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 ifa_ref(ifa); 1243 IF_ADDR_UNLOCK(SC2IFP(vh)); 1244 CARP_UNLOCK(cif); 1245 return (ifa); 1246 } 1247 } 1248 IF_ADDR_UNLOCK(SC2IFP(vh)); 1249 } 1250 CARP_UNLOCK(cif); 1251 1252 return (NULL); 1253 } 1254 1255 void * 1256 carp_macmatch6(void *v, struct mbuf *m, const struct in6_addr *taddr) 1257 { 1258 struct m_tag *mtag; 1259 struct carp_if *cif = v; 1260 struct carp_softc *sc; 1261 struct ifaddr *ifa; 1262 1263 CARP_LOCK(cif); 1264 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list) { 1265 IF_ADDR_LOCK(SC2IFP(sc)); 1266 TAILQ_FOREACH(ifa, &SC2IFP(sc)->if_addrlist, ifa_list) { 1267 if (IN6_ARE_ADDR_EQUAL(taddr, 1268 &ifatoia6(ifa)->ia_addr.sin6_addr) && 1269 (SC2IFP(sc)->if_flags & IFF_UP) && 1270 (SC2IFP(sc)->if_drv_flags & IFF_DRV_RUNNING)) { 1271 struct ifnet *ifp = SC2IFP(sc); 1272 mtag = m_tag_get(PACKET_TAG_CARP, 1273 sizeof(struct ifnet *), M_NOWAIT); 1274 if (mtag == NULL) { 1275 /* better a bit than nothing */ 1276 IF_ADDR_UNLOCK(SC2IFP(sc)); 1277 CARP_UNLOCK(cif); 1278 return (IF_LLADDR(sc->sc_ifp)); 1279 } 1280 bcopy(&ifp, (caddr_t)(mtag + 1), 1281 sizeof(struct ifnet *)); 1282 m_tag_prepend(m, mtag); 1283 1284 IF_ADDR_UNLOCK(SC2IFP(sc)); 1285 CARP_UNLOCK(cif); 1286 return (IF_LLADDR(sc->sc_ifp)); 1287 } 1288 } 1289 IF_ADDR_UNLOCK(SC2IFP(sc)); 1290 } 1291 CARP_UNLOCK(cif); 1292 1293 return (NULL); 1294 } 1295 #endif 1296 1297 struct ifnet * 1298 carp_forus(void *v, void *dhost) 1299 { 1300 struct carp_if *cif = v; 1301 struct carp_softc *vh; 1302 u_int8_t *ena = dhost; 1303 1304 if (ena[0] || ena[1] || ena[2] != 0x5e || ena[3] || ena[4] != 1) 1305 return (NULL); 1306 1307 CARP_LOCK(cif); 1308 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) 1309 if ((SC2IFP(vh)->if_flags & IFF_UP) && 1310 (SC2IFP(vh)->if_drv_flags & IFF_DRV_RUNNING) && 1311 vh->sc_state == MASTER && 1312 !bcmp(dhost, IF_LLADDR(vh->sc_ifp), ETHER_ADDR_LEN)) { 1313 CARP_UNLOCK(cif); 1314 return (SC2IFP(vh)); 1315 } 1316 1317 CARP_UNLOCK(cif); 1318 return (NULL); 1319 } 1320 1321 static void 1322 carp_master_down(void *v) 1323 { 1324 struct carp_softc *sc = v; 1325 1326 CARP_SCLOCK(sc); 1327 carp_master_down_locked(sc); 1328 CARP_SCUNLOCK(sc); 1329 } 1330 1331 static void 1332 carp_master_down_locked(struct carp_softc *sc) 1333 { 1334 if (sc->sc_carpdev) 1335 CARP_SCLOCK_ASSERT(sc); 1336 1337 switch (sc->sc_state) { 1338 case INIT: 1339 printf("%s: master_down event in INIT state\n", 1340 SC2IFP(sc)->if_xname); 1341 break; 1342 case MASTER: 1343 break; 1344 case BACKUP: 1345 carp_set_state(sc, MASTER); 1346 carp_send_ad_locked(sc); 1347 carp_send_arp(sc); 1348 #ifdef INET6 1349 carp_send_na(sc); 1350 #endif /* INET6 */ 1351 carp_setrun(sc, 0); 1352 carp_setroute(sc, RTM_ADD); 1353 break; 1354 } 1355 } 1356 1357 /* 1358 * When in backup state, af indicates whether to reset the master down timer 1359 * for v4 or v6. If it's set to zero, reset the ones which are already pending. 