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 #include <sys/cdefs.h>
28 __FBSDID("$FreeBSD$");
33 #include "opt_inet6.h"
35 #include <sys/types.h>
36 #include <sys/param.h>
37 #include <sys/systm.h>
39 #include <sys/kernel.h>
40 #include <sys/limits.h>
41 #include <sys/malloc.h>
43 #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>
53 #include <sys/socket.h>
54 #include <sys/vnode.h>
55 #include <sys/vimage.h>
57 #include <machine/stdarg.h>
60 #include <net/ethernet.h>
62 #include <net/iso88025.h>
64 #include <net/if_clone.h>
65 #include <net/if_dl.h>
66 #include <net/if_types.h>
67 #include <net/route.h>
70 #include <netinet/in.h>
71 #include <netinet/in_var.h>
72 #include <netinet/in_systm.h>
73 #include <netinet/ip.h>
74 #include <netinet/ip_var.h>
75 #include <netinet/if_ether.h>
76 #include <machine/in_cksum.h>
77 #include <netinet/vinet.h>
81 #include <netinet/icmp6.h>
82 #include <netinet/ip6.h>
83 #include <netinet6/ip6_var.h>
84 #include <netinet6/scope6_var.h>
85 #include <netinet6/nd6.h>
86 #include <netinet6/vinet6.h>
89 #include <crypto/sha1.h>
90 #include <netinet/ip_carp.h>
92 #define CARP_IFNAME "carp"
93 static MALLOC_DEFINE(M_CARP, "CARP", "CARP interfaces");
94 SYSCTL_DECL(_net_inet_carp);
97 struct ifnet *sc_ifp; /* Interface clue */
98 struct ifnet *sc_carpdev; /* Pointer to parent interface */
99 struct in_ifaddr *sc_ia; /* primary iface address */
100 struct ip_moptions sc_imo;
102 struct in6_ifaddr *sc_ia6; /* primary iface address v6 */
103 struct ip6_moptions sc_im6o;
105 TAILQ_ENTRY(carp_softc) sc_list;
107 enum { INIT = 0, BACKUP, MASTER } sc_state;
112 int sc_sendad_errors;
113 #define CARP_SENDAD_MAX_ERRORS 3
114 int sc_sendad_success;
115 #define CARP_SENDAD_MIN_SUCCESS 3
121 int sc_advbase; /* seconds */
123 u_int64_t sc_counter;
126 #define CARP_HMAC_PAD 64
127 unsigned char sc_key[CARP_KEY_LEN];
128 unsigned char sc_pad[CARP_HMAC_PAD];
131 struct callout sc_ad_tmo; /* advertisement timeout */
132 struct callout sc_md_tmo; /* master down timeout */
133 struct callout sc_md6_tmo; /* master down timeout */
135 LIST_ENTRY(carp_softc) sc_next; /* Interface clue */
137 #define SC2IFP(sc) ((sc)->sc_ifp)
139 int carp_suppress_preempt = 0;
140 int carp_opts[CARPCTL_MAXID] = { 0, 1, 0, 1, 0, 0 }; /* XXX for now */
141 SYSCTL_INT(_net_inet_carp, CARPCTL_ALLOW, allow, CTLFLAG_RW,
142 &carp_opts[CARPCTL_ALLOW], 0, "Accept incoming CARP packets");
143 SYSCTL_INT(_net_inet_carp, CARPCTL_PREEMPT, preempt, CTLFLAG_RW,
144 &carp_opts[CARPCTL_PREEMPT], 0, "high-priority backup preemption mode");
145 SYSCTL_INT(_net_inet_carp, CARPCTL_LOG, log, CTLFLAG_RW,
146 &carp_opts[CARPCTL_LOG], 0, "log bad carp packets");
147 SYSCTL_INT(_net_inet_carp, CARPCTL_ARPBALANCE, arpbalance, CTLFLAG_RW,
148 &carp_opts[CARPCTL_ARPBALANCE], 0, "balance arp responses");
149 SYSCTL_INT(_net_inet_carp, OID_AUTO, suppress_preempt, CTLFLAG_RD,
150 &carp_suppress_preempt, 0, "Preemption is suppressed");
152 struct carpstats carpstats;
153 SYSCTL_STRUCT(_net_inet_carp, CARPCTL_STATS, stats, CTLFLAG_RW,
154 &carpstats, carpstats,
155 "CARP statistics (struct carpstats, netinet/ip_carp.h)");
158 TAILQ_HEAD(, carp_softc) vhif_vrs;
161 struct ifnet *vhif_ifp;
165 /* Get carp_if from softc. Valid after carp_set_addr{,6}. */
166 #define SC2CIF(sc) ((struct carp_if *)(sc)->sc_carpdev->if_carp)
168 /* lock per carp_if queue */
169 #define CARP_LOCK_INIT(cif) mtx_init(&(cif)->vhif_mtx, "carp_if", \
171 #define CARP_LOCK_DESTROY(cif) mtx_destroy(&(cif)->vhif_mtx)
172 #define CARP_LOCK_ASSERT(cif) mtx_assert(&(cif)->vhif_mtx, MA_OWNED)
173 #define CARP_LOCK(cif) mtx_lock(&(cif)->vhif_mtx)
174 #define CARP_UNLOCK(cif) mtx_unlock(&(cif)->vhif_mtx)
176 #define CARP_SCLOCK(sc) mtx_lock(&SC2CIF(sc)->vhif_mtx)
177 #define CARP_SCUNLOCK(sc) mtx_unlock(&SC2CIF(sc)->vhif_mtx)
178 #define CARP_SCLOCK_ASSERT(sc) mtx_assert(&SC2CIF(sc)->vhif_mtx, MA_OWNED)
180 #define CARP_LOG(...) do { \
181 if (carp_opts[CARPCTL_LOG] > 0) \
182 log(LOG_INFO, __VA_ARGS__); \
185 #define CARP_DEBUG(...) do { \
186 if (carp_opts[CARPCTL_LOG] > 1) \
187 log(LOG_DEBUG, __VA_ARGS__); \
190 static void carp_hmac_prepare(struct carp_softc *);
191 static void carp_hmac_generate(struct carp_softc *, u_int32_t *,
193 static int carp_hmac_verify(struct carp_softc *, u_int32_t *,
195 static void carp_setroute(struct carp_softc *, int);
196 static void carp_input_c(struct mbuf *, struct carp_header *, sa_family_t);
197 static int carp_clone_create(struct if_clone *, int, caddr_t);
198 static void carp_clone_destroy(struct ifnet *);
199 static void carpdetach(struct carp_softc *, int);
200 static int carp_prepare_ad(struct mbuf *, struct carp_softc *,
201 struct carp_header *);
202 static void carp_send_ad_all(void);
203 static void carp_send_ad(void *);
204 static void carp_send_ad_locked(struct carp_softc *);
205 static void carp_send_arp(struct carp_softc *);
206 static void carp_master_down(void *);
207 static void carp_master_down_locked(struct carp_softc *);
208 static int carp_ioctl(struct ifnet *, u_long, caddr_t);
209 static int carp_looutput(struct ifnet *, struct mbuf *, struct sockaddr *,
211 static void carp_start(struct ifnet *);
212 static void carp_setrun(struct carp_softc *, sa_family_t);
213 static void carp_set_state(struct carp_softc *, int);
214 static int carp_addrcount(struct carp_if *, struct in_ifaddr *, int);
215 enum { CARP_COUNT_MASTER, CARP_COUNT_RUNNING };
217 static void carp_multicast_cleanup(struct carp_softc *);
218 static int carp_set_addr(struct carp_softc *, struct sockaddr_in *);
219 static int carp_del_addr(struct carp_softc *, struct sockaddr_in *);
220 static void carp_carpdev_state_locked(struct carp_if *);
221 static void carp_sc_state_locked(struct carp_softc *);
223 static void carp_send_na(struct carp_softc *);
224 static int carp_set_addr6(struct carp_softc *, struct sockaddr_in6 *);
225 static int carp_del_addr6(struct carp_softc *, struct sockaddr_in6 *);
226 static void carp_multicast6_cleanup(struct carp_softc *);
229 static LIST_HEAD(, carp_softc) carpif_list;
