1 /* $NetBSD: ip_carp.c,v 1.48 2012/03/27 17:48:17 bouyer Exp $ */ 2 /* $OpenBSD: ip_carp.c,v 1.113 2005/11/04 08:11:54 mcbride Exp $ */ 3 4 /* 5 * Copyright (c) 2002 Michael Shalayeff. All rights reserved. 6 * Copyright (c) 2003 Ryan McBride. All rights reserved. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR 18 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES 19 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. 20 * IN NO EVENT SHALL THE AUTHOR OR HIS RELATIVES BE LIABLE FOR ANY DIRECT, 21 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES 22 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR 23 * SERVICES; LOSS OF MIND, USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, 25 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING 26 * IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF 27 * THE POSSIBILITY OF SUCH DAMAGE. 28 */ 29 30 #include "opt_inet.h" 31 32 #include <sys/cdefs.h> 33 __KERNEL_RCSID(0, "$NetBSD: ip_carp.c,v 1.48 2012/03/27 17:48:17 bouyer Exp $"); 34 35 /* 36 * TODO: 37 * - iface reconfigure 38 * - support for hardware checksum calculations; 39 * 40 */ 41 42 #include <sys/param.h> 43 #include <sys/proc.h> 44 #include <sys/mbuf.h> 45 #include <sys/socket.h> 46 #include <sys/socketvar.h> 47 #include <sys/callout.h> 48 #include <sys/ioctl.h> 49 #include <sys/errno.h> 50 #include <sys/device.h> 51 #include <sys/time.h> 52 #include <sys/kernel.h> 53 #include <sys/kauth.h> 54 #include <sys/sysctl.h> 55 #include <sys/ucred.h> 56 #include <sys/syslog.h> 57 #include <sys/acct.h> 58 #include <sys/cprng.h> 59 60 #include <sys/cpu.h> 61 62 #include <net/if.h> 63 #include <net/pfil.h> 64 #include <net/if_types.h> 65 #include <net/if_ether.h> 66 #include <net/route.h> 67 #include <net/netisr.h> 68 #include <net/net_stats.h> 69 #include <netinet/if_inarp.h> 70 71 #if NFDDI > 0 72 #include <net/if_fddi.h> 73 #endif 74 #if NTOKEN > 0 75 #include <net/if_token.h> 76 #endif 77 78 #ifdef INET 79 #include <netinet/in.h> 80 #include <netinet/in_systm.h> 81 #include <netinet/in_var.h> 82 #include <netinet/ip.h> 83 #include <netinet/ip_var.h> 84 85 #include <net/if_dl.h> 86 #endif 87 88 #ifdef INET6 89 #include <netinet/icmp6.h> 90 #include <netinet/ip6.h> 91 #include <netinet6/ip6_var.h> 92 #include <netinet6/nd6.h> 93 #include <netinet6/scope6_var.h> 94 #endif 95 96 #include <net/bpf.h> 97 98 #include <sys/sha1.h> 99 100 #include <netinet/ip_carp.h> 101 102 struct carp_mc_entry { 103 LIST_ENTRY(carp_mc_entry) mc_entries; 104 union { 105 struct ether_multi *mcu_enm; 106 } mc_u; 107 struct sockaddr_storage mc_addr; 108 }; 109 #define mc_enm mc_u.mcu_enm 110 111 struct carp_softc { 112 struct ethercom sc_ac; 113 #define sc_if sc_ac.ec_if 114 #define sc_carpdev sc_ac.ec_if.if_carpdev 115 int ah_cookie; 116 int lh_cookie; 117 struct ip_moptions sc_imo; 118 #ifdef INET6 119 struct ip6_moptions sc_im6o; 120 #endif /* INET6 */ 121 TAILQ_ENTRY(carp_softc) sc_list; 122 123 enum { INIT = 0, BACKUP, MASTER } sc_state; 124 125 int sc_suppress; 126 int sc_bow_out; 127 128 int sc_sendad_errors; 129 #define CARP_SENDAD_MAX_ERRORS 3 130 int sc_sendad_success; 131 #define CARP_SENDAD_MIN_SUCCESS 3 132 133 int sc_vhid; 134 int sc_advskew; 135 int sc_naddrs; 136 int sc_naddrs6; 137 int sc_advbase; /* seconds */ 138 int sc_init_counter; 139 u_int64_t sc_counter; 140 141 /* authentication */ 142 #define CARP_HMAC_PAD 64 143 unsigned char sc_key[CARP_KEY_LEN]; 144 unsigned char sc_pad[CARP_HMAC_PAD]; 145 SHA1_CTX sc_sha1; 146 u_int32_t sc_hashkey[2]; 147 148 struct callout sc_ad_tmo; /* advertisement timeout */ 149 struct callout sc_md_tmo; /* master down timeout */ 150 struct callout sc_md6_tmo; /* master down timeout */ 151 152 LIST_HEAD(__carp_mchead, carp_mc_entry) carp_mc_listhead; 153 }; 154 155 int carp_suppress_preempt = 0; 156 int carp_opts[CARPCTL_MAXID] = { 0, 1, 0, 0, 0 }; /* XXX for now */ 157 158 static percpu_t *carpstat_percpu; 159 160 #define CARP_STATINC(x) _NET_STATINC(carpstat_percpu, x) 161 162 struct carp_if { 163 TAILQ_HEAD(, carp_softc) vhif_vrs; 164 int vhif_nvrs; 165 166 struct ifnet *vhif_ifp; 167 }; 168 169 #define CARP_LOG(sc, s) \ 170 if (carp_opts[CARPCTL_LOG]) { \ 171 if (sc) \ 172 log(LOG_INFO, "%s: ", \ 173 (sc)->sc_if.if_xname); \ 174 else \ 175 log(LOG_INFO, "carp: "); \ 176 addlog s; \ 177 addlog("\n"); \ 178 } 179 180 void carp_hmac_prepare(struct carp_softc *); 181 void carp_hmac_generate(struct carp_softc *, u_int32_t *, 182 unsigned char *); 183 int carp_hmac_verify(struct carp_softc *, u_int32_t *, 184 unsigned char *); 185 void carp_setroute(struct carp_softc *, int); 186 void carp_proto_input_c(struct mbuf *, struct carp_header *, sa_family_t); 187 void carpattach(int); 188 void carpdetach(struct carp_softc *); 189 int carp_prepare_ad(struct mbuf *, struct carp_softc *, 190 struct carp_header *); 191 void carp_send_ad_all(void); 192 void carp_send_ad(void *); 193 void carp_send_arp(struct carp_softc *); 194 void carp_master_down(void *); 195 int carp_ioctl(struct ifnet *, u_long, void *); 196 void carp_start(struct ifnet *); 197 void carp_setrun(struct carp_softc *, sa_family_t); 198 void carp_set_state(struct carp_softc *, int); 199 int carp_addrcount(struct carp_if *, struct in_ifaddr *, int); 200 enum { CARP_COUNT_MASTER, CARP_COUNT_RUNNING }; 201 202 void carp_multicast_cleanup(struct carp_softc *); 203 int carp_set_ifp(struct carp_softc *, struct ifnet *); 204 void carp_set_enaddr(struct carp_softc *); 205 void carp_addr_updated(void *); 206 u_int32_t carp_hash(struct carp_softc *, u_char *); 207 int carp_set_addr(struct carp_softc *, struct sockaddr_in *); 208 int carp_join_multicast(struct carp_softc *); 209 #ifdef INET6 210 void carp_send_na(struct carp_softc *); 211 int carp_set_addr6(struct carp_softc *, struct sockaddr_in6 *); 212 int carp_join_multicast6(struct carp_softc *); 213 #endif 214 int carp_clone_create(struct if_clone *, int); 215 int carp_clone_destroy(struct ifnet *); 216 int carp_ether_addmulti(struct carp_softc *, struct ifreq *); 217 int carp_ether_delmulti(struct carp_softc *, struct ifreq *); 218 void carp_ether_purgemulti(struct carp_softc *); 219 220 static void sysctl_net_inet_carp_setup(struct sysctllog **); 221 222 struct if_clone carp_cloner = 223 IF_CLONE_INITIALIZER("carp", carp_clone_create, carp_clone_destroy); 224 225 static __inline u_int16_t 226 carp_cksum(struct mbuf *m, int len) 227 { 228 return (in_cksum(m, len)); 229 } 230 231 void 232 carp_hmac_prepare(struct carp_softc *sc) 233 { 234 u_int8_t carp_version = CARP_VERSION, type = CARP_ADVERTISEMENT; 235 u_int8_t vhid = sc->sc_vhid & 0xff; 236 SHA1_CTX sha1ctx; 237 u_int32_t kmd[5]; 238 struct ifaddr *ifa; 239 int i, found; 240 struct in_addr last, cur, in; 241 #ifdef INET6 242 struct in6_addr last6, cur6, in6; 243 #endif /* INET6 */ 244 245 /* compute ipad from key */ 246 memset(sc->sc_pad, 0, sizeof(sc->sc_pad)); 247 memcpy(sc->sc_pad, sc->sc_key, sizeof(sc->sc_key)); 248 for (i = 0; i < sizeof(sc->sc_pad); i++) 249 sc->sc_pad[i] ^= 0x36; 250 251 /* precompute first part of inner hash */ 252 SHA1Init(&sc->sc_sha1); 253 SHA1Update(&sc->sc_sha1, sc->sc_pad, sizeof(sc->sc_pad)); 254 SHA1Update(&sc->sc_sha1, (void *)&carp_version, sizeof(carp_version)); 255 SHA1Update(&sc->sc_sha1, (void *)&type, sizeof(type)); 256 257 /* generate a key for the arpbalance hash, before the vhid is hashed */ 258 memcpy(&sha1ctx, &sc->sc_sha1, sizeof(sha1ctx)); 259 SHA1Final((unsigned char *)kmd, &sha1ctx); 260 sc->sc_hashkey[0] = kmd[0] ^ kmd[1]; 261 sc->sc_hashkey[1] = kmd[2] ^ kmd[3]; 262 263 /* the rest of the precomputation */ 264 SHA1Update(&sc->sc_sha1, (void *)&vhid, sizeof(vhid)); 265 266 /* Hash the addresses from