2 * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
4 * Copyright (c) 2002 Michael Shalayeff.
5 * Copyright (c) 2003 Ryan McBride.
6 * Copyright (c) 2011 Gleb Smirnoff <glebius@FreeBSD.org>
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
18 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
19 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
20 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
21 * IN NO EVENT SHALL THE AUTHOR OR HIS RELATIVES BE LIABLE FOR ANY DIRECT,
22 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
23 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
24 * SERVICES; LOSS OF MIND, USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
26 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
27 * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
28 * THE POSSIBILITY OF SUCH DAMAGE.
31 #include <sys/cdefs.h>
32 __FBSDID("$FreeBSD$");
36 #include "opt_inet6.h"
38 #include <sys/param.h>
39 #include <sys/systm.h>
42 #include <sys/kernel.h>
43 #include <sys/limits.h>
44 #include <sys/malloc.h>
46 #include <sys/module.h>
49 #include <sys/protosw.h>
50 #include <sys/socket.h>
51 #include <sys/sockio.h>
52 #include <sys/sysctl.h>
53 #include <sys/syslog.h>
54 #include <sys/taskqueue.h>
55 #include <sys/counter.h>
57 #include <net/ethernet.h>
59 #include <net/if_var.h>
60 #include <net/if_dl.h>
61 #include <net/if_llatbl.h>
62 #include <net/if_types.h>
63 #include <net/route.h>
66 #if defined(INET) || defined(INET6)
67 #include <netinet/in.h>
68 #include <netinet/in_var.h>
69 #include <netinet/ip_carp.h>
70 #include <netinet/ip.h>
71 #include <machine/in_cksum.h>
74 #include <netinet/ip_var.h>
75 #include <netinet/if_ether.h>
79 #include <netinet/icmp6.h>
80 #include <netinet/ip6.h>
81 #include <netinet6/in6_var.h>
82 #include <netinet6/ip6_var.h>
83 #include <netinet6/scope6_var.h>
84 #include <netinet6/nd6.h>
87 #include <crypto/sha1.h>
89 static MALLOC_DEFINE(M_CARP, "CARP", "CARP addresses");
92 struct ifnet *sc_carpdev; /* Pointer to parent ifnet. */
93 struct ifaddr **sc_ifas; /* Our ifaddrs. */
94 struct sockaddr_dl sc_addr; /* Our link level address. */
95 struct callout sc_ad_tmo; /* Advertising timeout. */
97 struct callout sc_md_tmo; /* Master down timeout. */
100 struct callout sc_md6_tmo; /* XXX: Master down timeout. */
111 enum { INIT = 0, BACKUP, MASTER } sc_state;
113 int sc_sendad_errors;
114 #define CARP_SENDAD_MAX_ERRORS 3
115 int sc_sendad_success;
116 #define CARP_SENDAD_MIN_SUCCESS 3
122 #define CARP_HMAC_PAD 64
123 unsigned char sc_key[CARP_KEY_LEN];
124 unsigned char sc_pad[CARP_HMAC_PAD];
127 TAILQ_ENTRY(carp_softc) sc_list; /* On the carp_if list. */
128 LIST_ENTRY(carp_softc) sc_next; /* On the global list. */
138 TAILQ_HEAD(, carp_softc) cif_vrs;
140 struct ip_moptions cif_imo;
143 struct ip6_moptions cif_im6o;
145 struct ifnet *cif_ifp;
148 #define CIF_PROMISC 0x00000001
153 static int proto_reg[] = {-1, -1};
156 * Brief design of carp(4).
158 * Any carp-capable ifnet may have a list of carp softcs hanging off
159 * its ifp->if_carp pointer. Each softc represents one unique virtual
160 * host id, or vhid. The softc has a back pointer to the ifnet. All
161 * softcs are joined in a global list, which has quite limited use.
163 * Any interface address that takes part in CARP negotiation has a
164 * pointer to the softc of its vhid, ifa->ifa_carp. That could be either
165 * AF_INET or AF_INET6 address.
167 * Although, one can get the softc's backpointer to ifnet and traverse
168 * through its ifp->if_addrhead queue to find all interface addresses
169 * involved in CARP, we keep a growable array of ifaddr pointers. This
170 * allows us to avoid grabbing the IF_ADDR_LOCK() in many traversals that
171 * do calls into the network stack, thus avoiding LORs.
175 * Each softc has a lock sc_mtx. It is used to synchronise carp_input_c(),
176 * callout-driven events and ioctl()s.
178 * To traverse the list of softcs on an ifnet we use CIF_LOCK() or carp_sx.
179 * To traverse the global list we use the mutex carp_mtx.
181 * Known issues with locking:
183 * - Sending ad, we put the pointer to the softc in an mtag, and no reference
184 * counting is done on the softc.
185 * - On module unload we may race (?) with packet processing thread
186 * dereferencing our function pointers.
189 /* Accept incoming CARP packets. */
190 static VNET_DEFINE(int, carp_allow) = 1;
191 #define V_carp_allow VNET(carp_allow)
193 /* Preempt slower nodes. */
194 static VNET_DEFINE(int, carp_preempt) = 0;
195 #define V_carp_preempt VNET(carp_preempt)
198 static VNET_DEFINE(int, carp_log) = 1;
199 #define V_carp_log VNET(carp_log)
201 /* Global advskew demotion. */
202 static VNET_DEFINE(int, carp_demotion) = 0;
203 #define V_carp_demotion VNET(carp_demotion)
205 /* Send error demotion factor. */
206 static VNET_DEFINE(int, carp_senderr_adj) = CARP_MAXSKEW;
207 #define V_carp_senderr_adj VNET(carp_senderr_adj)
209 /* Iface down demotion factor. */
210 static VNET_DEFINE(int, carp_ifdown_adj) = CARP_MAXSKEW;
211 #define V_carp_ifdown_adj VNET(carp_ifdown_adj)
213 static int carp_allow_sysctl(SYSCTL_HANDLER_ARGS);
214 static int carp_demote_adj_sysctl(SYSCTL_HANDLER_ARGS);
216 SYSCTL_NODE(_net_inet, IPPROTO_CARP, carp, CTLFLAG_RW, 0, "CARP");
217 SYSCTL_PROC(_net_inet_carp, OID_AUTO, allow,
218 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW, 0, 0, carp_allow_sysctl, "I",
219 "Accept incoming CARP packets");
220 SYSCTL_INT(_net_inet_carp, OID_AUTO, preempt, CTLFLAG_VNET | CTLFLAG_RW,
221 &VNET_NAME(carp_preempt), 0, "High-priority backup preemption mode");
222 SYSCTL_INT(_net_inet_carp, OID_AUTO, log, CTLFLAG_VNET | CTLFLAG_RW,
223 &VNET_NAME(carp_log), 0, "CARP log level");
224 SYSCTL_PROC(_net_inet_carp, OID_AUTO, demotion,
225 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW,
226 0, 0, carp_demote_adj_sysctl, "I",
227 "Adjust demotion factor (skew of advskew)");
228 SYSCTL_INT(_net_inet_carp, OID_AUTO, senderr_demotion_factor,
