2 * Copyright (c) 2002 Michael Shalayeff. All rights reserved.
3 * Copyright (c) 2003 Ryan McBride. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
15 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
16 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
17 * IN NO EVENT SHALL THE AUTHOR OR HIS RELATIVES BE LIABLE FOR ANY DIRECT,
18 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
19 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
20 * SERVICES; LOSS OF MIND, USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
22 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
23 * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
24 * THE POSSIBILITY OF SUCH DAMAGE.
27 * $FreeBSD: src/sys/netinet/ip_carp.c,v 1.48 2007/02/02 09:39:09 glebius Exp $
28 * $DragonFly: src/sys/netinet/ip_carp.c,v 1.10 2008/07/27 10:06:57 sephe Exp $
32 /*#include "opt_bpf.h"*/
34 #include "opt_inet6.h"
36 #include <sys/types.h>
37 #include <sys/param.h>
38 #include <sys/systm.h>
40 #include <sys/kernel.h>
41 #include <machine/limits.h>
42 #include <sys/malloc.h>
44 #include <sys/module.h>
47 #include <sys/sysctl.h>
48 #include <sys/syslog.h>
49 #include <sys/signalvar.h>
50 #include <sys/filio.h>
51 #include <sys/sockio.h>
52 #include <sys/in_cksum.h>
53 #include <sys/socket.h>
54 #include <sys/vnode.h>
56 #include <machine/stdarg.h>
59 #include <net/ethernet.h>
61 #include <net/if_dl.h>
62 #include <net/if_types.h>
63 #include <net/route.h>
64 #include <net/if_clone.h>
67 #include <netinet/in.h>
68 #include <netinet/in_var.h>
69 #include <netinet/in_systm.h>
70 #include <netinet/ip.h>
71 #include <netinet/ip_var.h>
72 #include <netinet/if_ether.h>
76 #include <netinet/icmp6.h>
77 #include <netinet/ip6.h>
78 #include <netinet6/ip6_var.h>
79 #include <netinet6/scope6_var.h>
80 #include <netinet6/nd6.h>
83 #include <crypto/sha1.h>
84 #include <netinet/ip_carp.h>
87 #define CARP_IFNAME "carp"
88 static MALLOC_DEFINE(M_CARP, "CARP", "CARP interfaces");
89 static MALLOC_DEFINE(M_IFNET, "IFNET", "IFNET CARP?");
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 */
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;
158 struct lock vhif_lock;
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 #define CARP_LOCK_INIT(cif) lockinit(&(cif)->vhif_lock, "carp_if", 0, 0);
165 #define CARP_LOCK_DESTROY(cif) ;
166 #define CARP_LOCK_ASSERT(cif) ;
167 #define CARP_LOCK(cif) lockmgr(&(cif)->vhif_lock, LK_EXCLUSIVE);
168 #define CARP_UNLOCK(cif) lockmgr(&(cif)->vhif_lock, LK_RELEASE);
170 #define CARP_SCLOCK(sc) lockmgr(&SC2CIF(sc)->vhif_lock, LK_EXCLUSIVE);
171 #define CARP_SCUNLOCK(sc) lockmgr(&SC2CIF(sc)->vhif_lock, LK_RELEASE);
172 #define CARP_SCLOCK_ASSERT(sc) ;
174 #define CARP_LOG(...) do { \
175 if (carp_opts[CARPCTL_LOG] > 0) \
176 log(LOG_INFO, __VA_ARGS__); \
179 #define CARP_DEBUG(...) do { \
180 if (carp_opts[CARPCTL_LOG] > 1) \
181 log(LOG_DEBUG, __VA_ARGS__); \
184 static void carp_hmac_prepare(struct carp_softc *);
185 static void carp_hmac_generate(struct carp_softc *, uint32_t *,
187 static int carp_hmac_verify(struct carp_softc *, uint32_t *,
189 static void carp_setroute(struct carp_softc *, int);
190 static void carp_input_c(struct mbuf *, struct carp_header *, sa_family_t);
191 static int carp_clone_create(struct if_clone *, int);
192 static void carp_clone_destroy(struct ifnet *);
193 static void carpdetach(struct carp_softc *, int);
194 static int carp_prepare_ad(struct mbuf *, struct carp_softc *,
195 struct carp_header *);
196 static void carp_send_ad_all(void);
197 static void carp_send_ad(void *);
198 static void carp_send_ad_locked(struct carp_softc *);
199 static void carp_send_arp(struct carp_softc *);
200 static void carp_master_down(void *);
201 static void carp_master_down_locked(struct carp_softc *);
202 static int carp_ioctl(struct ifnet *, u_long, caddr_t, struct ucred *);
203 static int carp_looutput(struct ifnet *, struct mbuf *, struct sockaddr *,
205 static void carp_start(struct ifnet *);
206 static void carp_setrun(struct carp_softc *, sa_family_t);
207 static void carp_set_state(struct carp_softc *, int);
208 static int carp_addrcount(struct carp_if *, struct in_ifaddr *, int);
209 enum { CARP_COUNT_MASTER, CARP_COUNT_RUNNING };
211 static void carp_multicast_cleanup(struct carp_softc *);
212 static int carp_set_addr(struct carp_softc *, struct sockaddr_in *);
213 static int carp_del_addr(struct carp_softc *, struct sockaddr_in *);
214 static void carp_carpdev_state_locked(struct carp_if *);
215 static void carp_sc_state_locked(struct carp_softc *);
217 static void carp_send_na(struct carp_softc *);
218 static int carp_set_addr6(struct carp_softc *, struct sockaddr_in6 *);
219 static int carp_del_addr6(struct carp_softc *, struct sockaddr_in6 *);
220 static void carp_multicast6_cleanup(struct carp_softc *);
223 static LIST_HEAD(, carp_softc) carpif_list;
225 struct if_clone carp_cloner = IF_CLONE_INITIALIZER(CARP_IFNAME, carp_clone_create, carp_clone_destroy, 0, IF_MAXUNIT);
227 static eventhandler_tag if_detach_event_tag;
229 static __inline uint16_t
230 carp_cksum(struct mbuf *m, int len)
232 return (in_cksum(m, len));
236 carp_hmac_prepare(struct carp_softc *sc)
238 uint8_t version = CARP_VERSION, type = CARP_ADVERTISEMENT;
239 uint8_t vhid = sc->sc_vhid & 0xff;
240 struct ifaddr_container *ifac;
249 /* XXX: possible race here */