1360 */ 1361 static void 1362 carp_setrun(struct carp_softc *sc, sa_family_t af) 1363 { 1364 struct timeval tv; 1365 1366 if (sc->sc_carpdev == NULL) { 1367 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING; 1368 carp_set_state(sc, INIT); 1369 return; 1370 } else 1371 CARP_SCLOCK_ASSERT(sc); 1372 1373 if (SC2IFP(sc)->if_flags & IFF_UP && 1374 sc->sc_vhid > 0 && (sc->sc_naddrs || sc->sc_naddrs6)) 1375 SC2IFP(sc)->if_drv_flags |= IFF_DRV_RUNNING; 1376 else { 1377 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING; 1378 carp_setroute(sc, RTM_DELETE); 1379 return; 1380 } 1381 1382 switch (sc->sc_state) { 1383 case INIT: 1384 if (carp_opts[CARPCTL_PREEMPT] && !carp_suppress_preempt) { 1385 carp_send_ad_locked(sc); 1386 carp_send_arp(sc); 1387 #ifdef INET6 1388 carp_send_na(sc); 1389 #endif /* INET6 */ 1390 CARP_DEBUG("%s: INIT -> MASTER (preempting)\n", 1391 SC2IFP(sc)->if_xname); 1392 carp_set_state(sc, MASTER); 1393 carp_setroute(sc, RTM_ADD); 1394 } else { 1395 CARP_DEBUG("%s: INIT -> BACKUP\n", SC2IFP(sc)->if_xname); 1396 carp_set_state(sc, BACKUP); 1397 carp_setroute(sc, RTM_DELETE); 1398 carp_setrun(sc, 0); 1399 } 1400 break; 1401 case BACKUP: 1402 callout_stop(&sc->sc_ad_tmo); 1403 tv.tv_sec = 3 * sc->sc_advbase; 1404 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1405 switch (af) { 1406 #ifdef INET 1407 case AF_INET: 1408 callout_reset(&sc->sc_md_tmo, tvtohz(&tv), 1409 carp_master_down, sc); 1410 break; 1411 #endif /* INET */ 1412 #ifdef INET6 1413 case AF_INET6: 1414 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv), 1415 carp_master_down, sc); 1416 break; 1417 #endif /* INET6 */ 1418 default: 1419 if (sc->sc_naddrs) 1420 callout_reset(&sc->sc_md_tmo, tvtohz(&tv), 1421 carp_master_down, sc); 1422 if (sc->sc_naddrs6) 1423 callout_reset(&sc->sc_md6_tmo, tvtohz(&tv), 1424 carp_master_down, sc); 1425 break; 1426 } 1427 break; 1428 case MASTER: 1429 tv.tv_sec = sc->sc_advbase; 1430 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1431 callout_reset(&sc->sc_ad_tmo, tvtohz(&tv), 1432 carp_send_ad, sc); 1433 break; 1434 } 1435 } 1436 1437 static void 1438 carp_multicast_cleanup(struct carp_softc *sc) 1439 { 1440 struct ip_moptions *imo = &sc->sc_imo; 1441 u_int16_t n = imo->imo_num_memberships; 1442 1443 /* Clean up our own multicast memberships */ 1444 while (n-- > 0) { 1445 if (imo->imo_membership[n] != NULL) { 1446 in_delmulti(imo->imo_membership[n]); 1447 imo->imo_membership[n] = NULL; 1448 } 1449 } 1450 KASSERT(imo->imo_mfilters == NULL, 1451 ("%s: imo_mfilters != NULL", __func__)); 1452 imo->imo_num_memberships = 0; 1453 imo->imo_multicast_ifp = NULL; 1454 } 1455 1456 #ifdef INET6 1457 static void 1458 carp_multicast6_cleanup(struct carp_softc *sc) 1459 { 1460 struct ip6_moptions *im6o = &sc->sc_im6o; 1461 u_int16_t n = im6o->im6o_num_memberships; 1462 1463 while (n-- > 0) { 1464 if (im6o->im6o_membership[n] != NULL) { 1465 in6_mc_leave(im6o->im6o_membership[n], NULL); 1466 im6o->im6o_membership[n] = NULL; 1467 } 1468 } 1469 KASSERT(im6o->im6o_mfilters == NULL, 1470 ("%s: im6o_mfilters != NULL", __func__)); 1471 im6o->im6o_num_memberships = 0; 1472 im6o->im6o_multicast_ifp = NULL; 1473 } 1474 #endif 1475 1476 static int 1477 carp_set_addr(struct carp_softc *sc, struct sockaddr_in *sin) 1478 { 1479 INIT_VNET_INET(curvnet); 1480 struct ifnet *ifp; 1481 struct carp_if *cif; 1482 struct in_ifaddr *ia, *ia_if; 1483 struct ip_moptions *imo = &sc->sc_imo; 1484 struct in_addr addr; 1485 u_long iaddr = htonl(sin->sin_addr.s_addr); 1486 int own, error; 1487 1488 if (sin->sin_addr.s_addr == 0) { 1489 if (!