1 /* $FreeBSD$ */ 2 3 /* 4 * Copyright (c) 2002 Michael Shalayeff. All rights reserved. 5 * Copyright (c) 2003 Ryan McBride. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 16 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 17 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 18 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 19 * IN NO EVENT SHALL THE AUTHOR OR HIS RELATIVES BE LIABLE FOR ANY DIRECT, 20 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES 21 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 22 * SERVICES; LOSS OF MIND, USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, 24 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING 25 * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF 26 * THE POSSIBILITY OF SUCH DAMAGE. 27 */ 28 29 #include "opt_carp.h" 30 #include "opt_bpf.h" 31 #include "opt_inet.h" 32 #include "opt_inet6.h" 33 34 #include <sys/types.h> 35 #include <sys/param.h> 36 #include <sys/systm.h> 37 #include <sys/conf.h> 38 #include <sys/kernel.h> 39 #include <sys/limits.h> 40 #include <sys/malloc.h> 41 #include <sys/mbuf.h> 42 #include <sys/module.h> 43 #include <sys/time.h> 44 #include <sys/proc.h> 45 #include <sys/sysctl.h> 46 #include <sys/syslog.h> 47 #include <sys/signalvar.h> 48 #include <sys/filio.h> 49 #include <sys/sockio.h> 50 51 #include <sys/socket.h> 52 #include <sys/vnode.h> 53 54 #include <machine/stdarg.h> 55 56 #include <net/bpf.h> 57 #include <net/ethernet.h> 58 #include <net/fddi.h> 59 #include <net/iso88025.h> 60 #include <net/if.h> 61 #include <net/if_clone.h> 62 #include <net/if_dl.h> 63 #include <net/if_types.h> 64 #include <net/route.h> 65 66 #ifdef INET 67 #include <netinet/in.h> 68 #include <netinet/in_var.h> 69 #include <netinet/in_systm.h> 70 #include <netinet/ip.h> 71 #include <netinet/ip_var.h> 72 #include <netinet/if_ether.h> 73 #include <machine/in_cksum.h> 74 #endif 75 76 #ifdef INET6 77 #include <netinet/icmp6.h> 78 #include <netinet/ip6.h> 79 #include <netinet6/ip6_var.h> 80 #include <netinet6/scope6_var.h> 81 #include <netinet6/nd6.h> 82 #endif 83 84 #include <crypto/sha1.h> 85 #include <netinet/ip_carp.h> 86 87 #define CARP_IFNAME "carp" 88 static MALLOC_DEFINE(M_CARP, "CARP", "CARP interfaces"); 89 SYSCTL_DECL(_net_inet_carp); 90 91 struct carp_softc { 92 struct ifnet *sc_ifp; /* Interface clue */ 93 struct ifnet *sc_carpdev; /* Pointer to parent interface */ 94 struct in_ifaddr *sc_ia; /* primary iface address */ 95 struct ip_moptions sc_imo; 96 #ifdef INET6 97 struct in6_ifaddr *sc_ia6; /* primary iface address v6 */ 98 struct ip6_moptions sc_im6o; 99 #endif /* INET6 */ 100 TAILQ_ENTRY(carp_softc) sc_list; 101 102 enum { INIT = 0, BACKUP, MASTER } sc_state; 103 104 int sc_flags_backup; 105 int sc_suppress; 106 107 int sc_sendad_errors; 108 #define CARP_SENDAD_MAX_ERRORS 3 109 int sc_sendad_success; 110 #define CARP_SENDAD_MIN_SUCCESS 3 111 112 int sc_vhid; 113 int sc_advskew; 114 int sc_naddrs; 115 int sc_naddrs6; 116 int sc_advbase; /* seconds */ 117 int sc_init_counter; 118 u_int64_t sc_counter; 119 120 /* authentication */ 121 #define CARP_HMAC_PAD 64 122 unsigned char sc_key[CARP_KEY_LEN]; 123 unsigned char sc_pad[CARP_HMAC_PAD]; 124 SHA1_CTX sc_sha1; 125 126 struct callout sc_ad_tmo; /* advertisement timeout */ 127 struct callout sc_md_tmo; /* master down timeout */ 128 struct callout sc_md6_tmo; /* master down timeout */ 129 130 LIST_ENTRY(carp_softc) sc_next; /* Interface clue */ 131 }; 132 #define SC2IFP(sc) ((sc)->sc_ifp) 133 134 int carp_suppress_preempt = 0; 135 int carp_opts[CARPCTL_MAXID] = { 0, 1, 0, 1, 0, 0 }; /* XXX for now */ 136 SYSCTL_INT(_net_inet_carp, CARPCTL_ALLOW, allow, CTLFLAG_RW, 137 &carp_opts[CARPCTL_ALLOW], 0, "Accept incoming CARP packets"); 138 SYSCTL_INT(_net_inet_carp, CARPCTL_PREEMPT, preempt, CTLFLAG_RW, 139 &carp_opts[CARPCTL_PREEMPT], 0, "high-priority backup preemption mode"); 140 SYSCTL_INT(_net_inet_carp, CARPCTL_LOG, log, CTLFLAG_RW, 141 &carp_opts[CARPCTL_LOG], 0, "log bad carp packets"); 142 SYSCTL_INT(_net_inet_carp, CARPCTL_ARPBALANCE, arpbalance, CTLFLAG_RW, 143 &carp_opts[CARPCTL_ARPBALANCE], 0, "balance arp responses"); 144 SYSCTL_INT(_net_inet_carp, OID_AUTO, suppress_preempt, CTLFLAG_RD, 145 &carp_suppress_preempt, 0, "Preemption is suppressed"); 146 147 struct carpstats carpstats; 148 SYSCTL_STRUCT(_net_inet_carp, CARPCTL_STATS, stats, CTLFLAG_RW, 149 &carpstats, carpstats, 150 "CARP statistics (struct carpstats, netinet/ip_carp.h)"); 151 152 struct carp_if { 153 TAILQ_HEAD(, carp_softc) vhif_vrs; 154 int vhif_nvrs; 155 156 struct ifnet *vhif_ifp; 157 struct mtx vhif_mtx; 158 }; 159 160 /* Get carp_if from softc. Valid after carp_set_addr{,6}. */ 161 #define SC2CIF(sc) ((struct carp_if *)(sc)->sc_carpdev->if_carp) 162 163 /* lock per carp_if queue */ 164 #define CARP_LOCK_INIT(cif) mtx_init(&(cif)->vhif_mtx, "carp_if", \ 165 NULL, MTX_DEF) 166 #define CARP_LOCK_DESTROY(cif) mtx_destroy(&(cif)->vhif_mtx) 167 #define CARP_LOCK_ASSERT(cif) mtx_assert(&(cif)->vhif_mtx, MA_OWNED) 168 #define CARP_LOCK(cif) mtx_lock(&(cif)->vhif_mtx) 169 #define CARP_UNLOCK(cif) mtx_unlock(&(cif)->vhif_mtx) 170 171 #define CARP_SCLOCK(sc) mtx_lock(&SC2CIF(sc)->vhif_mtx) 172 #define CARP_SCUNLOCK(sc) mtx_unlock(&SC2CIF(sc)->vhif_mtx) 173 #define CARP_SCLOCK_ASSERT(sc) mtx_assert(&SC2CIF(sc)->vhif_mtx, MA_OWNED) 174 175 #define CARP_LOG(...) do { \ 176 if (carp_opts[CARPCTL_LOG] > 0) \ 177 log(LOG_INFO, __VA_ARGS__); \ 178 } while (0) 179 180 #define CARP_DEBUG(...) do { \ 181 if (carp_opts[CARPCTL_LOG] > 1) \ 182 log(LOG_DEBUG, __VA_ARGS__); \ 183 } while (0) 184 185 static void carp_hmac_prepare(struct carp_softc *); 186 static void carp_hmac_generate(struct carp_softc *, u_int32_t *, 187 unsigned char *); 188 static int carp_hmac_verify(struct carp_softc *, u_int32_t *, 189 unsigned char *); 190 static void carp_setroute(struct carp_softc *, int); 191 static void carp_input_c(struct mbuf *, struct carp_header *, sa_family_t); 192 static int carp_clone_create(struct if_clone *, int, caddr_t); 193 static void carp_clone_destroy(struct ifnet *); 194 static void carpdetach(struct carp_softc *); 195 static int carp_prepare_ad(struct mbuf *, struct carp_softc *, 196 struct carp_header *); 197 static void carp_send_ad_all(void); 198 static void carp_send_ad(void *); 199 static void carp_send_ad_locked(struct carp_softc *); 200 static void carp_send_arp(struct carp_softc *); 201 static void carp_master_down(void *); 202 static void carp_master_down_locked(struct carp_softc *); 203 static int carp_ioctl(struct ifnet *, u_long, caddr_t); 204 static int carp_looutput(struct ifnet *, struct mbuf *, struct sockaddr *, 205 struct rtentry *); 206 static void carp_start(struct ifnet *); 207 static void carp_setrun(struct carp_softc *, sa_family_t); 208 static void carp_set_state(struct carp_softc *, int); 209 static int carp_addrcount(struct carp_if *, struct in_ifaddr *, int); 210 enum { CARP_COUNT_MASTER, CARP_COUNT_RUNNING }; 211 212 static void carp_multicast_cleanup(struct carp_softc *); 213 static int carp_set_addr(struct carp_softc *, struct sockaddr_in *); 214 static int carp_del_addr(struct carp_softc *, struct sockaddr_in *); 215 static void carp_carpdev_state_locked(struct carp_if *); 216 static void carp_sc_state_locked(struct carp_softc *); 217 #ifdef INET6 218 static void carp_send_na(struct carp_softc *); 219 static int carp_set_addr6(struct carp_softc *, struct sockaddr_in6 *); 220 static int carp_del_addr6(struct carp_softc *, struct sockaddr_in6 *); 221 #endif 222 223 static LIST_HEAD(, carp_softc) carpif_list; 224 static struct mtx carp_mtx; 225 IFC_SIMPLE_DECLARE(carp, 0); 226 227 static eventhandler_tag if_detach_event_tag; 228 229 static __inline u_int16_t 230 carp_cksum(struct mbuf *m, int len) 231 { 232 return (in_cksum(m, len)); 233 } 234 235 static void 236 carp_hmac_prepare(struct carp_softc *sc) 237 { 238 u_int8_t version = CARP_VERSION, type = CARP_ADVERTISEMENT; 239 u_int8_t vhid = sc->sc_vhid & 0xff; 240 struct ifaddr *ifa; 241 int i; 242 #ifdef INET6 243 struct in6_addr in6; 244 #endif 245 246 if (sc->sc_carpdev) 247 CARP_SCLOCK(sc); 248 249 /* XXX: possible race here */ 250 251 /* compute ipad from key */ 252 bzero(sc->sc_pad, sizeof(sc->sc_pad)); 253 bcopy(sc->sc_key, sc->sc_pad, sizeof(sc->sc_key)); 254 for (i = 0; i < sizeof(sc->sc_pad); i++) 255 sc->sc_pad[i] ^= 0x36; 256 257 /* precompute first part of inner hash */ 258 SHA1Init(&sc->sc_sha1); 259 SHA1Update(&sc->sc_sha1, sc->sc_pad, sizeof(sc->sc_pad)); 260 SHA1Update(&sc->sc_sha1, (void *)&version, sizeof(version)); 261 SHA1Update(&sc->sc_sha1, (void *)&type, sizeof(type)); 262 SHA1Update(&sc->sc_sha1, (void *)&vhid, sizeof(vhid)); 263 #ifdef INET 264 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) { 265 if (ifa->ifa_addr->sa_family == AF_INET) 266 SHA1Update(&sc->sc_sha1, 267 (void *)&ifatoia(ifa)->ia_addr.sin_addr.s_addr, 268 sizeof(struct in_addr)); 269 } 270 #endif /* INET */ 271 #ifdef INET6 272 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) { 273 if (ifa->ifa_addr->sa_family == AF_INET6) { 274 in6 = ifatoia6(ifa)->ia_addr.sin6_addr; 275 in6_clearscope(&in6); 276 SHA1Update(&sc->sc_sha1, (void *)&in6, sizeof(in6)); 277 } 278 } 279 #endif /* INET6 */ 280 281 /* convert ipad to opad */ 282 for (i = 0; i < sizeof(sc->sc_pad); i++) 283 sc->sc_pad[i] ^= 0x36 ^ 0x5c; 284 285 if (sc->sc_carpdev) 286 CARP_SCUNLOCK(sc); 287 } 288 289 static void 290 carp_hmac_generate(struct carp_softc *sc, u_int32_t counter[2], 291 unsigned char md[20]) 292 { 293 SHA1_CTX sha1ctx; 294 295 /* fetch first half of inner hash */ 296 bcopy(&sc->sc_sha1, &sha1ctx, sizeof(sha1ctx)); 297 298 SHA1Update(&sha1ctx, (void *)counter, sizeof(sc->sc_counter)); 299 SHA1Final(md, &sha1ctx); 300 301 /* outer hash */ 302 SHA1Init(&sha1ctx); 303 SHA1Update(&sha1ctx, sc->sc_pad, sizeof(sc->sc_pad)); 304 SHA1Update(&sha1ctx, md, 20); 305 SHA1Final(md, &sha1ctx); 306 } 307 308 static int 309 carp_hmac_verify(struct carp_softc *sc, u_int32_t counter[2], 310 unsigned char md[20]) 311 { 312 unsigned char md2[20]; 313 314 CARP_SCLOCK_ASSERT(sc); 315 316 carp_hmac_generate(sc, counter, md2); 317 318 return (bcmp(md, md2, sizeof(md2))); 319 } 320 321 static void 322 carp_setroute(struct carp_softc *sc, int cmd) 323 { 324 struct ifaddr *ifa; 325 int s; 326 327 if (sc->sc_carpdev) 328 CARP_SCLOCK_ASSERT(sc); 329 330 s = splnet(); 331 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) { 332 if (ifa->ifa_addr->sa_family == AF_INET && 333 sc->sc_carpdev != NULL) { 334 int count = carp_addrcount( 335 (struct carp_if *)sc->sc_carpdev->if_carp, 336 ifatoia(ifa), CARP_COUNT_MASTER); 337 338 if ((cmd == RTM_ADD && count == 1) || 339 (cmd == RTM_DELETE && count == 0)) 340 rtinit(ifa, cmd, RTF_UP | RTF_HOST); 341 } 342 #ifdef INET6 343 if (ifa->ifa_addr->sa_family == AF_INET6) { 344 if (cmd == RTM_ADD) 345 in6_ifaddloop(ifa); 346 else 347 in6_ifremloop(ifa); 348 } 349 #endif /* INET6 */ 350 } 351 splx(s); 352 } 353 354 static int 355 carp_clone_create(struct if_clone *ifc, int unit, caddr_t params) 356 { 357 358 struct carp_softc *sc; 359 struct ifnet *ifp; 360 361 MALLOC(sc, struct carp_softc *, sizeof(*sc), M_CARP, M_WAITOK|M_ZERO); 362 ifp = SC2IFP(sc) = if_alloc(IFT_ETHER); 363 if (ifp == NULL) { 364 FREE(sc, M_CARP); 365 return (ENOSPC); 366 } 367 368 sc->sc_flags_backup = 0; 369 sc->sc_suppress = 0; 370 sc->sc_advbase = CARP_DFLTINTV; 371 sc->sc_vhid = -1; /* required setting */ 372 sc->sc_advskew = 0; 373 sc->sc_init_counter = 1; 374 sc->sc_naddrs = sc->sc_naddrs6 = 0; /* M_ZERO? */ 375 #ifdef INET6 376 sc->sc_im6o.im6o_multicast_hlim = CARP_DFLTTL; 377 #endif 378 sc->sc_imo.imo_membership = (struct in_multi **)malloc( 379 (sizeof(struct in_multi *) * IP_MIN_MEMBERSHIPS), M_CARP, 380 M_WAITOK); 381 sc->sc_imo.imo_max_memberships = IP_MIN_MEMBERSHIPS; 382 sc->sc_imo.imo_multicast_vif = -1; 383 384 callout_init(&sc->sc_ad_tmo, NET_CALLOUT_MPSAFE); 385 callout_init(&sc->sc_md_tmo, NET_CALLOUT_MPSAFE); 386 callout_init(&sc->sc_md6_tmo, NET_CALLOUT_MPSAFE); 387 388 ifp->if_softc = sc; 389 if_initname(ifp, CARP_IFNAME, unit); 390 ifp->if_mtu = ETHERMTU; 391 ifp->if_flags = IFF_LOOPBACK; 392 ifp->if_ioctl = carp_ioctl; 393 ifp->if_output = carp_looutput; 394 ifp->if_start = carp_start; 395 ifp->if_type = IFT_CARP; 396 ifp->if_snd.ifq_maxlen = ifqmaxlen; 397 ifp->if_hdrlen = 0; 398 if_attach(ifp); 399 bpfattach(SC2IFP(sc), DLT_NULL, sizeof(u_int32_t)); 400 mtx_lock(&carp_mtx); 401 LIST_INSERT_HEAD(&carpif_list, sc, sc_next); 402 mtx_unlock(&carp_mtx); 403 return (0); 404 } 405 406 static void 407 carp_clone_destroy(struct ifnet *ifp) 408 { 409 struct carp_softc *sc = ifp->if_softc; 410 411 if (sc->sc_carpdev) 412 CARP_SCLOCK(sc); 413 carpdetach(sc); 414 if (sc->sc_carpdev) 415 CARP_SCUNLOCK(sc); 416 417 mtx_lock(&carp_mtx); 418 LIST_REMOVE(sc, sc_next); 419 mtx_unlock(&carp_mtx); 420 bpfdetach(ifp); 421 if_detach(ifp); 422 if_free_type(ifp, IFT_ETHER); 423 free(sc->sc_imo.imo_membership, M_CARP); 424 free(sc, M_CARP); 425 } 426 427 static void 428 carpdetach(struct carp_softc *sc) 429 { 430 struct carp_if *cif; 431 432 callout_stop(&sc->sc_ad_tmo); 433 callout_stop(&sc->sc_md_tmo); 434 callout_stop(&sc->sc_md6_tmo); 435 436 if (sc->sc_suppress) 437 carp_suppress_preempt--; 438 sc->sc_suppress = 0; 439 440 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) 441 carp_suppress_preempt--; 442 sc->sc_sendad_errors = 0; 443 444 carp_set_state(sc, INIT); 445 SC2IFP(sc)->if_flags &= ~IFF_UP; 446 carp_setrun(sc, 0); 447 carp_multicast_cleanup(sc); 448 449 if (sc->sc_carpdev != NULL) { 450 cif = (struct carp_if *)sc->sc_carpdev->if_carp; 451 CARP_LOCK_ASSERT(cif); 452 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list); 453 if (!--cif->vhif_nvrs) { 454 ifpromisc(sc->sc_carpdev, 0); 455 sc->sc_carpdev->if_carp = NULL; 456 CARP_LOCK_DESTROY(cif); 457 FREE(cif, M_IFADDR); 458 } 459 } 460 sc->sc_carpdev = NULL; 461 } 462 463 /* Detach an interface from the carp. */ 464 static void 465 carp_ifdetach(void *arg __unused, struct ifnet *ifp) 466 { 467 struct carp_if *cif = (struct carp_if *)ifp->if_carp; 468 struct carp_softc *sc, *nextsc; 469 470 if (cif == NULL) 471 return; 472 473 /* 474 * XXX: At the end of for() cycle the lock will be destroyed. 475 */ 476 CARP_LOCK(cif); 477 for (sc = TAILQ_FIRST(&cif->vhif_vrs); sc; sc = nextsc) { 478 nextsc = TAILQ_NEXT(sc, sc_list); 479 carpdetach(sc); 480 } 481 } 482 483 /* 484 * process input packet. 485 * we have rearranged checks order compared to the rfc, 486 * but it seems more efficient this way or not possible otherwise. 