2 * Copyright (c) 2002 Michael Shalayeff. All rights reserved.
3 * Copyright (c) 2003 Ryan McBride. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
15 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
16 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
17 * IN NO EVENT SHALL THE AUTHOR OR HIS RELATIVES BE LIABLE FOR ANY DIRECT,
18 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
19 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
20 * SERVICES; LOSS OF MIND, USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
22 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
23 * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
24 * THE POSSIBILITY OF SUCH DAMAGE.
27 * $FreeBSD: src/sys/netinet/ip_carp.c,v 1.48 2007/02/02 09:39:09 glebius Exp $
28 * $DragonFly: src/sys/netinet/ip_carp.c,v 1.4 2007/10/20 10:28:44 sephe Exp $
32 /*#include "opt_bpf.h"*/
34 #include "opt_inet6.h"
36 #include <sys/types.h>
37 #include <sys/param.h>
38 #include <sys/systm.h>
40 #include <sys/kernel.h>
41 #include <machine/limits.h>
42 #include <sys/malloc.h>
44 #include <sys/module.h>
47 #include <sys/sysctl.h>
48 #include <sys/syslog.h>
49 #include <sys/signalvar.h>
50 #include <sys/filio.h>
51 #include <sys/sockio.h>
52 #include <sys/in_cksum.h>
53 #include <sys/socket.h>
54 #include <sys/vnode.h>
56 #include <machine/stdarg.h>
59 #include <net/ethernet.h>
61 #include <net/if_dl.h>
62 #include <net/if_types.h>
63 #include <net/route.h>
66 #include <netinet/in.h>
67 #include <netinet/in_var.h>
68 #include <netinet/in_systm.h>
69 #include <netinet/ip.h>
70 #include <netinet/ip_var.h>
71 #include <netinet/if_ether.h>
75 #include <netinet/icmp6.h>
76 #include <netinet/ip6.h>
77 #include <netinet6/ip6_var.h>
78 #include <netinet6/scope6_var.h>
79 #include <netinet6/nd6.h>
82 #include <crypto/sha1.h>
83 #include <netinet/ip_carp.h>
86 #define CARP_IFNAME "carp"
87 static MALLOC_DEFINE(M_CARP, "CARP", "CARP interfaces");
88 static MALLOC_DEFINE(M_IFNET, "IFNET", "IFNET CARP?");
89 SYSCTL_DECL(_net_inet_carp);
92 struct ifnet *sc_ifp; /* Interface clue */
93 struct ifnet *sc_carpdev; /* Pointer to parent interface */
94 struct in_ifaddr *sc_ia; /* primary iface address */
95 struct ip_moptions sc_imo;
97 struct in6_ifaddr *sc_ia6; /* primary iface address v6 */
98 struct ip6_moptions sc_im6o;
100 TAILQ_ENTRY(carp_softc) sc_list;
102 enum { INIT = 0, BACKUP, MASTER } sc_state;
107 int sc_sendad_errors;
108 #define CARP_SENDAD_MAX_ERRORS 3
109 int sc_sendad_success;
110 #define CARP_SENDAD_MIN_SUCCESS 3
116 int sc_advbase; /* seconds */
118 u_int64_t sc_counter;
121 #define CARP_HMAC_PAD 64
122 unsigned char sc_key[CARP_KEY_LEN];
123 unsigned char sc_pad[CARP_HMAC_PAD];
126 struct callout sc_ad_tmo; /* advertisement timeout */
127 struct callout sc_md_tmo; /* master down timeout */
128 struct callout sc_md6_tmo; /* master down timeout */
130 LIST_ENTRY(carp_softc) sc_next; /* Interface clue */
132 #define SC2IFP(sc) ((sc)->sc_ifp)
134 int carp_suppress_preempt = 0;
135 int carp_opts[CARPCTL_MAXID] = { 0, 1, 0, 1, 0, 0 }; /* XXX for now */
136 SYSCTL_INT(_net_inet_carp, CARPCTL_ALLOW, allow, CTLFLAG_RW,
137 &carp_opts[CARPCTL_ALLOW], 0, "Accept incoming CARP packets");
138 SYSCTL_INT(_net_inet_carp, CARPCTL_PREEMPT, preempt, CTLFLAG_RW,
139 &carp_opts[CARPCTL_PREEMPT], 0, "high-priority backup preemption mode");
140 SYSCTL_INT(_net_inet_carp, CARPCTL_LOG, log, CTLFLAG_RW,
141 &carp_opts[CARPCTL_LOG], 0, "log bad carp packets");
142 SYSCTL_INT(_net_inet_carp, CARPCTL_ARPBALANCE, arpbalance, CTLFLAG_RW,
143 &carp_opts[CARPCTL_ARPBALANCE], 0, "balance arp responses");
144 SYSCTL_INT(_net_inet_carp, OID_AUTO, suppress_preempt, CTLFLAG_RD,
145 &carp_suppress_preempt, 0, "Preemption is suppressed");
147 struct carpstats carpstats;
148 SYSCTL_STRUCT(_net_inet_carp, CARPCTL_STATS, stats, CTLFLAG_RW,
149 &carpstats, carpstats,
150 "CARP statistics (struct carpstats, netinet/ip_carp.h)");
153 TAILQ_HEAD(, carp_softc) vhif_vrs;
156 struct ifnet *vhif_ifp;
157 struct lock vhif_lock;
160 /* Get carp_if from softc. Valid after carp_set_addr{,6}. */
161 #define SC2CIF(sc) ((struct carp_if *)(sc)->sc_carpdev->if_carp)
163 #define CARP_LOCK_INIT(cif) lockinit(&(cif)->vhif_lock, "carp_if", 0, LK_NOWAIT);
164 #define CARP_LOCK_DESTROY(cif) ;
165 #define CARP_LOCK_ASSERT(cif) ;
166 #define CARP_LOCK(cif) lockmgr(&(cif)->vhif_lock, LK_EXCLUSIVE);
167 #define CARP_UNLOCK(cif) lockmgr(&(cif)->vhif_lock, LK_RELEASE);
169 #define CARP_SCLOCK(sc) lockmgr(&SC2CIF(sc)->vhif_lock, LK_EXCLUSIVE);
170 #define CARP_SCUNLOCK(sc) lockmgr(&SC2CIF(sc)->vhif_lock, LK_RELEASE);
171 #define CARP_SCLOCK_ASSERT(sc) ;
173 #define CARP_LOG(...) do { \
174 if (carp_opts[CARPCTL_LOG] > 0) \
175 log(LOG_INFO, __VA_ARGS__); \
178 #define CARP_DEBUG(...) do { \
179 if (carp_opts[CARPCTL_LOG] > 1) \
180 log(LOG_DEBUG, __VA_ARGS__); \
183 static void carp_hmac_prepare(struct carp_softc *);
184 static void carp_hmac_generate(struct carp_softc *, u_int32_t *,
186 static int carp_hmac_verify(struct carp_softc *, u_int32_t *,
188 static void carp_setroute(struct carp_softc *, int);
189 static void carp_input_c(struct mbuf *, struct carp_header *, sa_family_t);
190 static int carp_clone_create(struct if_clone *, int);
191 static void carp_clone_destroy(struct ifnet *);
192 static void carpdetach(struct carp_softc *, int);
193 static int carp_prepare_ad(struct mbuf *, struct carp_softc *,
194 struct carp_header *);
195 static void carp_send_ad_all(void);
196 static void carp_send_ad(void *);
197 static void carp_send_ad_locked(struct carp_softc *);
198 static void carp_send_arp(struct carp_softc *);
199 static void carp_master_down(void *);
200 static void carp_master_down_locked(struct carp_softc *);
201 static int carp_ioctl(struct ifnet *, u_long, caddr_t, struct ucred *);
202 static int carp_looutput(struct ifnet *, struct mbuf *, struct sockaddr *,
204 static void carp_start(struct ifnet *);
205 static void carp_setrun(struct carp_softc *, sa_family_t);
206 static void carp_set_state(struct carp_softc *, int);
207 static int carp_addrcount(struct carp_if *, struct in_ifaddr *, int);
208 enum { CARP_COUNT_MASTER, CARP_COUNT_RUNNING };
