4 * Copyright (c) 2002 Michael Shalayeff. All rights reserved.
5 * Copyright (c) 2003 Ryan McBride. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
17 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
18 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
19 * IN NO EVENT SHALL THE AUTHOR OR HIS RELATIVES BE LIABLE FOR ANY DIRECT,
20 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
21 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
22 * SERVICES; LOSS OF MIND, USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
24 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
25 * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
26 * THE POSSIBILITY OF SUCH DAMAGE.
32 #include "opt_inet6.h"
34 #include <sys/types.h>
35 #include <sys/param.h>
36 #include <sys/systm.h>
38 #include <sys/kernel.h>
39 #include <sys/limits.h>
40 #include <sys/malloc.h>
42 #include <sys/module.h>
46 #include <sys/sysctl.h>
47 #include <sys/syslog.h>
48 #include <sys/signalvar.h>
49 #include <sys/filio.h>
50 #include <sys/sockio.h>
52 #include <sys/socket.h>
53 #include <sys/vnode.h>
55 #include <machine/stdarg.h>
58 #include <net/ethernet.h>
60 #include <net/iso88025.h>
62 #include <net/if_clone.h>
63 #include <net/if_dl.h>
64 #include <net/if_types.h>
65 #include <net/route.h>
68 #include <netinet/in.h>
69 #include <netinet/in_var.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/ip.h>
72 #include <netinet/ip_var.h>
73 #include <netinet/if_ether.h>
74 #include <machine/in_cksum.h>
78 #include <netinet/icmp6.h>
79 #include <netinet/ip6.h>
80 #include <netinet6/ip6_var.h>
81 #include <netinet6/scope6_var.h>
82 #include <netinet6/nd6.h>
85 #include <crypto/sha1.h>
86 #include <netinet/ip_carp.h>
88 #define CARP_IFNAME "carp"
89 static MALLOC_DEFINE(M_CARP, "CARP", "CARP interfaces");
90 SYSCTL_DECL(_net_inet_carp);
93 struct ifnet *sc_ifp; /* Interface clue */
94 struct ifnet *sc_carpdev; /* Pointer to parent interface */
95 struct in_ifaddr *sc_ia; /* primary iface address */
96 struct ip_moptions sc_imo;
98 struct in6_ifaddr *sc_ia6; /* primary iface address v6 */
99 struct ip6_moptions sc_im6o;
101 TAILQ_ENTRY(carp_softc) sc_list;
103 enum { INIT = 0, BACKUP, MASTER } sc_state;
108 int sc_sendad_errors;
109 #define CARP_SENDAD_MAX_ERRORS 3
110 int sc_sendad_success;
111 #define CARP_SENDAD_MIN_SUCCESS 3
117 int sc_advbase; /* seconds */
119 u_int64_t sc_counter;
122 #define CARP_HMAC_PAD 64
123 unsigned char sc_key[CARP_KEY_LEN];
124 unsigned char sc_pad[CARP_HMAC_PAD];
127 struct callout sc_ad_tmo; /* advertisement timeout */
128 struct callout sc_md_tmo; /* master down timeout */
129 struct callout sc_md6_tmo; /* master down timeout */
131 LIST_ENTRY(carp_softc) sc_next; /* Interface clue */
133 #define SC2IFP(sc) ((sc)->sc_ifp)
135 int carp_suppress_preempt = 0;
136 int carp_opts[CARPCTL_MAXID] = { 0, 1, 0, 1, 0, 0 }; /* XXX for now */
137 SYSCTL_INT(_net_inet_carp, CARPCTL_ALLOW, allow, CTLFLAG_RW,
138 &carp_opts[CARPCTL_ALLOW], 0, "Accept incoming CARP packets");
139 SYSCTL_INT(_net_inet_carp, CARPCTL_PREEMPT, preempt, CTLFLAG_RW,
140 &carp_opts[CARPCTL_PREEMPT], 0, "high-priority backup preemption mode");
141 SYSCTL_INT(_net_inet_carp, CARPCTL_LOG, log, CTLFLAG_RW,
142 &carp_opts[CARPCTL_LOG], 0, "log bad carp packets");
143 SYSCTL_INT(_net_inet_carp, CARPCTL_ARPBALANCE, arpbalance, CTLFLAG_RW,
144 &carp_opts[CARPCTL_ARPBALANCE], 0, "balance arp responses");
145 SYSCTL_INT(_net_inet_carp, OID_AUTO, suppress_preempt, CTLFLAG_RD,
146 &carp_suppress_preempt, 0, "Preemption is suppressed");
148 struct carpstats carpstats;
149 SYSCTL_STRUCT(_net_inet_carp, CARPCTL_STATS, stats, CTLFLAG_RW,
150 &carpstats, carpstats,
151 "CARP statistics (struct carpstats, netinet/ip_carp.h)");
154 TAILQ_HEAD(, carp_softc) vhif_vrs;
157 struct ifnet *vhif_ifp;
161 /* Get carp_if from softc. Valid after carp_set_addr{,6}. */
162 #define SC2CIF(sc) ((struct carp_if *)(sc)->sc_carpdev->if_carp)
164 /* lock per carp_if queue */
165 #define CARP_LOCK_INIT(cif) mtx_init(&(cif)->vhif_mtx, "carp_if", \
167 #define CARP_LOCK_DESTROY(cif) mtx_destroy(&(cif)->vhif_mtx)
168 #define CARP_LOCK_ASSERT(cif) mtx_assert(&(cif)->vhif_mtx, MA_OWNED)
169 #define CARP_LOCK(cif) mtx_lock(&(cif)->vhif_mtx)
170 #define CARP_UNLOCK(cif) mtx_unlock(&(cif)->vhif_mtx)
172 #define CARP_SCLOCK(sc) mtx_lock(&SC2CIF(sc)->vhif_mtx)
173 #define CARP_SCUNLOCK(sc) mtx_unlock(&SC2CIF(sc)->vhif_mtx)
174 #define CARP_SCLOCK_ASSERT(sc) mtx_assert(&SC2CIF(sc)->vhif_mtx, MA_OWNED)
176 #define CARP_LOG(...) do { \
177 if (carp_opts[CARPCTL_LOG] > 0) \
178 log(LOG_INFO, __VA_ARGS__); \
181 #define CARP_DEBUG(...) do { \
182 if (carp_opts[CARPCTL_LOG] > 1) \
183 log(LOG_DEBUG, __VA_ARGS__); \
186 static void carp_hmac_prepare(struct carp_softc *);
187 static void carp_hmac_generate(struct carp_softc *, u_int32_t *,
189 static int carp_hmac_verify(struct carp_softc *, u_int32_t *,
191 static void carp_setroute(struct carp_softc *, int);
192 static void carp_input_c(struct mbuf *, struct carp_header *, sa_family_t);
193 static int carp_clone_create(struct if_clone *, int, caddr_t);
194 static void carp_clone_destroy(struct ifnet *);
195 static void carpdetach(struct carp_softc *, int);
196 static int carp_prepare_ad(struct mbuf *, struct carp_softc *,
197 struct carp_header *);
198 static void carp_send_ad_all(void);
199 static void carp_send_ad(void *);
200 static void carp_send_ad_locked(struct carp_softc *);
201 static void carp_send_arp(struct carp_softc *);
202 static void carp_master_down(void *);
203 static void carp_master_down_locked(struct carp_softc *);
204 static int carp_ioctl(struct ifnet *, u_long, caddr_t);
205 static int carp_looutput(struct ifnet *, struct mbuf *, struct sockaddr *,
207 static void carp_start(struct ifnet *);
208 static void carp_setrun(struct carp_softc *, sa_family_t);
209 static void carp_set_state(struct carp_softc *, int);
210 static int carp_addrcount(struct carp_if *, struct in_ifaddr *, int);
211 enum { CARP_COUNT_MASTER, CARP_COUNT_RUNNING };
213 static void carp_multicast_cleanup(struct carp_softc *);
214 static int carp_set_addr(struct carp_softc *, struct sockaddr_in *);
215 static int carp_del_addr(struct carp_softc *, struct sockaddr_in *);
216 static void carp_carpdev_state_locked(struct carp_if *);
217 static void carp_sc_state_locked(struct carp_softc *);
219 static void carp_send_na(struct carp_softc *);
220 static int carp_set_addr6(struct carp_softc *, struct sockaddr_in6 *);
221 static int carp_del_addr6(struct carp_softc *, struct sockaddr_in6 *);
222 static void carp_multicast6_cleanup(struct carp_softc *);
225 static LIST_HEAD(, carp_softc) carpif_list;
226 static struct mtx carp_mtx;
227 IFC_SIMPLE_DECLARE(carp, 0);
229 static eventhandler_tag if_detach_event_tag;
231 static __inline u_int16_t
232 carp_cksum(struct mbuf *m, int len)
234 return (in_cksum(m, len));
238 carp_hmac_prepare(struct carp_softc *sc)
240 u_int8_t version = CARP_VERSION, type = CARP_ADVERTISEMENT;
241 u_int8_t vhid = sc->sc_vhid & 0xff;
251 /* XXX: possible race here */
253 /* compute ipad from key */
254 bzero(sc->sc_pad, sizeof(sc->sc_pad));
255 bcopy(sc->sc_key, sc->sc_pad, sizeof(sc->sc_key));
256 for (i = 0; i < sizeof(sc->sc_pad); i++)
257 sc->sc_pad[i] ^= 0x36;
259 /* precompute first part of inner hash */
260 SHA1Init(&sc->sc_sha1);
261 SHA1Update(&sc->sc_sha1, sc->sc_pad, sizeof(sc->sc_pad));
262 SHA1Update(&sc->sc_sha1, (void *)&version, sizeof(version));
263 SHA1Update(&sc->sc_sha1, (void *)&type, sizeof(type));
264 SHA1Update(&sc->sc_sha1, (void *)&vhid, sizeof(vhid));
266 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
267 if (ifa->ifa_addr->sa_family == AF_INET)
268 SHA1Update(&sc->sc_sha1,
269 (void *)&ifatoia(ifa)->ia_addr.sin_addr.s_addr,
270 sizeof(struct in_addr));
274 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
275 if (ifa->ifa_addr->sa_family == AF_INET6) {
276 in6 = ifatoia6(ifa)->ia_addr.sin6_addr;
277 in6_clearscope(&in6);
278 SHA1Update(&sc->sc_sha1, (void *)&in6, sizeof(in6));
283 /* convert ipad to opad */
284 for (i = 0; i < sizeof(sc->sc_pad); i++)
285 sc->sc_pad[i] ^= 0x36 ^ 0x5c;
292 carp_hmac_generate(struct carp_softc *sc, u_int32_t counter[2],
293 unsigned char md[20])
297 /* fetch first half of inner hash */
298 bcopy(&sc->sc_sha1, &sha1ctx, sizeof(sha1ctx));
300 SHA1Update(&sha1ctx, (void *)counter, sizeof(sc->sc_counter));
301 SHA1Final(md, &sha1ctx);
305 SHA1Update(&sha1ctx, sc->sc_pad, sizeof(sc->sc_pad));
306 SHA1Update(&sha1ctx, md, 20);
307 SHA1Final(md, &sha1ctx);
311 carp_hmac_verify(struct carp_softc *sc, u_int32_t counter[2],
312 unsigned char md[20])
314 unsigned char md2[20];
316 CARP_SCLOCK_ASSERT(sc);
318 carp_hmac_generate(sc, counter, md2);
320 return (bcmp(md, md2, sizeof(md2)));
324 carp_setroute(struct carp_softc *sc, int cmd)
330 CARP_SCLOCK_ASSERT(sc);
333 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
334 if (ifa->ifa_addr->sa_family == AF_INET &&
335 sc->sc_carpdev != NULL) {
336 int count = carp_addrcount(
337 (struct carp_if *)sc->sc_carpdev->if_carp,
338 ifatoia(ifa), CARP_COUNT_MASTER);
340 if ((cmd == RTM_ADD && count == 1) ||
341 (cmd == RTM_DELETE && count == 0))
342 rtinit(ifa, cmd, RTF_UP | RTF_HOST);
345 if (ifa->ifa_addr->sa_family == AF_INET6) {
357 carp_clone_create(struct if_clone *ifc, int unit, caddr_t params)
360 struct carp_softc *sc;
363 MALLOC(sc, struct carp_softc *, sizeof(*sc), M_CARP, M_WAITOK|M_ZERO);
364 ifp = SC2IFP(sc) = if_alloc(IFT_ETHER);
370 sc->sc_flags_backup = 0;
372 sc->sc_advbase = CARP_DFLTINTV;
373 sc->sc_vhid = -1; /* required setting */
375 sc->sc_init_counter = 1;
376 sc->sc_naddrs = sc->sc_naddrs6 = 0; /* M_ZERO? */
378 sc->sc_im6o.im6o_multicast_hlim = CARP_DFLTTL;
380 sc->sc_imo.imo_membership = (struct in_multi **)malloc(
381 (sizeof(struct in_multi *) * IP_MIN_MEMBERSHIPS), M_CARP,
383 sc->sc_imo.imo_mfilters = NULL;
384 sc->sc_imo.imo_max_memberships = IP_MIN_MEMBERSHIPS;
385 sc->sc_imo.imo_multicast_vif = -1;
387 callout_init(&sc->sc_ad_tmo, NET_CALLOUT_MPSAFE);
388 callout_init(&sc->sc_md_tmo, NET_CALLOUT_MPSAFE);
389 callout_init(&sc->sc_md6_tmo, NET_CALLOUT_MPSAFE);
392 if_initname(ifp, CARP_IFNAME, unit);
393 ifp->if_mtu = ETHERMTU;
394 ifp->if_flags = IFF_LOOPBACK;
395 ifp->if_ioctl = carp_ioctl;
396 ifp->if_output = carp_looutput;
397 ifp->if_start = carp_start;
398 ifp->if_type = IFT_CARP;
399 ifp->if_snd.ifq_maxlen = ifqmaxlen;
402 bpfattach(SC2IFP(sc), DLT_NULL, sizeof(u_int32_t));
404 LIST_INSERT_HEAD(&carpif_list, sc, sc_next);
405 mtx_unlock(&carp_mtx);
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);
420 mtx_unlock(&carp_mtx);
423 if_free_type(ifp, IFT_ETHER);
424 free(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);
500 * process input packet.
501 * we have rearranged checks order compared to the rfc,
502 * but it seems more efficient this way or not possible otherwise.
505 carp_input(struct mbuf *m, int hlen)
507 struct ip *ip = mtod(m, struct ip *);
508 struct carp_header *ch;
511 carpstats.carps_ipackets++;
513 if (!carp_opts[CARPCTL_ALLOW]) {
518 /* check if received on a valid carp interface */
519 if (m->m_pkthdr.rcvif->if_carp == NULL) {
520 carpstats.carps_badif++;
521 CARP_LOG("carp_input: packet received on non-carp "
523 m->m_pkthdr.rcvif->if_xname);
528 /* verify that the IP TTL is 255. */
529 if (ip->ip_ttl != CARP_DFLTTL) {
530 carpstats.carps_badttl++;
531 CARP_LOG("carp_input: received ttl %d != 255i on %s\n",
533 m->m_pkthdr.rcvif->if_xname);
538 iplen = ip->ip_hl << 2;
540 if (m->m_pkthdr.len < iplen + sizeof(*ch)) {
541 carpstats.carps_badlen++;
542 CARP_LOG("carp_input: received len %zd < "
543 "sizeof(struct carp_header)\n",
544 m->m_len - sizeof(struct ip));
549 if (iplen + sizeof(*ch) < m->m_len) {
550 if ((m = m_pullup(m, iplen + sizeof(*ch))) == NULL) {
551 carpstats.carps_hdrops++;
552 CARP_LOG("carp_input: pullup failed\n");
555 ip = mtod(m, struct ip *);
557 ch = (struct carp_header *)((char *)ip + iplen);
560 * verify that the received packet length is
561 * equal to the CARP header
563 len = iplen + sizeof(*ch);
564 if (len > m->m_pkthdr.len) {
565 carpstats.carps_badlen++;
566 CARP_LOG("carp_input: packet too short %d on %s\n",
568 m->m_pkthdr.rcvif->if_xname);
573 if ((m = m_pullup(m, len)) == NULL) {
574 carpstats.carps_hdrops++;
577 ip = mtod(m, struct ip *);
578 ch = (struct carp_header *)((char *)ip + iplen);
580 /* verify the CARP checksum */
582 if (carp_cksum(m, len - iplen)) {
583 carpstats.carps_badsum++;
584 CARP_LOG("carp_input: checksum failed on %s\n",
585 m->m_pkthdr.rcvif->if_xname);
591 carp_input_c(m, ch, AF_INET);
596 carp6_input(struct mbuf **mp, int *offp, int proto)
598 struct mbuf *m = *mp;
599 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
600 struct carp_header *ch;
603 carpstats.carps_ipackets6++;
605 if (!carp_opts[CARPCTL_ALLOW]) {
607 return (IPPROTO_DONE);
610 /* check if received on a valid carp interface */
611 if (m->m_pkthdr.rcvif->if_carp == NULL) {
612 carpstats.carps_badif++;
613 CARP_LOG("carp6_input: packet received on non-carp "
615 m->m_pkthdr.rcvif->if_xname);
617 return (IPPROTO_DONE);
620 /* verify that the IP TTL is 255 */
621 if (ip6->ip6_hlim != CARP_DFLTTL) {
622 carpstats.carps_badttl++;
623 CARP_LOG("carp6_input: received ttl %d != 255 on %s\n",
625 m->m_pkthdr.rcvif->if_xname);
627 return (IPPROTO_DONE);
630 /* verify that we have a complete carp packet */
632 IP6_EXTHDR_GET(ch, struct carp_header *, m, *offp, sizeof(*ch));
634 carpstats.carps_badlen++;
635 CARP_LOG("carp6_input: packet size %u too small\n", len);
636 return (IPPROTO_DONE);
640 /* verify the CARP checksum */
642 if (carp_cksum(m, sizeof(*ch))) {
643 carpstats.carps_badsum++;
644 CARP_LOG("carp6_input: checksum failed, on %s\n",
645 m->m_pkthdr.rcvif->if_xname);
647 return (IPPROTO_DONE);
651 carp_input_c(m, ch, AF_INET6);
652 return (IPPROTO_DONE);
657 carp_input_c(struct mbuf *m, struct carp_header *ch, sa_family_t af)
659 struct ifnet *ifp = m->m_pkthdr.rcvif;
660 struct carp_softc *sc;
661 u_int64_t tmp_counter;
662 struct timeval sc_tv, ch_tv;
664 /* verify that the VHID is valid on the receiving interface */
665 CARP_LOCK(ifp->if_carp);
666 TAILQ_FOREACH(sc, &((struct carp_if *)ifp->if_carp)->vhif_vrs, sc_list)
667 if (sc->sc_vhid == ch->carp_vhid)
670 if (!sc || !((SC2IFP(sc)->if_flags & IFF_UP) &&
671 (SC2IFP(sc)->if_drv_flags & IFF_DRV_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 (bpf_peers_present(SC2IFP(sc)->if_bpf)) {
683 struct ip *ip = mtod(m, struct ip *);
686 /* BPF wants net byte order */
687 ip->ip_len = htons(ip->ip_len + (ip->ip_hl << 2));
688 ip->ip_off = htons(ip->ip_off);
689 bpf_mtap2(SC2IFP(sc)->if_bpf, &af1, sizeof(af1), m);
692 /* verify the CARP version. */
693 if (ch->carp_version != CARP_VERSION) {
694 carpstats.carps_badver++;
695 SC2IFP(sc)->if_ierrors++;
696 CARP_UNLOCK(ifp->if_carp);
697 CARP_LOG("%s; invalid version %d\n",
698 SC2IFP(sc)->if_xname,
704 /* verify the hash */
705 if (carp_hmac_verify(sc, ch->carp_counter, ch->carp_md)) {
706 carpstats.carps_badauth++;
707 SC2IFP(sc)->if_ierrors++;
708 CARP_UNLOCK(ifp->if_carp);
709 CARP_LOG("%s: incorrect hash\n", SC2IFP(sc)->if_xname);
714 tmp_counter = ntohl(ch->carp_counter[0]);
715 tmp_counter = tmp_counter<<32;
716 tmp_counter += ntohl(ch->carp_counter[1]);
718 /* XXX Replay protection goes here */
720 sc->sc_init_counter = 0;
721 sc->sc_counter = tmp_counter;
723 sc_tv.tv_sec = sc->sc_advbase;
724 if (carp_suppress_preempt && sc->sc_advskew < 240)
725 sc_tv.tv_usec = 240 * 1000000 / 256;
727 sc_tv.tv_usec = sc->sc_advskew * 1000000 / 256;
728 ch_tv.tv_sec = ch->carp_advbase;
729 ch_tv.tv_usec = ch->carp_advskew * 1000000 / 256;
731 switch (sc->sc_state) {
736 * If we receive an advertisement from a master who's going to
737 * be more frequent than us, go into BACKUP state.
739 if (timevalcmp(&sc_tv, &ch_tv, >) ||
740 timevalcmp(&sc_tv, &ch_tv, ==)) {
741 callout_stop(&sc->sc_ad_tmo);
742 CARP_DEBUG("%s: MASTER -> BACKUP "
743 "(more frequent advertisement received)\n",
744 SC2IFP(sc)->if_xname);
745 carp_set_state(sc, BACKUP);
747 carp_setroute(sc, RTM_DELETE);
752 * If we're pre-empting masters who advertise slower than us,
753 * and this one claims to be slower, treat him as down.
755 if (carp_opts[CARPCTL_PREEMPT] &&
756 timevalcmp(&sc_tv, &ch_tv, <)) {
757 CARP_DEBUG("%s: BACKUP -> MASTER "
758 "(preempting a slower master)\n",
759 SC2IFP(sc)->if_xname);
760 carp_master_down_locked(sc);
765 * If the master is going to advertise at such a low frequency
766 * that he's guaranteed to time out, we'd might as well just
767 * treat him as timed out now.
769 sc_tv.tv_sec = sc->sc_advbase * 3;
770 if (timevalcmp(&sc_tv, &ch_tv, <)) {
771 CARP_DEBUG("%s: BACKUP -> MASTER "
772 "(master timed out)\n",
773 SC2IFP(sc)->if_xname);
774 carp_master_down_locked(sc);
779 * Otherwise, we reset the counter and wait for the next
786 CARP_UNLOCK(ifp->if_carp);
793 carp_prepare_ad(struct mbuf *m, struct carp_softc *sc, struct carp_header *ch)
796 struct ifnet *ifp = SC2IFP(sc);
798 if (sc->sc_init_counter) {
799 /* this could also be seconds since unix epoch */
800 sc->sc_counter = arc4random();
801 sc->sc_counter = sc->sc_counter << 32;
802 sc->sc_counter += arc4random();
806 ch->carp_counter[0] = htonl((sc->sc_counter>>32)&0xffffffff);
807 ch->carp_counter[1] = htonl(sc->sc_counter&0xffffffff);
809 carp_hmac_generate(sc, ch->carp_counter, ch->carp_md);
811 /* Tag packet for carp_output */
812 mtag = m_tag_get(PACKET_TAG_CARP, sizeof(struct ifnet *), M_NOWAIT);
815 SC2IFP(sc)->if_oerrors++;
818 bcopy(&ifp, (caddr_t)(mtag + 1), sizeof(struct ifnet *));
819 m_tag_prepend(m, mtag);
825 carp_send_ad_all(void)
827 struct carp_softc *sc;
830 LIST_FOREACH(sc, &carpif_list, sc_next) {
831 if (sc->sc_carpdev == NULL)
834 if ((SC2IFP(sc)->if_flags & IFF_UP) &&
835 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING) &&
836 sc->sc_state == MASTER)
837 carp_send_ad_locked(sc);
840 mtx_unlock(&carp_mtx);
844 carp_send_ad(void *v)
846 struct carp_softc *sc = v;
849 carp_send_ad_locked(sc);
854 carp_send_ad_locked(struct carp_softc *sc)
856 struct carp_header ch;
858 struct carp_header *ch_ptr;
860 int len, advbase, advskew;
862 CARP_SCLOCK_ASSERT(sc);
864 /* bow out if we've lost our UPness or RUNNINGuiness */
865 if (!((SC2IFP(sc)->if_flags & IFF_UP) &&
866 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING))) {
870 advbase = sc->sc_advbase;
871 if (!carp_suppress_preempt || sc->sc_advskew > 240)
872 advskew = sc->sc_advskew;
876 tv.tv_usec = advskew * 1000000 / 256;
879 ch.carp_version = CARP_VERSION;
880 ch.carp_type = CARP_ADVERTISEMENT;
881 ch.carp_vhid = sc->sc_vhid;
882 ch.carp_advbase = advbase;
883 ch.carp_advskew = advskew;
884 ch.carp_authlen = 7; /* XXX DEFINE */
885 ch.carp_pad1 = 0; /* must be zero */
892 MGETHDR(m, M_DONTWAIT, MT_HEADER);
894 SC2IFP(sc)->if_oerrors++;
895 carpstats.carps_onomem++;
896 /* XXX maybe less ? */
897 if (advbase != 255 || advskew != 255)
898 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv),
902 len = sizeof(*ip) + sizeof(ch);
903 m->m_pkthdr.len = len;
904 m->m_pkthdr.rcvif = NULL;
906 MH_ALIGN(m, m->m_len);
907 m->m_flags |= M_MCAST;
908 ip = mtod(m, struct ip *);
909 ip->ip_v = IPVERSION;
910 ip->ip_hl = sizeof(*ip) >> 2;
911 ip->ip_tos = IPTOS_LOWDELAY;
913 ip->ip_id = ip_newid();
915 ip->ip_ttl = CARP_DFLTTL;
916 ip->ip_p = IPPROTO_CARP;
918 ip->ip_src.s_addr = sc->sc_ia->ia_addr.sin_addr.s_addr;
919 ip->ip_dst.s_addr = htonl(INADDR_CARP_GROUP);
921 ch_ptr = (struct carp_header *)(&ip[1]);
922 bcopy(&ch, ch_ptr, sizeof(ch));
923 if (carp_prepare_ad(m, sc, ch_ptr))
926 m->m_data += sizeof(*ip);
927 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip));
928 m->m_data -= sizeof(*ip);
930 getmicrotime(&SC2IFP(sc)->if_lastchange);
931 SC2IFP(sc)->if_opackets++;
932 SC2IFP(sc)->if_obytes += len;
933 carpstats.carps_opackets++;
935 if (ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, NULL)) {
936 SC2IFP(sc)->if_oerrors++;
937 if (sc->sc_sendad_errors < INT_MAX)
938 sc->sc_sendad_errors++;
939 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) {
940 carp_suppress_preempt++;
941 if (carp_suppress_preempt == 1) {
947 sc->sc_sendad_success = 0;
949 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) {
950 if (++sc->sc_sendad_success >=
951 CARP_SENDAD_MIN_SUCCESS) {
952 carp_suppress_preempt--;
953 sc->sc_sendad_errors = 0;
956 sc->sc_sendad_errors = 0;
964 MGETHDR(m, M_DONTWAIT, MT_HEADER);
966 SC2IFP(sc)->if_oerrors++;
967 carpstats.carps_onomem++;
968 /* XXX maybe less ? */
969 if (advbase != 255 || advskew != 255)
970 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv),
974 len = sizeof(*ip6) + sizeof(ch);
975 m->m_pkthdr.len = len;
976 m->m_pkthdr.rcvif = NULL;
978 MH_ALIGN(m, m->m_len);
979 m->m_flags |= M_MCAST;
980 ip6 = mtod(m, struct ip6_hdr *);
981 bzero(ip6, sizeof(*ip6));
982 ip6->ip6_vfc |= IPV6_VERSION;
983 ip6->ip6_hlim = CARP_DFLTTL;
984 ip6->ip6_nxt = IPPROTO_CARP;
985 bcopy(&sc->sc_ia6->ia_addr.sin6_addr, &ip6->ip6_src,
986 sizeof(struct in6_addr));
987 /* set the multicast destination */
989 ip6->ip6_dst.s6_addr16[0] = htons(0xff02);
990 ip6->ip6_dst.s6_addr8[15] = 0x12;
991 if (in6_setscope(&ip6->ip6_dst, sc->sc_carpdev, NULL) != 0) {
992 SC2IFP(sc)->if_oerrors++;
994 CARP_LOG("%s: in6_setscope failed\n", __func__);
998 ch_ptr = (struct carp_header *)(&ip6[1]);
999 bcopy(&ch, ch_ptr, sizeof(ch));
1000 if (carp_prepare_ad(m, sc, ch_ptr))
1003 m->m_data += sizeof(*ip6);
1004 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip6));
1005 m->m_data -= sizeof(*ip6);
1007 getmicrotime(&SC2IFP(sc)->if_lastchange);