487 */ 488 void 489 carp_input(struct mbuf *m, int hlen) 490 { 491 struct ip *ip = mtod(m, struct ip *); 492 struct carp_header *ch; 493 int iplen, len; 494 495 carpstats.carps_ipackets++; 496 497 if (!carp_opts[CARPCTL_ALLOW]) { 498 m_freem(m); 499 return; 500 } 501 502 /* check if received on a valid carp interface */ 503 if (m->m_pkthdr.rcvif->if_carp == NULL) { 504 carpstats.carps_badif++; 505 CARP_LOG("carp_input: packet received on non-carp " 506 "interface: %s\n", 507 m->m_pkthdr.rcvif->if_xname); 508 m_freem(m); 509 return; 510 } 511 512 /* verify that the IP TTL is 255. */ 513 if (ip->ip_ttl != CARP_DFLTTL) { 514 carpstats.carps_badttl++; 515 CARP_LOG("carp_input: received ttl %d != 255i on %s\n", 516 ip->ip_ttl, 517 m->m_pkthdr.rcvif->if_xname); 518 m_freem(m); 519 return; 520 } 521 522 iplen = ip->ip_hl << 2; 523 524 if (m->m_pkthdr.len < iplen + sizeof(*ch)) { 525 carpstats.carps_badlen++; 526 CARP_LOG("carp_input: received len %zd < " 527 "sizeof(struct carp_header)\n", 528 m->m_len - sizeof(struct ip)); 529 m_freem(m); 530 return; 531 } 532 533 if (iplen + sizeof(*ch) < m->m_len) { 534 if ((m = m_pullup(m, iplen + sizeof(*ch))) == NULL) { 535 carpstats.carps_hdrops++; 536 CARP_LOG("carp_input: pullup failed\n"); 537 return; 538 } 539 ip = mtod(m, struct ip *); 540 } 541 ch = (struct carp_header *)((char *)ip + iplen); 542 543 /* 544 * verify that the received packet length is 545 * equal to the CARP header 546 */ 547 len = iplen + sizeof(*ch); 548 if (len > m->m_pkthdr.len) { 549 carpstats.carps_badlen++; 550 CARP_LOG("carp_input: packet too short %d on %s\n", 551 m->m_pkthdr.len, 552 m->m_pkthdr.rcvif->if_xname); 553 m_freem(m); 554 return; 555 } 556 557 if ((m = m_pullup(m, len)) == NULL) { 558 carpstats.carps_hdrops++; 559 return; 560 } 561 ip = mtod(m, struct ip *); 562 ch = (struct carp_header *)((char *)ip + iplen); 563 564 /* verify the CARP checksum */ 565 m->m_data += iplen; 566 if (carp_cksum(m, len - iplen)) { 567 carpstats.carps_badsum++; 568 CARP_LOG("carp_input: checksum failed on %s\n", 569 m->m_pkthdr.rcvif->if_xname); 570 m_freem(m); 571 return; 572 } 573 m->m_data -= iplen; 574 575 carp_input_c(m, ch, AF_INET); 576 } 577 578 #ifdef INET6 579 int 580 carp6_input(struct mbuf **mp, int *offp, int proto) 581 { 582 struct mbuf *m = *mp; 583 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 584 struct carp_header *ch; 585 u_int len; 586 587 carpstats.carps_ipackets6++; 588 589 if (!carp_opts[CARPCTL_ALLOW]) { 590 m_freem(m); 591 return (IPPROTO_DONE); 592 } 593 594 /* check if received on a valid carp interface */ 595 if (m->m_pkthdr.rcvif->if_carp == NULL) { 596 carpstats.carps_badif++; 597 CARP_LOG("carp6_input: packet received on non-carp " 598 "interface: %s\n", 599 m->m_pkthdr.rcvif->if_xname); 600 m_freem(m); 601 return (IPPROTO_DONE); 602 } 603 604 /* verify that the IP TTL is 255 */ 605 if (ip6->ip6_hlim != CARP_DFLTTL) { 606 carpstats.carps_badttl++; 607 CARP_LOG("carp6_input: received ttl %d != 255 on %s\n", 608 ip6->ip6_hlim, 609 m->m_pkthdr.rcvif->if_xname); 610 m_freem(m); 611 return (IPPROTO_DONE); 612 } 613 614 /* verify that we have a complete carp packet */ 615 len = m->m_len; 616 IP6_EXTHDR_GET(ch, struct carp_header *, m, *offp, sizeof(*ch)); 617 if (ch == NULL) { 618 carpstats.carps_badlen++; 619 CARP_LOG("carp6_input: packet size %u too small\n", len); 620 return (IPPROTO_DONE); 621 } 622 623 624 /* verify the CARP checksum */ 625 m->m_data += *offp; 626 if (carp_cksum(m, sizeof(*ch))) { 627 carpstats.carps_badsum++; 628 CARP_LOG("carp6_input: checksum failed, on %s\n", 629 m->m_pkthdr.rcvif->if_xname); 630 m_freem(m); 631 return (IPPROTO_DONE); 632 } 633 m->m_data -= *offp; 634 635 carp_input_c(m, ch, AF_INET6); 636 return (IPPROTO_DONE); 637 } 638 #endif /* INET6 */ 639 640 static void 641 carp_input_c(struct mbuf *m, struct carp_header *ch, sa_family_t af) 642 { 643 struct ifnet *ifp = m->m_pkthdr.rcvif; 644 struct carp_softc *sc; 645 u_int64_t tmp_counter; 646 struct timeval sc_tv, ch_tv; 647 648 /* verify that the VHID is valid on the receiving interface */ 649 CARP_LOCK(ifp->if_carp); 650 TAILQ_FOREACH(sc, &((struct carp_if *)ifp->if_carp)->vhif_vrs, sc_list) 651 if (sc->sc_vhid == ch->carp_vhid) 652 break; 653 654 if (!sc || !((SC2IFP(sc)->if_flags & IFF_UP) && 655 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING))) { 656 carpstats.carps_badvhid++; 657 CARP_UNLOCK(ifp->if_carp); 658 m_freem(m); 659 return; 660 } 661 662 getmicrotime(&SC2IFP(sc)->if_lastchange); 663 SC2IFP(sc)->if_ipackets++; 664 SC2IFP(sc)->if_ibytes += m->m_pkthdr.len; 665 666 if (bpf_peers_present(SC2IFP(sc)->if_bpf)) { 667 struct ip *ip = mtod(m, struct ip *); 668 uint32_t af1 = af; 669 670 /* BPF wants net byte order */ 671 ip->ip_len = htons(ip->ip_len + (ip->ip_hl << 2)); 672 ip->ip_off = htons(ip->ip_off); 673 bpf_mtap2(SC2IFP(sc)->if_bpf, &af1, sizeof(af1), m); 674 } 675 676 /* verify the CARP version. */ 677 if (ch->carp_version != CARP_VERSION) { 678 carpstats.carps_badver++; 679 SC2IFP(sc)->if_ierrors++; 680 CARP_UNLOCK(ifp->if_carp); 681 CARP_LOG("%s; invalid version %d\n", 682 SC2IFP(sc)->if_xname, 683 ch->carp_version); 684 m_freem(m); 685 return; 686 } 687 688 /* verify the hash */ 689 if (carp_hmac_verify(sc, ch->carp_counter, ch->carp_md)) { 690 carpstats.carps_badauth++; 691 SC2IFP(sc)->if_ierrors++; 692 CARP_UNLOCK(ifp->if_carp); 693 CARP_LOG("%s: incorrect hash\n", SC2IFP(sc)->if_xname); 694 m_freem(m); 695 return; 696 } 697 698 tmp_counter = ntohl(ch->carp_counter[0]); 699 tmp_counter = tmp_counter<<32; 700 tmp_counter += ntohl(ch->carp_counter[1]); 701 702 /* XXX Replay protection goes here */ 703 704 sc->sc_init_counter = 0; 705 sc->sc_counter = tmp_counter; 706 707 sc_tv.tv_sec = sc->sc_advbase; 708 if (carp_suppress_preempt && sc->sc_advskew < 240) 709 sc_tv.tv_usec = 240 * 1000000 / 256; 710 else 711 sc_tv.tv_usec = sc->sc_advskew * 1000000 / 256; 712 ch_tv.tv_sec = ch->carp_advbase; 713 ch_tv.tv_usec = ch->carp_advskew * 1000000 / 256; 714 715 switch (sc->sc_state) { 716 case INIT: 717 break; 718 case MASTER: 719 /* 720 * If we receive an advertisement from a master who's going to 721 * be more frequent than us, go into BACKUP state. 722 */ 723 if (timevalcmp(&sc_tv, &ch_tv, >) || 724 timevalcmp(&sc_tv, &ch_tv, ==)) { 725 callout_stop(&sc->sc_ad_tmo); 726 CARP_DEBUG("%s: MASTER -> BACKUP " 727 "(more frequent advertisement received)\n", 728 SC2IFP(sc)->if_xname); 729 carp_set_state(sc, BACKUP); 730 carp_setrun(sc, 0); 731 carp_setroute(sc, RTM_DELETE); 732 } 733 break; 734 case BACKUP: 735 /* 736 * If we're pre-empting masters who advertise slower than us, 737 * and this one claims to be slower, treat him as down. 738 */ 739 if (carp_opts[CARPCTL_PREEMPT] && 740 timevalcmp(&sc_tv, &ch_tv, <)) { 741 CARP_DEBUG("%s: BACKUP -> MASTER " 742 "(preempting a slower master)\n", 743 SC2IFP(sc)->if_xname); 744 carp_master_down_locked(sc); 745 break; 746 } 747 748 /* 749 * If the master is going to advertise at such a low frequency 750 * that he's guaranteed to time out, we'd might as well just 751 * treat him as timed out now. 752 */ 753 sc_tv.tv_sec = sc->sc_advbase * 3; 754 if (timevalcmp(&sc_tv, &ch_tv, <)) { 755 CARP_DEBUG("%s: BACKUP -> MASTER " 756 "(master timed out)\n", 757 SC2IFP(sc)->if_xname); 758 carp_master_down_locked(sc); 759 break; 760 } 761 762 /* 763 * Otherwise, we reset the counter and wait for the next 764 * advertisement. 