210 static void carp_multicast_cleanup(struct carp_softc *);
211 static int carp_set_addr(struct carp_softc *, struct sockaddr_in *);
212 static int carp_del_addr(struct carp_softc *, struct sockaddr_in *);
213 static void carp_carpdev_state_locked(struct carp_if *);
214 static void carp_sc_state_locked(struct carp_softc *);
216 static void carp_send_na(struct carp_softc *);
217 static int carp_set_addr6(struct carp_softc *, struct sockaddr_in6 *);
218 static int carp_del_addr6(struct carp_softc *, struct sockaddr_in6 *);
219 static void carp_multicast6_cleanup(struct carp_softc *);
222 static LIST_HEAD(, carp_softc) carpif_list;
224 struct if_clone carp_cloner = IF_CLONE_INITIALIZER(CARP_IFNAME, carp_clone_create, carp_clone_destroy, 0, IF_MAXUNIT);
226 static eventhandler_tag if_detach_event_tag;
228 static __inline u_int16_t
229 carp_cksum(struct mbuf *m, int len)
231 return (in_cksum(m, len));
235 carp_hmac_prepare(struct carp_softc *sc)
237 u_int8_t version = CARP_VERSION, type = CARP_ADVERTISEMENT;
238 u_int8_t vhid = sc->sc_vhid & 0xff;
248 /* XXX: possible race here */
250 /* compute ipad from key */
251 bzero(sc->sc_pad, sizeof(sc->sc_pad));
252 bcopy(sc->sc_key, sc->sc_pad, sizeof(sc->sc_key));
253 for (i = 0; i < sizeof(sc->sc_pad); i++)
254 sc->sc_pad[i] ^= 0x36;
256 /* precompute first part of inner hash */
257 SHA1Init(&sc->sc_sha1);
258 SHA1Update(&sc->sc_sha1, sc->sc_pad, sizeof(sc->sc_pad));
259 SHA1Update(&sc->sc_sha1, (void *)&version, sizeof(version));
260 SHA1Update(&sc->sc_sha1, (void *)&type, sizeof(type));
261 SHA1Update(&sc->sc_sha1, (void *)&vhid, sizeof(vhid));
263 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
264 if (ifa->ifa_addr->sa_family == AF_INET)
265 SHA1Update(&sc->sc_sha1,
266 (void *)&ifatoia(ifa)->ia_addr.sin_addr.s_addr,
267 sizeof(struct in_addr));
271 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
272 if (ifa->ifa_addr->sa_family == AF_INET6) {
273 in6 = ifatoia6(ifa)->ia_addr.sin6_addr;
274 in6_clearscope(&in6);
275 SHA1Update(&sc->sc_sha1, (void *)&in6, sizeof(in6));
280 /* convert ipad to opad */
281 for (i = 0; i < sizeof(sc->sc_pad); i++)
282 sc->sc_pad[i] ^= 0x36 ^ 0x5c;
289 carp_hmac_generate(struct carp_softc *sc, u_int32_t counter[2],
290 unsigned char md[20])
294 /* fetch first half of inner hash */
295 bcopy(&sc->sc_sha1, &sha1ctx, sizeof(sha1ctx));
297 SHA1Update(&sha1ctx, (void *)counter, sizeof(sc->sc_counter));
298 SHA1Final(md, &sha1ctx);
302 SHA1Update(&sha1ctx, sc->sc_pad, sizeof(sc->sc_pad));
303 SHA1Update(&sha1ctx, md, 20);
304 SHA1Final(md, &sha1ctx);
308 carp_hmac_verify(struct carp_softc *sc, u_int32_t counter[2],
309 unsigned char md[20])
311 unsigned char md2[20];
313 CARP_SCLOCK_ASSERT(sc);
315 carp_hmac_generate(sc, counter, md2);
317 return (bcmp(md, md2, sizeof(md2)));
321 carp_setroute(struct carp_softc *sc, int cmd)
326 CARP_SCLOCK_ASSERT(sc);
329 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
330 if (ifa->ifa_addr->sa_family == AF_INET &&
331 sc->sc_carpdev != NULL) {
332 int count = carp_addrcount(
333 (struct carp_if *)sc->sc_carpdev->if_carp,
334 ifatoia(ifa), CARP_COUNT_MASTER);
336 if ((cmd == RTM_ADD && count == 1) ||
337 (cmd == RTM_DELETE && count == 0))
338 rtinit(ifa, cmd, RTF_UP | RTF_HOST);
341 if (ifa->ifa_addr->sa_family == AF_INET6) {
354 carp_clone_create(struct if_clone *ifc, int unit)
357 struct carp_softc *sc;
360 MALLOC(sc, struct carp_softc *, sizeof(*sc), M_CARP, M_WAITOK|M_ZERO);
361 ifp = SC2IFP(sc) = kmalloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
368 sc->sc_flags_backup = 0;
370 sc->sc_advbase = CARP_DFLTINTV;
371 sc->sc_vhid = -1; /* required setting */
373 sc->sc_init_counter = 1;
374 sc->sc_naddrs = sc->sc_naddrs6 = 0; /* M_ZERO? */
377 sc->sc_im6o.im6o_multicast_hlim = CARP_DFLTTL;
380 /* sc->sc_imo.imo_membership = kmalloc((sizeof(struct in_multi) * IP_MAX_MEMBERSHIPS), M_CARP,M_WAITOK);*/
382 sc->sc_imo.imo_max_memberships = IP_MAX_MEMBERSHIPS;
383 sc->sc_imo.imo_multicast_vif = -1;
385 callout_init(&sc->sc_ad_tmo);
386 callout_init(&sc->sc_md_tmo);
387 callout_init(&sc->sc_md6_tmo);
390 if_initname(ifp, CARP_IFNAME, unit);
391 ifp->if_mtu = ETHERMTU;
392 ifp->if_flags = IFF_LOOPBACK;
393 ifp->if_ioctl = carp_ioctl;
394 ifp->if_output = carp_looutput;
395 ifp->if_start = carp_start;
396 ifp->if_type = IFT_CARP;
397 ifp->if_snd.ifq_maxlen = ifqmaxlen;
399 if_attach(ifp, NULL);
400 bpfattach(ifp, DLT_NULL, sizeof(u_int));
403 LIST_INSERT_HEAD(&carpif_list, sc, sc_next);
410 carp_clone_destroy(struct ifnet *ifp)
412 struct carp_softc *sc = ifp->if_softc;
416 carpdetach(sc, 1); /* Returns unlocked. */
419 LIST_REMOVE(sc, sc_next);
423 /* if_free_type(ifp, IFT_ETHER);*/
424 /* kfree(sc->sc_imo.imo_membership, M_CARP); */
429 * This function can be called on CARP interface destroy path,
430 * and in case of the removal of the underlying interface as
431 * well. We differentiate these two cases. In the latter case
432 * we do not cleanup our multicast memberships, since they
433 * are already freed. Also, in the latter case we do not
434 * release the lock on return, because the function will be
435 * called once more, for another CARP instance on the same
439 carpdetach(struct carp_softc *sc, int unlock)
443 callout_stop(&sc->sc_ad_tmo);
444 callout_stop(&sc->sc_md_tmo);
445 callout_stop(&sc->sc_md6_tmo);
448 carp_suppress_preempt--;
451 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS)
452 carp_suppress_preempt--;
453 sc->sc_sendad_errors = 0;
455 carp_set_state(sc, INIT);
456 SC2IFP(sc)->if_flags &= ~IFF_UP;
459 carp_multicast_cleanup(sc);
461 carp_multicast6_cleanup(sc);
464 if (sc->sc_carpdev != NULL) {
465 cif = (struct carp_if *)sc->sc_carpdev->if_carp;
466 CARP_LOCK_ASSERT(cif);
467 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list);
468 if (!--cif->vhif_nvrs) {
469 ifpromisc(sc->sc_carpdev, 0);
470 sc->sc_carpdev->if_carp = NULL;
471 CARP_LOCK_DESTROY(cif);
475 sc->sc_carpdev = NULL;
479 /* Detach an interface from the carp. */
481 carp_ifdetach(void *arg __unused, struct ifnet *ifp)
483 struct carp_if *cif = (struct carp_if *)ifp->if_carp;
484 struct carp_softc *sc, *nextsc;
490 * XXX: At the end of for() cycle the lock will be destroyed.