229 CTLFLAG_VNET | CTLFLAG_RW,
230 &VNET_NAME(carp_senderr_adj), 0, "Send error demotion factor adjustment");
231 SYSCTL_INT(_net_inet_carp, OID_AUTO, ifdown_demotion_factor,
232 CTLFLAG_VNET | CTLFLAG_RW,
233 &VNET_NAME(carp_ifdown_adj), 0,
234 "Interface down demotion factor adjustment");
236 VNET_PCPUSTAT_DEFINE(struct carpstats, carpstats);
237 VNET_PCPUSTAT_SYSINIT(carpstats);
238 VNET_PCPUSTAT_SYSUNINIT(carpstats);
240 #define CARPSTATS_ADD(name, val) \
241 counter_u64_add(VNET(carpstats)[offsetof(struct carpstats, name) / \
242 sizeof(uint64_t)], (val))
243 #define CARPSTATS_INC(name) CARPSTATS_ADD(name, 1)
245 SYSCTL_VNET_PCPUSTAT(_net_inet_carp, OID_AUTO, stats, struct carpstats,
246 carpstats, "CARP statistics (struct carpstats, netinet/ip_carp.h)");
248 #define CARP_LOCK_INIT(sc) mtx_init(&(sc)->sc_mtx, "carp_softc", \
250 #define CARP_LOCK_DESTROY(sc) mtx_destroy(&(sc)->sc_mtx)
251 #define CARP_LOCK_ASSERT(sc) mtx_assert(&(sc)->sc_mtx, MA_OWNED)
252 #define CARP_LOCK(sc) mtx_lock(&(sc)->sc_mtx)
253 #define CARP_UNLOCK(sc) mtx_unlock(&(sc)->sc_mtx)
254 #define CIF_LOCK_INIT(cif) mtx_init(&(cif)->cif_mtx, "carp_if", \
256 #define CIF_LOCK_DESTROY(cif) mtx_destroy(&(cif)->cif_mtx)
257 #define CIF_LOCK_ASSERT(cif) mtx_assert(&(cif)->cif_mtx, MA_OWNED)
258 #define CIF_LOCK(cif) mtx_lock(&(cif)->cif_mtx)
259 #define CIF_UNLOCK(cif) mtx_unlock(&(cif)->cif_mtx)
260 #define CIF_FREE(cif) do { \
262 if (TAILQ_EMPTY(&(cif)->cif_vrs)) \
268 #define CARP_LOG(...) do { \
269 if (V_carp_log > 0) \
270 log(LOG_INFO, "carp: " __VA_ARGS__); \
273 #define CARP_DEBUG(...) do { \
274 if (V_carp_log > 1) \
275 log(LOG_DEBUG, __VA_ARGS__); \
278 #define IFNET_FOREACH_IFA(ifp, ifa) \
279 CK_STAILQ_FOREACH((ifa), &(ifp)->if_addrhead, ifa_link) \
280 if ((ifa)->ifa_carp != NULL)
282 #define CARP_FOREACH_IFA(sc, ifa) \
283 CARP_LOCK_ASSERT(sc); \
285 _i < (sc)->sc_naddrs + (sc)->sc_naddrs6 && \
286 ((ifa) = sc->sc_ifas[_i]) != NULL; \
289 #define IFNET_FOREACH_CARP(ifp, sc) \
290 KASSERT(mtx_owned(&ifp->if_carp->cif_mtx) || \
291 sx_xlocked(&carp_sx), ("cif_vrs not locked")); \
292 TAILQ_FOREACH((sc), &(ifp)->if_carp->cif_vrs, sc_list)
294 #define DEMOTE_ADVSKEW(sc) \
295 (((sc)->sc_advskew + V_carp_demotion > CARP_MAXSKEW) ? \
296 CARP_MAXSKEW : ((sc)->sc_advskew + V_carp_demotion))
298 static void carp_input_c(struct mbuf *, struct carp_header *, sa_family_t);
299 static struct carp_softc
300 *carp_alloc(struct ifnet *);
301 static void carp_destroy(struct carp_softc *);
302 static struct carp_if
303 *carp_alloc_if(struct ifnet *);
304 static void carp_free_if(struct carp_if *);
305 static void carp_set_state(struct carp_softc *, int, const char* reason);
306 static void carp_sc_state(struct carp_softc *);
307 static void carp_setrun(struct carp_softc *, sa_family_t);
308 static void carp_master_down(void *);
309 static void carp_master_down_locked(struct carp_softc *,
311 static void carp_send_ad(void *);
312 static void carp_send_ad_locked(struct carp_softc *);
313 static void carp_addroute(struct carp_softc *);
314 static void carp_ifa_addroute(struct ifaddr *);
315 static void carp_delroute(struct carp_softc *);
316 static void carp_ifa_delroute(struct ifaddr *);
317 static void carp_send_ad_all(void *, int);
318 static void carp_demote_adj(int, char *);
320 static LIST_HEAD(, carp_softc) carp_list;
321 static struct mtx carp_mtx;
322 static struct sx carp_sx;
323 static struct task carp_sendall_task =
324 TASK_INITIALIZER(0, carp_send_ad_all, NULL);
327 carp_hmac_prepare(struct carp_softc *sc)
329 uint8_t version = CARP_VERSION, type = CARP_ADVERTISEMENT;
330 uint8_t vhid = sc->sc_vhid & 0xff;
334 struct in_addr last, cur, in;
337 struct in6_addr last6, cur6, in6;
340 CARP_LOCK_ASSERT(sc);
342 /* Compute ipad from key. */
343 bzero(sc->sc_pad, sizeof(sc->sc_pad));
344 bcopy(sc->sc_key, sc->sc_pad, sizeof(sc->sc_key));
345 for (i = 0; i < sizeof(sc->sc_pad); i++)
346 sc->sc_pad[i] ^= 0x36;
348 /* Precompute first part of inner hash. */
349 SHA1Init(&sc->sc_sha1);
350 SHA1Update(&sc->sc_sha1, sc->sc_pad, sizeof(sc->sc_pad));
351 SHA1Update(&sc->sc_sha1, (void *)&version, sizeof(version));
352 SHA1Update(&sc->sc_sha1, (void *)&type, sizeof(type));
353 SHA1Update(&sc->sc_sha1, (void *)&vhid, sizeof(vhid));
359 cur.s_addr = 0xffffffff;
360 CARP_FOREACH_IFA(sc, ifa) {
361 in.s_addr = ifatoia(ifa)->ia_addr.sin_addr.s_addr;
362 if (ifa->ifa_addr->sa_family == AF_INET &&
363 ntohl(in.s_addr) > ntohl(last.s_addr) &&
364 ntohl(in.s_addr) < ntohl(cur.s_addr)) {
365 cur.s_addr = in.s_addr;
370 SHA1Update(&sc->sc_sha1, (void *)&cur, sizeof(cur));
374 memset(&cur6, 0, sizeof(cur6));
378 memset(&cur6, 0xff, sizeof(cur6));
379 CARP_FOREACH_IFA(sc, ifa) {
380 in6 = ifatoia6(ifa)->ia_addr.sin6_addr;
381 if (IN6_IS_SCOPE_EMBED(&in6))
382 in6.s6_addr16[1] = 0;
383 if (ifa->ifa_addr->sa_family == AF_INET6 &&
384 memcmp(&in6, &last6, sizeof(in6)) > 0 &&
385 memcmp(&in6, &cur6, sizeof(in6)) < 0) {
391 SHA1Update(&sc->sc_sha1, (void *)&cur6, sizeof(cur6));
395 /* convert ipad to opad */
396 for (i = 0; i < sizeof(sc->sc_pad); i++)
397 sc->sc_pad[i] ^= 0x36 ^ 0x5c;
401 carp_hmac_generate(struct carp_softc *sc, uint32_t counter[2],
402 unsigned char md[20])
406 CARP_LOCK_ASSERT(sc);
408 /* fetch first half of inner hash */
409 bcopy(&sc->sc_sha1, &sha1ctx, sizeof(sha1ctx));
411 SHA1Update(&sha1ctx, (void *)counter, sizeof(sc->sc_counter));
412 SHA1Final(md, &sha1ctx);
416 SHA1Update(&sha1ctx, sc->sc_pad, sizeof(sc->sc_pad));
417 SHA1Update(&sha1ctx, md, 20);
418 SHA1Final(md, &sha1ctx);
422 carp_hmac_verify(struct carp_softc *sc, uint32_t counter[2],
423 unsigned char md[20])
425 unsigned char md2[20];
427 CARP_LOCK_ASSERT(sc);
429 carp_hmac_generate(sc, counter, md2);
431 return (bcmp(md, md2, sizeof(md2)));
435 * process input packet.
436 * we have rearranged checks order compared to the rfc,
437 * but it seems more efficient this way or not possible otherwise.
