1 /* 2 * Copyright (c) 2002 Michael Shalayeff. All rights reserved. 3 * Copyright (c) 2003 Ryan McBride. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 15 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 16 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 17 * IN NO EVENT SHALL THE AUTHOR OR HIS RELATIVES BE LIABLE FOR ANY DIRECT, 18 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES 19 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 20 * SERVICES; LOSS OF MIND, USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, 22 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING 23 * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF 24 * THE POSSIBILITY OF SUCH DAMAGE. 25 */ 26 27 #include <sys/cdefs.h> 28 __FBSDID("$FreeBSD$"); 29 30 #include "opt_carp.h" 31 #include "opt_bpf.h" 32 #include "opt_inet.h" 33 #include "opt_inet6.h" 34 35 #include <sys/types.h> 36 #include <sys/param.h> 37 #include <sys/systm.h> 38 #include <sys/conf.h> 39 #include <sys/kernel.h> 40 #include <sys/limits.h> 41 #include <sys/malloc.h> 42 #include <sys/mbuf.h> 43 #include <sys/module.h> 44 #include <sys/time.h> 45 #include <sys/priv.h> 46 #include <sys/proc.h> 47 #include <sys/sysctl.h> 48 #include <sys/syslog.h> 49 #include <sys/signalvar.h> 50 #include <sys/filio.h> 51 #include <sys/sockio.h> 52 53 #include <sys/socket.h> 54 #include <sys/vnode.h> 55 #include <sys/vimage.h> 56 57 #include <machine/stdarg.h> 58 59 #include <net/bpf.h> 60 #include <net/ethernet.h> 61 #include <net/fddi.h> 62 #include <net/iso88025.h> 63 #include <net/if.h> 64 #include <net/if_clone.h> 65 #include <net/if_dl.h> 66 #include <net/if_types.h> 67 #include <net/route.h> 68 69 #ifdef INET 70 #include <netinet/in.h> 71 #include <netinet/in_var.h> 72 #include <netinet/in_systm.h> 73 #include <netinet/ip.h> 74 #include <netinet/ip_var.h> 75 #include <netinet/if_ether.h> 76 #include <machine/in_cksum.h> 77 #include <netinet/vinet.h> 78 #endif 79 80 #ifdef INET6 81 #include <netinet/icmp6.h> 82 #include <netinet/ip6.h> 83 #include <netinet6/ip6_var.h> 84 #include <netinet6/scope6_var.h> 85 #include <netinet6/nd6.h> 86 #include <netinet6/vinet6.h> 87 #endif 88 89 #include <crypto/sha1.h> 90 #include <netinet/ip_carp.h> 91 92 #define CARP_IFNAME "carp" 93 static MALLOC_DEFINE(M_CARP, "CARP", "CARP interfaces"); 94 SYSCTL_DECL(_net_inet_carp); 95 96 struct carp_softc { 97 struct ifnet *sc_ifp; /* Interface clue */ 98 struct ifnet *sc_carpdev; /* Pointer to parent interface */ 99 struct in_ifaddr *sc_ia; /* primary iface address */ 100 struct ip_moptions sc_imo; 101 #ifdef INET6 102 struct in6_ifaddr *sc_ia6; /* primary iface address v6 */ 103 struct ip6_moptions sc_im6o; 104 #endif /* INET6 */ 105 TAILQ_ENTRY(carp_softc) sc_list; 106 107 enum { INIT = 0, BACKUP, MASTER } sc_state; 108 109 int sc_flags_backup; 110 int sc_suppress; 111 112 int sc_sendad_errors; 113 #define CARP_SENDAD_MAX_ERRORS 3 114 int sc_sendad_success; 115 #define CARP_SENDAD_MIN_SUCCESS 3 116 117 int sc_vhid; 118 int sc_advskew; 119 int sc_naddrs; 120 int sc_naddrs6; 121 int sc_advbase; /* seconds */ 122 int sc_init_counter; 123 u_int64_t sc_counter; 124 125 /* authentication */ 126 #define CARP_HMAC_PAD 64 127 unsigned char sc_key[CARP_KEY_LEN]; 128 unsigned char sc_pad[CARP_HMAC_PAD]; 129 SHA1_CTX sc_sha1; 130 131 struct callout sc_ad_tmo; /* advertisement timeout */ 132 struct callout sc_md_tmo; /* master down timeout */ 133 struct callout sc_md6_tmo; /* master down timeout */ 134 135 LIST_ENTRY(carp_softc) sc_next; /* Interface clue */ 136 }; 137 #define SC2IFP(sc) ((sc)->sc_ifp) 138 139 int carp_suppress_preempt = 0; 140 int carp_opts[CARPCTL_MAXID] = { 0, 1, 0, 1, 0, 0 }; /* XXX for now */ 141 SYSCTL_INT(_net_inet_carp, CARPCTL_ALLOW, allow, CTLFLAG_RW, 142 &carp_opts[CARPCTL_ALLOW], 0, "Accept incoming CARP packets"); 143 SYSCTL_INT(_net_inet_carp, CARPCTL_PREEMPT, preempt, CTLFLAG_RW, 144 &carp_opts[CARPCTL_PREEMPT], 0, "high-priority backup preemption mode"); 145 SYSCTL_INT(_net_inet_carp, CARPCTL_LOG, log, CTLFLAG_RW, 146 &carp_opts[CARPCTL_LOG], 0, "log bad carp packets"); 147 SYSCTL_INT(_net_inet_carp, CARPCTL_ARPBALANCE, arpbalance, CTLFLAG_RW, 148 &carp_opts[CARPCTL_ARPBALANCE], 0, "balance arp responses"); 149 SYSCTL_INT(_net_inet_carp, OID_AUTO, suppress_preempt, CTLFLAG_RD, 150 &carp_suppress_preempt, 0, "Preemption is suppressed"); 151 152 struct carpstats carpstats; 153 SYSCTL_STRUCT(_net_inet_carp, CARPCTL_STATS, stats, CTLFLAG_RW, 154 &carpstats, carpstats, 155 "CARP statistics (struct carpstats, netinet/ip_carp.h)"); 156 157 struct carp_if { 158 TAILQ_HEAD(, carp_softc) vhif_vrs; 159 int vhif_nvrs; 160 161 struct ifnet *vhif_ifp; 162 struct mtx vhif_mtx; 163 }; 164 165 /* Get carp_if from softc. Valid after carp_set_addr{,6}. */ 166 #define SC2CIF(sc) ((struct carp_if *)(sc)->sc_carpdev->if_carp) 167 168 /* lock per carp_if queue */ 169 #define CARP_LOCK_INIT(cif) mtx_init(&(cif)->vhif_mtx, "carp_if", \ 170 NULL, MTX_DEF) 171 #define CARP_LOCK_DESTROY(cif) mtx_destroy(&(cif)->vhif_mtx) 172 #define CARP_LOCK_ASSERT(cif) mtx_assert(&(cif)->vhif_mtx, MA_OWNED) 173 #define CARP_LOCK(cif) mtx_lock(&(cif)->vhif_mtx) 174 #define CARP_UNLOCK(cif) mtx_unlock(&(cif)->vhif_mtx) 175 176 #define CARP_SCLOCK(sc) mtx_lock(&SC2CIF(sc)->vhif_mtx) 177 #define CARP_SCUNLOCK(sc) mtx_unlock(&SC2CIF(sc)->vhif_mtx) 178 #define CARP_SCLOCK_ASSERT(sc) mtx_assert(&SC2CIF(sc)->vhif_mtx, MA_OWNED) 179 180 #define CARP_LOG(...) do { \ 181 if (carp_opts[CARPCTL_LOG] > 0) \ 182 log(LOG_INFO, __VA_ARGS__); \ 183 } while (0) 184 185 #define CARP_DEBUG(...) do { \ 186 if (carp_opts[CARPCTL_LOG] > 1) \ 187 log(LOG_DEBUG, __VA_ARGS__); \ 188 } while (0) 189 190 static void carp_hmac_prepare(struct carp_softc *); 191 static void carp_hmac_generate(struct carp_softc *, u_int32_t *, 192 unsigned char *); 193 static int carp_hmac_verify(struct carp_softc *, u_int32_t *, 194 unsigned char *); 195 static void carp_setroute(struct carp_softc *, int); 196 static void carp_input_c(struct mbuf *, struct carp_header *, sa_family_t); 197 static int carp_clone_create(struct if_clone *, int, caddr_t); 198 static void carp_clone_destroy(struct ifnet *); 199 static void carpdetach(struct carp_softc *, int); 200 static int carp_prepare_ad(struct mbuf *, struct carp_softc *, 201 struct carp_header *); 202 static void carp_send_ad_all(void); 203 static void carp_send_ad(void *); 204 static void carp_send_ad_locked(struct carp_softc *); 205 static void carp_send_arp(struct carp_softc *); 206 static void carp_master_down(void *); 207 static void carp_master_down_locked(struct carp_softc *); 208 static int carp_ioctl(struct ifnet *, u_long, caddr_t); 209 static int carp_looutput(struct ifnet *, struct mbuf *, struct sockaddr *, 210 struct route *); 211 static void carp_start(struct ifnet *); 212 static void carp_setrun(struct carp_softc *, sa_family_t); 213 static void carp_set_state(struct carp_softc *, int); 214 static int carp_addrcount(struct carp_if *, struct in_ifaddr *, int); 215 enum { CARP_COUNT_MASTER, CARP_COUNT_RUNNING }; 216 217 static void carp_multicast_cleanup(struct carp_softc *); 218 static int carp_set_addr(struct carp_softc *, struct sockaddr_in *); 219 static int carp_del_addr(struct carp_softc *, struct sockaddr_in *); 220 static void carp_carpdev_state_locked(struct carp_if *); 221 static void carp_sc_state_locked(struct carp_softc *); 222 #ifdef INET6 223 static void carp_send_na(struct carp_softc *); 224 static int carp_set_addr6(struct carp_softc *, struct sockaddr_in6 *); 225 static int carp_del_addr6(struct carp_softc *, struct sockaddr_in6 *); 226 static void carp_multicast6_cleanup(struct carp_softc *); 227 #endif 228 229 static LIST_HEAD(, carp_softc) carpif_list; 