230 static struct mtx carp_mtx;
231 IFC_SIMPLE_DECLARE(carp, 0);
233 static eventhandler_tag if_detach_event_tag;
235 static __inline u_int16_t
236 carp_cksum(struct mbuf *m, int len)
238 return (in_cksum(m, len));
242 carp_hmac_prepare(struct carp_softc *sc)
244 u_int8_t version = CARP_VERSION, type = CARP_ADVERTISEMENT;
245 u_int8_t vhid = sc->sc_vhid & 0xff;
249 struct in_addr last, cur, in;
252 struct in6_addr last6, cur6, in6;
258 /* XXX: possible race here */
260 /* compute ipad from key */
261 bzero(sc->sc_pad, sizeof(sc->sc_pad));
262 bcopy(sc->sc_key, sc->sc_pad, sizeof(sc->sc_key));
263 for (i = 0; i < sizeof(sc->sc_pad); i++)
264 sc->sc_pad[i] ^= 0x36;
266 /* precompute first part of inner hash */
267 SHA1Init(&sc->sc_sha1);
268 SHA1Update(&sc->sc_sha1, sc->sc_pad, sizeof(sc->sc_pad));
269 SHA1Update(&sc->sc_sha1, (void *)&version, sizeof(version));
270 SHA1Update(&sc->sc_sha1, (void *)&type, sizeof(type));
271 SHA1Update(&sc->sc_sha1, (void *)&vhid, sizeof(vhid));
277 cur.s_addr = 0xffffffff;
278 IF_ADDR_LOCK(SC2IFP(sc));
279 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
280 in.s_addr = ifatoia(ifa)->ia_addr.sin_addr.s_addr;
281 if (ifa->ifa_addr->sa_family == AF_INET &&
282 ntohl(in.s_addr) > ntohl(last.s_addr) &&
283 ntohl(in.s_addr) < ntohl(cur.s_addr)) {
284 cur.s_addr = in.s_addr;
288 IF_ADDR_UNLOCK(SC2IFP(sc));
290 SHA1Update(&sc->sc_sha1, (void *)&cur, sizeof(cur));
294 memset(&cur6, 0, sizeof(cur6));
298 memset(&cur6, 0xff, sizeof(cur6));
299 IF_ADDR_LOCK(SC2IFP(sc));
300 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
301 in6 = ifatoia6(ifa)->ia_addr.sin6_addr;
302 if (IN6_IS_SCOPE_EMBED(&in6))
303 in6.s6_addr16[1] = 0;
304 if (ifa->ifa_addr->sa_family == AF_INET6 &&
305 memcmp(&in6, &last6, sizeof(in6)) > 0 &&
306 memcmp(&in6, &cur6, sizeof(in6)) < 0) {
311 IF_ADDR_UNLOCK(SC2IFP(sc));
313 SHA1Update(&sc->sc_sha1, (void *)&cur6, sizeof(cur6));
317 /* convert ipad to opad */
318 for (i = 0; i < sizeof(sc->sc_pad); i++)
319 sc->sc_pad[i] ^= 0x36 ^ 0x5c;
326 carp_hmac_generate(struct carp_softc *sc, u_int32_t counter[2],
327 unsigned char md[20])
331 /* fetch first half of inner hash */
332 bcopy(&sc->sc_sha1, &sha1ctx, sizeof(sha1ctx));
334 SHA1Update(&sha1ctx, (void *)counter, sizeof(sc->sc_counter));
335 SHA1Final(md, &sha1ctx);
339 SHA1Update(&sha1ctx, sc->sc_pad, sizeof(sc->sc_pad));
340 SHA1Update(&sha1ctx, md, 20);
341 SHA1Final(md, &sha1ctx);
345 carp_hmac_verify(struct carp_softc *sc, u_int32_t counter[2],
346 unsigned char md[20])
348 unsigned char md2[20];
350 CARP_SCLOCK_ASSERT(sc);
352 carp_hmac_generate(sc, counter, md2);
354 return (bcmp(md, md2, sizeof(md2)));
358 carp_setroute(struct carp_softc *sc, int cmd)
364 CARP_SCLOCK_ASSERT(sc);
367 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
368 if (ifa->ifa_addr->sa_family == AF_INET &&
369 sc->sc_carpdev != NULL) {
370 int count = carp_addrcount(
371 (struct carp_if *)sc->sc_carpdev->if_carp,
372 ifatoia(ifa), CARP_COUNT_MASTER);
374 if ((cmd == RTM_ADD && count == 1) ||
375 (cmd == RTM_DELETE && count == 0))
376 rtinit(ifa, cmd, RTF_UP | RTF_HOST);
383 carp_clone_create(struct if_clone *ifc, int unit, caddr_t params)
386 struct carp_softc *sc;
389 sc = malloc(sizeof(*sc), M_CARP, M_WAITOK|M_ZERO);
390 ifp = SC2IFP(sc) = if_alloc(IFT_ETHER);
396 sc->sc_flags_backup = 0;
398 sc->sc_advbase = CARP_DFLTINTV;
399 sc->sc_vhid = -1; /* required setting */
401 sc->sc_init_counter = 1;
402 sc->sc_naddrs = sc->sc_naddrs6 = 0; /* M_ZERO? */
404 sc->sc_im6o.im6o_multicast_hlim = CARP_DFLTTL;
406 sc->sc_imo.imo_membership = (struct in_multi **)malloc(
407 (sizeof(struct in_multi *) * IP_MIN_MEMBERSHIPS), M_CARP,
409 sc->sc_imo.imo_mfilters = NULL;
410 sc->sc_imo.imo_max_memberships = IP_MIN_MEMBERSHIPS;
411 sc->sc_imo.imo_multicast_vif = -1;
413 callout_init(&sc->sc_ad_tmo, CALLOUT_MPSAFE);
414 callout_init(&sc->sc_md_tmo, CALLOUT_MPSAFE);
415 callout_init(&sc->sc_md6_tmo, CALLOUT_MPSAFE);
418 if_initname(ifp, CARP_IFNAME, unit);
419 ifp->if_mtu = ETHERMTU;
420 ifp->if_flags = IFF_LOOPBACK;
421 ifp->if_ioctl = carp_ioctl;
422 ifp->if_output = carp_looutput;
423 ifp->if_start = carp_start;
424 ifp->if_type = IFT_CARP;
425 ifp->if_snd.ifq_maxlen = ifqmaxlen;
428 bpfattach(SC2IFP(sc), DLT_NULL, sizeof(u_int32_t));
430 LIST_INSERT_HEAD(&carpif_list, sc, sc_next);
431 mtx_unlock(&carp_mtx);
436 carp_clone_destroy(struct ifnet *ifp)
438 struct carp_softc *sc = ifp->if_softc;
442 carpdetach(sc, 1); /* Returns unlocked. */
445 LIST_REMOVE(sc, sc_next);
446 mtx_unlock(&carp_mtx);
449 if_free_type(ifp, IFT_ETHER);
450 free(sc->sc_imo.imo_membership, M_CARP);
455 * This function can be called on CARP interface destroy path,
456 * and in case of the removal of the underlying interface as
457 * well. We differentiate these two cases. In the latter case
458 * we do not cleanup our multicast memberships, since they
459 * are already freed. Also, in the latter case we do not
460 * release the lock on return, because the function will be
461 * called once more, for another CARP instance on the same
465 carpdetach(struct carp_softc *sc, int unlock)
469 callout_stop(&sc->sc_ad_tmo);
470 callout_stop(&sc->sc_md_tmo);
471 callout_stop(&sc->sc_md6_tmo);
474 carp_suppress_preempt--;
477 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS)
478 carp_suppress_preempt--;
479 sc->sc_sendad_errors = 0;
481 carp_set_state(sc, INIT);
482 SC2IFP(sc)->if_flags &= ~IFF_UP;
485 carp_multicast_cleanup(sc);
487 carp_multicast6_cleanup(sc);
490 if (sc->sc_carpdev != NULL) {
491 cif = (struct carp_if *)sc->sc_carpdev->if_carp;
492 CARP_LOCK_ASSERT(cif);
493 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list);
494 if (!--cif->vhif_nvrs) {
495 ifpromisc(sc->sc_carpdev, 0);
496 sc->sc_carpdev->if_carp = NULL;
497 CARP_LOCK_DESTROY(cif);
501 sc->sc_carpdev = NULL;
505 /* Detach an interface from the carp. */
507 carp_ifdetach(void *arg __unused, struct ifnet *ifp)
509 struct carp_if *cif = (struct carp_if *)ifp->if_carp;
510 struct carp_softc *sc, *nextsc;
516 * XXX: At the end of for() cycle the lock will be destroyed.
