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>
40 #include <sys/devctl.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 VNET_DEFINE_STATIC(int, carp_allow) = 1;
191 #define V_carp_allow VNET(carp_allow)
193 /* Set DSCP in outgoing CARP packets. */
194 VNET_DEFINE_STATIC(int, carp_dscp) = 56;
195 #define V_carp_dscp VNET(carp_dscp)
197 /* Preempt slower nodes. */
198 VNET_DEFINE_STATIC(int, carp_preempt) = 0;
199 #define V_carp_preempt VNET(carp_preempt)
202 VNET_DEFINE_STATIC(int, carp_log) = 1;
203 #define V_carp_log VNET(carp_log)
205 /* Global advskew demotion. */
206 VNET_DEFINE_STATIC(int, carp_demotion) = 0;
207 #define V_carp_demotion VNET(carp_demotion)
209 /* Send error demotion factor. */
210 VNET_DEFINE_STATIC(int, carp_senderr_adj) = CARP_MAXSKEW;
211 #define V_carp_senderr_adj VNET(carp_senderr_adj)
213 /* Iface down demotion factor. */
214 VNET_DEFINE_STATIC(int, carp_ifdown_adj) = CARP_MAXSKEW;
215 #define V_carp_ifdown_adj VNET(carp_ifdown_adj)
217 static int carp_allow_sysctl(SYSCTL_HANDLER_ARGS);
218 static int carp_dscp_sysctl(SYSCTL_HANDLER_ARGS);
219 static int carp_demote_adj_sysctl(SYSCTL_HANDLER_ARGS);
221 SYSCTL_NODE(_net_inet, IPPROTO_CARP, carp, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
223 SYSCTL_PROC(_net_inet_carp, OID_AUTO, allow,
224 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
225 0, 0, carp_allow_sysctl, "I",
226 "Accept incoming CARP packets");
227 SYSCTL_PROC(_net_inet_carp, OID_AUTO, dscp,
228 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
229 0, 0, carp_dscp_sysctl, "I",
230 "DSCP value for carp packets");
231 SYSCTL_INT(_net_inet_carp, OID_AUTO, preempt, CTLFLAG_VNET | CTLFLAG_RW,
232 &VNET_NAME(carp_preempt), 0, "High-priority backup preemption mode");
233 SYSCTL_INT(_net_inet_carp, OID_AUTO, log, CTLFLAG_VNET | CTLFLAG_RW,
234 &VNET_NAME(carp_log), 0, "CARP log level");
235 SYSCTL_PROC(_net_inet_carp, OID_AUTO, demotion,
236 CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE,
237 0, 0, carp_demote_adj_sysctl, "I",
238 "Adjust demotion factor (skew of advskew)");
239 SYSCTL_INT(_net_inet_carp, OID_AUTO, senderr_demotion_factor,
240 CTLFLAG_VNET | CTLFLAG_RW,
241 &VNET_NAME(carp_senderr_adj), 0, "Send error demotion factor adjustment");
242 SYSCTL_INT(_net_inet_carp, OID_AUTO, ifdown_demotion_factor,
243 CTLFLAG_VNET | CTLFLAG_RW,
244 &VNET_NAME(carp_ifdown_adj), 0,
245 "Interface down demotion factor adjustment");
247 VNET_PCPUSTAT_DEFINE(struct carpstats, carpstats);
248 VNET_PCPUSTAT_SYSINIT(carpstats);
249 VNET_PCPUSTAT_SYSUNINIT(carpstats);
251 #define CARPSTATS_ADD(name, val) \
252 counter_u64_add(VNET(carpstats)[offsetof(struct carpstats, name) / \
253 sizeof(uint64_t)], (val))
254 #define CARPSTATS_INC(name) CARPSTATS_ADD(name, 1)
256 SYSCTL_VNET_PCPUSTAT(_net_inet_carp, OID_AUTO, stats, struct carpstats,
257 carpstats, "CARP statistics (struct carpstats, netinet/ip_carp.h)");
259 #define CARP_LOCK_INIT(sc) mtx_init(&(sc)->sc_mtx, "carp_softc", \
261 #define CARP_LOCK_DESTROY(sc) mtx_destroy(&(sc)->sc_mtx)
262 #define CARP_LOCK_ASSERT(sc) mtx_assert(&(sc)->sc_mtx, MA_OWNED)
263 #define CARP_LOCK(sc) mtx_lock(&(sc)->sc_mtx)
264 #define CARP_UNLOCK(sc) mtx_unlock(&(sc)->sc_mtx)
265 #define CIF_LOCK_INIT(cif) mtx_init(&(cif)->cif_mtx, "carp_if", \
267 #define CIF_LOCK_DESTROY(cif) mtx_destroy(&(cif)->cif_mtx)
268 #define CIF_LOCK_ASSERT(cif) mtx_assert(&(cif)->cif_mtx, MA_OWNED)
269 #define CIF_LOCK(cif) mtx_lock(&(cif)->cif_mtx)
270 #define CIF_UNLOCK(cif) mtx_unlock(&(cif)->cif_mtx)
271 #define CIF_FREE(cif) do { \
273 if (TAILQ_EMPTY(&(cif)->cif_vrs)) \
279 #define CARP_LOG(...) do { \
280 if (V_carp_log > 0) \
281 log(LOG_INFO, "carp: " __VA_ARGS__); \
284 #define CARP_DEBUG(...) do { \
285 if (V_carp_log > 1) \
286 log(LOG_DEBUG, __VA_ARGS__); \
289 #define IFNET_FOREACH_IFA(ifp, ifa) \
290 CK_STAILQ_FOREACH((ifa), &(ifp)->if_addrhead, ifa_link) \
291 if ((ifa)->ifa_carp != NULL)
293 #define CARP_FOREACH_IFA(sc, ifa) \
294 CARP_LOCK_ASSERT(sc); \
296 _i < (sc)->sc_naddrs + (sc)->sc_naddrs6 && \
297 ((ifa) = sc->sc_ifas[_i]) != NULL; \
300 #define IFNET_FOREACH_CARP(ifp, sc) \
301 KASSERT(mtx_owned(&ifp->if_carp->cif_mtx) || \
302 sx_xlocked(&carp_sx), ("cif_vrs not locked")); \
303 TAILQ_FOREACH((sc), &(ifp)->if_carp->cif_vrs, sc_list)
305 #define DEMOTE_ADVSKEW(sc) \
306 (((sc)->sc_advskew + V_carp_demotion > CARP_MAXSKEW) ? \
307 CARP_MAXSKEW : ((sc)->sc_advskew + V_carp_demotion))
309 static void carp_input_c(struct mbuf *, struct carp_header *, sa_family_t);
310 static struct carp_softc
311 *carp_alloc(struct ifnet *);
312 static void carp_destroy(struct carp_softc *);
313 static struct carp_if
314 *carp_alloc_if(struct ifnet *);
315 static void carp_free_if(struct carp_if *);
316 static void carp_set_state(struct carp_softc *, int, const char* reason);
317 static void carp_sc_state(struct carp_softc *);
318 static void carp_setrun(struct carp_softc *, sa_family_t);
319 static void carp_master_down(void *);
320 static void carp_master_down_locked(struct carp_softc *,
322 static void carp_send_ad(void *);
323 static void carp_send_ad_locked(struct carp_softc *);
324 static void carp_addroute(struct carp_softc *);
325 static void carp_ifa_addroute(struct ifaddr *);
326 static void carp_delroute(struct carp_softc *);
327 static void carp_ifa_delroute(struct ifaddr *);
328 static void carp_send_ad_all(void *, int);
329 static void carp_demote_adj(int, char *);
331 static LIST_HEAD(, carp_softc) carp_list;
332 static struct mtx carp_mtx;
333 static struct sx carp_sx;
334 static struct task carp_sendall_task =
335 TASK_INITIALIZER(0, carp_send_ad_all, NULL);
338 carp_hmac_prepare(struct carp_softc *sc)
340 uint8_t version = CARP_VERSION, type = CARP_ADVERTISEMENT;
341 uint8_t vhid = sc->sc_vhid & 0xff;
345 struct in_addr last, cur, in;