251 /* compute ipad from key */
252 bzero(sc->sc_pad, sizeof(sc->sc_pad));
253 bcopy(sc->sc_key, sc->sc_pad, sizeof(sc->sc_key));
254 for (i = 0; i < sizeof(sc->sc_pad); i++)
255 sc->sc_pad[i] ^= 0x36;
257 /* precompute first part of inner hash */
258 SHA1Init(&sc->sc_sha1);
259 SHA1Update(&sc->sc_sha1, sc->sc_pad, sizeof(sc->sc_pad));
260 SHA1Update(&sc->sc_sha1, (void *)&version, sizeof(version));
261 SHA1Update(&sc->sc_sha1, (void *)&type, sizeof(type));
262 SHA1Update(&sc->sc_sha1, (void *)&vhid, sizeof(vhid));
264 TAILQ_FOREACH(ifac, &SC2IFP(sc)->if_addrheads[mycpuid], ifa_link) {
265 struct ifaddr *ifa = ifac->ifa;
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(ifac, &SC2IFP(sc)->if_addrheads[mycpuid], ifa_link) {
275 struct ifaddr *ifa = ifac->ifa;
277 if (ifa->ifa_addr->sa_family == AF_INET6) {
278 in6 = ifatoia6(ifa)->ia_addr.sin6_addr;
279 in6_clearscope(&in6);
280 SHA1Update(&sc->sc_sha1, (void *)&in6, sizeof(in6));
285 /* convert ipad to opad */
286 for (i = 0; i < sizeof(sc->sc_pad); i++)
287 sc->sc_pad[i] ^= 0x36 ^ 0x5c;
294 carp_hmac_generate(struct carp_softc *sc, uint32_t counter[2],
295 unsigned char md[20])
299 /* fetch first half of inner hash */
300 bcopy(&sc->sc_sha1, &sha1ctx, sizeof(sha1ctx));
302 SHA1Update(&sha1ctx, (void *)counter, sizeof(sc->sc_counter));
303 SHA1Final(md, &sha1ctx);
307 SHA1Update(&sha1ctx, sc->sc_pad, sizeof(sc->sc_pad));
308 SHA1Update(&sha1ctx, md, 20);
309 SHA1Final(md, &sha1ctx);
313 carp_hmac_verify(struct carp_softc *sc, uint32_t counter[2],
314 unsigned char md[20])
316 unsigned char md2[20];
318 CARP_SCLOCK_ASSERT(sc);
320 carp_hmac_generate(sc, counter, md2);
322 return (bcmp(md, md2, sizeof(md2)));
326 carp_setroute(struct carp_softc *sc, int cmd)
328 struct ifaddr_container *ifac;
331 CARP_SCLOCK_ASSERT(sc);
334 TAILQ_FOREACH(ifac, &SC2IFP(sc)->if_addrheads[mycpuid], ifa_link) {
335 struct ifaddr *ifa = ifac->ifa;
337 if (ifa->ifa_addr->sa_family == AF_INET &&
338 sc->sc_carpdev != NULL) {
339 int count = carp_addrcount(
340 (struct carp_if *)sc->sc_carpdev->if_carp,
341 ifatoia(ifa), CARP_COUNT_MASTER);
343 if ((cmd == RTM_ADD && count == 1) ||
344 (cmd == RTM_DELETE && count == 0))
345 rtinit(ifa, cmd, RTF_UP | RTF_HOST);
348 if (ifa->ifa_addr->sa_family == AF_INET6) {
361 carp_clone_create(struct if_clone *ifc, int unit)
364 struct carp_softc *sc;
367 MALLOC(sc, struct carp_softc *, sizeof(*sc), M_CARP, M_WAITOK|M_ZERO);
368 ifp = SC2IFP(sc) = kmalloc(sizeof(struct ifnet), M_IFNET, M_WAITOK|M_ZERO);
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? */
379 sc->sc_im6o.im6o_multicast_hlim = CARP_DFLTTL;
382 /* sc->sc_imo.imo_membership = kmalloc((sizeof(struct in_multi) * IP_MAX_MEMBERSHIPS), M_CARP,M_WAITOK);*/
384 sc->sc_imo.imo_max_memberships = IP_MAX_MEMBERSHIPS;
385 sc->sc_imo.imo_multicast_vif = -1;
387 callout_init(&sc->sc_ad_tmo);
388 callout_init(&sc->sc_md_tmo);
389 callout_init(&sc->sc_md6_tmo);
392 if_initname(ifp, CARP_IFNAME, unit);
393 ifp->if_mtu = ETHERMTU;
394 ifp->if_flags = IFF_LOOPBACK;
395 ifp->if_ioctl = carp_ioctl;
396 ifp->if_output = carp_looutput;
397 ifp->if_start = carp_start;
398 ifp->if_type = IFT_CARP;
399 ifp->if_snd.ifq_maxlen = ifqmaxlen;
401 if_attach(ifp, NULL);
402 bpfattach(ifp, DLT_NULL, sizeof(u_int));
405 LIST_INSERT_HEAD(&carpif_list, sc, sc_next);
412 carp_clone_destroy(struct ifnet *ifp)
414 struct carp_softc *sc = ifp->if_softc;
418 carpdetach(sc, 1); /* Returns unlocked. */
421 LIST_REMOVE(sc, sc_next);
425 /* if_free_type(ifp, IFT_ETHER);*/
426 /* kfree(sc->sc_imo.imo_membership, M_CARP); */
431 * This function can be called on CARP interface destroy path,
432 * and in case of the removal of the underlying interface as
433 * well. We differentiate these two cases. In the latter case
434 * we do not cleanup our multicast memberships, since they
435 * are already freed. Also, in the latter case we do not
436 * release the lock on return, because the function will be
437 * called once more, for another CARP instance on the same
441 carpdetach(struct carp_softc *sc, int unlock)
445 callout_stop(&sc->sc_ad_tmo);
446 callout_stop(&sc->sc_md_tmo);
447 callout_stop(&sc->sc_md6_tmo);
450 carp_suppress_preempt--;
453 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS)
454 carp_suppress_preempt--;
455 sc->sc_sendad_errors = 0;
457 carp_set_state(sc, INIT);
458 SC2IFP(sc)->if_flags &= ~IFF_UP;
461 carp_multicast_cleanup(sc);
463 carp_multicast6_cleanup(sc);
466 if (sc->sc_carpdev != NULL) {
467 cif = (struct carp_if *)sc->sc_carpdev->if_carp;
468 CARP_LOCK_ASSERT(cif);
469 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list);
470 if (!--cif->vhif_nvrs) {
471 ifpromisc(sc->sc_carpdev, 0);
472 sc->sc_carpdev->if_carp = NULL;
473 CARP_LOCK_DESTROY(cif);
477 sc->sc_carpdev = NULL;
481 /* Detach an interface from the carp. */
483 carp_ifdetach(void *arg __unused, struct ifnet *ifp)
485 struct carp_if *cif = (struct carp_if *)ifp->if_carp;
486 struct carp_softc *sc, *nextsc;
492 * XXX: At the end of for() cycle the lock will be destroyed.
495 for (sc = TAILQ_FIRST(&cif->vhif_vrs); sc; sc = nextsc) {
496 nextsc = TAILQ_NEXT(sc, sc_list);
503 * process input packet.
504 * we have rearranged checks order compared to the rfc,
505 * but it seems more efficient this way or not possible otherwise.
508 carp_input(struct mbuf *m, ...)