1008 SC2IFP(sc)->if_opackets++;
1009 SC2IFP(sc)->if_obytes += len;
1010 carpstats.carps_opackets6++;
1012 if (ip6_output(m, NULL, NULL, 0, &sc->sc_im6o, NULL, NULL)) {
1013 SC2IFP(sc)->if_oerrors++;
1014 if (sc->sc_sendad_errors < INT_MAX)
1015 sc->sc_sendad_errors++;
1016 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) {
1017 carp_suppress_preempt++;
1018 if (carp_suppress_preempt == 1) {
1024 sc->sc_sendad_success = 0;
1026 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) {
1027 if (++sc->sc_sendad_success >=
1028 CARP_SENDAD_MIN_SUCCESS) {
1029 carp_suppress_preempt--;
1030 sc->sc_sendad_errors = 0;
1033 sc->sc_sendad_errors = 0;
1038 if (advbase != 255 || advskew != 255)
1039 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv),
1045 * Broadcast a gratuitous ARP request containing
1046 * the virtual router MAC address for each IP address
1047 * associated with the virtual router.
1050 carp_send_arp(struct carp_softc *sc)
1054 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
1056 if (ifa->ifa_addr->sa_family != AF_INET)
1059 /* arprequest(sc->sc_carpdev, &in, &in, IF_LLADDR(sc->sc_ifp)); */
1060 arp_ifinit2(sc->sc_carpdev, ifa, IF_LLADDR(sc->sc_ifp));
1062 DELAY(1000); /* XXX */
1068 carp_send_na(struct carp_softc *sc)
1071 struct in6_addr *in6;
1072 static struct in6_addr mcast = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
1074 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
1076 if (ifa->ifa_addr->sa_family != AF_INET6)
1079 in6 = &ifatoia6(ifa)->ia_addr.sin6_addr;
1080 nd6_na_output(sc->sc_carpdev, &mcast, in6,
1081 ND_NA_FLAG_OVERRIDE, 1, NULL);
1082 DELAY(1000); /* XXX */
1088 carp_addrcount(struct carp_if *cif, struct in_ifaddr *ia, int type)
1090 struct carp_softc *vh;
1094 CARP_LOCK_ASSERT(cif);
1096 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1097 if ((type == CARP_COUNT_RUNNING &&
1098 (SC2IFP(vh)->if_flags & IFF_UP) &&
1099 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING)) ||
1100 (type == CARP_COUNT_MASTER && vh->sc_state == MASTER)) {
1101 TAILQ_FOREACH(ifa, &SC2IFP(vh)->if_addrlist,
1103 if (ifa->ifa_addr->sa_family == AF_INET &&
1104 ia->ia_addr.sin_addr.s_addr ==
1105 ifatoia(ifa)->ia_addr.sin_addr.s_addr)
1114 carp_iamatch(void *v, struct in_ifaddr *ia,
1115 struct in_addr *isaddr, u_int8_t **enaddr)
1117 struct carp_if *cif = v;
1118 struct carp_softc *vh;
1119 int index, count = 0;
1124 if (carp_opts[CARPCTL_ARPBALANCE]) {
1126 * XXX proof of concept implementation.
1127 * We use the source ip to decide which virtual host should
1128 * handle the request. If we're master of that virtual host,
1129 * then we respond, otherwise, just drop the arp packet on
1132 count = carp_addrcount(cif, ia, CARP_COUNT_RUNNING);
1134 /* should never reach this */
1139 /* this should be a hash, like pf_hash() */
1140 index = ntohl(isaddr->s_addr) % count;
1143 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1144 if ((SC2IFP(vh)->if_flags & IFF_UP) &&
1145 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING)) {
1146 TAILQ_FOREACH(ifa, &SC2IFP(vh)->if_addrlist,
1148 if (ifa->ifa_addr->sa_family ==
1150 ia->ia_addr.sin_addr.s_addr ==
1151 ifatoia(ifa)->ia_addr.sin_addr.s_addr) {
1152 if (count == index) {
1155 *enaddr = IF_LLADDR(vh->sc_ifp);
1169 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1170 if ((SC2IFP(vh)->if_flags & IFF_UP) &&
1171 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING) &&
1172 ia->ia_ifp == SC2IFP(vh) &&
1173 vh->sc_state == MASTER) {
1174 *enaddr = IF_LLADDR(vh->sc_ifp);
1186 carp_iamatch6(void *v, struct in6_addr *taddr)
1188 struct carp_if *cif = v;
1189 struct carp_softc *vh;
1193 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1194 TAILQ_FOREACH(ifa, &SC2IFP(vh)->if_addrlist, ifa_list) {
1195 if (IN6_ARE_ADDR_EQUAL(taddr,
1196 &ifatoia6(ifa)->ia_addr.sin6_addr) &&
1197 (SC2IFP(vh)->if_flags & IFF_UP) &&
1198 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING) &&
1199 vh->sc_state == MASTER) {
1211 carp_macmatch6(void *v, struct mbuf *m, const struct in6_addr *taddr)
1214 struct carp_if *cif = v;
1215 struct carp_softc *sc;
1219 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list) {
1220 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
1221 if (IN6_ARE_ADDR_EQUAL(taddr,
1222 &ifatoia6(ifa)->ia_addr.sin6_addr) &&
1223 (SC2IFP(sc)->if_flags & IFF_UP) &&
1224 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING)) {
1225 struct ifnet *ifp = SC2IFP(sc);
1226 mtag = m_tag_get(PACKET_TAG_CARP,
1227 sizeof(struct ifnet *), M_NOWAIT);
1229 /* better a bit than nothing */
1231 return (IF_LLADDR(sc->sc_ifp));
1233 bcopy(&ifp, (caddr_t)(mtag + 1),
1234 sizeof(struct ifnet *));
1235 m_tag_prepend(m, mtag);
1238 return (IF_LLADDR(sc->sc_ifp));
1249 carp_forus(void *v, void *dhost)
1251 struct carp_if *cif = v;
1252 struct carp_softc *vh;
1253 u_int8_t *ena = dhost;
1255 if (ena[0] || ena[1] || ena[2] != 0x5e || ena[3] || ena[4] != 1)
1259 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list)
1260 if ((SC2IFP(vh)->if_flags & IFF_UP) &&
1261 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING) &&
1262 vh->sc_state == MASTER &&
1263 !bcmp(dhost, IF_LLADDR(vh->sc_ifp), ETHER_ADDR_LEN)) {
1265 return (SC2IFP(vh));
1273 carp_master_down(void *v)
1275 struct carp_softc *sc = v;
1278 carp_master_down_locked(sc);
1283 carp_master_down_locked(struct carp_softc *sc)
1286 CARP_SCLOCK_ASSERT(sc);
1288 switch (sc->sc_state) {
1290 printf("%s: master_down event in INIT state\n",
1291 SC2IFP(sc)->if_xname);
1296 carp_set_state(sc, MASTER);
1297 carp_send_ad_locked(sc);
1303 carp_setroute(sc, RTM_ADD);
1309 * When in backup state, af indicates whether to reset the master down timer
1310 * for v4 or v6. If it's set to zero, reset the ones which are already pending.