230 static struct mtx carp_mtx; 231 IFC_SIMPLE_DECLARE(carp, 0); 232 233 static eventhandler_tag if_detach_event_tag; 234 235 static __inline u_int16_t 236 carp_cksum(struct mbuf *m, int len) 237 { 238 return (in_cksum(m, len)); 239 } 240 241 static void 242 carp_hmac_prepare(struct carp_softc *sc) 243 { 244 u_int8_t version = CARP_VERSION, type = CARP_ADVERTISEMENT; 245 u_int8_t vhid = sc->sc_vhid & 0xff; 246 struct ifaddr *ifa; 247 int i, found; 248 #ifdef INET 249 struct in_addr last, cur, in; 250 #endif 251 #ifdef INET6 252 struct in6_addr last6, cur6, in6; 253 #endif 254 255 if (sc->sc_carpdev) 256 CARP_SCLOCK(sc); 257 258 /* XXX: possible race here */ 259 260 /* compute ipad from key */ 261 bzero(sc->sc_pad, sizeof(sc->sc_pad)); 262 bcopy(sc->sc_key, sc->sc_pad, sizeof(sc->sc_key)); 263 for (i = 0; i < sizeof(sc->sc_pad); i++) 264 sc->sc_pad[i] ^= 0x36; 265 266 /* precompute first part of inner hash */ 267 SHA1Init(&sc->sc_sha1); 268 SHA1Update(&sc->sc_sha1, sc->sc_pad, sizeof(sc->sc_pad)); 269 SHA1Update(&sc->sc_sha1, (void *)&version, sizeof(version)); 270 SHA1Update(&sc->sc_sha1, (void *)&type, sizeof(type)); 271 SHA1Update(&sc->sc_sha1, (void *)&vhid, sizeof(vhid)); 272 #ifdef INET 273 cur.s_addr = 0; 274 do { 275 found = 0; 276 last = cur; 277 cur.s_addr = 0xffffffff; 278 IF_ADDR_LOCK(SC2IFP(sc)); 279 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) { 280 in.s_addr = ifatoia(ifa)->ia_addr.sin_addr.s_addr; 281 if (ifa->ifa_addr->sa_family == AF_INET && 282 ntohl(in.s_addr) > ntohl(last.s_addr) && 283 ntohl(in.s_addr) < ntohl(cur.s_addr)) { 284 cur.s_addr = in.s_addr; 285 found++; 286 } 287 } 288 IF_ADDR_UNLOCK(SC2IFP(sc)); 289 if (found) 290 SHA1Update(&sc->sc_sha1, (void *)&cur, sizeof(cur)); 291 } while (found); 292 #endif /* INET */ 293 #ifdef INET6 294 memset(&cur6, 0, sizeof(cur6)); 295 do { 296 found = 0; 297 last6 = cur6; 298 memset(&cur6, 0xff, sizeof(cur6)); 299 IF_ADDR_LOCK(SC2IFP(sc)); 300 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) { 301 in6 = ifatoia6(ifa)->ia_addr.sin6_addr; 302 if (IN6_IS_SCOPE_EMBED(&in6)) 303 in6.s6_addr16[1] = 0; 304 if (ifa->ifa_addr->sa_family == AF_INET6 && 305 memcmp(&in6, &last6, sizeof(in6)) > 0 && 306 memcmp(&in6, &cur6, sizeof(in6)) < 0) { 307 cur6 = in6; 308 found++; 309 } 310 } 311 IF_ADDR_UNLOCK(SC2IFP(sc)); 312 if (found) 313 SHA1Update(&sc->sc_sha1, (void *)&cur6, sizeof(cur6)); 314 } while (found); 315 #endif /* INET6 */ 316 317 /* convert ipad to opad */ 318 for (i = 0; i < sizeof(sc->sc_pad); i++) 319 sc->sc_pad[i] ^= 0x36 ^ 0x5c; 320 321 if (sc->sc_carpdev) 322 CARP_SCUNLOCK(sc); 323 } 324 325 static void 326 carp_hmac_generate(struct carp_softc *sc, u_int32_t counter[2], 327 unsigned char md[20]) 328 { 329 SHA1_CTX sha1ctx; 330 331 /* fetch first half of inner hash */ 332 bcopy(&sc->sc_sha1, &sha1ctx, sizeof(sha1ctx)); 333 334 SHA1Update(&sha1ctx, (void *)counter, sizeof(sc->sc_counter)); 335 SHA1Final(md, &sha1ctx); 336 337 /* outer hash */ 338 SHA1Init(&sha1ctx); 339 SHA1Update(&sha1ctx, sc->sc_pad, sizeof(sc->sc_pad)); 340 SHA1Update(&sha1ctx, md, 20); 341 SHA1Final(md, &sha1ctx); 342 } 343 344 static int 345 carp_hmac_verify(struct carp_softc *sc, u_int32_t counter[2], 346 unsigned char md[20]) 347 { 348 unsigned char md2[20]; 349 350 CARP_SCLOCK_ASSERT(sc); 351 352 carp_hmac_generate(sc, counter, md2); 353 354 return (bcmp(md, md2, sizeof(md2))); 355 } 356 357 static void 358 carp_setroute(struct carp_softc *sc, int cmd) 359 { 360 struct ifaddr *ifa; 361 int s; 362 363 if (sc->sc_carpdev) 364 CARP_SCLOCK_ASSERT(sc); 365 366 s = splnet(); 367 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) { 368 if (ifa->ifa_addr->sa_family == AF_INET && 369 sc->sc_carpdev != NULL) { 370 int count = carp_addrcount( 371 (struct carp_if *)sc->sc_carpdev->if_carp, 372 ifatoia(ifa), CARP_COUNT_MASTER); 373 374 if ((cmd == RTM_ADD && count == 1) || 375 (cmd == RTM_DELETE && count == 0)) 376 rtinit(ifa, cmd, RTF_UP | RTF_HOST); 377 } 378 } 379 splx(s); 380 } 381 382 static int 383 carp_clone_create(struct if_clone *ifc, int unit, caddr_t params) 384 { 385 386 struct carp_softc *sc; 387 struct ifnet *ifp; 388 389 sc = malloc(sizeof(*sc), M_CARP, M_WAITOK|M_ZERO); 390 ifp = SC2IFP(sc) = if_alloc(IFT_ETHER); 391 if (ifp == NULL) { 392 free(sc, M_CARP); 393 return (ENOSPC); 394 } 395 396 sc->sc_flags_backup = 0; 397 sc->sc_suppress = 0; 398 sc->sc_advbase = CARP_DFLTINTV; 399 sc->sc_vhid = -1; /* required setting */ 400 sc->sc_advskew = 0; 401 sc->sc_init_counter = 1; 402 sc->sc_naddrs = sc->sc_naddrs6 = 0; /* M_ZERO? */ 403 sc->sc_imo.imo_membership = (struct in_multi **)malloc( 404 (sizeof(struct in_multi *) * IP_MIN_MEMBERSHIPS), M_CARP, 405 M_WAITOK); 406 sc->sc_imo.imo_mfilters = NULL; 407 sc->sc_imo.imo_max_memberships = IP_MIN_MEMBERSHIPS; 408 sc->sc_imo.imo_multicast_vif = -1; 409 #ifdef INET6 410 sc->sc_im6o.im6o_membership = (struct in6_multi **)malloc( 411 (sizeof(struct in6_multi *) * IPV6_MIN_MEMBERSHIPS), M_CARP, 412 M_WAITOK); 413 sc->sc_im6o.im6o_mfilters = NULL; 414 sc->sc_im6o.im6o_max_memberships = IPV6_MIN_MEMBERSHIPS; 415 sc->sc_im6o.im6o_multicast_hlim = CARP_DFLTTL; 416 #endif 417 418 callout_init(&sc->sc_ad_tmo, CALLOUT_MPSAFE); 419 callout_init(&sc->sc_md_tmo, CALLOUT_MPSAFE); 420 callout_init(&sc->sc_md6_tmo, CALLOUT_MPSAFE); 421 422 ifp->if_softc = sc; 423 if_initname(ifp, CARP_IFNAME, unit); 424 ifp->if_mtu = ETHERMTU; 425 ifp->if_flags = IFF_LOOPBACK; 426 ifp->if_ioctl = carp_ioctl; 427 ifp->if_output = carp_looutput; 428 ifp->if_start = carp_start; 429 ifp->if_type = IFT_CARP; 430 ifp->if_snd.ifq_maxlen = ifqmaxlen; 431 ifp->if_hdrlen = 0; 432 if_attach(ifp); 433 bpfattach(SC2IFP(sc), DLT_NULL, sizeof(u_int32_t)); 434 mtx_lock(&carp_mtx); 435 LIST_INSERT_HEAD(&carpif_list, sc, sc_next); 436 mtx_unlock(&carp_mtx); 437 return (0); 438 } 439 440 static void 441 carp_clone_destroy(struct ifnet *ifp) 442 { 443 struct carp_softc *sc = ifp->if_softc; 444 445 if (sc->sc_carpdev) 446 CARP_SCLOCK(sc); 447 carpdetach(sc, 1); /* Returns unlocked. */ 448 449 mtx_lock(&carp_mtx); 450 LIST_REMOVE(sc, sc_next); 451 mtx_unlock(&carp_mtx); 452 bpfdetach(ifp); 453 if_detach(ifp); 454 if_free_type(ifp, IFT_ETHER); 455 free(sc->sc_imo.imo_membership, M_CARP); 456 #ifdef INET6 457 free(sc->sc_im6o.im6o_membership, M_CARP); 458 #endif 459 free(sc, M_CARP); 460 } 461 462 /* 463 * This function can be called on CARP interface destroy path, 464 * and in case of the removal of the underlying interface as 465 * well. We differentiate these two cases. In the latter case 466 * we do not cleanup our multicast memberships, since they 467 * are already freed. Also, in the latter case we do not 468 * release the lock on return, because the function will be 469 * called once more, for another CARP instance on the same 470 * interface. 471 */ 472 static void 473 carpdetach(struct carp_softc *sc, int unlock) 474 { 475 struct carp_if *cif; 476 477 callout_stop(&sc->sc_ad_tmo); 478 callout_stop(&sc->sc_md_tmo); 479 callout_stop(&sc->sc_md6_tmo); 480 481 if (sc->sc_suppress) 482 carp_suppress_preempt--; 483 sc->sc_suppress = 0; 484 485 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) 486 carp_suppress_preempt--; 487 sc->sc_sendad_errors = 0; 488 489 carp_set_state(sc, INIT); 490 SC2IFP(sc)->if_flags &= ~IFF_UP; 491 carp_setrun(sc, 0); 492 if (unlock) 493 carp_multicast_cleanup(sc); 494 #ifdef INET6 495 carp_multicast6_cleanup(sc); 496 #endif 497 498 if (sc->sc_carpdev != NULL) { 499 cif = (struct carp_if *)sc->sc_carpdev->if_carp; 500 CARP_LOCK_ASSERT(cif); 501 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list); 502 if (!