506	/* prefixsets for filters in the RDE */
507	while ((ps = SIMPLEQ_FIRST(conf->prefixsets)) != NULL) {
508		SIMPLEQ_REMOVE_HEAD(conf->prefixsets, entry);
509		if (imsg_compose(ibuf_rde, IMSG_RECONF_PREFIXSET, 0, 0, -1,
510		    ps->name, sizeof(ps->name)) == -1)
511			return (-1);
512		while ((psi = SIMPLEQ_FIRST(&ps->psitems)) != NULL) {
513			SIMPLEQ_REMOVE_HEAD(&ps->psitems, entry);
514			if (imsg_compose(ibuf_rde, IMSG_RECONF_PREFIXSETITEM, 0,
515			    0, -1, psi, sizeof(*psi)) == -1)
516				return (-1);
517			free(psi);
518		}
519		free(ps);
520	}
521
522	/* as-sets for filters in the RDE */
523	if (as_sets_send(ibuf_rde, conf->as_sets) == -1)
524		return (-1);
525	as_sets_free(conf->as_sets);
526	conf->as_sets = NULL;
527
528	/* filters for the RDE */
529	while ((r = TAILQ_FIRST(conf->filters)) != NULL) {
530		TAILQ_REMOVE(conf->filters, r, entry);
531		if (imsg_compose(ibuf_rde, IMSG_RECONF_FILTER, 0, 0, -1,
532		    r, sizeof(struct filter_rule)) == -1)
533			return (-1);
534		if (send_filterset(ibuf_rde, &r->set) == -1)
535			return (-1);
536		filterset_free(&r->set);
537		free(r);
538	}
539
540	while ((rd = SIMPLEQ_FIRST(&conf->rdomains)) != NULL) {
541		SIMPLEQ_REMOVE_HEAD(&conf->rdomains, entry);
542		if (ktable_update(rd->rtableid, rd->descr, rd->ifmpe,
543		    rd->flags, conf->fib_priority) == -1) {
544			log_warnx("failed to load rdomain %d",
545			    rd->rtableid);
546			return (-1);
547		}
548		/* networks go via kroute to the RDE */
549		if (kr_net_reload(rd->rtableid, &rd->net_l))
550			return (-1);
551
552		if (imsg_compose(ibuf_rde, IMSG_RECONF_RDOMAIN, 0, 0, -1,
553		    rd, sizeof(*rd)) == -1)
554			return (-1);
555
556		/* export targets */
557		if (imsg_compose(ibuf_rde, IMSG_RECONF_RDOMAIN_EXPORT, 0, 0,
558		    -1, NULL, 0) == -1)
559			return (-1);
560		if (send_filterset(ibuf_rde, &rd->export) == -1)
561			return (-1);
562		filterset_free(&rd->export);
563
564		/* import targets */
565		if (imsg_compose(ibuf_rde, IMSG_RECONF_RDOMAIN_IMPORT, 0, 0,
566		    -1, NULL, 0) == -1)
567			return (-1);
568		if (send_filterset(ibuf_rde, &rd->import) == -1)
569			return (-1);
570		filterset_free(&rd->import);
571
572		if (imsg_compose(ibuf_rde, IMSG_RECONF_RDOMAIN_DONE, 0, 0,
573		    -1, NULL, 0) == -1)
574			return (-1);
575
576		free(rd);
577	}
578
579	/* signal the SE first then the RDE to activate the new config */
580	if (imsg_compose(ibuf_se, IMSG_RECONF_DONE, 0, 0, -1, NULL, 0) == -1)
581		return (-1);
582
583	/* mrt changes can be sent out of bound */
584	mrt_reconfigure(conf->mrt);
585	return (0);
586}
587
588int
589dispatch_imsg(struct imsgbuf *ibuf, int idx, struct bgpd_config *conf)
590{
591	struct imsg		 imsg;
592	ssize_t			 n;
593	int			 rv, verbose;
594
595	rv = 0;
596	while (ibuf) {
597		if ((n = imsg_get(ibuf, &imsg)) == -1)
598			return (-1);
599
600		if (n == 0)
601			break;
602
603		switch (imsg.hdr.type) {
604		case IMSG_KROUTE_CHANGE:
605			if (idx != PFD_PIPE_ROUTE)
606				log_warnx("route request not from RDE");
607			else if (imsg.hdr.len != IMSG_HEADER_SIZE +
608			    sizeof(struct kroute_full))
609				log_warnx("wrong imsg len");