510 struct ip *ip = mtod(m, struct ip *);
511 struct carp_header *ch;
512 int iplen, len, hlen;
516 hlen = __va_arg(ap, int);
519 carpstats.carps_ipackets++;
521 if (!carp_opts[CARPCTL_ALLOW]) {
526 /* check if received on a valid carp interface */
527 if (m->m_pkthdr.rcvif->if_carp == NULL) {
528 carpstats.carps_badif++;
529 CARP_LOG("carp_input: packet received on non-carp "
531 m->m_pkthdr.rcvif->if_xname);
536 /* verify that the IP TTL is 255. */
537 if (ip->ip_ttl != CARP_DFLTTL) {
538 carpstats.carps_badttl++;
539 CARP_LOG("carp_input: received ttl %d != 255i on %s\n",
541 m->m_pkthdr.rcvif->if_xname);
546 iplen = ip->ip_hl << 2;
548 if (m->m_pkthdr.len < iplen + sizeof(*ch)) {
549 carpstats.carps_badlen++;
550 CARP_LOG("carp_input: received len %zd < "
551 "sizeof(struct carp_header)\n",
552 m->m_len - sizeof(struct ip));
557 if (iplen + sizeof(*ch) < m->m_len) {
558 if ((m = m_pullup(m, iplen + sizeof(*ch))) == NULL) {
559 carpstats.carps_hdrops++;
560 CARP_LOG("carp_input: pullup failed\n");
563 ip = mtod(m, struct ip *);
565 ch = (struct carp_header *)((char *)ip + iplen);
568 * verify that the received packet length is
569 * equal to the CARP header
571 len = iplen + sizeof(*ch);
572 if (len > m->m_pkthdr.len) {
573 carpstats.carps_badlen++;
574 CARP_LOG("carp_input: packet too short %d on %s\n",
576 m->m_pkthdr.rcvif->if_xname);
581 if ((m = m_pullup(m, len)) == NULL) {
582 carpstats.carps_hdrops++;
585 ip = mtod(m, struct ip *);
586 ch = (struct carp_header *)((char *)ip + iplen);
588 /* verify the CARP checksum */
590 if (carp_cksum(m, len - iplen)) {
591 carpstats.carps_badsum++;
592 CARP_LOG("carp_input: checksum failed on %s\n",
593 m->m_pkthdr.rcvif->if_xname);
599 carp_input_c(m, ch, AF_INET);
604 carp6_input(struct mbuf **mp, int *offp, int proto)
606 struct mbuf *m = *mp;
607 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
608 struct carp_header *ch;
611 carpstats.carps_ipackets6++;
613 if (!carp_opts[CARPCTL_ALLOW]) {
615 return (IPPROTO_DONE);
618 /* check if received on a valid carp interface */
619 if (m->m_pkthdr.rcvif->if_carp == NULL) {
620 carpstats.carps_badif++;
621 CARP_LOG("carp6_input: packet received on non-carp "
623 m->m_pkthdr.rcvif->if_xname);
625 return (IPPROTO_DONE);
628 /* verify that the IP TTL is 255 */
629 if (ip6->ip6_hlim != CARP_DFLTTL) {
630 carpstats.carps_badttl++;
631 CARP_LOG("carp6_input: received ttl %d != 255 on %s\n",
633 m->m_pkthdr.rcvif->if_xname);
635 return (IPPROTO_DONE);
638 /* verify that we have a complete carp packet */
640 IP6_EXTHDR_GET(ch, struct carp_header *, m, *offp, sizeof(*ch));
642 carpstats.carps_badlen++;
643 CARP_LOG("carp6_input: packet size %u too small\n", len);
644 return (IPPROTO_DONE);
648 /* verify the CARP checksum */
650 if (carp_cksum(m, sizeof(*ch))) {
651 carpstats.carps_badsum++;
652 CARP_LOG("carp6_input: checksum failed, on %s\n",
653 m->m_pkthdr.rcvif->if_xname);
655 return (IPPROTO_DONE);
659 carp_input_c(m, ch, AF_INET6);
660 return (IPPROTO_DONE);
665 carp_input_c(struct mbuf *m, struct carp_header *ch, sa_family_t af)
667 struct ifnet *ifp = m->m_pkthdr.rcvif;
668 struct carp_softc *sc;
669 uint64_t tmp_counter;
670 struct timeval sc_tv, ch_tv;
672 /* verify that the VHID is valid on the receiving interface */
673 CARP_LOCK(ifp->if_carp);
674 TAILQ_FOREACH(sc, &((struct carp_if *)ifp->if_carp)->vhif_vrs, sc_list)
675 if (sc->sc_vhid == ch->carp_vhid)
678 if (!sc || !((SC2IFP(sc)->if_flags & IFF_UP) && (SC2IFP(sc)->if_flags & IFF_RUNNING))) {
679 carpstats.carps_badvhid++;
680 CARP_UNLOCK(ifp->if_carp);
685 getmicrotime(&SC2IFP(sc)->if_lastchange);
686 SC2IFP(sc)->if_ipackets++;
687 SC2IFP(sc)->if_ibytes += m->m_pkthdr.len;
689 if (SC2IFP(sc)->if_bpf) {
690 struct ip *ip = mtod(m, struct ip *);
692 /* BPF wants net byte order */
693 ip->ip_len = htons(ip->ip_len + (ip->ip_hl << 2));
694 ip->ip_off = htons(ip->ip_off);
695 bpf_mtap(SC2IFP(sc)->if_bpf, m);
698 /* verify the CARP version. */
699 if (ch->carp_version != CARP_VERSION) {
700 carpstats.carps_badver++;
701 SC2IFP(sc)->if_ierrors++;
702 CARP_UNLOCK(ifp->if_carp);
703 CARP_LOG("%s; invalid version %d\n",
704 SC2IFP(sc)->if_xname,
710 /* verify the hash */
711 if (carp_hmac_verify(sc, ch->carp_counter, ch->carp_md)) {
712 carpstats.carps_badauth++;
713 SC2IFP(sc)->if_ierrors++;
714 CARP_UNLOCK(ifp->if_carp);
715 CARP_LOG("%s: incorrect hash\n", SC2IFP(sc)->if_xname);
720 tmp_counter = ntohl(ch->carp_counter[0]);
721 tmp_counter = tmp_counter<<32;
722 tmp_counter += ntohl(ch->carp_counter[1]);
724 /* XXX Replay protection goes here */
726 sc->sc_init_counter = 0;
727 sc->sc_counter = tmp_counter;
729 sc_tv.tv_sec = sc->sc_advbase;
730 if (carp_suppress_preempt && sc->sc_advskew < 240)
731 sc_tv.tv_usec = 240 * 1000000 / 256;
733 sc_tv.tv_usec = sc->sc_advskew * 1000000 / 256;
734 ch_tv.tv_sec = ch->carp_advbase;
735 ch_tv.tv_usec = ch->carp_advskew * 1000000 / 256;
737 switch (sc->sc_state) {
742 * If we receive an advertisement from a master who's going to
743 * be more frequent than us, go into BACKUP state.
745 if (timevalcmp(&sc_tv, &ch_tv, >) ||
746 timevalcmp(&sc_tv, &ch_tv, ==)) {
747 callout_stop(&sc->sc_ad_tmo);
748 CARP_DEBUG("%s: MASTER -> BACKUP "
749 "(more frequent advertisement received)\n",
750 SC2IFP(sc)->if_xname);
751 carp_set_state(sc, BACKUP);
753 carp_setroute(sc, RTM_DELETE);
758 * If we're pre-empting masters who advertise slower than us,
759 * and this one claims to be slower, treat him as down.
761 if (carp_opts[CARPCTL_PREEMPT] &&
762 timevalcmp(&sc_tv, &ch_tv, <)) {
763 CARP_DEBUG("%s: BACKUP -> MASTER "
764 "(preempting a slower master)\n",
765 SC2IFP(sc)->if_xname);
766 carp_master_down_locked(sc);
771 * If the master is going to advertise at such a low frequency
772 * that he's guaranteed to time out, we'd might as well just
773 * treat him as timed out now.