348 struct in6_addr last6, cur6, in6;
351 CARP_LOCK_ASSERT(sc);
353 /* Compute ipad from key. */
354 bzero(sc->sc_pad, sizeof(sc->sc_pad));
355 bcopy(sc->sc_key, sc->sc_pad, sizeof(sc->sc_key));
356 for (i = 0; i < sizeof(sc->sc_pad); i++)
357 sc->sc_pad[i] ^= 0x36;
359 /* Precompute first part of inner hash. */
360 SHA1Init(&sc->sc_sha1);
361 SHA1Update(&sc->sc_sha1, sc->sc_pad, sizeof(sc->sc_pad));
362 SHA1Update(&sc->sc_sha1, (void *)&version, sizeof(version));
363 SHA1Update(&sc->sc_sha1, (void *)&type, sizeof(type));
364 SHA1Update(&sc->sc_sha1, (void *)&vhid, sizeof(vhid));
370 cur.s_addr = 0xffffffff;
371 CARP_FOREACH_IFA(sc, ifa) {
372 in.s_addr = ifatoia(ifa)->ia_addr.sin_addr.s_addr;
373 if (ifa->ifa_addr->sa_family == AF_INET &&
374 ntohl(in.s_addr) > ntohl(last.s_addr) &&
375 ntohl(in.s_addr) < ntohl(cur.s_addr)) {
376 cur.s_addr = in.s_addr;
381 SHA1Update(&sc->sc_sha1, (void *)&cur, sizeof(cur));
385 memset(&cur6, 0, sizeof(cur6));
389 memset(&cur6, 0xff, sizeof(cur6));
390 CARP_FOREACH_IFA(sc, ifa) {
391 in6 = ifatoia6(ifa)->ia_addr.sin6_addr;
392 if (IN6_IS_SCOPE_EMBED(&in6))
393 in6.s6_addr16[1] = 0;
394 if (ifa->ifa_addr->sa_family == AF_INET6 &&
395 memcmp(&in6, &last6, sizeof(in6)) > 0 &&
396 memcmp(&in6, &cur6, sizeof(in6)) < 0) {
402 SHA1Update(&sc->sc_sha1, (void *)&cur6, sizeof(cur6));
406 /* convert ipad to opad */
407 for (i = 0; i < sizeof(sc->sc_pad); i++)
408 sc->sc_pad[i] ^= 0x36 ^ 0x5c;
412 carp_hmac_generate(struct carp_softc *sc, uint32_t counter[2],
413 unsigned char md[20])
417 CARP_LOCK_ASSERT(sc);
419 /* fetch first half of inner hash */
420 bcopy(&sc->sc_sha1, &sha1ctx, sizeof(sha1ctx));
422 SHA1Update(&sha1ctx, (void *)counter, sizeof(sc->sc_counter));
423 SHA1Final(md, &sha1ctx);
427 SHA1Update(&sha1ctx, sc->sc_pad, sizeof(sc->sc_pad));
428 SHA1Update(&sha1ctx, md, 20);
429 SHA1Final(md, &sha1ctx);
433 carp_hmac_verify(struct carp_softc *sc, uint32_t counter[2],
434 unsigned char md[20])
436 unsigned char md2[20];
438 CARP_LOCK_ASSERT(sc);
440 carp_hmac_generate(sc, counter, md2);
442 return (bcmp(md, md2, sizeof(md2)));
446 * process input packet.
447 * we have rearranged checks order compared to the rfc,
448 * but it seems more efficient this way or not possible otherwise.
452 carp_input(struct mbuf **mp, int *offp, int proto)
454 struct mbuf *m = *mp;
455 struct ip *ip = mtod(m, struct ip *);
456 struct carp_header *ch;
462 CARPSTATS_INC(carps_ipackets);
466 return (IPPROTO_DONE);
469 /* verify that the IP TTL is 255. */
470 if (ip->ip_ttl != CARP_DFLTTL) {
471 CARPSTATS_INC(carps_badttl);
472 CARP_DEBUG("%s: received ttl %d != 255 on %s\n", __func__,
474 m->m_pkthdr.rcvif->if_xname);
476 return (IPPROTO_DONE);
479 iplen = ip->ip_hl << 2;
481 if (m->m_pkthdr.len < iplen + sizeof(*ch)) {
482 CARPSTATS_INC(carps_badlen);
483 CARP_DEBUG("%s: received len %zd < sizeof(struct carp_header) "
484 "on %s\n", __func__, m->m_len - sizeof(struct ip),
485 m->m_pkthdr.rcvif->if_xname);
487 return (IPPROTO_DONE);
490 if (iplen + sizeof(*ch) < m->m_len) {
491 if ((m = m_pullup(m, iplen + sizeof(*ch))) == NULL) {
492 CARPSTATS_INC(carps_hdrops);
493 CARP_DEBUG("%s: pullup failed\n", __func__);
494 return (IPPROTO_DONE);
496 ip = mtod(m, struct ip *);
498 ch = (struct carp_header *)((char *)ip + iplen);
501 * verify that the received packet length is
502 * equal to the CARP header
504 len = iplen + sizeof(*ch);
505 if (len > m->m_pkthdr.len) {
506 CARPSTATS_INC(carps_badlen);
507 CARP_DEBUG("%s: packet too short %d on %s\n", __func__,
509 m->m_pkthdr.rcvif->if_xname);
511 return (IPPROTO_DONE);
514 if ((m = m_pullup(m, len)) == NULL) {
515 CARPSTATS_INC(carps_hdrops);
516 return (IPPROTO_DONE);
518 ip = mtod(m, struct ip *);
519 ch = (struct carp_header *)((char *)ip + iplen);
521 /* verify the CARP checksum */
523 if (in_cksum(m, len - iplen)) {
524 CARPSTATS_INC(carps_badsum);
525 CARP_DEBUG("%s: checksum failed on %s\n", __func__,
526 m->m_pkthdr.rcvif->if_xname);
528 return (IPPROTO_DONE);
532 carp_input_c(m, ch, AF_INET);
533 return (IPPROTO_DONE);
539 carp6_input(struct mbuf **mp, int *offp, int proto)
541 struct mbuf *m = *mp;
542 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
543 struct carp_header *ch;
546 CARPSTATS_INC(carps_ipackets6);
550 return (IPPROTO_DONE);
553 /* check if received on a valid carp interface */
554 if (m->m_pkthdr.rcvif->if_carp == NULL) {
555 CARPSTATS_INC(carps_badif);
556 CARP_DEBUG("%s: packet received on non-carp interface: %s\n",
557 __func__, m->m_pkthdr.rcvif->if_xname);
559 return (IPPROTO_DONE);
562 /* verify that the IP TTL is 255 */
563 if (ip6->ip6_hlim != CARP_DFLTTL) {
564 CARPSTATS_INC(carps_badttl);
565 CARP_DEBUG("%s: received ttl %d != 255 on %s\n", __func__,
566 ip6->ip6_hlim, m->m_pkthdr.rcvif->if_xname);
568 return (IPPROTO_DONE);
571 /* verify that we have a complete carp packet */
572 if (m->m_len < *offp + sizeof(*ch)) {
574 m = m_pullup(m, *offp + sizeof(*ch));
576 CARPSTATS_INC(carps_badlen);
577 CARP_DEBUG("%s: packet size %u too small\n", __func__, len);
578 return (IPPROTO_DONE);
581 ch = (struct carp_header *)(mtod(m, char *) + *offp);
583 /* verify the CARP checksum */
585 if (in_cksum(m, sizeof(*ch))) {
586 CARPSTATS_INC(carps_badsum);
587 CARP_DEBUG("%s: checksum failed, on %s\n", __func__,
588 m->m_pkthdr.rcvif->if_xname);
590 return (IPPROTO_DONE);
594 carp_input_c(m, ch, AF_INET6);
595 return (IPPROTO_DONE);
600 * This routine should not be necessary at all, but some switches
601 * (VMWare ESX vswitches) can echo our own packets back at us,
602 * and we must ignore them or they will cause us to drop out of
605 * We cannot catch all cases of network loops. Instead, what we
606 * do here is catch any packet that arrives with a carp header
607 * with a VHID of 0, that comes from an address that is our own.