(SC2IFP(sc)->if_flags & IFF_UP)) 1490 carp_set_state(sc, INIT); 1491 if (sc->sc_naddrs) 1492 SC2IFP(sc)->if_flags |= IFF_UP; 1493 if (sc->sc_carpdev) 1494 CARP_SCLOCK(sc); 1495 carp_setrun(sc, 0); 1496 if (sc->sc_carpdev) 1497 CARP_SCUNLOCK(sc); 1498 return (0); 1499 } 1500 1501 /* we have to do it by hands to check we won't match on us */ 1502 ia_if = NULL; own = 0; 1503 IN_IFADDR_RLOCK(); 1504 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 1505 /* and, yeah, we need a multicast-capable iface too */ 1506 if (ia->ia_ifp != SC2IFP(sc) && 1507 (ia->ia_ifp->if_flags & IFF_MULTICAST) && 1508 (iaddr & ia->ia_subnetmask) == ia->ia_subnet) { 1509 if (!ia_if) 1510 ia_if = ia; 1511 if (sin->sin_addr.s_addr == 1512 ia->ia_addr.sin_addr.s_addr) 1513 own++; 1514 } 1515 } 1516 1517 if (!ia_if) { 1518 IN_IFADDR_RUNLOCK(); 1519 return (EADDRNOTAVAIL); 1520 } 1521 1522 ia = ia_if; 1523 ifa_ref(&ia->ia_ifa); 1524 IN_IFADDR_RUNLOCK(); 1525 1526 ifp = ia->ia_ifp; 1527 1528 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0 || 1529 (imo->imo_multicast_ifp && imo->imo_multicast_ifp != ifp)) { 1530 ifa_free(&ia->ia_ifa); 1531 return (EADDRNOTAVAIL); 1532 } 1533 1534 if (imo->imo_num_memberships == 0) { 1535 addr.s_addr = htonl(INADDR_CARP_GROUP); 1536 if ((imo->imo_membership[0] = in_addmulti(&addr, ifp)) == 1537 NULL) { 1538 ifa_free(&ia->ia_ifa); 1539 return (ENOBUFS); 1540 } 1541 imo->imo_num_memberships++; 1542 imo->imo_multicast_ifp = ifp; 1543 imo->imo_multicast_ttl = CARP_DFLTTL; 1544 imo->imo_multicast_loop = 0; 1545 } 1546 1547 if (!ifp->if_carp) { 1548 1549 cif = malloc(sizeof(*cif), M_CARP, 1550 M_WAITOK|M_ZERO); 1551 if (!cif) { 1552 error = ENOBUFS; 1553 goto cleanup; 1554 } 1555 if ((error = ifpromisc(ifp, 1))) { 1556 free(cif, M_CARP); 1557 goto cleanup; 1558 } 1559 1560 CARP_LOCK_INIT(cif); 1561 CARP_LOCK(cif); 1562 cif->vhif_ifp = ifp; 1563 TAILQ_INIT(&cif->vhif_vrs); 1564 ifp->if_carp = cif; 1565 1566 } else { 1567 struct carp_softc *vr; 1568 1569 cif = (struct carp_if *)ifp->if_carp; 1570 CARP_LOCK(cif); 1571 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) 1572 if (vr != sc && vr->sc_vhid == sc->sc_vhid) { 1573 CARP_UNLOCK(cif); 1574 error = EEXIST; 1575 goto cleanup; 1576 } 1577 } 1578 sc->sc_ia = ia; 1579 sc->sc_carpdev = ifp; 1580 1581 { /* XXX prevent endless loop if already in queue */ 1582 struct carp_softc *vr, *after = NULL; 1583 int myself = 0; 1584 cif = (struct carp_if *)ifp->if_carp; 1585 1586 /* XXX: cif should not change, right? So we still hold the lock */ 1587 CARP_LOCK_ASSERT(cif); 1588 1589 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) { 1590 if (vr == sc) 1591 myself = 1; 1592 if (vr->sc_vhid < sc->sc_vhid) 1593 after = vr; 1594 } 1595 1596 if (!myself) { 1597 /* We're trying to keep things in order */ 1598 if (after == NULL) { 1599 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list); 1600 } else { 1601 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, sc, sc_list); 1602 } 1603 cif->vhif_nvrs++; 1604 } 1605 } 1606 1607 sc->sc_naddrs++; 1608 SC2IFP(sc)->if_flags |= IFF_UP; 1609 if (own) 1610 sc->sc_advskew = 0; 1611 carp_sc_state_locked(sc); 1612 carp_setrun(sc, 0); 1613 1614 CARP_UNLOCK(cif); 1615 ifa_free(&ia->ia_ifa); /* XXXRW: should hold reference for softc. */ 1616 1617 return (0); 1618 1619 cleanup: 1620 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]); 1621 ifa_free(&ia->ia_ifa); 1622 return (error); 1623 } 1624 1625 static int 1626 carp_del_addr(struct carp_softc *sc, struct sockaddr_in *sin) 1627 { 1628 int error = 0; 1629 1630 if (!