775 sc_tv.tv_sec = sc->sc_advbase * 3;
776 if (timevalcmp(&sc_tv, &ch_tv, <)) {
777 CARP_DEBUG("%s: BACKUP -> MASTER "
778 "(master timed out)\n",
779 SC2IFP(sc)->if_xname);
780 carp_master_down_locked(sc);
785 * Otherwise, we reset the counter and wait for the next
792 CARP_UNLOCK(ifp->if_carp);
799 carp_prepare_ad(struct mbuf *m, struct carp_softc *sc, struct carp_header *ch)
802 struct ifnet *ifp = SC2IFP(sc);
804 if (sc->sc_init_counter) {
805 /* this could also be seconds since unix epoch */
806 sc->sc_counter = karc4random();
807 sc->sc_counter = sc->sc_counter << 32;
808 sc->sc_counter += karc4random();
812 ch->carp_counter[0] = htonl((sc->sc_counter>>32)&0xffffffff);
813 ch->carp_counter[1] = htonl(sc->sc_counter&0xffffffff);
815 carp_hmac_generate(sc, ch->carp_counter, ch->carp_md);
817 /* Tag packet for carp_output */
818 mtag = m_tag_get(PACKET_TAG_CARP, sizeof(struct ifnet *), MB_DONTWAIT);
821 SC2IFP(sc)->if_oerrors++;
824 bcopy(&ifp, (caddr_t)(mtag + 1), sizeof(struct ifnet *));
825 m_tag_prepend(m, mtag);
831 carp_send_ad_all(void)
833 struct carp_softc *sc;
835 LIST_FOREACH(sc, &carpif_list, sc_next) {
836 if (sc->sc_carpdev == NULL)
839 if ((SC2IFP(sc)->if_flags & IFF_UP) && (SC2IFP(sc)->if_flags & IFF_RUNNING) &&
840 sc->sc_state == MASTER)
841 carp_send_ad_locked(sc);
847 carp_send_ad(void *v)
849 struct carp_softc *sc = v;
852 carp_send_ad_locked(sc);
857 carp_send_ad_locked(struct carp_softc *sc)
859 struct carp_header ch;
861 struct carp_header *ch_ptr;
863 int len, advbase, advskew;
866 /* bow out if we've lost our UPness or RUNNINGuiness */
867 if (!((SC2IFP(sc)->if_flags & IFF_UP) && (SC2IFP(sc)->if_flags & IFF_RUNNING))) {
871 advbase = sc->sc_advbase;
872 if (!carp_suppress_preempt || sc->sc_advskew > 240)
873 advskew = sc->sc_advskew;
877 tv.tv_usec = advskew * 1000000 / 256;
880 ch.carp_version = CARP_VERSION;
881 ch.carp_type = CARP_ADVERTISEMENT;
882 ch.carp_vhid = sc->sc_vhid;
883 ch.carp_advbase = advbase;
884 ch.carp_advskew = advskew;
885 ch.carp_authlen = 7; /* XXX DEFINE */
886 ch.carp_pad1 = 0; /* must be zero */
893 MGETHDR(m, M_NOWAIT, MT_HEADER);
895 SC2IFP(sc)->if_oerrors++;
896 carpstats.carps_onomem++;
897 /* XXX maybe less ? */
898 if (advbase != 255 || advskew != 255)
899 callout_reset(&sc->sc_ad_tmo, tvtohz_high(&tv),
903 len = sizeof(*ip) + sizeof(ch);
904 m->m_pkthdr.len = len;
905 m->m_pkthdr.rcvif = NULL;
907 MH_ALIGN(m, m->m_len);
908 m->m_flags |= M_MCAST;
909 ip = mtod(m, struct ip *);
910 ip->ip_v = IPVERSION;
911 ip->ip_hl = sizeof(*ip) >> 2;
912 ip->ip_tos = IPTOS_LOWDELAY;
914 ip->ip_id = ip_newid();
916 ip->ip_ttl = CARP_DFLTTL;
917 ip->ip_p = IPPROTO_CARP;
919 ip->ip_src.s_addr = sc->sc_ia->ia_addr.sin_addr.s_addr;
920 ip->ip_dst.s_addr = htonl(INADDR_CARP_GROUP);
922 ch_ptr = (struct carp_header *)(&ip[1]);
923 bcopy(&ch, ch_ptr, sizeof(ch));
924 if (carp_prepare_ad(m, sc, ch_ptr))
927 m->m_data += sizeof(*ip);
928 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip));
929 m->m_data -= sizeof(*ip);
931 getmicrotime(&SC2IFP(sc)->if_lastchange);
932 SC2IFP(sc)->if_opackets++;
933 SC2IFP(sc)->if_obytes += len;
934 carpstats.carps_opackets++;
936 if (ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, NULL)) {
937 SC2IFP(sc)->if_oerrors++;
938 if (sc->sc_sendad_errors < INT_MAX)
939 sc->sc_sendad_errors++;
940 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) {
941 carp_suppress_preempt++;
942 if (carp_suppress_preempt == 1) {
948 sc->sc_sendad_success = 0;
950 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) {
951 if (++sc->sc_sendad_success >=
952 CARP_SENDAD_MIN_SUCCESS) {
953 carp_suppress_preempt--;
954 sc->sc_sendad_errors = 0;
957 sc->sc_sendad_errors = 0;
965 MGETHDR(m, M_NOWAIT, MT_HEADER);
967 SC2IFP(sc)->if_oerrors++;
968 carpstats.carps_onomem++;
969 /* XXX maybe less ? */
970 if (advbase != 255 || advskew != 255)
971 callout_reset(&sc->sc_ad_tmo, tvtohz_high(&tv),
975 len = sizeof(*ip6) + sizeof(ch);
976 m->m_pkthdr.len = len;
977 m->m_pkthdr.rcvif = NULL;
979 MH_ALIGN(m, m->m_len);
980 m->m_flags |= M_MCAST;
981 ip6 = mtod(m, struct ip6_hdr *);
982 bzero(ip6, sizeof(*ip6));
983 ip6->ip6_vfc |= IPV6_VERSION;
984 ip6->ip6_hlim = CARP_DFLTTL;
985 ip6->ip6_nxt = IPPROTO_CARP;
986 bcopy(&sc->sc_ia6->ia_addr.sin6_addr, &ip6->ip6_src,
987 sizeof(struct in6_addr));
988 /* set the multicast destination */
990 ip6->ip6_dst.s6_addr16[0] = htons(0xff02);
991 ip6->ip6_dst.s6_addr8[15] = 0x12;
992 if (in6_setscope(&ip6->ip6_dst, sc->sc_carpdev, NULL) != 0) {
993 SC2IFP(sc)->if_oerrors++;
995 CARP_LOG("%s: in6_setscope failed\n", __func__);
999 ch_ptr = (struct carp_header *)(&ip6[1]);
1000 bcopy(&ch, ch_ptr, sizeof(ch));
1001 if (carp_prepare_ad(m, sc, ch_ptr))
1004 m->m_data += sizeof(*ip6);
1005 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip6));
1006 m->m_data -= sizeof(*ip6);
1008 getmicrotime(&SC2IFP(sc)->if_lastchange);
1009 SC2IFP(sc)->if_opackets++;
1010 SC2IFP(sc)->if_obytes += len;
1011 carpstats.carps_opackets6++;
1013 if (ip6_output(m, NULL, NULL, 0, &sc->sc_im6o, NULL, NULL)) {
1014 SC2IFP(sc)->if_oerrors++;
1015 if (sc->sc_sendad_errors < INT_MAX)
1016 sc->sc_sendad_errors++;
1017 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) {
1018 carp_suppress_preempt++;
1019 if (carp_suppress_preempt == 1) {
1025 sc->sc_sendad_success = 0;
1027 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) {
1028 if (++sc->sc_sendad_success >=
1029 CARP_SENDAD_MIN_SUCCESS) {
1030 carp_suppress_preempt--;
1031 sc->sc_sendad_errors = 0;
1034 sc->sc_sendad_errors = 0;
1039 if (advbase != 255 || advskew != 255)
1040 callout_reset(&sc->sc_ad_tmo, tvtohz_high(&tv),
1046 * Broadcast a gratuitous ARP request containing
1047 * the virtual router MAC address for each IP address
1048 * associated with the virtual router.