1313 carp_setrun(struct carp_softc *sc, sa_family_t af)
1317 if (sc->sc_carpdev == NULL) {
1318 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING;
1319 carp_set_state(sc, INIT);
1322 CARP_SCLOCK_ASSERT(sc);
1324 if (SC2IFP(sc)->if_flags & IFF_UP &&
1325 sc->sc_vhid > 0 && (sc->sc_naddrs || sc->sc_naddrs6))
1326 SC2IFP(sc)->if_drv_flags |= IFF_DRV_RUNNING;
1328 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING;
1329 carp_setroute(sc, RTM_DELETE);
1333 switch (sc->sc_state) {
1335 if (carp_opts[CARPCTL_PREEMPT] && !carp_suppress_preempt) {
1336 carp_send_ad_locked(sc);
1341 CARP_DEBUG("%s: INIT -> MASTER (preempting)\n",
1342 SC2IFP(sc)->if_xname);
1343 carp_set_state(sc, MASTER);
1344 carp_setroute(sc, RTM_ADD);
1346 CARP_DEBUG("%s: INIT -> BACKUP\n", SC2IFP(sc)->if_xname);
1347 carp_set_state(sc, BACKUP);
1348 carp_setroute(sc, RTM_DELETE);
1353 callout_stop(&sc->sc_ad_tmo);
1354 tv.tv_sec = 3 * sc->sc_advbase;
1355 tv.tv_usec = sc->sc_advskew * 1000000 / 256;
1359 callout_reset(&sc->sc_md_tmo, tvtohz(&tv),
1360 carp_master_down, sc);
1365 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv),
1366 carp_master_down, sc);
1371 callout_reset(&sc->sc_md_tmo, tvtohz(&tv),
1372 carp_master_down, sc);
1374 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv),
1375 carp_master_down, sc);
1380 tv.tv_sec = sc->sc_advbase;
1381 tv.tv_usec = sc->sc_advskew * 1000000 / 256;
1382 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv),
1389 carp_multicast_cleanup(struct carp_softc *sc)
1391 struct ip_moptions *imo = &sc->sc_imo;
1392 u_int16_t n = imo->imo_num_memberships;
1394 /* Clean up our own multicast memberships */
1396 if (imo->imo_membership[n] != NULL) {
1397 in_delmulti(imo->imo_membership[n]);
1398 imo->imo_membership[n] = NULL;
1401 KASSERT(imo->imo_mfilters == NULL,
1402 ("%s: imo_mfilters != NULL", __func__));
1403 imo->imo_num_memberships = 0;
1404 imo->imo_multicast_ifp = NULL;
1409 carp_multicast6_cleanup(struct carp_softc *sc)
1411 struct ip6_moptions *im6o = &sc->sc_im6o;
1413 while (!LIST_EMPTY(&im6o->im6o_memberships)) {
1414 struct in6_multi_mship *imm =
1415 LIST_FIRST(&im6o->im6o_memberships);
1417 LIST_REMOVE(imm, i6mm_chain);
1418 in6_leavegroup(imm);
1420 im6o->im6o_multicast_ifp = NULL;
1425 carp_set_addr(struct carp_softc *sc, struct sockaddr_in *sin)
1428 struct carp_if *cif;
1429 struct in_ifaddr *ia, *ia_if;
1430 struct ip_moptions *imo = &sc->sc_imo;
1431 struct in_addr addr;
1432 u_long iaddr = htonl(sin->sin_addr.s_addr);
1435 if (sin->sin_addr.s_addr == 0) {
1436 if (!(SC2IFP(sc)->if_flags & IFF_UP))
1437 carp_set_state(sc, INIT);
1439 SC2IFP(sc)->if_flags |= IFF_UP;
1444 /* we have to do it by hands to check we won't match on us */
1445 ia_if = NULL; own = 0;
1446 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
1447 /* and, yeah, we need a multicast-capable iface too */
1448 if (ia->ia_ifp != SC2IFP(sc) &&
1449 (ia->ia_ifp->if_flags & IFF_MULTICAST) &&
1450 (iaddr & ia->ia_subnetmask) == ia->ia_subnet) {
1453 if (sin->sin_addr.s_addr ==
1454 ia->ia_addr.sin_addr.s_addr)
1460 return (EADDRNOTAVAIL);
1465 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0 ||
1466 (imo->imo_multicast_ifp && imo->imo_multicast_ifp != ifp))
1467 return (EADDRNOTAVAIL);
1469 if (imo->imo_num_memberships == 0) {
1470 addr.s_addr = htonl(INADDR_CARP_GROUP);
1471 if ((imo->imo_membership[0] = in_addmulti(&addr, ifp)) == NULL)
1473 imo->imo_num_memberships++;
1474 imo->imo_multicast_ifp = ifp;
1475 imo->imo_multicast_ttl = CARP_DFLTTL;
1476 imo->imo_multicast_loop = 0;
1479 if (!ifp->if_carp) {
1481 MALLOC(cif, struct carp_if *, sizeof(*cif), M_CARP,
1487 if ((error = ifpromisc(ifp, 1))) {
1492 CARP_LOCK_INIT(cif);
1494 cif->vhif_ifp = ifp;
1495 TAILQ_INIT(&cif->vhif_vrs);
1499 struct carp_softc *vr;
1501 cif = (struct carp_if *)ifp->if_carp;
1503 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list)
1504 if (vr != sc && vr->sc_vhid == sc->sc_vhid) {
1511 sc->sc_carpdev = ifp;
1513 { /* XXX prevent endless loop if already in queue */
1514 struct carp_softc *vr, *after = NULL;
1516 cif = (struct carp_if *)ifp->if_carp;
1518 /* XXX: cif should not change, right? So we still hold the lock */
1519 CARP_LOCK_ASSERT(cif);
1521 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) {
1524 if (vr->sc_vhid < sc->sc_vhid)
1529 /* We're trying to keep things in order */
1530 if (after == NULL) {
1531 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list);
1533 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, sc, sc_list);
1540 SC2IFP(sc)->if_flags |= IFF_UP;
1543 carp_sc_state_locked(sc);
1551 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]);
1556 carp_del_addr(struct carp_softc *sc, struct sockaddr_in *sin)
1560 if (!--sc->sc_naddrs) {
1561 struct carp_if *cif = (struct carp_if *)sc->sc_carpdev->if_carp;
1562 struct ip_moptions *imo = &sc->sc_imo;
1565 callout_stop(&sc->sc_ad_tmo);
1566 SC2IFP(sc)->if_flags &= ~IFF_UP;
1567 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING;
1569 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]);
1570 imo->imo_multicast_ifp = NULL;
1571 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list);
1572 if (!--cif->vhif_nvrs) {
1573 sc->sc_carpdev->if_carp = NULL;
1574 CARP_LOCK_DESTROY(cif);
1575 FREE(cif, M_IFADDR);
1586 carp_set_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6)
1589 struct carp_if *cif;