765 */ 766 carp_setrun(sc, af); 767 break; 768 } 769 770 CARP_UNLOCK(ifp->if_carp); 771 772 m_freem(m); 773 return; 774 } 775 776 static int 777 carp_prepare_ad(struct mbuf *m, struct carp_softc *sc, struct carp_header *ch) 778 { 779 struct m_tag *mtag; 780 struct ifnet *ifp = SC2IFP(sc); 781 782 if (sc->sc_init_counter) { 783 /* this could also be seconds since unix epoch */ 784 sc->sc_counter = arc4random(); 785 sc->sc_counter = sc->sc_counter << 32; 786 sc->sc_counter += arc4random(); 787 } else 788 sc->sc_counter++; 789 790 ch->carp_counter[0] = htonl((sc->sc_counter>>32)&0xffffffff); 791 ch->carp_counter[1] = htonl(sc->sc_counter&0xffffffff); 792 793 carp_hmac_generate(sc, ch->carp_counter, ch->carp_md); 794 795 /* Tag packet for carp_output */ 796 mtag = m_tag_get(PACKET_TAG_CARP, sizeof(struct ifnet *), M_NOWAIT); 797 if (mtag == NULL) { 798 m_freem(m); 799 SC2IFP(sc)->if_oerrors++; 800 return (ENOMEM); 801 } 802 bcopy(&ifp, (caddr_t)(mtag + 1), sizeof(struct ifnet *)); 803 m_tag_prepend(m, mtag); 804 805 return (0); 806 } 807 808 static void 809 carp_send_ad_all(void) 810 { 811 struct carp_softc *sc; 812 813 mtx_lock(&carp_mtx); 814 LIST_FOREACH(sc, &carpif_list, sc_next) { 815 if (sc->sc_carpdev == NULL) 816 continue; 817 CARP_SCLOCK(sc); 818 if ((SC2IFP(sc)->if_flags & IFF_UP) && 819 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING) && 820 sc->sc_state == MASTER) 821 carp_send_ad_locked(sc); 822 CARP_SCUNLOCK(sc); 823 } 824 mtx_unlock(&carp_mtx); 825 } 826 827 static void 828 carp_send_ad(void *v) 829 { 830 struct carp_softc *sc = v; 831 832 CARP_SCLOCK(sc); 833 carp_send_ad_locked(sc); 834 CARP_SCUNLOCK(sc); 835 } 836 837 static void 838 carp_send_ad_locked(struct carp_softc *sc) 839 { 840 struct carp_header ch; 841 struct timeval tv; 842 struct carp_header *ch_ptr; 843 struct mbuf *m; 844 int len, advbase, advskew; 845 846 CARP_SCLOCK_ASSERT(sc); 847 848 /* bow out if we've lost our UPness or RUNNINGuiness */ 849 if (!((SC2IFP(sc)->if_flags & IFF_UP) && 850 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING))) { 851 advbase = 255; 852 advskew = 255; 853 } else { 854 advbase = sc->sc_advbase; 855 if (!carp_suppress_preempt || sc->sc_advskew > 240) 856 advskew = sc->sc_advskew; 857 else 858 advskew = 240; 859 tv.tv_sec = advbase; 860 tv.tv_usec = advskew * 1000000 / 256; 861 } 862 863 ch.carp_version = CARP_VERSION; 864 ch.carp_type = CARP_ADVERTISEMENT; 865 ch.carp_vhid = sc->sc_vhid; 866 ch.carp_advbase = advbase; 867 ch.carp_advskew = advskew; 868 ch.carp_authlen = 7; /* XXX DEFINE */ 869 ch.carp_pad1 = 0; /* must be zero */ 870 ch.carp_cksum = 0; 871 872 #ifdef INET 873 if (sc->sc_ia) { 874 struct ip *ip; 875 876 MGETHDR(m, M_DONTWAIT, MT_HEADER); 877 if (m == NULL) { 878 SC2IFP(sc)->if_oerrors++; 879 carpstats.carps_onomem++; 880 /* XXX maybe less ? */ 881 if (advbase != 255 || advskew != 255) 882 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv), 883 carp_send_ad, sc); 884 return; 885 } 886 len = sizeof(*ip) + sizeof(ch); 887 m->m_pkthdr.len = len; 888 m->m_pkthdr.rcvif = NULL; 889 m->m_len = len; 890 MH_ALIGN(m, m->m_len); 891 m->m_flags |= M_MCAST; 892 ip = mtod(m, struct ip *); 893 ip->ip_v = IPVERSION; 894 ip->ip_hl = sizeof(*ip) >> 2; 895 ip->ip_tos = IPTOS_LOWDELAY; 896 ip->ip_len = len; 897 ip->ip_id = ip_newid(); 898 ip->ip_off = IP_DF; 899 ip->ip_ttl = CARP_DFLTTL; 900 ip->ip_p = IPPROTO_CARP; 901 ip->ip_sum = 0; 902 ip->ip_src.s_addr = sc->sc_ia->ia_addr.sin_addr.s_addr; 903 ip->ip_dst.s_addr = htonl(INADDR_CARP_GROUP); 904 905 ch_ptr = (struct carp_header *)(&ip[1]); 906 bcopy(&ch, ch_ptr, sizeof(ch)); 907 if (carp_prepare_ad(m, sc, ch_ptr)) 908 return; 909 910 m->m_data += sizeof(*ip); 911 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip)); 912 m->m_data -= sizeof(*ip); 913 914 getmicrotime(&SC2IFP(sc)->if_lastchange); 915 SC2IFP(sc)->if_opackets++; 916 SC2IFP(sc)->if_obytes += len; 917 carpstats.carps_opackets++; 918 919 if (ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, NULL)) { 920 SC2IFP(sc)->if_oerrors++; 921 if (sc->sc_sendad_errors < INT_MAX) 922 sc->sc_sendad_errors++; 923 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) { 924 carp_suppress_preempt++; 925 if (carp_suppress_preempt == 1) { 926 CARP_SCUNLOCK(sc); 927 carp_send_ad_all(); 928 CARP_SCLOCK(sc); 929 } 930 } 931 sc->sc_sendad_success = 0; 932 } else { 933 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) { 934 if (++sc->sc_sendad_success >= 935 CARP_SENDAD_MIN_SUCCESS) { 936 carp_suppress_preempt--; 937 sc->sc_sendad_errors = 0; 938 } 939 } else 940 sc->sc_sendad_errors = 0; 941 } 942 } 943 #endif /* INET */ 944 #ifdef INET6 945 if (sc->sc_ia6) { 946 struct ip6_hdr *ip6; 947 948 MGETHDR(m, M_DONTWAIT, MT_HEADER); 949 if (m == NULL) { 950 SC2IFP(sc)->if_oerrors++; 951 carpstats.carps_onomem++; 952 /* XXX maybe less ? */ 953 if (advbase != 255 || advskew != 255) 954 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv), 955 carp_send_ad, sc); 956 return; 957 } 958 len = sizeof(*ip6) + sizeof(ch); 959 m->m_pkthdr.len = len; 960 m->m_pkthdr.rcvif = NULL; 961 m->m_len = len; 962 MH_ALIGN(m, m->m_len); 963 m->m_flags |= M_MCAST; 964 ip6 = mtod(m, struct ip6_hdr *); 965 bzero(ip6, sizeof(*ip6)); 966 ip6->ip6_vfc |= IPV6_VERSION; 967 ip6->ip6_hlim = CARP_DFLTTL; 968 ip6->ip6_nxt = IPPROTO_CARP; 969 bcopy(&sc->sc_ia6->ia_addr.sin6_addr, &ip6->ip6_src, 970 sizeof(struct in6_addr)); 971 /* set the multicast destination */ 972 973 ip6->ip6_dst.s6_addr16[0] = htons(0xff02); 974 ip6->ip6_dst.s6_addr8[15] = 0x12; 975 if (in6_setscope(&ip6->ip6_dst, sc->sc_carpdev, NULL) != 0) { 976 SC2IFP(sc)->if_oerrors++; 977 m_freem(m); 978 CARP_LOG("%s: in6_setscope failed\n", __func__); 979 return; 980 } 981 982 ch_ptr = (struct carp_header *)(&ip6[1]); 983 bcopy(&ch, ch_ptr, sizeof(ch)); 984 if (carp_prepare_ad(m, sc, ch_ptr)) 985 return; 986 987 m->m_data += sizeof(*ip6); 988 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip6)); 989 m->m_data -= sizeof(*ip6); 990 991 getmicrotime(&SC2IFP(sc)->if_lastchange); 992 SC2IFP(sc)->if_opackets++; 993 SC2IFP(sc)->if_obytes += len; 994 carpstats.carps_opackets6++; 995 996 if (ip6_output(m, NULL, NULL, 0, &sc->sc_im6o, NULL, NULL)) { 997 SC2IFP(sc)->if_oerrors++; 998 if (sc->sc_sendad_errors < INT_MAX) 999 sc->sc_sendad_errors++; 1000 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) { 1001 carp_suppress_preempt++; 1002 if (carp_suppress_preempt == 1) { 1003 CARP_SCUNLOCK(sc); 1004 carp_send_ad_all(); 1005 CARP_SCLOCK(sc); 1006 } 1007 } 1008 sc->sc_sendad_success = 0; 1009 } else { 1010 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) { 1011 if (++sc->sc_sendad_success >= 1012 CARP_SENDAD_MIN_SUCCESS) { 1013 carp_suppress_preempt--; 1014 sc->sc_sendad_errors = 0; 1015 } 1016 } else 1017 sc->sc_sendad_errors = 0; 1018 } 1019 } 1020 #endif /* INET6 */ 1021 1022 if (advbase != 255 || advskew != 255) 1023 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv), 1024 carp_send_ad, sc); 1025 1026 } 1027 1028 /* 1029 * Broadcast a gratuitous ARP request containing 1030 * the virtual router MAC address for each IP address 1031 * associated with the virtual router. 