493 for (sc = TAILQ_FIRST(&cif->vhif_vrs); sc; sc = nextsc) {
494 nextsc = TAILQ_NEXT(sc, sc_list);
501 * process input packet.
502 * we have rearranged checks order compared to the rfc,
503 * but it seems more efficient this way or not possible otherwise.
506 carp_input(struct mbuf *m, int hlen)
508 struct ip *ip = mtod(m, struct ip *);
509 struct carp_header *ch;
512 carpstats.carps_ipackets++;
514 if (!carp_opts[CARPCTL_ALLOW]) {
519 /* check if received on a valid carp interface */
520 if (m->m_pkthdr.rcvif->if_carp == NULL) {
521 carpstats.carps_badif++;
522 CARP_LOG("carp_input: packet received on non-carp "
524 m->m_pkthdr.rcvif->if_xname);
529 /* verify that the IP TTL is 255. */
530 if (ip->ip_ttl != CARP_DFLTTL) {
531 carpstats.carps_badttl++;
532 CARP_LOG("carp_input: received ttl %d != 255i on %s\n",
534 m->m_pkthdr.rcvif->if_xname);
539 iplen = ip->ip_hl << 2;
541 if (m->m_pkthdr.len < iplen + sizeof(*ch)) {
542 carpstats.carps_badlen++;
543 CARP_LOG("carp_input: received len %zd < "
544 "sizeof(struct carp_header)\n",
545 m->m_len - sizeof(struct ip));
550 if (iplen + sizeof(*ch) < m->m_len) {
551 if ((m = m_pullup(m, iplen + sizeof(*ch))) == NULL) {
552 carpstats.carps_hdrops++;
553 CARP_LOG("carp_input: pullup failed\n");
556 ip = mtod(m, struct ip *);
558 ch = (struct carp_header *)((char *)ip + iplen);
561 * verify that the received packet length is
562 * equal to the CARP header
564 len = iplen + sizeof(*ch);
565 if (len > m->m_pkthdr.len) {
566 carpstats.carps_badlen++;
567 CARP_LOG("carp_input: packet too short %d on %s\n",
569 m->m_pkthdr.rcvif->if_xname);
574 if ((m = m_pullup(m, len)) == NULL) {
575 carpstats.carps_hdrops++;
578 ip = mtod(m, struct ip *);
579 ch = (struct carp_header *)((char *)ip + iplen);
581 /* verify the CARP checksum */
583 if (carp_cksum(m, len - iplen)) {
584 carpstats.carps_badsum++;
585 CARP_LOG("carp_input: checksum failed on %s\n",
586 m->m_pkthdr.rcvif->if_xname);
592 carp_input_c(m, ch, AF_INET);
597 carp6_input(struct mbuf **mp, int *offp, int proto)
599 struct mbuf *m = *mp;
600 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
601 struct carp_header *ch;
604 carpstats.carps_ipackets6++;
606 if (!carp_opts[CARPCTL_ALLOW]) {
608 return (IPPROTO_DONE);
611 /* check if received on a valid carp interface */
612 if (m->m_pkthdr.rcvif->if_carp == NULL) {
613 carpstats.carps_badif++;
614 CARP_LOG("carp6_input: packet received on non-carp "
616 m->m_pkthdr.rcvif->if_xname);
618 return (IPPROTO_DONE);
621 /* verify that the IP TTL is 255 */
622 if (ip6->ip6_hlim != CARP_DFLTTL) {
623 carpstats.carps_badttl++;
624 CARP_LOG("carp6_input: received ttl %d != 255 on %s\n",
626 m->m_pkthdr.rcvif->if_xname);
628 return (IPPROTO_DONE);
631 /* verify that we have a complete carp packet */
633 IP6_EXTHDR_GET(ch, struct carp_header *, m, *offp, sizeof(*ch));
635 carpstats.carps_badlen++;
636 CARP_LOG("carp6_input: packet size %u too small\n", len);
637 return (IPPROTO_DONE);
641 /* verify the CARP checksum */
643 if (carp_cksum(m, sizeof(*ch))) {
644 carpstats.carps_badsum++;
645 CARP_LOG("carp6_input: checksum failed, on %s\n",
646 m->m_pkthdr.rcvif->if_xname);
648 return (IPPROTO_DONE);
652 carp_input_c(m, ch, AF_INET6);
653 return (IPPROTO_DONE);
658 carp_input_c(struct mbuf *m, struct carp_header *ch, sa_family_t af)
660 struct ifnet *ifp = m->m_pkthdr.rcvif;
661 struct carp_softc *sc;
662 u_int64_t tmp_counter;
663 struct timeval sc_tv, ch_tv;
665 /* verify that the VHID is valid on the receiving interface */
666 CARP_LOCK(ifp->if_carp);
667 TAILQ_FOREACH(sc, &((struct carp_if *)ifp->if_carp)->vhif_vrs, sc_list)
668 if (sc->sc_vhid == ch->carp_vhid)
671 if (!sc || !((SC2IFP(sc)->if_flags & IFF_UP) && (SC2IFP(sc)->if_flags & IFF_RUNNING))) {
672 carpstats.carps_badvhid++;
673 CARP_UNLOCK(ifp->if_carp);
678 getmicrotime(&SC2IFP(sc)->if_lastchange);
679 SC2IFP(sc)->if_ipackets++;
680 SC2IFP(sc)->if_ibytes += m->m_pkthdr.len;
682 if (SC2IFP(sc)->if_bpf) {
683 struct ip *ip = mtod(m, struct ip *);
685 /* BPF wants net byte order */
686 ip->ip_len = htons(ip->ip_len + (ip->ip_hl << 2));
687 ip->ip_off = htons(ip->ip_off);
688 bpf_mtap(SC2IFP(sc)->if_bpf, m);
691 /* verify the CARP version. */
692 if (ch->carp_version != CARP_VERSION) {
693 carpstats.carps_badver++;
694 SC2IFP(sc)->if_ierrors++;
695 CARP_UNLOCK(ifp->if_carp);
696 CARP_LOG("%s; invalid version %d\n",
697 SC2IFP(sc)->if_xname,
703 /* verify the hash */
704 if (carp_hmac_verify(sc, ch->carp_counter, ch->carp_md)) {
705 carpstats.carps_badauth++;
706 SC2IFP(sc)->if_ierrors++;
707 CARP_UNLOCK(ifp->if_carp);
708 CARP_LOG("%s: incorrect hash\n", SC2IFP(sc)->if_xname);
713 tmp_counter = ntohl(ch->carp_counter[0]);
714 tmp_counter = tmp_counter<<32;
715 tmp_counter += ntohl(ch->carp_counter[1]);
717 /* XXX Replay protection goes here */
719 sc->sc_init_counter = 0;
720 sc->sc_counter = tmp_counter;
722 sc_tv.tv_sec = sc->sc_advbase;
723 if (carp_suppress_preempt && sc->sc_advskew < 240)
724 sc_tv.tv_usec = 240 * 1000000 / 256;
726 sc_tv.tv_usec = sc->sc_advskew * 1000000 / 256;
727 ch_tv.tv_sec = ch->carp_advbase;
728 ch_tv.tv_usec = ch->carp_advskew * 1000000 / 256;
730 switch (sc->sc_state) {
735 * If we receive an advertisement from a master who's going to
736 * be more frequent than us, go into BACKUP state.