1051 carp_send_arp(struct carp_softc *sc)
1053 struct ifaddr_container *ifac;
1055 TAILQ_FOREACH(ifac, &SC2IFP(sc)->if_addrheads[mycpuid], ifa_link) {
1056 struct ifaddr *ifa = ifac->ifa;
1058 if (ifa->ifa_addr->sa_family != AF_INET)
1060 arp_ifinit2(sc->sc_carpdev, ifa, IF_LLADDR(sc->sc_ifp));
1062 DELAY(1000); /* XXX */
1068 carp_send_na(struct carp_softc *sc)
1070 struct ifaddr_container *ifac;
1071 struct in6_addr *in6;
1072 static struct in6_addr mcast = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
1074 TAILQ_FOREACH(ifac, &SC2IFP(sc)->if_addrheads[mycpuid], ifa_link) {
1075 struct ifaddr *ifa = ifac->ifa;
1077 if (ifa->ifa_addr->sa_family != AF_INET6)
1080 in6 = &ifatoia6(ifa)->ia_addr.sin6_addr;
1081 nd6_na_output(sc->sc_carpdev, &mcast, in6,
1082 ND_NA_FLAG_OVERRIDE, 1, NULL);
1083 DELAY(1000); /* XXX */
1089 carp_addrcount(struct carp_if *cif, struct in_ifaddr *ia, int type)
1091 struct carp_softc *vh;
1094 CARP_LOCK_ASSERT(cif);
1096 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1097 if ((type == CARP_COUNT_RUNNING &&
1098 (SC2IFP(vh)->if_flags & IFF_UP) && (SC2IFP(vh)->if_flags & IFF_RUNNING)) ||
1099 (type == CARP_COUNT_MASTER && vh->sc_state == MASTER)) {
1100 struct ifaddr_container *ifac;
1102 TAILQ_FOREACH(ifac, &SC2IFP(vh)->if_addrheads[mycpuid],
1104 struct ifaddr *ifa = ifac->ifa;
1106 if (ifa->ifa_addr->sa_family == AF_INET &&
1107 ia->ia_addr.sin_addr.s_addr ==
1108 ifatoia(ifa)->ia_addr.sin_addr.s_addr)
1117 carp_iamatch(void *v, struct in_ifaddr *ia,
1118 struct in_addr *isaddr, uint8_t **enaddr)
1120 struct carp_if *cif = v;
1121 struct carp_softc *vh;
1122 int index, count = 0;
1126 if (carp_opts[CARPCTL_ARPBALANCE]) {
1128 * XXX proof of concept implementation.
1129 * We use the source ip to decide which virtual host should
1130 * handle the request. If we're master of that virtual host,
1131 * then we respond, otherwise, just drop the arp packet on
1134 count = carp_addrcount(cif, ia, CARP_COUNT_RUNNING);
1136 /* should never reach this */
1141 /* this should be a hash, like pf_hash() */
1142 index = ntohl(isaddr->s_addr) % count;
1145 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1146 if ((SC2IFP(vh)->if_flags & IFF_UP) && (SC2IFP(vh)->if_flags & IFF_RUNNING)) {
1147 struct ifaddr_container *ifac;
1149 TAILQ_FOREACH(ifac, &SC2IFP(vh)->if_addrheads[mycpuid], ifa_link) {
1150 struct ifaddr *ifa = ifac->ifa;
1152 if (ifa->ifa_addr->sa_family ==
1154 ia->ia_addr.sin_addr.s_addr ==
1155 ifatoia(ifa)->ia_addr.sin_addr.s_addr) {
1156 if (count == index) {
1159 *enaddr = IF_LLADDR(vh->sc_ifp);
1173 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1174 if ((SC2IFP(vh)->if_flags & IFF_UP) && (SC2IFP(vh)->if_flags & IFF_RUNNING) &&
1175 vh->sc_state == MASTER) {
1176 *enaddr = IF_LLADDR(vh->sc_ifp);
1188 carp_iamatch6(void *v, struct in6_addr *taddr)
1190 struct carp_if *cif = v;
1191 struct carp_softc *vh;
1194 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) {
1195 struct ifaddr_container *ifac;
1197 TAILQ_FOREACH(ifac, &SC2IFP(vh)->if_addrheads[mycpuid], ifa_link) {
1198 struct ifaddr *ifa = ifac->ifa;
1200 if (IN6_ARE_ADDR_EQUAL(taddr,
1201 &ifatoia6(ifa)->ia_addr.sin6_addr) &&
1202 (SC2IFP(vh)->if_flags & IFF_UP) && (SC2IFP(vh)->if_flags & IFF_RUNNING) &&
1203 vh->sc_state == MASTER) {
1215 carp_macmatch6(void *v, struct mbuf *m, const struct in6_addr *taddr)
1218 struct carp_if *cif = v;
1219 struct carp_softc *sc;
1222 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list) {
1223 struct ifaddr_container *ifac;
1225 TAILQ_FOREACH(ifac, &SC2IFP(sc)->if_addrheads[mycpuid], ifa_link) {
1226 struct ifaddr *ifa = ifac->ifa;
1228 if (IN6_ARE_ADDR_EQUAL(taddr,
1229 &ifatoia6(ifa)->ia_addr.sin6_addr) &&
1230 (SC2IFP(sc)->if_flags & IFF_UP) && (SC2IFP(sc)->if_flags & IFF_RUNNING)) {
1231 struct ifnet *ifp = SC2IFP(sc);
1232 mtag = m_tag_get(PACKET_TAG_CARP,
1233 sizeof(struct ifnet *), MB_DONTWAIT);
1235 /* better a bit than nothing */
1237 return (IF_LLADDR(sc->sc_ifp));
1239 bcopy(&ifp, (caddr_t)(mtag + 1),
1240 sizeof(struct ifnet *));
1241 m_tag_prepend(m, mtag);
1244 return (IF_LLADDR(sc->sc_ifp));
1255 carp_forus(void *v, void *dhost)
1257 struct carp_if *cif = v;
1258 struct carp_softc *vh;
1259 uint8_t *ena = dhost;
1262 * XXX: See here for check on MAC adr is not for virtual use
1266 if (ena[0] || ena[1] || ena[2] != 0x5e || ena[3] || ena[4] != 1)
1272 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list)
1273 if ((SC2IFP(vh)->if_flags & IFF_UP) && (SC2IFP(vh)->if_flags & IFF_RUNNING) &&
1274 vh->sc_state == MASTER &&
1275 !bcmp(dhost, IF_LLADDR(vh->sc_ifp), ETHER_ADDR_LEN)) {
1277 return (SC2IFP(vh));
1285 carp_master_down(void *v)
1287 struct carp_softc *sc = v;
1289 lwkt_serialize_enter(sc->sc_ifp->if_serializer);
1290 carp_master_down_locked(sc);
1291 lwkt_serialize_exit(sc->sc_ifp->if_serializer);
1295 carp_master_down_locked(struct carp_softc *sc)
1298 CARP_SCLOCK_ASSERT(sc);
1300 switch (sc->sc_state) {
1302 kprintf("%s: master_down event in INIT state\n",
1303 SC2IFP(sc)->if_xname);
1308 carp_set_state(sc, MASTER);
1309 carp_send_ad_locked(sc);
1315 carp_setroute(sc, RTM_ADD);
1321 * When in backup state, af indicates whether to reset the master down timer
1322 * for v4 or v6. If it's set to zero, reset the ones which are already pending.