--sc->sc_naddrs) { 1631 struct carp_if *cif = (struct carp_if *)sc->sc_carpdev->if_carp; 1632 struct ip_moptions *imo = &sc->sc_imo; 1633 1634 CARP_LOCK(cif); 1635 callout_stop(&sc->sc_ad_tmo); 1636 SC2IFP(sc)->if_flags &= ~IFF_UP; 1637 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING; 1638 sc->sc_vhid = -1; 1639 in_delmulti(imo->imo_membership[--imo->imo_num_memberships]); 1640 imo->imo_multicast_ifp = NULL; 1641 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list); 1642 if (!--cif->vhif_nvrs) { 1643 sc->sc_carpdev->if_carp = NULL; 1644 CARP_LOCK_DESTROY(cif); 1645 free(cif, M_IFADDR); 1646 } else { 1647 CARP_UNLOCK(cif); 1648 } 1649 } 1650 1651 return (error); 1652 } 1653 1654 #ifdef INET6 1655 static int 1656 carp_set_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6) 1657 { 1658 INIT_VNET_INET6(curvnet); 1659 struct ifnet *ifp; 1660 struct carp_if *cif; 1661 struct in6_ifaddr *ia, *ia_if; 1662 struct ip6_moptions *im6o = &sc->sc_im6o; 1663 struct in6_addr in6; 1664 int own, error; 1665 1666 error = 0; 1667 1668 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1669 if (!(SC2IFP(sc)->if_flags & IFF_UP)) 1670 carp_set_state(sc, INIT); 1671 if (sc->sc_naddrs6) 1672 SC2IFP(sc)->if_flags |= IFF_UP; 1673 if (sc->sc_carpdev) 1674 CARP_SCLOCK(sc); 1675 carp_setrun(sc, 0); 1676 if (sc->sc_carpdev) 1677 CARP_SCUNLOCK(sc); 1678 return (0); 1679 } 1680 1681 /* we have to do it by hands to check we won't match on us */ 1682 ia_if = NULL; own = 0; 1683 IN6_IFADDR_RLOCK(); 1684 TAILQ_FOREACH(ia, &V_in6_ifaddrhead, ia_link) { 1685 int i; 1686 1687 for (i = 0; i < 4; i++) { 1688 if ((sin6->sin6_addr.s6_addr32[i] & 1689 ia->ia_prefixmask.sin6_addr.s6_addr32[i]) != 1690 (ia->ia_addr.sin6_addr.s6_addr32[i] & 1691 ia->ia_prefixmask.sin6_addr.s6_addr32[i])) 1692 break; 1693 } 1694 /* and, yeah, we need a multicast-capable iface too */ 1695 if (ia->ia_ifp != SC2IFP(sc) && 1696 (ia->ia_ifp->if_flags & IFF_MULTICAST) && 1697 (i == 4)) { 1698 if (!ia_if) 1699 ia_if = ia; 1700 if (IN6_ARE_ADDR_EQUAL(&sin6->sin6_addr, 1701 &ia->ia_addr.sin6_addr)) 1702 own++; 1703 } 1704 } 1705 1706 if (!ia_if) { 1707 IN6_IFADDR_RUNLOCK(); 1708 return (EADDRNOTAVAIL); 1709 } 1710 ia = ia_if; 1711 ifa_ref(&ia->ia_ifa); 1712 IN6_IFADDR_RUNLOCK(); 1713 ifp = ia->ia_ifp; 1714 1715 if (ifp == NULL || (ifp->if_flags & IFF_MULTICAST) == 0 || 1716 (im6o->im6o_multicast_ifp && im6o->im6o_multicast_ifp != ifp)) { 1717 ifa_free(&ia->ia_ifa); 1718 return (EADDRNOTAVAIL); 1719 } 1720 1721 if (!sc->sc_naddrs6) { 1722 struct in6_multi *in6m; 1723 1724 im6o->im6o_multicast_ifp = ifp; 1725 1726 /* join CARP multicast address */ 1727 bzero(&in6, sizeof(in6)); 1728 in6.s6_addr16[0] = htons(0xff02); 1729 in6.s6_addr8[15] = 0x12; 1730 if (in6_setscope(&in6, ifp, NULL) != 0) 1731 goto cleanup; 1732 in6m = NULL; 1733 error = in6_mc_join(ifp, &in6, NULL, &in6m, 0); 1734 if (error) 1735 goto cleanup; 1736 im6o->im6o_membership[0] = in6m; 1737 im6o->im6o_num_memberships++; 1738 1739 /* join solicited multicast address */ 1740 bzero(&in6, sizeof(in6)); 1741 in6.s6_addr16[0] = htons(0xff02); 1742 in6.s6_addr32[1] = 0; 1743 in6.s6_addr32[2] = htonl(1); 1744 in6.s6_addr32[3] = sin6->sin6_addr.s6_addr32[3]; 1745 in6.s6_addr8[12] = 0xff; 1746 if (in6_setscope(&in6, ifp, NULL) != 0) 1747 goto cleanup; 1748 in6m = NULL; 1749 error = in6_mc_join(ifp, &in6, NULL, &in6m, 0); 1750 if (error) 1751 goto cleanup; 1752 