1032 */ 1033 static void 1034 carp_send_arp(struct carp_softc *sc) 1035 { 1036 struct ifaddr *ifa; 1037 1038 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) { 1039 1040 if (ifa->ifa_addr->sa_family != AF_INET) 1041 continue; 1042 1043 /* arprequest(sc->sc_carpdev, &in, &in, IF_LLADDR(sc->sc_ifp)); */ 1044 arp_ifinit2(sc->sc_carpdev, ifa, IF_LLADDR(sc->sc_ifp)); 1045 1046 DELAY(1000); /* XXX */ 1047 } 1048 } 1049 1050 #ifdef INET6 1051 static void 1052 carp_send_na(struct carp_softc *sc) 1053 { 1054 struct ifaddr *ifa; 1055 struct in6_addr *in6; 1056 static struct in6_addr mcast = IN6ADDR_LINKLOCAL_ALLNODES_INIT; 1057 1058 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) { 1059 1060 if (ifa->ifa_addr->sa_family != AF_INET6) 1061 continue; 1062 1063 in6 = &ifatoia6(ifa)->ia_addr.sin6_addr; 1064 nd6_na_output(sc->sc_carpdev, &mcast, in6, 1065 ND_NA_FLAG_OVERRIDE, 1, NULL); 1066 DELAY(1000); /* XXX */ 1067 } 1068 } 1069 #endif /* INET6 */ 1070 1071 static int 1072 carp_addrcount(struct carp_if *cif, struct in_ifaddr *ia, int type) 1073 { 1074 struct carp_softc *vh; 1075 struct ifaddr *ifa; 1076 int count = 0; 1077 1078 CARP_LOCK_ASSERT(cif); 1079 1080 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1081 if ((type == CARP_COUNT_RUNNING && 1082 (SC2IFP(vh)->if_flags & IFF_UP) && 1083 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING)) || 1084 (type == CARP_COUNT_MASTER && vh->sc_state == MASTER)) { 1085 TAILQ_FOREACH(ifa, &SC2IFP(vh)->if_addrlist, 1086 ifa_list) { 1087 if (ifa->ifa_addr->sa_family == AF_INET && 1088 ia->ia_addr.sin_addr.s_addr == 1089 ifatoia(ifa)->ia_addr.sin_addr.s_addr) 1090 count++; 1091 } 1092 } 1093 } 1094 return (count); 1095 } 1096 1097 int 1098 carp_iamatch(void *v, struct in_ifaddr *ia, 1099 struct in_addr *isaddr, u_int8_t **enaddr) 1100 { 1101 struct carp_if *cif = v; 1102 struct carp_softc *vh; 1103 int index, count = 0; 1104 struct ifaddr *ifa; 1105 1106 CARP_LOCK(cif); 1107 1108 if (carp_opts[CARPCTL_ARPBALANCE]) { 1109 /* 1110 * XXX proof of concept implementation. 1111 * We use the source ip to decide which virtual host should 1112 * handle the request. If we're master of that virtual host, 1113 * then we respond, otherwise, just drop the arp packet on 1114 * the floor. 1115 */ 1116 count = carp_addrcount(cif, ia, CARP_COUNT_RUNNING); 1117 if (count == 0) { 1118 /* should never reach this */ 1119 CARP_UNLOCK(cif); 1120 return (0); 1121 } 1122 1123 /* this should be a hash, like pf_hash() */ 1124 index = ntohl(isaddr->s_addr) % count; 1125 count = 0; 1126 1127 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1128 if ((SC2IFP(vh)->if_flags & IFF_UP) && 1129 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING)) { 1130 TAILQ_FOREACH(ifa, &SC2IFP(vh)->if_addrlist, 1131 ifa_list) { 1132 if (ifa->ifa_addr->sa_family == 1133 AF_INET && 1134 ia->ia_addr.sin_addr.s_addr == 1135 ifatoia(ifa)->ia_addr.sin_addr.s_addr) { 1136 if (count == index) { 1137 if (vh->sc_state == 1138 MASTER) { 1139 *enaddr = IF_LLADDR(vh->sc_ifp); 1140 CARP_UNLOCK(cif); 1141 return (1); 1142 } else { 1143 CARP_UNLOCK(cif); 1144 return (0); 1145 } 1146 } 1147 count++; 1148 } 1149 } 1150 } 1151 } 1152 } else { 1153 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1154 if ((SC2IFP(vh)->if_flags & IFF_UP) && 1155 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING) && 1156 ia->ia_ifp == SC2IFP(vh) && 1157 vh->sc_state == MASTER) { 1158 *enaddr = IF_LLADDR(vh->sc_ifp); 1159 CARP_UNLOCK(cif); 1160 return (1); 1161 } 1162 } 1163 } 1164 CARP_UNLOCK(cif); 1165 return (0); 1166 } 1167 1168 #ifdef INET6 1169 struct ifaddr * 1170 carp_iamatch6(void *v, struct in6_addr *taddr) 1171 { 1172 struct carp_if *cif = v; 1173 struct carp_softc *vh; 1174 struct ifaddr *ifa; 1175 1176 CARP_LOCK(cif); 1177 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1178 TAILQ_FOREACH(ifa, &SC2IFP(vh)->if_addrlist, ifa_list) { 1179 if (IN6_ARE_ADDR_EQUAL(taddr, 1180 &ifatoia6(ifa)->ia_addr.sin6_addr) && 1181 (SC2IFP(vh)->if_flags & IFF_UP) && 1182 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING) && 1183 vh->sc_state == MASTER) { 1184 CARP_UNLOCK(cif); 1185 return (ifa); 1186 } 1187 } 1188 } 1189 CARP_UNLOCK(cif); 1190 1191 return (NULL); 1192 } 1193 1194 void * 1195 carp_macmatch6(void *v, struct mbuf *m, const struct in6_addr *taddr) 1196 { 1197 struct m_tag *mtag; 1198 struct carp_if *cif = v; 1199 struct carp_softc *sc; 1200 struct ifaddr *ifa; 1201 1202 CARP_LOCK(cif); 1203 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list) { 1204 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) { 1205 if (IN6_ARE_ADDR_EQUAL(taddr, 1206 &ifatoia6(ifa)->ia_addr.sin6_addr) && 1207 (SC2IFP(sc)->if_flags & IFF_UP) && 1208 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING)) { 1209 struct ifnet *ifp = SC2IFP(sc); 1210 mtag = m_tag_get(PACKET_TAG_CARP, 1211 sizeof(struct ifnet *), M_NOWAIT); 1212 if (mtag == NULL) { 1213 /* better a bit than nothing */ 1214 CARP_UNLOCK(cif); 1215 return (IF_LLADDR(sc->sc_ifp)); 1216 } 1217 bcopy(&ifp, (caddr_t)(mtag + 1), 1218 sizeof(struct ifnet *)); 1219 m_tag_prepend(m, mtag); 1220 1221 CARP_UNLOCK(cif); 1222 return (IF_LLADDR(sc->sc_ifp)); 1223 } 1224 } 1225 } 1226 CARP_UNLOCK(cif); 1227 1228 return (NULL); 1229 } 1230 #endif 1231 1232 struct ifnet * 1233 carp_forus(void *v, void *dhost) 1234 { 1235 struct carp_if *cif = v; 1236 struct carp_softc *vh; 1237 u_int8_t *ena = dhost; 1238 1239 if (ena[0] || ena[1] || ena[2] != 0x5e || ena[3] || ena[4] != 1) 1240 return (NULL); 1241 1242 CARP_LOCK(cif); 1243 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) 1244 if ((SC2IFP(vh)->if_flags & IFF_UP) && 1245 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING) && 1246 vh->sc_state == MASTER && 1247 !bcmp(dhost, IF_LLADDR(vh->sc_ifp), ETHER_ADDR_LEN)) { 1248 CARP_UNLOCK(cif); 1249 return (SC2IFP(vh)); 1250 } 1251 1252 CARP_UNLOCK(cif); 1253 return (NULL); 1254 } 1255 1256 static void 1257 carp_master_down(void *v) 1258 { 1259 struct carp_softc *sc = v; 1260 1261 CARP_SCLOCK(sc); 1262 carp_master_down_locked(sc); 1263 CARP_SCUNLOCK(sc); 1264 } 1265 1266 static void 1267 carp_master_down_locked(struct carp_softc *sc) 1268 { 1269 if (sc->sc_carpdev) 1270 CARP_SCLOCK_ASSERT(sc); 1271 1272 switch (sc->sc_state) { 1273 case INIT: 1274 printf("%s: master_down event in INIT state\n", 1275 SC2IFP(sc)->if_xname); 1276 break; 1277 case MASTER: 1278 break; 1279 case BACKUP: 1280 carp_set_state(sc, MASTER); 1281 carp_send_ad_locked(sc); 1282 carp_send_arp(sc); 1283 #ifdef INET6 1284 carp_send_na(sc); 1285 #endif /* INET6 */ 1286 carp_setrun(sc, 0); 1287 carp_setroute(sc, RTM_ADD); 1288 break; 1289 } 1290 } 1291 1292 /* 1293 * When in backup state, af indicates whether to reset the master down timer 1294 * for v4 or v6. If it's set to zero, reset the ones which are already pending. 