441 carp_input(struct mbuf **mp, int *offp, int proto)
443 struct mbuf *m = *mp;
444 struct ip *ip = mtod(m, struct ip *);
445 struct carp_header *ch;
451 CARPSTATS_INC(carps_ipackets);
455 return (IPPROTO_DONE);
458 /* verify that the IP TTL is 255. */
459 if (ip->ip_ttl != CARP_DFLTTL) {
460 CARPSTATS_INC(carps_badttl);
461 CARP_DEBUG("%s: received ttl %d != 255 on %s\n", __func__,
463 m->m_pkthdr.rcvif->if_xname);
465 return (IPPROTO_DONE);
468 iplen = ip->ip_hl << 2;
470 if (m->m_pkthdr.len < iplen + sizeof(*ch)) {
471 CARPSTATS_INC(carps_badlen);
472 CARP_DEBUG("%s: received len %zd < sizeof(struct carp_header) "
473 "on %s\n", __func__, m->m_len - sizeof(struct ip),
474 m->m_pkthdr.rcvif->if_xname);
476 return (IPPROTO_DONE);
479 if (iplen + sizeof(*ch) < m->m_len) {
480 if ((m = m_pullup(m, iplen + sizeof(*ch))) == NULL) {
481 CARPSTATS_INC(carps_hdrops);
482 CARP_DEBUG("%s: pullup failed\n", __func__);
483 return (IPPROTO_DONE);
485 ip = mtod(m, struct ip *);
487 ch = (struct carp_header *)((char *)ip + iplen);
490 * verify that the received packet length is
491 * equal to the CARP header
493 len = iplen + sizeof(*ch);
494 if (len > m->m_pkthdr.len) {
495 CARPSTATS_INC(carps_badlen);
496 CARP_DEBUG("%s: packet too short %d on %s\n", __func__,
498 m->m_pkthdr.rcvif->if_xname);
500 return (IPPROTO_DONE);
503 if ((m = m_pullup(m, len)) == NULL) {
504 CARPSTATS_INC(carps_hdrops);
505 return (IPPROTO_DONE);
507 ip = mtod(m, struct ip *);
508 ch = (struct carp_header *)((char *)ip + iplen);
510 /* verify the CARP checksum */
512 if (in_cksum(m, len - iplen)) {
513 CARPSTATS_INC(carps_badsum);
514 CARP_DEBUG("%s: checksum failed on %s\n", __func__,
515 m->m_pkthdr.rcvif->if_xname);
517 return (IPPROTO_DONE);
521 carp_input_c(m, ch, AF_INET);
522 return (IPPROTO_DONE);
528 carp6_input(struct mbuf **mp, int *offp, int proto)
530 struct mbuf *m = *mp;
531 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
532 struct carp_header *ch;
535 CARPSTATS_INC(carps_ipackets6);
539 return (IPPROTO_DONE);
542 /* check if received on a valid carp interface */
543 if (m->m_pkthdr.rcvif->if_carp == NULL) {
544 CARPSTATS_INC(carps_badif);
545 CARP_DEBUG("%s: packet received on non-carp interface: %s\n",
546 __func__, m->m_pkthdr.rcvif->if_xname);
548 return (IPPROTO_DONE);
551 /* verify that the IP TTL is 255 */
552 if (ip6->ip6_hlim != CARP_DFLTTL) {
553 CARPSTATS_INC(carps_badttl);
554 CARP_DEBUG("%s: received ttl %d != 255 on %s\n", __func__,
555 ip6->ip6_hlim, m->m_pkthdr.rcvif->if_xname);
557 return (IPPROTO_DONE);
560 /* verify that we have a complete carp packet */
562 IP6_EXTHDR_GET(ch, struct carp_header *, m, *offp, sizeof(*ch));
564 CARPSTATS_INC(carps_badlen);
565 CARP_DEBUG("%s: packet size %u too small\n", __func__, len);
566 return (IPPROTO_DONE);
570 /* verify the CARP checksum */
572 if (in_cksum(m, sizeof(*ch))) {
573 CARPSTATS_INC(carps_badsum);
574 CARP_DEBUG("%s: checksum failed, on %s\n", __func__,
575 m->m_pkthdr.rcvif->if_xname);
577 return (IPPROTO_DONE);
581 carp_input_c(m, ch, AF_INET6);
582 return (IPPROTO_DONE);
587 * This routine should not be necessary at all, but some switches
588 * (VMWare ESX vswitches) can echo our own packets back at us,
589 * and we must ignore them or they will cause us to drop out of
592 * We cannot catch all cases of network loops. Instead, what we
593 * do here is catch any packet that arrives with a carp header
594 * with a VHID of 0, that comes from an address that is our own.
595 * These packets are by definition "from us" (even if they are from
596 * a misconfigured host that is pretending to be us).
598 * The VHID test is outside this mini-function.
601 carp_source_is_self(struct mbuf *m, struct ifaddr *ifa, sa_family_t af)
615 ip4 = mtod(m, struct ip *);
616 in4 = ifatoia(ifa)->ia_addr.sin_addr;
617 return (in4.s_addr == ip4->ip_src.s_addr);
621 ip6 = mtod(m, struct ip6_hdr *);
622 in6 = ifatoia6(ifa)->ia_addr.sin6_addr;
623 return (memcmp(&in6, &ip6->ip6_src, sizeof(in6)) == 0);
632 carp_input_c(struct mbuf *m, struct carp_header *ch, sa_family_t af)
634 struct ifnet *ifp = m->m_pkthdr.rcvif;
635 struct ifaddr *ifa, *match;
636 struct carp_softc *sc;
637 uint64_t tmp_counter;
638 struct timeval sc_tv, ch_tv;
642 * Verify that the VHID is valid on the receiving interface.
644 * There should be just one match. If there are none
645 * the VHID is not valid and we drop the packet. If
646 * there are multiple VHID matches, take just the first
647 * one, for compatibility with previous code. While we're
648 * scanning, check for obvious loops in the network topology
649 * (these should never happen, and as noted above, we may
650 * miss real loops; this is just a double-check).
655 IFNET_FOREACH_IFA(ifp, ifa) {
656 if (match == NULL && ifa->ifa_carp != NULL &&
657 ifa->ifa_addr->sa_family == af &&
658 ifa->ifa_carp->sc_vhid == ch->carp_vhid)
660 if (ch->carp_vhid == 0 && carp_source_is_self(m, ifa, af))
663 ifa = error ? NULL : match;
666 IF_ADDR_RUNLOCK(ifp);
669 if (error == ELOOP) {
670 CARP_DEBUG("dropping looped packet on interface %s\n",
672 CARPSTATS_INC(carps_badif); /* ??? */
674 CARPSTATS_INC(carps_badvhid);
680 /* verify the CARP version. */
681 if (ch->carp_version != CARP_VERSION) {
682 CARPSTATS_INC(carps_badver);
683 CARP_DEBUG("%s: invalid version %d\n", ifp->if_xname,
694 if (carp_hmac_verify(sc, ch->carp_counter, ch->carp_md)) {
695 CARPSTATS_INC(carps_badauth);
696 CARP_DEBUG("%s: incorrect hash for VHID %u@%s\n", __func__,
697 sc->sc_vhid, ifp->if_xname);
701 tmp_counter = ntohl(ch->carp_counter[0]);
702 tmp_counter = tmp_counter<<32;
703 tmp_counter += ntohl(ch->carp_counter[1]);
705 /* XXX Replay protection goes here */
707 sc->sc_init_counter = 0;
708 sc->sc_counter = tmp_counter;
710 sc_tv.tv_sec = sc->sc_advbase;
711 sc_tv.tv_usec = DEMOTE_ADVSKEW(sc) * 1000000 / 256;
712 ch_tv.tv_sec = ch->carp_advbase;
713 ch_tv.tv_usec = ch->carp_advskew * 1000000 / 256;
715 switch (sc->sc_state) {
720 * If we receive an advertisement from a master who's going to
721 * be more frequent than us, go into BACKUP state.