738 if (timevalcmp(&sc_tv, &ch_tv, >) ||
739 timevalcmp(&sc_tv, &ch_tv, ==)) {
740 callout_stop(&sc->sc_ad_tmo);
741 CARP_DEBUG("%s: MASTER -> BACKUP "
742 "(more frequent advertisement received)\n",
743 SC2IFP(sc)->if_xname);
744 carp_set_state(sc, BACKUP);
746 carp_setroute(sc, RTM_DELETE);
751 * If we're pre-empting masters who advertise slower than us,
752 * and this one claims to be slower, treat him as down.
754 if (carp_opts[CARPCTL_PREEMPT] &&
755 timevalcmp(&sc_tv, &ch_tv, <)) {
756 CARP_DEBUG("%s: BACKUP -> MASTER "
757 "(preempting a slower master)\n",
758 SC2IFP(sc)->if_xname);
759 carp_master_down_locked(sc);
764 * If the master is going to advertise at such a low frequency
765 * that he's guaranteed to time out, we'd might as well just
766 * treat him as timed out now.
768 sc_tv.tv_sec = sc->sc_advbase * 3;
769 if (timevalcmp(&sc_tv, &ch_tv, <)) {
770 CARP_DEBUG("%s: BACKUP -> MASTER "
771 "(master timed out)\n",
772 SC2IFP(sc)->if_xname);
773 carp_master_down_locked(sc);
778 * Otherwise, we reset the counter and wait for the next
785 CARP_UNLOCK(ifp->if_carp);
792 carp_prepare_ad(struct mbuf *m, struct carp_softc *sc, struct carp_header *ch)
795 struct ifnet *ifp = SC2IFP(sc);
797 if (sc->sc_init_counter) {
798 /* this could also be seconds since unix epoch */
799 sc->sc_counter = karc4random();
800 sc->sc_counter = sc->sc_counter << 32;
801 sc->sc_counter += karc4random();
805 ch->carp_counter[0] = htonl((sc->sc_counter>>32)&0xffffffff);
806 ch->carp_counter[1] = htonl(sc->sc_counter&0xffffffff);
808 carp_hmac_generate(sc, ch->carp_counter, ch->carp_md);
810 /* Tag packet for carp_output */
811 mtag = m_tag_get(PACKET_TAG_CARP, sizeof(struct ifnet *), MB_DONTWAIT);
814 SC2IFP(sc)->if_oerrors++;
817 bcopy(&ifp, (caddr_t)(mtag + 1), sizeof(struct ifnet *));
818 m_tag_prepend(m, mtag);
824 carp_send_ad_all(void)
826 struct carp_softc *sc;
828 LIST_FOREACH(sc, &carpif_list, sc_next) {
829 if (sc->sc_carpdev == NULL)
832 if ((SC2IFP(sc)->if_flags & IFF_UP) && (SC2IFP(sc)->if_flags & IFF_RUNNING) &&
833 sc->sc_state == MASTER)
834 carp_send_ad_locked(sc);
840 carp_send_ad(void *v)
842 struct carp_softc *sc = v;
845 carp_send_ad_locked(sc);
850 carp_send_ad_locked(struct carp_softc *sc)
852 struct carp_header ch;
854 struct carp_header *ch_ptr;
856 int len, advbase, advskew;
859 /* bow out if we've lost our UPness or RUNNINGuiness */
860 if (!((SC2IFP(sc)->if_flags & IFF_UP) && (SC2IFP(sc)->if_flags & IFF_RUNNING))) {
864 advbase = sc->sc_advbase;
865 if (!carp_suppress_preempt || sc->sc_advskew > 240)
866 advskew = sc->sc_advskew;
870 tv.tv_usec = advskew * 1000000 / 256;
873 ch.carp_version = CARP_VERSION;
874 ch.carp_type = CARP_ADVERTISEMENT;
875 ch.carp_vhid = sc->sc_vhid;
876 ch.carp_advbase = advbase;
877 ch.carp_advskew = advskew;
878 ch.carp_authlen = 7; /* XXX DEFINE */
879 ch.carp_pad1 = 0; /* must be zero */
886 MGETHDR(m, M_NOWAIT, MT_HEADER);
888 SC2IFP(sc)->if_oerrors++;
889 carpstats.carps_onomem++;
890 /* XXX maybe less ? */
891 if (advbase != 255 || advskew != 255)
892 callout_reset(&sc->sc_ad_tmo, tvtohz_high(&tv),
896 len = sizeof(*ip) + sizeof(ch);
897 m->m_pkthdr.len = len;
898 m->m_pkthdr.rcvif = NULL;
900 MH_ALIGN(m, m->m_len);
901 m->m_flags |= M_MCAST;
902 ip = mtod(m, struct ip *);
903 ip->ip_v = IPVERSION;
904 ip->ip_hl = sizeof(*ip) >> 2;
905 ip->ip_tos = IPTOS_LOWDELAY;
907 ip->ip_id = ip_newid();
909 ip->ip_ttl = CARP_DFLTTL;
910 ip->ip_p = IPPROTO_CARP;
912 ip->ip_src.s_addr = sc->sc_ia->ia_addr.sin_addr.s_addr;
913 ip->ip_dst.s_addr = htonl(INADDR_CARP_GROUP);
915 ch_ptr = (struct carp_header *)(&ip[1]);
916 bcopy(&ch, ch_ptr, sizeof(ch));
917 if (carp_prepare_ad(m, sc, ch_ptr))
920 m->m_data += sizeof(*ip);
921 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip));
922 m->m_data -= sizeof(*ip);
924 getmicrotime(&SC2IFP(sc)->if_lastchange);
925 SC2IFP(sc)->if_opackets++;
926 SC2IFP(sc)->if_obytes += len;
927 carpstats.carps_opackets++;
929 if (ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, NULL)) {
930 SC2IFP(sc)->if_oerrors++;
931 if (sc->sc_sendad_errors < INT_MAX)
932 sc->sc_sendad_errors++;
933 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) {
934 carp_suppress_preempt++;
935 if (carp_suppress_preempt == 1) {
941 sc->sc_sendad_success = 0;
943 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) {
944 if (++sc->sc_sendad_success >=
945 CARP_SENDAD_MIN_SUCCESS) {
946 carp_suppress_preempt--;
947 sc->sc_sendad_errors = 0;
950 sc->sc_sendad_errors = 0;
958 MGETHDR(m, M_NOWAIT, MT_HEADER);
960 SC2IFP(sc)->if_oerrors++;
961 carpstats.carps_onomem++;
962 /* XXX maybe less ? */
963 if (advbase != 255 || advskew != 255)
964 callout_reset(&sc->sc_ad_tmo, tvtohz_high(&tv),
968 len = sizeof(*ip6) + sizeof(ch);
969 m->m_pkthdr.len = len;
970 m->m_pkthdr.rcvif = NULL;
972 MH_ALIGN(m, m->m_len);
973 m->m_flags |= M_MCAST;
974 ip6 = mtod(m, struct ip6_hdr *);
975 bzero(ip6, sizeof(*ip6));
976 ip6->ip6_vfc |= IPV6_VERSION;
977 ip6->ip6_hlim = CARP_DFLTTL;
978 ip6->ip6_nxt = IPPROTO_CARP;
979 bcopy(&sc->sc_ia6->ia_addr.sin6_addr, &ip6->ip6_src,
980 sizeof(struct in6_addr));
981 /* set the multicast destination */
983 ip6->ip6_dst.s6_addr16[0] = htons(0xff02);
984 ip6->ip6_dst.s6_addr8[15] = 0x12;
985 if (in6_setscope(&ip6->ip6_dst, sc->sc_carpdev, NULL) != 0) {
986 SC2IFP(sc)->if_oerrors++;
988 CARP_LOG("%s: in6_setscope failed\n", __func__);
992 ch_ptr = (struct carp_header *)(&ip6[1]);
993 bcopy(&ch, ch_ptr, sizeof(ch));
994 if (carp_prepare_ad(m, sc, ch_ptr))
997 m->m_data += sizeof(*ip6);
998 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip6));
999 m->m_data -= sizeof(*ip6);
1001 getmicrotime(&SC2IFP(sc)->if_lastchange);
1002 SC2IFP(sc)->if_opackets++;
1003 SC2IFP(sc)->if_obytes += len;
