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_max_memberships = IP_MIN_MEMBERSHIPS;
384 sc->sc_imo.imo_multicast_vif = -1;
386 callout_init(&sc->sc_ad_tmo, NET_CALLOUT_MPSAFE);
387 callout_init(&sc->sc_md_tmo, NET_CALLOUT_MPSAFE);
388 callout_init(&sc->sc_md6_tmo, NET_CALLOUT_MPSAFE);
391 if_initname(ifp, CARP_IFNAME, unit);
392 ifp->if_mtu = ETHERMTU;
393 ifp->if_flags = IFF_LOOPBACK;
394 ifp->if_ioctl = carp_ioctl;
395 ifp->if_output = carp_looutput;
396 ifp->if_start = carp_start;
397 ifp->if_type = IFT_CARP;
398 ifp->if_snd.ifq_maxlen = ifqmaxlen;
401 bpfattach(SC2IFP(sc), DLT_NULL, sizeof(u_int32_t));
403 LIST_INSERT_HEAD(&carpif_list, sc, sc_next);
404 mtx_unlock(&carp_mtx);
409 carp_clone_destroy(struct ifnet *ifp)
411 struct carp_softc *sc = ifp->if_softc;
415 carpdetach(sc, 1); /* Returns unlocked. */
418 LIST_REMOVE(sc, sc_next);
419 mtx_unlock(&carp_mtx);
422 if_free_type(ifp, IFT_ETHER);
423 free(sc->sc_imo.imo_membership, M_CARP);
428 * This function can be called on CARP interface destroy path,
429 * and in case of the removal of the underlying interface as
430 * well. We differentiate these two cases. In the latter case
431 * we do not cleanup our multicast memberships, since they
432 * are already freed. Also, in the latter case we do not
433 * release the lock on return, because the function will be
434 * called once more, for another CARP instance on the same
438 carpdetach(struct carp_softc *sc, int unlock)
442 callout_stop(&sc->sc_ad_tmo);
443 callout_stop(&sc->sc_md_tmo);
444 callout_stop(&sc->sc_md6_tmo);
447 carp_suppress_preempt--;
450 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS)
451 carp_suppress_preempt--;
452 sc->sc_sendad_errors = 0;
454 carp_set_state(sc, INIT);
455 SC2IFP(sc)->if_flags &= ~IFF_UP;
458 carp_multicast_cleanup(sc);
460 carp_multicast6_cleanup(sc);
463 if (sc->sc_carpdev != NULL) {
464 cif = (struct carp_if *)sc->sc_carpdev->if_carp;
465 CARP_LOCK_ASSERT(cif);
466 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list);
467 if (!--cif->vhif_nvrs) {
468 ifpromisc(sc->sc_carpdev, 0);
469 sc->sc_carpdev->if_carp = NULL;
470 CARP_LOCK_DESTROY(cif);
474 sc->sc_carpdev = NULL;
478 /* Detach an interface from the carp. */
480 carp_ifdetach(void *arg __unused, struct ifnet *ifp)
482 struct carp_if *cif = (struct carp_if *)ifp->if_carp;
483 struct carp_softc *sc, *nextsc;
489 * XXX: At the end of for() cycle the lock will be destroyed.
492 for (sc = TAILQ_FIRST(&cif->vhif_vrs); sc; sc = nextsc) {
493 nextsc = TAILQ_NEXT(sc, sc_list);
499 * process input packet.
500 * we have rearranged checks order compared to the rfc,
501 * but it seems more efficient this way or not possible otherwise.
504 carp_input(struct mbuf *m, int hlen)
506 struct ip *ip = mtod(m, struct ip *);
507 struct carp_header *ch;
510 carpstats.carps_ipackets++;
512 if (!carp_opts[CARPCTL_ALLOW]) {
517 /* check if received on a valid carp interface */
518 if (m->m_pkthdr.rcvif->if_carp == NULL) {
519 carpstats.carps_badif++;
520 CARP_LOG("carp_input: packet received on non-carp "
522 m->m_pkthdr.rcvif->if_xname);
527 /* verify that the IP TTL is 255. */
528 if (ip->ip_ttl != CARP_DFLTTL) {
529 carpstats.carps_badttl++;
530 CARP_LOG("carp_input: received ttl %d != 255i on %s\n",
532 m->m_pkthdr.rcvif->if_xname);
537 iplen = ip->ip_hl << 2;
539 if (m->m_pkthdr.len < iplen + sizeof(*ch)) {
540 carpstats.carps_badlen++;
541 CARP_LOG("carp_input: received len %zd < "
542 "sizeof(struct carp_header)\n",
543 m->m_len - sizeof(struct ip));
548 if (iplen + sizeof(*ch) < m->m_len) {
549 if ((m = m_pullup(m, iplen + sizeof(*ch))) == NULL) {
550 carpstats.carps_hdrops++;
551 CARP_LOG("carp_input: pullup failed\n");
554 ip = mtod(m, struct ip *);
556 ch = (struct carp_header *)((char *)ip + iplen);
559 * verify that the received packet length is
560 * equal to the CARP header
562 len = iplen + sizeof(*ch);