smallest to largest, not interface order */ 267 #ifdef INET 268 cur.s_addr = 0; 269 do { 270 found = 0; 271 last = cur; 272 cur.s_addr = 0xffffffff; 273 IFADDR_FOREACH(ifa, &sc->sc_if) { 274 in.s_addr = ifatoia(ifa)->ia_addr.sin_addr.s_addr; 275 if (ifa->ifa_addr->sa_family == AF_INET && 276 ntohl(in.s_addr) > ntohl(last.s_addr) && 277 ntohl(in.s_addr) < ntohl(cur.s_addr)) { 278 cur.s_addr = in.s_addr; 279 found++; 280 } 281 } 282 if (found) 283 SHA1Update(&sc->sc_sha1, (void *)&cur, sizeof(cur)); 284 } while (found); 285 #endif /* INET */ 286 287 #ifdef INET6 288 memset(&cur6, 0x00, sizeof(cur6)); 289 do { 290 found = 0; 291 last6 = cur6; 292 memset(&cur6, 0xff, sizeof(cur6)); 293 IFADDR_FOREACH(ifa, &sc->sc_if) { 294 in6 = ifatoia6(ifa)->ia_addr.sin6_addr; 295 if (IN6_IS_ADDR_LINKLOCAL(&in6)) 296 in6.s6_addr16[1] = 0; 297 if (ifa->ifa_addr->sa_family == AF_INET6 && 298 memcmp(&in6, &last6, sizeof(in6)) > 0 && 299 memcmp(&in6, &cur6, sizeof(in6)) < 0) { 300 cur6 = in6; 301 found++; 302 } 303 } 304 if (found) 305 SHA1Update(&sc->sc_sha1, (void *)&cur6, sizeof(cur6)); 306 } while (found); 307 #endif /* INET6 */ 308 309 /* convert ipad to opad */ 310 for (i = 0; i < sizeof(sc->sc_pad); i++) 311 sc->sc_pad[i] ^= 0x36 ^ 0x5c; 312 } 313 314 void 315 carp_hmac_generate(struct carp_softc *sc, u_int32_t counter[2], 316 unsigned char md[20]) 317 { 318 SHA1_CTX sha1ctx; 319 320 /* fetch first half of inner hash */ 321 memcpy(&sha1ctx, &sc->sc_sha1, sizeof(sha1ctx)); 322 323 SHA1Update(&sha1ctx, (void *)counter, sizeof(sc->sc_counter)); 324 SHA1Final(md, &sha1ctx); 325 326 /* outer hash */ 327 SHA1Init(&sha1ctx); 328 SHA1Update(&sha1ctx, sc->sc_pad, sizeof(sc->sc_pad)); 329 SHA1Update(&sha1ctx, md, 20); 330 SHA1Final(md, &sha1ctx); 331 } 332 333 int 334 carp_hmac_verify(struct carp_softc *sc, u_int32_t counter[2], 335 unsigned char md[20]) 336 { 337 unsigned char md2[20]; 338 339 carp_hmac_generate(sc, counter, md2); 340 341 return (memcmp(md, md2, sizeof(md2))); 342 } 343 344 void 345 carp_setroute(struct carp_softc *sc, int cmd) 346 { 347 struct ifaddr *ifa; 348 int s; 349 350 s = splsoftnet(); 351 IFADDR_FOREACH(ifa, &sc->sc_if) { 352 switch (ifa->ifa_addr->sa_family) { 353 case AF_INET: { 354 int count = 0; 355 struct rtentry *rt; 356 int hr_otherif, nr_ourif; 357 358 /* 359 * Avoid screwing with the routes if there are other 360 * carp interfaces which are master and have the same 361 * address. 362 */ 363 if (sc->sc_carpdev != NULL && 364 sc->sc_carpdev->if_carp != NULL) { 365 count = carp_addrcount( 366 (struct carp_if *)sc->sc_carpdev->if_carp, 367 ifatoia(ifa), CARP_COUNT_MASTER); 368 if ((cmd == RTM_ADD && count != 1) || 369 (cmd == RTM_DELETE && count != 0)) 370 continue; 371 } 372 373 /* Remove the existing host route, if any */ 374 rtrequest(RTM_DELETE, ifa->ifa_addr, 375 ifa->ifa_addr, ifa->ifa_netmask, 376 RTF_HOST, NULL); 377 378 rt = NULL; 379 (void)rtrequest(RTM_GET, ifa->ifa_addr, ifa->ifa_addr, 380 ifa->ifa_netmask, RTF_HOST, &rt); 381 hr_otherif = (rt && rt->rt_ifp != &sc->sc_if && 382 rt->rt_flags & (RTF_CLONING|RTF_CLONED)); 383 if (rt != NULL) { 384 RTFREE(rt); 385 rt = NULL; 386 } 387 388 /* Check for a network route on our interface */ 389 390 rt = NULL; 391 (void)rtrequest(RTM_GET, ifa->ifa_addr, ifa->ifa_addr, 392 ifa->ifa_netmask, 0, &rt); 393 nr_ourif = (rt && rt->rt_ifp == &sc->sc_if); 394 395 switch (cmd) { 396 case RTM_ADD: 397 if (hr_otherif) { 398 ifa->ifa_rtrequest = NULL; 399 ifa->ifa_flags &= ~RTF_CLONING; 400 401 rtrequest(RTM_ADD, ifa->ifa_addr, 402 ifa->ifa_addr, ifa->ifa_netmask, 403 RTF_UP | RTF_HOST, NULL); 404 } 405 if (!hr_otherif || nr_ourif || !rt) { 406 if (nr_ourif && !(rt->rt_flags & 407 RTF_CLONING)) 408 rtrequest(RTM_DELETE, 409 ifa->ifa_addr, 410 ifa->ifa_addr, 411 ifa->ifa_netmask, 0, NULL); 412 413 ifa->ifa_rtrequest = arp_rtrequest; 414 ifa->ifa_flags |= RTF_CLONING; 415 416 if (rtrequest(RTM_ADD, ifa->ifa_addr, 417 ifa->ifa_addr, ifa->ifa_netmask, 0, 418 NULL) == 0) 419 ifa->ifa_flags |= IFA_ROUTE; 420 } 421 break; 422 case RTM_DELETE: 423 break; 424 default: 425 break; 426 } 427 if (rt != NULL) { 428 RTFREE(rt); 429 rt = NULL; 430 } 431 break; 432 } 433 434 #ifdef INET6 435 case AF_INET6: 436 if (cmd == RTM_ADD) 437 in6_ifaddloop(ifa); 438 else 439 in6_ifremloop(ifa); 440 break; 441 #endif /* INET6 */ 442 default: 443 break; 444 } 445 } 446 splx(s); 447 } 448 449 /* 450 * process input packet. 451 * we have rearranged checks order compared to the rfc, 452 * but it seems more efficient this way or not possible otherwise. 453 */ 454 void 455 carp_proto_input(struct mbuf *m, ...) 456 { 457 struct ip *ip = mtod(m, struct ip *); 458 struct carp_softc *sc = NULL; 459 struct carp_header *ch; 460 int iplen, len, hlen; 461 va_list ap; 462 463 va_start(ap, m); 464 hlen = va_arg(ap, int); 465 va_end(ap); 466 467 CARP_STATINC(CARP_STAT_IPACKETS); 468 469 if (!carp_opts[CARPCTL_ALLOW]) { 470 m_freem(m); 471 return; 472 } 473 474 /* check if received on a valid carp interface */ 475 if (m->m_pkthdr.rcvif->if_type != IFT_CARP) { 476 CARP_STATINC(CARP_STAT_BADIF); 477 CARP_LOG(sc, ("packet received on non-carp interface: %s", 478 m->m_pkthdr.rcvif->if_xname)); 479 m_freem(m); 480 return; 481 } 482 483 /* verify that the IP TTL is 255. */ 484 if (ip->ip_ttl != CARP_DFLTTL) { 485 CARP_STATINC(CARP_STAT_BADTTL); 486 CARP_LOG(sc, ("received ttl %d != %d on %s", ip->ip_ttl, 487 CARP_DFLTTL, m->m_pkthdr.rcvif->if_xname)); 488 m_freem(m); 489 return; 490 } 491 492 /* 493 * verify that the received packet length is 494 * equal to the CARP header 495 */ 496 iplen = ip->ip_hl << 2; 497 len = iplen + sizeof(*ch); 498 if (len > m->m_pkthdr.len) { 499 CARP_STATINC(CARP_STAT_BADLEN); 500 CARP_LOG(sc, ("packet too short %d on %s", m->m_pkthdr.len, 501 m->m_pkthdr.rcvif->if_xname)); 502 m_freem(m); 503 return; 504 } 505 506 if ((m = m_pullup(m, len)) == NULL) { 507 CARP_STATINC(CARP_STAT_HDROPS); 508 return; 509 } 510 ip = mtod(m, struct ip *); 511 ch = (struct carp_header *)((char *)ip + iplen); 512 /* verify the CARP checksum */ 513 m->m_data += iplen; 514 if (carp_cksum(m, len - iplen)) { 515 CARP_STATINC(CARP_STAT_BADSUM); 516 CARP_LOG(sc, ("checksum failed on %s", 517 m->m_pkthdr.rcvif->if_xname)); 518 m_freem(m); 519 return; 520 } 521 m->m_data -= iplen; 522 523 carp_proto_input_c(m, ch, AF_INET); 524 } 525 526 #ifdef INET6 527 int 528 carp6_proto_input(struct mbuf **mp, int *offp, int proto) 529 { 530 struct mbuf *m = *mp; 531 struct carp_softc *sc = NULL; 532 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 533 struct carp_header *ch; 534 u_int len; 535 536 CARP_STATINC(CARP_STAT_IPACKETS6); 537 538 if (!carp_opts[CARPCTL_ALLOW]) { 539 m_freem(m); 540 return (IPPROTO_DONE); 541 } 542 543 /* check if received on a valid carp interface */ 544 if (m->m_pkthdr.rcvif->if_type != IFT_CARP) { 545 CARP_STATINC(CARP_STAT_BADIF); 546 CARP_LOG(sc, ("packet received on non-carp interface: %s", 547 m->m_pkthdr.rcvif->if_xname)); 548 m_freem(m); 549 return (IPPROTO_DONE); 550 } 551 552 /* verify that the IP TTL is 255 */ 553 if (ip6->ip6_hlim != CARP_DFLTTL) { 554 CARP_STATINC(CARP_STAT_BADTTL); 555 CARP_LOG(sc, ("received ttl %d != %d on %s", ip6->ip6_hlim, 556 CARP_DFLTTL, m->m_pkthdr.rcvif->if_xname)); 557 m_freem(m); 558 return (IPPROTO_DONE); 559 } 560 561 /* verify that we have a complete carp packet */ 562 len = m->m_len; 563 IP6_EXTHDR_GET(ch, struct carp_header *, m, *offp, sizeof(*ch)); 564 if (ch == NULL) { 565 CARP_STATINC(CARP_STAT_BADLEN); 566 CARP_LOG(sc, ("packet size %u too small", len)); 567 return (IPPROTO_DONE); 568 } 569 570 571 /* verify the CARP checksum */ 572 m->m_data += *offp; 573 if (carp_cksum(m, sizeof(*ch))) { 574 CARP_STATINC(CARP_STAT_BADSUM); 575 CARP_LOG(sc, ("checksum failed, on %s", 576 m->m_pkthdr.rcvif->if_xname)); 577 m_freem(m); 578 return (IPPROTO_DONE); 579 } 580 m->m_data -= *offp; 581 582 carp_proto_input_c(m, ch, AF_INET6); 583 return (IPPROTO_DONE); 584 } 585 #endif /* INET6 */ 586 587 void 588 carp_proto_input_c(struct mbuf *m, struct carp_header *ch, sa_family_t af) 589 { 590 struct carp_softc *sc; 591 u_int64_t tmp_counter; 592 struct timeval sc_tv, ch_tv; 593 594 TAILQ_FOREACH(sc, &((struct carp_if *) 595 m->m_pkthdr.rcvif->if_carpdev->if_carp)->vhif_vrs, sc_list) 596 if (sc->sc_vhid == ch->carp_vhid) 597 break; 598 599 if (!sc || (sc->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) != 600 (IFF_UP|IFF_RUNNING)) { 601 CARP_STATINC(CARP_STAT_BADVHID); 602 m_freem(m); 603 return; 604 } 605 606 /* 607 * Check if our own advertisement was duplicated 608 * from a non simplex interface. 