519 for (sc = TAILQ_FIRST(&cif->vhif_vrs); sc; sc = nextsc) {
520 nextsc = TAILQ_NEXT(sc, sc_list);
526 * process input packet.
527 * we have rearranged checks order compared to the rfc,
528 * but it seems more efficient this way or not possible otherwise.
531 carp_input(struct mbuf *m, int hlen)
533 struct ip *ip = mtod(m, struct ip *);
534 struct carp_header *ch;
537 CARPSTATS_INC(carps_ipackets);
539 if (!carp_opts[CARPCTL_ALLOW]) {
544 /* check if received on a valid carp interface */
545 if (m->m_pkthdr.rcvif->if_carp == NULL) {
546 CARPSTATS_INC(carps_badif);
547 CARP_LOG("carp_input: packet received on non-carp "
549 m->m_pkthdr.rcvif->if_xname);
554 /* verify that the IP TTL is 255. */
555 if (ip->ip_ttl != CARP_DFLTTL) {
556 CARPSTATS_INC(carps_badttl);
557 CARP_LOG("carp_input: received ttl %d != 255i on %s\n",
559 m->m_pkthdr.rcvif->if_xname);
564 iplen = ip->ip_hl << 2;
566 if (m->m_pkthdr.len < iplen + sizeof(*ch)) {
567 CARPSTATS_INC(carps_badlen);
568 CARP_LOG("carp_input: received len %zd < "
569 "sizeof(struct carp_header)\n",
570 m->m_len - sizeof(struct ip));
575 if (iplen + sizeof(*ch) < m->m_len) {
576 if ((m = m_pullup(m, iplen + sizeof(*ch))) == NULL) {
577 CARPSTATS_INC(carps_hdrops);
578 CARP_LOG("carp_input: pullup failed\n");
581 ip = mtod(m, struct ip *);
583 ch = (struct carp_header *)((char *)ip + iplen);
586 * verify that the received packet length is
587 * equal to the CARP header
589 len = iplen + sizeof(*ch);
590 if (len > m->m_pkthdr.len) {
591 CARPSTATS_INC(carps_badlen);
592 CARP_LOG("carp_input: packet too short %d on %s\n",
594 m->m_pkthdr.rcvif->if_xname);
599 if ((m = m_pullup(m, len)) == NULL) {
600 CARPSTATS_INC(carps_hdrops);
603 ip = mtod(m, struct ip *);
604 ch = (struct carp_header *)((char *)ip + iplen);
606 /* verify the CARP checksum */
608 if (carp_cksum(m, len - iplen)) {
609 CARPSTATS_INC(carps_badsum);
610 CARP_LOG("carp_input: checksum failed on %s\n",
611 m->m_pkthdr.rcvif->if_xname);
617 carp_input_c(m, ch, AF_INET);
622 carp6_input(struct mbuf **mp, int *offp, int proto)
624 struct mbuf *m = *mp;
625 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
626 struct carp_header *ch;
629 CARPSTATS_INC(carps_ipackets6);
631 if (!carp_opts[CARPCTL_ALLOW]) {
633 return (IPPROTO_DONE);
636 /* check if received on a valid carp interface */
637 if (m->m_pkthdr.rcvif->if_carp == NULL) {
638 CARPSTATS_INC(carps_badif);
639 CARP_LOG("carp6_input: packet received on non-carp "
641 m->m_pkthdr.rcvif->if_xname);
643 return (IPPROTO_DONE);
646 /* verify that the IP TTL is 255 */
647 if (ip6->ip6_hlim != CARP_DFLTTL) {
648 CARPSTATS_INC(carps_badttl);
649 CARP_LOG("carp6_input: received ttl %d != 255 on %s\n",
651 m->m_pkthdr.rcvif->if_xname);
653 return (IPPROTO_DONE);
656 /* verify that we have a complete carp packet */
658 IP6_EXTHDR_GET(ch, struct carp_header *, m, *offp, sizeof(*ch));
660 CARPSTATS_INC(carps_badlen);
661 CARP_LOG("carp6_input: packet size %u too small\n", len);
662 return (IPPROTO_DONE);
666 /* verify the CARP checksum */
668 if (carp_cksum(m, sizeof(*ch))) {
669 CARPSTATS_INC(carps_badsum);
670 CARP_LOG("carp6_input: checksum failed, on %s\n",
671 m->m_pkthdr.rcvif->if_xname);
673 return (IPPROTO_DONE);
677 carp_input_c(m, ch, AF_INET6);
678 return (IPPROTO_DONE);
683 carp_input_c(struct mbuf *m, struct carp_header *ch, sa_family_t af)
685 struct ifnet *ifp = m->m_pkthdr.rcvif;
686 struct carp_softc *sc;
687 u_int64_t tmp_counter;
688 struct timeval sc_tv, ch_tv;
690 /* verify that the VHID is valid on the receiving interface */
691 CARP_LOCK(ifp->if_carp);
692 TAILQ_FOREACH(sc, &((struct carp_if *)ifp->if_carp)->vhif_vrs, sc_list)
693 if (sc->sc_vhid == ch->carp_vhid)
696 if (!sc || !((SC2IFP(sc)->if_flags & IFF_UP) &&
697 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING))) {
698 CARPSTATS_INC(carps_badvhid);
699 CARP_UNLOCK(ifp->if_carp);
704 getmicrotime(&SC2IFP(sc)->if_lastchange);
705 SC2IFP(sc)->if_ipackets++;
706 SC2IFP(sc)->if_ibytes += m->m_pkthdr.len;
708 if (bpf_peers_present(SC2IFP(sc)->if_bpf)) {
709 struct ip *ip = mtod(m, struct ip *);
712 /* BPF wants net byte order */
713 ip->ip_len = htons(ip->ip_len + (ip->ip_hl << 2));
714 ip->ip_off = htons(ip->ip_off);
715 bpf_mtap2(SC2IFP(sc)->if_bpf, &af1, sizeof(af1), m);
718 /* verify the CARP version. */
719 if (ch->carp_version != CARP_VERSION) {
720 CARPSTATS_INC(carps_badver);
721 SC2IFP(sc)->if_ierrors++;
722 CARP_UNLOCK(ifp->if_carp);
723 CARP_LOG("%s; invalid version %d\n",
724 SC2IFP(sc)->if_xname,
730 /* verify the hash */
731 if (carp_hmac_verify(sc, ch->carp_counter, ch->carp_md)) {
732 CARPSTATS_INC(carps_badauth);
733 SC2IFP(sc)->if_ierrors++;
734 CARP_UNLOCK(ifp->if_carp);
735 CARP_LOG("%s: incorrect hash\n", SC2IFP(sc)->if_xname);
740 tmp_counter = ntohl(ch->carp_counter[0]);
741 tmp_counter = tmp_counter<<32;
742 tmp_counter += ntohl(ch->carp_counter[1]);
744 /* XXX Replay protection goes here */
746 sc->sc_init_counter = 0;
747 sc->sc_counter = tmp_counter;
749 sc_tv.tv_sec = sc->sc_advbase;
750 if (carp_suppress_preempt && sc->sc_advskew < 240)
751 sc_tv.tv_usec = 240 * 1000000 / 256;
753 sc_tv.tv_usec = sc->sc_advskew * 1000000 / 256;
754 ch_tv.tv_sec = ch->carp_advbase;
755 ch_tv.tv_usec = ch->carp_advskew * 1000000 / 256;
757 switch (sc->sc_state) {
762 * If we receive an advertisement from a master who's going to
763 * be more frequent than us, go into BACKUP state.
765 if (timevalcmp(&sc_tv, &ch_tv, >) ||
766 timevalcmp(&sc_tv, &ch_tv, ==)) {
767 callout_stop(&sc->sc_ad_tmo);
768 CARP_DEBUG("%s: MASTER -> BACKUP "
769 "(more frequent advertisement received)\n",
770 SC2IFP(sc)->if_xname);
771 carp_set_state(sc, BACKUP);
773 carp_setroute(sc, RTM_DELETE);
778 * If we're pre-empting masters who advertise slower than us,
779 * and this one claims to be slower, treat him as down.
781 if (carp_opts[CARPCTL_PREEMPT] &&
782 timevalcmp(&sc_tv, &ch_tv, <)) {
783 CARP_DEBUG("%s: BACKUP -> MASTER "
784 "(preempting a slower master)\n",
785 SC2IFP(sc)->if_xname);
786 carp_master_down_locked(sc);
791 * If the master is going to advertise at such a low frequency
792 * that he's guaranteed to time out, we'd might as well just
793 * treat him as timed out now.
795 sc_tv.tv_sec = sc->sc_advbase * 3;
796 if (timevalcmp(&sc_tv, &ch_tv, <)) {
797 CARP_DEBUG("%s: BACKUP -> MASTER "
798 "(master timed out)\n",
799 SC2IFP(sc)->if_xname);
800 carp_master_down_locked(sc);
805 * Otherwise, we reset the counter and wait for the next
812 CARP_UNLOCK(ifp->if_carp);
819 carp_prepare_ad(struct mbuf *m, struct carp_softc *sc, struct carp_header *ch)
822 struct ifnet *ifp = SC2IFP(sc);
824 if (sc->sc_init_counter) {
825 /* this could also be seconds since unix epoch */
826 sc->sc_counter = arc4random();
827 sc->sc_counter = sc->sc_counter << 32;
828 sc->sc_counter += arc4random();
832 ch->carp_counter[0] = htonl((sc->sc_counter>>32)&0xffffffff);
833 ch->carp_counter[1] = htonl(sc->sc_counter&0xffffffff);
835 carp_hmac_generate(sc, ch->carp_counter, ch->carp_md);
837 /* Tag packet for carp_output */
838 mtag = m_tag_get(PACKET_TAG_CARP, sizeof(struct ifnet *), M_NOWAIT);
841 SC2IFP(sc)->if_oerrors++;
844 bcopy(&ifp, (caddr_t)(mtag + 1), sizeof(struct ifnet *));
845 m_tag_prepend(m, mtag);
851 carp_send_ad_all(void)
853 struct carp_softc *sc;
856 LIST_FOREACH(sc, &carpif_list, sc_next) {
857 if (sc->sc_carpdev == NULL)
860 if ((SC2IFP(sc)->if_flags & IFF_UP) &&
861 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING) &&
862 sc->sc_state == MASTER)
863 carp_send_ad_locked(sc);
866 mtx_unlock(&carp_mtx);
870 carp_send_ad(void *v)
872 struct carp_softc *sc = v;
875 carp_send_ad_locked(sc);
880 carp_send_ad_locked(struct carp_softc *sc)
882 struct carp_header ch;
884 struct carp_header *ch_ptr;
886 int len, advbase, advskew;
888 CARP_SCLOCK_ASSERT(sc);
890 /* bow out if we've lost our UPness or RUNNINGuiness */
891 if (!((SC2IFP(sc)->if_flags & IFF_UP) &&
892 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING))) {
896 advbase = sc->sc_advbase;
897 if (!carp_suppress_preempt || sc->sc_advskew > 240)
898 advskew = sc->sc_advskew;
902 tv.tv_usec = advskew * 1000000 / 256;
905 ch.carp_version = CARP_VERSION;
906 ch.carp_type = CARP_ADVERTISEMENT;
907 ch.carp_vhid = sc->sc_vhid;
908 ch.carp_advbase = advbase;
909 ch.carp_advskew = advskew;
910 ch.carp_authlen = 7; /* XXX DEFINE */
911 ch.carp_pad1 = 0; /* must be zero */
915 INIT_VNET_INET(curvnet);
919 MGETHDR(m, M_DONTWAIT, MT_HEADER);
921 SC2IFP(sc)->if_oerrors++;
922 CARPSTATS_INC(carps_onomem);
923 /* XXX maybe less ? */
924 if (advbase != 255 || advskew != 255)
925 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv),
929 len = sizeof(*ip) + sizeof(ch);
930 m->m_pkthdr.len = len;
931 m->m_pkthdr.rcvif = NULL;
933 MH_ALIGN(m, m->m_len);
934 m->m_flags |= M_MCAST;
935 ip = mtod(m, struct ip *);
936 ip->ip_v = IPVERSION;
937 ip->ip_hl = sizeof(*ip) >> 2;
938 ip->ip_tos = IPTOS_LOWDELAY;
940 ip->ip_id = ip_newid();
942 ip->ip_ttl = CARP_DFLTTL;
943 ip->ip_p = IPPROTO_CARP;
945 ip->ip_src.s_addr = sc->sc_ia->ia_addr.sin_addr.s_addr;
946 ip->ip_dst.s_addr = htonl(INADDR_CARP_GROUP);
948 ch_ptr = (struct carp_header *)(&ip[1]);
949 bcopy(&ch, ch_ptr, sizeof(ch));
950 if (carp_prepare_ad(m, sc, ch_ptr))
953 m->m_data += sizeof(*ip);
954 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip));
955 m->m_data -= sizeof(*ip);
957 getmicrotime(&SC2IFP(sc)->if_lastchange);
958 SC2IFP(sc)->if_opackets++;
959 SC2IFP(sc)->if_obytes += len;
960 CARPSTATS_INC(carps_opackets);
962 if (ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, NULL)) {
963 SC2IFP(sc)->if_oerrors++;
964 if (sc->sc_sendad_errors < INT_MAX)
965 sc->sc_sendad_errors++;
966 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) {
967 carp_suppress_preempt++;
968 if (carp_suppress_preempt == 1) {
974 sc->sc_sendad_success = 0;
976 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) {
977 if (++sc->sc_sendad_success >=
978 CARP_SENDAD_MIN_SUCCESS) {
979 carp_suppress_preempt--;
980 sc->sc_sendad_errors = 0;
983 sc->sc_sendad_errors = 0;
991 MGETHDR(m, M_DONTWAIT, MT_HEADER);
993 SC2IFP(sc)->if_oerrors++;
994 CARPSTATS_INC(carps_onomem);
995 /* XXX maybe less ? */
996 if (advbase != 255 || advskew != 255)
997 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv),
1001 len = sizeof(*ip6) + sizeof(ch);
1002 m->m_pkthdr.len = len;
1003 m->m_pkthdr.rcvif = NULL;
1005 MH_ALIGN(m, m->m_len);
1006 m->m_flags |= M_MCAST;
1007 ip6 = mtod(m, struct ip6_hdr *);
1008 bzero(ip6, sizeof(*ip6));
1009 ip6->ip6_vfc |= IPV6_VERSION;
1010 ip6->ip6_hlim = CARP_DFLTTL;
1011 ip6->ip6_nxt = IPPROTO_CARP;
1012 bcopy(&sc->sc_ia6->ia_addr.sin6_addr, &ip6->ip6_src,
1013 sizeof(struct in6_addr));
1014 /* set the multicast destination */
1016 ip6->ip6_dst.s6_addr16[0] = htons(0xff02);
1017 ip6->ip6_dst.s6_addr8[15] = 0x12;
1018 if (in6_setscope(&ip6->ip6_dst, sc->sc_carpdev, NULL) != 0) {
1019 SC2IFP(sc)->if_oerrors++;
1021 CARP_LOG("%s: in6_setscope failed\n", __func__);
1025 ch_ptr = (struct carp_header *)(&ip6[1]);
1026 bcopy(&ch, ch_ptr, sizeof(ch));
1027 if (carp_prepare_ad(m, sc, ch_ptr))
1030 m->m_data += sizeof(*ip6);
1031 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip6));
1032 m->m_data -= sizeof(*ip6);
1034 getmicrotime(&SC2IFP(sc)->if_lastchange);
1035 SC2IFP(sc)->if_opackets++;
1036 SC2IFP(sc)->if_obytes += len;
1037 CARPSTATS_INC(carps_opackets6);
1039 if (ip6_output(m, NULL, NULL, 0, &sc->sc_im6o, NULL, NULL)) {
1040 SC2IFP(sc)->if_oerrors++;
1041 if (sc->sc_sendad_errors < INT_MAX)
1042 sc->sc_sendad_errors++;
1043 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) {
1044 carp_suppress_preempt++;
1045 if (carp_suppress_preempt == 1) {
1051 sc->sc_sendad_success = 0;
1053 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) {
1054 if (++sc->sc_sendad_success >=
1055 CARP_SENDAD_MIN_SUCCESS) {
1056 carp_suppress_preempt--;
1057 sc->sc_sendad_errors = 0;
1060 sc->sc_sendad_errors = 0;
1065 if (advbase != 255 || advskew != 255)
1066 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv),
1072 * Broadcast a gratuitous ARP request containing
1073 * the virtual router MAC address for each IP address
1074 * associated with the virtual router.