563 if (len > m->m_pkthdr.len) {
564 carpstats.carps_badlen++;
565 CARP_LOG("carp_input: packet too short %d on %s\n",
567 m->m_pkthdr.rcvif->if_xname);
572 if ((m = m_pullup(m, len)) == NULL) {
573 carpstats.carps_hdrops++;
576 ip = mtod(m, struct ip *);
577 ch = (struct carp_header *)((char *)ip + iplen);
579 /* verify the CARP checksum */
581 if (carp_cksum(m, len - iplen)) {
582 carpstats.carps_badsum++;
583 CARP_LOG("carp_input: checksum failed on %s\n",
584 m->m_pkthdr.rcvif->if_xname);
590 carp_input_c(m, ch, AF_INET);
595 carp6_input(struct mbuf **mp, int *offp, int proto)
597 struct mbuf *m = *mp;
598 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
599 struct carp_header *ch;
602 carpstats.carps_ipackets6++;
604 if (!carp_opts[CARPCTL_ALLOW]) {
606 return (IPPROTO_DONE);
609 /* check if received on a valid carp interface */
610 if (m->m_pkthdr.rcvif->if_carp == NULL) {
611 carpstats.carps_badif++;
612 CARP_LOG("carp6_input: packet received on non-carp "
614 m->m_pkthdr.rcvif->if_xname);
616 return (IPPROTO_DONE);
619 /* verify that the IP TTL is 255 */
620 if (ip6->ip6_hlim != CARP_DFLTTL) {
621 carpstats.carps_badttl++;
622 CARP_LOG("carp6_input: received ttl %d != 255 on %s\n",
624 m->m_pkthdr.rcvif->if_xname);
626 return (IPPROTO_DONE);
629 /* verify that we have a complete carp packet */
631 IP6_EXTHDR_GET(ch, struct carp_header *, m, *offp, sizeof(*ch));
633 carpstats.carps_badlen++;
634 CARP_LOG("carp6_input: packet size %u too small\n", len);
635 return (IPPROTO_DONE);
639 /* verify the CARP checksum */
641 if (carp_cksum(m, sizeof(*ch))) {
642 carpstats.carps_badsum++;
643 CARP_LOG("carp6_input: checksum failed, on %s\n",
644 m->m_pkthdr.rcvif->if_xname);
646 return (IPPROTO_DONE);
650 carp_input_c(m, ch, AF_INET6);
651 return (IPPROTO_DONE);
656 carp_input_c(struct mbuf *m, struct carp_header *ch, sa_family_t af)
658 struct ifnet *ifp = m->m_pkthdr.rcvif;
659 struct carp_softc *sc;
660 u_int64_t tmp_counter;
661 struct timeval sc_tv, ch_tv;
663 /* verify that the VHID is valid on the receiving interface */
664 CARP_LOCK(ifp->if_carp);
665 TAILQ_FOREACH(sc, &((struct carp_if *)ifp->if_carp)->vhif_vrs, sc_list)
666 if (sc->sc_vhid == ch->carp_vhid)
669 if (!sc || !((SC2IFP(sc)->if_flags & IFF_UP) &&
670 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING))) {
671 carpstats.carps_badvhid++;
672 CARP_UNLOCK(ifp->if_carp);
677 getmicrotime(&SC2IFP(sc)->if_lastchange);
678 SC2IFP(sc)->if_ipackets++;
679 SC2IFP(sc)->if_ibytes += m->m_pkthdr.len;
681 if (bpf_peers_present(SC2IFP(sc)->if_bpf)) {
682 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_mtap2(SC2IFP(sc)->if_bpf, &af1, sizeof(af1), 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 = arc4random();
800 sc->sc_counter = sc->sc_counter << 32;
801 sc->sc_counter += arc4random();
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 *), M_NOWAIT);
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;
829 LIST_FOREACH(sc, &carpif_list, sc_next) {
830 if (sc->sc_carpdev == NULL)
833 if ((SC2IFP(sc)->if_flags & IFF_UP) &&
834 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING) &&
835 sc->sc_state == MASTER)
836 carp_send_ad_locked(sc);
839 mtx_unlock(&carp_mtx);
843 carp_send_ad(void *v)
845 struct carp_softc *sc = v;
848 carp_send_ad_locked(sc);
853 carp_send_ad_locked(struct carp_softc *sc)
855 struct carp_header ch;
857 struct carp_header *ch_ptr;
859 int len, advbase, advskew;
861 CARP_SCLOCK_ASSERT(sc);
863 /* bow out if we've lost our UPness or RUNNINGuiness */
864 if (!((SC2IFP(sc)->if_flags & IFF_UP) &&
865 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING))) {
869 advbase = sc->sc_advbase;
870 if (!carp_suppress_preempt || sc->sc_advskew > 240)
871 advskew = sc->sc_advskew;
875 tv.tv_usec = advskew * 1000000 / 256;
878 ch.carp_version = CARP_VERSION;
879 ch.carp_type = CARP_ADVERTISEMENT;
880 ch.carp_vhid = sc->sc_vhid;