723 if (timevalcmp(&sc_tv, &ch_tv, >) ||
724 timevalcmp(&sc_tv, &ch_tv, ==)) {
725 callout_stop(&sc->sc_ad_tmo);
726 carp_set_state(sc, BACKUP,
727 "more frequent advertisement received");
734 * If we're pre-empting masters who advertise slower than us,
735 * and this one claims to be slower, treat him as down.
737 if (V_carp_preempt && timevalcmp(&sc_tv, &ch_tv, <)) {
738 carp_master_down_locked(sc,
739 "preempting a slower master");
744 * If the master is going to advertise at such a low frequency
745 * that he's guaranteed to time out, we'd might as well just
746 * treat him as timed out now.
748 sc_tv.tv_sec = sc->sc_advbase * 3;
749 if (timevalcmp(&sc_tv, &ch_tv, <)) {
750 carp_master_down_locked(sc, "master will time out");
755 * Otherwise, we reset the counter and wait for the next
768 carp_prepare_ad(struct mbuf *m, struct carp_softc *sc, struct carp_header *ch)
772 if (sc->sc_init_counter) {
773 /* this could also be seconds since unix epoch */
774 sc->sc_counter = arc4random();
775 sc->sc_counter = sc->sc_counter << 32;
776 sc->sc_counter += arc4random();
780 ch->carp_counter[0] = htonl((sc->sc_counter>>32)&0xffffffff);
781 ch->carp_counter[1] = htonl(sc->sc_counter&0xffffffff);
783 carp_hmac_generate(sc, ch->carp_counter, ch->carp_md);
785 /* Tag packet for carp_output */
786 if ((mtag = m_tag_get(PACKET_TAG_CARP, sizeof(struct carp_softc *),
787 M_NOWAIT)) == NULL) {
789 CARPSTATS_INC(carps_onomem);
792 bcopy(&sc, mtag + 1, sizeof(sc));
793 m_tag_prepend(m, mtag);
799 * To avoid LORs and possible recursions this function shouldn't
800 * be called directly, but scheduled via taskqueue.
803 carp_send_ad_all(void *ctx __unused, int pending __unused)
805 struct carp_softc *sc;
808 LIST_FOREACH(sc, &carp_list, sc_next)
809 if (sc->sc_state == MASTER) {
811 CURVNET_SET(sc->sc_carpdev->if_vnet);
812 carp_send_ad_locked(sc);
816 mtx_unlock(&carp_mtx);
819 /* Send a periodic advertisement, executed in callout context. */
821 carp_send_ad(void *v)
823 struct carp_softc *sc = v;
825 CARP_LOCK_ASSERT(sc);
826 CURVNET_SET(sc->sc_carpdev->if_vnet);
827 carp_send_ad_locked(sc);
833 carp_send_ad_error(struct carp_softc *sc, int error)
837 if (sc->sc_sendad_errors < INT_MAX)
838 sc->sc_sendad_errors++;
839 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) {
840 static const char fmt[] = "send error %d on %s";
841 char msg[sizeof(fmt) + IFNAMSIZ];
843 sprintf(msg, fmt, error, sc->sc_carpdev->if_xname);
844 carp_demote_adj(V_carp_senderr_adj, msg);
846 sc->sc_sendad_success = 0;
848 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS &&
849 ++sc->sc_sendad_success >= CARP_SENDAD_MIN_SUCCESS) {
850 static const char fmt[] = "send ok on %s";
851 char msg[sizeof(fmt) + IFNAMSIZ];
853 sprintf(msg, fmt, sc->sc_carpdev->if_xname);
854 carp_demote_adj(-V_carp_senderr_adj, msg);
855 sc->sc_sendad_errors = 0;
857 sc->sc_sendad_errors = 0;
862 * Pick the best ifaddr on the given ifp for sending CARP
865 * "Best" here is defined by ifa_preferred(). This function is much
866 * much like ifaof_ifpforaddr() except that we just use ifa_preferred().
868 * (This could be simplified to return the actual address, except that
869 * it has a different format in AF_INET and AF_INET6.)
871 static struct ifaddr *
872 carp_best_ifa(int af, struct ifnet *ifp)
874 struct ifaddr *ifa, *best;
880 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
881 if (ifa->ifa_addr->sa_family == af &&
882 (best == NULL || ifa_preferred(best, ifa)))
885 IF_ADDR_RUNLOCK(ifp);
892 carp_send_ad_locked(struct carp_softc *sc)
894 struct carp_header ch;
897 struct carp_header *ch_ptr;
901 CARP_LOCK_ASSERT(sc);
903 advskew = DEMOTE_ADVSKEW(sc);
904 tv.tv_sec = sc->sc_advbase;
905 tv.tv_usec = advskew * 1000000 / 256;
907 ch.carp_version = CARP_VERSION;
908 ch.carp_type = CARP_ADVERTISEMENT;
909 ch.carp_vhid = sc->sc_vhid;
910 ch.carp_advbase = sc->sc_advbase;
911 ch.carp_advskew = advskew;
912 ch.carp_authlen = 7; /* XXX DEFINE */
913 ch.carp_pad1 = 0; /* must be zero */
916 /* XXXGL: OpenBSD picks first ifaddr with needed family. */
922 m = m_gethdr(M_NOWAIT, MT_DATA);
924 CARPSTATS_INC(carps_onomem);
927 len = sizeof(*ip) + sizeof(ch);
928 m->m_pkthdr.len = len;
929 m->m_pkthdr.rcvif = NULL;
931 M_ALIGN(m, m->m_len);
932 m->m_flags |= M_MCAST;
933 ip = mtod(m, struct ip *);
934 ip->ip_v = IPVERSION;
935 ip->ip_hl = sizeof(*ip) >> 2;
936 ip->ip_tos = IPTOS_LOWDELAY;
937 ip->ip_len = htons(len);
938 ip->ip_off = htons(IP_DF);
939 ip->ip_ttl = CARP_DFLTTL;
940 ip->ip_p = IPPROTO_CARP;
944 ifa = carp_best_ifa(AF_INET, sc->sc_carpdev);
947 ifatoia(ifa)->ia_addr.sin_addr.s_addr;
950 ip->ip_src.s_addr = 0;
951 ip->ip_dst.s_addr = htonl(INADDR_CARP_GROUP);
953 ch_ptr = (struct carp_header *)(&ip[1]);
954 bcopy(&ch, ch_ptr, sizeof(ch));
955 if (carp_prepare_ad(m, sc, ch_ptr))
958 m->m_data += sizeof(*ip);
959 ch_ptr->carp_cksum = in_cksum(m, len - sizeof(*ip));
960 m->m_data -= sizeof(*ip);
962 CARPSTATS_INC(carps_opackets);
964 carp_send_ad_error(sc, ip_output(m, NULL, NULL, IP_RAWOUTPUT,
965 &sc->sc_carpdev->if_carp->cif_imo, NULL));
969 if (sc->sc_naddrs6) {
972 m = m_gethdr(M_NOWAIT, MT_DATA);
974 CARPSTATS_INC(carps_onomem);
977 len = sizeof(*ip6) + sizeof(ch);
978 m->m_pkthdr.len = len;
979 m->m_pkthdr.rcvif = NULL;
981 M_ALIGN(m, m->m_len);
982 m->m_flags |= M_MCAST;
983 ip6 = mtod(m, struct ip6_hdr *);
984 bzero(ip6, sizeof(*ip6));
985 ip6->ip6_vfc |= IPV6_VERSION;
986 ip6->ip6_hlim = CARP_DFLTTL;
987 ip6->ip6_nxt = IPPROTO_CARP;
989 /* set the source address */
990 ifa = carp_best_ifa(AF_INET6, sc->sc_carpdev);
992 bcopy(IFA_IN6(ifa), &ip6->ip6_src,
993 sizeof(struct in6_addr));
996 /* This should never happen with IPv6. */
997 bzero(&ip6->ip6_src, sizeof(struct in6_addr));
999 /* Set the multicast destination. */