1325 carp_setrun(struct carp_softc *sc, sa_family_t af)
1329 if (sc->sc_carpdev == NULL) {
1330 SC2IFP(sc)->if_flags &= ~IFF_RUNNING;
1331 carp_set_state(sc, INIT);
1335 if (SC2IFP(sc)->if_flags & IFF_UP &&
1336 sc->sc_vhid > 0 && (sc->sc_naddrs || sc->sc_naddrs6))
1337 SC2IFP(sc)->if_flags |= IFF_RUNNING;
1339 SC2IFP(sc)->if_flags &= ~IFF_RUNNING;
1340 carp_setroute(sc, RTM_DELETE);
1344 switch (sc->sc_state) {
1346 if (carp_opts[CARPCTL_PREEMPT] && !carp_suppress_preempt) {
1347 carp_send_ad_locked(sc);
1352 CARP_DEBUG("%s: INIT -> MASTER (preempting)\n",
1353 SC2IFP(sc)->if_xname);
1354 carp_set_state(sc, MASTER);
1355 carp_setroute(sc, RTM_ADD);
1357 CARP_DEBUG("%s: INIT -> BACKUP\n", SC2IFP(sc)->if_xname);
1358 carp_set_state(sc, BACKUP);
1359 carp_setroute(sc, RTM_DELETE);
1364 callout_stop(&sc->sc_ad_tmo);
1365 tv.tv_sec = 3 * sc->sc_advbase;
1366 tv.tv_usec = sc->sc_advskew * 1000000 / 256;
1370 callout_reset(&sc->sc_md_tmo, tvtohz_high(&tv),
1371 carp_master_down, sc);
1376 callout_reset(&sc->sc_md6_tmo, tvtohz_high(&tv),
1377 carp_master_down, sc);
1382 callout_reset(&sc->sc_md_tmo, tvtohz_high(&tv),
1383 carp_master_down, sc);
1385 callout_reset(&sc->sc_md6_tmo, tvtohz_high(&tv),
1386 carp_master_down, sc);
1391 tv.tv_sec = sc->sc_advbase;
1392 tv.tv_usec = sc->sc_advskew * 1000000 / 256;
1393 callout_reset(&sc->sc_ad_tmo, tvtohz_high(&tv),
1400 carp_multicast_cleanup(struct carp_softc *sc)
1402 struct ip_moptions *imo = &sc->sc_imo;
1403 uint16_t n = imo->imo_num_memberships;
1405 /* Clean up our own multicast memberships */
1407 if (imo->imo_membership[n] != NULL) {
1408 in_delmulti(imo->imo_membership[n]);
1409 imo->imo_membership[n] = NULL;
1412 imo->imo_num_memberships = 0;
1413 imo->imo_multicast_ifp = NULL;
1418 carp_multicast6_cleanup(struct carp_softc *sc)
1420 struct ip6_moptions *im6o = &sc->sc_im6o;
1422 while (!LIST_EMPTY(&im6o->im6o_memberships)) {
1423 struct in6_multi_mship *imm =
1424 LIST_FIRST(&im6o->im6o_memberships);
1426 LIST_REMOVE(imm, i6mm_chain);
1427 in6_leavegroup(imm);
1429 im6o->im6o_multicast_ifp = NULL;
1434 carp_set_addr(struct carp_softc *sc, struct sockaddr_in *sin)
1437 struct carp_if *cif;
1438 struct in_ifaddr *ia, *ia_if;
1439 struct in_ifaddr_container *iac;
1440 struct ip_moptions *imo = &sc->sc_imo;
1441 struct in_addr addr;
1442 u_long iaddr = htonl(sin->sin_addr.s_addr);
1445 if (sin->sin_addr.s_addr == 0)
1447 if (!(SC2IFP(sc)->if_flags & IFF_UP))
1449 carp_set_state(sc, INIT);
1453 SC2IFP(sc)->if_flags |= IFF_UP;
1458 /* we have to do it by hands to check we won't match on us */
1459 ia_if = NULL; own = 0;
1460 TAILQ_FOREACH(iac, &in_ifaddrheads[mycpuid], ia_link) {
1463 /* and, yeah, we need a multicast-capable iface too */
1464 if (ia->ia_ifp != SC2IFP(sc) &&
1465 (ia->ia_ifp->if_flags & IFF_MULTICAST) &&
1466 (iaddr & ia->ia_subnetmask) == ia->ia_subnet) {
1469 if (sin->sin_addr.s_addr ==
1470 ia->ia_addr.sin_addr.s_addr)
1477 return (EADDRNOTAVAIL);
1482 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0 ||
1483 (imo->imo_multicast_ifp && imo->imo_multicast_ifp != ifp))
1484 return (EADDRNOTAVAIL);
1486 if (imo->imo_num_memberships == 0) {
1487 addr.s_addr = htonl(INADDR_CARP_GROUP);
1488 if ((imo->imo_membership[0] = in_addmulti(&addr, ifp)) == NULL)
1490 imo->imo_num_memberships++;
1491 imo->imo_multicast_ifp = ifp;
1492 imo->imo_multicast_ttl = CARP_DFLTTL;
1493 imo->imo_multicast_loop = 0;
1496 if (!ifp->if_carp) {
1498 MALLOC(cif, struct carp_if *, sizeof(*cif), M_CARP,
1500 if ((error = ifpromisc(ifp, 1))) {
1505 CARP_LOCK_INIT(cif);
1507 cif->vhif_ifp = ifp;
1508 TAILQ_INIT(&cif->vhif_vrs);
1512 struct carp_softc *vr;
1514 cif = (struct carp_if *)ifp->if_carp;
1516 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list)
1517 if (vr != sc && vr->sc_vhid == sc->sc_vhid) {
1524 sc->sc_carpdev = ifp;
1526 { /* XXX prevent endless loop if already in queue */
1527 struct carp_softc *vr, *after = NULL;
1529 cif = (struct carp_if *)ifp->if_carp;
1531 /* XXX: cif should not change, right? So we still hold the lock */
1532 CARP_LOCK_ASSERT(cif);
1534 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) {
1537 if (vr->sc_vhid < sc->sc_vhid)
1542 /* We're trying to keep things in order */
1543 if (after == NULL) {
1544 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list);
1546 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, sc, sc_list);
1553 SC2IFP(sc)->if_flags |= IFF_UP;
1558 carp_sc_state_locked(sc);
1566 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]);
1572 carp_del_addr(struct carp_softc *sc, struct sockaddr_in *sin)
1576 if (!--sc->sc_naddrs) {
1577 struct carp_if *cif = (struct carp_if *)sc->sc_carpdev->if_carp;
1578 struct ip_moptions *imo = &sc->sc_imo;
1581 callout_stop(&sc->sc_ad_tmo);
1582 SC2IFP(sc)->if_flags &= ~IFF_UP;
1583 SC2IFP(sc)->if_flags &= ~IFF_RUNNING;
1585 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]);
1586 imo->imo_multicast_ifp = NULL;
1587 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list);
1588 if (!--cif->vhif_nvrs) {
1589 sc->sc_carpdev->if_carp = NULL;
1590 CARP_LOCK_DESTROY(cif);
1591 FREE(cif, M_IFADDR);
1602 carp_set_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6)
1605 struct carp_if *cif;
1606 struct in6_ifaddr *ia, *ia_if;
1607 struct ip6_moptions *im6o = &sc->sc_im6o;
1608 struct in6_multi_mship *imm;
1609 struct in6_addr in6;
1612 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1613 if (!(SC2IFP(sc)->if_flags & IFF_UP))
1614 carp_set_state(sc, INIT);
1616 SC2IFP(sc)->if_flags |= IFF_UP;
1621 /* we have to do it by hands to check we won't match on us */
1622 ia_if = NULL; own = 0;
1623 for (ia = in6_ifaddr; ia; ia = ia->ia_next) {
1626 for (i = 0; i < 4; i++) {
1627 if ((sin6->sin6_addr.s6_addr32[i] &
1628 ia->ia_prefixmask.sin6_addr.s6_addr32[i]) !=
1629 (ia->ia_addr.sin6_addr.s6_addr32[i] &
1630 ia->ia_prefixmask.sin6_addr.s6_addr32[i]))
1633 /* and, yeah, we need a multicast-capable iface too */
1634 if (ia->ia_ifp != SC2IFP(sc) &&
1635 (ia->ia_ifp->if_flags & IFF_MULTICAST) &&
1639 if (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr,
1640 &ia->ia_addr.sin6_addr))
1646 return (EADDRNOTAVAIL);
1650 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0 ||
1651 (im6o->im6o_multicast_ifp && im6o->im6o_multicast_ifp != ifp))
1652 return (EADDRNOTAVAIL);
1654 if (!sc->sc_naddrs6) {
1655 im6o->im6o_multicast_ifp = ifp;
1657 /* join CARP multicast address */
1658 bzero(&in6, sizeof(in6));
1659 in6.s6_addr16[0] = htons(0xff02);
1660 in6.s6_addr8[15] = 0x12;
1661 if (in6_setscope(&in6, ifp, NULL) != 0)
1663 if ((imm = in6_joingroup(ifp, &in6, &error)) == NULL)
1665 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1667 /* join solicited multicast address */
1668 bzero(&in6, sizeof(in6));
1669 in6.s6_addr16[0] = htons(0xff02);
1670 in6.s6_addr32[1] = 0;
1671 in6.s6_addr32[2] = htonl(1);
1672 in6.s6_addr32[3] = sin6->sin6_addr.s6_addr32[3];
1673 in6.s6_addr8[12] = 0xff;
1674 if (in6_setscope(&in6, ifp, NULL) != 0)
1676 if ((imm = in6_joingroup(ifp, &in6, &error)) == NULL)
1678 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain);
1681 if (!ifp->if_carp) {
1682 MALLOC(cif, struct carp_if *, sizeof(*cif), M_CARP,
1684 if ((error = ifpromisc(ifp, 1))) {
1689 CARP_LOCK_INIT(cif);
1691 cif->vhif_ifp = ifp;
1692 TAILQ_INIT(&cif->vhif_vrs);
1696 struct carp_softc *vr;
1698 cif = (struct carp_if *)ifp->if_carp;
1700 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list)
1701 if (vr != sc && vr->sc_vhid == sc->sc_vhid) {
1708 sc->sc_carpdev = ifp;
1710 { /* XXX prevent endless loop if already in queue */
1711 struct carp_softc *vr, *after = NULL;
1713 cif = (struct carp_if *)ifp->if_carp;
1714 CARP_LOCK_ASSERT(cif);
1716 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) {
1719 if (vr->sc_vhid < sc->sc_vhid)
1724 /* We're trying to keep things in order */
1725 if (after == NULL) {
1726 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list);
1728 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, sc, sc_list);
1735 SC2IFP(sc)->if_flags |= IFF_UP;
1738 carp_sc_state_locked(sc);
1746 /* clean up multicast memberships */
1747 if (!sc->sc_naddrs6) {
1748 while (!LIST_EMPTY(&im6o->im6o_memberships)) {
1749 imm = LIST_FIRST(&im6o->im6o_memberships);
1750 LIST_REMOVE(imm, i6mm_chain);
1751 in6_leavegroup(imm);
1758 carp_del_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6)
1762 if (!--sc->sc_naddrs6) {
1763 struct carp_if *cif = (struct carp_if *)sc->sc_carpdev->if_carp;
1764 struct ip6_moptions *im6o = &sc->sc_im6o;
1767 callout_stop(&sc->sc_ad_tmo);
1768 SC2IFP(sc)->if_flags &= ~IFF_UP;
1769 SC2IFP(sc)->if_flags &= ~IFF_RUNNING;
1771 while (!LIST_EMPTY(&im6o->im6o_memberships)) {
1772 struct in6_multi_mship *imm =
1773 LIST_FIRST(&im6o->im6o_memberships);
1775 LIST_REMOVE(imm, i6mm_chain);
1776 in6_leavegroup(imm);
1778 im6o->im6o_multicast_ifp = NULL;
1779 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list);
1780 if (!--cif->vhif_nvrs) {
1781 CARP_LOCK_DESTROY(cif);
1782 sc->sc_carpdev->if_carp = NULL;
1783 FREE(cif, M_IFADDR);
1793 carp_ioctl(struct ifnet *ifp, u_long cmd, caddr_t addr, struct ucred *creds)
1795 struct carp_softc *sc = ifp->if_softc, *vr;
1796 struct carpreq carpr;
1799 struct ifaliasreq *ifra;
1800 int locked = 0, error = 0;
1802 ifa = (struct ifaddr *)addr;
1803 ifra = (struct ifaliasreq *)addr;
1804 ifr = (struct ifreq *)addr;
1809 switch (ifa->ifa_addr->sa_family) {
1812 SC2IFP(sc)->if_flags |= IFF_UP;
1813 bcopy(ifa->ifa_addr, ifa->ifa_dstaddr,
1814 sizeof(struct sockaddr));
1815 error = carp_set_addr(sc, satosin(ifa->ifa_addr));
1820 SC2IFP(sc)->if_flags |= IFF_UP;
1821 error = carp_set_addr6(sc, satosin6(ifa->ifa_addr));
1825 error = EAFNOSUPPORT;
1831 switch (ifa->ifa_addr->sa_family) {
1834 SC2IFP(sc)->if_flags |= IFF_UP;
1835 bcopy(ifa->ifa_addr, ifa->ifa_dstaddr,
1836 sizeof(struct sockaddr));
1837 error = carp_set_addr(sc, satosin(&ifra->ifra_addr));
1842 SC2IFP(sc)->if_flags |= IFF_UP;
1843 error = carp_set_addr6(sc, satosin6(&ifra->ifra_addr));
1847 error = EAFNOSUPPORT;
1853 switch (ifa->ifa_addr->sa_family) {
1856 error = carp_del_addr(sc, satosin(&ifra->ifra_addr));
1861 error = carp_del_addr6(sc, satosin6(&ifra->ifra_addr));
1865 error = EAFNOSUPPORT;
1871 if (sc->sc_carpdev) {
1875 if (sc->sc_state != INIT && !(ifr->ifr_flags & IFF_UP)) {
1876 callout_stop(&sc->sc_ad_tmo);
1877 callout_stop(&sc->sc_md_tmo);
1878 callout_stop(&sc->sc_md6_tmo);
1879 if (sc->sc_state == MASTER)
1880 carp_send_ad_locked(sc);
1881 carp_set_state(sc, INIT);
1883 } else if (sc->sc_state == INIT && (ifr->ifr_flags & IFF_UP)) {
1884 SC2IFP(sc)->if_flags |= IFF_UP;
1890 error = suser(curthread);
1893 if ((error = copyin(ifr->ifr_data, &carpr, sizeof carpr)))
1896 if (sc->sc_carpdev) {
1900 if (sc->sc_state != INIT && carpr.carpr_state != sc->sc_state) {
1901 switch (carpr.carpr_state) {
1903 callout_stop(&sc->sc_ad_tmo);
1904 carp_set_state(sc, BACKUP);
1906 carp_setroute(sc, RTM_DELETE);
1909 carp_master_down_locked(sc);
1915 if (carpr.carpr_vhid > 0) {
1916 if (carpr.carpr_vhid > 255) {
1920 if (sc->sc_carpdev) {
1921 struct carp_if *cif;
1922 cif = (struct carp_if *)sc->sc_carpdev->if_carp;
1923 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list)
1925 vr->sc_vhid == carpr.carpr_vhid)
1928 sc->sc_vhid = carpr.carpr_vhid;
1929 IF_LLADDR(sc->sc_ifp)[0] = 0;
1930 IF_LLADDR(sc->sc_ifp)[1] = 0;
1931 IF_LLADDR(sc->sc_ifp)[2] = 0x5e;
1932 IF_LLADDR(sc->sc_ifp)[3] = 0;
1933 IF_LLADDR(sc->sc_ifp)[4] = 1;
1934 IF_LLADDR(sc->sc_ifp)[5] = sc->sc_vhid;
1937 if (carpr.carpr_advbase > 0 || carpr.carpr_advskew > 0) {
1938 if (carpr.carpr_advskew >= 255) {
1942 if (carpr.carpr_advbase > 255) {
1946 sc->sc_advbase = carpr.carpr_advbase;
1947 sc->sc_advskew = carpr.carpr_advskew;
1950 bcopy(carpr.carpr_key, sc->sc_key, sizeof(sc->sc_key));
1960 /* XXX: lockless read */
1961 bzero(&carpr, sizeof(carpr));
1962 carpr.carpr_state = sc->sc_state;
1963 carpr.carpr_vhid = sc->sc_vhid;
1964 carpr.carpr_advbase = sc->sc_advbase;
1965 carpr.carpr_advskew = sc->sc_advskew;
1966 error = suser(curthread);
1968 bcopy(sc->sc_key, carpr.carpr_key,
1969 sizeof(carpr.carpr_key));
1970 error = copyout(&carpr, ifr->ifr_data, sizeof(carpr));
1980 carp_hmac_prepare(sc);
1986 * XXX: this is looutput. We should eventually use it from there.
1989 carp_looutput(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
1994 M_ASSERTPKTHDR(m); /* check if we have the packet header */
1996 if (rt && rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) {
1998 return (rt->rt_flags & RTF_BLACKHOLE ? 0 :
1999 rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH);
2003 ifp->if_obytes += m->m_pkthdr.len;
2005 /* BPF writes need to be handled specially. */
2006 if (dst->sa_family == AF_UNSPEC) {
2007 bcopy(dst->sa_data, &af, sizeof(af));
2008 dst->sa_family = af;
2012 switch (dst->sa_family) {
2020 return (EAFNOSUPPORT);
2023 return(if_simloop(ifp, m, dst->sa_family, 0));
2027 * Start output on carp interface. This function should never be called.
2030 carp_start(struct ifnet *ifp)
2033 kprintf("%s: start called\n", ifp->if_xname);
2038 carp_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2042 struct carp_softc *sc;
2043 struct ifnet *carp_ifp;
2048 switch (sa->sa_family) {
2061 mtag = m_tag_find(m, PACKET_TAG_CARP, NULL);
2065 bcopy(mtag + 1, &carp_ifp, sizeof(struct ifnet *));
2066 sc = carp_ifp->if_softc;
2068 /* Set the source MAC address to Virtual Router MAC Address */
2069 switch (ifp->if_type) {
2072 struct ether_header *eh;
2074 eh = mtod(m, struct ether_header *);
2075 eh->ether_shost[0] = 0;
2076 eh->ether_shost[1] = 0;
2077 eh->ether_shost[2] = 0x5e;
2078 eh->ether_shost[3] = 0;
2079 eh->ether_shost[4] = 1;
2080 eh->ether_shost[5] = sc->sc_vhid;
2084 kprintf("%s: carp is not supported for this interface type\n",
2086 return (EOPNOTSUPP);
2094 carp_set_state(struct carp_softc *sc, int state)
2098 CARP_SCLOCK_ASSERT(sc);
2100 if (sc->sc_state == state)
2103 sc->sc_state = state;
2106 SC2IFP(sc)->if_link_state = LINK_STATE_DOWN;
2109 SC2IFP(sc)->if_link_state = LINK_STATE_UP;
2112 SC2IFP(sc)->if_link_state = LINK_STATE_UNKNOWN;
2115 rt_ifmsg(SC2IFP(sc));
2119 carp_carpdev_state(void *v)
2121 struct carp_if *cif = v;
2124 carp_carpdev_state_locked(cif);
2129 carp_carpdev_state_locked(struct carp_if *cif)
2131 struct carp_softc *sc;
2133 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list)
2134 carp_sc_state_locked(sc);
2138 carp_sc_state_locked(struct carp_softc *sc)
2140 CARP_SCLOCK_ASSERT(sc);
2142 if ( !(sc->sc_carpdev->if_flags & IFF_UP)) {
2143 sc->sc_flags_backup = SC2IFP(sc)->if_flags;
2144 SC2IFP(sc)->if_flags &= ~IFF_UP;
2145 SC2IFP(sc)->if_flags &= ~IFF_RUNNING;
2146 callout_stop(&sc->sc_ad_tmo);
2147 callout_stop(&sc->sc_md_tmo);
2148 callout_stop(&sc->sc_md6_tmo);
2149 carp_set_state(sc, INIT);
2151 if (!sc->sc_suppress) {
2152 carp_suppress_preempt++;
2153 if (carp_suppress_preempt == 1) {
2159 sc->sc_suppress = 1;
2161 SC2IFP(sc)->if_flags |= sc->sc_flags_backup;
2162 carp_set_state(sc, INIT);
2164 if (sc->sc_suppress)
2165 carp_suppress_preempt--;
2166 sc->sc_suppress = 0;
2173 carp_modevent(module_t mod, int type, void *data)
2177 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2178 carp_ifdetach, NULL, EVENTHANDLER_PRI_ANY);
2179 if (if_detach_event_tag == NULL)
2182 LIST_INIT(&carpif_list);
2183 if_clone_attach(&carp_cloner);
2187 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2188 if_clone_detach(&carp_cloner);
2198 static moduledata_t carp_mod = {
2204 DECLARE_MODULE(carp, carp_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);