1295 */ 1296 static void 1297 carp_setrun(struct carp_softc *sc, sa_family_t af) 1298 { 1299 struct timeval tv; 1300 1301 if (sc->sc_carpdev == NULL) { 1302 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING; 1303 carp_set_state(sc, INIT); 1304 return; 1305 } else 1306 CARP_SCLOCK_ASSERT(sc); 1307 1308 if (SC2IFP(sc)->if_flags & IFF_UP && 1309 sc->sc_vhid > 0 && (sc->sc_naddrs || sc->sc_naddrs6)) 1310 SC2IFP(sc)->if_drv_flags |= IFF_DRV_RUNNING; 1311 else { 1312 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING; 1313 carp_setroute(sc, RTM_DELETE); 1314 return; 1315 } 1316 1317 switch (sc->sc_state) { 1318 case INIT: 1319 if (carp_opts[CARPCTL_PREEMPT] && !carp_suppress_preempt) { 1320 carp_send_ad_locked(sc); 1321 carp_send_arp(sc); 1322 #ifdef INET6 1323 carp_send_na(sc); 1324 #endif /* INET6 */ 1325 CARP_DEBUG("%s: INIT -> MASTER (preempting)\n", 1326 SC2IFP(sc)->if_xname); 1327 carp_set_state(sc, MASTER); 1328 carp_setroute(sc, RTM_ADD); 1329 } else { 1330 CARP_DEBUG("%s: INIT -> BACKUP\n", SC2IFP(sc)->if_xname); 1331 carp_set_state(sc, BACKUP); 1332 carp_setroute(sc, RTM_DELETE); 1333 carp_setrun(sc, 0); 1334 } 1335 break; 1336 case BACKUP: 1337 callout_stop(&sc->sc_ad_tmo); 1338 tv.tv_sec = 3 * sc->sc_advbase; 1339 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1340 switch (af) { 1341 #ifdef INET 1342 case AF_INET: 1343 callout_reset(&sc->sc_md_tmo, tvtohz(&tv), 1344 carp_master_down, sc); 1345 break; 1346 #endif /* INET */ 1347 #ifdef INET6 1348 case AF_INET6: 1349 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv), 1350 carp_master_down, sc); 1351 break; 1352 #endif /* INET6 */ 1353 default: 1354 if (sc->sc_naddrs) 1355 callout_reset(&sc->sc_md_tmo, tvtohz(&tv), 1356 carp_master_down, sc); 1357 if (sc->sc_naddrs6) 1358 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv), 1359 carp_master_down, sc); 1360 break; 1361 } 1362 break; 1363 case MASTER: 1364 tv.tv_sec = sc->sc_advbase; 1365 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1366 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv), 1367 carp_send_ad, sc); 1368 break; 1369 } 1370 } 1371 1372 void 1373 carp_multicast_cleanup(struct carp_softc *sc) 1374 { 1375 struct ip_moptions *imo = &sc->sc_imo; 1376 #ifdef INET6 1377 struct ip6_moptions *im6o = &sc->sc_im6o; 1378 #endif 1379 u_int16_t n = imo->imo_num_memberships; 1380 1381 /* Clean up our own multicast memberships */ 1382 while (n-- > 0) { 1383 if (imo->imo_membership[n] != NULL) { 1384 in_delmulti(imo->imo_membership[n]); 1385 imo->imo_membership[n] = NULL; 1386 } 1387 } 1388 imo->imo_num_memberships = 0; 1389 imo->imo_multicast_ifp = NULL; 1390 1391 #ifdef INET6 1392 while (!LIST_EMPTY(&im6o->im6o_memberships)) { 1393 struct in6_multi_mship *imm = 1394 LIST_FIRST(&im6o->im6o_memberships); 1395 1396 LIST_REMOVE(imm, i6mm_chain); 1397 in6_leavegroup(imm); 1398 } 1399 im6o->im6o_multicast_ifp = NULL; 1400 #endif 1401 } 1402 1403 static int 1404 carp_set_addr(struct carp_softc *sc, struct sockaddr_in *sin) 1405 { 1406 struct ifnet *ifp; 1407 struct carp_if *cif; 1408 struct in_ifaddr *ia, *ia_if; 1409 struct ip_moptions *imo = &sc->sc_imo; 1410 struct in_addr addr; 1411 u_long iaddr = htonl(sin->sin_addr.s_addr); 1412 int own, error; 1413 1414 if (sin->sin_addr.s_addr == 0) { 1415 if (!(SC2IFP(sc)->if_flags & IFF_UP)) 1416 carp_set_state(sc, INIT); 1417 if (sc->sc_naddrs) 1418 SC2IFP(sc)->if_flags |= IFF_UP; 1419 carp_setrun(sc, 0); 1420 return (0); 1421 } 1422 1423 /* we have to do it by hands to check we won't match on us */ 1424 ia_if = NULL; own = 0; 1425 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) { 1426 /* and, yeah, we need a multicast-capable iface too */ 1427 if (ia->ia_ifp != SC2IFP(sc) && 1428 (ia->ia_ifp->if_flags & IFF_MULTICAST) && 1429 (iaddr & ia->ia_subnetmask) == ia->ia_subnet) { 1430 if (!ia_if) 1431 ia_if = ia; 1432 if (sin->sin_addr.s_addr == 1433 ia->ia_addr.sin_addr.s_addr) 1434 own++; 1435 } 1436 } 1437 1438 if (!ia_if) 1439 return (EADDRNOTAVAIL); 1440 1441 ia = ia_if; 1442 ifp = ia->ia_ifp; 1443 1444 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0 || 1445 (imo->imo_multicast_ifp && imo->imo_multicast_ifp != ifp)) 1446 return (EADDRNOTAVAIL); 1447 1448 if (imo->imo_num_memberships == 0) { 1449 addr.s_addr = htonl(INADDR_CARP_GROUP); 1450 if ((imo->imo_membership[0] = in_addmulti(&addr, ifp)) == NULL) 1451 return (ENOBUFS); 1452 imo->imo_num_memberships++; 1453 imo->imo_multicast_ifp = ifp; 1454 imo->imo_multicast_ttl = CARP_DFLTTL; 1455 imo->imo_multicast_loop = 0; 1456 } 1457 1458 if (!ifp->if_carp) { 1459 1460 MALLOC(cif, struct carp_if *, sizeof(*cif), M_CARP, 1461 M_WAITOK|M_ZERO); 1462 if (!cif) { 1463 error = ENOBUFS; 1464 goto cleanup; 1465 } 1466 if ((error = ifpromisc(ifp, 1))) { 1467 FREE(cif, M_CARP); 1468 goto cleanup; 1469 } 1470 1471 CARP_LOCK_INIT(cif); 1472 CARP_LOCK(cif); 1473 cif->vhif_ifp = ifp; 1474 TAILQ_INIT(&cif->vhif_vrs); 1475 ifp->if_carp = cif; 1476 1477 } else { 1478 struct carp_softc *vr; 1479 1480 cif = (struct carp_if *)ifp->if_carp; 1481 CARP_LOCK(cif); 1482 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) 1483 if (vr != sc && vr->sc_vhid == sc->sc_vhid) { 1484 CARP_UNLOCK(cif); 1485 error = EINVAL; 1486 goto cleanup; 1487 } 1488 } 1489 sc->sc_ia = ia; 1490 sc->sc_carpdev = ifp; 1491 1492 { /* XXX prevent endless loop if already in queue */ 1493 struct carp_softc *vr, *after = NULL; 1494 int myself = 0; 1495 cif = (struct carp_if *)ifp->if_carp; 1496 1497 /* XXX: cif should not change, right? So we still hold the lock */ 1498 CARP_LOCK_ASSERT(cif); 1499 1500 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) { 1501 if (vr == sc) 1502 myself = 1; 1503 if (vr->sc_vhid < sc->sc_vhid) 1504 after = vr; 1505 } 1506 1507 if (!myself) { 1508 /* We're trying to keep things in order */ 1509 if (after == NULL) { 1510 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list); 1511 } else { 1512 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, sc, sc_list); 1513 } 1514 cif->vhif_nvrs++; 1515 } 1516 } 1517 1518 sc->sc_naddrs++; 1519 SC2IFP(sc)->if_flags |= IFF_UP; 1520 if (own) 1521 sc->sc_advskew = 0; 1522 carp_sc_state_locked(sc); 1523 carp_setrun(sc, 0); 1524 1525 CARP_UNLOCK(cif); 1526 1527 return (0); 1528 1529 cleanup: 1530 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]); 1531 return (error); 1532 } 1533 1534 static int 1535 carp_del_addr(struct carp_softc *sc, struct sockaddr_in *sin) 1536 { 1537 int error = 0; 1538 1539 if (!--sc->sc_naddrs) { 1540 struct carp_if *cif = (struct carp_if *)sc->sc_carpdev->if_carp; 1541 struct ip_moptions *imo = &sc->sc_imo; 1542 1543 CARP_LOCK(cif); 1544 callout_stop(&sc->sc_ad_tmo); 1545 SC2IFP(sc)->if_flags &= ~IFF_UP; 1546 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING; 1547 sc->sc_vhid = -1; 1548 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]); 1549 imo->imo_multicast_ifp = NULL; 1550 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list); 1551 if (!--cif->vhif_nvrs) { 1552 sc->sc_carpdev->if_carp = NULL; 1553 CARP_LOCK_DESTROY(cif); 1554 FREE(cif, M_IFADDR); 1555 } else { 1556 CARP_UNLOCK(cif); 1557 } 1558 } 1559 1560 return (error); 1561 } 1562 1563 #ifdef INET6 1564 static int 1565 carp_set_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6) 1566 { 1567 struct ifnet *ifp; 1568 struct carp_if *cif; 1569 struct in6_ifaddr *ia, *ia_if; 1570 struct ip6_moptions *im6o = &sc->sc_im6o; 1571 struct in6_multi_mship *imm; 1572 struct in6_addr in6; 1573 int own, error; 1574 1575 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1576 if (!