1000 ip6->ip6_dst.s6_addr16[0] = htons(0xff02);
1001 ip6->ip6_dst.s6_addr8[15] = 0x12;
1002 if (in6_setscope(&ip6->ip6_dst, sc->sc_carpdev, NULL) != 0) {
1004 CARP_DEBUG("%s: in6_setscope failed\n", __func__);
1008 ch_ptr = (struct carp_header *)(&ip6[1]);
1009 bcopy(&ch, ch_ptr, sizeof(ch));
1010 if (carp_prepare_ad(m, sc, ch_ptr))
1013 m->m_data += sizeof(*ip6);
1014 ch_ptr->carp_cksum = in_cksum(m, len - sizeof(*ip6));
1015 m->m_data -= sizeof(*ip6);
1017 CARPSTATS_INC(carps_opackets6);
1019 carp_send_ad_error(sc, ip6_output(m, NULL, NULL, 0,
1020 &sc->sc_carpdev->if_carp->cif_im6o, NULL, NULL));
1025 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv), carp_send_ad, sc);
1029 carp_addroute(struct carp_softc *sc)
1033 CARP_FOREACH_IFA(sc, ifa)
1034 carp_ifa_addroute(ifa);
1038 carp_ifa_addroute(struct ifaddr *ifa)
1041 switch (ifa->ifa_addr->sa_family) {
1044 in_addprefix(ifatoia(ifa), RTF_UP);
1045 ifa_add_loopback_route(ifa,
1046 (struct sockaddr *)&ifatoia(ifa)->ia_addr);
1051 ifa_add_loopback_route(ifa,
1052 (struct sockaddr *)&ifatoia6(ifa)->ia_addr);
1053 nd6_add_ifa_lle(ifatoia6(ifa));
1060 carp_delroute(struct carp_softc *sc)
1064 CARP_FOREACH_IFA(sc, ifa)
1065 carp_ifa_delroute(ifa);
1069 carp_ifa_delroute(struct ifaddr *ifa)
1072 switch (ifa->ifa_addr->sa_family) {
1075 ifa_del_loopback_route(ifa,
1076 (struct sockaddr *)&ifatoia(ifa)->ia_addr);
1077 in_scrubprefix(ifatoia(ifa), LLE_STATIC);
1082 ifa_del_loopback_route(ifa,
1083 (struct sockaddr *)&ifatoia6(ifa)->ia_addr);
1084 nd6_rem_ifa_lle(ifatoia6(ifa), 1);
1091 carp_master(struct ifaddr *ifa)
1093 struct carp_softc *sc = ifa->ifa_carp;
1095 return (sc->sc_state == MASTER);
1100 * Broadcast a gratuitous ARP request containing
1101 * the virtual router MAC address for each IP address
1102 * associated with the virtual router.
1105 carp_send_arp(struct carp_softc *sc)
1108 struct in_addr addr;
1110 CARP_FOREACH_IFA(sc, ifa) {
1111 if (ifa->ifa_addr->sa_family != AF_INET)
1113 addr = ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr;
1114 arp_announce_ifaddr(sc->sc_carpdev, addr, LLADDR(&sc->sc_addr));
1119 carp_iamatch(struct ifaddr *ifa, uint8_t **enaddr)
1121 struct carp_softc *sc = ifa->ifa_carp;
1123 if (sc->sc_state == MASTER) {
1124 *enaddr = LLADDR(&sc->sc_addr);
1134 carp_send_na(struct carp_softc *sc)
1136 static struct in6_addr mcast = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
1138 struct in6_addr *in6;
1140 CARP_FOREACH_IFA(sc, ifa) {
1141 if (ifa->ifa_addr->sa_family != AF_INET6)
1145 nd6_na_output(sc->sc_carpdev, &mcast, in6,
1146 ND_NA_FLAG_OVERRIDE, 1, NULL);
1147 DELAY(1000); /* XXX */
1152 * Returns ifa in case it's a carp address and it is MASTER, or if the address
1153 * matches and is not a carp address. Returns NULL otherwise.
1156 carp_iamatch6(struct ifnet *ifp, struct in6_addr *taddr)
1162 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1163 if (ifa->ifa_addr->sa_family != AF_INET6)
1165 if (!IN6_ARE_ADDR_EQUAL(taddr, IFA_IN6(ifa)))
1167 if (ifa->ifa_carp && ifa->ifa_carp->sc_state != MASTER)
1173 IF_ADDR_RUNLOCK(ifp);
1179 carp_macmatch6(struct ifnet *ifp, struct mbuf *m, const struct in6_addr *taddr)
1184 IFNET_FOREACH_IFA(ifp, ifa)
1185 if (ifa->ifa_addr->sa_family == AF_INET6 &&
1186 IN6_ARE_ADDR_EQUAL(taddr, IFA_IN6(ifa))) {
1187 struct carp_softc *sc = ifa->ifa_carp;
1190 IF_ADDR_RUNLOCK(ifp);
1192 mtag = m_tag_get(PACKET_TAG_CARP,
1193 sizeof(struct carp_softc *), M_NOWAIT);
1195 /* Better a bit than nothing. */
1196 return (LLADDR(&sc->sc_addr));
1198 bcopy(&sc, mtag + 1, sizeof(sc));
1199 m_tag_prepend(m, mtag);
1201 return (LLADDR(&sc->sc_addr));
1203 IF_ADDR_RUNLOCK(ifp);
1210 carp_forus(struct ifnet *ifp, u_char *dhost)
1212 struct carp_softc *sc;
1213 uint8_t *ena = dhost;
1215 if (ena[0] || ena[1] || ena[2] != 0x5e || ena[3] || ena[4] != 1)
1218 CIF_LOCK(ifp->if_carp);
1219 IFNET_FOREACH_CARP(ifp, sc) {
1221 if (sc->sc_state == MASTER && !bcmp(dhost, LLADDR(&sc->sc_addr),
1224 CIF_UNLOCK(ifp->if_carp);
1229 CIF_UNLOCK(ifp->if_carp);
1234 /* Master down timeout event, executed in callout context. */
1236 carp_master_down(void *v)
1238 struct carp_softc *sc = v;
1240 CARP_LOCK_ASSERT(sc);
1242 CURVNET_SET(sc->sc_carpdev->if_vnet);
1243 if (sc->sc_state == BACKUP) {
1244 carp_master_down_locked(sc, "master timed out");
1252 carp_master_down_locked(struct carp_softc *sc, const char *reason)
1255 CARP_LOCK_ASSERT(sc);
1257 switch (sc->sc_state) {
1259 carp_set_state(sc, MASTER, reason);
1260 carp_send_ad_locked(sc);
1273 panic("carp: VHID %u@%s: master_down event in %s state\n",
1275 sc->sc_carpdev->if_xname,
1276 sc->sc_state ? "MASTER" : "INIT");
1283 * When in backup state, af indicates whether to reset the master down timer
1284 * for v4 or v6. If it's set to zero, reset the ones which are already pending.
1287 carp_setrun(struct carp_softc *sc, sa_family_t af)
1291 CARP_LOCK_ASSERT(sc);
1293 if ((sc->sc_carpdev->if_flags & IFF_UP) == 0 ||
1294 sc->sc_carpdev->if_link_state != LINK_STATE_UP ||
1295 (sc->sc_naddrs == 0 && sc->sc_naddrs6 == 0) ||
1299 switch (sc->sc_state) {
1301 carp_set_state(sc, BACKUP, "initialization complete");
1305 callout_stop(&sc->sc_ad_tmo);
1306 tv.tv_sec = 3 * sc->sc_advbase;
1307 tv.tv_usec = sc->sc_advskew * 1000000 / 256;
1311 callout_reset(&sc->sc_md_tmo, tvtohz(&tv),
1312 carp_master_down, sc);
1317 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv),
1318 carp_master_down, sc);
1324 callout_reset(&sc->sc_md_tmo, tvtohz(&tv),
1325 carp_master_down, sc);
1329 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv),
1330 carp_master_down, sc);
1336 tv.tv_sec = sc->sc_advbase;
1337 tv.tv_usec = sc->sc_advskew * 1000000 / 256;
1338 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv),
1345 * Setup multicast structures.