im6o->im6o_membership[1] = in6m; 1753 im6o->im6o_num_memberships++; 1754 } 1755 1756 if (!ifp->if_carp) { 1757 cif = malloc(sizeof(*cif), M_CARP, 1758 M_WAITOK|M_ZERO); 1759 if (!cif) { 1760 error = ENOBUFS; 1761 goto cleanup; 1762 } 1763 if ((error = ifpromisc(ifp, 1))) { 1764 free(cif, M_CARP); 1765 goto cleanup; 1766 } 1767 1768 CARP_LOCK_INIT(cif); 1769 CARP_LOCK(cif); 1770 cif->vhif_ifp = ifp; 1771 TAILQ_INIT(&cif->vhif_vrs); 1772 ifp->if_carp = cif; 1773 1774 } else { 1775 struct carp_softc *vr; 1776 1777 cif = (struct carp_if *)ifp->if_carp; 1778 CARP_LOCK(cif); 1779 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) 1780 if (vr != sc && vr->sc_vhid == sc->sc_vhid) { 1781 CARP_UNLOCK(cif); 1782 error = EINVAL; 1783 goto cleanup; 1784 } 1785 } 1786 sc->sc_ia6 = ia; 1787 sc->sc_carpdev = ifp; 1788 1789 { /* XXX prevent endless loop if already in queue */ 1790 struct carp_softc *vr, *after = NULL; 1791 int myself = 0; 1792 cif = (struct carp_if *)ifp->if_carp; 1793 CARP_LOCK_ASSERT(cif); 1794 1795 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) { 1796 if (vr == sc) 1797 myself = 1; 1798 if (vr->sc_vhid < sc->sc_vhid) 1799 after = vr; 1800 } 1801 1802 if (!myself) { 1803 /* We're trying to keep things in order */ 1804 if (after == NULL) { 1805 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list); 1806 } else { 1807 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, sc, sc_list); 1808 } 1809 cif->vhif_nvrs++; 1810 } 1811 } 1812 1813 sc->sc_naddrs6++; 1814 SC2IFP(sc)->if_flags |= IFF_UP; 1815 if (own) 1816 sc->sc_advskew = 0; 1817 carp_sc_state_locked(sc); 1818 carp_setrun(sc, 0); 1819 1820 CARP_UNLOCK(cif); 1821 ifa_free(&ia->ia_ifa); /* XXXRW: should hold reference for softc. */ 1822 1823 return (0); 1824 1825 cleanup: 1826 if (!sc->sc_naddrs6) 1827 carp_multicast6_cleanup(sc); 1828 ifa_free(&ia->ia_ifa); 1829 return (error); 1830 } 1831 1832 static int 1833 carp_del_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6) 1834 { 1835 int error = 0; 1836 1837 if (!--sc->sc_naddrs6) { 1838 struct carp_if *cif = (struct carp_if *)sc->sc_carpdev->if_carp; 1839 1840 CARP_LOCK(cif); 1841 callout_stop(&sc->sc_ad_tmo); 1842 SC2IFP(sc)->if_flags &= ~IFF_UP; 1843 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING; 1844 sc->sc_vhid = -1; 1845 carp_multicast6_cleanup(sc); 1846 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list); 1847 if (!--cif->vhif_nvrs) { 1848 CARP_LOCK_DESTROY(cif); 1849 sc->sc_carpdev->if_carp = NULL; 1850 free(cif, M_IFADDR); 1851 } else 1852 CARP_UNLOCK(cif); 1853 } 1854 1855 return (error); 1856 } 1857 #endif /* INET6 */ 1858 1859 static int 1860 carp_ioctl(struct ifnet *ifp, u_long cmd, caddr_t addr) 1861 { 1862 struct carp_softc *sc = ifp->if_softc, *vr; 1863 struct carpreq carpr; 1864 struct ifaddr *ifa; 1865 struct ifreq *ifr; 1866 struct ifaliasreq *ifra; 1867 int locked = 0, error = 0; 1868 1869 ifa = (struct ifaddr *)addr; 1870 ifra = (struct ifaliasreq *)addr; 1871 ifr = (struct ifreq *)addr; 1872 1873 switch (cmd) { 1874 case SIOCSIFADDR: 1875 switch (ifa->ifa_addr->sa_family) { 1876 #ifdef INET 1877 case AF_INET: 1878 SC2IFP(sc)->if_flags |= IFF_UP; 1879 bcopy(ifa->ifa_addr, ifa->ifa_dstaddr, 1880 sizeof(struct sockaddr)); 1881 error = carp_set_addr(sc, satosin(ifa->ifa_addr)); 1882 break; 1883 #endif /* INET */ 1884 #ifdef INET6 1885 case AF_INET6: 1886 SC2IFP(sc)->if_flags |= IFF_UP; 1887 error = carp_set_addr6(sc, satosin6(ifa->ifa_addr)); 1888 break; 1889 #endif /* INET6 */ 1890 default: 1891 error = EAFNOSUPPORT; 1892 break; 1893 } 1894 break; 1895 1896 case SIOCAIFADDR: 1897 switch (ifa->ifa_addr->sa_family) { 1898 #ifdef INET 1899 case AF_INET: 1900 SC2IFP(sc)->if_flags |= IFF_UP; 1901 bcopy(ifa->ifa_addr, ifa->ifa_dstaddr, 1902 sizeof(struct sockaddr)); 1903 error = carp_set_addr(sc, satosin(&ifra->ifra_addr)); 1904 break; 1905 #endif /* INET */ 1906 #ifdef INET6 1907 case AF_INET6: 1908 SC2IFP(sc)->if_flags |= IFF_UP; 1909 error = carp_set_addr6(sc, satosin6(&ifra->ifra_addr)); 1910 break; 1911 #endif /* INET6 */ 1912 default: 1913 error = EAFNOSUPPORT; 1914 break; 1915 } 1916 break; 1917 1918 case SIOCDIFADDR: 1919 switch (ifa->ifa_addr->sa_family) { 1920 #ifdef INET 1921 case AF_INET: 1922 error = carp_del_addr(sc, satosin(&ifra->ifra_addr)); 1923 break; 1924 #endif /* INET */ 1925 #ifdef INET6 1926 case AF_INET6: 1927 error = carp_del_addr6(sc, satosin6(&ifra->ifra_addr)); 1928 break; 1929 #endif /* INET6 */ 1930 default: 1931 error = EAFNOSUPPORT; 1932 break; 1933 } 1934 break; 1935 1936 case SIOCSIFFLAGS: 1937 if (sc->sc_carpdev) { 1938 locked = 1; 1939 CARP_SCLOCK(sc); 1940 } 1941 if (sc->sc_state != INIT && !(ifr->ifr_flags & IFF_UP)) { 1942 callout_stop(&sc->sc_ad_tmo); 1943 callout_stop(&sc->sc_md_tmo); 1944 callout_stop(&sc->sc_md6_tmo); 1945 if (sc->sc_state == MASTER) 1946 carp_send_ad_locked(sc); 1947 carp_set_state(sc, INIT); 1948 carp_setrun(sc, 0); 1949 } else if (sc->sc_state == INIT && (ifr->ifr_flags & IFF_UP)) { 1950 SC2IFP(sc)->if_flags |= IFF_UP; 1951 carp_setrun(sc, 0); 1952 } 1953 break; 1954 1955 case SIOCSVH: 1956 error = priv_check(curthread, PRIV_NETINET_CARP); 1957 if (error) 1958 break; 1959 if ((error = copyin(ifr->ifr_data, &carpr, sizeof carpr))) 1960 break; 1961 error = 1; 1962 if (sc->sc_carpdev) { 1963 locked = 1; 1964 CARP_SCLOCK(sc); 1965 } 1966 if (sc->sc_state != INIT && carpr.carpr_state != sc->sc_state) { 1967 switch (carpr.carpr_state) { 1968 case BACKUP: 1969 callout_stop(&sc->sc_ad_tmo); 1970 carp_set_state(sc, BACKUP); 1971 carp_setrun(sc, 0); 1972 carp_setroute(sc, RTM_DELETE); 1973 break; 1974 case MASTER: 1975 carp_master_down_locked(sc); 1976 break; 1977 default: 1978 break; 1979 } 1980 } 1981 if (carpr.carpr_vhid > 0) { 1982 if (carpr.carpr_vhid > 255) { 1983 error = EINVAL; 1984 break; 1985 } 1986 if (sc->sc_carpdev) { 1987 struct carp_if *cif; 1988 cif = (struct carp_if *)sc->sc_carpdev->if_carp; 1989 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) 1990 if (vr != sc && 1991 vr->sc_vhid == carpr.carpr_vhid) { 1992 error = EEXIST; 1993 break; 1994 } 1995 if (error == EEXIST) 1996 break; 1997 } 1998 sc->sc_vhid = carpr.carpr_vhid; 1999 IF_LLADDR(sc->sc_ifp)[0] = 0; 2000 IF_LLADDR(sc->sc_ifp)[1] = 0; 2001 IF_LLADDR(sc->sc_ifp)[2] = 0x5e; 2002 IF_LLADDR(sc->sc_ifp)[3] = 0; 2003 IF_LLADDR(sc->sc_ifp)[4] = 1; 2004 IF_LLADDR(sc->sc_ifp)[5] = sc->sc_vhid; 2005 error--; 2006 } 2007 if (carpr.carpr_advbase > 0 || carpr.carpr_advskew > 0) { 2008 if (carpr.carpr_advskew >= 255) { 2009 error = EINVAL; 2010 break; 2011 } 2012 if (carpr.carpr_advbase > 255) { 2013 error = EINVAL; 2014 break; 2015 } 2016 sc->sc_advbase = carpr.carpr_advbase; 2017 sc->sc_advskew = carpr.carpr_advskew; 2018 error--; 2019 } 2020 bcopy(carpr.carpr_key, sc->sc_key, sizeof(sc->sc_key)); 