608 * These packets are by definition "from us" (even if they are from
609 * a misconfigured host that is pretending to be us).
611 * The VHID test is outside this mini-function.
614 carp_source_is_self(struct mbuf *m, struct ifaddr *ifa, sa_family_t af)
628 ip4 = mtod(m, struct ip *);
629 in4 = ifatoia(ifa)->ia_addr.sin_addr;
630 return (in4.s_addr == ip4->ip_src.s_addr);
634 ip6 = mtod(m, struct ip6_hdr *);
635 in6 = ifatoia6(ifa)->ia_addr.sin6_addr;
636 return (memcmp(&in6, &ip6->ip6_src, sizeof(in6)) == 0);
645 carp_input_c(struct mbuf *m, struct carp_header *ch, sa_family_t af)
647 struct ifnet *ifp = m->m_pkthdr.rcvif;
648 struct ifaddr *ifa, *match;
649 struct carp_softc *sc;
650 uint64_t tmp_counter;
651 struct timeval sc_tv, ch_tv;
657 * Verify that the VHID is valid on the receiving interface.
659 * There should be just one match. If there are none
660 * the VHID is not valid and we drop the packet. If
661 * there are multiple VHID matches, take just the first
662 * one, for compatibility with previous code. While we're
663 * scanning, check for obvious loops in the network topology
664 * (these should never happen, and as noted above, we may
665 * miss real loops; this is just a double-check).
669 IFNET_FOREACH_IFA(ifp, ifa) {
670 if (match == NULL && ifa->ifa_carp != NULL &&
671 ifa->ifa_addr->sa_family == af &&
672 ifa->ifa_carp->sc_vhid == ch->carp_vhid)
674 if (ch->carp_vhid == 0 && carp_source_is_self(m, ifa, af))
677 ifa = error ? NULL : match;
682 if (error == ELOOP) {
683 CARP_DEBUG("dropping looped packet on interface %s\n",
685 CARPSTATS_INC(carps_badif); /* ??? */
687 CARPSTATS_INC(carps_badvhid);
693 /* verify the CARP version. */
694 if (ch->carp_version != CARP_VERSION) {
695 CARPSTATS_INC(carps_badver);
696 CARP_DEBUG("%s: invalid version %d\n", ifp->if_xname,
707 if (carp_hmac_verify(sc, ch->carp_counter, ch->carp_md)) {
708 CARPSTATS_INC(carps_badauth);
709 CARP_DEBUG("%s: incorrect hash for VHID %u@%s\n", __func__,
710 sc->sc_vhid, ifp->if_xname);
714 tmp_counter = ntohl(ch->carp_counter[0]);
715 tmp_counter = tmp_counter<<32;
716 tmp_counter += ntohl(ch->carp_counter[1]);
718 /* XXX Replay protection goes here */
720 sc->sc_init_counter = 0;
721 sc->sc_counter = tmp_counter;
723 sc_tv.tv_sec = sc->sc_advbase;
724 sc_tv.tv_usec = DEMOTE_ADVSKEW(sc) * 1000000 / 256;
725 ch_tv.tv_sec = ch->carp_advbase;
726 ch_tv.tv_usec = ch->carp_advskew * 1000000 / 256;
728 switch (sc->sc_state) {
733 * If we receive an advertisement from a master who's going to
734 * be more frequent than us, go into BACKUP state.
736 if (timevalcmp(&sc_tv, &ch_tv, >) ||
737 timevalcmp(&sc_tv, &ch_tv, ==)) {
738 callout_stop(&sc->sc_ad_tmo);
739 carp_set_state(sc, BACKUP,
740 "more frequent advertisement received");
747 * If we're pre-empting masters who advertise slower than us,
748 * and this one claims to be slower, treat him as down.
750 if (V_carp_preempt && timevalcmp(&sc_tv, &ch_tv, <)) {
751 carp_master_down_locked(sc,
752 "preempting a slower master");
757 * If the master is going to advertise at such a low frequency
758 * that he's guaranteed to time out, we'd might as well just
759 * treat him as timed out now.
761 sc_tv.tv_sec = sc->sc_advbase * 3;
762 if (timevalcmp(&sc_tv, &ch_tv, <)) {
763 carp_master_down_locked(sc, "master will time out");
768 * Otherwise, we reset the counter and wait for the next
781 carp_prepare_ad(struct mbuf *m, struct carp_softc *sc, struct carp_header *ch)
785 if (sc->sc_init_counter) {
786 /* this could also be seconds since unix epoch */
787 sc->sc_counter = arc4random();
788 sc->sc_counter = sc->sc_counter << 32;
789 sc->sc_counter += arc4random();
793 ch->carp_counter[0] = htonl((sc->sc_counter>>32)&0xffffffff);
794 ch->carp_counter[1] = htonl(sc->sc_counter&0xffffffff);
796 carp_hmac_generate(sc, ch->carp_counter, ch->carp_md);
798 /* Tag packet for carp_output */
799 if ((mtag = m_tag_get(PACKET_TAG_CARP, sizeof(struct carp_softc *),
800 M_NOWAIT)) == NULL) {
802 CARPSTATS_INC(carps_onomem);
805 bcopy(&sc, mtag + 1, sizeof(sc));
806 m_tag_prepend(m, mtag);
812 * To avoid LORs and possible recursions this function shouldn't
813 * be called directly, but scheduled via taskqueue.