1004 carpstats.carps_opackets6++;
1006 if (ip6_output(m, NULL, NULL, 0, &sc->sc_im6o, NULL, NULL)) {
1007 SC2IFP(sc)->if_oerrors++;
1008 if (sc->sc_sendad_errors < INT_MAX)
1009 sc->sc_sendad_errors++;
1010 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) {
1011 carp_suppress_preempt++;
1012 if (carp_suppress_preempt == 1) {
1018 sc->sc_sendad_success = 0;
1020 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) {
1021 if (++sc->sc_sendad_success >=
1022 CARP_SENDAD_MIN_SUCCESS) {
1023 carp_suppress_preempt--;
1024 sc->sc_sendad_errors = 0;
1027 sc->sc_sendad_errors = 0;
1032 if (advbase != 255 || advskew != 255)
1033 callout_reset(&sc->sc_ad_tmo, tvtohz_high(&tv),
1039 * Broadcast a gratuitous ARP request containing
1040 * the virtual router MAC address for each IP address
1041 * associated with the virtual router.
1044 carp_send_arp(struct carp_softc *sc)
1048 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
1050 if (ifa->ifa_addr->sa_family != AF_INET)
1052 lwkt_serialize_enter(sc->sc_carpdev->if_serializer);
1053 arp_ifinit2(sc->sc_carpdev, ifa, IF_LLADDR(sc->sc_ifp));
1054 lwkt_serialize_exit(sc->sc_carpdev->if_serializer);
1056 DELAY(1000); /* XXX */
1062 carp_send_na(struct carp_softc *sc)
1065 struct in6_addr *in6;
1066 static struct in6_addr mcast = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
1068 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
1070 if (ifa->ifa_addr->sa_family != AF_INET6)
1073 in6 = &ifatoia6(ifa)->ia_addr.sin6_addr;
1074 nd6_na_output(sc->sc_carpdev, &mcast, in6,
1075 ND_NA_FLAG_OVERRIDE, 1, NULL);
1076 DELAY(1000); /* XXX */
1082 carp_addrcount(struct carp_if *cif, struct in_ifaddr *ia, int type)
1084 struct carp_softc *vh;
1088 CARP_LOCK_ASSERT(cif);
1090 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1091 if ((type == CARP_COUNT_RUNNING &&
1092 (SC2IFP(vh)->if_flags & IFF_UP) && (SC2IFP(vh)->if_flags & IFF_RUNNING)) ||
1093 (type == CARP_COUNT_MASTER && vh->sc_state == MASTER)) {
1094 TAILQ_FOREACH(ifa, &SC2IFP(vh)->if_addrlist,
1096 if (ifa->ifa_addr->sa_family == AF_INET &&
1097 ia->ia_addr.sin_addr.s_addr ==
1098 ifatoia(ifa)->ia_addr.sin_addr.s_addr)
1107 carp_iamatch(void *v, struct in_ifaddr *ia,
1108 struct in_addr *isaddr, u_int8_t **enaddr)
1110 struct carp_if *cif = v;
1111 struct carp_softc *vh;
1112 int index, count = 0;
1117 if (carp_opts[CARPCTL_ARPBALANCE]) {
1119 * XXX proof of concept implementation.
1120 * We use the source ip to decide which virtual host should
1121 * handle the request. If we're master of that virtual host,
1122 * then we respond, otherwise, just drop the arp packet on
1125 count = carp_addrcount(cif, ia, CARP_COUNT_RUNNING);
1127 /* should never reach this */
1132 /* this should be a hash, like pf_hash() */
1133 index = ntohl(isaddr->s_addr) % count;
1136 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1137 if ((SC2IFP(vh)->if_flags & IFF_UP) && (SC2IFP(vh)->if_flags & IFF_RUNNING)) {
1138 TAILQ_FOREACH(ifa, &SC2IFP(vh)->if_addrlist,
1140 if (ifa->ifa_addr->sa_family ==
1142 ia->ia_addr.sin_addr.s_addr ==
1143 ifatoia(ifa)->ia_addr.sin_addr.s_addr) {
1144 if (count == index) {
1147 *enaddr = IF_LLADDR(vh->sc_ifp);
1161 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1162 if ((SC2IFP(vh)->if_flags & IFF_UP) && (SC2IFP(vh)->if_flags & IFF_RUNNING) &&
1163 vh->sc_state == MASTER) {
1164 *enaddr = IF_LLADDR(vh->sc_ifp);
1176 carp_iamatch6(void *v, struct in6_addr *taddr)
1178 struct carp_if *cif = v;
1179 struct carp_softc *vh;
1183 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1184 TAILQ_FOREACH(ifa, &SC2IFP(vh)->if_addrlist, ifa_list) {
1185 if (IN6_ARE_ADDR_EQUAL(taddr,
1186 &ifatoia6(ifa)->ia_addr.sin6_addr) &&
1187 (SC2IFP(vh)->if_flags & IFF_UP) && (SC2IFP(vh)->if_flags & IFF_RUNNING) &&
1188 vh->sc_state == MASTER) {
1200 carp_macmatch6(void *v, struct mbuf *m, const struct in6_addr *taddr)
1203 struct carp_if *cif = v;
1204 struct carp_softc *sc;
1208 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list) {
1209 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
1210 if (IN6_ARE_ADDR_EQUAL(taddr,
1211 &ifatoia6(ifa)->ia_addr.sin6_addr) &&
1212 (SC2IFP(sc)->if_flags & IFF_UP) && (SC2IFP(sc)->if_flags & IFF_RUNNING)) {
1213 struct ifnet *ifp = SC2IFP(sc);
1214 mtag = m_tag_get(PACKET_TAG_CARP,
1215 sizeof(struct ifnet *), MB_DONTWAIT);
1217 /* better a bit than nothing */
1219 return (IF_LLADDR(sc->sc_ifp));
1221 bcopy(&ifp, (caddr_t)(mtag + 1),
1222 sizeof(struct ifnet *));
1223 m_tag_prepend(m, mtag);
1226 return (IF_LLADDR(sc->sc_ifp));
1237 carp_forus(void *v, void *dhost)
1239 struct carp_if *cif = v;
1240 struct carp_softc *vh;
1241 u_int8_t *ena = dhost;
1244 * XXX: See here for check on MAC adr is not for virtual use
1248 if (ena[0] || ena[1] || ena[2] != 0x5e || ena[3] || ena[4] != 1)
1254 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list)
1255 if ((SC2IFP(vh)->if_flags & IFF_UP) && (SC2IFP(vh)->if_flags & IFF_RUNNING) &&
1256 vh->sc_state == MASTER &&
1257 !bcmp(dhost, IF_LLADDR(vh->sc_ifp), ETHER_ADDR_LEN)) {
1259 return (SC2IFP(vh));
1267 carp_master_down(void *v)
1269 struct carp_softc *sc = v;
1271 lwkt_serialize_enter(sc->sc_ifp->if_serializer);
1272 carp_master_down_locked(sc);
1273 lwkt_serialize_exit(sc->sc_ifp->if_serializer);
1277 carp_master_down_locked(struct carp_softc *sc)
1280 CARP_SCLOCK_ASSERT(sc);
1282 switch (sc->sc_state) {
1284 kprintf("%s: master_down event in INIT state\n",
1285 SC2IFP(sc)->if_xname);
1290 carp_set_state(sc, MASTER);
1291 carp_send_ad_locked(sc);
1297 carp_setroute(sc, RTM_ADD);
1303 * When in backup state, af indicates whether to reset the master down timer
1304 * for v4 or v6. If it's set to zero, reset the ones which are already pending.