2021 if (error > 0) 2022 error = EINVAL; 2023 else { 2024 error = 0; 2025 carp_setrun(sc, 0); 2026 } 2027 break; 2028 2029 case SIOCGVH: 2030 /* XXX: lockless read */ 2031 bzero(&carpr, sizeof(carpr)); 2032 carpr.carpr_state = sc->sc_state; 2033 carpr.carpr_vhid = sc->sc_vhid; 2034 carpr.carpr_advbase = sc->sc_advbase; 2035 carpr.carpr_advskew = sc->sc_advskew; 2036 error = priv_check(curthread, PRIV_NETINET_CARP); 2037 if (error == 0) 2038 bcopy(sc->sc_key, carpr.carpr_key, 2039 sizeof(carpr.carpr_key)); 2040 error = copyout(&carpr, ifr->ifr_data, sizeof(carpr)); 2041 break; 2042 2043 default: 2044 error = EINVAL; 2045 } 2046 2047 if (locked) 2048 CARP_SCUNLOCK(sc); 2049 2050 carp_hmac_prepare(sc); 2051 2052 return (error); 2053 } 2054 2055 /* 2056 * XXX: this is looutput. We should eventually use it from there. 2057 */ 2058 static int 2059 carp_looutput(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst, 2060 struct route *ro) 2061 { 2062 u_int32_t af; 2063 struct rtentry *rt = NULL; 2064 2065 M_ASSERTPKTHDR(m); /* check if we have the packet header */ 2066 2067 if (ro != NULL) 2068 rt = ro->ro_rt; 2069 if (rt && rt->rt_flags & (RTF_REJECT|RTF_BLACKHOLE)) { 2070 m_freem(m); 2071 return (rt->rt_flags & RTF_BLACKHOLE ? 0 : 2072 rt->rt_flags & RTF_HOST ? EHOSTUNREACH : ENETUNREACH); 2073 } 2074 2075 ifp->if_opackets++; 2076 ifp->if_obytes += m->m_pkthdr.len; 2077 2078 /* BPF writes need to be handled specially. */ 2079 if (dst->sa_family == AF_UNSPEC) { 2080 bcopy(dst->sa_data, &af, sizeof(af)); 2081 dst->sa_family = af; 2082 } 2083 2084 #if 1 /* XXX */ 2085 switch (dst->sa_family) { 2086 case AF_INET: 2087 case AF_INET6: 2088 case AF_IPX: 2089 case AF_APPLETALK: 2090 break; 2091 default: 2092 printf("carp_looutput: af=%d unexpected\n", dst->sa_family); 2093 m_freem(m); 2094 return (EAFNOSUPPORT); 2095 } 2096 #endif 2097 return(if_simloop(ifp, m, dst->sa_family, 0)); 2098 } 2099 2100 /* 2101 * Start output on carp interface. This function should never be called. 2102 */ 2103 static void 2104 carp_start(struct ifnet *ifp) 2105 { 2106 #ifdef DEBUG 2107 printf("%s: start called\n", ifp->if_xname); 2108 #endif 2109 } 2110 2111 int 2112 carp_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa, 2113 struct rtentry *rt) 2114 { 2115 struct m_tag *mtag; 2116 struct carp_softc *sc; 2117 struct ifnet *carp_ifp; 2118 2119 if (!sa) 2120 return (0); 2121 2122 switch (sa->sa_family) { 2123 #ifdef INET 2124 case AF_INET: 2125 break; 2126 #endif /* INET */ 2127 #ifdef INET6 2128 case AF_INET6: 2129 break; 2130 #endif /* INET6 */ 2131 default: 2132 return (0); 2133 } 2134 2135 mtag = m_tag_find(m, PACKET_TAG_CARP, NULL); 2136 if (mtag == NULL) 2137 return (0); 2138 2139 bcopy(mtag + 1, &carp_ifp, sizeof(struct ifnet *)); 2140 sc = carp_ifp->if_softc; 2141 2142 /* Set the source MAC address to Virtual Router MAC Address */ 2143 switch (ifp->if_type) { 2144 case IFT_ETHER: 2145 case IFT_L2VLAN: { 2146 struct ether_header *eh; 2147 2148 eh = mtod(m, struct ether_header *); 2149 eh->ether_shost[0] = 0; 2150 eh->ether_shost[1] = 0; 2151 eh->ether_shost[2] = 0x5e; 2152 eh->ether_shost[3] = 0; 2153 eh->ether_shost[4] = 1; 2154 eh->ether_shost[5] = sc->sc_vhid; 2155 } 2156 break; 2157 case IFT_FDDI: { 2158 struct fddi_header *fh; 2159 2160 fh = mtod(m, struct fddi_header *); 2161 fh->fddi_shost[0] = 0; 2162 fh->fddi_shost[1] = 0; 