816 carp_send_ad_all(void *ctx __unused, int pending __unused)
818 struct carp_softc *sc;
819 struct epoch_tracker et;
823 LIST_FOREACH(sc, &carp_list, sc_next)
824 if (sc->sc_state == MASTER) {
826 CURVNET_SET(sc->sc_carpdev->if_vnet);
827 carp_send_ad_locked(sc);
831 mtx_unlock(&carp_mtx);
835 /* Send a periodic advertisement, executed in callout context. */
837 carp_send_ad(void *v)
839 struct carp_softc *sc = v;
840 struct epoch_tracker et;
843 CARP_LOCK_ASSERT(sc);
844 CURVNET_SET(sc->sc_carpdev->if_vnet);
845 carp_send_ad_locked(sc);
852 carp_send_ad_error(struct carp_softc *sc, int error)
856 if (sc->sc_sendad_errors < INT_MAX)
857 sc->sc_sendad_errors++;
858 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) {
859 static const char fmt[] = "send error %d on %s";
860 char msg[sizeof(fmt) + IFNAMSIZ];
862 sprintf(msg, fmt, error, sc->sc_carpdev->if_xname);
863 carp_demote_adj(V_carp_senderr_adj, msg);
865 sc->sc_sendad_success = 0;
867 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS &&
868 ++sc->sc_sendad_success >= CARP_SENDAD_MIN_SUCCESS) {
869 static const char fmt[] = "send ok on %s";
870 char msg[sizeof(fmt) + IFNAMSIZ];
872 sprintf(msg, fmt, sc->sc_carpdev->if_xname);
873 carp_demote_adj(-V_carp_senderr_adj, msg);
874 sc->sc_sendad_errors = 0;
876 sc->sc_sendad_errors = 0;
881 * Pick the best ifaddr on the given ifp for sending CARP
884 * "Best" here is defined by ifa_preferred(). This function is much
885 * much like ifaof_ifpforaddr() except that we just use ifa_preferred().
887 * (This could be simplified to return the actual address, except that
888 * it has a different format in AF_INET and AF_INET6.)
890 static struct ifaddr *
891 carp_best_ifa(int af, struct ifnet *ifp)
893 struct ifaddr *ifa, *best;
900 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
901 if (ifa->ifa_addr->sa_family == af &&
902 (best == NULL || ifa_preferred(best, ifa)))
911 carp_send_ad_locked(struct carp_softc *sc)
913 struct carp_header ch;
916 struct carp_header *ch_ptr;
921 CARP_LOCK_ASSERT(sc);
923 advskew = DEMOTE_ADVSKEW(sc);
924 tv.tv_sec = sc->sc_advbase;
925 tv.tv_usec = advskew * 1000000 / 256;
927 ch.carp_version = CARP_VERSION;
928 ch.carp_type = CARP_ADVERTISEMENT;
929 ch.carp_vhid = sc->sc_vhid;
930 ch.carp_advbase = sc->sc_advbase;
931 ch.carp_advskew = advskew;
932 ch.carp_authlen = 7; /* XXX DEFINE */
933 ch.carp_pad1 = 0; /* must be zero */
936 /* XXXGL: OpenBSD picks first ifaddr with needed family. */
942 m = m_gethdr(M_NOWAIT, MT_DATA);
944 CARPSTATS_INC(carps_onomem);
947 len = sizeof(*ip) + sizeof(ch);
948 m->m_pkthdr.len = len;
949 m->m_pkthdr.rcvif = NULL;
951 M_ALIGN(m, m->m_len);
952 m->m_flags |= M_MCAST;
953 ip = mtod(m, struct ip *);
954 ip->ip_v = IPVERSION;
955 ip->ip_hl = sizeof(*ip) >> 2;
956 ip->ip_tos = V_carp_dscp << IPTOS_DSCP_OFFSET;
957 ip->ip_len = htons(len);
958 ip->ip_off = htons(IP_DF);
959 ip->ip_ttl = CARP_DFLTTL;
960 ip->ip_p = IPPROTO_CARP;
964 ifa = carp_best_ifa(AF_INET, sc->sc_carpdev);
967 ifatoia(ifa)->ia_addr.sin_addr.s_addr;
970 ip->ip_src.s_addr = 0;
971 ip->ip_dst.s_addr = htonl(INADDR_CARP_GROUP);
973 ch_ptr = (struct carp_header *)(&ip[1]);
974 bcopy(&ch, ch_ptr, sizeof(ch));
975 if (carp_prepare_ad(m, sc, ch_ptr))
978 m->m_data += sizeof(*ip);
979 ch_ptr->carp_cksum = in_cksum(m, len - sizeof(*ip));
980 m->m_data -= sizeof(*ip);
982 CARPSTATS_INC(carps_opackets);
984 carp_send_ad_error(sc, ip_output(m, NULL, NULL, IP_RAWOUTPUT,
985 &sc->sc_carpdev->if_carp->cif_imo, NULL));
989 if (sc->sc_naddrs6) {
992 m = m_gethdr(M_NOWAIT, MT_DATA);
994 CARPSTATS_INC(carps_onomem);
997 len = sizeof(*ip6) + sizeof(ch);
998 m->m_pkthdr.len = len;
999 m->m_pkthdr.rcvif = NULL;
1001 M_ALIGN(m, m->m_len);
1002 m->m_flags |= M_MCAST;
1003 ip6 = mtod(m, struct ip6_hdr *);
1004 bzero(ip6, sizeof(*ip6));
1005 ip6->ip6_vfc |= IPV6_VERSION;
1006 /* Traffic class isn't defined in ip6 struct instead
1007 * it gets offset into flowid field */
1008 ip6->ip6_flow |= htonl(V_carp_dscp << (IPV6_FLOWLABEL_LEN +
1009 IPTOS_DSCP_OFFSET));
1010 ip6->ip6_hlim = CARP_DFLTTL;
1011 ip6->ip6_nxt = IPPROTO_CARP;
1013 /* set the source address */
1014 ifa = carp_best_ifa(AF_INET6, sc->sc_carpdev);
1016 bcopy(IFA_IN6(ifa), &ip6->ip6_src,
1017 sizeof(struct in6_addr));
1020 /* This should never happen with IPv6. */
1021 bzero(&ip6->ip6_src, sizeof(struct in6_addr));
1023 /* Set the multicast destination. */
1024 ip6->ip6_dst.s6_addr16[0] = htons(0xff02);
1025 ip6->ip6_dst.s6_addr8[15] = 0x12;
1026 if (in6_setscope(&ip6->ip6_dst, sc->sc_carpdev, NULL) != 0) {
1028 CARP_DEBUG("%s: in6_setscope failed\n", __func__);
1032 ch_ptr = (struct carp_header *)(&ip6[1]);
1033 bcopy(&ch, ch_ptr, sizeof(ch));
1034 if (carp_prepare_ad(m, sc, ch_ptr))
1037 m->m_data += sizeof(*ip6);
1038 ch_ptr->carp_cksum = in_cksum(m, len - sizeof(*ip6));
1039 m->m_data -= sizeof(*ip6);
1041 CARPSTATS_INC(carps_opackets6);
1043 carp_send_ad_error(sc, ip6_output(m, NULL, NULL, 0,
1044 &sc->sc_carpdev->if_carp->cif_im6o, NULL, NULL));
1049 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv), carp_send_ad, sc);
1053 carp_addroute(struct carp_softc *sc)
1057 CARP_FOREACH_IFA(sc, ifa)
1058 carp_ifa_addroute(ifa);
1062 carp_ifa_addroute(struct ifaddr *ifa)
1065 switch (ifa->ifa_addr->sa_family) {
1068 in_addprefix(ifatoia(ifa), RTF_UP);
1069 ifa_add_loopback_route(ifa,
1070 (struct sockaddr *)&ifatoia(ifa)->ia_addr);
1075 ifa_add_loopback_route(ifa,
1076 (struct sockaddr *)&ifatoia6(ifa)->ia_addr);
1077 nd6_add_ifa_lle(ifatoia6(ifa));
1084 carp_delroute(struct carp_softc *sc)
1088 CARP_FOREACH_IFA(sc, ifa)
1089 carp_ifa_delroute(ifa);
1093 carp_ifa_delroute(struct ifaddr *ifa)
1096 switch (ifa->ifa_addr->sa_family) {
1099 ifa_del_loopback_route(ifa,
1100 (struct sockaddr *)&ifatoia(ifa)->ia_addr);
1101 in_scrubprefix(ifatoia(ifa), LLE_STATIC);
1106 ifa_del_loopback_route(ifa,
1107 (struct sockaddr *)&ifatoia6(ifa)->ia_addr);
1108 nd6_rem_ifa_lle(ifatoia6(ifa), 1);
1115 carp_master(struct ifaddr *ifa)
1117 struct carp_softc *sc = ifa->ifa_carp;
1119 return (sc->sc_state == MASTER);
1124 * Broadcast a gratuitous ARP request containing
1125 * the virtual router MAC address for each IP address
1126 * associated with the virtual router.