1307 carp_setrun(struct carp_softc *sc, sa_family_t af)
1311 if (sc->sc_carpdev == NULL) {
1312 SC2IFP(sc)->if_flags &= ~IFF_RUNNING;
1313 carp_set_state(sc, INIT);
1317 if (SC2IFP(sc)->if_flags & IFF_UP &&
1318 sc->sc_vhid > 0 && (sc->sc_naddrs || sc->sc_naddrs6))
1319 SC2IFP(sc)->if_flags |= IFF_RUNNING;
1321 SC2IFP(sc)->if_flags &= ~IFF_RUNNING;
1322 carp_setroute(sc, RTM_DELETE);
1326 switch (sc->sc_state) {
1328 if (carp_opts[CARPCTL_PREEMPT] && !carp_suppress_preempt) {
1329 carp_send_ad_locked(sc);
1334 CARP_DEBUG("%s: INIT -> MASTER (preempting)\n",
1335 SC2IFP(sc)->if_xname);
1336 carp_set_state(sc, MASTER);
1337 carp_setroute(sc, RTM_ADD);
1339 CARP_DEBUG("%s: INIT -> BACKUP\n", SC2IFP(sc)->if_xname);
1340 carp_set_state(sc, BACKUP);
1341 carp_setroute(sc, RTM_DELETE);
1346 callout_stop(&sc->sc_ad_tmo);
1347 tv.tv_sec = 3 * sc->sc_advbase;
1348 tv.tv_usec = sc->sc_advskew * 1000000 / 256;
1352 callout_reset(&sc->sc_md_tmo, tvtohz_high(&tv),
1353 carp_master_down, sc);
1358 callout_reset(&sc->sc_md6_tmo, tvtohz_high(&tv),
1359 carp_master_down, sc);
1364 callout_reset(&sc->sc_md_tmo, tvtohz_high(&tv),
1365 carp_master_down, sc);
1367 callout_reset(&sc->sc_md6_tmo, tvtohz_high(&tv),
1368 carp_master_down, sc);
1373 tv.tv_sec = sc->sc_advbase;
1374 tv.tv_usec = sc->sc_advskew * 1000000 / 256;
1375 callout_reset(&sc->sc_ad_tmo, tvtohz_high(&tv),
1382 carp_multicast_cleanup(struct carp_softc *sc)
1384 struct ip_moptions *imo = &sc->sc_imo;
1385 u_int16_t n = imo->imo_num_memberships;
1387 /* Clean up our own multicast memberships */
1389 if (imo->imo_membership[n] != NULL) {
1390 in_delmulti(imo->imo_membership[n]);
1391 imo->imo_membership[n] = NULL;
1394 imo->imo_num_memberships = 0;
1395 imo->imo_multicast_ifp = NULL;
1400 carp_multicast6_cleanup(struct carp_softc *sc)
1402 struct ip6_moptions *im6o = &sc->sc_im6o;
1404 while (!LIST_EMPTY(&im6o->im6o_memberships)) {
1405 struct in6_multi_mship *imm =
1406 LIST_FIRST(&im6o->im6o_memberships);
1408 LIST_REMOVE(imm, i6mm_chain);
1409 in6_leavegroup(imm);
1411 im6o->im6o_multicast_ifp = NULL;
1416 carp_set_addr(struct carp_softc *sc, struct sockaddr_in *sin)
1419 struct carp_if *cif;
1420 struct in_ifaddr *ia, *ia_if;
1421 struct ip_moptions *imo = &sc->sc_imo;
1422 struct in_addr addr;
1423 u_long iaddr = htonl(sin->sin_addr.s_addr);
1426 if (sin->sin_addr.s_addr == 0)
1428 if (!(SC2IFP(sc)->if_flags & IFF_UP))
1430 carp_set_state(sc, INIT);
1434 SC2IFP(sc)->if_flags |= IFF_UP;
1439 /* we have to do it by hands to check we won't match on us */
1440 ia_if = NULL; own = 0;
1441 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
1442 /* and, yeah, we need a multicast-capable iface too */
1443 if (ia->ia_ifp != SC2IFP(sc) &&
1444 (ia->ia_ifp->if_flags & IFF_MULTICAST) &&
1445 (iaddr & ia->ia_subnetmask) == ia->ia_subnet) {
1448 if (sin->sin_addr.s_addr ==
1449 ia->ia_addr.sin_addr.s_addr)
1456 return (EADDRNOTAVAIL);
1461 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0 ||
1462 (imo->imo_multicast_ifp && imo->imo_multicast_ifp != ifp))
1463 return (EADDRNOTAVAIL);
1465 if (imo->imo_num_memberships == 0) {
1466 addr.s_addr = htonl(INADDR_CARP_GROUP);
1467 if ((imo->imo_membership[0] = in_addmulti(&addr, ifp)) == NULL)
1469 imo->imo_num_memberships++;
1470 imo->imo_multicast_ifp = ifp;
1471 imo->imo_multicast_ttl = CARP_DFLTTL;
1472 imo->imo_multicast_loop = 0;
1475 if (!ifp->if_carp) {
1477 MALLOC(cif, struct carp_if *, sizeof(*cif), M_CARP,
1483 if ((error = ifpromisc(ifp, 1))) {
1488 CARP_LOCK_INIT(cif);
1490 cif->vhif_ifp = ifp;
1491 TAILQ_INIT(&cif->vhif_vrs);
1495 struct carp_softc *vr;
1497 cif = (struct carp_if *)ifp->if_carp;
1499 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list)
1500 if (vr != sc && vr->sc_vhid == sc->sc_vhid) {
1507 sc->sc_carpdev = ifp;
1509 { /* XXX prevent endless loop if already in queue */
1510 struct carp_softc *vr, *after = NULL;
1512 cif = (struct carp_if *)ifp->if_carp;
1514 /* XXX: cif should not change, right? So we still hold the lock */
1515 CARP_LOCK_ASSERT(cif);
1517 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) {
1520 if (vr->sc_vhid < sc->sc_vhid)
1525 /* We're trying to keep things in order */
1526 if (after == NULL) {
1527 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list);
1529 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, sc, sc_list);
1536 SC2IFP(sc)->if_flags |= IFF_UP;
1541 carp_sc_state_locked(sc);
1549 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]);
1555 carp_del_addr(struct carp_softc *sc, struct sockaddr_in *sin)
1559 if (!--sc->sc_naddrs) {
1560 struct carp_if *cif = (struct carp_if *)sc->sc_carpdev->if_carp;
1561 struct ip_moptions *imo = &sc->sc_imo;
1564 callout_stop(&sc->sc_ad_tmo);
1565 SC2IFP(sc)->if_flags &= ~IFF_UP;
1566 SC2IFP(sc)->if_flags &= ~IFF_RUNNING;
1568 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]);
1569 imo->imo_multicast_ifp = NULL;
1570 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list);
1571 if (!--cif->vhif_nvrs) {
1572 sc->sc_carpdev->if_carp = NULL;
1573 CARP_LOCK_DESTROY(cif);
1574 FREE(cif, M_IFADDR);
1585 carp_set_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6)
1588 struct carp_if *cif;