610			else if (kr_change(imsg.hdr.peerid, imsg.data,
611			    conf->fib_priority))
612				rv = -1;
613			break;
614		case IMSG_KROUTE_DELETE:
615			if (idx != PFD_PIPE_ROUTE)
616				log_warnx("route request not from RDE");
617			else if (imsg.hdr.len != IMSG_HEADER_SIZE +
618			    sizeof(struct kroute_full))
619				log_warnx("wrong imsg len");
620			else if (kr_delete(imsg.hdr.peerid, imsg.data,
621			    conf->fib_priority))
622				rv = -1;
623			break;
624		case IMSG_NEXTHOP_ADD:
625			if (idx != PFD_PIPE_ROUTE)
626				log_warnx("nexthop request not from RDE");
627			else if (imsg.hdr.len != IMSG_HEADER_SIZE +
628			    sizeof(struct bgpd_addr))
629				log_warnx("wrong imsg len");
630			else if (kr_nexthop_add(imsg.hdr.peerid, imsg.data,
631			    conf) == -1)
632				rv = -1;
633			break;
634		case IMSG_NEXTHOP_REMOVE:
635			if (idx != PFD_PIPE_ROUTE)
636				log_warnx("nexthop request not from RDE");
637			else if (imsg.hdr.len != IMSG_HEADER_SIZE +
638			    sizeof(struct bgpd_addr))
639				log_warnx("wrong imsg len");
640			else
641				kr_nexthop_delete(imsg.hdr.peerid, imsg.data,
642				    conf);
643			break;
644		case IMSG_PFTABLE_ADD:
645			if (idx != PFD_PIPE_ROUTE)
646				log_warnx("pftable request not from RDE");
647			else
648				if (imsg.hdr.len != IMSG_HEADER_SIZE +
649				    sizeof(struct pftable_msg))
650					log_warnx("wrong imsg len");
651				else if (pftable_addr_add(imsg.data) != 0)
652					rv = -1;
653			break;
654		case IMSG_PFTABLE_REMOVE:
655			if (idx != PFD_PIPE_ROUTE)
656				log_warnx("pftable request not from RDE");
657			else
658				if (imsg.hdr.len != IMSG_HEADER_SIZE +
659				    sizeof(struct pftable_msg))
660					log_warnx("wrong imsg len");
661				else if (pftable_addr_remove(imsg.data) != 0)
662					rv = -1;
663			break;
664		case IMSG_PFTABLE_COMMIT:
665			if (idx != PFD_PIPE_ROUTE)
666				log_warnx("pftable request not from RDE");
667			else
668				if (imsg.hdr.len != IMSG_HEADER_SIZE)
669					log_warnx("wrong imsg len");
670				else if (pftable_commit() != 0)
671					rv = -1;
672			break;
673		case IMSG_CTL_RELOAD:
674			if (idx != PFD_PIPE_SESSION)
675				log_warnx("reload request not from SE");
676			else {
677				reconfig = 1;
678				reconfpid = imsg.hdr.pid;
679			}
680			break;
681		case IMSG_CTL_FIB_COUPLE:
682			if (idx != PFD_PIPE_SESSION)
683				log_warnx("couple request not from SE");
684			else
685				kr_fib_couple(imsg.hdr.peerid,
686				    conf->fib_priority);
687			break;
688		case IMSG_CTL_FIB_DECOUPLE:
689			if (idx != PFD_PIPE_SESSION)
690				log_warnx("decouple request not from SE");
691			else
692				kr_fib_decouple(imsg.hdr.peerid,
693				    conf->fib_priority);
694			break;
695		case IMSG_CTL_KROUTE:
696		case IMSG_CTL_KROUTE_ADDR:
697		case IMSG_CTL_SHOW_NEXTHOP:
698		case IMSG_CTL_SHOW_INTERFACE:
699		case IMSG_CTL_SHOW_FIB_TABLES:
700			if (idx != PFD_PIPE_SESSION)
701				log_warnx("kroute request not from SE");
702			else
703				kr_show_route(&imsg);
704			break;
705		case IMSG_IFINFO:
706			if (idx != PFD_PIPE_SESSION)
707				log_warnx("IFINFO request not from SE");
708			else if (imsg.hdr.len != IMSG_HEADER_SIZE + IFNAMSIZ)