609 * XXX If there is no address on our physical interface 610 * there is no way to distinguish our ads from the ones 611 * another carp host might have sent us. 612 */ 613 if ((sc->sc_carpdev->if_flags & IFF_SIMPLEX) == 0) { 614 struct sockaddr sa; 615 struct ifaddr *ifa; 616 617 memset(&sa, 0, sizeof(sa)); 618 sa.sa_family = af; 619 ifa = ifaof_ifpforaddr(&sa, sc->sc_carpdev); 620 621 if (ifa && af == AF_INET) { 622 struct ip *ip = mtod(m, struct ip *); 623 if (ip->ip_src.s_addr == 624 ifatoia(ifa)->ia_addr.sin_addr.s_addr) { 625 m_freem(m); 626 return; 627 } 628 } 629 #ifdef INET6 630 if (ifa && af == AF_INET6) { 631 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 632 struct in6_addr in6_src, in6_found; 633 634 in6_src = ip6->ip6_src; 635 in6_found = ifatoia6(ifa)->ia_addr.sin6_addr; 636 if (IN6_IS_ADDR_LINKLOCAL(&in6_src)) 637 in6_src.s6_addr16[1] = 0; 638 if (IN6_IS_ADDR_LINKLOCAL(&in6_found)) 639 in6_found.s6_addr16[1] = 0; 640 if (IN6_ARE_ADDR_EQUAL(&in6_src, &in6_found)) { 641 m_freem(m); 642 return; 643 } 644 } 645 #endif /* INET6 */ 646 } 647 648 nanotime(&sc->sc_if.if_lastchange); 649 sc->sc_if.if_ipackets++; 650 sc->sc_if.if_ibytes += m->m_pkthdr.len; 651 652 /* verify the CARP version. */ 653 if (ch->carp_version != CARP_VERSION) { 654 CARP_STATINC(CARP_STAT_BADVER); 655 sc->sc_if.if_ierrors++; 656 CARP_LOG(sc, ("invalid version %d != %d", 657 ch->carp_version, CARP_VERSION)); 658 m_freem(m); 659 return; 660 } 661 662 /* verify the hash */ 663 if (carp_hmac_verify(sc, ch->carp_counter, ch->carp_md)) { 664 CARP_STATINC(CARP_STAT_BADAUTH); 665 sc->sc_if.if_ierrors++; 666 CARP_LOG(sc, ("incorrect hash")); 667 m_freem(m); 668 return; 669 } 670 671 tmp_counter = ntohl(ch->carp_counter[0]); 672 tmp_counter = tmp_counter<<32; 673 tmp_counter += ntohl(ch->carp_counter[1]); 674 675 /* XXX Replay protection goes here */ 676 677 sc->sc_init_counter = 0; 678 sc->sc_counter = tmp_counter; 679 680 681 sc_tv.tv_sec = sc->sc_advbase; 682 if (carp_suppress_preempt && sc->sc_advskew < 240) 683 sc_tv.tv_usec = 240 * 1000000 / 256; 684 else 685 sc_tv.tv_usec = sc->sc_advskew * 1000000 / 256; 686 ch_tv.tv_sec = ch->carp_advbase; 687 ch_tv.tv_usec = ch->carp_advskew * 1000000 / 256; 688 689 switch (sc->sc_state) { 690 case INIT: 691 break; 692 case MASTER: 693 /* 694 * If we receive an advertisement from a backup who's going to 695 * be more frequent than us, go into BACKUP state. 696 */ 697 if (timercmp(&sc_tv, &ch_tv, >) || 698 timercmp(&sc_tv, &ch_tv, ==)) { 699 callout_stop(&sc->sc_ad_tmo); 700 CARP_LOG(sc, ("MASTER -> BACKUP (more frequent advertisement received)")); 701 carp_set_state(sc, BACKUP); 702 carp_setrun(sc, 0); 703 carp_setroute(sc, RTM_DELETE); 704 } 705 break; 706 case BACKUP: 707 /* 708 * If we're pre-empting masters who advertise slower than us, 709 * and this one claims to be slower, treat him as down. 710 */ 711 if (carp_opts[CARPCTL_PREEMPT] && timercmp(&sc_tv, &ch_tv, <)) { 712 CARP_LOG(sc, ("BACKUP -> MASTER (preempting a slower master)")); 713 carp_master_down(sc); 714 break; 715 } 716 717 /* 718 * If the master is going to advertise at such a low frequency 719 * that he's guaranteed to time out, we'd might as well just 720 * treat him as timed out now. 721 */ 722 sc_tv.tv_sec = sc->sc_advbase * 3; 723 if (timercmp(&sc_tv, &ch_tv, <)) { 724 CARP_LOG(sc, ("BACKUP -> MASTER (master timed out)")); 725 carp_master_down(sc); 726 break; 727 } 728 729 /* 730 * Otherwise, we reset the counter and wait for the next 731 * advertisement. 732 */ 733 carp_setrun(sc, af); 734 break; 735 } 736 737 m_freem(m); 738 return; 739 } 740 741 /* 742 * Interface side of the CARP implementation. 743 */ 744 745 /* ARGSUSED */ 746 void 747 carpattach(int n) 748 { 749 if_clone_attach(&carp_cloner); 750 751 carpstat_percpu = percpu_alloc(sizeof(uint64_t) * CARP_NSTATS); 752 } 753 754 int 755 carp_clone_create(struct if_clone *ifc, int unit) 756 { 757 extern int ifqmaxlen; 758 struct carp_softc *sc; 759 struct ifnet *ifp; 760 761 sc = malloc(sizeof(*sc), M_DEVBUF, M_NOWAIT|M_ZERO); 762 if (!sc) 763 return (ENOMEM); 764 765 sc->sc_suppress = 0; 766 sc->sc_advbase = CARP_DFLTINTV; 767 sc->sc_vhid = -1; /* required setting */ 768 sc->sc_advskew = 0; 769 sc->sc_init_counter = 1; 770 sc->sc_naddrs = sc->sc_naddrs6 = 0; 771 #ifdef INET6 772 sc->sc_im6o.im6o_multicast_hlim = CARP_DFLTTL; 773 #endif /* INET6 */ 774 775 callout_init(&sc->sc_ad_tmo, 0); 776 callout_init(&sc->sc_md_tmo, 0); 777 callout_init(&sc->sc_md6_tmo, 0); 778 779 callout_setfunc(&sc->sc_ad_tmo, carp_send_ad, sc); 780 callout_setfunc(&sc->sc_md_tmo, carp_master_down, sc); 781 callout_setfunc(&sc->sc_md6_tmo, carp_master_down, sc); 782 783 LIST_INIT(&sc->carp_mc_listhead); 784 ifp = &sc->sc_if; 785 ifp->if_softc = sc; 786 snprintf(ifp->if_xname, sizeof ifp->if_xname, "%s%d", ifc->ifc_name, 787 unit); 788 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; 789 ifp->if_ioctl = carp_ioctl; 790 ifp->if_start = carp_start; 791 ifp->if_output = carp_output; 792 ifp->if_type = IFT_CARP; 793 ifp->if_addrlen = ETHER_ADDR_LEN; 794 ifp->if_hdrlen = ETHER_HDR_LEN; 795 ifp->if_mtu = ETHERMTU; 796 IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen); 797 IFQ_SET_READY(&ifp->if_snd); 798 if_attach(ifp); 799 800 if_alloc_sadl(ifp); 801 ifp->if_broadcastaddr = etherbroadcastaddr; 802 carp_set_enaddr(sc); 803 LIST_INIT(&sc->sc_ac.ec_multiaddrs); 804 bpf_attach(ifp, DLT_EN10MB, ETHER_HDR_LEN); 805 return (0); 806 } 807 808 int 809 carp_clone_destroy(struct ifnet *ifp) 810 { 811 struct carp_softc *sc = ifp->if_softc; 812 813 carpdetach(ifp->if_softc); 814 ether_ifdetach(ifp); 815 if_detach(ifp); 816 callout_destroy(&sc->sc_ad_tmo); 817 callout_destroy(&sc->sc_md_tmo); 818 callout_destroy(&sc->sc_md6_tmo); 819 free(ifp->if_softc, M_DEVBUF); 820 821 return (0); 822 } 823 824 void 825 carpdetach(struct carp_softc *sc) 826 { 827 struct carp_if *cif; 828 int s; 829 830 callout_stop(&sc->sc_ad_tmo); 831 callout_stop(&sc->sc_md_tmo); 832 callout_stop(&sc->sc_md6_tmo); 833 834 if (sc->sc_suppress) 835 carp_suppress_preempt--; 836 sc->sc_suppress = 0; 837 838 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) 839 carp_suppress_preempt--; 840 sc->sc_sendad_errors = 0; 841 842 carp_set_state(sc, INIT); 843 sc->sc_if.if_flags &= ~IFF_UP; 844 carp_setrun(sc, 0); 845 carp_multicast_cleanup(sc); 846 847 s = splnet(); 848 if (sc->sc_carpdev != NULL) { 849 /* XXX linkstatehook removal */ 850 cif = (struct carp_if *)sc->sc_carpdev->if_carp; 851 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list); 852 if (!