1590 struct in6_ifaddr *ia, *ia_if;
1591 struct ip6_moptions *im6o = &sc->sc_im6o;
1592 struct in6_multi_mship *imm;
1593 struct in6_addr in6;
1596 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1597 if (!(SC2IFP(sc)->if_flags & IFF_UP))
1598 carp_set_state(sc, INIT);
1600 SC2IFP(sc)->if_flags |= IFF_UP;
1605 /* we have to do it by hands to check we won't match on us */
1606 ia_if = NULL; own = 0;
1607 for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1610 for (i = 0; i < 4; i++) {
1611 if ((sin6->sin6_addr.s6_addr32[i] &
1612 ia->ia_prefixmask.sin6_addr.s6_addr32[i]) !=
1613 (ia->ia_addr.sin6_addr.s6_addr32[i] &
1614 ia->ia_prefixmask.sin6_addr.s6_addr32[i]))
1617 /* and, yeah, we need a multicast-capable iface too */
1618 if (ia->ia_ifp != SC2IFP(sc) &&
1619 (ia->ia_ifp->if_flags & IFF_MULTICAST) &&
1623 if (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
1624 &ia->ia_addr.sin6_addr))
1630 return (EADDRNOTAVAIL);
1634 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0 ||
1635 (im6o->im6o_multicast_ifp && im6o->im6o_multicast_ifp != ifp))
1636 return (EADDRNOTAVAIL);
1638 if (!sc->sc_naddrs6) {
1639 im6o->im6o_multicast_ifp = ifp;
1641 /* join CARP multicast address */
1642 bzero(&in6, sizeof(in6));
1643 in6.s6_addr16[0] = htons(0xff02);
1644 in6.s6_addr8[15] = 0x12;
1645 if (in6_setscope(&in6, ifp, NULL) != 0)
1647 if ((imm = in6_joingroup(ifp, &in6, &error, 0)) == NULL)
1649 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1651 /* join solicited multicast address */
1652 bzero(&in6, sizeof(in6));
1653 in6.s6_addr16[0] = htons(0xff02);
1654 in6.s6_addr32[1] = 0;
1655 in6.s6_addr32[2] = htonl(1);
1656 in6.s6_addr32[3] = sin6->sin6_addr.s6_addr32[3];
1657 in6.s6_addr8[12] = 0xff;
1658 if (in6_setscope(&in6, ifp, NULL) != 0)
1660 if ((imm = in6_joingroup(ifp, &in6, &error, 0)) == NULL)
1662 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1665 if (!ifp->if_carp) {
1666 MALLOC(cif, struct carp_if *, sizeof(*cif), M_CARP,
1672 if ((error = ifpromisc(ifp, 1))) {
1677 CARP_LOCK_INIT(cif);
1679 cif->vhif_ifp = ifp;
1680 TAILQ_INIT(&cif->vhif_vrs);
1684 struct carp_softc *vr;
1686 cif = (struct carp_if *)ifp->if_carp;
1688 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list)
1689 if (vr != sc && vr->sc_vhid == sc->sc_vhid) {
1696 sc->sc_carpdev = ifp;
1698 { /* XXX prevent endless loop if already in queue */
1699 struct carp_softc *vr, *after = NULL;
1701 cif = (struct carp_if *)ifp->if_carp;
1702 CARP_LOCK_ASSERT(cif);
1704 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) {
1707 if (vr->sc_vhid < sc->sc_vhid)
1712 /* We're trying to keep things in order */
1713 if (after == NULL) {
1714 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list);
1716 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, sc, sc_list);
1723 SC2IFP(sc)->if_flags |= IFF_UP;
1726 carp_sc_state_locked(sc);
1734 /* clean up multicast memberships */
1735 if (!sc->sc_naddrs6) {
1736 while (!LIST_EMPTY(&im6o->im6o_memberships)) {
1737 imm = LIST_FIRST(&im6o->im6o_memberships);
1738 LIST_REMOVE(imm, i6mm_chain);
1739 in6_leavegroup(imm);
1746 carp_del_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6)
1750 if (!--sc->sc_naddrs6) {
1751 struct carp_if *cif = (struct carp_if *)sc->sc_carpdev->if_carp;
1752 struct ip6_moptions *im6o = &sc->sc_im6o;
1755 callout_stop(&sc->sc_ad_tmo);
1756 SC2IFP(sc)->if_flags &= ~IFF_UP;
1757 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING;
1759 while (!LIST_EMPTY(&im6o->im6o_memberships)) {
1760 struct in6_multi_mship *imm =
1761 LIST_FIRST(&im6o->im6o_memberships);
1763 LIST_REMOVE(imm, i6mm_chain);
1764 in6_leavegroup(imm);
1766 im6o->im6o_multicast_ifp = NULL;
1767 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list);
1768 if (!--cif->vhif_nvrs) {
1769 CARP_LOCK_DESTROY(cif);
1770 sc->sc_carpdev->if_carp = NULL;
1771 FREE(cif, M_IFADDR);
1781 carp_ioctl(struct ifnet *ifp, u_long cmd, caddr_t addr)
1783 struct carp_softc *sc = ifp->if_softc, *vr;
1784 struct carpreq carpr;
1787 struct ifaliasreq *ifra;
1788 int locked = 0, error = 0;
1790 ifa = (struct ifaddr *)addr;
1791 ifra = (struct ifaliasreq *)addr;
1792 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 = priv_check(curthread, PRIV_NETINET_CARP);
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) {
1916 if (error == EEXIST)
1919 sc->sc_vhid = carpr.carpr_vhid;
1920 IF_LLADDR(sc->sc_ifp)[0] = 0;
1921 IF_LLADDR(sc->sc_ifp)[1] = 0;
1922 IF_LLADDR(sc->sc_ifp)[2] = 0x5e;
1923 IF_LLADDR(sc->sc_ifp)[3] = 0;
1924 IF_LLADDR(sc->sc_ifp)[4] = 1;
1925 IF_LLADDR(sc->sc_ifp)[5] = sc->sc_vhid;
1928 if (carpr.carpr_advbase > 0 || carpr.carpr_advskew > 0) {
1929 if (carpr.carpr_advskew >= 255) {
1933 if (carpr.carpr_advbase > 255) {
1937 sc->sc_advbase = carpr.carpr_advbase;
1938 sc->sc_advskew = carpr.carpr_advskew;
1941 bcopy(carpr.carpr_key, sc->sc_key, sizeof(sc->sc_key));
1951 /* XXX: lockless read */
1952 bzero(&carpr, sizeof(carpr));
1953 carpr.carpr_state = sc->sc_state;
1954 carpr.carpr_vhid = sc->sc_vhid;
1955 carpr.carpr_advbase = sc->sc_advbase;
1956 carpr.carpr_advskew = sc->sc_advskew;
1957 error = priv_check(curthread, PRIV_NETINET_CARP);
1959 bcopy(sc->sc_key, carpr.carpr_key,
1960 sizeof(carpr.carpr_key));
1961 error = copyout(&carpr, ifr->ifr_data, sizeof(carpr));
1971 carp_hmac_prepare(sc);
1977 * XXX: this is looutput. We should eventually use it from there.