1589 struct in6_ifaddr *ia, *ia_if;
1590 struct ip6_moptions *im6o = &sc->sc_im6o;
1591 struct in6_multi_mship *imm;
1592 struct in6_addr in6;
1595 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1596 if (!(SC2IFP(sc)->if_flags & IFF_UP))
1597 carp_set_state(sc, INIT);
1599 SC2IFP(sc)->if_flags |= IFF_UP;
1604 /* we have to do it by hands to check we won't match on us */
1605 ia_if = NULL; own = 0;
1606 for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1609 for (i = 0; i < 4; i++) {
1610 if ((sin6->sin6_addr.s6_addr32[i] &
1611 ia->ia_prefixmask.sin6_addr.s6_addr32[i]) !=
1612 (ia->ia_addr.sin6_addr.s6_addr32[i] &
1613 ia->ia_prefixmask.sin6_addr.s6_addr32[i]))
1616 /* and, yeah, we need a multicast-capable iface too */
1617 if (ia->ia_ifp != SC2IFP(sc) &&
1618 (ia->ia_ifp->if_flags & IFF_MULTICAST) &&
1622 if (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
1623 &ia->ia_addr.sin6_addr))
1629 return (EADDRNOTAVAIL);
1633 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0 ||
1634 (im6o->im6o_multicast_ifp && im6o->im6o_multicast_ifp != ifp))
1635 return (EADDRNOTAVAIL);
1637 if (!sc->sc_naddrs6) {
1638 im6o->im6o_multicast_ifp = ifp;
1640 /* join CARP multicast address */
1641 bzero(&in6, sizeof(in6));
1642 in6.s6_addr16[0] = htons(0xff02);
1643 in6.s6_addr8[15] = 0x12;
1644 if (in6_setscope(&in6, ifp, NULL) != 0)
1646 if ((imm = in6_joingroup(ifp, &in6, &error)) == NULL)
1648 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1650 /* join solicited multicast address */
1651 bzero(&in6, sizeof(in6));
1652 in6.s6_addr16[0] = htons(0xff02);
1653 in6.s6_addr32[1] = 0;
1654 in6.s6_addr32[2] = htonl(1);
1655 in6.s6_addr32[3] = sin6->sin6_addr.s6_addr32[3];
1656 in6.s6_addr8[12] = 0xff;
1657 if (in6_setscope(&in6, ifp, NULL) != 0)
1659 if ((imm = in6_joingroup(ifp, &in6, &error)) == NULL)
1661 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1664 if (!ifp->if_carp) {
1665 MALLOC(cif, struct carp_if *, sizeof(*cif), M_CARP,
1671 if ((error = ifpromisc(ifp, 1))) {
1676 CARP_LOCK_INIT(cif);
1678 cif->vhif_ifp = ifp;
1679 TAILQ_INIT(&cif->vhif_vrs);
1683 struct carp_softc *vr;
1685 cif = (struct carp_if *)ifp->if_carp;
1687 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list)
1688 if (vr != sc && vr->sc_vhid == sc->sc_vhid) {
1695 sc->sc_carpdev = ifp;
1697 { /* XXX prevent endless loop if already in queue */
1698 struct carp_softc *vr, *after = NULL;
1700 cif = (struct carp_if *)ifp->if_carp;
1701 CARP_LOCK_ASSERT(cif);
1703 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) {
1706 if (vr->sc_vhid < sc->sc_vhid)
1711 /* We're trying to keep things in order */
1712 if (after == NULL) {
1713 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list);
1715 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, sc, sc_list);
1722 SC2IFP(sc)->if_flags |= IFF_UP;
1725 carp_sc_state_locked(sc);
1733 /* clean up multicast memberships */
1734 if (!sc->sc_naddrs6) {
1735 while (!LIST_EMPTY(&im6o->im6o_memberships)) {
1736 imm = LIST_FIRST(&im6o->im6o_memberships);
1737 LIST_REMOVE(imm, i6mm_chain);
1738 in6_leavegroup(imm);
1745 carp_del_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6)
1749 if (!--sc->sc_naddrs6) {
1750 struct carp_if *cif = (struct carp_if *)sc->sc_carpdev->if_carp;
1751 struct ip6_moptions *im6o = &sc->sc_im6o;
1754 callout_stop(&sc->sc_ad_tmo);
1755 SC2IFP(sc)->if_flags &= ~IFF_UP;
1756 SC2IFP(sc)->if_flags &= ~IFF_RUNNING;
1758 while (!LIST_EMPTY(&im6o->im6o_memberships)) {
1759 struct in6_multi_mship *imm =
1760 LIST_FIRST(&im6o->im6o_memberships);
1762 LIST_REMOVE(imm, i6mm_chain);
1763 in6_leavegroup(imm);
1765 im6o->im6o_multicast_ifp = NULL;
1766 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list);
1767 if (!--cif->vhif_nvrs) {
1768 CARP_LOCK_DESTROY(cif);
1769 sc->sc_carpdev->if_carp = NULL;
1770 FREE(cif, M_IFADDR);
1780 carp_ioctl(struct ifnet *ifp, u_long cmd, caddr_t addr, struct ucred *creds)
1782 struct carp_softc *sc = ifp->if_softc, *vr;
1783 struct carpreq carpr;
1786 struct ifaliasreq *ifra;
1787 int locked = 0, error = 0;
1789 ifa = (struct ifaddr *)addr;
1790 ifra = (struct ifaliasreq *)addr;
1791 ifr = (struct ifreq *)addr;
1796 switch (ifa->ifa_addr->sa_family) {
1799 SC2IFP(sc)->if_flags |= IFF_UP;
1800 bcopy(ifa->ifa_addr, ifa->ifa_dstaddr,
1801 sizeof(struct sockaddr));
1802 error = carp_set_addr(sc, satosin(ifa->ifa_addr));
1807 SC2IFP(sc)->if_flags |= IFF_UP;
1808 error = carp_set_addr6(sc, satosin6(ifa->ifa_addr));
1812 error = EAFNOSUPPORT;
1818 switch (ifa->ifa_addr->sa_family) {
1821 SC2IFP(sc)->if_flags |= IFF_UP;
1822 bcopy(ifa->ifa_addr, ifa->ifa_dstaddr,
1823 sizeof(struct sockaddr));
1824 error = carp_set_addr(sc, satosin(&ifra->ifra_addr));
1829 SC2IFP(sc)->if_flags |= IFF_UP;
1830 error = carp_set_addr6(sc, satosin6(&ifra->ifra_addr));
1834 error = EAFNOSUPPORT;
1840 switch (ifa->ifa_addr->sa_family) {
1843 error = carp_del_addr(sc, satosin(&ifra->ifra_addr));
1848 error = carp_del_addr6(sc, satosin6(&ifra->ifra_addr));
1852 error = EAFNOSUPPORT;
1858 if (sc->sc_carpdev) {
1862 if (sc->sc_state != INIT && !(ifr->ifr_flags & IFF_UP)) {
1863 callout_stop(&sc->sc_ad_tmo);
1864 callout_stop(&sc->sc_md_tmo);