--cif->vhif_nvrs) { 853 ifpromisc(sc->sc_carpdev, 0); 854 sc->sc_carpdev->if_carp = NULL; 855 free(cif, M_IFADDR); 856 } 857 } 858 sc->sc_carpdev = NULL; 859 splx(s); 860 } 861 862 /* Detach an interface from the carp. */ 863 void 864 carp_ifdetach(struct ifnet *ifp) 865 { 866 struct carp_softc *sc, *nextsc; 867 struct carp_if *cif = (struct carp_if *)ifp->if_carp; 868 869 for (sc = TAILQ_FIRST(&cif->vhif_vrs); sc; sc = nextsc) { 870 nextsc = TAILQ_NEXT(sc, sc_list); 871 carpdetach(sc); 872 } 873 } 874 875 int 876 carp_prepare_ad(struct mbuf *m, struct carp_softc *sc, 877 struct carp_header *ch) 878 { 879 if (sc->sc_init_counter) { 880 /* this could also be seconds since unix epoch */ 881 sc->sc_counter = cprng_fast64(); 882 } else 883 sc->sc_counter++; 884 885 ch->carp_counter[0] = htonl((sc->sc_counter>>32)&0xffffffff); 886 ch->carp_counter[1] = htonl(sc->sc_counter&0xffffffff); 887 888 carp_hmac_generate(sc, ch->carp_counter, ch->carp_md); 889 890 return (0); 891 } 892 893 void 894 carp_send_ad_all(void) 895 { 896 struct ifnet *ifp; 897 struct carp_if *cif; 898 struct carp_softc *vh; 899 900 TAILQ_FOREACH(ifp, &ifnet, if_list) { 901 if (ifp->if_carp == NULL || ifp->if_type == IFT_CARP) 902 continue; 903 904 cif = (struct carp_if *)ifp->if_carp; 905 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 906 if ((vh->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) == 907 (IFF_UP|IFF_RUNNING) && vh->sc_state == MASTER) 908 carp_send_ad(vh); 909 } 910 } 911 } 912 913 914 void 915 carp_send_ad(void *v) 916 { 917 struct carp_header ch; 918 struct timeval tv; 919 struct carp_softc *sc = v; 920 struct carp_header *ch_ptr; 921 struct mbuf *m; 922 int error, len, advbase, advskew, s; 923 struct ifaddr *ifa; 924 struct sockaddr sa; 925 926 s = splsoftnet(); 927 928 advbase = advskew = 0; /* Sssssh compiler */ 929 if (sc->sc_carpdev == NULL) { 930 sc->sc_if.if_oerrors++; 931 goto retry_later; 932 } 933 934 /* bow out if we've gone to backup (the carp interface is going down) */ 935 if (sc->sc_bow_out) { 936 sc->sc_bow_out = 0; 937 advbase = 255; 938 advskew = 255; 939 } else { 940 advbase = sc->sc_advbase; 941 if (!carp_suppress_preempt || sc->sc_advskew > 240) 942 advskew = sc->sc_advskew; 943 else 944 advskew = 240; 945 tv.tv_sec = advbase; 946 tv.tv_usec = advskew * 1000000 / 256; 947 } 948 949 ch.carp_version = CARP_VERSION; 950 ch.carp_type = CARP_ADVERTISEMENT; 951 ch.carp_vhid = sc->sc_vhid; 952 ch.carp_advbase = advbase; 953 ch.carp_advskew = advskew; 954 ch.carp_authlen = 7; /* XXX DEFINE */ 955 ch.carp_pad1 = 0; /* must be zero */ 956 ch.carp_cksum = 0; 957 958 959 #ifdef INET 960 if (sc->sc_naddrs) { 961 struct ip *ip; 962 963 MGETHDR(m, M_DONTWAIT, MT_HEADER); 964 if (m == NULL) { 965 sc->sc_if.if_oerrors++; 966 CARP_STATINC(CARP_STAT_ONOMEM); 967 /* XXX maybe less ? */ 968 goto retry_later; 969 } 970 len = sizeof(*ip) + sizeof(ch); 971 m->m_pkthdr.len = len; 972 m->m_pkthdr.rcvif = NULL; 973 m->m_len = len; 974 MH_ALIGN(m, m->m_len); 975 m->m_flags |= M_MCAST; 976 ip = mtod(m, struct ip *); 977 ip->ip_v = IPVERSION; 978 ip->ip_hl = sizeof(*ip) >> 2; 979 ip->ip_tos = IPTOS_LOWDELAY; 980 ip->ip_len = htons(len); 981 ip->ip_id = 0; /* no need for id, we don't support fragments */ 982 ip->ip_off = htons(IP_DF); 983 ip->ip_ttl = CARP_DFLTTL; 984 ip->ip_p = IPPROTO_CARP; 985 ip->ip_sum = 0; 986 987 memset(&sa, 0, sizeof(sa)); 988 sa.sa_family = AF_INET; 989 ifa = ifaof_ifpforaddr(&sa, sc->sc_carpdev); 990 if (ifa == NULL) 991 ip->ip_src.s_addr = 0; 992 else 993 ip->ip_src.s_addr = 994 ifatoia(ifa)->ia_addr.sin_addr.s_addr; 995 ip->ip_dst.s_addr = INADDR_CARP_GROUP; 996 997 ch_ptr = (struct carp_header *)(&ip[1]); 998 memcpy(ch_ptr, &ch, sizeof(ch)); 999 if (carp_prepare_ad(m, sc, ch_ptr)) 1000 goto retry_later; 1001 1002 m->m_data += sizeof(*ip); 1003 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip)); 1004 m->m_data -= sizeof(*ip); 1005 1006 nanotime(&sc->sc_if.if_lastchange); 1007 sc->sc_if.if_opackets++; 1008 sc->sc_if.if_obytes += len; 1009 CARP_STATINC(CARP_STAT_OPACKETS); 1010 1011 error = ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, 1012 NULL); 1013 if (error) { 1014 if (error == ENOBUFS) 1015 CARP_STATINC(CARP_STAT_ONOMEM); 1016 else 1017 CARP_LOG(sc, ("ip_output failed: %d", error)); 1018 sc->sc_if.if_oerrors++; 1019 if (sc->sc_sendad_errors < INT_MAX) 1020 sc->sc_sendad_errors++; 1021 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) { 1022 carp_suppress_preempt++; 1023 if (carp_suppress_preempt == 1) 1024 carp_send_ad_all(); 1025 } 1026 sc->sc_sendad_success = 0; 1027 } else { 1028 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) { 1029 if (++sc->sc_sendad_success >= 1030 CARP_SENDAD_MIN_SUCCESS) { 1031 carp_suppress_preempt--; 1032 sc->sc_sendad_errors = 0; 1033 } 1034 } else 1035 sc->sc_sendad_errors = 0; 1036 } 1037 } 1038 #endif /* INET */ 1039 #ifdef INET6 1040 if (sc->sc_naddrs6) { 1041 struct ip6_hdr *ip6; 1042 1043 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1044 if (m == NULL) { 1045 sc->sc_if.if_oerrors++; 1046 CARP_STATINC(CARP_STAT_ONOMEM); 1047 /* XXX maybe less ? */ 1048 goto retry_later; 1049 } 1050 len = sizeof(*ip6) + sizeof(ch); 1051 m->m_pkthdr.len = len; 1052 m->m_pkthdr.rcvif = NULL; 1053 m->m_len = len; 1054 MH_ALIGN(m, m->m_len); 1055 m->m_flags |= M_MCAST; 1056 ip6 = mtod(m, struct ip6_hdr *); 1057 memset(ip6, 0, sizeof(*ip6)); 1058 ip6->ip6_vfc |= IPV6_VERSION; 1059 ip6->ip6_hlim = CARP_DFLTTL; 1060 ip6->ip6_nxt = IPPROTO_CARP; 1061 1062 /* set the source address */ 1063 memset(&sa, 0, sizeof(sa)); 1064 sa.sa_family = AF_INET6; 1065 ifa = ifaof_ifpforaddr(&sa, sc->sc_carpdev); 1066 if (ifa == NULL) /* This should never happen with IPv6 */ 1067 memset(&ip6->ip6_src, 0, sizeof(struct in6_addr)); 1068 else 1069 bcopy(ifatoia6(ifa)->ia_addr.sin6_addr.s6_addr, 1070 &ip6->ip6_src, sizeof(struct in6_addr)); 1071 /* set the multicast destination */ 1072 1073 ip6->ip6_dst.s6_addr16[0] = htons(0xff02); 1074 ip6->ip6_dst.s6_addr8[15] = 0x12; 1075 if (in6_setscope(&ip6->ip6_dst, sc->sc_carpdev, NULL) != 0) { 1076 sc->sc_if.if_oerrors++; 1077 m_freem(m); 1078 CARP_LOG(sc, ("in6_setscope failed")); 1079 goto retry_later; 1080 } 1081 1082 ch_ptr = (struct carp_header *)(&ip6[1]); 1083 memcpy(ch_ptr, &ch, sizeof(ch)); 1084 if (carp_prepare_ad(m, sc, ch_ptr)) 1085 goto retry_later; 1086 1087 m->m_data += sizeof(*ip6); 1088 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip6)); 1089 m->m_data -= sizeof(*ip6); 1090 1091 nanotime(&sc->sc_if.if_lastchange); 1092 sc->sc_if.if_opackets++; 1093 sc->sc_if.if_obytes += len; 1094 CARP_STATINC(CARP_STAT_OPACKETS6); 1095 1096 error = ip6_output(m, NULL, NULL, 0, &sc->sc_im6o, NULL, NULL); 1097 if (error) { 1098 if (error == ENOBUFS) 1099 CARP_STATINC(CARP_STAT_ONOMEM); 1100 else 1101 CARP_LOG(sc, ("ip6_output failed: %d", error)); 1102 sc->sc_if.if_oerrors++; 1103 if (sc->sc_sendad_errors < INT_MAX) 1104 sc->sc_sendad_errors++; 1105 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) { 1106 carp_suppress_preempt++; 1107 if (carp_suppress_preempt == 1) 1108 carp_send_ad_all(); 1109 } 1110 sc->sc_sendad_success = 0; 1111 } else { 1112 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) { 1113 if (++sc->sc_sendad_success >= 1114 CARP_SENDAD_MIN_SUCCESS) { 1115 carp_suppress_preempt--; 1116 sc->sc_sendad_errors = 0; 1117 } 1118 } else 1119 sc->sc_sendad_errors = 0; 1120 } 1121 } 1122 #endif /* INET6 */ 1123 1124 retry_later: 1125 splx(s); 1126 if (advbase != 255 || advskew != 255) 1127 callout_schedule(&sc->sc_ad_tmo, tvtohz(&tv)); 1128 } 1129 1130 /* 1131 * Broadcast a gratuitous ARP request containing 1132 * the virtual router MAC address for each IP address 1133 * associated with the virtual router. 