1129 carp_send_arp(struct carp_softc *sc)
1132 struct in_addr addr;
1136 CARP_FOREACH_IFA(sc, ifa) {
1137 if (ifa->ifa_addr->sa_family != AF_INET)
1139 addr = ((struct sockaddr_in *)ifa->ifa_addr)->sin_addr;
1140 arp_announce_ifaddr(sc->sc_carpdev, addr, LLADDR(&sc->sc_addr));
1145 carp_iamatch(struct ifaddr *ifa, uint8_t **enaddr)
1147 struct carp_softc *sc = ifa->ifa_carp;
1149 if (sc->sc_state == MASTER) {
1150 *enaddr = LLADDR(&sc->sc_addr);
1160 carp_send_na(struct carp_softc *sc)
1162 static struct in6_addr mcast = IN6ADDR_LINKLOCAL_ALLNODES_INIT;
1164 struct in6_addr *in6;
1166 CARP_FOREACH_IFA(sc, ifa) {
1167 if (ifa->ifa_addr->sa_family != AF_INET6)
1171 nd6_na_output(sc->sc_carpdev, &mcast, in6,
1172 ND_NA_FLAG_OVERRIDE, 1, NULL);
1173 DELAY(1000); /* XXX */
1178 * Returns ifa in case it's a carp address and it is MASTER, or if the address
1179 * matches and is not a carp address. Returns NULL otherwise.
1182 carp_iamatch6(struct ifnet *ifp, struct in6_addr *taddr)
1189 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
1190 if (ifa->ifa_addr->sa_family != AF_INET6)
1192 if (!IN6_ARE_ADDR_EQUAL(taddr, IFA_IN6(ifa)))
1194 if (ifa->ifa_carp && ifa->ifa_carp->sc_state != MASTER)
1205 carp_macmatch6(struct ifnet *ifp, struct mbuf *m, const struct in6_addr *taddr)
1211 IFNET_FOREACH_IFA(ifp, ifa)
1212 if (ifa->ifa_addr->sa_family == AF_INET6 &&
1213 IN6_ARE_ADDR_EQUAL(taddr, IFA_IN6(ifa))) {
1214 struct carp_softc *sc = ifa->ifa_carp;
1217 mtag = m_tag_get(PACKET_TAG_CARP,
1218 sizeof(struct carp_softc *), M_NOWAIT);
1220 /* Better a bit than nothing. */
1221 return (LLADDR(&sc->sc_addr));
1223 bcopy(&sc, mtag + 1, sizeof(sc));
1224 m_tag_prepend(m, mtag);
1226 return (LLADDR(&sc->sc_addr));
1234 carp_forus(struct ifnet *ifp, u_char *dhost)
1236 struct carp_softc *sc;
1237 uint8_t *ena = dhost;
1239 if (ena[0] || ena[1] || ena[2] != 0x5e || ena[3] || ena[4] != 1)
1242 CIF_LOCK(ifp->if_carp);
1243 IFNET_FOREACH_CARP(ifp, sc) {
1245 * CARP_LOCK() is not here, since would protect nothing, but
1246 * cause deadlock with if_bridge, calling this under its lock.
1248 if (sc->sc_state == MASTER && !bcmp(dhost, LLADDR(&sc->sc_addr),
1250 CIF_UNLOCK(ifp->if_carp);
1254 CIF_UNLOCK(ifp->if_carp);
1259 /* Master down timeout event, executed in callout context. */
1261 carp_master_down(void *v)
1263 struct carp_softc *sc = v;
1264 struct epoch_tracker et;
1266 NET_EPOCH_ENTER(et);
1267 CARP_LOCK_ASSERT(sc);
1269 CURVNET_SET(sc->sc_carpdev->if_vnet);
1270 if (sc->sc_state == BACKUP) {
1271 carp_master_down_locked(sc, "master timed out");
1280 carp_master_down_locked(struct carp_softc *sc, const char *reason)
1284 CARP_LOCK_ASSERT(sc);
1286 switch (sc->sc_state) {
1288 carp_set_state(sc, MASTER, reason);
1289 carp_send_ad_locked(sc);
1302 panic("carp: VHID %u@%s: master_down event in %s state\n",
1304 sc->sc_carpdev->if_xname,
1305 sc->sc_state ? "MASTER" : "INIT");
1312 * When in backup state, af indicates whether to reset the master down timer
1313 * for v4 or v6. If it's set to zero, reset the ones which are already pending.
1316 carp_setrun(struct carp_softc *sc, sa_family_t af)
1320 CARP_LOCK_ASSERT(sc);
1322 if ((sc->sc_carpdev->if_flags & IFF_UP) == 0 ||
1323 sc->sc_carpdev->if_link_state != LINK_STATE_UP ||
1324 (sc->sc_naddrs == 0 && sc->sc_naddrs6 == 0) ||
1328 switch (sc->sc_state) {
1330 carp_set_state(sc, BACKUP, "initialization complete");
1334 callout_stop(&sc->sc_ad_tmo);
1335 tv.tv_sec = 3 * sc->sc_advbase;
1336 tv.tv_usec = sc->sc_advskew * 1000000 / 256;
1340 callout_reset(&sc->sc_md_tmo, tvtohz(&tv),
1341 carp_master_down, sc);
1346 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv),
1347 carp_master_down, sc);
1353 callout_reset(&sc->sc_md_tmo, tvtohz(&tv),
1354 carp_master_down, sc);
1358 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv),
1359 carp_master_down, sc);
1365 tv.tv_sec = sc->sc_advbase;
1366 tv.tv_usec = sc->sc_advskew * 1000000 / 256;
1367 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv),
1374 * Setup multicast structures.