1865 callout_stop(&sc->sc_md6_tmo);
1866 if (sc->sc_state == MASTER)
1867 carp_send_ad_locked(sc);
1868 carp_set_state(sc, INIT);
1870 } else if (sc->sc_state == INIT && (ifr->ifr_flags & IFF_UP)) {
1871 SC2IFP(sc)->if_flags |= IFF_UP;
1877 error = suser(curthread);
1880 if ((error = copyin(ifr->ifr_data, &carpr, sizeof carpr)))
1883 if (sc->sc_carpdev) {
1887 if (sc->sc_state != INIT && carpr.carpr_state != sc->sc_state) {
1888 switch (carpr.carpr_state) {
1890 callout_stop(&sc->sc_ad_tmo);
1891 carp_set_state(sc, BACKUP);
1893 carp_setroute(sc, RTM_DELETE);
1896 carp_master_down_locked(sc);
1902 if (carpr.carpr_vhid > 0) {
1903 if (carpr.carpr_vhid > 255) {
1907 if (sc->sc_carpdev) {
1908 struct carp_if *cif;
1909 cif = (struct carp_if *)sc->sc_carpdev->if_carp;
1910 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list)
1912 vr->sc_vhid == carpr.carpr_vhid)
1915 sc->sc_vhid = carpr.carpr_vhid;
1916 IF_LLADDR(sc->sc_ifp)[0] = 0;
1917 IF_LLADDR(sc->sc_ifp)[1] = 0;
1918 IF_LLADDR(sc->sc_ifp)[2] = 0x5e;
1919 IF_LLADDR(sc->sc_ifp)[3] = 0;
1920 IF_LLADDR(sc->sc_ifp)[4] = 1;
1921 IF_LLADDR(sc->sc_ifp)[5] = sc->sc_vhid;
1924 if (carpr.carpr_advbase > 0 || carpr.carpr_advskew > 0) {
1925 if (carpr.carpr_advskew >= 255) {
1929 if (carpr.carpr_advbase > 255) {
1933 sc->sc_advbase = carpr.carpr_advbase;
1934 sc->sc_advskew = carpr.carpr_advskew;
1937 bcopy(carpr.carpr_key, sc->sc_key, sizeof(sc->sc_key));
1947 /* XXX: lockless read */
1948 bzero(&carpr, sizeof(carpr));
1949 carpr.carpr_state = sc->sc_state;
1950 carpr.carpr_vhid = sc->sc_vhid;
1951 carpr.carpr_advbase = sc->sc_advbase;
1952 carpr.carpr_advskew = sc->sc_advskew;
1953 error = suser(curthread);
1955 bcopy(sc->sc_key, carpr.carpr_key,
1956 sizeof(carpr.carpr_key));
1957 error = copyout(&carpr, ifr->ifr_data, sizeof(carpr));
1967 carp_hmac_prepare(sc);
1973 * XXX: this is looutput. We should eventually use it from there.
1976 carp_looutput(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
1981 M_ASSERTPKTHDR(m); /* check if we have the packet header */
1983 if (rt && rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
1985 return (rt->rt_flags & RTF_BLACKHOLE ? 0 :
1986 rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
1990 ifp->if_obytes += m->m_pkthdr.len;
1992 /* BPF writes need to be handled specially. */
1993 if (dst->sa_family == AF_UNSPEC) {
1994 bcopy(dst->sa_data, &af, sizeof(af));
1995 dst->sa_family = af;
1999 switch (dst->sa_family) {
2007 return (EAFNOSUPPORT);
2010 return(if_simloop(ifp, m, dst->sa_family, 0));
2014 * Start output on carp interface. This function should never be called.
2017 carp_start(struct ifnet *ifp)
2020 kprintf("%s: start called\n", ifp->if_xname);
2025 carp_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2029 struct carp_softc *sc;
2030 struct ifnet *carp_ifp;
2035 switch (sa->sa_family) {
2048 mtag = m_tag_find(m, PACKET_TAG_CARP, NULL);
2052 bcopy(mtag + 1, &carp_ifp, sizeof(struct ifnet *));
2053 sc = carp_ifp->if_softc;
2055 /* Set the source MAC address to Virtual Router MAC Address */
2056 switch (ifp->if_type) {
2059 struct ether_header *eh;
2061 eh = mtod(m, struct ether_header *);
2062 eh->ether_shost[0] = 0;
2063 eh->ether_shost[1] = 0;
2064 eh->ether_shost[2] = 0x5e;
2065 eh->ether_shost[3] = 0;
2066 eh->ether_shost[4] = 1;
2067 eh->ether_shost[5] = sc->sc_vhid;
2071 kprintf("%s: carp is not supported for this interface type\n",
2073 return (EOPNOTSUPP);
2081 carp_set_state(struct carp_softc *sc, int state)
2085 CARP_SCLOCK_ASSERT(sc);
2087 if (sc->sc_state == state)
2090 sc->sc_state = state;
2093 SC2IFP(sc)->if_link_state = LINK_STATE_DOWN;
2096 SC2IFP(sc)->if_link_state = LINK_STATE_UP;
2099 SC2IFP(sc)->if_link_state = LINK_STATE_UNKNOWN;
2102 rt_ifmsg(SC2IFP(sc));
2106 carp_carpdev_state(void *v)
2108 struct carp_if *cif = v;
2111 carp_carpdev_state_locked(cif);
2116 carp_carpdev_state_locked(struct carp_if *cif)
2118 struct carp_softc *sc;
2120 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list)
2121 carp_sc_state_locked(sc);
2125 carp_sc_state_locked(struct carp_softc *sc)
2127 CARP_SCLOCK_ASSERT(sc);
2129 if ( !(sc->sc_carpdev->if_flags & IFF_UP)) {
2130 sc->sc_flags_backup = SC2IFP(sc)->if_flags;
2131 SC2IFP(sc)->if_flags &= ~IFF_UP;
2132 SC2IFP(sc)->if_flags &= ~IFF_RUNNING;
2133 callout_stop(&sc->sc_ad_tmo);
2134 callout_stop(&sc->sc_md_tmo);
2135 callout_stop(&sc->sc_md6_tmo);
2136 carp_set_state(sc, INIT);
2138 if (!sc->sc_suppress) {
2139 carp_suppress_preempt++;
2140 if (carp_suppress_preempt == 1) {
2146 sc->sc_suppress = 1;
2148 SC2IFP(sc)->if_flags |= sc->sc_flags_backup;
2149 carp_set_state(sc, INIT);
2151 if (sc->sc_suppress)
2152 carp_suppress_preempt--;
2153 sc->sc_suppress = 0;
2160 carp_modevent(module_t mod, int type, void *data)
2164 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2165 carp_ifdetach, NULL, EVENTHANDLER_PRI_ANY);
2166 if (if_detach_event_tag == NULL)
2169 LIST_INIT(&carpif_list);
2170 if_clone_attach(&carp_cloner);
2174 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2175 if_clone_detach(&carp_cloner);
2185 static moduledata_t carp_mod = {
2191 DECLARE_MODULE(carp, carp_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);