1348 carp_multicast_setup(struct carp_if *cif, sa_family_t sa)
1350 struct ifnet *ifp = cif->cif_ifp;
1357 struct ip_moptions *imo = &cif->cif_imo;
1358 struct in_addr addr;
1360 if (imo->imo_membership)
1363 imo->imo_membership = (struct in_multi **)malloc(
1364 (sizeof(struct in_multi *) * IP_MIN_MEMBERSHIPS), M_CARP,
1366 imo->imo_mfilters = NULL;
1367 imo->imo_max_memberships = IP_MIN_MEMBERSHIPS;
1368 imo->imo_multicast_vif = -1;
1370 addr.s_addr = htonl(INADDR_CARP_GROUP);
1371 if ((error = in_joingroup(ifp, &addr, NULL,
1372 &imo->imo_membership[0])) != 0) {
1373 free(imo->imo_membership, M_CARP);
1376 imo->imo_num_memberships++;
1377 imo->imo_multicast_ifp = ifp;
1378 imo->imo_multicast_ttl = CARP_DFLTTL;
1379 imo->imo_multicast_loop = 0;
1386 struct ip6_moptions *im6o = &cif->cif_im6o;
1387 struct in6_addr in6;
1388 struct in6_multi *in6m;
1390 if (im6o->im6o_membership)
1393 im6o->im6o_membership = (struct in6_multi **)malloc(
1394 (sizeof(struct in6_multi *) * IPV6_MIN_MEMBERSHIPS), M_CARP,
1396 im6o->im6o_mfilters = NULL;
1397 im6o->im6o_max_memberships = IPV6_MIN_MEMBERSHIPS;
1398 im6o->im6o_multicast_hlim = CARP_DFLTTL;
1399 im6o->im6o_multicast_ifp = ifp;
1401 /* Join IPv6 CARP multicast group. */
1402 bzero(&in6, sizeof(in6));
1403 in6.s6_addr16[0] = htons(0xff02);
1404 in6.s6_addr8[15] = 0x12;
1405 if ((error = in6_setscope(&in6, ifp, NULL)) != 0) {
1406 free(im6o->im6o_membership, M_CARP);
1410 if ((error = in6_joingroup(ifp, &in6, NULL, &in6m, 0)) != 0) {
1411 free(im6o->im6o_membership, M_CARP);
1414 im6o->im6o_membership[0] = in6m;
1415 im6o->im6o_num_memberships++;
1417 /* Join solicited multicast address. */
1418 bzero(&in6, sizeof(in6));
1419 in6.s6_addr16[0] = htons(0xff02);
1420 in6.s6_addr32[1] = 0;
1421 in6.s6_addr32[2] = htonl(1);
1422 in6.s6_addr32[3] = 0;
1423 in6.s6_addr8[12] = 0xff;
1424 if ((error = in6_setscope(&in6, ifp, NULL)) != 0) {
1425 in6_leavegroup(im6o->im6o_membership[0], NULL);
1426 free(im6o->im6o_membership, M_CARP);
1430 if ((error = in6_joingroup(ifp, &in6, NULL, &in6m, 0)) != 0) {
1431 in6_leavegroup(im6o->im6o_membership[0], NULL);
1432 free(im6o->im6o_membership, M_CARP);
1435 im6o->im6o_membership[1] = in6m;
1436 im6o->im6o_num_memberships++;
1446 * Free multicast structures.
1449 carp_multicast_cleanup(struct carp_if *cif, sa_family_t sa)
1452 sx_assert(&carp_sx, SA_XLOCKED);
1457 if (cif->cif_naddrs == 0) {
1458 struct ip_moptions *imo = &cif->cif_imo;
1460 in_leavegroup(imo->imo_membership[0], NULL);
1461 KASSERT(imo->imo_mfilters == NULL,
1462 ("%s: imo_mfilters != NULL", __func__));
1463 free(imo->imo_membership, M_CARP);
1464 imo->imo_membership = NULL;
1471 if (cif->cif_naddrs6 == 0) {
1472 struct ip6_moptions *im6o = &cif->cif_im6o;
1474 in6_leavegroup(im6o->im6o_membership[0], NULL);
1475 in6_leavegroup(im6o->im6o_membership[1], NULL);
1476 KASSERT(im6o->im6o_mfilters == NULL,
1477 ("%s: im6o_mfilters != NULL", __func__));
1478 free(im6o->im6o_membership, M_CARP);
1479 im6o->im6o_membership = NULL;
1487 carp_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa)
1490 struct carp_softc *sc;
1495 switch (sa->sa_family) {
1508 mtag = m_tag_find(m, PACKET_TAG_CARP, NULL);
1512 bcopy(mtag + 1, &sc, sizeof(sc));
1514 /* Set the source MAC address to the Virtual Router MAC Address. */
1515 switch (ifp->if_type) {
1519 struct ether_header *eh;
1521 eh = mtod(m, struct ether_header *);
1522 eh->ether_shost[0] = 0;
1523 eh->ether_shost[1] = 0;
1524 eh->ether_shost[2] = 0x5e;
1525 eh->ether_shost[3] = 0;
1526 eh->ether_shost[4] = 1;
1527 eh->ether_shost[5] = sc->sc_vhid;
1531 printf("%s: carp is not supported for the %d interface type\n",
1532 ifp->if_xname, ifp->if_type);
1533 return (EOPNOTSUPP);
1539 static struct carp_softc*
1540 carp_alloc(struct ifnet *ifp)
1542 struct carp_softc *sc;
1543 struct carp_if *cif;
1545 sx_assert(&carp_sx, SA_XLOCKED);
1547 if ((cif = ifp->if_carp) == NULL)
1548 cif = carp_alloc_if(ifp);
1550 sc = malloc(sizeof(*sc), M_CARP, M_WAITOK|M_ZERO);
1552 sc->sc_advbase = CARP_DFLTINTV;
1553 sc->sc_vhid = -1; /* required setting */
1554 sc->sc_init_counter = 1;
1555 sc->sc_state = INIT;
1557 sc->sc_ifasiz = sizeof(struct ifaddr *);
1558 sc->sc_ifas = malloc(sc->sc_ifasiz, M_CARP, M_WAITOK|M_ZERO);
1559 sc->sc_carpdev = ifp;
1563 callout_init_mtx(&sc->sc_md_tmo, &sc->sc_mtx, CALLOUT_RETURNUNLOCKED);
1566 callout_init_mtx(&sc->sc_md6_tmo, &sc->sc_mtx, CALLOUT_RETURNUNLOCKED);
1568 callout_init_mtx(&sc->sc_ad_tmo, &sc->sc_mtx, CALLOUT_RETURNUNLOCKED);
1571 TAILQ_INSERT_TAIL(&cif->cif_vrs, sc, sc_list);
1574 mtx_lock(&carp_mtx);
1575 LIST_INSERT_HEAD(&carp_list, sc, sc_next);
1576 mtx_unlock(&carp_mtx);
1582 carp_grow_ifas(struct carp_softc *sc)
1584 struct ifaddr **new;
1586 new = malloc(sc->sc_ifasiz * 2, M_CARP, M_WAITOK | M_ZERO);
1588 bcopy(sc->sc_ifas, new, sc->sc_ifasiz);
1589 free(sc->sc_ifas, M_CARP);
1596 carp_destroy(struct carp_softc *sc)
1598 struct ifnet *ifp = sc->sc_carpdev;
1599 struct carp_if *cif = ifp->if_carp;
1601 sx_assert(&carp_sx, SA_XLOCKED);
1603 if (sc->sc_suppress)
1604 carp_demote_adj(-V_carp_ifdown_adj, "vhid removed");
1608 TAILQ_REMOVE(&cif->cif_vrs, sc, sc_list);
1611 mtx_lock(&carp_mtx);
1612 LIST_REMOVE(sc, sc_next);