709				log_warnx("IFINFO request with wrong len");
710			else
711				kr_ifinfo(imsg.data);
712			break;
713		case IMSG_DEMOTE:
714			if (idx != PFD_PIPE_SESSION)
715				log_warnx("demote request not from SE");
716			else if (imsg.hdr.len != IMSG_HEADER_SIZE +
717			    sizeof(struct demote_msg))
718				log_warnx("DEMOTE request with wrong len");
719			else {
720				struct demote_msg	*msg;
721
722				msg = imsg.data;
723				carp_demote_set(msg->demote_group, msg->level);
724			}
725			break;
726		case IMSG_CTL_LOG_VERBOSE:
727			/* already checked by SE */
728			memcpy(&verbose, imsg.data, sizeof(verbose));
729			log_setverbose(verbose);
730			break;
731		case IMSG_RECONF_DONE:
732			if (reconfpending == 0)
733				log_warnx("unexpected RECONF_DONE received");
734			else if (reconfpending == 2) {
735				imsg_compose(ibuf_rde, IMSG_RECONF_DONE, 0,
736				    0, -1, NULL, 0);
737
738				/* finally fix kroute information */
739				ktable_postload(conf->fib_priority);
740
741				/* redistribute list needs to be reloaded too */
742				kr_reload();
743			}
744			reconfpending--;
745			break;
746		default:
747			break;
748		}
749		imsg_free(&imsg);
750		if (rv != 0)
751			return (rv);
752	}
753	return (0);
754}
755
756void
757send_nexthop_update(struct kroute_nexthop *msg)
758{
759	char	*gw = NULL;
760
761	if (msg->gateway.aid)
762		if (asprintf(&gw, ": via %s",
763		    log_addr(&msg->gateway)) == -1) {
764			log_warn("send_nexthop_update");
765			quit = 1;
766		}
767
768	log_debug("nexthop %s now %s%s%s", log_addr(&msg->nexthop),
769	    msg->valid ? "valid" : "invalid",
770	    msg->connected ? ": directly connected" : "",
771	    msg->gateway.aid ? gw : "");
772
773	free(gw);
774
775	if (imsg_compose(ibuf_rde, IMSG_NEXTHOP_UPDATE, 0, 0, -1,
776	    msg, sizeof(struct kroute_nexthop)) == -1)
777		quit = 1;
778}
779
780void
781send_imsg_session(int type, pid_t pid, void *data, u_int16_t datalen)
782{
783	imsg_compose(ibuf_se, type, 0, pid, -1, data, datalen);
784}
785
786int
787send_network(int type, struct network_config *net, struct filter_set_head *h)
788{
789	if (imsg_compose(ibuf_rde, type, 0, 0, -1, net,
790	    sizeof(struct network_config)) == -1)
791		return (-1);
792	/* networks that get deleted don't need to send the filter set */
793	if (type == IMSG_NETWORK_REMOVE)
794		return (0);
795	if (send_filterset(ibuf_rde, h) == -1)
796		return (-1);
797	if (imsg_compose(ibuf_rde, IMSG_NETWORK_DONE, 0, 0, -1, NULL, 0) == -1)
798		return (-1);
799
800	return (0);
801}
802
803int
804bgpd_filternexthop(struct kroute *kr, struct kroute6 *kr6)
805{
806	/* kernel routes are never filtered */
807	if (kr && kr->flags & F_KERNEL && kr->prefixlen != 0)
808		return (0);
809	if (kr6 && kr6->flags & F_KERNEL && kr6->prefixlen != 0)
810		return (0);
811
812	if (cflags & BGPD_FLAG_NEXTHOP_BGP) {
813		if (kr && kr->flags & F_BGPD_INSERTED)
814			return (0);
815		if (kr6 && kr6->flags & F_BGPD_INSERTED)
816			return (0);
817	}
818
819	if (cflags & BGPD_FLAG_NEXTHOP_DEFAULT) {
820		if (kr && kr->prefixlen == 0)
821			return (0);
822		if (kr6 && kr6->prefixlen == 0)
823			return (0);