881 ch.carp_advbase = advbase;
882 ch.carp_advskew = advskew;
883 ch.carp_authlen = 7; /* XXX DEFINE */
884 ch.carp_pad1 = 0; /* must be zero */
891 MGETHDR(m, M_DONTWAIT, MT_HEADER);
893 SC2IFP(sc)->if_oerrors++;
894 carpstats.carps_onomem++;
895 /* XXX maybe less ? */
896 if (advbase != 255 || advskew != 255)
897 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv),
901 len = sizeof(*ip) + sizeof(ch);
902 m->m_pkthdr.len = len;
903 m->m_pkthdr.rcvif = NULL;
905 MH_ALIGN(m, m->m_len);
906 m->m_flags |= M_MCAST;
907 ip = mtod(m, struct ip *);
908 ip->ip_v = IPVERSION;
909 ip->ip_hl = sizeof(*ip) >> 2;
910 ip->ip_tos = IPTOS_LOWDELAY;
912 ip->ip_id = ip_newid();
914 ip->ip_ttl = CARP_DFLTTL;
915 ip->ip_p = IPPROTO_CARP;
917 ip->ip_src.s_addr = sc->sc_ia->ia_addr.sin_addr.s_addr;
918 ip->ip_dst.s_addr = htonl(INADDR_CARP_GROUP);
920 ch_ptr = (struct carp_header *)(&ip[1]);
921 bcopy(&ch, ch_ptr, sizeof(ch));
922 if (carp_prepare_ad(m, sc, ch_ptr))
925 m->m_data += sizeof(*ip);
926 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip));
927 m->m_data -= sizeof(*ip);
929 getmicrotime(&SC2IFP(sc)->if_lastchange);
930 SC2IFP(sc)->if_opackets++;
931 SC2IFP(sc)->if_obytes += len;
932 carpstats.carps_opackets++;
934 if (ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, NULL)) {
935 SC2IFP(sc)->if_oerrors++;
936 if (sc->sc_sendad_errors < INT_MAX)
937 sc->sc_sendad_errors++;
938 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) {
939 carp_suppress_preempt++;
940 if (carp_suppress_preempt == 1) {
946 sc->sc_sendad_success = 0;
948 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) {
949 if (++sc->sc_sendad_success >=
950 CARP_SENDAD_MIN_SUCCESS) {
951 carp_suppress_preempt--;
952 sc->sc_sendad_errors = 0;
955 sc->sc_sendad_errors = 0;
963 MGETHDR(m, M_DONTWAIT, MT_HEADER);
965 SC2IFP(sc)->if_oerrors++;
966 carpstats.carps_onomem++;
967 /* XXX maybe less ? */
968 if (advbase != 255 || advskew != 255)
969 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv),
973 len = sizeof(*ip6) + sizeof(ch);
974 m->m_pkthdr.len = len;
975 m->m_pkthdr.rcvif = NULL;
977 MH_ALIGN(m, m->m_len);
978 m->m_flags |= M_MCAST;
979 ip6 = mtod(m, struct ip6_hdr *);
980 bzero(ip6, sizeof(*ip6));
981 ip6->ip6_vfc |= IPV6_VERSION;
982 ip6->ip6_hlim = CARP_DFLTTL;
983 ip6->ip6_nxt = IPPROTO_CARP;
984 bcopy(&sc->sc_ia6->ia_addr.sin6_addr, &ip6->ip6_src,
985 sizeof(struct in6_addr));
986 /* set the multicast destination */
988 ip6->ip6_dst.s6_addr16[0] = htons(0xff02);
989 ip6->ip6_dst.s6_addr8[15] = 0x12;
990 if (in6_setscope(&ip6->ip6_dst, sc->sc_carpdev, NULL) != 0) {
991 SC2IFP(sc)->if_oerrors++;
993 CARP_LOG("%s: in6_setscope failed\n", __func__);
997 ch_ptr = (struct carp_header *)(&ip6[1]);
998 bcopy(&ch, ch_ptr, sizeof(ch));
999 if (carp_prepare_ad(m, sc, ch_ptr))
1002 m->m_data += sizeof(*ip6);
1003 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip6));
1004 m->m_data -= sizeof(*ip6);
1006 getmicrotime(&SC2IFP(sc)->if_lastchange);
1007 SC2IFP(sc)->if_opackets++;
1008 SC2IFP(sc)->if_obytes += len;
1009 carpstats.carps_opackets6++;
1011 if (ip6_output(m, NULL, NULL, 0, &sc->sc_im6o, NULL, NULL)) {
1012 SC2IFP(sc)->if_oerrors++;
1013 if (sc->sc_sendad_errors < INT_MAX)
1014 sc->sc_sendad_errors++;
1015 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) {
1016 carp_suppress_preempt++;
1017 if (carp_suppress_preempt == 1) {
1023 sc->sc_sendad_success = 0;
1025 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) {
1026 if (++sc->sc_sendad_success >=
1027 CARP_SENDAD_MIN_SUCCESS) {
1028 carp_suppress_preempt--;
1029 sc->sc_sendad_errors = 0;
1032 sc->sc_sendad_errors = 0;
1037 if (advbase != 255 || advskew != 255)
1038 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv),
1044 * Broadcast a gratuitous ARP request containing
1045 * the virtual router MAC address for each IP address
1046 * associated with the virtual router.
1049 carp_send_arp(struct carp_softc *sc)
1053 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
1055 if (ifa->ifa_addr->sa_family != AF_INET)
1058 /* arprequest(sc->sc_carpdev, &in, &in, IF_LLADDR(sc->sc_ifp)); */
1059 arp_ifinit2(sc->sc_carpdev, ifa, IF_LLADDR(sc->sc_ifp));
1061 DELAY(1000); /* XXX */
1067 carp_send_na(struct carp_softc *sc)
1070 struct in6_addr *in6;
1071 static struct in6_addr mcast = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
1073 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
1075 if (ifa->ifa_addr->sa_family != AF_INET6)
1078 in6 = &ifatoia6(ifa)->ia_addr.sin6_addr;
1079 nd6_na_output(sc->sc_carpdev, &mcast, in6,
1080 ND_NA_FLAG_OVERRIDE, 1, NULL);
1081 DELAY(1000); /* XXX */
1087 carp_addrcount(struct carp_if *cif, struct in_ifaddr *ia, int type)
1089 struct carp_softc *vh;
1093 CARP_LOCK_ASSERT(cif);
1095 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1096 if ((type == CARP_COUNT_RUNNING &&
1097 (SC2IFP(vh)->if_flags & IFF_UP) &&
1098 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING)) ||
1099 (type == CARP_COUNT_MASTER && vh->sc_state == MASTER)) {
1100 TAILQ_FOREACH(ifa, &SC2IFP(vh)->if_addrlist,
1102 if (ifa->ifa_addr->sa_family == AF_INET &&
1103 ia->ia_addr.sin_addr.s_addr ==
1104 ifatoia(ifa)->ia_addr.sin_addr.s_addr)
1113 carp_iamatch(void *v, struct in_ifaddr *ia,
1114 struct in_addr *isaddr, u_int8_t **enaddr)
1116 struct carp_if *cif = v;
1117 struct carp_softc *vh;
1118 int index, count = 0;
1123 if (carp_opts[CARPCTL_ARPBALANCE]) {
1125 * XXX proof of concept implementation.
1126 * We use the source ip to decide which virtual host should
1127 * handle the request. If we're master of that virtual host,
1128 * then we respond, otherwise, just drop the arp packet on
1131 count = carp_addrcount(cif, ia, CARP_COUNT_RUNNING);
1133 /* should never reach this */
1138 /* this should be a hash, like pf_hash() */
1139 index = ntohl(isaddr->s_addr) % count;
1142 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1143 if ((SC2IFP(vh)->if_flags & IFF_UP) &&
1144 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING)) {
1145 TAILQ_FOREACH(ifa, &SC2IFP(vh)->if_addrlist,
1147 if (ifa->ifa_addr->sa_family ==
1149 ia->ia_addr.sin_addr.s_addr ==
1150 ifatoia(ifa)->ia_addr.sin_addr.s_addr) {
1151 if (count == index) {
1154 *enaddr = IF_LLADDR(vh->sc_ifp);
1168 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1169 if ((SC2IFP(vh)->if_flags & IFF_UP) &&
1170 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING) &&
1171 ia->ia_ifp == SC2IFP(vh) &&
1172 vh->sc_state == MASTER) {
1173 *enaddr = IF_LLADDR(vh->sc_ifp);
1185 carp_iamatch6(void *v, struct in6_addr *taddr)
1187 struct carp_if *cif = v;
1188 struct carp_softc *vh;
1192 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1193 TAILQ_FOREACH(ifa, &SC2IFP(vh)->if_addrlist, ifa_list) {
1194 if (IN6_ARE_ADDR_EQUAL(taddr,
1195 &ifatoia6(ifa)->ia_addr.sin6_addr) &&
1196 (SC2IFP(vh)->if_flags & IFF_UP) &&
1197 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING) &&
1198 vh->sc_state == MASTER) {
1210 carp_macmatch6(void *v, struct mbuf *m, const struct in6_addr *taddr)
1213 struct carp_if *cif = v;
1214 struct carp_softc *sc;
1218 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list) {
1219 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
1220 if (IN6_ARE_ADDR_EQUAL(taddr,
1221 &ifatoia6(ifa)->ia_addr.sin6_addr) &&
1222 (SC2IFP(sc)->if_flags & IFF_UP) &&
1223 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING)) {
1224 struct ifnet *ifp = SC2IFP(sc);
1225 mtag = m_tag_get(PACKET_TAG_CARP,
1226 sizeof(struct ifnet *), M_NOWAIT);
1228 /* better a bit than nothing */
1230 return (IF_LLADDR(sc->sc_ifp));
1232 bcopy(&ifp, (caddr_t)(mtag + 1),
1233 sizeof(struct ifnet *));
1234 m_tag_prepend(m, mtag);
1237 return (IF_LLADDR(sc->sc_ifp));
1248 carp_forus(void *v, void *dhost)
1250 struct carp_if *cif = v;
1251 struct carp_softc *vh;
1252 u_int8_t *ena = dhost;
1254 if (ena[0] || ena[1] || ena[2] != 0x5e || ena[3] || ena[4] != 1)
1258 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list)
1259 if ((SC2IFP(vh)->if_flags & IFF_UP) &&
1260 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING) &&
1261 vh->sc_state == MASTER &&
1262 !bcmp(dhost, IF_LLADDR(vh->sc_ifp), ETHER_ADDR_LEN)) {
1264 return (SC2IFP(vh));
1272 carp_master_down(void *v)
1274 struct carp_softc *sc = v;
1277 carp_master_down_locked(sc);
1282 carp_master_down_locked(struct carp_softc *sc)
1285 CARP_SCLOCK_ASSERT(sc);
1287 switch (sc->sc_state) {
1289 printf("%s: master_down event in INIT state\n",
1290 SC2IFP(sc)->if_xname);
1295 carp_set_state(sc, MASTER);
1296 carp_send_ad_locked(sc);
1302 carp_setroute(sc, RTM_ADD);
1308 * When in backup state, af indicates whether to reset the master down timer
1309 * for v4 or v6. If it's set to zero, reset the ones which are already pending.
1312 carp_setrun(struct carp_softc *sc, sa_family_t af)
1316 if (sc->sc_carpdev == NULL) {
1317 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING;
1318 carp_set_state(sc, INIT);
1321 CARP_SCLOCK_ASSERT(sc);
1323 if (SC2IFP(sc)->if_flags & IFF_UP &&
1324 sc->sc_vhid > 0 && (sc->sc_naddrs || sc->sc_naddrs6))
1325 SC2IFP(sc)->if_drv_flags |= IFF_DRV_RUNNING;
1327 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING;
1328 carp_setroute(sc, RTM_DELETE);
1332 switch (sc->sc_state) {
1334 if (carp_opts[CARPCTL_PREEMPT] && !carp_suppress_preempt) {
1335 carp_send_ad_locked(sc);
1340 CARP_DEBUG("%s: INIT -> MASTER (preempting)\n",
1341 SC2IFP(sc)->if_xname);
1342 carp_set_state(sc, MASTER);
1343 carp_setroute(sc, RTM_ADD);
1345 CARP_DEBUG("%s: INIT -> BACKUP\n", SC2IFP(sc)->if_xname);
1346 carp_set_state(sc, BACKUP);
1347 carp_setroute(sc, RTM_DELETE);
1352 callout_stop(&sc->sc_ad_tmo);
1353 tv.tv_sec = 3 * sc->sc_advbase;
1354 tv.tv_usec = sc->sc_advskew * 1000000 / 256;
1358 callout_reset(&sc->sc_md_tmo, tvtohz(&tv),
1359 carp_master_down, sc);
1364 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv),
1365 carp_master_down, sc);
1370 callout_reset(&sc->sc_md_tmo, tvtohz(&tv),
1371 carp_master_down, sc);
1373 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv),
1374 carp_master_down, sc);
1379 tv.tv_sec = sc->sc_advbase;
1380 tv.tv_usec = sc->sc_advskew * 1000000 / 256;
1381 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv),
1388 carp_multicast_cleanup(struct carp_softc *sc)
1390 struct ip_moptions *imo = &sc->sc_imo;
1391 u_int16_t n = imo->imo_num_memberships;
1393 /* Clean up our own multicast memberships */
1395 if (imo->imo_membership[n] != NULL) {
1396 in_delmulti(imo->imo_membership[n]);
1397 imo->imo_membership[n] = NULL;
1400 imo->imo_num_memberships = 0;
1401 imo->imo_multicast_ifp = NULL;
1406 carp_multicast6_cleanup(struct carp_softc *sc)
1408 struct ip6_moptions *im6o = &sc->sc_im6o;
1410 while (!LIST_EMPTY(&im6o->im6o_memberships)) {
1411 struct in6_multi_mship *imm =
1412 LIST_FIRST(&im6o->im6o_memberships);
1414 LIST_REMOVE(imm, i6mm_chain);
1415 in6_leavegroup(imm);
1417 im6o->im6o_multicast_ifp = NULL;
1422 carp_set_addr(struct carp_softc *sc, struct sockaddr_in *sin)
1425 struct carp_if *cif;
1426 struct in_ifaddr *ia, *ia_if;
1427 struct ip_moptions *imo = &sc->sc_imo;
1428 struct in_addr addr;
1429 u_long iaddr = htonl(sin->sin_addr.s_addr);
1432 if (sin->sin_addr.s_addr == 0) {
1433 if (!(SC2IFP(sc)->if_flags & IFF_UP))
1434 carp_set_state(sc, INIT);
1436 SC2IFP(sc)->if_flags |= IFF_UP;
1441 /* we have to do it by hands to check we won't match on us */
1442 ia_if = NULL; own = 0;
1443 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) {
1444 /* and, yeah, we need a multicast-capable iface too */
1445 if (ia->ia_ifp != SC2IFP(sc) &&
1446 (ia->ia_ifp->if_flags & IFF_MULTICAST) &&
1447 (iaddr & ia->ia_subnetmask) == ia->ia_subnet) {
1450 if (sin->sin_addr.s_addr ==
1451 ia->ia_addr.sin_addr.s_addr)
1457 return (EADDRNOTAVAIL);
1462 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0 ||
1463 (imo->imo_multicast_ifp && imo->imo_multicast_ifp != ifp))
1464 return (EADDRNOTAVAIL);
1466 if (imo->imo_num_memberships == 0) {
1467 addr.s_addr = htonl(INADDR_CARP_GROUP);
1468 if ((imo->imo_membership[0] = in_addmulti(&addr, ifp)) == NULL)
1470 imo->imo_num_memberships++;
1471 imo->imo_multicast_ifp = ifp;
1472 imo->imo_multicast_ttl = CARP_DFLTTL;
1473 imo->imo_multicast_loop = 0;
1476 if (!ifp->if_carp) {
1478 MALLOC(cif, struct carp_if *, sizeof(*cif), M_CARP,
1484 if ((error = ifpromisc(ifp, 1))) {
1489 CARP_LOCK_INIT(cif);
1491 cif->vhif_ifp = ifp;
1492 TAILQ_INIT(&cif->vhif_vrs);
1496 struct carp_softc *vr;
1498 cif = (struct carp_if *)ifp->if_carp;
1500 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list)
1501 if (vr != sc && vr->sc_vhid == sc->sc_vhid) {
1508 sc->sc_carpdev = ifp;
1510 { /* XXX prevent endless loop if already in queue */
1511 struct carp_softc *vr, *after = NULL;
1513 cif = (struct carp_if *)ifp->if_carp;
1515 /* XXX: cif should not change, right? So we still hold the lock */
1516 CARP_LOCK_ASSERT(cif);
1518 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) {
1521 if (vr->sc_vhid < sc->sc_vhid)
1526 /* We're trying to keep things in order */
1527 if (after == NULL) {
1528 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list);
1530 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, sc, sc_list);
1537 SC2IFP(sc)->if_flags |= IFF_UP;
1540 carp_sc_state_locked(sc);
1548 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]);
1553 carp_del_addr(struct carp_softc *sc, struct sockaddr_in *sin)
1557 if (!--sc->sc_naddrs) {
1558 struct carp_if *cif = (struct carp_if *)sc->sc_carpdev->if_carp;
1559 struct ip_moptions *imo = &sc->sc_imo;
1562 callout_stop(&sc->sc_ad_tmo);
1563 SC2IFP(sc)->if_flags &= ~IFF_UP;
1564 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING;
1566 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]);
1567 imo->imo_multicast_ifp = NULL;
1568 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list);
1569 if (!--cif->vhif_nvrs) {
1570 sc->sc_carpdev->if_carp = NULL;
1571 CARP_LOCK_DESTROY(cif);
1572 FREE(cif, M_IFADDR);
1583 carp_set_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6)
1586 struct carp_if *cif;
1587 struct in6_ifaddr *ia, *ia_if;
1588 struct ip6_moptions *im6o = &sc->sc_im6o;
1589 struct in6_multi_mship *imm;
1590 struct in6_addr in6;
1593 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1594 if (!(SC2IFP(sc)->if_flags & IFF_UP))
1595 carp_set_state(sc, INIT);
1597 SC2IFP(sc)->if_flags |= IFF_UP;
1602 /* we have to do it by hands to check we won't match on us */
1603 ia_if = NULL; own = 0;
1604 for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1607 for (i = 0; i < 4; i++) {
1608 if ((sin6->sin6_addr.s6_addr32[i] &
1609 ia->ia_prefixmask.sin6_addr.s6_addr32[i]) !=
1610 (ia->ia_addr.sin6_addr.s6_addr32[i] &
1611 ia->ia_prefixmask.sin6_addr.s6_addr32[i]))
1614 /* and, yeah, we need a multicast-capable iface too */
1615 if (ia->ia_ifp != SC2IFP(sc) &&
1616 (ia->ia_ifp->if_flags & IFF_MULTICAST) &&
1620 if (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
1621 &ia->ia_addr.sin6_addr))
1627 return (EADDRNOTAVAIL);
1631 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0 ||
1632 (im6o->im6o_multicast_ifp && im6o->im6o_multicast_ifp != ifp))
1633 return (EADDRNOTAVAIL);
1635 if (!sc->sc_naddrs6) {
1636 im6o->im6o_multicast_ifp = ifp;
1638 /* join CARP multicast address */
1639 bzero(&in6, sizeof(in6));
1640 in6.s6_addr16[0] = htons(0xff02);
1641 in6.s6_addr8[15] = 0x12;
1642 if (in6_setscope(&in6, ifp, NULL) != 0)
1644 if ((imm = in6_joingroup(ifp, &in6, &error, 0)) == NULL)
1646 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1648 /* join solicited multicast address */
1649 bzero(&in6, sizeof(in6));
1650 in6.s6_addr16[0] = htons(0xff02);
1651 in6.s6_addr32[1] = 0;
1652 in6.s6_addr32[2] = htonl(1);
1653 in6.s6_addr32[3] = sin6->sin6_addr.s6_addr32[3];
1654 in6.s6_addr8[12] = 0xff;
1655 if (in6_setscope(&in6, ifp, NULL) != 0)
1657 if ((imm = in6_joingroup(ifp, &in6, &error, 0)) == NULL)
1659 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1662 if (!ifp->if_carp) {
1663 MALLOC(cif, struct carp_if *, sizeof(*cif), M_CARP,
1669 if ((error = ifpromisc(ifp, 1))) {
1674 CARP_LOCK_INIT(cif);
1676 cif->vhif_ifp = ifp;
1677 TAILQ_INIT(&cif->vhif_vrs);
1681 struct carp_softc *vr;
1683 cif = (struct carp_if *)ifp->if_carp;
1685 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list)
1686 if (vr != sc && vr->sc_vhid == sc->sc_vhid) {
1693 sc->sc_carpdev = ifp;
1695 { /* XXX prevent endless loop if already in queue */
1696 struct carp_softc *vr, *after = NULL;
1698 cif = (struct carp_if *)ifp->if_carp;
1699 CARP_LOCK_ASSERT(cif);
1701 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) {
1704 if (vr->sc_vhid < sc->sc_vhid)
1709 /* We're trying to keep things in order */
1710 if (after == NULL) {
1711 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list);
1713 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, sc, sc_list);
1720 SC2IFP(sc)->if_flags |= IFF_UP;
1723 carp_sc_state_locked(sc);
1731 /* clean up multicast memberships */
1732 if (!sc->sc_naddrs6) {
1733 while (!LIST_EMPTY(&im6o->im6o_memberships)) {
1734 imm = LIST_FIRST(&im6o->im6o_memberships);
1735 LIST_REMOVE(imm, i6mm_chain);
1736 in6_leavegroup(imm);
1743 carp_del_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6)
1747 if (!--sc->sc_naddrs6) {
1748 struct carp_if *cif = (struct carp_if *)sc->sc_carpdev->if_carp;
1749 struct ip6_moptions *im6o = &sc->sc_im6o;
1752 callout_stop(&sc->sc_ad_tmo);
1753 SC2IFP(sc)->if_flags &= ~IFF_UP;
1754 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING;
1756 while (!LIST_EMPTY(&im6o->im6o_memberships)) {
1757 struct in6_multi_mship *imm =
1758 LIST_FIRST(&im6o->im6o_memberships);
1760 LIST_REMOVE(imm, i6mm_chain);
1761 in6_leavegroup(imm);
1763 im6o->im6o_multicast_ifp = NULL;
1764 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list);
1765 if (!--cif->vhif_nvrs) {
1766 CARP_LOCK_DESTROY(cif);
1767 sc->sc_carpdev->if_carp = NULL;
1768 FREE(cif, M_IFADDR);
1778 carp_ioctl(struct ifnet *ifp, u_long cmd, caddr_t addr)
1780 struct carp_softc *sc = ifp->if_softc, *vr;
1781 struct carpreq carpr;
1784 struct ifaliasreq *ifra;
1785 int locked = 0, error = 0;
1787 ifa = (struct ifaddr *)addr;
1788 ifra = (struct ifaliasreq *)addr;
1789 ifr = (struct ifreq *)addr;
1793 switch (ifa->ifa_addr->sa_family) {
1796 SC2IFP(sc)->if_flags |= IFF_UP;
1797 bcopy(ifa->ifa_addr, ifa->ifa_dstaddr,
1798 sizeof(struct sockaddr));
1799 error = carp_set_addr(sc, satosin(ifa->ifa_addr));
1804 SC2IFP(sc)->if_flags |= IFF_UP;
1805 error = carp_set_addr6(sc, satosin6(ifa->ifa_addr));
1809 error = EAFNOSUPPORT;
1815 switch (ifa->ifa_addr->sa_family) {
1818 SC2IFP(sc)->if_flags |= IFF_UP;
1819 bcopy(ifa->ifa_addr, ifa->ifa_dstaddr,
1820 sizeof(struct sockaddr));
1821 error = carp_set_addr(sc, satosin(&ifra->ifra_addr));
1826 SC2IFP(sc)->if_flags |= IFF_UP;
1827 error = carp_set_addr6(sc, satosin6(&ifra->ifra_addr));
1831 error = EAFNOSUPPORT;
1837 switch (ifa->ifa_addr->sa_family) {
1840 error = carp_del_addr(sc, satosin(&ifra->ifra_addr));
1845 error = carp_del_addr6(sc, satosin6(&ifra->ifra_addr));
1849 error = EAFNOSUPPORT;
1855 if (sc->sc_carpdev) {
1859 if (sc->sc_state != INIT && !(ifr->ifr_flags & IFF_UP)) {
1860 callout_stop(&sc->sc_ad_tmo);
1861 callout_stop(&sc->sc_md_tmo);
1862 callout_stop(&sc->sc_md6_tmo);
1863 if (sc->sc_state == MASTER)
1864 carp_send_ad_locked(sc);
1865 carp_set_state(sc, INIT);
1867 } else if (sc->sc_state == INIT && (ifr->ifr_flags & IFF_UP)) {
1868 SC2IFP(sc)->if_flags |= IFF_UP;
1874 error = priv_check(curthread, PRIV_NETINET_CARP);
1877 if ((error = copyin(ifr->ifr_data, &carpr, sizeof carpr)))
1880 if (sc->sc_carpdev) {
1884 if (sc->sc_state != INIT && carpr.carpr_state != sc->sc_state) {
1885 switch (carpr.carpr_state) {
1887 callout_stop(&sc->sc_ad_tmo);
1888 carp_set_state(sc, BACKUP);
1890 carp_setroute(sc, RTM_DELETE);
1893 carp_master_down_locked(sc);
1899 if (carpr.carpr_vhid > 0) {
1900 if (carpr.carpr_vhid > 255) {
1904 if (sc->sc_carpdev) {
1905 struct carp_if *cif;
1906 cif = (struct carp_if *)sc->sc_carpdev->if_carp;
1907 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list)
1909 vr->sc_vhid == carpr.carpr_vhid)
1912 sc->sc_vhid = carpr.carpr_vhid;
1913 IF_LLADDR(sc->sc_ifp)[0] = 0;
1914 IF_LLADDR(sc->sc_ifp)[1] = 0;
1915 IF_LLADDR(sc->sc_ifp)[2] = 0x5e;
1916 IF_LLADDR(sc->sc_ifp)[3] = 0;
1917 IF_LLADDR(sc->sc_ifp)[4] = 1;
1918 IF_LLADDR(sc->sc_ifp)[5] = sc->sc_vhid;
1921 if (carpr.carpr_advbase > 0 || carpr.carpr_advskew > 0) {
1922 if (carpr.carpr_advskew >= 255) {
1926 if (carpr.carpr_advbase > 255) {
1930 sc->sc_advbase = carpr.carpr_advbase;
1931 sc->sc_advskew = carpr.carpr_advskew;
1934 bcopy(carpr.carpr_key, sc->sc_key, sizeof(sc->sc_key));
1944 /* XXX: lockless read */
1945 bzero(&carpr, sizeof(carpr));
1946 carpr.carpr_state = sc->sc_state;
1947 carpr.carpr_vhid = sc->sc_vhid;
1948 carpr.carpr_advbase = sc->sc_advbase;
1949 carpr.carpr_advskew = sc->sc_advskew;
1950 error = priv_check(curthread, PRIV_NETINET_CARP);
1952 bcopy(sc->sc_key, carpr.carpr_key,
1953 sizeof(carpr.carpr_key));
1954 error = copyout(&carpr, ifr->ifr_data, sizeof(carpr));
1964 carp_hmac_prepare(sc);
1970 * XXX: this is looutput. We should eventually use it from there.
1973 carp_looutput(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
1978 M_ASSERTPKTHDR(m); /* check if we have the packet header */
1980 if (rt && rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
1982 return (rt->rt_flags & RTF_BLACKHOLE ? 0 :
1983 rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
1987 ifp->if_obytes += m->m_pkthdr.len;
1989 /* BPF writes need to be handled specially. */
1990 if (dst->sa_family == AF_UNSPEC) {
1991 bcopy(dst->sa_data, &af, sizeof(af));
1992 dst->sa_family = af;
1996 switch (dst->sa_family) {
2003 printf("carp_looutput: af=%d unexpected\n", dst->sa_family);
2005 return (EAFNOSUPPORT);
2008 return(if_simloop(ifp, m, dst->sa_family, 0));
2012 * Start output on carp interface. This function should never be called.
2015 carp_start(struct ifnet *ifp)
2018 printf("%s: start called\n", ifp->if_xname);
2023 carp_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2027 struct carp_softc *sc;
2028 struct ifnet *carp_ifp;
2033 switch (sa->sa_family) {
2046 mtag = m_tag_find(m, PACKET_TAG_CARP, NULL);
2050 bcopy(mtag + 1, &carp_ifp, sizeof(struct ifnet *));
2051 sc = carp_ifp->if_softc;
2053 /* Set the source MAC address to Virtual Router MAC Address */
2054 switch (ifp->if_type) {
2057 struct ether_header *eh;
2059 eh = mtod(m, struct ether_header *);
2060 eh->ether_shost[0] = 0;
2061 eh->ether_shost[1] = 0;
2062 eh->ether_shost[2] = 0x5e;
2063 eh->ether_shost[3] = 0;
2064 eh->ether_shost[4] = 1;
2065 eh->ether_shost[5] = sc->sc_vhid;
2069 struct fddi_header *fh;
2071 fh = mtod(m, struct fddi_header *);
2072 fh->fddi_shost[0] = 0;
2073 fh->fddi_shost[1] = 0;
2074 fh->fddi_shost[2] = 0x5e;
2075 fh->fddi_shost[3] = 0;
2076 fh->fddi_shost[4] = 1;
2077 fh->fddi_shost[5] = sc->sc_vhid;
2080 case IFT_ISO88025: {
2081 struct iso88025_header *th;
2082 th = mtod(m, struct iso88025_header *);
2083 th->iso88025_shost[0] = 3;
2084 th->iso88025_shost[1] = 0;
2085 th->iso88025_shost[2] = 0x40 >> (sc->sc_vhid - 1);
2086 th->iso88025_shost[3] = 0x40000 >> (sc->sc_vhid - 1);
2087 th->iso88025_shost[4] = 0;
2088 th->iso88025_shost[5] = 0;
2092 printf("%s: carp is not supported for this interface type\n",
2094 return (EOPNOTSUPP);
2101 carp_set_state(struct carp_softc *sc, int state)
2105 CARP_SCLOCK_ASSERT(sc);
2107 if (sc->sc_state == state)
2110 sc->sc_state = state;
2113 SC2IFP(sc)->if_link_state = LINK_STATE_DOWN;
2116 SC2IFP(sc)->if_link_state = LINK_STATE_UP;
2119 SC2IFP(sc)->if_link_state = LINK_STATE_UNKNOWN;
2122 rt_ifmsg(SC2IFP(sc));
2126 carp_carpdev_state(void *v)
2128 struct carp_if *cif = v;
2131 carp_carpdev_state_locked(cif);
2136 carp_carpdev_state_locked(struct carp_if *cif)
2138 struct carp_softc *sc;
2140 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list)
2141 carp_sc_state_locked(sc);
2145 carp_sc_state_locked(struct carp_softc *sc)
2147 CARP_SCLOCK_ASSERT(sc);
2149 if (sc->sc_carpdev->if_link_state != LINK_STATE_UP ||
2150 !(sc->sc_carpdev->if_flags & IFF_UP)) {
2151 sc->sc_flags_backup = SC2IFP(sc)->if_flags;
2152 SC2IFP(sc)->if_flags &= ~IFF_UP;
2153 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING;
2154 callout_stop(&sc->sc_ad_tmo);
2155 callout_stop(&sc->sc_md_tmo);
2156 callout_stop(&sc->sc_md6_tmo);
2157 carp_set_state(sc, INIT);
2159 if (!sc->sc_suppress) {
2160 carp_suppress_preempt++;
2161 if (carp_suppress_preempt == 1) {
2167 sc->sc_suppress = 1;
2169 SC2IFP(sc)->if_flags |= sc->sc_flags_backup;
2170 carp_set_state(sc, INIT);
2172 if (sc->sc_suppress)
2173 carp_suppress_preempt--;
2174 sc->sc_suppress = 0;
2181 carp_modevent(module_t mod, int type, void *data)
2185 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2186 carp_ifdetach, NULL, EVENTHANDLER_PRI_ANY);
2187 if (if_detach_event_tag == NULL)
2189 mtx_init(&carp_mtx, "carp_mtx", NULL, MTX_DEF);
2190 LIST_INIT(&carpif_list);
2191 if_clone_attach(&carp_cloner);
2195 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2196 if_clone_detach(&carp_cloner);
2197 mtx_destroy(&carp_mtx);
2207 static moduledata_t carp_mod = {
2213 DECLARE_MODULE(carp, carp_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);