1613 mtx_unlock(&carp_mtx);
1615 callout_drain(&sc->sc_ad_tmo);
1617 callout_drain(&sc->sc_md_tmo);
1620 callout_drain(&sc->sc_md6_tmo);
1622 CARP_LOCK_DESTROY(sc);
1624 free(sc->sc_ifas, M_CARP);
1628 static struct carp_if*
1629 carp_alloc_if(struct ifnet *ifp)
1631 struct carp_if *cif;
1634 cif = malloc(sizeof(*cif), M_CARP, M_WAITOK|M_ZERO);
1636 if ((error = ifpromisc(ifp, 1)) != 0)
1637 printf("%s: ifpromisc(%s) failed: %d\n",
1638 __func__, ifp->if_xname, error);
1640 cif->cif_flags |= CIF_PROMISC;
1644 TAILQ_INIT(&cif->cif_vrs);
1649 IF_ADDR_WUNLOCK(ifp);
1655 carp_free_if(struct carp_if *cif)
1657 struct ifnet *ifp = cif->cif_ifp;
1659 CIF_LOCK_ASSERT(cif);
1660 KASSERT(TAILQ_EMPTY(&cif->cif_vrs), ("%s: softc list not empty",
1664 ifp->if_carp = NULL;
1665 IF_ADDR_WUNLOCK(ifp);
1667 CIF_LOCK_DESTROY(cif);
1669 if (cif->cif_flags & CIF_PROMISC)
1677 carp_carprcp(struct carpreq *carpr, struct carp_softc *sc, int priv)
1681 carpr->carpr_state = sc->sc_state;
1682 carpr->carpr_vhid = sc->sc_vhid;
1683 carpr->carpr_advbase = sc->sc_advbase;
1684 carpr->carpr_advskew = sc->sc_advskew;
1686 bcopy(sc->sc_key, carpr->carpr_key, sizeof(carpr->carpr_key));
1688 bzero(carpr->carpr_key, sizeof(carpr->carpr_key));
1693 carp_ioctl(struct ifreq *ifr, u_long cmd, struct thread *td)
1695 struct carpreq carpr;
1697 struct carp_softc *sc = NULL;
1698 int error = 0, locked = 0;
1700 if ((error = copyin(ifr_data_get_ptr(ifr), &carpr, sizeof carpr)))
1703 ifp = ifunit_ref(ifr->ifr_name);
1707 switch (ifp->if_type) {
1717 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
1718 error = EADDRNOTAVAIL;
1725 if ((error = priv_check(td, PRIV_NETINET_CARP)))
1727 if (carpr.carpr_vhid <= 0 || carpr.carpr_vhid > CARP_MAXVHID ||
1728 carpr.carpr_advbase < 0 || carpr.carpr_advskew < 0) {
1734 IFNET_FOREACH_CARP(ifp, sc)
1735 if (sc->sc_vhid == carpr.carpr_vhid)
1739 sc = carp_alloc(ifp);
1741 sc->sc_vhid = carpr.carpr_vhid;
1742 LLADDR(&sc->sc_addr)[0] = 0;
1743 LLADDR(&sc->sc_addr)[1] = 0;
1744 LLADDR(&sc->sc_addr)[2] = 0x5e;
1745 LLADDR(&sc->sc_addr)[3] = 0;
1746 LLADDR(&sc->sc_addr)[4] = 1;
1747 LLADDR(&sc->sc_addr)[5] = sc->sc_vhid;
1751 if (carpr.carpr_advbase > 0) {
1752 if (carpr.carpr_advbase > 255 ||
1753 carpr.carpr_advbase < CARP_DFLTINTV) {
1757 sc->sc_advbase = carpr.carpr_advbase;
1759 if (carpr.carpr_advskew >= 255) {
1763 sc->sc_advskew = carpr.carpr_advskew;
1764 if (carpr.carpr_key[0] != '\0') {
1765 bcopy(carpr.carpr_key, sc->sc_key, sizeof(sc->sc_key));
1766 carp_hmac_prepare(sc);
1768 if (sc->sc_state != INIT &&
1769 carpr.carpr_state != sc->sc_state) {
1770 switch (carpr.carpr_state) {
1772 callout_stop(&sc->sc_ad_tmo);
1773 carp_set_state(sc, BACKUP,
1774 "user requested via ifconfig");
1779 carp_master_down_locked(sc,
1780 "user requested via ifconfig");
1792 if (carpr.carpr_vhid < 0 || carpr.carpr_vhid > CARP_MAXVHID) {
1796 if (carpr.carpr_count < 1) {
1800 if (ifp->if_carp == NULL) {
1805 priveleged = (priv_check(td, PRIV_NETINET_CARP) == 0);
1806 if (carpr.carpr_vhid != 0) {
1807 IFNET_FOREACH_CARP(ifp, sc)
1808 if (sc->sc_vhid == carpr.carpr_vhid)
1814 carp_carprcp(&carpr, sc, priveleged);
1815 error = copyout(&carpr, ifr_data_get_ptr(ifr),
1821 IFNET_FOREACH_CARP(ifp, sc)
1824 if (count > carpr.carpr_count) {
1825 CIF_UNLOCK(ifp->if_carp);
1831 IFNET_FOREACH_CARP(ifp, sc) {
1832 carp_carprcp(&carpr, sc, priveleged);
1833 carpr.carpr_count = count;
1834 error = copyout(&carpr,
1835 (caddr_t)ifr_data_get_ptr(ifr) +
1836 (i * sizeof(carpr)), sizeof(carpr));
1838 CIF_UNLOCK(ifp->if_carp);
1849 sx_xunlock(&carp_sx);
1860 carp_get_vhid(struct ifaddr *ifa)
1863 if (ifa == NULL || ifa->ifa_carp == NULL)
1866 return (ifa->ifa_carp->sc_vhid);
1870 carp_attach(struct ifaddr *ifa, int vhid)
1872 struct ifnet *ifp = ifa->ifa_ifp;
1873 struct carp_if *cif = ifp->if_carp;
1874 struct carp_softc *sc;
1877 KASSERT(ifa->ifa_carp == NULL, ("%s: ifa %p attached", __func__, ifa));
1879 switch (ifa->ifa_addr->sa_family) {
1888 return (EPROTOTYPE);
1892 if (ifp->if_carp == NULL) {
1893 sx_xunlock(&carp_sx);
1894 return (ENOPROTOOPT);
1897 IFNET_FOREACH_CARP(ifp, sc)
1898 if (sc->sc_vhid == vhid)
1901 sx_xunlock(&carp_sx);
1905 error = carp_multicast_setup(cif, ifa->ifa_addr->sa_family);
1908 sx_xunlock(&carp_sx);
1912 index = sc->sc_naddrs + sc->sc_naddrs6 + 1;
1913 if (index > sc->sc_ifasiz / sizeof(struct ifaddr *))
1916 switch (ifa->ifa_addr->sa_family) {
1934 sc->sc_ifas[index - 1] = ifa;
1936 carp_hmac_prepare(sc);
1940 sx_xunlock(&carp_sx);
1946 carp_detach(struct ifaddr *ifa, bool keep_cif)
1948 struct ifnet *ifp = ifa->ifa_ifp;
1949 struct carp_if *cif = ifp->if_carp;
1950 struct carp_softc *sc = ifa->ifa_carp;
1953 KASSERT(sc != NULL, ("%s: %p not attached", __func__, ifa));
1959 index = sc->sc_naddrs + sc->sc_naddrs6;
1960 for (i = 0; i < index; i++)
1961 if (sc->sc_ifas[i] == ifa)
1963 KASSERT(i < index, ("%s: %p no backref", __func__, ifa));
1964 for (; i < index - 1; i++)
1965 sc->sc_ifas[i] = sc->sc_ifas[i+1];
1966 sc->sc_ifas[index - 1] = NULL;
1968 switch (ifa->ifa_addr->sa_family) {
1983 carp_ifa_delroute(ifa);
1984 carp_multicast_cleanup(cif, ifa->ifa_addr->sa_family);
1986 ifa->ifa_carp = NULL;