(SC2IFP(sc)->if_flags & IFF_UP)) 1577 carp_set_state(sc, INIT); 1578 if (sc->sc_naddrs6) 1579 SC2IFP(sc)->if_flags |= IFF_UP; 1580 carp_setrun(sc, 0); 1581 return (0); 1582 } 1583 1584 /* we have to do it by hands to check we won't match on us */ 1585 ia_if = NULL; own = 0; 1586 for (ia = in6_ifaddr; ia; ia = ia->ia_next) { 1587 int i; 1588 1589 for (i = 0; i < 4; i++) { 1590 if ((sin6->sin6_addr.s6_addr32[i] & 1591 ia->ia_prefixmask.sin6_addr.s6_addr32[i]) != 1592 (ia->ia_addr.sin6_addr.s6_addr32[i] & 1593 ia->ia_prefixmask.sin6_addr.s6_addr32[i])) 1594 break; 1595 } 1596 /* and, yeah, we need a multicast-capable iface too */ 1597 if (ia->ia_ifp != SC2IFP(sc) && 1598 (ia->ia_ifp->if_flags & IFF_MULTICAST) && 1599 (i == 4)) { 1600 if (!ia_if) 1601 ia_if = ia; 1602 if (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, 1603 &ia->ia_addr.sin6_addr)) 1604 own++; 1605 } 1606 } 1607 1608 if (!ia_if) 1609 return (EADDRNOTAVAIL); 1610 ia = ia_if; 1611 ifp = ia->ia_ifp; 1612 1613 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0 || 1614 (im6o->im6o_multicast_ifp && im6o->im6o_multicast_ifp != ifp)) 1615 return (EADDRNOTAVAIL); 1616 1617 if (!sc->sc_naddrs6) { 1618 im6o->im6o_multicast_ifp = ifp; 1619 1620 /* join CARP multicast address */ 1621 bzero(&in6, sizeof(in6)); 1622 in6.s6_addr16[0] = htons(0xff02); 1623 in6.s6_addr8[15] = 0x12; 1624 if (in6_setscope(&in6, ifp, NULL) != 0) 1625 goto cleanup; 1626 if ((imm = in6_joingroup(ifp, &in6, &error, 0)) == NULL) 1627 goto cleanup; 1628 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 1629 1630 /* join solicited multicast address */ 1631 bzero(&in6, sizeof(in6)); 1632 in6.s6_addr16[0] = htons(0xff02); 1633 in6.s6_addr32[1] = 0; 1634 in6.s6_addr32[2] = htonl(1); 1635 in6.s6_addr32[3] = sin6->sin6_addr.s6_addr32[3]; 1636 in6.s6_addr8[12] = 0xff; 1637 if (in6_setscope(&in6, ifp, NULL) != 0) 1638 goto cleanup; 1639 if ((imm = in6_joingroup(ifp, &in6, &error, 0)) == NULL) 1640 goto cleanup; 1641 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, i6mm_chain); 1642 } 1643 1644 if (!ifp->if_carp) { 1645 MALLOC(cif, struct carp_if *, sizeof(*cif), M_CARP, 1646 M_WAITOK|M_ZERO); 1647 if (!cif) { 1648 error = ENOBUFS; 1649 goto cleanup; 1650 } 1651 if ((error = ifpromisc(ifp, 1))) { 1652 FREE(cif, M_CARP); 1653 goto cleanup; 1654 } 1655 1656 CARP_LOCK_INIT(cif); 1657 CARP_LOCK(cif); 1658 cif->vhif_ifp = ifp; 1659 TAILQ_INIT(&cif->vhif_vrs); 1660 ifp->if_carp = cif; 1661 1662 } else { 1663 struct carp_softc *vr; 1664 1665 cif = (struct carp_if *)ifp->if_carp; 1666 CARP_LOCK(cif); 1667 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) 1668 if (vr != sc && vr->sc_vhid == sc->sc_vhid) { 1669 CARP_UNLOCK(cif); 1670 error = EINVAL; 1671 goto cleanup; 1672 } 1673 } 1674 sc->sc_ia6 = ia; 1675 sc->sc_carpdev = ifp; 1676 1677 { /* XXX prevent endless loop if already in queue */ 1678 struct carp_softc *vr, *after = NULL; 1679 int myself = 0; 1680 cif = (struct carp_if *)ifp->if_carp; 1681 CARP_LOCK_ASSERT(cif); 1682 1683 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) { 1684 if (vr == sc) 1685 myself = 1; 1686 if (vr->sc_vhid < sc->sc_vhid) 1687 after = vr; 1688 } 1689 1690 if (!myself) { 1691 /* We're trying to keep things in order */ 1692 if (after == NULL) { 1693 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list); 1694 } else { 1695 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, sc, sc_list); 1696 } 1697 cif->vhif_nvrs++; 1698 } 1699 } 1700 1701 sc->sc_naddrs6++; 1702 SC2IFP(sc)->if_flags |= IFF_UP; 1703 if (own) 1704 sc->sc_advskew = 0; 1705 carp_sc_state_locked(sc); 1706 carp_setrun(sc, 0); 1707 1708 CARP_UNLOCK(cif); 1709 1710 return (0); 1711 1712 cleanup: 1713 /* clean up multicast memberships */ 1714 if (!sc->sc_naddrs6) { 1715 while (!LIST_EMPTY(&im6o->im6o_memberships)) { 1716 imm = LIST_FIRST(&im6o->im6o_memberships); 1717 LIST_REMOVE(imm, i6mm_chain); 1718 in6_leavegroup(imm); 1719 } 1720 } 1721 return (error); 1722 } 1723 1724 static int 1725 carp_del_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6) 1726 { 1727 int error = 0; 1728 1729 if (!--sc->sc_naddrs6) { 1730 struct carp_if *cif = (struct carp_if *)sc->sc_carpdev->if_carp; 1731 struct ip6_moptions *im6o = &sc->sc_im6o; 1732 1733 CARP_LOCK(cif); 1734 callout_stop(&sc->sc_ad_tmo); 1735 SC2IFP(sc)->if_flags &= ~IFF_UP; 1736 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING; 1737 sc->sc_vhid = -1; 1738 while (!LIST_EMPTY(&im6o->im6o_memberships)) { 1739 struct in6_multi_mship *imm = 1740 LIST_FIRST(&im6o->im6o_memberships); 1741 1742 LIST_REMOVE(imm, i6mm_chain); 1743 in6_leavegroup(imm); 1744 } 1745 im6o->im6o_multicast_ifp = NULL; 1746 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list); 1747 if (!--cif->vhif_nvrs) { 1748 CARP_LOCK_DESTROY(cif); 1749 sc->sc_carpdev->if_carp = NULL; 1750 FREE(cif, M_IFADDR); 1751 } else 1752 CARP_UNLOCK(cif); 1753 } 1754 1755 return (error); 1756 } 1757 #endif /* INET6 */ 1758 1759 static int 1760 carp_ioctl(struct ifnet *ifp, u_long cmd, caddr_t addr) 1761 { 1762 struct carp_softc *sc = ifp->if_softc, *vr; 1763 struct carpreq carpr; 1764 struct ifaddr *ifa; 1765 struct ifreq *ifr; 1766 struct ifaliasreq *ifra; 1767 int locked = 0, error = 0; 1768 1769 ifa = (struct ifaddr *)addr; 1770 ifra = (struct ifaliasreq *)addr; 1771 ifr = (struct ifreq *)addr; 1772 1773 switch (cmd) { 1774 case SIOCSIFADDR: 1775 switch (ifa->ifa_addr->sa_family) { 1776 #ifdef INET 1777 case AF_INET: 1778 SC2IFP(sc)->if_flags |= IFF_UP; 1779 bcopy(ifa->ifa_addr, ifa->ifa_dstaddr, 1780 sizeof(struct sockaddr)); 1781 error = carp_set_addr(sc, satosin(ifa->ifa_addr)); 1782 break; 1783 #endif /* INET */ 1784 #ifdef INET6 1785 case AF_INET6: 1786 SC2IFP(sc)->if_flags |= IFF_UP; 1787 error = carp_set_addr6(sc, satosin6(ifa->ifa_addr)); 1788 break; 1789 #endif /* INET6 */ 1790 default: 1791 error = EAFNOSUPPORT; 1792 break; 1793 } 1794 break; 1795 1796 case SIOCAIFADDR: 1797 switch (ifa->ifa_addr->sa_family) { 1798 #ifdef INET 1799 case AF_INET: 1800 SC2IFP(sc)->if_flags |= IFF_UP; 1801 bcopy(ifa->ifa_addr, ifa->ifa_dstaddr, 1802 sizeof(struct sockaddr)); 1803 error = carp_set_addr(sc, satosin(&ifra->ifra_addr)); 1804 break; 1805 #endif /* INET */ 1806 #ifdef INET6 1807 case AF_INET6: 1808 SC2IFP(sc)->if_flags |= IFF_UP; 1809 error = carp_set_addr6(sc, satosin6(&ifra->ifra_addr)); 1810 break; 1811 #endif /* INET6 */ 1812 default: 1813 error = EAFNOSUPPORT; 1814 break; 1815 } 1816 break; 1817 1818 case SIOCDIFADDR: 1819 switch (ifa->ifa_addr->sa_family) { 1820 #ifdef INET 1821 case AF_INET: 1822 error = carp_del_addr(sc, satosin(&ifra->ifra_addr)); 1823 break; 1824 #endif /* INET */ 1825 #ifdef INET6 1826 case AF_INET6: 1827 error = carp_del_addr6(sc, satosin6(&ifra->ifra_addr)); 1828 break; 1829 #endif /* INET6 */ 1830 default: 1831 error = EAFNOSUPPORT; 1832 break; 1833 } 1834 break; 1835 1836 case SIOCSIFFLAGS: 1837 if (sc->sc_carpdev) { 1838 locked = 1; 1839 CARP_SCLOCK(sc); 1840 } 1841 if (sc->sc_state != INIT && !(ifr->ifr_flags & IFF_UP)) { 1842 callout_stop(&sc->sc_ad_tmo); 1843 callout_stop(&sc->sc_md_tmo); 1844 callout_stop(&sc->sc_md6_tmo); 1845 if (sc->sc_state == MASTER) 1846 carp_send_ad_locked(sc); 1847 carp_set_state(sc, INIT); 1848 carp_setrun(sc, 0); 1849 } else if (sc->sc_state == INIT && (ifr->ifr_flags & IFF_UP)) { 1850 SC2IFP(sc)->if_flags |= IFF_UP; 1851 carp_setrun(sc, 0); 1852 } 1853 break; 1854 1855 case SIOCSVH: 1856 if ((error = suser(curthread)) != 0) 1857 break; 1858 if ((error = copyin(ifr->ifr_data, &carpr, sizeof carpr))) 1859 break; 1860 error = 1; 1861 if (sc->sc_carpdev) { 1862 locked = 1; 1863 CARP_SCLOCK(sc); 1864 } 1865 if (sc->sc_state != INIT && carpr.carpr_state != sc->sc_state) { 1866 switch (carpr.carpr_state) { 1867 case BACKUP: 1868 callout_stop(&sc->sc_ad_tmo); 1869 carp_set_state(sc, BACKUP); 1870 carp_setrun(sc, 0); 1871 carp_setroute(sc, RTM_DELETE); 1872 break; 1873 case MASTER: 1874 carp_master_down_locked(sc); 1875 break; 1876 default: 1877 break; 1878 } 1879 } 1880 if (carpr.carpr_vhid > 0) { 1881 if (carpr.carpr_vhid > 255) { 1882 error = EINVAL; 1883 break; 1884 } 1885 if (sc->sc_carpdev) { 1886 struct carp_if *cif; 1887 cif = (struct carp_if *)sc->sc_carpdev->if_carp; 1888 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) 1889 if (vr != sc && 1890 vr->sc_vhid == carpr.carpr_vhid) 1891 return EEXIST; 1892 } 1893 sc->sc_vhid = carpr.carpr_vhid; 1894 IF_LLADDR(sc->sc_ifp)[0] = 0; 1895 IF_LLADDR(sc->sc_ifp)[1] = 0; 1896 IF_LLADDR(sc->sc_ifp)[2] = 0x5e; 1897 IF_LLADDR(sc->sc_ifp)[3] = 0; 1898 IF_LLADDR(sc->sc_ifp)[4] = 1; 1899 IF_LLADDR(sc->sc_ifp)[5] = sc->sc_vhid; 1900 error--; 1901 } 1902 if (carpr.carpr_advbase > 0 || carpr.carpr_advskew > 0) { 1903 if (carpr.carpr_advskew >= 255) { 1904 error = EINVAL; 1905 break; 1906 } 1907 if (carpr.carpr_advbase > 255) { 1908 error = EINVAL; 1909 break; 1910 } 1911 sc->sc_advbase = carpr.carpr_advbase; 1912 sc->sc_advskew = carpr.carpr_advskew; 1913 error--; 1914 } 1915 bcopy(carpr.carpr_key, sc->sc_key, sizeof(sc->sc_key)); 1916 if (error > 0) 1917 error = EINVAL; 1918 else { 1919 error = 0; 1920 carp_setrun(sc, 0); 1921 } 1922 break; 1923 1924 case SIOCGVH: 1925 /* XXX: lockless read */ 1926 bzero(&carpr, sizeof(carpr)); 1927 carpr.carpr_state = sc->sc_state; 1928 carpr.carpr_vhid = sc->sc_vhid; 1929 carpr.carpr_advbase = sc->sc_advbase; 1930 carpr.carpr_advskew = sc->sc_advskew; 1931 if (suser(curthread) == 0) 1932 bcopy(sc->sc_key, carpr.carpr_key, 1933 sizeof(carpr.carpr_key)); 1934 error = copyout(&carpr, ifr->ifr_data, sizeof(carpr)); 1935 break; 1936 1937 default: 1938 error = EINVAL; 1939 } 1940 1941 if (locked) 1942 CARP_SCUNLOCK(sc); 1943 1944 carp_hmac_prepare(sc); 1945 1946 return (error); 1947 } 1948 1949 /* 1950 * XXX: this is looutput. We should eventually use it from there. 1951 */ 1952 static int 1953 carp_looutput(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, 1954 struct rtentry *rt) 1955 { 1956 u_int32_t af; 1957 1958 M_ASSERTPKTHDR(m); /* check if we have the packet header */ 1959 1960 if (rt && rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 1961 m_freem(m); 1962 return (rt->rt_flags & RTF_BLACKHOLE ? 0 : 1963 rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); 1964 } 1965 1966 ifp->if_opackets++; 1967 ifp->if_obytes += m->m_pkthdr.len; 1968 1969 /* BPF writes need to be handled specially. */ 1970 if (dst->sa_family == AF_UNSPEC) { 1971 bcopy(dst->sa_data, &af, sizeof(af)); 1972 dst->sa_family = af; 1973 } 1974 1975 #if 1 /* XXX */ 1976 switch (dst->sa_family) { 1977 case AF_INET: 1978 case AF_INET6: 1979 case AF_IPX: 1980 case AF_APPLETALK: 1981 break; 1982 default: 1983 printf("carp_looutput: af=%d unexpected\n", dst->sa_family); 1984 m_freem(m); 1985 return (EAFNOSUPPORT); 1986 } 1987 #endif 1988 return(if_simloop(ifp, m, dst->sa_family, 0)); 1989 } 1990 1991 /* 1992 * Start output on carp interface. This function should never be called. 1993 */ 1994 static void 1995 carp_start(struct ifnet *ifp) 1996 { 1997 #ifdef DEBUG 1998 printf("%s: start called\n", ifp->if_xname); 1999 #endif 2000 } 2001 2002 int 2003 carp_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa, 2004 struct rtentry *rt) 2005 { 2006 struct m_tag *mtag; 2007 struct carp_softc *sc; 2008 struct ifnet *carp_ifp; 2009 2010 if (!sa) 2011 return (0); 2012 2013 switch (sa->sa_family) { 2014 #ifdef INET 2015 case AF_INET: 2016 break; 2017 #endif /* INET */ 2018 #ifdef INET6 2019 case AF_INET6: 2020 break; 2021 #endif /* INET6 */ 2022 default: 2023 return (0); 2024 } 2025 2026 mtag = m_tag_find(m, PACKET_TAG_CARP, NULL); 2027 if (mtag == NULL) 2028 return (0); 2029 2030 bcopy(mtag + 1, &carp_ifp, sizeof(struct ifnet *)); 2031 sc = carp_ifp->if_softc; 2032 2033 /* Set the source MAC address to Virtual Router MAC Address */ 2034 switch (ifp->if_type) { 2035 case IFT_ETHER: 2036 case IFT_L2VLAN: { 2037 struct ether_header *eh; 2038 2039 eh = mtod(m, struct ether_header *); 2040 eh->ether_shost[0] = 0; 2041 eh->ether_shost[1] = 0; 2042 eh->ether_shost[2] = 0x5e; 2043 eh->ether_shost[3] = 0; 2044 eh->ether_shost[4] = 1; 2045 eh->ether_shost[5] = sc->sc_vhid; 2046 } 2047 break; 2048 case IFT_FDDI: { 2049 struct fddi_header *fh; 2050 2051 fh = mtod(m, struct fddi_header *); 2052 fh->fddi_shost[0] = 0; 2053 fh->fddi_shost[1] = 0; 2054 fh->fddi_shost[2] = 0x5e; 2055 fh->fddi_shost[3] = 0; 2056 fh->fddi_shost[4] = 1; 2057 fh->fddi_shost[5] = sc->sc_vhid; 2058 } 2059 break; 2060 case IFT_ISO88025: { 2061 struct iso88025_header *th; 2062 th = mtod(m, struct iso88025_header *); 2063 th->iso88025_shost[0] = 3; 2064 th->iso88025_shost[1] = 0; 2065 th->iso88025_shost[2] = 0x40 >> (sc->sc_vhid - 1); 2066 th->iso88025_shost[3] = 0x40000 >> (sc->sc_vhid - 1); 2067 th->iso88025_shost[4] = 0; 2068 th->iso88025_shost[5] = 0; 2069 } 2070 break; 2071 default: 2072 printf("%s: carp is not supported for this interface type\n", 2073 ifp->if_xname); 2074 return (EOPNOTSUPP); 2075 } 2076 2077 return (0); 2078 } 2079 2080 static void 2081 carp_set_state(struct carp_softc *sc, int state) 2082 { 2083 2084 if (sc->sc_carpdev) 2085 CARP_SCLOCK_ASSERT(sc); 2086 2087 if (sc->sc_state == state) 2088 return; 2089 2090 sc->sc_state = state; 2091 switch (state) { 2092 case BACKUP: 2093 SC2IFP(sc)->if_link_state = LINK_STATE_DOWN; 2094 break; 2095 case MASTER: 2096 SC2IFP(sc)->if_link_state = LINK_STATE_UP; 2097 break; 2098 default: 2099 SC2IFP(sc)->if_link_state = LINK_STATE_UNKNOWN; 2100 break; 2101 } 2102 rt_ifmsg(SC2IFP(sc)); 2103 } 2104 2105 void 2106 carp_carpdev_state(void *v) 2107 { 2108 struct carp_if *cif = v; 2109 2110 CARP_LOCK(cif); 2111 carp_carpdev_state_locked(cif); 2112 CARP_UNLOCK(cif); 2113 } 2114 2115 static void 2116 carp_carpdev_state_locked(struct carp_if *cif) 2117 { 2118 struct carp_softc *sc; 2119 2120 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list) 2121 carp_sc_state_locked(sc); 2122 } 2123 2124 static void 2125 carp_sc_state_locked(struct carp_softc *sc) 2126 { 2127 CARP_SCLOCK_ASSERT(sc); 2128 2129 if (sc->sc_carpdev->if_link_state != LINK_STATE_UP || 2130 !(sc->sc_carpdev->if_flags & IFF_UP)) { 2131 sc->sc_flags_backup = SC2IFP(sc)->if_flags; 2132 SC2IFP(sc)->if_flags &= ~IFF_UP; 2133 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING; 2134 callout_stop(&sc->sc_ad_tmo); 2135 callout_stop(&sc->sc_md_tmo); 2136 callout_stop(&sc->sc_md6_tmo); 2137 carp_set_state(sc, INIT); 2138 carp_setrun(sc, 0); 2139 if (!sc->sc_suppress) { 2140 carp_suppress_preempt++; 2141 if (carp_suppress_preempt == 1) { 2142 CARP_SCUNLOCK(sc); 2143 carp_send_ad_all(); 2144 CARP_SCLOCK(sc); 2145 } 2146 } 2147 sc->sc_suppress = 1; 2148 } else { 2149 SC2IFP(sc)->if_flags |= sc->sc_flags_backup; 2150 carp_set_state(sc, INIT); 2151 carp_setrun(sc, 0); 2152 if (sc->sc_suppress) 2153 carp_suppress_preempt--; 2154 sc->sc_suppress = 0; 2155 } 2156 2157 return; 2158 } 2159 2160 static int 2161 carp_modevent(module_t mod, int type, void *data) 2162 { 2163 switch (type) { 2164 case MOD_LOAD: 2165 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event, 2166 carp_ifdetach, NULL, EVENTHANDLER_PRI_ANY); 2167 if (if_detach_event_tag == NULL) 2168 return (ENOMEM); 2169 mtx_init(&carp_mtx, "carp_mtx", NULL, MTX_DEF); 2170 LIST_INIT(&carpif_list); 2171 if_clone_attach(&carp_cloner); 2172 break; 2173 2174 case MOD_UNLOAD: 2175 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag); 2176 if_clone_detach(&carp_cloner); 2177 mtx_destroy(&carp_mtx); 2178 break; 2179 2180 default: 2181 return (EINVAL); 2182 } 2183 2184 return (0); 2185 } 2186 2187 static moduledata_t carp_mod = { 2188 "carp", 2189 carp_modevent, 2190 0 2191 }; 2192 2193 DECLARE_MODULE(carp, carp_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 2194