1980 carp_looutput(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
1985 M_ASSERTPKTHDR(m); /* check if we have the packet header */
1987 if (rt && rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
1989 return (rt->rt_flags & RTF_BLACKHOLE ? 0 :
1990 rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
1994 ifp->if_obytes += m->m_pkthdr.len;
1996 /* BPF writes need to be handled specially. */
1997 if (dst->sa_family == AF_UNSPEC) {
1998 bcopy(dst->sa_data, &af, sizeof(af));
1999 dst->sa_family = af;
2003 switch (dst->sa_family) {
2010 printf("carp_looutput: af=%d unexpected\n", dst->sa_family);
2012 return (EAFNOSUPPORT);
2015 return(if_simloop(ifp, m, dst->sa_family, 0));
2019 * Start output on carp interface. This function should never be called.
2022 carp_start(struct ifnet *ifp)
2025 printf("%s: start called\n", ifp->if_xname);
2030 carp_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2034 struct carp_softc *sc;
2035 struct ifnet *carp_ifp;
2040 switch (sa->sa_family) {
2053 mtag = m_tag_find(m, PACKET_TAG_CARP, NULL);
2057 bcopy(mtag + 1, &carp_ifp, sizeof(struct ifnet *));
2058 sc = carp_ifp->if_softc;
2060 /* Set the source MAC address to Virtual Router MAC Address */
2061 switch (ifp->if_type) {
2064 struct ether_header *eh;
2066 eh = mtod(m, struct ether_header *);
2067 eh->ether_shost[0] = 0;
2068 eh->ether_shost[1] = 0;
2069 eh->ether_shost[2] = 0x5e;
2070 eh->ether_shost[3] = 0;
2071 eh->ether_shost[4] = 1;
2072 eh->ether_shost[5] = sc->sc_vhid;
2076 struct fddi_header *fh;
2078 fh = mtod(m, struct fddi_header *);
2079 fh->fddi_shost[0] = 0;
2080 fh->fddi_shost[1] = 0;
2081 fh->fddi_shost[2] = 0x5e;
2082 fh->fddi_shost[3] = 0;
2083 fh->fddi_shost[4] = 1;
2084 fh->fddi_shost[5] = sc->sc_vhid;
2087 case IFT_ISO88025: {
2088 struct iso88025_header *th;
2089 th = mtod(m, struct iso88025_header *);
2090 th->iso88025_shost[0] = 3;
2091 th->iso88025_shost[1] = 0;
2092 th->iso88025_shost[2] = 0x40 >> (sc->sc_vhid - 1);
2093 th->iso88025_shost[3] = 0x40000 >> (sc->sc_vhid - 1);
2094 th->iso88025_shost[4] = 0;
2095 th->iso88025_shost[5] = 0;
2099 printf("%s: carp is not supported for this interface type\n",
2101 return (EOPNOTSUPP);
2108 carp_set_state(struct carp_softc *sc, int state)
2112 CARP_SCLOCK_ASSERT(sc);
2114 if (sc->sc_state == state)
2117 sc->sc_state = state;
2120 SC2IFP(sc)->if_link_state = LINK_STATE_DOWN;
2123 SC2IFP(sc)->if_link_state = LINK_STATE_UP;
2126 SC2IFP(sc)->if_link_state = LINK_STATE_UNKNOWN;
2129 rt_ifmsg(SC2IFP(sc));
2133 carp_carpdev_state(void *v)
2135 struct carp_if *cif = v;
2138 carp_carpdev_state_locked(cif);
2143 carp_carpdev_state_locked(struct carp_if *cif)
2145 struct carp_softc *sc;
2147 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list)
2148 carp_sc_state_locked(sc);
2152 carp_sc_state_locked(struct carp_softc *sc)
2154 CARP_SCLOCK_ASSERT(sc);
2156 if (sc->sc_carpdev->if_link_state != LINK_STATE_UP ||
2157 !(sc->sc_carpdev->if_flags & IFF_UP)) {
2158 sc->sc_flags_backup = SC2IFP(sc)->if_flags;
2159 SC2IFP(sc)->if_flags &= ~IFF_UP;
2160 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING;
2161 callout_stop(&sc->sc_ad_tmo);
2162 callout_stop(&sc->sc_md_tmo);
2163 callout_stop(&sc->sc_md6_tmo);
2164 carp_set_state(sc, INIT);
2166 if (!sc->sc_suppress) {
2167 carp_suppress_preempt++;
2168 if (carp_suppress_preempt == 1) {
2174 sc->sc_suppress = 1;
2176 SC2IFP(sc)->if_flags |= sc->sc_flags_backup;
2177 carp_set_state(sc, INIT);
2179 if (sc->sc_suppress)
2180 carp_suppress_preempt--;
2181 sc->sc_suppress = 0;
2188 carp_modevent(module_t mod, int type, void *data)
2192 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2193 carp_ifdetach, NULL, EVENTHANDLER_PRI_ANY);
2194 if (if_detach_event_tag == NULL)
2196 mtx_init(&carp_mtx, "carp_mtx", NULL, MTX_DEF);
2197 LIST_INIT(&carpif_list);
2198 if_clone_attach(&carp_cloner);
2202 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2203 if_clone_detach(&carp_cloner);
2204 mtx_destroy(&carp_mtx);
2214 static moduledata_t carp_mod = {
2220 DECLARE_MODULE(carp, carp_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);