1134 */ 1135 void 1136 carp_send_arp(struct carp_softc *sc) 1137 { 1138 struct ifaddr *ifa; 1139 struct in_addr *in; 1140 int s = splsoftnet(); 1141 1142 IFADDR_FOREACH(ifa, &sc->sc_if) { 1143 1144 if (ifa->ifa_addr->sa_family != AF_INET) 1145 continue; 1146 1147 in = &ifatoia(ifa)->ia_addr.sin_addr; 1148 arprequest(sc->sc_carpdev, in, in, CLLADDR(sc->sc_if.if_sadl)); 1149 } 1150 splx(s); 1151 } 1152 1153 #ifdef INET6 1154 void 1155 carp_send_na(struct carp_softc *sc) 1156 { 1157 struct ifaddr *ifa; 1158 struct in6_addr *in6; 1159 static struct in6_addr mcast = IN6ADDR_LINKLOCAL_ALLNODES_INIT; 1160 int s = splsoftnet(); 1161 1162 IFADDR_FOREACH(ifa, &sc->sc_if) { 1163 1164 if (ifa->ifa_addr->sa_family != AF_INET6) 1165 continue; 1166 1167 in6 = &ifatoia6(ifa)->ia_addr.sin6_addr; 1168 nd6_na_output(sc->sc_carpdev, &mcast, in6, 1169 ND_NA_FLAG_OVERRIDE, 1, NULL); 1170 } 1171 splx(s); 1172 } 1173 #endif /* INET6 */ 1174 1175 /* 1176 * Based on bridge_hash() in if_bridge.c 1177 */ 1178 #define mix(a,b,c) \ 1179 do { \ 1180 a -= b; a -= c; a ^= (c >> 13); \ 1181 b -= c; b -= a; b ^= (a << 8); \ 1182 c -= a; c -= b; c ^= (b >> 13); \ 1183 a -= b; a -= c; a ^= (c >> 12); \ 1184 b -= c; b -= a; b ^= (a << 16); \ 1185 c -= a; c -= b; c ^= (b >> 5); \ 1186 a -= b; a -= c; a ^= (c >> 3); \ 1187 b -= c; b -= a; b ^= (a << 10); \ 1188 c -= a; c -= b; c ^= (b >> 15); \ 1189 } while (0) 1190 1191 u_int32_t 1192 carp_hash(struct carp_softc *sc, u_char *src) 1193 { 1194 u_int32_t a = 0x9e3779b9, b = sc->sc_hashkey[0], c = sc->sc_hashkey[1]; 1195 1196 c += sc->sc_key[3] << 24; 1197 c += sc->sc_key[2] << 16; 1198 c += sc->sc_key[1] << 8; 1199 c += sc->sc_key[0]; 1200 b += src[5] << 8; 1201 b += src[4]; 1202 a += src[3] << 24; 1203 a += src[2] << 16; 1204 a += src[1] << 8; 1205 a += src[0]; 1206 1207 mix(a, b, c); 1208 return (c); 1209 } 1210 1211 int 1212 carp_addrcount(struct carp_if *cif, struct in_ifaddr *ia, int type) 1213 { 1214 struct carp_softc *vh; 1215 struct ifaddr *ifa; 1216 int count = 0; 1217 1218 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1219 if ((type == CARP_COUNT_RUNNING && 1220 (vh->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) == 1221 (IFF_UP|IFF_RUNNING)) || 1222 (type == CARP_COUNT_MASTER && vh->sc_state == MASTER)) { 1223 IFADDR_FOREACH(ifa, &vh->sc_if) { 1224 if (ifa->ifa_addr->sa_family == AF_INET && 1225 ia->ia_addr.sin_addr.s_addr == 1226 ifatoia(ifa)->ia_addr.sin_addr.s_addr) 1227 count++; 1228 } 1229 } 1230 } 1231 return (count); 1232 } 1233 1234 int 1235 carp_iamatch(struct in_ifaddr *ia, u_char *src, 1236 u_int32_t *count, u_int32_t index) 1237 { 1238 struct carp_softc *sc = ia->ia_ifp->if_softc; 1239 1240 if (carp_opts[CARPCTL_ARPBALANCE]) { 1241 /* 1242 * We use the source ip to decide which virtual host should 1243 * handle the request. If we're master of that virtual host, 1244 * then we respond, otherwise, just drop the arp packet on 1245 * the floor. 1246 */ 1247 1248 /* Count the elegible carp interfaces with this address */ 1249 if (*count == 0) 1250 *count = carp_addrcount( 1251 (struct carp_if *)ia->ia_ifp->if_carpdev->if_carp, 1252 ia, CARP_COUNT_RUNNING); 1253 1254 /* This should never happen, but... */ 1255 if (*count == 0) 1256 return (0); 1257 1258 if (carp_hash(sc, src) % *count == index - 1 && 1259 sc->sc_state == MASTER) { 1260 return (1); 1261 } 1262 } else { 1263 if (sc->sc_state == MASTER) 1264 return (1); 1265 } 1266 1267 return (0); 1268 } 1269 1270 #ifdef INET6 1271 struct ifaddr * 1272 carp_iamatch6(void *v, struct in6_addr *taddr) 1273 { 1274 struct carp_if *cif = v; 1275 struct carp_softc *vh; 1276 struct ifaddr *ifa; 1277 1278 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1279 IFADDR_FOREACH(ifa, &vh->sc_if) { 1280 if (IN6_ARE_ADDR_EQUAL(taddr, 1281 &ifatoia6(ifa)->ia_addr.sin6_addr) && 1282 ((vh->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) == 1283 (IFF_UP|IFF_RUNNING)) && vh->sc_state == MASTER) 1284 return (ifa); 1285 } 1286 } 1287 1288 return (NULL); 1289 } 1290 #endif /* INET6 */ 1291 1292 struct ifnet * 1293 carp_ourether(void *v, struct ether_header *eh, u_char iftype, int src) 1294 { 1295 struct carp_if *cif = (struct carp_if *)v; 1296 struct carp_softc *vh; 1297 u_int8_t *ena; 1298 1299 if (src) 1300 ena = (u_int8_t *)&eh->ether_shost; 1301 else 1302 ena = (u_int8_t *)&eh->ether_dhost; 1303 1304 switch (iftype) { 1305 case IFT_ETHER: 1306 case IFT_FDDI: 1307 if (ena[0] || ena[1] || ena[2] != 0x5e || ena[3] || ena[4] != 1) 1308 return (NULL); 1309 break; 1310 case IFT_ISO88025: 1311 if (ena[0] != 3 || ena[1] || ena[4] || ena[5]) 1312 return (NULL); 1313 break; 1314 default: 1315 return (NULL); 1316 break; 1317 } 1318 1319 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) 1320 if ((vh->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) == 1321 (IFF_UP|IFF_RUNNING) && vh->sc_state == MASTER && 1322 !memcmp(ena, CLLADDR(vh->sc_if.if_sadl), 1323 ETHER_ADDR_LEN)) { 1324 return (&vh->sc_if); 1325 } 1326 1327 return (NULL); 1328 } 1329 1330 int 1331 carp_input(struct mbuf *m, u_int8_t *shost, u_int8_t *dhost, u_int16_t etype) 1332 { 1333 struct ether_header eh; 1334 struct carp_if *cif = (struct carp_if *)m->m_pkthdr.rcvif->if_carp; 1335 struct ifnet *ifp; 1336 1337 memcpy(&eh.ether_shost, shost, sizeof(eh.ether_shost)); 1338 memcpy(&eh.ether_dhost, dhost, sizeof(eh.ether_dhost)); 1339 eh.ether_type = etype; 1340 1341 if (m->m_flags & (M_BCAST|M_MCAST)) { 1342 struct carp_softc *vh; 1343 struct mbuf *m0; 1344 1345 /* 1346 * XXX Should really check the list of multicast addresses 1347 * for each CARP interface _before_ copying. 1348 */ 1349 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1350 m0 = m_copym(m, 0, M_COPYALL, M_DONTWAIT); 1351 if (m0 == NULL) 1352 continue; 1353 m0->m_pkthdr.rcvif = &vh->sc_if; 1354 ether_input(&vh->sc_if, m0); 1355 } 1356 return (1); 1357 } 1358 1359 ifp = carp_ourether(cif, &eh, m->m_pkthdr.rcvif->if_type, 0); 1360 if (ifp == NULL) { 1361 return (1); 1362 } 1363 1364 m->m_pkthdr.rcvif = ifp; 1365 1366 bpf_mtap(ifp, m); 1367 ifp->if_ipackets++; 1368 ether_input(ifp, m); 1369 return (0); 1370 } 1371 1372 void 1373 carp_master_down(void *v) 1374 { 1375 struct carp_softc *sc = v; 1376 1377 switch (sc->sc_state) { 1378 case INIT: 1379 printf("%s: master_down event in INIT state\n", 1380 sc->sc_if.if_xname); 1381 break; 1382 case MASTER: 1383 break; 1384 case BACKUP: 1385 CARP_LOG(sc, ("INIT -> MASTER (preempting)")); 1386 carp_set_state(sc, MASTER); 1387 carp_send_ad(sc); 1388 carp_send_arp(sc); 1389 #ifdef INET6 1390 carp_send_na(sc); 1391 #endif /* INET6 */ 1392 carp_setrun(sc, 0); 1393 carp_setroute(sc, RTM_ADD); 1394 break; 1395 } 1396 } 1397 1398 /* 1399 * When in backup state, af indicates whether to reset the master down timer 1400 * for v4 or v6. If it's set to zero, reset the ones which are already pending. 1401 */ 1402 void 1403 carp_setrun(struct carp_softc *sc, sa_family_t af) 1404 { 1405 struct timeval tv; 1406 1407 if (sc->sc_carpdev == NULL) { 1408 sc->sc_if.if_flags &= ~IFF_RUNNING; 1409 carp_set_state(sc, INIT); 1410 return; 1411 } 1412 1413 if (sc->sc_if.if_flags & IFF_UP && sc->sc_vhid > 0 && 1414 (sc->sc_naddrs || sc->sc_naddrs6) && !sc->sc_suppress) { 1415 sc->sc_if.if_flags |= IFF_RUNNING; 1416 } else { 1417 sc->sc_if.if_flags &= ~IFF_RUNNING; 1418 carp_setroute(sc, RTM_DELETE); 1419 return; 1420 } 1421 1422 switch (sc->sc_state) { 1423 case INIT: 1424 carp_set_state(sc, BACKUP); 1425 carp_setroute(sc, RTM_DELETE); 1426 carp_setrun(sc, 0); 1427 break; 1428 case BACKUP: 1429 callout_stop(&sc->sc_ad_tmo); 1430 tv.tv_sec = 3 * sc->sc_advbase; 1431 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1432 switch (af) { 1433 #ifdef INET 1434 case AF_INET: 1435 callout_schedule(&sc->sc_md_tmo, tvtohz(&tv)); 1436 break; 1437 #endif /* INET */ 1438 #ifdef INET6 1439 case AF_INET6: 1440 callout_schedule(&sc->sc_md6_tmo, tvtohz(&tv)); 1441 break; 1442 #endif /* INET6 */ 1443 default: 1444 if (sc->sc_naddrs) 1445 callout_schedule(&sc->sc_md_tmo, tvtohz(&tv)); 1446 if (sc->sc_naddrs6) 1447 callout_schedule(&sc->sc_md6_tmo, tvtohz(&tv)); 1448 break; 1449 } 1450 break; 1451 case MASTER: 1452 tv.tv_sec = sc->sc_advbase; 1453 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1454 callout_schedule(&sc->sc_ad_tmo, tvtohz(&tv)); 1455 break; 1456 } 1457 } 1458 1459 void 1460 carp_multicast_cleanup(struct carp_softc *sc) 1461 { 1462 struct ip_moptions *imo = &sc->sc_imo; 1463 #ifdef INET6 1464 struct ip6_moptions *im6o = &sc->sc_im6o; 1465 #endif 1466 u_int16_t n = imo->imo_num_memberships; 1467 1468 /* Clean up our own multicast memberships */ 1469 while (n-- > 0) { 1470 if (imo->imo_membership[n] != NULL) { 1471 in_delmulti(imo->imo_membership[n]); 1472 imo->imo_membership[n] = NULL; 1473 } 1474 } 1475 imo->imo_num_memberships = 0; 1476 imo->imo_multicast_ifp = NULL; 1477 1478 #ifdef INET6 1479 while (!LIST_EMPTY(&im6o->im6o_memberships)) { 1480 struct in6_multi_mship *imm = 1481 LIST_FIRST(&im6o->im6o_memberships); 1482 1483 LIST_REMOVE(imm, i6mm_chain); 1484 in6_leavegroup(imm); 1485 } 1486 im6o->im6o_multicast_ifp = NULL; 1487 #endif 1488 1489 /* And any other multicast memberships */ 1490 carp_ether_purgemulti(sc); 1491 } 1492 1493 int 1494 carp_set_ifp(struct carp_softc *sc, struct ifnet *ifp) 1495 { 1496 struct carp_if *cif, *ncif = NULL; 1497 struct carp_softc *vr, *after = NULL; 1498 int myself = 0, error = 0; 1499 int s; 1500 1501 if (ifp == sc->sc_carpdev) 1502 return (0); 1503 1504 if (ifp != NULL) { 1505 if ((ifp->if_flags & IFF_MULTICAST) == 0) 1506 return (EADDRNOTAVAIL); 1507 1508 if (ifp->if_type == IFT_CARP) 1509 return (EINVAL); 1510 1511 if (ifp->if_carp == NULL) { 1512 ncif = malloc(sizeof(*cif), M_IFADDR, M_NOWAIT); 1513 if (ncif == NULL) 1514 return (ENOBUFS); 1515 if ((error = ifpromisc(ifp, 1))) { 1516 free(ncif, M_IFADDR); 1517 return (error); 1518 } 1519 1520 ncif->vhif_ifp = ifp; 1521 TAILQ_INIT(&ncif->vhif_vrs); 1522 } else { 1523 cif = (struct carp_if *)ifp->if_carp; 1524 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) 1525 if (vr != sc && vr->sc_vhid == sc->sc_vhid) 1526 return (EINVAL); 1527 } 1528 1529 /* detach from old interface */ 1530 if (sc->sc_carpdev != NULL) 1531 carpdetach(sc); 1532 1533 /* join multicast groups */ 1534 if (sc->sc_naddrs < 0 && 1535 (error = carp_join_multicast(sc)) != 0) { 1536 if (ncif != NULL) 1537 free(ncif, M_IFADDR); 1538 return (error); 1539 } 1540 1541 #ifdef INET6 1542 if (sc->sc_naddrs6 < 0 && 1543 (error = carp_join_multicast6(sc)) != 0) { 1544 if (ncif != NULL) 1545 free(ncif, M_IFADDR); 1546 carp_multicast_cleanup(sc); 1547 return (error); 1548 } 1549 #endif 1550 1551 /* attach carp interface to physical interface */ 1552 if (ncif != NULL) 1553 ifp->if_carp = (void *)ncif; 1554 sc->sc_carpdev = ifp; 1555 cif = (struct carp_if *)ifp->if_carp; 1556 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) { 1557 if (vr == sc) 1558 myself = 1; 1559 if (vr->sc_vhid < sc->sc_vhid) 1560 after = vr; 1561 } 1562 1563 if (!myself) { 1564 /* We're trying to keep things in order */ 1565 if (after == NULL) { 1566 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list); 1567 } else { 1568 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, 1569 sc, sc_list); 1570 } 1571 cif->vhif_nvrs++; 1572 } 1573 if (sc->sc_naddrs || sc->sc_naddrs6) 1574 sc->sc_if.if_flags |= IFF_UP; 1575 carp_set_enaddr(sc); 1576 s = splnet(); 1577 /* XXX linkstatehooks establish */ 1578 carp_carpdev_state(ifp); 1579 splx(s); 1580 } else { 1581 carpdetach(sc); 1582 sc->sc_if.if_flags &= ~(IFF_UP|IFF_RUNNING); 1583 } 1584 return (0); 1585 } 1586 1587 void 1588 carp_set_enaddr(struct carp_softc *sc) 1589 { 1590 uint8_t enaddr[ETHER_ADDR_LEN]; 1591 if (sc->sc_carpdev && sc->sc_carpdev->if_type == IFT_ISO88025) { 1592 enaddr[0] = 3; 1593 enaddr[1] = 0; 1594 enaddr[2] = 0x40 >> (sc->sc_vhid - 1); 1595 enaddr[3] = 0x40000 >> (sc->sc_vhid - 1); 1596 enaddr[4] = 0; 1597 enaddr[5] = 0; 1598 } else { 1599 enaddr[0] = 0; 1600 enaddr[1] = 0; 1601 enaddr[2] = 0x5e; 1602 enaddr[3] = 0; 1603 enaddr[4] = 1; 1604 enaddr[5] = sc->sc_vhid; 1605 } 1606 if_set_sadl(&sc->sc_if, enaddr, sizeof(enaddr), false); 1607 } 1608 1609 void 1610 carp_addr_updated(void *v) 1611 { 1612 struct carp_softc *sc = (struct carp_softc *) v; 1613 struct ifaddr *ifa; 1614 int new_naddrs = 0, new_naddrs6 = 0; 1615 1616 IFADDR_FOREACH(ifa, &sc->sc_if) { 1617 if (ifa->ifa_addr->sa_family == AF_INET) 1618 new_naddrs++; 1619 else if (ifa->ifa_addr->sa_family == AF_INET6) 1620 new_naddrs6++; 1621 } 1622 1623 /* Handle a callback after SIOCDIFADDR */ 1624 if (new_naddrs < sc->sc_naddrs || new_naddrs6 < sc->sc_naddrs6) { 1625 struct in_addr mc_addr; 1626 struct in_multi *inm; 1627 1628 sc->sc_naddrs = new_naddrs; 1629 sc->sc_naddrs6 = new_naddrs6; 1630 1631 /* Re-establish multicast membership removed by in_control */ 1632 mc_addr.s_addr = INADDR_CARP_GROUP; 1633 IN_LOOKUP_MULTI(mc_addr, &sc->sc_if, inm); 1634 if (inm == NULL) { 1635 memset(&sc->sc_imo, 0, sizeof(sc->sc_imo)); 1636 1637 if (sc->sc_carpdev != NULL && sc->sc_naddrs > 0) 1638 carp_join_multicast(sc); 1639 } 1640 1641 if (sc->sc_naddrs == 0 && sc->sc_naddrs6 == 0) { 1642 sc->sc_if.if_flags &= ~IFF_UP; 1643 carp_set_state(sc, INIT); 1644 } else 1645 carp_hmac_prepare(sc); 1646 } 1647 1648 carp_setrun(sc, 0); 1649 } 1650 1651 int 1652 carp_set_addr(struct carp_softc *sc, struct sockaddr_in *sin) 1653 { 1654 struct ifnet *ifp = sc->sc_carpdev; 1655 struct in_ifaddr *ia, *ia_if; 1656 int error = 0; 1657 1658 if (sin->sin_addr.s_addr == 0) { 1659 if (!(sc->sc_if.if_flags & IFF_UP)) 1660 carp_set_state(sc, INIT); 1661 if (sc->sc_naddrs) 1662 sc->sc_if.if_flags |= IFF_UP; 1663 carp_setrun(sc, 0); 1664 return (0); 1665 } 1666 1667 /* we have to do this by hand to ensure we don't match on ourselves */ 1668 ia_if = NULL; 1669 for (ia = TAILQ_FIRST(&in_ifaddrhead); ia; 1670 ia = TAILQ_NEXT(ia, ia_list)) { 1671 1672 /* and, yeah, we need a multicast-capable iface too */ 1673 if (ia->ia_ifp != &sc->sc_if && 1674 ia->ia_ifp->if_type != IFT_CARP && 1675 (ia->ia_ifp->if_flags & IFF_MULTICAST) && 1676 (sin->sin_addr.s_addr & ia->ia_subnetmask) == 1677 ia->ia_subnet) { 1678 if (!ia_if) 1679 ia_if = ia; 1680 } 1681 } 1682 1683 if (ia_if) { 1684 ia = ia_if; 1685 if (ifp) { 1686 if (ifp != ia->ia_ifp) 1687 return (EADDRNOTAVAIL); 1688 } else { 1689 ifp = ia->ia_ifp; 1690 } 1691 } 1692 1693 if ((error = carp_set_ifp(sc, ifp))) 1694 return (error); 1695 1696 if (sc->sc_carpdev == NULL) 1697 return (EADDRNOTAVAIL); 1698 1699 if (sc->sc_naddrs == 0 && (error = carp_join_multicast(sc)) != 0) 1700 return (error); 1701 1702 sc->sc_naddrs++; 1703 if (sc->sc_carpdev != NULL) 1704 sc->sc_if.if_flags |= IFF_UP; 1705 1706 carp_set_state(sc, INIT); 1707 carp_setrun(sc, 0); 1708 1709 /* 1710 * Hook if_addrhooks so that we get a callback after in_ifinit has run, 1711 * to correct any inappropriate routes that it inserted. 1712 */ 1713 if (sc->ah_cookie == 0) { 1714 /* XXX link address hook */ 1715 } 1716 1717 return (0); 1718 } 1719 1720 int 1721 carp_join_multicast(struct carp_softc *sc) 1722 { 1723 struct ip_moptions *imo = &sc->sc_imo, tmpimo; 1724 struct in_addr addr; 1725 1726 memset(&tmpimo, 0, sizeof(tmpimo)); 1727 addr.s_addr = INADDR_CARP_GROUP; 1728 if ((tmpimo.imo_membership[0] = 1729 in_addmulti(&addr, &sc->sc_if)) == NULL) { 1730 return (ENOBUFS); 1731 } 1732 1733 imo->imo_membership[0] = tmpimo.imo_membership[0]; 1734 imo->imo_num_memberships = 1; 1735 imo->imo_multicast_ifp = &sc->sc_if; 1736 imo->imo_multicast_ttl = CARP_DFLTTL; 1737 imo->imo_multicast_loop = 0; 1738 return (0); 1739 } 1740 1741 1742 #ifdef INET6 1743 int 1744 carp_set_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6) 1745 { 1746 struct ifnet *ifp = sc->sc_carpdev; 1747 struct in6_ifaddr *ia, *ia_if; 1748 int error = 0; 1749 1750 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1751 if (!