2163 fh->fddi_shost[2] = 0x5e; 2164 fh->fddi_shost[3] = 0; 2165 fh->fddi_shost[4] = 1; 2166 fh->fddi_shost[5] = sc->sc_vhid; 2167 } 2168 break; 2169 case IFT_ISO88025: { 2170 struct iso88025_header *th; 2171 th = mtod(m, struct iso88025_header *); 2172 th->iso88025_shost[0] = 3; 2173 th->iso88025_shost[1] = 0; 2174 th->iso88025_shost[2] = 0x40 >> (sc->sc_vhid - 1); 2175 th->iso88025_shost[3] = 0x40000 >> (sc->sc_vhid - 1); 2176 th->iso88025_shost[4] = 0; 2177 th->iso88025_shost[5] = 0; 2178 } 2179 break; 2180 default: 2181 printf("%s: carp is not supported for this interface type\n", 2182 ifp->if_xname); 2183 return (EOPNOTSUPP); 2184 } 2185 2186 return (0); 2187 } 2188 2189 static void 2190 carp_set_state(struct carp_softc *sc, int state) 2191 { 2192 int link_state; 2193 2194 if (sc->sc_carpdev) 2195 CARP_SCLOCK_ASSERT(sc); 2196 2197 if (sc->sc_state == state) 2198 return; 2199 2200 sc->sc_state = state; 2201 switch (state) { 2202 case BACKUP: 2203 link_state = LINK_STATE_DOWN; 2204 break; 2205 case MASTER: 2206 link_state = LINK_STATE_UP; 2207 break; 2208 default: 2209 link_state = LINK_STATE_UNKNOWN; 2210 break; 2211 } 2212 if_link_state_change(SC2IFP(sc), link_state); 2213 } 2214 2215 void 2216 carp_carpdev_state(void *v) 2217 { 2218 struct carp_if *cif = v; 2219 2220 CARP_LOCK(cif); 2221 carp_carpdev_state_locked(cif); 2222 CARP_UNLOCK(cif); 2223 } 2224 2225 static void 2226 carp_carpdev_state_locked(struct carp_if *cif) 2227 { 2228 struct carp_softc *sc; 2229 2230 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list) 2231 carp_sc_state_locked(sc); 2232 } 2233 2234 static void 2235 carp_sc_state_locked(struct carp_softc *sc) 2236 { 2237 CARP_SCLOCK_ASSERT(sc); 2238 2239 if (sc->sc_carpdev->if_link_state != LINK_STATE_UP || 2240 !(sc->sc_carpdev->if_flags & IFF_UP)) { 2241 sc->sc_flags_backup = SC2IFP(sc)->if_flags; 2242 SC2IFP(sc)->if_flags &= ~IFF_UP; 2243 SC2IFP(sc)->if_drv_flags &= ~IFF_DRV_RUNNING; 2244 callout_stop(&sc->sc_ad_tmo); 2245 callout_stop(&sc->sc_md_tmo); 2246 callout_stop(&sc->sc_md6_tmo); 2247 carp_set_state(sc, INIT); 2248 carp_setrun(sc, 0); 2249 if (!sc->sc_suppress) { 2250 carp_suppress_preempt++; 2251 if (carp_suppress_preempt == 1) { 2252 CARP_SCUNLOCK(sc); 2253 carp_send_ad_all(); 2254 CARP_SCLOCK(sc); 2255 } 2256 } 2257 sc->sc_suppress = 1; 2258 } else { 2259 SC2IFP(sc)->if_flags |= sc->sc_flags_backup; 2260 carp_set_state(sc, INIT); 2261 carp_setrun(sc, 0); 2262 if (sc->sc_suppress) 2263 carp_suppress_preempt--; 2264 sc->sc_suppress = 0; 2265 } 2266 2267 return; 2268 } 2269 2270 static int 2271 carp_modevent(module_t mod, int type, void *data) 2272 { 2273 switch (type) { 2274 case MOD_LOAD: 2275 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event, 2276 carp_ifdetach, NULL, EVENTHANDLER_PRI_ANY); 2277 if (if_detach_event_tag == NULL) 2278 return (ENOMEM); 2279 mtx_init(&carp_mtx, "carp_mtx", NULL, MTX_DEF); 2280 LIST_INIT(&carpif_list); 2281 if_clone_attach(&carp_cloner); 2282 break; 2283 2284 case MOD_UNLOAD: 2285 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag); 2286 if_clone_detach(&carp_cloner); 2287 mtx_destroy(&carp_mtx); 2288 break; 2289 2290 default: 2291 return (EINVAL); 2292 } 2293 2294 return (0); 2295 } 2296 2297 static moduledata_t carp_mod = { 2298 "carp", 2299 carp_modevent, 2300 0 2301 }; 2302 2303 DECLARE_MODULE(carp, carp_mod, SI_SUB_PSEUDO, SI_ORDER_ANY); 2304