--cif->vhif_nvrs) { 503 ifpromisc(sc->sc_carpdev, 0); 504 sc->sc_carpdev->if_carp = NULL; 505 CARP_LOCK_DESTROY(cif); 506 free(cif, M_IFADDR); 507 } else if (unlock) 508 CARP_UNLOCK(cif); 509 sc->sc_carpdev = NULL; 510 } 511 } 512 513 /* Detach an interface from the carp. */ 514 static void 515 carp_ifdetach(void *arg __unused, struct ifnet *ifp) 516 { 517 struct carp_if *cif = (struct carp_if *)ifp->if_carp; 518 struct carp_softc *sc, *nextsc; 519 520 if (cif == NULL) 521 return; 522 523 /* 524 * XXX: At the end of for() cycle the lock will be destroyed. 525 */ 526 CARP_LOCK(cif); 527 for (sc = TAILQ_FIRST(&cif->vhif_vrs); sc; sc = nextsc) { 528 nextsc = TAILQ_NEXT(sc, sc_list); 529 carpdetach(sc, 0); 530 } 531 } 532 533 /* 534 * process input packet. 535 * we have rearranged checks order compared to the rfc, 536 * but it seems more efficient this way or not possible otherwise. 537 */ 538 void 539 carp_input(struct mbuf *m, int hlen) 540 { 541 struct ip *ip = mtod(m, struct ip *); 542 struct carp_header *ch; 543 int iplen, len; 544 545 CARPSTATS_INC(carps_ipackets); 546 547 if (!carp_opts[CARPCTL_ALLOW]) { 548 m_freem(m); 549 return; 550 } 551 552 /* check if received on a valid carp interface */ 553 if (m->m_pkthdr.rcvif->if_carp == NULL) { 554 CARPSTATS_INC(carps_badif); 555 CARP_LOG("carp_input: packet received on non-carp " 556 "interface: %s\n", 557 m->m_pkthdr.rcvif->if_xname); 558 m_freem(m); 559 return; 560 } 561 562 /* verify that the IP TTL is 255. */ 563 if (ip->ip_ttl != CARP_DFLTTL) { 564 CARPSTATS_INC(carps_badttl); 565 CARP_LOG("carp_input: received ttl %d != 255i on %s\n", 566 ip->ip_ttl, 567 m->m_pkthdr.rcvif->if_xname); 568 m_freem(m); 569 return; 570 } 571 572 iplen = ip->ip_hl << 2; 573 574 if (m->m_pkthdr.len < iplen + sizeof(*ch)) { 575 CARPSTATS_INC(carps_badlen); 576 CARP_LOG("carp_input: received len %zd < " 577 "sizeof(struct carp_header)\n", 578 m->m_len - sizeof(struct ip)); 579 m_freem(m); 580 return; 581 } 582 583 if (iplen + sizeof(*ch) < m->m_len) { 584 if ((m = m_pullup(m, iplen + sizeof(*ch))) == NULL) { 585 CARPSTATS_INC(carps_hdrops); 586 CARP_LOG("carp_input: pullup failed\n"); 587 return; 588 } 589 ip = mtod(m, struct ip *); 590 } 591 ch = (struct carp_header *)((char *)ip + iplen); 592 593 /* 594 * verify that the received packet length is 595 * equal to the CARP header 596 */ 597 len = iplen + sizeof(*ch); 598 if (len > m->m_pkthdr.len) { 599 CARPSTATS_INC(carps_badlen); 600 CARP_LOG("carp_input: packet too short %d on %s\n", 601 m->m_pkthdr.len, 602 m->m_pkthdr.rcvif->if_xname); 603 m_freem(m); 604 return; 605 } 606 607 if ((m = m_pullup(m, len)) == NULL) { 608 CARPSTATS_INC(carps_hdrops); 609 return; 610 } 611 ip = mtod(m, struct ip *); 612 ch = (struct carp_header *)((char *)ip + iplen); 613 614 /* verify the CARP checksum */ 615 m->m_data += iplen; 616 if (carp_cksum(m, len - iplen)) { 617 CARPSTATS_INC(carps_badsum); 618 CARP_LOG("carp_input: checksum failed on %s\n", 619 m->m_pkthdr.rcvif->if_xname); 620 m_freem(m); 621 return; 622 } 623 m->m_data -= iplen; 624 625 carp_input_c(m, ch, AF_INET); 626 } 627 628 #ifdef INET6 629 int 630 carp6_input(struct mbuf **mp, int *offp, int proto) 631 { 632 struct mbuf *m = *mp; 633 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 634 struct carp_header *ch; 635 u_int len; 636 637 CARPSTATS_INC(carps_ipackets6); 638 639 if (!carp_opts[CARPCTL_ALLOW]) { 640 m_freem(m); 641 return (IPPROTO_DONE); 642 } 643 644 /* check if received on a valid carp interface */ 645 if (m->m_pkthdr.rcvif->if_carp == NULL) { 646 CARPSTATS_INC(carps_badif); 647 CARP_LOG("carp6_input: packet received on non-carp " 648 "interface: %s\n", 649 m->m_pkthdr.rcvif->if_xname); 650 m_freem(m); 651 return (IPPROTO_DONE); 652 } 653 654 /* verify that the IP TTL is 255 */ 655 if (ip6->ip6_hlim != CARP_DFLTTL) { 656 CARPSTATS_INC(carps_badttl); 657 CARP_LOG("carp6_input: received ttl %d != 255 on %s\n", 658 ip6->ip6_hlim, 659 m->m_pkthdr.rcvif->if_xname); 660 m_freem(m); 661 return (IPPROTO_DONE); 662 } 663 664 /* verify that we have a complete carp packet */ 665 len = m->m_len; 666 IP6_EXTHDR_GET(ch, struct carp_header *, m, *offp, sizeof(*ch)); 667 if (ch == NULL) { 668 CARPSTATS_INC(carps_badlen); 669 CARP_LOG("carp6_input: packet size %u too small\n", len); 670 return (IPPROTO_DONE); 671 } 672 673 674 /* verify the CARP checksum */ 675 m->m_data += *offp; 676 if (carp_cksum(m, sizeof(*ch))) { 677 CARPSTATS_INC(carps_badsum); 678 CARP_LOG("carp6_input: checksum failed, on %s\n", 679 m->m_pkthdr.rcvif->if_xname); 680 m_freem(m); 681 return (IPPROTO_DONE); 682 } 683 m->m_data -= *offp; 684 685 carp_input_c(m, ch, AF_INET6); 686 return (IPPROTO_DONE); 687 } 688 #endif /* INET6 */ 689 690 static void 691 carp_input_c(struct mbuf *m, struct carp_header *ch, sa_family_t af) 692 { 693 struct ifnet *ifp = m->m_pkthdr.rcvif; 694 struct carp_softc *sc; 695 u_int64_t tmp_counter; 696 struct timeval sc_tv, ch_tv; 697 698 /* verify that the VHID is valid on the receiving interface */ 699 CARP_LOCK(ifp->if_carp); 700 TAILQ_FOREACH(sc, &((struct carp_if *)ifp->if_carp)->vhif_vrs, sc_list) 701 if (sc->sc_vhid == ch->carp_vhid) 702 break; 703 704 if (!sc || !((SC2IFP(sc)->if_flags & IFF_UP) && 705 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING))) { 706 CARPSTATS_INC(carps_badvhid); 707 CARP_UNLOCK(ifp->if_carp); 708 m_freem(m); 709 return; 710 } 711 712 getmicrotime(&SC2IFP(sc)->if_lastchange); 713 SC2IFP(sc)->if_ipackets++; 714 SC2IFP(sc)->if_ibytes += m->m_pkthdr.len; 715 716 if (bpf_peers_present(SC2IFP(sc)->if_bpf)) { 717 struct ip *ip = mtod(m, struct ip *); 718 uint32_t af1 = af; 719 720 /* BPF wants net byte order */ 721 ip->ip_len = htons(ip->ip_len + (ip->ip_hl << 2)); 722 ip->ip_off = htons(ip->ip_off); 723 bpf_mtap2(SC2IFP(sc)->if_bpf, &af1, sizeof(af1), m); 724 } 725 726 /* verify the CARP version. */ 727 if (ch->carp_version != CARP_VERSION) { 728 CARPSTATS_INC(carps_badver); 729 SC2IFP(sc)->if_ierrors++; 730 CARP_UNLOCK(ifp->if_carp); 731 CARP_LOG("%s; invalid version %d\n", 732 SC2IFP(sc)->if_xname, 733 ch->carp_version); 734 m_freem(m); 735 return; 736 } 737 738 /* verify the hash */ 739 if (carp_hmac_verify(sc, ch->carp_counter, ch->carp_md)) { 740 CARPSTATS_INC(carps_badauth); 741 SC2IFP(sc)->if_ierrors++; 742 CARP_UNLOCK(ifp->if_carp); 743 CARP_LOG("%s: incorrect hash\n", SC2IFP(sc)->if_xname); 744 m_freem(m); 745 return; 746 } 747 748 tmp_counter = ntohl(ch->carp_counter[0]); 749 tmp_counter = tmp_counter<<32; 750 tmp_counter += ntohl(ch->carp_counter[1]); 751 752 /* XXX Replay protection goes here */ 753 754 sc->sc_init_counter = 0; 755 sc->sc_counter = tmp_counter; 756 757 sc_tv.tv_sec = sc->sc_advbase; 758 if (carp_suppress_preempt && sc->sc_advskew < 240) 759 sc_tv.tv_usec = 240 * 1000000 / 256; 760 else 761 sc_tv.tv_usec = sc->sc_advskew * 1000000 / 256; 762 ch_tv.tv_sec = ch->carp_advbase; 763 ch_tv.tv_usec = ch->carp_advskew * 1000000 / 256; 764 765 switch (sc->sc_state) { 766 case INIT: 767 break; 768 case MASTER: 769 /* 770 * If we receive an advertisement from a master who's going to 771 * be more frequent than us, go into BACKUP state. 772 */ 773 if (timevalcmp(&sc_tv, &ch_tv, >) || 774 timevalcmp(&sc_tv, &ch_tv, ==)) { 775 callout_stop(&sc->sc_ad_tmo); 776 CARP_DEBUG("%s: MASTER -> BACKUP " 777 "(more frequent advertisement received)\n", 778 SC2IFP(sc)->if_xname); 779 carp_set_state(sc, BACKUP); 780 carp_setrun(sc, 0); 781 carp_setroute(sc, RTM_DELETE); 782 } 783 break; 784 case BACKUP: 785 /* 786 * If we're pre-empting masters who advertise slower than us, 787 * and this one claims to be slower, treat him as down. 788 */ 789 if (carp_opts[CARPCTL_PREEMPT] && 790 timevalcmp(&sc_tv, &ch_tv, <)) { 791 CARP_DEBUG("%s: BACKUP -> MASTER " 792 "(preempting a slower master)\n", 793 SC2IFP(sc)->if_xname); 794 carp_master_down_locked(sc); 795 break; 796 } 797 798 /* 799 * If the master is going to advertise at such a low frequency 800 * that he's guaranteed to time out, we'd might as well just 801 * treat him as timed out now. 802 */ 803 sc_tv.tv_sec = sc->sc_advbase * 3; 804 if (timevalcmp(&sc_tv, &ch_tv, <)) { 805 CARP_DEBUG("%s: BACKUP -> MASTER " 806 "(master timed out)\n", 807 SC2IFP(sc)->if_xname); 808 carp_master_down_locked(sc); 809 break; 810 } 811 812 /* 813 * Otherwise, we reset the counter and wait for the next 814 * advertisement. 