824	}
825
826	return (1);
827}
828
829int
830control_setup(struct bgpd_config *conf)
831{
832	int fd, restricted;
833
834	/* control socket is outside chroot */
835	if (!cname || strcmp(cname, conf->csock)) {
836		if (cname) {
837			control_cleanup(cname);
838			free(cname);
839		}
840		if ((cname = strdup(conf->csock)) == NULL)
841			fatal("strdup");
842		if ((fd = control_init(0, cname)) == -1)
843			fatalx("control socket setup failed");
844		if (control_listen(fd) == -1)
845			fatalx("control socket setup failed");
846		restricted = 0;
847		if (imsg_compose(ibuf_se, IMSG_RECONF_CTRL, 0, 0, fd,
848		    &restricted, sizeof(restricted)) == -1)
849			return (-1);
850	}
851	if (!conf->rcsock) {
852		/* remove restricted socket */
853		control_cleanup(rcname);
854		free(rcname);
855		rcname = NULL;
856	} else if (!rcname || strcmp(rcname, conf->rcsock)) {
857		if (rcname) {
858			control_cleanup(rcname);
859			free(rcname);
860		}
861		if ((rcname = strdup(conf->rcsock)) == NULL)
862			fatal("strdup");
863		if ((fd = control_init(1, rcname)) == -1)
864			fatalx("control socket setup failed");
865		if (control_listen(fd) == -1)
866			fatalx("control socket setup failed");
867		restricted = 1;
868		if (imsg_compose(ibuf_se, IMSG_RECONF_CTRL, 0, 0, fd,
869		    &restricted, sizeof(restricted)) == -1)
870			return (-1);
871	}
872	return (0);
873}
874
875void
876set_pollfd(struct pollfd *pfd, struct imsgbuf *i)
877{
878	if (i == NULL || i->fd == -1) {
879		pfd->fd = -1;
880		return;
881	}
882	pfd->fd = i->fd;
883	pfd->events = POLLIN;
884	if (i->w.queued > 0)
885		pfd->events |= POLLOUT;
886}
887
888int
889handle_pollfd(struct pollfd *pfd, struct imsgbuf *i)
890{
891	ssize_t n;
892
893	if (i == NULL)
894		return (0);
895
896	if (pfd->revents & POLLOUT)
897		if (msgbuf_write(&i->w) <= 0 && errno != EAGAIN) {
898			log_warn("imsg write error");
899			close(i->fd);
900			i->fd = -1;
901			return (-1);
902		}
903
904	if (pfd->revents & POLLIN) {
905		if ((n = imsg_read(i)) == -1 && errno != EAGAIN) {
906			log_warn("imsg read error");
907			close(i->fd);
908			i->fd = -1;
909			return (-1);
910		}
911		if (n == 0) {
912			log_warnx("peer closed imsg connection");
913			close(i->fd);
914			i->fd = -1;
915			return (-1);
916		}
917	}
918	return (0);
919}
920
921int
922imsg_send_sockets(struct imsgbuf *se, struct imsgbuf *rde)
923{
924	int pipe_s2r[2];
925	int pipe_s2r_ctl[2];
926
927	if (socketpair(AF_UNIX, SOCK_STREAM | SOCK_CLOEXEC | SOCK_NONBLOCK,
928	     PF_UNSPEC, pipe_s2r) == -1)
929		return (-1);
930	if (socketpair(AF_UNIX, SOCK_STREAM | SOCK_CLOEXEC | SOCK_NONBLOCK,
931	     PF_UNSPEC, pipe_s2r_ctl) == -1)
932		return (-1);
933
934	if (imsg_compose(se, IMSG_SOCKET_CONN, 0, 0, pipe_s2r[0],
935	    NULL, 0) == -1)
936		return (-1);
937	if (imsg_compose(rde, IMSG_SOCKET_CONN, 0, 0, pipe_s2r[1],
938	    NULL, 0) == -1)
939		return (-1);
940
941	if (imsg_compose(se, IMSG_SOCKET_CONN_CTL, 0, 0, pipe_s2r_ctl[0],
942	    NULL, 0) == -1)
943		return (-1);
944	if (imsg_compose(rde, IMSG_SOCKET_CONN_CTL, 0, 0, pipe_s2r_ctl[1],
945	    NULL, 0) == -1)
946		return (-1);
947
948	return (0);
949}
950