1377 carp_multicast_setup(struct carp_if *cif, sa_family_t sa)
1379 struct ifnet *ifp = cif->cif_ifp;
1386 struct ip_moptions *imo = &cif->cif_imo;
1387 struct in_mfilter *imf;
1388 struct in_addr addr;
1390 if (ip_mfilter_first(&imo->imo_head) != NULL)
1393 imf = ip_mfilter_alloc(M_WAITOK, 0, 0);
1394 ip_mfilter_init(&imo->imo_head);
1395 imo->imo_multicast_vif = -1;
1397 addr.s_addr = htonl(INADDR_CARP_GROUP);
1398 if ((error = in_joingroup(ifp, &addr, NULL,
1399 &imf->imf_inm)) != 0) {
1400 ip_mfilter_free(imf);
1404 ip_mfilter_insert(&imo->imo_head, imf);
1405 imo->imo_multicast_ifp = ifp;
1406 imo->imo_multicast_ttl = CARP_DFLTTL;
1407 imo->imo_multicast_loop = 0;
1414 struct ip6_moptions *im6o = &cif->cif_im6o;
1415 struct in6_mfilter *im6f[2];
1416 struct in6_addr in6;
1418 if (ip6_mfilter_first(&im6o->im6o_head))
1421 im6f[0] = ip6_mfilter_alloc(M_WAITOK, 0, 0);
1422 im6f[1] = ip6_mfilter_alloc(M_WAITOK, 0, 0);
1424 ip6_mfilter_init(&im6o->im6o_head);
1425 im6o->im6o_multicast_hlim = CARP_DFLTTL;
1426 im6o->im6o_multicast_ifp = ifp;
1428 /* Join IPv6 CARP multicast group. */
1429 bzero(&in6, sizeof(in6));
1430 in6.s6_addr16[0] = htons(0xff02);
1431 in6.s6_addr8[15] = 0x12;
1432 if ((error = in6_setscope(&in6, ifp, NULL)) != 0) {
1433 ip6_mfilter_free(im6f[0]);
1434 ip6_mfilter_free(im6f[1]);
1437 if ((error = in6_joingroup(ifp, &in6, NULL, &im6f[0]->im6f_in6m, 0)) != 0) {
1438 ip6_mfilter_free(im6f[0]);
1439 ip6_mfilter_free(im6f[1]);
1443 /* Join solicited multicast address. */
1444 bzero(&in6, sizeof(in6));
1445 in6.s6_addr16[0] = htons(0xff02);
1446 in6.s6_addr32[1] = 0;
1447 in6.s6_addr32[2] = htonl(1);
1448 in6.s6_addr32[3] = 0;
1449 in6.s6_addr8[12] = 0xff;
1451 if ((error = in6_setscope(&in6, ifp, NULL)) != 0) {
1452 ip6_mfilter_free(im6f[0]);
1453 ip6_mfilter_free(im6f[1]);
1457 if ((error = in6_joingroup(ifp, &in6, NULL, &im6f[1]->im6f_in6m, 0)) != 0) {
1458 in6_leavegroup(im6f[0]->im6f_in6m, NULL);
1459 ip6_mfilter_free(im6f[0]);
1460 ip6_mfilter_free(im6f[1]);
1463 ip6_mfilter_insert(&im6o->im6o_head, im6f[0]);
1464 ip6_mfilter_insert(&im6o->im6o_head, im6f[1]);
1474 * Free multicast structures.
1477 carp_multicast_cleanup(struct carp_if *cif, sa_family_t sa)
1480 struct ip_moptions *imo = &cif->cif_imo;
1481 struct in_mfilter *imf;
1484 struct ip6_moptions *im6o = &cif->cif_im6o;
1485 struct in6_mfilter *im6f;
1487 sx_assert(&carp_sx, SA_XLOCKED);
1492 if (cif->cif_naddrs != 0)
1495 while ((imf = ip_mfilter_first(&imo->imo_head)) != NULL) {
1496 ip_mfilter_remove(&imo->imo_head, imf);
1497 in_leavegroup(imf->imf_inm, NULL);
1498 ip_mfilter_free(imf);
1504 if (cif->cif_naddrs6 != 0)
1507 while ((im6f = ip6_mfilter_first(&im6o->im6o_head)) != NULL) {
1508 ip6_mfilter_remove(&im6o->im6o_head, im6f);
1509 in6_leavegroup(im6f->im6f_in6m, NULL);
1510 ip6_mfilter_free(im6f);
1518 carp_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa)
1521 struct carp_softc *sc;
1526 switch (sa->sa_family) {
1539 mtag = m_tag_find(m, PACKET_TAG_CARP, NULL);
1543 bcopy(mtag + 1, &sc, sizeof(sc));
1545 /* Set the source MAC address to the Virtual Router MAC Address. */
1546 switch (ifp->if_type) {
1550 struct ether_header *eh;
1552 eh = mtod(m, struct ether_header *);
1553 eh->ether_shost[0] = 0;
1554 eh->ether_shost[1] = 0;
1555 eh->ether_shost[2] = 0x5e;
1556 eh->ether_shost[3] = 0;
1557 eh->ether_shost[4] = 1;
1558 eh->ether_shost[5] = sc->sc_vhid;
1562 printf("%s: carp is not supported for the %d interface type\n",
1563 ifp->if_xname, ifp->if_type);
1564 return (EOPNOTSUPP);
1570 static struct carp_softc*
1571 carp_alloc(struct ifnet *ifp)
1573 struct carp_softc *sc;
1574 struct carp_if *cif;
1576 sx_assert(&carp_sx, SA_XLOCKED);
1578 if ((cif = ifp->if_carp) == NULL)
1579 cif = carp_alloc_if(ifp);
1581 sc = malloc(sizeof(*sc), M_CARP, M_WAITOK|M_ZERO);
1583 sc->sc_advbase = CARP_DFLTINTV;
1584 sc->sc_vhid = -1; /* required setting */
1585 sc->sc_init_counter = 1;
1586 sc->sc_state = INIT;
1588 sc->sc_ifasiz = sizeof(struct ifaddr *);
1589 sc->sc_ifas = malloc(sc->sc_ifasiz, M_CARP, M_WAITOK|M_ZERO);
1590 sc->sc_carpdev = ifp;
1594 callout_init_mtx(&sc->sc_md_tmo, &sc->sc_mtx, CALLOUT_RETURNUNLOCKED);
1597 callout_init_mtx(&sc->sc_md6_tmo, &sc->sc_mtx, CALLOUT_RETURNUNLOCKED);
1599 callout_init_mtx(&sc->sc_ad_tmo, &sc->sc_mtx, CALLOUT_RETURNUNLOCKED);
1602 TAILQ_INSERT_TAIL(&cif->cif_vrs, sc, sc_list);
1605 mtx_lock(&carp_mtx);
1606 LIST_INSERT_HEAD(&carp_list, sc, sc_next);
1607 mtx_unlock(&carp_mtx);
1613 carp_grow_ifas(struct carp_softc *sc)
1615 struct ifaddr **new;
1617 new = malloc(sc->sc_ifasiz * 2, M_CARP, M_WAITOK | M_ZERO);
1619 bcopy(sc->sc_ifas, new, sc->sc_ifasiz);
1620 free(sc->sc_ifas, M_CARP);
1627 carp_destroy(struct carp_softc *sc)
1629 struct ifnet *ifp = sc->sc_carpdev;
1630 struct carp_if *cif = ifp->if_carp;
1632 sx_assert(&carp_sx, SA_XLOCKED);
1634 if (sc->sc_suppress)
1635 carp_demote_adj(-V_carp_ifdown_adj, "vhid removed");
1639 TAILQ_REMOVE(&cif->cif_vrs, sc, sc_list);
1642 mtx_lock(&carp_mtx);
1643 LIST_REMOVE(sc, sc_next);
1644 mtx_unlock(&carp_mtx);
1646 callout_drain(&sc->sc_ad_tmo);
1648 callout_drain(&sc->sc_md_tmo);
1651 callout_drain(&sc->sc_md6_tmo);
1653 CARP_LOCK_DESTROY(sc);
1655 free(sc->sc_ifas, M_CARP);
1659 static struct carp_if*
1660 carp_alloc_if(struct ifnet *ifp)
1662 struct carp_if *cif;
1665 cif = malloc(sizeof(*cif), M_CARP, M_WAITOK|M_ZERO);
1667 if ((error = ifpromisc(ifp, 1)) != 0)
1668 printf("%s: ifpromisc(%s) failed: %d\n",
1669 __func__, ifp->if_xname, error);
1671 cif->cif_flags |= CIF_PROMISC;
1675 TAILQ_INIT(&cif->cif_vrs);
1680 IF_ADDR_WUNLOCK(ifp);
1686 carp_free_if(struct carp_if *cif)
1688 struct ifnet *ifp = cif->cif_ifp;
1690 CIF_LOCK_ASSERT(cif);
1691 KASSERT(TAILQ_EMPTY(&cif->cif_vrs), ("%s: softc list not empty",
1695 ifp->if_carp = NULL;
1696 IF_ADDR_WUNLOCK(ifp);
1698 CIF_LOCK_DESTROY(cif);
1700 if (cif->cif_flags & CIF_PROMISC)
1708 carp_carprcp(struct carpreq *carpr, struct carp_softc *sc, int priv)
1712 carpr->carpr_state = sc->sc_state;
1713 carpr->carpr_vhid = sc->sc_vhid;
1714 carpr->carpr_advbase = sc->sc_advbase;
1715 carpr->carpr_advskew = sc->sc_advskew;
1717 bcopy(sc->sc_key, carpr->carpr_key, sizeof(carpr->carpr_key));
1719 bzero(carpr->carpr_key, sizeof(carpr->carpr_key));
1724 carp_ioctl(struct ifreq *ifr, u_long cmd, struct thread *td)
1726 struct carpreq carpr;
1728 struct carp_softc *sc = NULL;
1729 int error = 0, locked = 0;
1731 if ((error = copyin(ifr_data_get_ptr(ifr), &carpr, sizeof carpr)))
1734 ifp = ifunit_ref(ifr->ifr_name);
1738 switch (ifp->if_type) {
1748 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
1749 error = EADDRNOTAVAIL;
1756 if ((error = priv_check(td, PRIV_NETINET_CARP)))
1758 if (carpr.carpr_vhid <= 0 || carpr.carpr_vhid > CARP_MAXVHID ||
1759 carpr.carpr_advbase < 0 || carpr.carpr_advskew < 0) {
1765 IFNET_FOREACH_CARP(ifp, sc)
1766 if (sc->sc_vhid == carpr.carpr_vhid)
1770 sc = carp_alloc(ifp);
1772 sc->sc_vhid = carpr.carpr_vhid;
1773 LLADDR(&sc->sc_addr)[0] = 0;
1774 LLADDR(&sc->sc_addr)[1] = 0;
1775 LLADDR(&sc->sc_addr)[2] = 0x5e;
1776 LLADDR(&sc->sc_addr)[3] = 0;
1777 LLADDR(&sc->sc_addr)[4] = 1;
1778 LLADDR(&sc->sc_addr)[5] = sc->sc_vhid;
1782 if (carpr.carpr_advbase > 0) {
1783 if (carpr.carpr_advbase > 255 ||
1784 carpr.carpr_advbase < CARP_DFLTINTV) {
1788 sc->sc_advbase = carpr.carpr_advbase;
1790 if (carpr.carpr_advskew >= 255) {
1794 sc->sc_advskew = carpr.carpr_advskew;
1795 if (carpr.carpr_key[0] != '\0') {
1796 bcopy(carpr.carpr_key, sc->sc_key, sizeof(sc->sc_key));
1797 carp_hmac_prepare(sc);
1799 if (sc->sc_state != INIT &&
1800 carpr.carpr_state != sc->sc_state) {
1801 switch (carpr.carpr_state) {
1803 callout_stop(&sc->sc_ad_tmo);
1804 carp_set_state(sc, BACKUP,
1805 "user requested via ifconfig");
1810 carp_master_down_locked(sc,
1811 "user requested via ifconfig");
1823 if (carpr.carpr_vhid < 0 || carpr.carpr_vhid > CARP_MAXVHID) {
1827 if (carpr.carpr_count < 1) {
1831 if (ifp->if_carp == NULL) {
1836 priveleged = (priv_check(td, PRIV_NETINET_CARP) == 0);
1837 if (carpr.carpr_vhid != 0) {
1838 IFNET_FOREACH_CARP(ifp, sc)
1839 if (sc->sc_vhid == carpr.carpr_vhid)
1845 carp_carprcp(&carpr, sc, priveleged);
1846 error = copyout(&carpr, ifr_data_get_ptr(ifr),
1852 IFNET_FOREACH_CARP(ifp, sc)
1855 if (count > carpr.carpr_count) {
1856 CIF_UNLOCK(ifp->if_carp);
1862 IFNET_FOREACH_CARP(ifp, sc) {
1863 carp_carprcp(&carpr, sc, priveleged);
1864 carpr.carpr_count = count;
1865 error = copyout(&carpr,
1866 (char *)ifr_data_get_ptr(ifr) +
1867 (i * sizeof(carpr)), sizeof(carpr));
1869 CIF_UNLOCK(ifp->if_carp);
1880 sx_xunlock(&carp_sx);
1891 carp_get_vhid(struct ifaddr *ifa)
1894 if (ifa == NULL || ifa->ifa_carp == NULL)
1897 return (ifa->ifa_carp->sc_vhid);
1901 carp_attach(struct ifaddr *ifa, int vhid)
1903 struct ifnet *ifp = ifa->ifa_ifp;
1904 struct carp_if *cif = ifp->if_carp;
1905 struct carp_softc *sc;
1908 KASSERT(ifa->ifa_carp == NULL, ("%s: ifa %p attached", __func__, ifa));
1910 switch (ifa->ifa_addr->sa_family) {
1919 return (EPROTOTYPE);
1923 if (ifp->if_carp == NULL) {
1924 sx_xunlock(&carp_sx);
1925 return (ENOPROTOOPT);
1928 IFNET_FOREACH_CARP(ifp, sc)
1929 if (sc->sc_vhid == vhid)
1932 sx_xunlock(&carp_sx);
1936 error = carp_multicast_setup(cif, ifa->ifa_addr->sa_family);
1939 sx_xunlock(&carp_sx);
1943 index = sc->sc_naddrs + sc->sc_naddrs6 + 1;
1944 if (index > sc->sc_ifasiz / sizeof(struct ifaddr *))
1947 switch (ifa->ifa_addr->sa_family) {
1965 sc->sc_ifas[index - 1] = ifa;
1967 carp_hmac_prepare(sc);
1971 sx_xunlock(&carp_sx);
1977 carp_detach(struct ifaddr *ifa, bool keep_cif)
1979 struct ifnet *ifp = ifa->ifa_ifp;
1980 struct carp_if *cif = ifp->if_carp;
1981 struct carp_softc *sc = ifa->ifa_carp;
1984 KASSERT(sc != NULL, ("%s: %p not attached", __func__, ifa));
1990 index = sc->sc_naddrs + sc->sc_naddrs6;
1991 for (i = 0; i < index; i++)
1992 if (sc->sc_ifas[i] == ifa)
1994 KASSERT(i < index, ("%s: %p no backref", __func__, ifa));
1995 for (; i < index - 1; i++)
1996 sc->sc_ifas[i] = sc->sc_ifas[i+1];
1997 sc->sc_ifas[index - 1] = NULL;
1999 switch (ifa->ifa_addr->sa_family) {
2014 carp_ifa_delroute(ifa);
2015 carp_multicast_cleanup(cif, ifa->ifa_addr->sa_family);
2017 ifa->ifa_carp = NULL;
2020 carp_hmac_prepare(sc);
2023 if (!keep_cif && sc->sc_naddrs == 0 && sc->sc_naddrs6 == 0)
2031 sx_xunlock(&carp_sx);
2035 carp_set_state(struct carp_softc *sc, int state, const char *reason)