815 */ 816 carp_setrun(sc, af); 817 break; 818 } 819 820 CARP_UNLOCK(ifp->if_carp); 821 822 m_freem(m); 823 return; 824 } 825 826 static int 827 carp_prepare_ad(struct mbuf *m, struct carp_softc *sc, struct carp_header *ch) 828 { 829 struct m_tag *mtag; 830 struct ifnet *ifp = SC2IFP(sc); 831 832 if (sc->sc_init_counter) { 833 /* this could also be seconds since unix epoch */ 834 sc->sc_counter = arc4random(); 835 sc->sc_counter = sc->sc_counter << 32; 836 sc->sc_counter += arc4random(); 837 } else 838 sc->sc_counter++; 839 840 ch->carp_counter[0] = htonl((sc->sc_counter>>32)&0xffffffff); 841 ch->carp_counter[1] = htonl(sc->sc_counter&0xffffffff); 842 843 carp_hmac_generate(sc, ch->carp_counter, ch->carp_md); 844 845 /* Tag packet for carp_output */ 846 mtag = m_tag_get(PACKET_TAG_CARP, sizeof(struct ifnet *), M_NOWAIT); 847 if (mtag == NULL) { 848 m_freem(m); 849 SC2IFP(sc)->if_oerrors++; 850 return (ENOMEM); 851 } 852 bcopy(&ifp, (caddr_t)(mtag + 1), sizeof(struct ifnet *)); 853 m_tag_prepend(m, mtag); 854 855 return (0); 856 } 857 858 static void 859 carp_send_ad_all(void) 860 { 861 struct carp_softc *sc; 862 863 mtx_lock(&carp_mtx); 864 LIST_FOREACH(sc, &carpif_list, sc_next) { 865 if (sc->sc_carpdev == NULL) 866 continue; 867 CARP_SCLOCK(sc); 868 if ((SC2IFP(sc)->if_flags & IFF_UP) && 869 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING) && 870 sc->sc_state == MASTER) 871 carp_send_ad_locked(sc); 872 CARP_SCUNLOCK(sc); 873 } 874 mtx_unlock(&carp_mtx); 875 } 876 877 static void 878 carp_send_ad(void *v) 879 { 880 struct carp_softc *sc = v; 881 882 CARP_SCLOCK(sc); 883 carp_send_ad_locked(sc); 884 CARP_SCUNLOCK(sc); 885 } 886 887 static void 888 carp_send_ad_locked(struct carp_softc *sc) 889 { 890 struct carp_header ch; 891 struct timeval tv; 892 struct carp_header *ch_ptr; 893 struct mbuf *m; 894 int len, advbase, advskew; 895 896 CARP_SCLOCK_ASSERT(sc); 897 898 /* bow out if we've lost our UPness or RUNNINGuiness */ 899 if (!((SC2IFP(sc)->if_flags & IFF_UP) && 900 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING))) { 901 advbase = 255; 902 advskew = 255; 903 } else { 904 advbase = sc->sc_advbase; 905 if (!carp_suppress_preempt || sc->sc_advskew > 240) 906 advskew = sc->sc_advskew; 907 else 908 advskew = 240; 909 tv.tv_sec = advbase; 910 tv.tv_usec = advskew * 1000000 / 256; 911 } 912 913 ch.carp_version = CARP_VERSION; 914 ch.carp_type = CARP_ADVERTISEMENT; 915 ch.carp_vhid = sc->sc_vhid; 916 ch.carp_advbase = advbase; 917 ch.carp_advskew = advskew; 918 ch.carp_authlen = 7; /* XXX DEFINE */ 919 ch.carp_pad1 = 0; /* must be zero */ 920 ch.carp_cksum = 0; 921 922 #ifdef INET 923 INIT_VNET_INET(curvnet); 924 if (sc->sc_ia) { 925 struct ip *ip; 926 927 MGETHDR(m, M_DONTWAIT, MT_HEADER); 928 if (m == NULL) { 929 SC2IFP(sc)->if_oerrors++; 930 CARPSTATS_INC(carps_onomem); 931 /* XXX maybe less ? */ 932 if (advbase != 255 || advskew != 255) 933 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv), 934 carp_send_ad, sc); 935 return; 936 } 937 len = sizeof(*ip) + sizeof(ch); 938 m->m_pkthdr.len = len; 939 m->m_pkthdr.rcvif = NULL; 940 m->m_len = len; 941 MH_ALIGN(m, m->m_len); 942 m->m_flags |= M_MCAST; 943 ip = mtod(m, struct ip *); 944 ip->ip_v = IPVERSION; 945 ip->ip_hl = sizeof(*ip) >> 2; 946 ip->ip_tos = IPTOS_LOWDELAY; 947 ip->ip_len = len; 948 ip->ip_id = ip_newid(); 949 ip->ip_off = IP_DF; 950 ip->ip_ttl = CARP_DFLTTL; 951 ip->ip_p = IPPROTO_CARP; 952 ip->ip_sum = 0; 953 ip->ip_src.s_addr = sc->sc_ia->ia_addr.sin_addr.s_addr; 954 ip->ip_dst.s_addr = htonl(INADDR_CARP_GROUP); 955 956 ch_ptr = (struct carp_header *)(&ip[1]); 957 bcopy(&ch, ch_ptr, sizeof(ch)); 958 if (carp_prepare_ad(m, sc, ch_ptr)) 959 return; 960 961 m->m_data += sizeof(*ip); 962 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip)); 963 m->m_data -= sizeof(*ip); 964 965 getmicrotime(&SC2IFP(sc)->if_lastchange); 966 SC2IFP(sc)->if_opackets++; 967 SC2IFP(sc)->if_obytes += len; 968 CARPSTATS_INC(carps_opackets); 969 970 if (ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, NULL)) { 971 SC2IFP(sc)->if_oerrors++; 972 if (sc->sc_sendad_errors < INT_MAX) 973 sc->sc_sendad_errors++; 974 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) { 975 carp_suppress_preempt++; 976 if (carp_suppress_preempt == 1) { 977 CARP_SCUNLOCK(sc); 978 carp_send_ad_all(); 979 CARP_SCLOCK(sc); 980 } 981 } 982 sc->sc_sendad_success = 0; 983 } else { 984 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) { 985 if (++sc->sc_sendad_success >= 986 CARP_SENDAD_MIN_SUCCESS) { 987 carp_suppress_preempt--; 988 sc->sc_sendad_errors = 0; 989 } 990 } else 991 sc->sc_sendad_errors = 0; 992 } 993 } 994 #endif /* INET */ 995 #ifdef INET6 996 if (sc->sc_ia6) { 997 struct ip6_hdr *ip6; 998 999 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1000 if (m == NULL) { 1001 SC2IFP(sc)->if_oerrors++; 1002 CARPSTATS_INC(carps_onomem); 1003 /* XXX maybe less ? */ 1004 if (advbase != 255 || advskew != 255) 1005 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv), 1006 carp_send_ad, sc); 1007 return; 1008 } 1009 len = sizeof(*ip6) + sizeof(ch); 1010 m->m_pkthdr.len = len; 1011 m->m_pkthdr.rcvif = NULL; 1012 m->m_len = len; 1013 MH_ALIGN(m, m->m_len); 1014 m->m_flags |= M_MCAST; 1015 ip6 = mtod(m, struct ip6_hdr *); 1016 bzero(ip6, sizeof(*ip6)); 1017 ip6->ip6_vfc |= IPV6_VERSION; 1018 ip6->ip6_hlim = CARP_DFLTTL; 1019 ip6->ip6_nxt = IPPROTO_CARP; 1020 bcopy(&sc->sc_ia6->ia_addr.sin6_addr, &ip6->ip6_src, 1021 sizeof(struct in6_addr)); 1022 /* set the multicast destination */ 1023 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) { 1027 SC2IFP(sc)->if_oerrors++; 1028 m_freem(m); 1029 CARP_LOG("%s: in6_setscope failed\n", __func__); 1030 return; 1031 } 1032 1033 ch_ptr = (struct carp_header *)(&ip6[1]); 1034 bcopy(&ch, ch_ptr, sizeof(ch)); 1035 if (carp_prepare_ad(m, sc, ch_ptr)) 1036 return; 1037 1038 m->m_data += sizeof(*ip6); 1039 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip6)); 1040 m->m_data -= sizeof(*ip6); 1041 1042 getmicrotime(&SC2IFP(sc)->if_lastchange); 1043 SC2IFP(sc)->if_opackets++; 1044 SC2IFP(sc)->if_obytes += len; 1045 CARPSTATS_INC(carps_opackets6); 1046 1047 if (ip6_output(m, NULL, NULL, 0, &sc->sc_im6o, NULL, NULL)) { 1048 SC2IFP(sc)->if_oerrors++; 1049 if (sc->sc_sendad_errors < INT_MAX) 1050 sc->sc_sendad_errors++; 1051 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) { 1052 carp_suppress_preempt++; 1053 if (carp_suppress_preempt == 1) { 1054 CARP_SCUNLOCK(sc); 1055 carp_send_ad_all(); 1056 CARP_SCLOCK(sc); 1057 } 1058 } 1059 sc->sc_sendad_success = 0; 1060 } else { 1061 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) { 1062 if (++sc->sc_sendad_success >= 1063 CARP_SENDAD_MIN_SUCCESS) { 1064 carp_suppress_preempt--; 1065 sc->sc_sendad_errors = 0; 1066 } 1067 } else 1068 sc->sc_sendad_errors = 0; 1069 } 1070 } 1071 #endif /* INET6 */ 1072 1073 if (advbase != 255 || advskew != 255) 1074 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv), 1075 carp_send_ad, sc); 1076 1077 } 1078 1079 /* 1080 * Broadcast a gratuitous ARP request containing 1081 * the virtual router MAC address for each IP address 1082 * associated with the virtual router. 1083 */ 1084 static void 1085 carp_send_arp(struct carp_softc *sc) 1086 { 1087 struct ifaddr *ifa; 1088 1089 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) { 1090 1091 if (ifa->ifa_addr->sa_family != AF_INET) 1092 continue; 1093 1094 /* arprequest(sc->sc_carpdev, &in, &in, IF_LLADDR(sc->sc_ifp)); */ 1095 arp_ifinit2(sc->sc_carpdev, ifa, IF_LLADDR(sc->sc_ifp)); 1096 1097 DELAY(1000); /* XXX */ 1098 } 1099 } 1100 1101 #ifdef INET6 1102 static void 1103 carp_send_na(struct carp_softc *sc) 1104 { 1105 struct ifaddr *ifa; 1106 struct in6_addr *in6; 1107 static struct in6_addr mcast = IN6ADDR_LINKLOCAL_ALLNODES_INIT; 1108 1109 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) { 1110 1111 if (ifa->ifa_addr->sa_family != AF_INET6) 1112 continue; 1113 1114 in6 = &ifatoia6(ifa)->ia_addr.sin6_addr; 1115 nd6_na_output(sc->sc_carpdev, &mcast, in6, 1116 ND_NA_FLAG_OVERRIDE, 1, NULL); 1117 DELAY(1000); /* XXX */ 1118 } 1119 } 1120 #endif /* INET6 */ 1121 1122 static int 1123 carp_addrcount(struct carp_if *cif, struct in_ifaddr *ia, int type) 1124 { 1125 struct carp_softc *vh; 1126 struct ifaddr *ifa; 1127 int count = 0; 1128 1129 CARP_LOCK_ASSERT(cif); 1130 1131 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1132 if ((type == CARP_COUNT_RUNNING && 1133 (SC2IFP(vh)->if_flags & IFF_UP) && 1134 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING)) || 1135 (type == CARP_COUNT_MASTER && vh->sc_state == MASTER)) { 1136 IF_ADDR_LOCK(SC2IFP(vh)); 1137 TAILQ_FOREACH(ifa, &SC2IFP(vh)->if_addrlist, 1138 ifa_list) { 1139 if (ifa->ifa_addr->sa_family == AF_INET && 1140 ia->ia_addr.sin_addr.s_addr == 1141 ifatoia(ifa)->ia_addr.sin_addr.s_addr) 1142 count++; 1143 } 1144 IF_ADDR_UNLOCK(SC2IFP(vh)); 1145 } 1146 } 1147 return (count); 1148 } 1149 1150 int 1151 carp_iamatch(void *v, struct in_ifaddr *ia, 1152 struct in_addr *isaddr, u_int8_t **enaddr) 1153 { 1154 struct carp_if *cif = v; 1155 struct carp_softc *vh; 1156 int index, count = 0; 1157 struct ifaddr *ifa; 1158 1159 CARP_LOCK(cif); 1160 1161 if (carp_opts[CARPCTL_ARPBALANCE]) { 1162 /* 1163 * XXX proof of concept implementation. 1164 * We use the source ip to decide which virtual host should 1165 * handle the request. If we're master of that virtual host, 1166 * then we respond, otherwise, just drop the arp packet on 1167 * the floor. 1168 */ 1169 count = carp_addrcount(cif, ia, CARP_COUNT_RUNNING); 1170 if (count == 0) { 1171 /* should never reach this */ 1172 CARP_UNLOCK(cif); 1173 return (0); 1174 } 1175 1176 /* this should be a hash, like pf_hash() */ 1177 index = ntohl(isaddr->s_addr) % count; 1178 count = 0; 1179 1180 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1181 if ((SC2IFP(vh)->if_flags & IFF_UP) && 1182 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING)) { 1183 IF_ADDR_LOCK(SC2IFP(vh)); 1184 TAILQ_FOREACH(ifa, &SC2IFP(vh)->if_addrlist, 1185 ifa_list) { 1186 if (ifa->ifa_addr->sa_family == 1187 AF_INET && 1188 ia->ia_addr.sin_addr.s_addr == 1189 ifatoia(ifa)->ia_addr.sin_addr.s_addr) { 1190 if (count == index) { 1191 if (vh->sc_state == 1192 MASTER) { 1193 *enaddr = IF_LLADDR(vh->sc_ifp); 1194 IF_ADDR_UNLOCK(SC2IFP(vh)); 1195 CARP_UNLOCK(cif); 1196 return (1); 1197 } else { 1198 IF_ADDR_UNLOCK(SC2IFP(vh)); 1199 CARP_UNLOCK(cif); 1200 return (0); 1201 } 1202 } 1203 count++; 1204 } 1205 } 1206 IF_ADDR_UNLOCK(SC2IFP(vh)); 1207 } 1208 } 1209 } else { 1210 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1211 if ((SC2IFP(vh)->if_flags & IFF_UP) && 1212 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING) && 1213 ia->ia_ifp == SC2IFP(vh) && 1214 vh->sc_state == MASTER) { 1215 *enaddr = IF_LLADDR(vh->sc_ifp); 1216 CARP_UNLOCK(cif); 1217 return (1); 1218 } 1219 } 1220 } 1221 CARP_UNLOCK(cif); 1222 return (0); 1223 } 1224 1225 #ifdef INET6 1226 struct ifaddr * 1227 carp_iamatch6(void *v, struct in6_addr *taddr) 1228 { 1229 struct carp_if *cif = v; 1230 struct carp_softc *vh; 1231 struct ifaddr *ifa; 1232 1233 CARP_LOCK(cif); 1234 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1235 IF_ADDR_LOCK(SC2IFP(vh)); 1236 TAILQ_FOREACH(ifa, &SC2IFP(vh)->if_addrlist, ifa_list) { 1237 if (IN6_ARE_ADDR_EQUAL(taddr, 1238 &ifatoia6(ifa)->ia_addr.sin6_addr) && 1239 (SC2IFP(vh)->if_flags & IFF_UP) && 1240 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING) && 1241 vh->sc_state == MASTER) { 1242 IF_ADDR_UNLOCK(SC2IFP(vh)); 1243 CARP_UNLOCK(cif); 1244 return (ifa); 1245 } 1246 } 1247 IF_ADDR_UNLOCK(SC2IFP(vh)); 1248 } 1249 CARP_UNLOCK(cif); 1250 1251 return (NULL); 1252 } 1253 1254 void * 1255 carp_macmatch6(void *v, struct mbuf *m, const struct in6_addr *taddr) 1256 { 1257 struct m_tag *mtag; 1258 struct carp_if *cif = v; 1259 struct carp_softc *sc; 1260 struct ifaddr *ifa; 1261 1262 CARP_LOCK(cif); 1263 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list) { 1264 IF_ADDR_LOCK(SC2IFP(sc)); 1265 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) { 1266 if (IN6_ARE_ADDR_EQUAL(taddr, 1267 &ifatoia6(ifa)->ia_addr.sin6_addr) && 1268 (SC2IFP(sc)->if_flags & IFF_UP) && 1269 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING)) { 1270 struct ifnet *ifp = SC2IFP(sc); 1271 mtag = m_tag_get(PACKET_TAG_CARP, 1272 sizeof(struct ifnet *), M_NOWAIT); 1273 if (mtag == NULL) { 1274 /* better a bit than nothing */ 1275 IF_ADDR_UNLOCK(SC2IFP(sc)); 1276 CARP_UNLOCK(cif); 1277 return (IF_LLADDR(sc->sc_ifp)); 1278 } 1279 bcopy(&ifp, (caddr_t)(mtag + 1), 1280 sizeof(struct ifnet *)); 1281 m_tag_prepend(m, mtag); 1282 1283 IF_ADDR_UNLOCK(SC2IFP(sc)); 1284 CARP_UNLOCK(cif); 1285 return (IF_LLADDR(sc->sc_ifp)); 1286 } 1287 } 1288 IF_ADDR_UNLOCK(SC2IFP(sc)); 1289 } 1290 CARP_UNLOCK(cif); 1291 1292 return (NULL); 1293 } 1294 #endif 1295 1296 struct ifnet * 1297 carp_forus(void *v, void *dhost) 1298 { 1299 struct carp_if *cif = v; 1300 struct carp_softc *vh; 1301 u_int8_t *ena = dhost; 1302 1303 if (ena[0] || ena[1] || ena[2] != 0x5e || ena[3] || ena[4] != 1) 1304 return (NULL); 1305 1306 CARP_LOCK(cif); 1307 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) 1308 if ((SC2IFP(vh)->if_flags & IFF_UP) && 1309 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING) && 1310 vh->sc_state == MASTER && 1311 !bcmp(dhost, IF_LLADDR(vh->sc_ifp), ETHER_ADDR_LEN)) { 1312 CARP_UNLOCK(cif); 1313 return (SC2IFP(vh)); 1314 } 1315 1316 CARP_UNLOCK(cif); 1317 return (NULL); 1318 } 1319 1320 static void 1321 carp_master_down(void *v) 1322 { 1323 struct carp_softc *sc = v; 1324 1325 CARP_SCLOCK(sc); 1326 carp_master_down_locked(sc); 1327 CARP_SCUNLOCK(sc); 1328 } 1329 1330 static void 1331 carp_master_down_locked(struct carp_softc *sc) 1332 { 1333 if (sc->sc_carpdev) 1334 CARP_SCLOCK_ASSERT(sc); 1335 1336 switch (sc->sc_state) { 1337 case INIT: 1338 printf("%s: master_down event in INIT state\n", 1339 SC2IFP(sc)->if_xname); 1340 break; 1341 case MASTER: 1342 break; 1343 case BACKUP: 1344 carp_set_state(sc, MASTER); 1345 carp_send_ad_locked(sc); 1346 carp_send_arp(sc); 1347 #ifdef INET6 1348 carp_send_na(sc); 1349 #endif /* INET6 */ 1350 carp_setrun(sc, 0); 1351 carp_setroute(sc, RTM_ADD); 1352 break; 1353 } 1354 } 1355 1356 /* 1357 * When in backup state, af indicates whether to reset the master down timer 1358 * for v4 or v6. If it's set to zero, reset the ones which are already pending. 1359 */ 1360 static void 1361 carp_setrun(struct carp_softc *sc, sa_family_t af) 1362 { 1363 struct timeval tv; 1364 1365 if (sc->sc_carpdev == NULL) { 1366 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING; 1367 carp_set_state(sc, INIT); 1368 return; 1369 } else 1370 CARP_SCLOCK_ASSERT(sc); 1371 1372 if (SC2IFP(sc)->if_flags & IFF_UP && 1373 sc->sc_vhid > 0 && (sc->sc_naddrs || sc->sc_naddrs6)) 1374 SC2IFP(sc)->if_drv_flags |= IFF_DRV_RUNNING; 1375 else { 1376 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING; 1377 carp_setroute(sc, RTM_DELETE); 1378 return; 1379 } 1380 1381 switch (sc->sc_state) { 1382 case INIT: 1383 if (carp_opts[CARPCTL_PREEMPT] && !carp_suppress_preempt) { 1384 carp_send_ad_locked(sc); 1385 carp_send_arp(sc); 1386 #ifdef INET6 1387 carp_send_na(sc); 1388 #endif /* INET6 */ 1389 CARP_DEBUG("%s: INIT -> MASTER (preempting)\n", 1390 SC2IFP(sc)->if_xname); 1391 carp_set_state(sc, MASTER); 1392 carp_setroute(sc, RTM_ADD); 1393 } else { 1394 CARP_DEBUG("%s: INIT -> BACKUP\n", SC2IFP(sc)->if_xname); 1395 carp_set_state(sc, BACKUP); 1396 carp_setroute(sc, RTM_DELETE); 1397 carp_setrun(sc, 0); 1398 } 1399 break; 1400 case BACKUP: 1401 callout_stop(&sc->sc_ad_tmo); 1402 tv.tv_sec = 3 * sc->sc_advbase; 1403 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1404 switch (af) { 1405 #ifdef INET 1406 case AF_INET: 1407 callout_reset(&sc->sc_md_tmo, tvtohz(&tv), 1408 carp_master_down, sc); 1409 break; 1410 #endif /* INET */ 1411 #ifdef INET6 1412 case AF_INET6: 1413 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv), 1414 carp_master_down, sc); 1415 break; 1416 #endif /* INET6 */ 1417 default: 1418 if (sc->sc_naddrs) 1419 callout_reset(&sc->sc_md_tmo, tvtohz(&tv), 1420 carp_master_down, sc); 1421 if (sc->sc_naddrs6) 1422 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv), 1423 carp_master_down, sc); 1424 break; 1425 } 1426 break; 1427 case MASTER: 1428 tv.tv_sec = sc->sc_advbase; 1429 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1430 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv), 1431 carp_send_ad, sc); 1432 break; 1433 } 1434 } 1435 1436 static void 1437 carp_multicast_cleanup(struct carp_softc *sc) 1438 { 1439 struct ip_moptions *imo = &sc->sc_imo; 1440 u_int16_t n = imo->imo_num_memberships; 1441 1442 /* Clean up our own multicast memberships */ 1443 while (n-- > 0) { 1444 if (imo->imo_membership[n] != NULL) { 1445 in_delmulti(imo->imo_membership[n]); 1446 imo->imo_membership[n] = NULL; 1447 } 1448 } 1449 KASSERT(imo->imo_mfilters == NULL, 1450 ("%s: imo_mfilters != NULL", __func__)); 1451 imo->imo_num_memberships = 0; 1452 imo->imo_multicast_ifp = NULL; 1453 } 1454 1455 #ifdef INET6 1456 static void 1457 carp_multicast6_cleanup(struct carp_softc *sc) 1458 { 1459 struct ip6_moptions *im6o = &sc->sc_im6o; 1460 u_int16_t n = im6o->im6o_num_memberships; 1461 1462 while (n-- > 0) { 1463 if (im6o->im6o_membership[n] != NULL) { 1464 in6_mc_leave(im6o->im6o_membership[n], NULL); 1465 im6o->im6o_membership[n] = NULL; 1466 } 1467 } 1468 KASSERT(im6o->im6o_mfilters == NULL, 1469 ("%s: im6o_mfilters != NULL", __func__)); 1470 im6o->im6o_num_memberships = 0; 1471 im6o->im6o_multicast_ifp = NULL; 1472 } 1473 #endif 1474 1475 static int 1476 carp_set_addr(struct carp_softc *sc, struct sockaddr_in *sin) 1477 { 1478 INIT_VNET_INET(curvnet); 1479 struct ifnet *ifp; 1480 struct carp_if *cif; 1481 struct in_ifaddr *ia, *ia_if; 1482 struct ip_moptions *imo = &sc->sc_imo; 1483 struct in_addr addr; 1484 u_long iaddr = htonl(sin->sin_addr.s_addr); 1485 int own, error; 1486 1487 if (sin->sin_addr.s_addr == 0) { 1488 if (!(SC2IFP(sc)->if_flags & IFF_UP)) 1489 carp_set_state(sc, INIT); 1490 if (sc->sc_naddrs) 1491 SC2IFP(sc)->if_flags |= IFF_UP; 1492 if (sc->sc_carpdev) 1493 CARP_SCLOCK(sc); 1494 carp_setrun(sc, 0); 1495 if (sc->sc_carpdev) 1496 CARP_SCUNLOCK(sc); 1497 return (0); 1498 } 1499 1500 /* we have to do it by hands to check we won't match on us */ 1501 ia_if = NULL; own = 0; 1502 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 1503 /* and, yeah, we need a multicast-capable iface too */ 1504 if (ia->ia_ifp != SC2IFP(sc) && 1505 (ia->ia_ifp->if_flags & IFF_MULTICAST) && 1506 (iaddr & ia->ia_subnetmask) == ia->ia_subnet) { 1507 if (!ia_if) 1508 ia_if = ia; 1509 if (sin->sin_addr.s_addr == 1510 ia->ia_addr.sin_addr.s_addr) 1511 own++; 1512 } 1513 } 1514 1515 if (!ia_if) 1516 return (EADDRNOTAVAIL); 1517 1518 ia = ia_if; 1519 ifp = ia->ia_ifp; 1520 1521 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0 || 1522 (imo->imo_multicast_ifp && imo->imo_multicast_ifp != ifp)) 1523 return (EADDRNOTAVAIL); 1524 1525 if (imo->imo_num_memberships == 0) { 1526 addr.s_addr = htonl(INADDR_CARP_GROUP); 1527 if ((imo->imo_membership[0] = in_addmulti(&addr, ifp)) == NULL) 1528 return (ENOBUFS); 1529 imo->imo_num_memberships++; 1530 imo->imo_multicast_ifp = ifp; 1531 imo->imo_multicast_ttl = CARP_DFLTTL; 1532 imo->imo_multicast_loop = 0; 1533 } 1534 1535 if (!ifp->if_carp) { 1536 1537 cif = malloc(sizeof(*cif), M_CARP, 1538 M_WAITOK|M_ZERO); 1539 if (!cif) { 1540 error = ENOBUFS; 1541 goto cleanup; 1542 } 1543 if ((error = ifpromisc(ifp, 1))) { 1544 free(cif, M_CARP); 1545 goto cleanup; 1546 } 1547 1548 CARP_LOCK_INIT(cif); 1549 CARP_LOCK(cif); 1550 cif->vhif_ifp = ifp; 1551 TAILQ_INIT(&cif->vhif_vrs); 1552 ifp->if_carp = cif; 1553 1554 } else { 1555 struct carp_softc *vr; 1556 1557 cif = (struct carp_if *)ifp->if_carp; 1558 CARP_LOCK(cif); 1559 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) 1560 if (vr != sc && vr->sc_vhid == sc->sc_vhid) { 1561 CARP_UNLOCK(cif); 1562 error = EEXIST; 1563 goto cleanup; 1564 } 1565 } 1566 sc->sc_ia = ia; 1567 sc->sc_carpdev = ifp; 1568 1569 { /* XXX prevent endless loop if already in queue */ 1570 struct carp_softc *vr, *after = NULL; 1571 int myself = 0; 1572 cif = (struct carp_if *)ifp->if_carp; 1573 1574 /* XXX: cif should not change, right? So we still hold the lock */ 1575 CARP_LOCK_ASSERT(cif); 1576 1577 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) { 1578 if (vr == sc) 1579 myself = 1; 1580 if (vr->sc_vhid < sc->sc_vhid) 1581 after = vr; 1582 } 1583 1584 if (!myself) { 1585 /* We're trying to keep things in order */ 1586 if (after == NULL) { 1587 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list); 1588 } else { 1589 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, sc, sc_list); 1590 } 1591 cif->vhif_nvrs++; 1592 } 1593 } 1594 1595 sc->sc_naddrs++; 1596 SC2IFP(sc)->if_flags |= IFF_UP; 1597 if (own) 1598 sc->sc_advskew = 0; 1599 carp_sc_state_locked(sc); 1600 carp_setrun(sc, 0); 1601 1602 CARP_UNLOCK(cif); 1603 1604 return (0); 1605 1606 cleanup: 1607 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]); 1608 return (error); 1609 } 1610 1611 static int 1612 carp_del_addr(struct carp_softc *sc, struct sockaddr_in *sin) 1613 { 1614 int error = 0; 1615 1616 if (!--sc->sc_naddrs) { 1617 struct carp_if *cif = (struct carp_if *)sc->sc_carpdev->if_carp; 1618 struct ip_moptions *imo = &sc->sc_imo; 1619 1620 CARP_LOCK(cif); 1621 callout_stop(&sc->sc_ad_tmo); 1622 SC2IFP(sc)->if_flags &= ~IFF_UP; 1623 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING; 1624 sc->sc_vhid = -1; 1625 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]); 1626 imo->imo_multicast_ifp = NULL; 1627 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list); 1628 if (!--cif->vhif_nvrs) { 1629 sc->sc_carpdev->if_carp = NULL; 1630 CARP_LOCK_DESTROY(cif); 1631 free(cif, M_IFADDR); 1632 } else { 1633 CARP_UNLOCK(cif); 1634 } 1635 } 1636 1637 return (error); 1638 } 1639 1640 #ifdef INET6 1641 static int 1642 carp_set_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6) 1643 { 1644 INIT_VNET_INET6(curvnet); 1645 struct ifnet *ifp; 1646 struct carp_if *cif; 1647 struct in6_ifaddr *ia, *ia_if; 1648 struct ip6_moptions *im6o = &sc->sc_im6o; 1649 struct in6_addr in6; 1650 int own, error; 1651 1652 error = 0; 1653 1654 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1655 if (!(SC2IFP(sc)->if_flags & IFF_UP)) 1656 carp_set_state(sc, INIT); 1657 if (sc->sc_naddrs6) 1658 SC2IFP(sc)->if_flags |= IFF_UP; 1659 if (sc->sc_carpdev) 1660 CARP_SCLOCK(sc); 1661 carp_setrun(sc, 0); 1662 if (sc->sc_carpdev) 1663 CARP_SCUNLOCK(sc); 1664 return (0); 1665 } 1666 1667 /* we have to do it by hands to check we won't match on us */ 1668 ia_if = NULL; own = 0; 1669 for (ia = V_in6_ifaddr; ia; ia = ia->ia_next) { 1670 int i; 1671 1672 for (i = 0; i < 4; i++) { 1673 if ((sin6->sin6_addr.s6_addr32[i] & 1674 ia->ia_prefixmask.sin6_addr.s6_addr32[i]) != 1675 (ia->ia_addr.sin6_addr.s6_addr32[i] & 1676 ia->ia_prefixmask.sin6_addr.s6_addr32[i])) 1677 break; 1678 } 1679 /* and, yeah, we need a multicast-capable iface too */ 1680 if (ia->ia_ifp != SC2IFP(sc) && 1681 (ia->ia_ifp->if_flags & IFF_MULTICAST) && 1682 (i == 4)) { 1683 if (!ia_if) 1684 ia_if = ia; 1685 if (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, 1686 &ia->ia_addr.sin6_addr)) 1687 own++; 1688 } 1689 } 1690 1691 if (!ia_if) 1692 return (EADDRNOTAVAIL); 1693 ia = ia_if; 1694 ifp = ia->ia_ifp; 1695 1696 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0 || 1697 (im6o->im6o_multicast_ifp && im6o->im6o_multicast_ifp != ifp)) 1698 return (EADDRNOTAVAIL); 1699 1700 if (!sc->sc_naddrs6) { 1701 struct in6_multi *in6m; 1702 1703 im6o->im6o_multicast_ifp = ifp; 1704 1705 /* join CARP multicast address */ 1706 bzero(&in6, sizeof(in6)); 1707 in6.s6_addr16[0] = htons(0xff02); 1708 in6.s6_addr8[15] = 0x12; 1709 if (in6_setscope(&in6, ifp, NULL) != 0) 1710 goto cleanup; 1711 in6m = NULL; 1712 error = in6_mc_join(ifp, &in6, NULL, &in6m, 0); 1713 if (error) 1714 goto cleanup; 1715 im6o->im6o_membership[0] = in6m; 1716 im6o->im6o_num_memberships++; 1717 1718 /* join solicited multicast address */ 1719 bzero(&in6, sizeof(in6)); 1720 in6.s6_addr16[0] = htons(0xff02); 1721 in6.s6_addr32[1] = 0; 1722 in6.s6_addr32[2] = htonl(1); 1723 in6.s6_addr32[3] = sin6->sin6_addr.s6_addr32[3]; 1724 in6.s6_addr8[12] = 0xff; 1725 if (in6_setscope(&in6, ifp, NULL) != 0) 1726 goto cleanup; 1727 in6m = NULL; 1728 error = in6_mc_join(ifp, &in6, NULL, &in6m, 0); 1729 if (error) 1730 goto cleanup; 1731 im6o->im6o_membership[1] = in6m; 1732 im6o->im6o_num_memberships++; 1733 } 1734 1735 if (!ifp->if_carp) { 1736 cif = malloc(sizeof(*cif), M_CARP, 1737 M_WAITOK|M_ZERO); 1738 if (!cif) { 1739 error = ENOBUFS; 1740 goto cleanup; 1741 } 1742 if ((error = ifpromisc(ifp, 1))) { 1743 free(cif, M_CARP); 1744 goto cleanup; 1745 } 1746 1747 CARP_LOCK_INIT(cif); 1748 CARP_LOCK(cif); 1749 cif->vhif_ifp = ifp; 1750 TAILQ_INIT(&cif->vhif_vrs); 1751 ifp->if_carp = cif; 1752 1753 } else { 1754 struct carp_softc *vr; 1755 1756 cif = (struct carp_if *)ifp->if_carp; 1757 CARP_LOCK(cif); 1758 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) 1759 if (vr != sc && vr->sc_vhid == sc->sc_vhid) { 1760 CARP_UNLOCK(cif); 1761 error = EINVAL; 1762 goto cleanup; 1763 } 1764 } 1765 sc->sc_ia6 = ia; 1766 sc->sc_carpdev = ifp; 1767 1768 { /* XXX prevent endless loop if already in queue */ 1769 struct carp_softc *vr, *after = NULL; 1770 int myself = 0; 1771 cif = (struct carp_if *)ifp->if_carp; 1772 CARP_LOCK_ASSERT(cif); 1773 1774 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) { 1775 if (vr == sc) 1776 myself = 1; 1777 if (vr->sc_vhid < sc->sc_vhid) 1778 after = vr; 1779 } 1780 1781 if (!myself) { 1782 /* We're trying to keep things in order */ 1783 if (after == NULL) { 1784 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list); 1785 } else { 1786 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, sc, sc_list); 1787 } 1788 cif->vhif_nvrs++; 1789 } 1790 } 1791 1792 sc->sc_naddrs6++; 1793 SC2IFP(sc)->if_flags |= IFF_UP; 1794 if (own) 1795 sc->sc_advskew = 0; 1796 carp_sc_state_locked(sc); 1797 carp_setrun(sc, 0); 1798 1799 CARP_UNLOCK(cif); 1800 1801 return (0); 1802 1803 cleanup: 1804 if (!sc->sc_naddrs6) 1805 carp_multicast6_cleanup(sc); 1806 return (error); 1807 } 1808 1809 static int 1810 carp_del_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6) 1811 { 1812 int error = 0; 1813 1814 if (!--sc->sc_naddrs6) { 1815 struct carp_if *cif = (struct carp_if *)sc->sc_carpdev->if_carp; 1816 1817 CARP_LOCK(cif); 1818 callout_stop(&sc->sc_ad_tmo); 1819 SC2IFP(sc)->if_flags &= ~IFF_UP; 1820 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING; 1821 sc->sc_vhid = -1; 1822 carp_multicast6_cleanup(sc); 1823 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list); 1824 if (!--cif->vhif_nvrs) { 1825 CARP_LOCK_DESTROY(cif); 1826 sc->sc_carpdev->if_carp = NULL; 1827 free(cif, M_IFADDR); 1828 } else 1829 CARP_UNLOCK(cif); 1830 } 1831 1832 return (error); 1833 } 1834 #endif /* INET6 */ 1835 1836 static int 1837 carp_ioctl(struct ifnet *ifp, u_long cmd, caddr_t addr) 1838 { 1839 struct carp_softc *sc = ifp->if_softc, *vr; 1840 struct carpreq carpr; 1841 struct ifaddr *ifa; 1842 struct ifreq *ifr; 1843 struct ifaliasreq *ifra; 1844 int locked = 0, error = 0; 1845 1846 ifa = (struct ifaddr *)addr; 1847 ifra = (struct ifaliasreq *)addr; 1848 ifr = (struct ifreq *)addr; 1849 1850 switch (cmd) { 1851 case SIOCSIFADDR: 1852 switch (ifa->ifa_addr->sa_family) { 1853 #ifdef INET 1854 case AF_INET: 1855 SC2IFP(sc)->if_flags |= IFF_UP; 1856 bcopy(ifa->ifa_addr, ifa->ifa_dstaddr, 1857 sizeof(struct sockaddr)); 1858 error = carp_set_addr(sc, satosin(ifa->ifa_addr)); 1859 break; 1860 #endif /* INET */ 1861 #ifdef INET6 1862 case AF_INET6: 1863 SC2IFP(sc)->if_flags |= IFF_UP; 1864 error = carp_set_addr6(sc, satosin6(ifa->ifa_addr)); 1865 break; 1866 #endif /* INET6 */ 1867 default: 1868 error = EAFNOSUPPORT; 1869 break; 1870 } 1871 break; 1872 1873 case SIOCAIFADDR: 1874 switch (ifa->ifa_addr->sa_family) { 1875 #ifdef INET 1876 case AF_INET: 1877 SC2IFP(sc)->if_flags |= IFF_UP; 1878 bcopy(ifa->ifa_addr, ifa->ifa_dstaddr, 1879 sizeof(struct sockaddr)); 1880 error = carp_set_addr(sc, satosin(&ifra->ifra_addr)); 1881 break; 1882 #endif /* INET */ 1883 #ifdef INET6 1884 case AF_INET6: 1885 SC2IFP(sc)->if_flags |= IFF_UP; 1886 error = carp_set_addr6(sc, satosin6(&ifra->ifra_addr)); 1887 break; 1888 #endif /* INET6 */ 1889 default: 1890 error = EAFNOSUPPORT; 1891 break; 1892 } 1893 break; 1894 1895 case SIOCDIFADDR: 1896 switch (ifa->ifa_addr->sa_family) { 1897 #ifdef INET 1898 case AF_INET: 1899 error = carp_del_addr(sc, satosin(&ifra->ifra_addr)); 1900 break; 1901 #endif /* INET */ 1902 #ifdef INET6 1903 case AF_INET6: 1904 error = carp_del_addr6(sc, satosin6(&ifra->ifra_addr)); 1905 break; 1906 #endif /* INET6 */ 1907 default: 1908 error = EAFNOSUPPORT; 1909 break; 1910 } 1911 break; 1912 1913 case SIOCSIFFLAGS: 1914 if (sc->sc_carpdev) { 1915 locked = 1; 1916 CARP_SCLOCK(sc); 1917 } 1918 if (sc->sc_state != INIT && !(ifr->ifr_flags & IFF_UP)) { 1919 callout_stop(&sc->sc_ad_tmo); 1920 callout_stop(&sc->sc_md_tmo); 1921 callout_stop(&sc->sc_md6_tmo); 1922 if (sc->sc_state == MASTER) 1923 carp_send_ad_locked(sc); 1924 carp_set_state(sc, INIT); 1925 carp_setrun(sc, 0); 1926 } else if (sc->sc_state == INIT && (ifr->ifr_flags & IFF_UP)) { 1927 SC2IFP(sc)->if_flags |= IFF_UP; 1928 carp_setrun(sc, 0); 1929 } 1930 break; 1931 1932 case SIOCSVH: 1933 error = priv_check(curthread, PRIV_NETINET_CARP); 1934 if (error) 1935 break; 1936 if ((error = copyin(ifr->ifr_data, &carpr, sizeof carpr))) 1937 break; 1938 error = 1; 1939 if (sc->sc_carpdev) { 1940 locked = 1; 1941 CARP_SCLOCK(sc); 1942 } 1943 if (sc->sc_state != INIT && carpr.carpr_state != sc->sc_state) { 1944 switch (carpr.carpr_state) { 1945 case BACKUP: 1946 callout_stop(&sc->sc_ad_tmo); 1947 carp_set_state(sc, BACKUP); 1948 carp_setrun(sc, 0); 1949 carp_setroute(sc, RTM_DELETE); 1950 break; 1951 case MASTER: 1952 carp_master_down_locked(sc); 1953 break; 1954 default: 1955 break; 1956 } 1957 } 1958 if (carpr.carpr_vhid > 0) { 1959 if (carpr.carpr_vhid > 255) { 1960 error = EINVAL; 1961 break; 1962 } 1963 if (sc->sc_carpdev) { 1964 struct carp_if *cif; 1965 cif = (struct carp_if *)sc->sc_carpdev->if_carp; 1966 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) 1967 if (vr != sc && 1968 vr->sc_vhid == carpr.carpr_vhid) { 1969 error = EEXIST; 1970 break; 1971 } 1972 if (error == EEXIST) 1973 break; 1974 } 1975 sc->sc_vhid = carpr.carpr_vhid; 1976 IF_LLADDR(sc->sc_ifp)[0] = 0; 1977 IF_LLADDR(sc->sc_ifp)[1] = 0; 1978 IF_LLADDR(sc->sc_ifp)[2] = 0x5e; 1979 IF_LLADDR(sc->sc_ifp)[3] = 0; 1980 IF_LLADDR(sc->sc_ifp)[4] = 1; 1981 IF_LLADDR(sc->sc_ifp)[5] = sc->sc_vhid; 1982 error--; 1983 } 1984 if (carpr.carpr_advbase > 0 || carpr.carpr_advskew > 0) { 1985 if (carpr.carpr_advskew >= 255) { 1986 error = EINVAL; 1987 break; 1988 } 1989 if (carpr.carpr_advbase > 255) { 1990 error = EINVAL; 1991 break; 1992 } 1993 sc->sc_advbase = carpr.carpr_advbase; 1994 sc->sc_advskew = carpr.carpr_advskew; 1995 error--; 1996 } 1997 bcopy(carpr.carpr_key, sc->sc_key, sizeof(sc->sc_key)); 1998 if (error > 0) 1999 error = EINVAL; 2000 else { 2001 error = 0; 2002 carp_setrun(sc, 0); 2003 } 2004 break; 2005 2006 case SIOCGVH: 2007 /* XXX: lockless read */ 2008 bzero(&carpr, sizeof(carpr)); 2009 carpr.carpr_state = sc->sc_state; 2010 carpr.carpr_vhid = sc->sc_vhid; 2011 carpr.carpr_advbase = sc->sc_advbase; 2012 carpr.carpr_advskew = sc->sc_advskew; 2013 error = priv_check(curthread, PRIV_NETINET_CARP); 2014 if (error == 0) 2015 bcopy(sc->sc_key, carpr.carpr_key, 2016 sizeof(carpr.carpr_key)); 2017 error = copyout(&carpr, ifr->ifr_data, sizeof(carpr)); 2018 break; 2019 2020 default: 2021 error = EINVAL; 2022 } 2023 2024 if (locked) 2025 CARP_SCUNLOCK(sc); 2026 2027 carp_hmac_prepare(sc); 2028 2029 return (error); 2030 } 2031 2032 /* 2033 * XXX: this is looutput. We should eventually use it from there. 2034 */ 2035 static int 2036 carp_looutput(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, 2037 struct route *ro) 2038 { 2039 u_int32_t af; 2040 struct rtentry *rt = NULL; 2041 2042 M_ASSERTPKTHDR(m); /* check if we have the packet header */ 2043 2044 if (ro != NULL) 2045 rt = ro->ro_rt; 2046 if (rt && rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 2047 m_freem(m); 2048 return (rt->rt_flags & RTF_BLACKHOLE ? 0 : 2049 rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); 2050 } 2051 2052 ifp->if_opackets++; 2053 ifp->if_obytes += m->m_pkthdr.len; 2054 2055 /* BPF writes need to be handled specially. */ 2056 if (dst->sa_family == AF_UNSPEC) { 2057 bcopy(dst->sa_data, &af, sizeof(af)); 2058 dst->sa_family = af; 2059 } 2060 2061 #if 1 /* XXX */ 2062 switch (dst->sa_family) { 2063 case AF_INET: 2064 case AF_INET6: 2065 case AF_IPX: 2066 case AF_APPLETALK: 2067 break; 2068 default: 2069 printf("carp_looutput: af=%d unexpected\n", dst->sa_family); 2070 m_freem(m); 2071 return (EAFNOSUPPORT); 2072 } 2073 #endif 2074 return(if_simloop(ifp, m, dst->sa_family, 0)); 2075 } 2076 2077 /* 2078 * Start output on carp interface. This function should never be called. 2079 */ 2080 static void 2081 carp_start(struct ifnet *ifp) 2082 { 2083 #ifdef DEBUG 2084 printf("%s: start called\n", ifp->if_xname); 2085 #endif 2086 } 2087 2088 int 2089 carp_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa, 2090 struct rtentry *rt) 2091 { 2092 struct m_tag *mtag; 2093 struct carp_softc *sc; 2094 struct ifnet *carp_ifp; 2095 2096 if (!sa) 2097 return (0); 2098 2099 switch (sa->sa_family) { 2100 #ifdef INET 2101 case AF_INET: 2102 break; 2103 #endif /* INET */ 2104 #ifdef INET6 2105 case AF_INET6: 2106 break; 2107 #endif /* INET6 */ 2108 default: 2109 return (0); 2110 } 2111 2112 mtag = m_tag_find(m, PACKET_TAG_CARP, NULL); 2113 if (mtag == NULL) 2114 return (0); 2115 2116 bcopy(mtag + 1, &carp_ifp, sizeof(struct ifnet *)); 2117 sc = carp_ifp->if_softc; 2118 2119 /* Set the source MAC address to Virtual Router MAC Address */ 2120 switch (ifp->if_type) { 2121 case IFT_ETHER: 2122 case IFT_L2VLAN: { 2123 struct ether_header *eh; 2124 2125 eh = mtod(m, struct ether_header *); 2126 eh->ether_shost[0] = 0; 2127 eh->ether_shost[1] = 0; 2128 eh->ether_shost[2] = 0x5e; 2129 eh->ether_shost[3] = 0; 2130 eh->ether_shost[4] = 1; 2131 eh->ether_shost[5] = sc->sc_vhid; 2132 } 2133 break; 2134 case IFT_FDDI: { 2135 struct fddi_header *fh; 2136 2137 fh = mtod(m, struct fddi_header *); 2138 fh->fddi_shost[0] = 0; 2139 fh->fddi_shost[1] = 0; 2140 fh->fddi_shost[2] = 0x5e; 2141 fh->fddi_shost[3] = 0; 2142 fh->fddi_shost[4] = 1; 2143 fh->fddi_shost[5] = sc->sc_vhid; 2144 } 2145 break; 2146 case IFT_ISO88025: { 2147 struct iso88025_header *th; 2148 th = mtod(m, struct iso88025_header *); 2149 th->iso88025_shost[0] = 3; 2150 th->iso88025_shost[1] = 0; 2151 th->iso88025_shost[2] = 0x40 >> (sc->sc_vhid - 1); 2152 th->iso88025_shost[3] = 0x40000 >> (sc->sc_vhid - 1); 2153 th->iso88025_shost[4] = 0; 2154 th->iso88025_shost[5] = 0; 2155 } 2156 break; 2157 default: 2158 printf("%s: carp is not supported for this interface type\n", 2159 ifp->if_xname); 2160 return (EOPNOTSUPP); 2161 } 2162 2163 return (0); 2164 } 2165 2166 static void 2167 carp_set_state(struct carp_softc *sc, int state) 2168 { 2169 int link_state; 2170 2171 if (sc->sc_carpdev) 2172 CARP_SCLOCK_ASSERT(sc); 2173 2174 if (sc->sc_state == state) 2175 return; 2176 2177 sc->sc_state = state; 2178 switch (state) { 2179 case BACKUP: 2180 link_state = LINK_STATE_DOWN; 2181 break; 2182 case MASTER: 2183 link_state = LINK_STATE_UP; 2184 break; 2185 default: 2186 link_state = LINK_STATE_UNKNOWN; 2187 break; 2188 } 2189 if_link_state_change(SC2IFP(sc), link_state); 2190 } 2191 2192 void 2193 carp_carpdev_state(void *v) 2194 { 2195 struct carp_if *cif = v; 2196 2197 CARP_LOCK(cif); 2198 carp_carpdev_state_locked(cif); 2199 CARP_UNLOCK(cif); 2200 } 2201 2202 static void 2203 carp_carpdev_state_locked(struct carp_if *cif) 2204 { 2205 struct carp_softc *sc; 2206 2207 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list) 2208 carp_sc_state_locked(sc); 2209 } 2210 2211 static void 2212 carp_sc_state_locked(struct carp_softc *sc) 2213 { 2214 CARP_SCLOCK_ASSERT(sc); 2215 2216 if (sc->sc_carpdev->if_link_state != LINK_STATE_UP || 2217 !(sc->sc_carpdev->if_flags & IFF_UP)) { 2218 sc->sc_flags_backup = SC2IFP(sc)->if_flags; 2219 SC2IFP(sc)->if_flags &= ~IFF_UP; 2220 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING; 2221 callout_stop(&sc->sc_ad_tmo); 2222 callout_stop(&sc->sc_md_tmo); 2223 callout_stop(&sc->sc_md6_tmo); 2224 carp_set_state(sc, INIT); 2225 carp_setrun(sc, 0); 2226 if (!sc->sc_suppress) { 2227 carp_suppress_preempt++; 2228 if (carp_suppress_preempt == 1) { 2229 CARP_SCUNLOCK(sc); 2230 carp_send_ad_all(); 2231 CARP_SCLOCK(sc); 2232 } 2233 } 2234 sc->sc_suppress = 1; 2235 } else { 2236 SC2IFP(sc)->if_flags |= sc->sc_flags_backup; 2237 carp_set_state(sc, INIT); 2238 carp_setrun(sc, 0); 2239 if (sc->sc_suppress) 2240 carp_suppress_preempt--; 2241 sc->sc_suppress = 0; 2242 } 2243 2244 return; 2245 } 2246 2247 static int 2248 carp_modevent(module_t mod, int type, void *data) 2249 { 2250 switch (type) { 2251 case MOD_LOAD: 2252 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event, 2253 carp_ifdetach, NULL, EVENTHANDLER_PRI_ANY); 2254 if (if_detach_event_tag == NULL) 2255 return (ENOMEM); 2256 mtx_init(&carp_mtx, "carp_mtx", NULL, MTX_DEF); 2257 LIST_INIT(&carpif_list); 2258 if_clone_attach(&carp_cloner); 2259 break; 2260 2261 case MOD_UNLOAD: 2262 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag); 2263 if_clone_detach(&carp_cloner); 2264 mtx_destroy(&carp_mtx); 2265 break; 2266 2267 default: 2268 return (EINVAL); 2269 } 2270 2271 return (0); 2272 } 2273 2274 static moduledata_t carp_mod = { 2275 "carp", 2276 carp_modevent, 2277 0 2278 }; 2279 2280 DECLARE_MODULE(carp, carp_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 2281