1077 carp_send_arp(struct carp_softc *sc)
1081 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
1083 if (ifa->ifa_addr->sa_family != AF_INET)
1086 /* arprequest(sc->sc_carpdev, &in, &in, IF_LLADDR(sc->sc_ifp)); */
1087 arp_ifinit2(sc->sc_carpdev, ifa, IF_LLADDR(sc->sc_ifp));
1089 DELAY(1000); /* XXX */
1095 carp_send_na(struct carp_softc *sc)
1098 struct in6_addr *in6;
1099 static struct in6_addr mcast = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
1101 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
1103 if (ifa->ifa_addr->sa_family != AF_INET6)
1106 in6 = &ifatoia6(ifa)->ia_addr.sin6_addr;
1107 nd6_na_output(sc->sc_carpdev, &mcast, in6,
1108 ND_NA_FLAG_OVERRIDE, 1, NULL);
1109 DELAY(1000); /* XXX */
1115 carp_addrcount(struct carp_if *cif, struct in_ifaddr *ia, int type)
1117 struct carp_softc *vh;
1121 CARP_LOCK_ASSERT(cif);
1123 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1124 if ((type == CARP_COUNT_RUNNING &&
1125 (SC2IFP(vh)->if_flags & IFF_UP) &&
1126 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING)) ||
1127 (type == CARP_COUNT_MASTER && vh->sc_state == MASTER)) {
1128 IF_ADDR_LOCK(SC2IFP(vh));
1129 TAILQ_FOREACH(ifa, &SC2IFP(vh)->if_addrlist,
1131 if (ifa->ifa_addr->sa_family == AF_INET &&
1132 ia->ia_addr.sin_addr.s_addr ==
1133 ifatoia(ifa)->ia_addr.sin_addr.s_addr)
1136 IF_ADDR_UNLOCK(SC2IFP(vh));
1143 carp_iamatch(void *v, struct in_ifaddr *ia,
1144 struct in_addr *isaddr, u_int8_t **enaddr)
1146 struct carp_if *cif = v;
1147 struct carp_softc *vh;
1148 int index, count = 0;
1153 if (carp_opts[CARPCTL_ARPBALANCE]) {
1155 * XXX proof of concept implementation.
1156 * We use the source ip to decide which virtual host should
1157 * handle the request. If we're master of that virtual host,
1158 * then we respond, otherwise, just drop the arp packet on
1161 count = carp_addrcount(cif, ia, CARP_COUNT_RUNNING);
1163 /* should never reach this */
1168 /* this should be a hash, like pf_hash() */
1169 index = ntohl(isaddr->s_addr) % count;
1172 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1173 if ((SC2IFP(vh)->if_flags & IFF_UP) &&
1174 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING)) {
1175 IF_ADDR_LOCK(SC2IFP(vh));
1176 TAILQ_FOREACH(ifa, &SC2IFP(vh)->if_addrlist,
1178 if (ifa->ifa_addr->sa_family ==
1180 ia->ia_addr.sin_addr.s_addr ==
1181 ifatoia(ifa)->ia_addr.sin_addr.s_addr) {
1182 if (count == index) {
1185 *enaddr = IF_LLADDR(vh->sc_ifp);
1186 IF_ADDR_UNLOCK(SC2IFP(vh));
1190 IF_ADDR_UNLOCK(SC2IFP(vh));
1198 IF_ADDR_UNLOCK(SC2IFP(vh));
1202 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1203 if ((SC2IFP(vh)->if_flags & IFF_UP) &&
1204 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING) &&
1205 ia->ia_ifp == SC2IFP(vh) &&
1206 vh->sc_state == MASTER) {
1207 *enaddr = IF_LLADDR(vh->sc_ifp);
1219 carp_iamatch6(void *v, struct in6_addr *taddr)
1221 struct carp_if *cif = v;
1222 struct carp_softc *vh;
1226 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1227 IF_ADDR_LOCK(SC2IFP(vh));
1228 TAILQ_FOREACH(ifa, &SC2IFP(vh)->if_addrlist, ifa_list) {
1229 if (IN6_ARE_ADDR_EQUAL(taddr,
1230 &ifatoia6(ifa)->ia_addr.sin6_addr) &&
1231 (SC2IFP(vh)->if_flags & IFF_UP) &&
1232 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING) &&
1233 vh->sc_state == MASTER) {
1234 IF_ADDR_UNLOCK(SC2IFP(vh));
1239 IF_ADDR_UNLOCK(SC2IFP(vh));
1247 carp_macmatch6(void *v, struct mbuf *m, const struct in6_addr *taddr)
1250 struct carp_if *cif = v;
1251 struct carp_softc *sc;
1255 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list) {
1256 IF_ADDR_LOCK(SC2IFP(sc));
1257 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) {
1258 if (IN6_ARE_ADDR_EQUAL(taddr,
1259 &ifatoia6(ifa)->ia_addr.sin6_addr) &&
1260 (SC2IFP(sc)->if_flags & IFF_UP) &&
1261 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING)) {
1262 struct ifnet *ifp = SC2IFP(sc);
1263 mtag = m_tag_get(PACKET_TAG_CARP,
1264 sizeof(struct ifnet *), M_NOWAIT);
1266 /* better a bit than nothing */
1267 IF_ADDR_UNLOCK(SC2IFP(sc));
1269 return (IF_LLADDR(sc->sc_ifp));
1271 bcopy(&ifp, (caddr_t)(mtag + 1),
1272 sizeof(struct ifnet *));
1273 m_tag_prepend(m, mtag);
1275 IF_ADDR_UNLOCK(SC2IFP(sc));
1277 return (IF_LLADDR(sc->sc_ifp));
1280 IF_ADDR_UNLOCK(SC2IFP(sc));
1289 carp_forus(void *v, void *dhost)
1291 struct carp_if *cif = v;
1292 struct carp_softc *vh;
1293 u_int8_t *ena = dhost;
1295 if (ena[0] || ena[1] || ena[2] != 0x5e || ena[3] || ena[4] != 1)
1299 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list)
1300 if ((SC2IFP(vh)->if_flags & IFF_UP) &&
1301 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING) &&
1302 vh->sc_state == MASTER &&
1303 !bcmp(dhost, IF_LLADDR(vh->sc_ifp), ETHER_ADDR_LEN)) {
1305 return (SC2IFP(vh));
1313 carp_master_down(void *v)
1315 struct carp_softc *sc = v;
1318 carp_master_down_locked(sc);
1323 carp_master_down_locked(struct carp_softc *sc)
1326 CARP_SCLOCK_ASSERT(sc);
1328 switch (sc->sc_state) {
1330 printf("%s: master_down event in INIT state\n",
1331 SC2IFP(sc)->if_xname);
1336 carp_set_state(sc, MASTER);
1337 carp_send_ad_locked(sc);
1343 carp_setroute(sc, RTM_ADD);
1349 * When in backup state, af indicates whether to reset the master down timer
1350 * for v4 or v6. If it's set to zero, reset the ones which are already pending.