(sc->sc_if.if_flags & IFF_UP)) 1752 carp_set_state(sc, INIT); 1753 if (sc->sc_naddrs6) 1754 sc->sc_if.if_flags |= IFF_UP; 1755 carp_setrun(sc, 0); 1756 return (0); 1757 } 1758 1759 /* we have to do this by hand to ensure we don't match on ourselves */ 1760 ia_if = NULL; 1761 for (ia = in6_ifaddr; ia; ia = ia->ia_next) { 1762 int i; 1763 1764 for (i = 0; i < 4; i++) { 1765 if ((sin6->sin6_addr.s6_addr32[i] & 1766 ia->ia_prefixmask.sin6_addr.s6_addr32[i]) != 1767 (ia->ia_addr.sin6_addr.s6_addr32[i] & 1768 ia->ia_prefixmask.sin6_addr.s6_addr32[i])) 1769 break; 1770 } 1771 /* and, yeah, we need a multicast-capable iface too */ 1772 if (ia->ia_ifp != &sc->sc_if && 1773 ia->ia_ifp->if_type != IFT_CARP && 1774 (ia->ia_ifp->if_flags & IFF_MULTICAST) && 1775 (i == 4)) { 1776 if (!ia_if) 1777 ia_if = ia; 1778 } 1779 } 1780 1781 if (ia_if) { 1782 ia = ia_if; 1783 if (sc->sc_carpdev) { 1784 if (sc->sc_carpdev != ia->ia_ifp) 1785 return (EADDRNOTAVAIL); 1786 } else { 1787 ifp = ia->ia_ifp; 1788 } 1789 } 1790 1791 if ((error = carp_set_ifp(sc, ifp))) 1792 return (error); 1793 1794 if (sc->sc_carpdev == NULL) 1795 return (EADDRNOTAVAIL); 1796 1797 if (sc->sc_naddrs6 == 0 && (error = carp_join_multicast6(sc)) != 0) 1798 return (error); 1799 1800 sc->sc_naddrs6++; 1801 if (sc->sc_carpdev != NULL) 1802 sc->sc_if.if_flags |= IFF_UP; 1803 carp_set_state(sc, INIT); 1804 carp_setrun(sc, 0); 1805 1806 return (0); 1807 } 1808 1809 int 1810 carp_join_multicast6(struct carp_softc *sc) 1811 { 1812 struct in6_multi_mship *imm, *imm2; 1813 struct ip6_moptions *im6o = &sc->sc_im6o; 1814 struct sockaddr_in6 addr6; 1815 int error; 1816 1817 /* Join IPv6 CARP multicast group */ 1818 memset(&addr6, 0, sizeof(addr6)); 1819 addr6.sin6_family = AF_INET6; 1820 addr6.sin6_len = sizeof(addr6); 1821 addr6.sin6_addr.s6_addr16[0] = htons(0xff02); 1822 addr6.sin6_addr.s6_addr16[1] = htons(sc->sc_if.if_index); 1823 addr6.sin6_addr.s6_addr8[15] = 0x12; 1824 if ((imm = in6_joingroup(&sc->sc_if, 1825 &addr6.sin6_addr, &error, 0)) == NULL) { 1826 return (error); 1827 } 1828 /* join solicited multicast address */ 1829 memset(&addr6.sin6_addr, 0, sizeof(addr6.sin6_addr)); 1830 addr6.sin6_addr.s6_addr16[0] = htons(0xff02); 1831 addr6.sin6_addr.s6_addr16[1] = htons(sc->sc_if.if_index); 1832 addr6.sin6_addr.s6_addr32[1] = 0; 1833 addr6.sin6_addr.s6_addr32[2] = htonl(1); 1834 addr6.sin6_addr.s6_addr32[3] = 0; 1835 addr6.sin6_addr.s6_addr8[12] = 0xff; 1836 if ((imm2 = in6_joingroup(&sc->sc_if, 1837 &addr6.sin6_addr, &error, 0)) == NULL) { 1838 in6_leavegroup(imm); 1839 return (error); 1840 } 1841 1842 /* apply v6 multicast membership */ 1843 im6o->im6o_multicast_ifp = &sc->sc_if; 1844 if (imm) 1845 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, 1846 i6mm_chain); 1847 if (imm2) 1848 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm2, 1849 i6mm_chain); 1850 1851 return (0); 1852 } 1853 1854 #endif /* INET6 */ 1855 1856 int 1857 carp_ioctl(struct ifnet *ifp, u_long cmd, void *data) 1858 { 1859 struct lwp *l = curlwp; /* XXX */ 1860 struct carp_softc *sc = ifp->if_softc, *vr; 1861 struct carpreq carpr; 1862 struct ifaddr *ifa; 1863 struct ifreq *ifr; 1864 struct ifnet *cdev = NULL; 1865 int error = 0; 1866 1867 ifa = (struct ifaddr *)data; 1868 ifr = (struct ifreq *)data; 1869 1870 switch (cmd) { 1871 case SIOCINITIFADDR: 1872 switch (ifa->ifa_addr->sa_family) { 1873 #ifdef INET 1874 case AF_INET: 1875 sc->sc_if.if_flags |= IFF_UP; 1876 memcpy(ifa->ifa_dstaddr, ifa->ifa_addr, 1877 sizeof(struct sockaddr)); 1878 error = carp_set_addr(sc, satosin(ifa->ifa_addr)); 1879 break; 1880 #endif /* INET */ 1881 #ifdef INET6 1882 case AF_INET6: 1883 sc->sc_if.if_flags|= IFF_UP; 1884 error = carp_set_addr6(sc, satosin6(ifa->ifa_addr)); 1885 break; 1886 #endif /* INET6 */ 1887 default: 1888 error = EAFNOSUPPORT; 1889 break; 1890 } 1891 break; 1892 1893 case SIOCSIFFLAGS: 1894 if ((error = ifioctl_common(ifp, cmd, data)) != 0) 1895 break; 1896 if (sc->sc_state != INIT && !(ifr->ifr_flags & IFF_UP)) { 1897 callout_stop(&sc->sc_ad_tmo); 1898 callout_stop(&sc->sc_md_tmo); 1899 callout_stop(&sc->sc_md6_tmo); 1900 if (sc->sc_state == MASTER) { 1901 /* we need the interface up to bow out */ 1902 sc->sc_if.if_flags |= IFF_UP; 1903 sc->sc_bow_out = 1; 1904 carp_send_ad(sc); 1905 } 1906 sc->sc_if.if_flags &= ~IFF_UP; 1907 carp_set_state(sc, INIT); 1908 carp_setrun(sc, 0); 1909 } else if (sc->sc_state == INIT && (ifr->ifr_flags & IFF_UP)) { 1910 sc->sc_if.if_flags |= IFF_UP; 1911 carp_setrun(sc, 0); 1912 } 1913 break; 1914 1915 case SIOCSVH: 1916 if (l == NULL) 1917 break; 1918 if ((error = kauth_authorize_network(l->l_cred, 1919 KAUTH_NETWORK_INTERFACE, 1920 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 1921 NULL)) != 0) 1922 break; 1923 if ((error = copyin(ifr->ifr_data, &carpr, sizeof carpr))) 1924 break; 1925 error = 1; 1926 if (carpr.carpr_carpdev[0] != '\0' && 1927 (cdev = ifunit(carpr.carpr_carpdev)) == NULL) 1928 return (EINVAL); 1929 if ((error = carp_set_ifp(sc, cdev))) 1930 return (error); 1931 if (sc->sc_state != INIT && carpr.carpr_state != sc->sc_state) { 1932 switch (carpr.carpr_state) { 1933 case BACKUP: 1934 callout_stop(&sc->sc_ad_tmo); 1935 carp_set_state(sc, BACKUP); 1936 carp_setrun(sc, 0); 1937 carp_setroute(sc, RTM_DELETE); 1938 break; 1939 case MASTER: 1940 carp_master_down(sc); 1941 break; 1942 default: 1943 break; 1944 } 1945 } 1946 if (carpr.carpr_vhid > 0) { 1947 if (carpr.carpr_vhid > 255) { 1948 error = EINVAL; 1949 break; 1950 } 1951 if (sc->sc_carpdev) { 1952 struct carp_if *cif; 1953 cif = (struct carp_if *)sc->sc_carpdev->if_carp; 1954 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) 1955 if (vr != sc && 1956 vr->sc_vhid == carpr.carpr_vhid) 1957 return (EINVAL); 1958 } 1959 sc->sc_vhid = carpr.carpr_vhid; 1960 carp_set_enaddr(sc); 1961 carp_set_state(sc, INIT); 1962 error--; 1963 } 1964 if (carpr.carpr_advbase > 0 || carpr.carpr_advskew > 0) { 1965 if (carpr.carpr_advskew > 254) { 1966 error = EINVAL; 1967 break; 1968 } 1969 if (carpr.carpr_advbase > 255) { 1970 error = EINVAL; 1971 break; 1972 } 1973 sc->sc_advbase = carpr.carpr_advbase; 1974 sc->sc_advskew = carpr.carpr_advskew; 1975 error--; 1976 } 1977 memcpy(sc->sc_key, carpr.carpr_key, sizeof(sc->sc_key)); 1978 if (error > 0) 1979 error = EINVAL; 1980 else { 1981 error = 0; 1982 carp_setrun(sc, 0); 1983 } 1984 break; 1985 1986 case SIOCGVH: 1987 memset(&carpr, 0, sizeof(carpr)); 1988 if (sc->sc_carpdev != NULL) 1989 strlcpy(carpr.carpr_carpdev, sc->sc_carpdev->if_xname, 1990 IFNAMSIZ); 1991 carpr.carpr_state = sc->sc_state; 1992 carpr.carpr_vhid = sc->sc_vhid; 1993 carpr.carpr_advbase = sc->sc_advbase; 1994 carpr.carpr_advskew = sc->sc_advskew; 1995 1996 if ((l != NULL) && (error = kauth_authorize_network(l->l_cred, 1997 KAUTH_NETWORK_INTERFACE, 1998 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 1999 NULL)) == 0) 2000 memcpy(carpr.carpr_key, sc->sc_key, 2001 sizeof(carpr.carpr_key)); 2002 error = copyout(&carpr, ifr->ifr_data, sizeof(carpr)); 2003 break; 2004 2005 case SIOCADDMULTI: 2006 error = carp_ether_addmulti(sc, ifr); 2007 break; 2008 2009 case SIOCDELMULTI: 2010 error = carp_ether_delmulti(sc, ifr); 2011 break; 2012 2013 default: 2014 error = ether_ioctl(ifp, cmd, data); 2015 } 2016 2017 carp_hmac_prepare(sc); 2018 return (error); 2019 } 2020 2021 2022 /* 2023 * Start output on carp interface. This function should never be called. 