2038 CARP_LOCK_ASSERT(sc);
2040 if (sc->sc_state != state) {
2041 const char *carp_states[] = { CARP_STATES };
2042 char subsys[IFNAMSIZ+5];
2044 snprintf(subsys, IFNAMSIZ+5, "%u@%s", sc->sc_vhid,
2045 sc->sc_carpdev->if_xname);
2047 CARP_LOG("%s: %s -> %s (%s)\n", subsys,
2048 carp_states[sc->sc_state], carp_states[state], reason);
2050 sc->sc_state = state;
2052 devctl_notify("CARP", subsys, carp_states[state], NULL);
2057 carp_linkstate(struct ifnet *ifp)
2059 struct carp_softc *sc;
2061 CIF_LOCK(ifp->if_carp);
2062 IFNET_FOREACH_CARP(ifp, sc) {
2067 CIF_UNLOCK(ifp->if_carp);
2071 carp_sc_state(struct carp_softc *sc)
2074 CARP_LOCK_ASSERT(sc);
2076 if (sc->sc_carpdev->if_link_state != LINK_STATE_UP ||
2077 !(sc->sc_carpdev->if_flags & IFF_UP) ||
2079 callout_stop(&sc->sc_ad_tmo);
2081 callout_stop(&sc->sc_md_tmo);
2084 callout_stop(&sc->sc_md6_tmo);
2086 carp_set_state(sc, INIT, "hardware interface down");
2088 if (!sc->sc_suppress)
2089 carp_demote_adj(V_carp_ifdown_adj, "interface down");
2090 sc->sc_suppress = 1;
2092 carp_set_state(sc, INIT, "hardware interface up");
2094 if (sc->sc_suppress)
2095 carp_demote_adj(-V_carp_ifdown_adj, "interface up");
2096 sc->sc_suppress = 0;
2101 carp_demote_adj(int adj, char *reason)
2103 atomic_add_int(&V_carp_demotion, adj);
2104 CARP_LOG("demoted by %d to %d (%s)\n", adj, V_carp_demotion, reason);
2105 taskqueue_enqueue(taskqueue_swi, &carp_sendall_task);
2109 carp_allow_sysctl(SYSCTL_HANDLER_ARGS)
2112 struct carp_softc *sc;
2115 error = sysctl_handle_int(oidp, &new, 0, req);
2116 if (error || !req->newptr)
2119 if (V_carp_allow != new) {
2122 mtx_lock(&carp_mtx);
2123 LIST_FOREACH(sc, &carp_list, sc_next) {
2125 if (curvnet == sc->sc_carpdev->if_vnet)
2129 mtx_unlock(&carp_mtx);
2136 carp_dscp_sysctl(SYSCTL_HANDLER_ARGS)
2141 error = sysctl_handle_int(oidp, &new, 0, req);
2142 if (error || !req->newptr)
2145 if (new < 0 || new > 63)
2154 carp_demote_adj_sysctl(SYSCTL_HANDLER_ARGS)
2158 new = V_carp_demotion;
2159 error = sysctl_handle_int(oidp, &new, 0, req);
2160 if (error || !req->newptr)
2163 carp_demote_adj(new, "sysctl");
2169 extern struct domain inetdomain;
2170 static struct protosw in_carp_protosw = {
2171 .pr_type = SOCK_RAW,
2172 .pr_domain = &inetdomain,
2173 .pr_protocol = IPPROTO_CARP,
2174 .pr_flags = PR_ATOMIC|PR_ADDR,
2175 .pr_input = carp_input,
2176 .pr_output = rip_output,
2177 .pr_ctloutput = rip_ctloutput,
2178 .pr_usrreqs = &rip_usrreqs
2183 extern struct domain inet6domain;
2184 static struct protosw in6_carp_protosw = {
2185 .pr_type = SOCK_RAW,
2186 .pr_domain = &inet6domain,
2187 .pr_protocol = IPPROTO_CARP,
2188 .pr_flags = PR_ATOMIC|PR_ADDR,
2189 .pr_input = carp6_input,
2190 .pr_output = rip6_output,
2191 .pr_ctloutput = rip6_ctloutput,
2192 .pr_usrreqs = &rip6_usrreqs
2197 carp_mod_cleanup(void)
2201 if (proto_reg[CARP_INET] == 0) {
2202 (void)ipproto_unregister(IPPROTO_CARP);
2203 pf_proto_unregister(PF_INET, IPPROTO_CARP, SOCK_RAW);
2204 proto_reg[CARP_INET] = -1;
2206 carp_iamatch_p = NULL;
2209 if (proto_reg[CARP_INET6] == 0) {
2210 (void)ip6proto_unregister(IPPROTO_CARP);
2211 pf_proto_unregister(PF_INET6, IPPROTO_CARP, SOCK_RAW);
2212 proto_reg[CARP_INET6] = -1;
2214 carp_iamatch6_p = NULL;
2215 carp_macmatch6_p = NULL;
2217 carp_ioctl_p = NULL;
2218 carp_attach_p = NULL;
2219 carp_detach_p = NULL;
2220 carp_get_vhid_p = NULL;
2221 carp_linkstate_p = NULL;
2222 carp_forus_p = NULL;
2223 carp_output_p = NULL;
2224 carp_demote_adj_p = NULL;
2225 carp_master_p = NULL;
2226 mtx_unlock(&carp_mtx);
2227 taskqueue_drain(taskqueue_swi, &carp_sendall_task);
2228 mtx_destroy(&carp_mtx);
2229 sx_destroy(&carp_sx);
2237 mtx_init(&carp_mtx, "carp_mtx", NULL, MTX_DEF);
2238 sx_init(&carp_sx, "carp_sx");
2239 LIST_INIT(&carp_list);
2240 carp_get_vhid_p = carp_get_vhid;
2241 carp_forus_p = carp_forus;
2242 carp_output_p = carp_output;
2243 carp_linkstate_p = carp_linkstate;
2244 carp_ioctl_p = carp_ioctl;
2245 carp_attach_p = carp_attach;
2246 carp_detach_p = carp_detach;
2247 carp_demote_adj_p = carp_demote_adj;
2248 carp_master_p = carp_master;
2250 carp_iamatch6_p = carp_iamatch6;
2251 carp_macmatch6_p = carp_macmatch6;
2252 proto_reg[CARP_INET6] = pf_proto_register(PF_INET6,
2253 (struct protosw *)&in6_carp_protosw);
2254 if (proto_reg[CARP_INET6]) {
2255 printf("carp: error %d attaching to PF_INET6\n",
2256 proto_reg[CARP_INET6]);
2258 return (proto_reg[CARP_INET6]);
2260 err = ip6proto_register(IPPROTO_CARP);
2262 printf("carp: error %d registering with INET6\n", err);
2268 carp_iamatch_p = carp_iamatch;
2269 proto_reg[CARP_INET] = pf_proto_register(PF_INET, &in_carp_protosw);
2270 if (proto_reg[CARP_INET]) {
2271 printf("carp: error %d attaching to PF_INET\n",
2272 proto_reg[CARP_INET]);
2274 return (proto_reg[CARP_INET]);
2276 err = ipproto_register(IPPROTO_CARP);
2278 printf("carp: error %d registering with INET\n", err);
2287 carp_modevent(module_t mod, int type, void *data)
2291 return carp_mod_load();
2294 mtx_lock(&carp_mtx);
2295 if (LIST_EMPTY(&carp_list))
2298 mtx_unlock(&carp_mtx);
2310 static moduledata_t carp_mod = {
2316 DECLARE_MODULE(carp, carp_mod, SI_SUB_PROTO_DOMAIN, SI_ORDER_ANY);