1353 carp_setrun(struct carp_softc *sc, sa_family_t af)
1357 if (sc->sc_carpdev == NULL) {
1358 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING;
1359 carp_set_state(sc, INIT);
1362 CARP_SCLOCK_ASSERT(sc);
1364 if (SC2IFP(sc)->if_flags & IFF_UP &&
1365 sc->sc_vhid > 0 && (sc->sc_naddrs || sc->sc_naddrs6))
1366 SC2IFP(sc)->if_drv_flags |= IFF_DRV_RUNNING;
1368 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING;
1369 carp_setroute(sc, RTM_DELETE);
1373 switch (sc->sc_state) {
1375 if (carp_opts[CARPCTL_PREEMPT] && !carp_suppress_preempt) {
1376 carp_send_ad_locked(sc);
1381 CARP_DEBUG("%s: INIT -> MASTER (preempting)\n",
1382 SC2IFP(sc)->if_xname);
1383 carp_set_state(sc, MASTER);
1384 carp_setroute(sc, RTM_ADD);
1386 CARP_DEBUG("%s: INIT -> BACKUP\n", SC2IFP(sc)->if_xname);
1387 carp_set_state(sc, BACKUP);
1388 carp_setroute(sc, RTM_DELETE);
1393 callout_stop(&sc->sc_ad_tmo);
1394 tv.tv_sec = 3 * sc->sc_advbase;
1395 tv.tv_usec = sc->sc_advskew * 1000000 / 256;
1399 callout_reset(&sc->sc_md_tmo, tvtohz(&tv),
1400 carp_master_down, sc);
1405 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv),
1406 carp_master_down, sc);
1411 callout_reset(&sc->sc_md_tmo, tvtohz(&tv),
1412 carp_master_down, sc);
1414 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv),
1415 carp_master_down, sc);
1420 tv.tv_sec = sc->sc_advbase;
1421 tv.tv_usec = sc->sc_advskew * 1000000 / 256;
1422 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv),
1429 carp_multicast_cleanup(struct carp_softc *sc)
1431 struct ip_moptions *imo = &sc->sc_imo;
1432 u_int16_t n = imo->imo_num_memberships;
1434 /* Clean up our own multicast memberships */
1436 if (imo->imo_membership[n] != NULL) {
1437 in_delmulti(imo->imo_membership[n]);
1438 imo->imo_membership[n] = NULL;
1441 KASSERT(imo->imo_mfilters == NULL,
1442 ("%s: imo_mfilters != NULL", __func__));
1443 imo->imo_num_memberships = 0;
1444 imo->imo_multicast_ifp = NULL;
1449 carp_multicast6_cleanup(struct carp_softc *sc)
1451 struct ip6_moptions *im6o = &sc->sc_im6o;
1453 while (!LIST_EMPTY(&im6o->im6o_memberships)) {
1454 struct in6_multi_mship *imm =
1455 LIST_FIRST(&im6o->im6o_memberships);
1457 LIST_REMOVE(imm, i6mm_chain);
1458 in6_leavegroup(imm);
1460 im6o->im6o_multicast_ifp = NULL;
1465 carp_set_addr(struct carp_softc *sc, struct sockaddr_in *sin)
1467 INIT_VNET_INET(curvnet);
1469 struct carp_if *cif;
1470 struct in_ifaddr *ia, *ia_if;
1471 struct ip_moptions *imo = &sc->sc_imo;
1472 struct in_addr addr;
1473 u_long iaddr = htonl(sin->sin_addr.s_addr);
1476 if (sin->sin_addr.s_addr == 0) {
1477 if (!(SC2IFP(sc)->if_flags & IFF_UP))
1478 carp_set_state(sc, INIT);
1480 SC2IFP(sc)->if_flags |= IFF_UP;
1489 /* we have to do it by hands to check we won't match on us */
1490 ia_if = NULL; own = 0;
1491 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
1492 /* and, yeah, we need a multicast-capable iface too */
1493 if (ia->ia_ifp != SC2IFP(sc) &&
1494 (ia->ia_ifp->if_flags & IFF_MULTICAST) &&
1495 (iaddr & ia->ia_subnetmask) == ia->ia_subnet) {
1498 if (sin->sin_addr.s_addr ==
1499 ia->ia_addr.sin_addr.s_addr)
1505 return (EADDRNOTAVAIL);
1510 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0 ||
1511 (imo->imo_multicast_ifp && imo->imo_multicast_ifp != ifp))
1512 return (EADDRNOTAVAIL);
1514 if (imo->imo_num_memberships == 0) {
1515 addr.s_addr = htonl(INADDR_CARP_GROUP);
1516 if ((imo->imo_membership[0] = in_addmulti(&addr, ifp)) == NULL)
1518 imo->imo_num_memberships++;
1519 imo->imo_multicast_ifp = ifp;
1520 imo->imo_multicast_ttl = CARP_DFLTTL;
1521 imo->imo_multicast_loop = 0;
1524 if (!ifp->if_carp) {
1526 cif = malloc(sizeof(*cif), M_CARP,
1532 if ((error = ifpromisc(ifp, 1))) {
1537 CARP_LOCK_INIT(cif);
1539 cif->vhif_ifp = ifp;
1540 TAILQ_INIT(&cif->vhif_vrs);
1544 struct carp_softc *vr;
1546 cif = (struct carp_if *)ifp->if_carp;
1548 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list)
1549 if (vr != sc && vr->sc_vhid == sc->sc_vhid) {
1556 sc->sc_carpdev = ifp;
1558 { /* XXX prevent endless loop if already in queue */
1559 struct carp_softc *vr, *after = NULL;
1561 cif = (struct carp_if *)ifp->if_carp;
1563 /* XXX: cif should not change, right? So we still hold the lock */
1564 CARP_LOCK_ASSERT(cif);
1566 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) {
1569 if (vr->sc_vhid < sc->sc_vhid)
1574 /* We're trying to keep things in order */
1575 if (after == NULL) {
1576 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list);
1578 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, sc, sc_list);
1585 SC2IFP(sc)->if_flags |= IFF_UP;
1588 carp_sc_state_locked(sc);
1596 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]);
1601 carp_del_addr(struct carp_softc *sc, struct sockaddr_in *sin)
1605 if (!--sc->sc_naddrs) {
1606 struct carp_if *cif = (struct carp_if *)sc->sc_carpdev->if_carp;
1607 struct ip_moptions *imo = &sc->sc_imo;
1610 callout_stop(&sc->sc_ad_tmo);
1611 SC2IFP(sc)->if_flags &= ~IFF_UP;
1612 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING;
1614 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]);
1615 imo->imo_multicast_ifp = NULL;
1616 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list);
1617 if (!--cif->vhif_nvrs) {
1618 sc->sc_carpdev->if_carp = NULL;
1619 CARP_LOCK_DESTROY(cif);
1620 free(cif, M_IFADDR);
1631 carp_set_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6)
1633 INIT_VNET_INET6(curvnet);
1635 struct carp_if *cif;
1636 struct in6_ifaddr *ia, *ia_if;
1637 struct ip6_moptions *im6o = &sc->sc_im6o;
1638 struct in6_multi_mship *imm;
1639 struct in6_addr in6;
1642 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1643 if (!(SC2IFP(sc)->if_flags & IFF_UP))
1644 carp_set_state(sc, INIT);
1646 SC2IFP(sc)->if_flags |= IFF_UP;
1655 /* we have to do it by hands to check we won't match on us */
1656 ia_if = NULL; own = 0;
1657 for (ia = V_in6_ifaddr; ia; ia = ia->ia_next) {
1660 for (i = 0; i < 4; i++) {
1661 if ((sin6->sin6_addr.s6_addr32[i] &
1662 ia->ia_prefixmask.sin6_addr.s6_addr32[i]) !=
1663 (ia->ia_addr.sin6_addr.s6_addr32[i] &
1664 ia->ia_prefixmask.sin6_addr.s6_addr32[i]))
1667 /* and, yeah, we need a multicast-capable iface too */
1668 if (ia->ia_ifp != SC2IFP(sc) &&
1669 (ia->ia_ifp->if_flags & IFF_MULTICAST) &&
1673 if (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
1674 &ia->ia_addr.sin6_addr))
1680 return (EADDRNOTAVAIL);
1684 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0 ||
1685 (im6o->im6o_multicast_ifp && im6o->im6o_multicast_ifp != ifp))
1686 return (EADDRNOTAVAIL);
1688 if (!sc->sc_naddrs6) {
1689 im6o->im6o_multicast_ifp = ifp;
1691 /* join CARP multicast address */
1692 bzero(&in6, sizeof(in6));
1693 in6.s6_addr16[0] = htons(0xff02);
1694 in6.s6_addr8[15] = 0x12;
1695 if (in6_setscope(&in6, ifp, NULL) != 0)
1697 if ((imm = in6_joingroup(ifp, &in6, &error, 0)) == NULL)
1699 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1701 /* join solicited multicast address */
1702 bzero(&in6, sizeof(in6));
1703 in6.s6_addr16[0] = htons(0xff02);
1704 in6.s6_addr32[1] = 0;
1705 in6.s6_addr32[2] = htonl(1);
1706 in6.s6_addr32[3] = sin6->sin6_addr.s6_addr32[3];
1707 in6.s6_addr8[12] = 0xff;
1708 if (in6_setscope(&in6, ifp, NULL) != 0)
1710 if ((imm = in6_joingroup(ifp, &in6, &error, 0)) == NULL)
1712 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1715 if (!ifp->if_carp) {
1716 cif = malloc(sizeof(*cif), M_CARP,
1722 if ((error = ifpromisc(ifp, 1))) {
1727 CARP_LOCK_INIT(cif);
1729 cif->vhif_ifp = ifp;
1730 TAILQ_INIT(&cif->vhif_vrs);
1734 struct carp_softc *vr;
1736 cif = (struct carp_if *)ifp->if_carp;
1738 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list)
1739 if (vr != sc && vr->sc_vhid == sc->sc_vhid) {
1746 sc->sc_carpdev = ifp;
1748 { /* XXX prevent endless loop if already in queue */
1749 struct carp_softc *vr, *after = NULL;
1751 cif = (struct carp_if *)ifp->if_carp;
1752 CARP_LOCK_ASSERT(cif);
1754 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) {
1757 if (vr->sc_vhid < sc->sc_vhid)
1762 /* We're trying to keep things in order */
1763 if (after == NULL) {
1764 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list);
1766 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, sc, sc_list);
1773 SC2IFP(sc)->if_flags |= IFF_UP;
1776 carp_sc_state_locked(sc);
1784 /* clean up multicast memberships */
1785 if (!sc->sc_naddrs6) {
1786 while (!LIST_EMPTY(&im6o->im6o_memberships)) {
1787 imm = LIST_FIRST(&im6o->im6o_memberships);
1788 LIST_REMOVE(imm, i6mm_chain);
1789 in6_leavegroup(imm);
1796 carp_del_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6)
1800 if (!--sc->sc_naddrs6) {
1801 struct carp_if *cif = (struct carp_if *)sc->sc_carpdev->if_carp;
1802 struct ip6_moptions *im6o = &sc->sc_im6o;
1805 callout_stop(&sc->sc_ad_tmo);
1806 SC2IFP(sc)->if_flags &= ~IFF_UP;
1807 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING;
1809 while (!LIST_EMPTY(&im6o->im6o_memberships)) {
1810 struct in6_multi_mship *imm =
1811 LIST_FIRST(&im6o->im6o_memberships);
1813 LIST_REMOVE(imm, i6mm_chain);
1814 in6_leavegroup(imm);
1816 im6o->im6o_multicast_ifp = NULL;
1817 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list);
1818 if (!--cif->vhif_nvrs) {
1819 CARP_LOCK_DESTROY(cif);
1820 sc->sc_carpdev->if_carp = NULL;
1821 free(cif, M_IFADDR);
1831 carp_ioctl(struct ifnet *ifp, u_long cmd, caddr_t addr)
1833 struct carp_softc *sc = ifp->if_softc, *vr;
1834 struct carpreq carpr;
1837 struct ifaliasreq *ifra;
1838 int locked = 0, error = 0;
1840 ifa = (struct ifaddr *)addr;
1841 ifra = (struct ifaliasreq *)addr;
1842 ifr = (struct ifreq *)addr;
1846 switch (ifa->ifa_addr->sa_family) {
1849 SC2IFP(sc)->if_flags |= IFF_UP;
1850 bcopy(ifa->ifa_addr, ifa->ifa_dstaddr,
1851 sizeof(struct sockaddr));
1852 error = carp_set_addr(sc, satosin(ifa->ifa_addr));
1857 SC2IFP(sc)->if_flags |= IFF_UP;
1858 error = carp_set_addr6(sc, satosin6(ifa->ifa_addr));
1862 error = EAFNOSUPPORT;
1868 switch (ifa->ifa_addr->sa_family) {
1871 SC2IFP(sc)->if_flags |= IFF_UP;
1872 bcopy(ifa->ifa_addr, ifa->ifa_dstaddr,
1873 sizeof(struct sockaddr));
1874 error = carp_set_addr(sc, satosin(&ifra->ifra_addr));
1879 SC2IFP(sc)->if_flags |= IFF_UP;
1880 error = carp_set_addr6(sc, satosin6(&ifra->ifra_addr));
1884 error = EAFNOSUPPORT;
1890 switch (ifa->ifa_addr->sa_family) {
1893 error = carp_del_addr(sc, satosin(&ifra->ifra_addr));
1898 error = carp_del_addr6(sc, satosin6(&ifra->ifra_addr));
1902 error = EAFNOSUPPORT;
1908 if (sc->sc_carpdev) {
1912 if (sc->sc_state != INIT && !(ifr->ifr_flags & IFF_UP)) {
1913 callout_stop(&sc->sc_ad_tmo);
1914 callout_stop(&sc->sc_md_tmo);
1915 callout_stop(&sc->sc_md6_tmo);
1916 if (sc->sc_state == MASTER)
1917 carp_send_ad_locked(sc);
1918 carp_set_state(sc, INIT);
1920 } else if (sc->sc_state == INIT && (ifr->ifr_flags & IFF_UP)) {
1921 SC2IFP(sc)->if_flags |= IFF_UP;
1927 error = priv_check(curthread, PRIV_NETINET_CARP);
1930 if ((error = copyin(ifr->ifr_data, &carpr, sizeof carpr)))
1933 if (sc->sc_carpdev) {
1937 if (sc->sc_state != INIT && carpr.carpr_state != sc->sc_state) {
1938 switch (carpr.carpr_state) {
1940 callout_stop(&sc->sc_ad_tmo);
1941 carp_set_state(sc, BACKUP);
1943 carp_setroute(sc, RTM_DELETE);
1946 carp_master_down_locked(sc);
1952 if (carpr.carpr_vhid > 0) {
1953 if (carpr.carpr_vhid > 255) {
1957 if (sc->sc_carpdev) {
1958 struct carp_if *cif;
1959 cif = (struct carp_if *)sc->sc_carpdev->if_carp;
1960 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list)
1962 vr->sc_vhid == carpr.carpr_vhid) {
1966 if (error == EEXIST)
1969 sc->sc_vhid = carpr.carpr_vhid;
1970 IF_LLADDR(sc->sc_ifp)[0] = 0;
1971 IF_LLADDR(sc->sc_ifp)[1] = 0;
1972 IF_LLADDR(sc->sc_ifp)[2] = 0x5e;
1973 IF_LLADDR(sc->sc_ifp)[3] = 0;
1974 IF_LLADDR(sc->sc_ifp)[4] = 1;
1975 IF_LLADDR(sc->sc_ifp)[5] = sc->sc_vhid;
1978 if (carpr.carpr_advbase > 0 || carpr.carpr_advskew > 0) {
1979 if (carpr.carpr_advskew >= 255) {
1983 if (carpr.carpr_advbase > 255) {
1987 sc->sc_advbase = carpr.carpr_advbase;
1988 sc->sc_advskew = carpr.carpr_advskew;
1991 bcopy(carpr.carpr_key, sc->sc_key, sizeof(sc->sc_key));
2001 /* XXX: lockless read */
2002 bzero(&carpr, sizeof(carpr));
2003 carpr.carpr_state = sc->sc_state;
2004 carpr.carpr_vhid = sc->sc_vhid;
2005 carpr.carpr_advbase = sc->sc_advbase;
2006 carpr.carpr_advskew = sc->sc_advskew;
2007 error = priv_check(curthread, PRIV_NETINET_CARP);
2009 bcopy(sc->sc_key, carpr.carpr_key,
2010 sizeof(carpr.carpr_key));
2011 error = copyout(&carpr, ifr->ifr_data, sizeof(carpr));
2021 carp_hmac_prepare(sc);
2027 * XXX: this is looutput. We should eventually use it from there.
2030 carp_looutput(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
2034 struct rtentry *rt = NULL;
2036 M_ASSERTPKTHDR(m); /* check if we have the packet header */
2040 if (rt && rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
2042 return (rt->rt_flags & RTF_BLACKHOLE ? 0 :
2043 rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
2047 ifp->if_obytes += m->m_pkthdr.len;
2049 /* BPF writes need to be handled specially. */
2050 if (dst->sa_family == AF_UNSPEC) {
2051 bcopy(dst->sa_data, &af, sizeof(af));
2052 dst->sa_family = af;
2056 switch (dst->sa_family) {
2063 printf("carp_looutput: af=%d unexpected\n", dst->sa_family);
2065 return (EAFNOSUPPORT);
2068 return(if_simloop(ifp, m, dst->sa_family, 0));
2072 * Start output on carp interface. This function should never be called.
2075 carp_start(struct ifnet *ifp)
2078 printf("%s: start called\n", ifp->if_xname);
2083 carp_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2087 struct carp_softc *sc;
2088 struct ifnet *carp_ifp;
2093 switch (sa->sa_family) {
2106 mtag = m_tag_find(m, PACKET_TAG_CARP, NULL);
2110 bcopy(mtag + 1, &carp_ifp, sizeof(struct ifnet *));
2111 sc = carp_ifp->if_softc;
2113 /* Set the source MAC address to Virtual Router MAC Address */
2114 switch (ifp->if_type) {
2117 struct ether_header *eh;
2119 eh = mtod(m, struct ether_header *);
2120 eh->ether_shost[0] = 0;
2121 eh->ether_shost[1] = 0;
2122 eh->ether_shost[2] = 0x5e;
2123 eh->ether_shost[3] = 0;
2124 eh->ether_shost[4] = 1;
2125 eh->ether_shost[5] = sc->sc_vhid;
2129 struct fddi_header *fh;
2131 fh = mtod(m, struct fddi_header *);
2132 fh->fddi_shost[0] = 0;
2133 fh->fddi_shost[1] = 0;
2134 fh->fddi_shost[2] = 0x5e;
2135 fh->fddi_shost[3] = 0;
2136 fh->fddi_shost[4] = 1;
2137 fh->fddi_shost[5] = sc->sc_vhid;
2140 case IFT_ISO88025: {
2141 struct iso88025_header *th;
2142 th = mtod(m, struct iso88025_header *);
2143 th->iso88025_shost[0] = 3;
2144 th->iso88025_shost[1] = 0;
2145 th->iso88025_shost[2] = 0x40 >> (sc->sc_vhid - 1);
2146 th->iso88025_shost[3] = 0x40000 >> (sc->sc_vhid - 1);
2147 th->iso88025_shost[4] = 0;
2148 th->iso88025_shost[5] = 0;
2152 printf("%s: carp is not supported for this interface type\n",
2154 return (EOPNOTSUPP);
2161 carp_set_state(struct carp_softc *sc, int state)
2166 CARP_SCLOCK_ASSERT(sc);
2168 if (sc->sc_state == state)
2171 sc->sc_state = state;
2174 link_state = LINK_STATE_DOWN;
2177 link_state = LINK_STATE_UP;
2180 link_state = LINK_STATE_UNKNOWN;
2183 if_link_state_change(SC2IFP(sc), link_state);
2187 carp_carpdev_state(void *v)
2189 struct carp_if *cif = v;
2192 carp_carpdev_state_locked(cif);
2197 carp_carpdev_state_locked(struct carp_if *cif)
2199 struct carp_softc *sc;
2201 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list)
2202 carp_sc_state_locked(sc);
2206 carp_sc_state_locked(struct carp_softc *sc)
2208 CARP_SCLOCK_ASSERT(sc);
2210 if (sc->sc_carpdev->if_link_state != LINK_STATE_UP ||
2211 !(sc->sc_carpdev->if_flags & IFF_UP)) {
2212 sc->sc_flags_backup = SC2IFP(sc)->if_flags;
2213 SC2IFP(sc)->if_flags &= ~IFF_UP;
2214 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING;
2215 callout_stop(&sc->sc_ad_tmo);
2216 callout_stop(&sc->sc_md_tmo);
2217 callout_stop(&sc->sc_md6_tmo);
2218 carp_set_state(sc, INIT);
2220 if (!sc->sc_suppress) {
2221 carp_suppress_preempt++;
2222 if (carp_suppress_preempt == 1) {
2228 sc->sc_suppress = 1;
2230 SC2IFP(sc)->if_flags |= sc->sc_flags_backup;
2231 carp_set_state(sc, INIT);
2233 if (sc->sc_suppress)
2234 carp_suppress_preempt--;
2235 sc->sc_suppress = 0;
2242 carp_modevent(module_t mod, int type, void *data)
2246 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2247 carp_ifdetach, NULL, EVENTHANDLER_PRI_ANY);
2248 if (if_detach_event_tag == NULL)
2250 mtx_init(&carp_mtx, "carp_mtx", NULL, MTX_DEF);
2251 LIST_INIT(&carpif_list);
2252 if_clone_attach(&carp_cloner);
2256 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2257 if_clone_detach(&carp_cloner);
2258 mtx_destroy(&carp_mtx);
2268 static moduledata_t carp_mod = {
2274 DECLARE_MODULE(carp, carp_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);