2024 */ 2025 void 2026 carp_start(struct ifnet *ifp) 2027 { 2028 #ifdef DEBUG 2029 printf("%s: start called\n", ifp->if_xname); 2030 #endif 2031 } 2032 2033 int 2034 carp_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa, 2035 struct rtentry *rt) 2036 { 2037 struct carp_softc *sc = ((struct carp_softc *)ifp->if_softc); 2038 2039 if (sc->sc_carpdev != NULL && sc->sc_state == MASTER) { 2040 return (sc->sc_carpdev->if_output(ifp, m, sa, rt)); 2041 } else { 2042 m_freem(m); 2043 return (ENETUNREACH); 2044 } 2045 } 2046 2047 void 2048 carp_set_state(struct carp_softc *sc, int state) 2049 { 2050 static const char *carp_states[] = { CARP_STATES }; 2051 if (sc->sc_state == state) 2052 return; 2053 2054 CARP_LOG(sc, ("state transition from: %s -> to: %s", carp_states[sc->sc_state], carp_states[state])); 2055 2056 sc->sc_state = state; 2057 switch (state) { 2058 case BACKUP: 2059 sc->sc_if.if_link_state = LINK_STATE_DOWN; 2060 break; 2061 case MASTER: 2062 sc->sc_if.if_link_state = LINK_STATE_UP; 2063 break; 2064 default: 2065 sc->sc_if.if_link_state = LINK_STATE_UNKNOWN; 2066 break; 2067 } 2068 rt_ifmsg(&sc->sc_if); 2069 } 2070 2071 void 2072 carp_carpdev_state(void *v) 2073 { 2074 struct carp_if *cif; 2075 struct carp_softc *sc; 2076 struct ifnet *ifp = v; 2077 2078 if (ifp->if_type == IFT_CARP) 2079 return; 2080 2081 cif = (struct carp_if *)ifp->if_carp; 2082 2083 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list) { 2084 int suppressed = sc->sc_suppress; 2085 2086 if (sc->sc_carpdev->if_link_state == LINK_STATE_DOWN || 2087 !(sc->sc_carpdev->if_flags & IFF_UP)) { 2088 sc->sc_if.if_flags &= ~IFF_RUNNING; 2089 callout_stop(&sc->sc_ad_tmo); 2090 callout_stop(&sc->sc_md_tmo); 2091 callout_stop(&sc->sc_md6_tmo); 2092 carp_set_state(sc, INIT); 2093 sc->sc_suppress = 1; 2094 carp_setrun(sc, 0); 2095 if (!suppressed) { 2096 carp_suppress_preempt++; 2097 if (carp_suppress_preempt == 1) 2098 carp_send_ad_all(); 2099 } 2100 } else { 2101 carp_set_state(sc, INIT); 2102 sc->sc_suppress = 0; 2103 carp_setrun(sc, 0); 2104 if (suppressed) 2105 carp_suppress_preempt--; 2106 } 2107 } 2108 } 2109 2110 int 2111 carp_ether_addmulti(struct carp_softc *sc, struct ifreq *ifr) 2112 { 2113 const struct sockaddr *sa = ifreq_getaddr(SIOCADDMULTI, ifr); 2114 struct ifnet *ifp; 2115 struct carp_mc_entry *mc; 2116 u_int8_t addrlo[ETHER_ADDR_LEN], addrhi[ETHER_ADDR_LEN]; 2117 int error; 2118 2119 ifp = sc->sc_carpdev; 2120 if (ifp == NULL) 2121 return (EINVAL); 2122 2123 error = ether_addmulti(sa, &sc->sc_ac); 2124 if (error != ENETRESET) 2125 return (error); 2126 2127 /* 2128 * This is new multicast address. We have to tell parent 2129 * about it. Also, remember this multicast address so that 2130 * we can delete them on unconfigure. 2131 */ 2132 mc = malloc(sizeof(struct carp_mc_entry), M_DEVBUF, M_NOWAIT); 2133 if (mc == NULL) { 2134 error = ENOMEM; 2135 goto alloc_failed; 2136 } 2137 2138 /* 2139 * As ether_addmulti() returns ENETRESET, following two 2140 * statement shouldn't fail. 2141 */ 2142 (void)ether_multiaddr(sa, addrlo, addrhi); 2143 ETHER_LOOKUP_MULTI(addrlo, addrhi, &sc->sc_ac, mc->mc_enm); 2144 memcpy(&mc->mc_addr, sa, sa->sa_len); 2145 LIST_INSERT_HEAD(&sc->carp_mc_listhead, mc, mc_entries); 2146 2147 error = if_mcast_op(ifp, SIOCADDMULTI, sa); 2148 if (error != 0) 2149 goto ioctl_failed; 2150 2151 return (error); 2152 2153 ioctl_failed: 2154 LIST_REMOVE(mc, mc_entries); 2155 free(mc, M_DEVBUF); 2156 alloc_failed: 2157 (void)ether_delmulti(sa, &sc->sc_ac); 2158 2159 return (error); 2160 } 2161 2162 int 2163 carp_ether_delmulti(struct carp_softc *sc, struct ifreq *ifr) 2164 { 2165 const struct sockaddr *sa = ifreq_getaddr(SIOCDELMULTI, ifr); 2166 struct ifnet *ifp; 2167 struct ether_multi *enm; 2168 struct carp_mc_entry *mc; 2169 u_int8_t addrlo[ETHER_ADDR_LEN], addrhi[ETHER_ADDR_LEN]; 2170 int error; 2171 2172 ifp = sc->sc_carpdev; 2173 if (ifp == NULL) 2174 return (EINVAL); 2175 2176 /* 2177 * Find a key to lookup carp_mc_entry. We have to do this 2178 * before calling ether_delmulti for obvious reason. 2179 */ 2180 if ((error = ether_multiaddr(sa, addrlo, addrhi)) != 0) 2181 return (error); 2182 ETHER_LOOKUP_MULTI(addrlo, addrhi, &sc->sc_ac, enm); 2183 if (enm == NULL) 2184 return (EINVAL); 2185 2186 LIST_FOREACH(mc, &sc->carp_mc_listhead, mc_entries) 2187 if (mc->mc_enm == enm) 2188 break; 2189 2190 /* We won't delete entries we didn't add */ 2191 if (mc == NULL) 2192 return (EINVAL); 2193 2194 error = ether_delmulti(sa, &sc->sc_ac); 2195 if (error != ENETRESET) 2196 return (error); 2197 2198 /* We no longer use this multicast address. Tell parent so. */ 2199 error = if_mcast_op(ifp, SIOCDELMULTI, sa); 2200 if (error == 0) { 2201 /* And forget about this address. */ 2202 LIST_REMOVE(mc, mc_entries); 2203 free(mc, M_DEVBUF); 2204 } else 2205 (void)ether_addmulti(sa, &sc->sc_ac); 2206 return (error); 2207 } 2208 2209 /* 2210 * Delete any multicast address we have asked to add from parent 2211 * interface. Called when the carp is being unconfigured. 2212 */ 2213 void 2214 carp_ether_purgemulti(struct carp_softc *sc) 2215 { 2216 struct ifnet *ifp = sc->sc_carpdev; /* Parent. */ 2217 struct carp_mc_entry *mc; 2218 2219 if (ifp == NULL) 2220 return; 2221 2222 while ((mc = LIST_FIRST(&sc->carp_mc_listhead)) != NULL) { 2223 (void)if_mcast_op(ifp, SIOCDELMULTI, sstosa(&mc->mc_addr)); 2224 LIST_REMOVE(mc, mc_entries); 2225 free(mc, M_DEVBUF); 2226 } 2227 } 2228 2229 static int 2230 sysctl_net_inet_carp_stats(SYSCTLFN_ARGS) 2231 { 2232 2233 return (NETSTAT_SYSCTL(carpstat_percpu, CARP_NSTATS)); 2234 } 2235 2236 void 2237 carp_init(void) 2238 { 2239 2240 sysctl_net_inet_carp_setup(NULL); 2241 } 2242 2243 static void 2244 sysctl_net_inet_carp_setup(struct sysctllog **clog) 2245 { 2246 2247 sysctl_createv(clog, 0, NULL, NULL, 2248 CTLFLAG_PERMANENT, 2249 CTLTYPE_NODE, "net", NULL, 2250 NULL, 0, NULL, 0, 2251 CTL_NET, CTL_EOL); 2252 sysctl_createv(clog, 0, NULL, NULL, 2253 CTLFLAG_PERMANENT, 2254 CTLTYPE_NODE, "inet", NULL, 2255 NULL, 0, NULL, 0, 2256 CTL_NET, PF_INET, CTL_EOL); 2257 sysctl_createv(clog, 0, NULL, NULL, 2258 CTLFLAG_PERMANENT, 2259 CTLTYPE_NODE, "carp", 2260 SYSCTL_DESCR("CARP related settings"), 2261 NULL, 0, NULL, 0, 2262 CTL_NET, PF_INET, IPPROTO_CARP, CTL_EOL); 2263 2264 sysctl_createv(clog, 0, NULL, NULL, 2265 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2266 CTLTYPE_INT, "preempt", 2267 SYSCTL_DESCR("Enable CARP Preempt"), 2268 NULL, 0, &carp_opts[CARPCTL_PREEMPT], 0, 2269 CTL_NET, PF_INET, IPPROTO_CARP, 2270 CTL_CREATE, CTL_EOL); 2271 sysctl_createv(clog, 0, NULL, NULL, 2272 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2273 CTLTYPE_INT, "arpbalance", 2274 SYSCTL_DESCR("Enable ARP balancing"), 2275 NULL, 0, &carp_opts[CARPCTL_ARPBALANCE], 0, 2276 CTL_NET, PF_INET, IPPROTO_CARP, 2277 CTL_CREATE, CTL_EOL); 2278 sysctl_createv(clog, 0, NULL, NULL, 2279 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2280 CTLTYPE_INT, "allow", 2281 SYSCTL_DESCR("Enable CARP"), 2282 NULL, 0, &carp_opts[CARPCTL_ALLOW], 0, 2283 CTL_NET, PF_INET, IPPROTO_CARP, 2284 CTL_CREATE, CTL_EOL); 2285 sysctl_createv(clog, 0, NULL, NULL, 2286 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2287 CTLTYPE_INT, "log", 2288 SYSCTL_DESCR("CARP logging"), 2289 NULL, 0, &carp_opts[CARPCTL_LOG], 0, 2290 CTL_NET, PF_INET, IPPROTO_CARP, 2291 CTL_CREATE, CTL_EOL); 2292 sysctl_createv(clog, 0, NULL, NULL, 2293 CTLFLAG_PERMANENT, 2294 CTLTYPE_STRUCT, "stats", 2295 SYSCTL_DESCR("CARP statistics"), 2296 sysctl_net_inet_carp_stats, 0, NULL, 0, 2297 CTL_NET, PF_INET, IPPROTO_CARP, CARPCTL_STATS, 2298 CTL_EOL); 2299 } 2300