1989 carp_hmac_prepare(sc);
1992 if (!keep_cif && sc->sc_naddrs == 0 && sc->sc_naddrs6 == 0)
2000 sx_xunlock(&carp_sx);
2004 carp_set_state(struct carp_softc *sc, int state, const char *reason)
2007 CARP_LOCK_ASSERT(sc);
2009 if (sc->sc_state != state) {
2010 const char *carp_states[] = { CARP_STATES };
2011 char subsys[IFNAMSIZ+5];
2013 snprintf(subsys, IFNAMSIZ+5, "%u@%s", sc->sc_vhid,
2014 sc->sc_carpdev->if_xname);
2016 CARP_LOG("%s: %s -> %s (%s)\n", subsys,
2017 carp_states[sc->sc_state], carp_states[state], reason);
2019 sc->sc_state = state;
2021 devctl_notify("CARP", subsys, carp_states[state], NULL);
2026 carp_linkstate(struct ifnet *ifp)
2028 struct carp_softc *sc;
2030 CIF_LOCK(ifp->if_carp);
2031 IFNET_FOREACH_CARP(ifp, sc) {
2036 CIF_UNLOCK(ifp->if_carp);
2040 carp_sc_state(struct carp_softc *sc)
2043 CARP_LOCK_ASSERT(sc);
2045 if (sc->sc_carpdev->if_link_state != LINK_STATE_UP ||
2046 !(sc->sc_carpdev->if_flags & IFF_UP) ||
2048 callout_stop(&sc->sc_ad_tmo);
2050 callout_stop(&sc->sc_md_tmo);
2053 callout_stop(&sc->sc_md6_tmo);
2055 carp_set_state(sc, INIT, "hardware interface down");
2057 if (!sc->sc_suppress)
2058 carp_demote_adj(V_carp_ifdown_adj, "interface down");
2059 sc->sc_suppress = 1;
2061 carp_set_state(sc, INIT, "hardware interface up");
2063 if (sc->sc_suppress)
2064 carp_demote_adj(-V_carp_ifdown_adj, "interface up");
2065 sc->sc_suppress = 0;
2070 carp_demote_adj(int adj, char *reason)
2072 atomic_add_int(&V_carp_demotion, adj);
2073 CARP_LOG("demoted by %d to %d (%s)\n", adj, V_carp_demotion, reason);
2074 taskqueue_enqueue(taskqueue_swi, &carp_sendall_task);
2078 carp_allow_sysctl(SYSCTL_HANDLER_ARGS)
2081 struct carp_softc *sc;
2084 error = sysctl_handle_int(oidp, &new, 0, req);
2085 if (error || !req->newptr)
2088 if (V_carp_allow != new) {
2091 mtx_lock(&carp_mtx);
2092 LIST_FOREACH(sc, &carp_list, sc_next) {
2094 if (curvnet == sc->sc_carpdev->if_vnet)
2098 mtx_unlock(&carp_mtx);
2105 carp_demote_adj_sysctl(SYSCTL_HANDLER_ARGS)
2109 new = V_carp_demotion;
2110 error = sysctl_handle_int(oidp, &new, 0, req);
2111 if (error || !req->newptr)
2114 carp_demote_adj(new, "sysctl");
2120 extern struct domain inetdomain;
2121 static struct protosw in_carp_protosw = {
2122 .pr_type = SOCK_RAW,
2123 .pr_domain = &inetdomain,
2124 .pr_protocol = IPPROTO_CARP,
2125 .pr_flags = PR_ATOMIC|PR_ADDR,
2126 .pr_input = carp_input,
2127 .pr_output = rip_output,
2128 .pr_ctloutput = rip_ctloutput,
2129 .pr_usrreqs = &rip_usrreqs
2134 extern struct domain inet6domain;
2135 static struct protosw in6_carp_protosw = {
2136 .pr_type = SOCK_RAW,
2137 .pr_domain = &inet6domain,
2138 .pr_protocol = IPPROTO_CARP,
2139 .pr_flags = PR_ATOMIC|PR_ADDR,
2140 .pr_input = carp6_input,
2141 .pr_output = rip6_output,
2142 .pr_ctloutput = rip6_ctloutput,
2143 .pr_usrreqs = &rip6_usrreqs
2148 carp_mod_cleanup(void)
2152 if (proto_reg[CARP_INET] == 0) {
2153 (void)ipproto_unregister(IPPROTO_CARP);
2154 pf_proto_unregister(PF_INET, IPPROTO_CARP, SOCK_RAW);
2155 proto_reg[CARP_INET] = -1;
2157 carp_iamatch_p = NULL;
2160 if (proto_reg[CARP_INET6] == 0) {
2161 (void)ip6proto_unregister(IPPROTO_CARP);
2162 pf_proto_unregister(PF_INET6, IPPROTO_CARP, SOCK_RAW);
2163 proto_reg[CARP_INET6] = -1;
2165 carp_iamatch6_p = NULL;
2166 carp_macmatch6_p = NULL;
2168 carp_ioctl_p = NULL;
2169 carp_attach_p = NULL;
2170 carp_detach_p = NULL;
2171 carp_get_vhid_p = NULL;
2172 carp_linkstate_p = NULL;
2173 carp_forus_p = NULL;
2174 carp_output_p = NULL;
2175 carp_demote_adj_p = NULL;
2176 carp_master_p = NULL;
2177 mtx_unlock(&carp_mtx);
2178 taskqueue_drain(taskqueue_swi, &carp_sendall_task);
2179 mtx_destroy(&carp_mtx);
2180 sx_destroy(&carp_sx);
2188 mtx_init(&carp_mtx, "carp_mtx", NULL, MTX_DEF);
2189 sx_init(&carp_sx, "carp_sx");
2190 LIST_INIT(&carp_list);
2191 carp_get_vhid_p = carp_get_vhid;
2192 carp_forus_p = carp_forus;
2193 carp_output_p = carp_output;
2194 carp_linkstate_p = carp_linkstate;
2195 carp_ioctl_p = carp_ioctl;
2196 carp_attach_p = carp_attach;
2197 carp_detach_p = carp_detach;
2198 carp_demote_adj_p = carp_demote_adj;
2199 carp_master_p = carp_master;
2201 carp_iamatch6_p = carp_iamatch6;
2202 carp_macmatch6_p = carp_macmatch6;
2203 proto_reg[CARP_INET6] = pf_proto_register(PF_INET6,
2204 (struct protosw *)&in6_carp_protosw);
2205 if (proto_reg[CARP_INET6]) {
2206 printf("carp: error %d attaching to PF_INET6\n",
2207 proto_reg[CARP_INET6]);
2209 return (proto_reg[CARP_INET6]);
2211 err = ip6proto_register(IPPROTO_CARP);
2213 printf("carp: error %d registering with INET6\n", err);
2219 carp_iamatch_p = carp_iamatch;
2220 proto_reg[CARP_INET] = pf_proto_register(PF_INET, &in_carp_protosw);
2221 if (proto_reg[CARP_INET]) {
2222 printf("carp: error %d attaching to PF_INET\n",
2223 proto_reg[CARP_INET]);
2225 return (proto_reg[CARP_INET]);
2227 err = ipproto_register(IPPROTO_CARP);
2229 printf("carp: error %d registering with INET\n", err);
2238 carp_modevent(module_t mod, int type, void *data)
2242 return carp_mod_load();
2245 mtx_lock(&carp_mtx);
2246 if (LIST_EMPTY(&carp_list))
2249 mtx_unlock(&carp_mtx);
2261 static moduledata_t carp_mod = {
2267 DECLARE_MODULE(carp, carp_mod, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY);