1 /* $NetBSD: ip_carp.c,v 1.57 2014/06/06 01:02:47 rmind 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 #include "opt_mbuftrace.h" 32 33 #include <sys/cdefs.h> 34 __KERNEL_RCSID(0, "$NetBSD: ip_carp.c,v 1.57 2014/06/06 01:02:47 rmind Exp $"); 35 36 /* 37 * TODO: 38 * - iface reconfigure 39 * - support for hardware checksum calculations; 40 * 41 */ 42 43 #include <sys/param.h> 44 #include <sys/proc.h> 45 #include <sys/mbuf.h> 46 #include <sys/socket.h> 47 #include <sys/socketvar.h> 48 #include <sys/callout.h> 49 #include <sys/ioctl.h> 50 #include <sys/errno.h> 51 #include <sys/device.h> 52 #include <sys/time.h> 53 #include <sys/kernel.h> 54 #include <sys/kauth.h> 55 #include <sys/sysctl.h> 56 #include <sys/ucred.h> 57 #include <sys/syslog.h> 58 #include <sys/acct.h> 59 #include <sys/cprng.h> 60 61 #include <sys/cpu.h> 62 63 #include <net/if.h> 64 #include <net/pfil.h> 65 #include <net/if_types.h> 66 #include <net/if_ether.h> 67 #include <net/route.h> 68 #include <net/netisr.h> 69 #include <net/net_stats.h> 70 #include <netinet/if_inarp.h> 71 72 #if NFDDI > 0 73 #include <net/if_fddi.h> 74 #endif 75 #if NTOKEN > 0 76 #include <net/if_token.h> 77 #endif 78 79 #ifdef INET 80 #include <netinet/in.h> 81 #include <netinet/in_systm.h> 82 #include <netinet/in_var.h> 83 #include <netinet/ip.h> 84 #include <netinet/ip_var.h> 85 86 #include <net/if_dl.h> 87 #endif 88 89 #ifdef INET6 90 #include <netinet/icmp6.h> 91 #include <netinet/ip6.h> 92 #include <netinet6/ip6_var.h> 93 #include <netinet6/nd6.h> 94 #include <netinet6/scope6_var.h> 95 #endif 96 97 #include <net/bpf.h> 98 99 #include <sys/sha1.h> 100 101 #include <netinet/ip_carp.h> 102 103 struct carp_mc_entry { 104 LIST_ENTRY(carp_mc_entry) mc_entries; 105 union { 106 struct ether_multi *mcu_enm; 107 } mc_u; 108 struct sockaddr_storage mc_addr; 109 }; 110 #define mc_enm mc_u.mcu_enm 111 112 struct carp_softc { 113 struct ethercom sc_ac; 114 #define sc_if sc_ac.ec_if 115 #define sc_carpdev sc_ac.ec_if.if_carpdev 116 int ah_cookie; 117 int lh_cookie; 118 struct ip_moptions sc_imo; 119 #ifdef INET6 120 struct ip6_moptions sc_im6o; 121 #endif /* INET6 */ 122 TAILQ_ENTRY(carp_softc) sc_list; 123 124 enum { INIT = 0, BACKUP, MASTER } sc_state; 125 126 int sc_suppress; 127 int sc_bow_out; 128 129 int sc_sendad_errors; 130 #define CARP_SENDAD_MAX_ERRORS 3 131 int sc_sendad_success; 132 #define CARP_SENDAD_MIN_SUCCESS 3 133 134 int sc_vhid; 135 int sc_advskew; 136 int sc_naddrs; 137 int sc_naddrs6; 138 int sc_advbase; /* seconds */ 139 int sc_init_counter; 140 u_int64_t sc_counter; 141 142 /* authentication */ 143 #define CARP_HMAC_PAD 64 144 unsigned char sc_key[CARP_KEY_LEN]; 145 unsigned char sc_pad[CARP_HMAC_PAD]; 146 SHA1_CTX sc_sha1; 147 u_int32_t sc_hashkey[2]; 148 149 struct callout sc_ad_tmo; /* advertisement timeout */ 150 struct callout sc_md_tmo; /* master down timeout */ 151 struct callout sc_md6_tmo; /* master down timeout */ 152 153 LIST_HEAD(__carp_mchead, carp_mc_entry) carp_mc_listhead; 154 }; 155 156 int carp_suppress_preempt = 0; 157 int carp_opts[CARPCTL_MAXID] = { 0, 1, 0, 0, 0 }; /* XXX for now */ 158 159 static percpu_t *carpstat_percpu; 160 161 #define CARP_STATINC(x) _NET_STATINC(carpstat_percpu, x) 162 163 #ifdef MBUFTRACE 164 static struct mowner carp_proto_mowner_rx = MOWNER_INIT("carp", "rx"); 165 static struct mowner carp_proto_mowner_tx = MOWNER_INIT("carp", "tx"); 166 static struct mowner carp_proto6_mowner_rx = MOWNER_INIT("carp6", "rx"); 167 static struct mowner carp_proto6_mowner_tx = MOWNER_INIT("carp6", "tx"); 168 #endif 169 170 struct carp_if { 171 TAILQ_HEAD(, carp_softc) vhif_vrs; 172 int vhif_nvrs; 173 174 struct ifnet *vhif_ifp; 175 }; 176 177 #define CARP_LOG(sc, s) \ 178 if (carp_opts[CARPCTL_LOG]) { \ 179 if (sc) \ 180 log(LOG_INFO, "%s: ", \ 181 (sc)->sc_if.if_xname); \ 182 else \ 183 log(LOG_INFO, "carp: "); \ 184 addlog s; \ 185 addlog("\n"); \ 186 } 187 188 void carp_hmac_prepare(struct carp_softc *); 189 void carp_hmac_generate(struct carp_softc *, u_int32_t *, 190 unsigned char *); 191 int carp_hmac_verify(struct carp_softc *, u_int32_t *, 192 unsigned char *); 193 void carp_setroute(struct carp_softc *, int); 194 void carp_proto_input_c(struct mbuf *, struct carp_header *, sa_family_t); 195 void carpattach(int); 196 void carpdetach(struct carp_softc *); 197 int carp_prepare_ad(struct mbuf *, struct carp_softc *, 198 struct carp_header *); 199 void carp_send_ad_all(void); 200 void carp_send_ad(void *); 201 void carp_send_arp(struct carp_softc *); 202 void carp_master_down(void *); 203 int carp_ioctl(struct ifnet *, u_long, void *); 204 void carp_start(struct ifnet *); 205 void carp_setrun(struct carp_softc *, sa_family_t); 206 void carp_set_state(struct carp_softc *, int); 207 int carp_addrcount(struct carp_if *, struct in_ifaddr *, int); 208 enum { CARP_COUNT_MASTER, CARP_COUNT_RUNNING }; 209 210 void carp_multicast_cleanup(struct carp_softc *); 211 int carp_set_ifp(struct carp_softc *, struct ifnet *); 212 void carp_set_enaddr(struct carp_softc *); 213 void carp_addr_updated(void *); 214 u_int32_t carp_hash(struct carp_softc *, u_char *); 215 int carp_set_addr(struct carp_softc *, struct sockaddr_in *); 216 int carp_join_multicast(struct carp_softc *); 217 #ifdef INET6 218 void carp_send_na(struct carp_softc *); 219 int carp_set_addr6(struct carp_softc *, struct sockaddr_in6 *); 220 int carp_join_multicast6(struct carp_softc *); 221 #endif 222 int carp_clone_create(struct if_clone *, int); 223 int carp_clone_destroy(struct ifnet *); 224 int carp_ether_addmulti(struct carp_softc *, struct ifreq *); 225 int carp_ether_delmulti(struct carp_softc *, struct ifreq *); 226 void carp_ether_purgemulti(struct carp_softc *); 227 228 static void sysctl_net_inet_carp_setup(struct sysctllog **); 229 230 struct if_clone carp_cloner = 231 IF_CLONE_INITIALIZER("carp", carp_clone_create, carp_clone_destroy); 232 233 static __inline u_int16_t 234 carp_cksum(struct mbuf *m, int len) 235 { 236 return (in_cksum(m, len)); 237 } 238 239 void 240 carp_hmac_prepare(struct carp_softc *sc) 241 { 242 u_int8_t carp_version = CARP_VERSION, type = CARP_ADVERTISEMENT; 243 u_int8_t vhid = sc->sc_vhid & 0xff; 244 SHA1_CTX sha1ctx; 245 u_int32_t kmd[5]; 246 struct ifaddr *ifa; 247 int i, found; 248 struct in_addr last, cur, in; 249 #ifdef INET6 250 struct in6_addr last6, cur6, in6; 251 #endif /* INET6 */ 252 253 /* compute ipad from key */ 254 memset(sc->sc_pad, 0, sizeof(sc->sc_pad)); 255 memcpy(sc->sc_pad, sc->sc_key, sizeof(sc->sc_key)); 256 for (i = 0; i < sizeof(sc->sc_pad); i++) 257 sc->sc_pad[i] ^= 0x36; 258 259 /* precompute first part of inner hash */ 260 SHA1Init(&sc->sc_sha1); 261 SHA1Update(&sc->sc_sha1, sc->sc_pad, sizeof(sc->sc_pad)); 262 SHA1Update(&sc->sc_sha1, (void *)&carp_version, sizeof(carp_version)); 263 SHA1Update(&sc->sc_sha1, (void *)&type, sizeof(type)); 264 265 /* generate a key for the arpbalance hash, before the vhid is hashed */ 266 memcpy(&sha1ctx, &sc->sc_sha1, sizeof(sha1ctx)); 267 SHA1Final((unsigned char *)kmd, &sha1ctx); 268 sc->sc_hashkey[0] = kmd[0] ^ kmd[1]; 269 sc->sc_hashkey[1] = kmd[2] ^ kmd[3]; 270 271 /* the rest of the precomputation */ 272 SHA1Update(&sc->sc_sha1, (void *)&vhid, sizeof(vhid)); 273 274 /* Hash the addresses from smallest to largest, not interface order */ 275 #ifdef INET 276 cur.s_addr = 0; 277 do { 278 found = 0; 279 last = cur; 280 cur.s_addr = 0xffffffff; 281 IFADDR_FOREACH(ifa, &sc->sc_if) { 282 in.s_addr = ifatoia(ifa)->ia_addr.sin_addr.s_addr; 283 if (ifa->ifa_addr->sa_family == AF_INET && 284 ntohl(in.s_addr) > ntohl(last.s_addr) && 285 ntohl(in.s_addr) < ntohl(cur.s_addr)) { 286 cur.s_addr = in.s_addr; 287 found++; 288 } 289 } 290 if (found) 291 SHA1Update(&sc->sc_sha1, (void *)&cur, sizeof(cur)); 292 } while (found); 293 #endif /* INET */ 294 295 #ifdef INET6 296 memset(&cur6, 0x00, sizeof(cur6)); 297 do { 298 found = 0; 299 last6 = cur6; 300 memset(&cur6, 0xff, sizeof(cur6)); 301 IFADDR_FOREACH(ifa, &sc->sc_if) { 302 in6 = ifatoia6(ifa)->ia_addr.sin6_addr; 303 if (IN6_IS_ADDR_LINKLOCAL(&in6)) 304 in6.s6_addr16[1] = 0; 305 if (ifa->ifa_addr->sa_family == AF_INET6 && 306 memcmp(&in6, &last6, sizeof(in6)) > 0 && 307 memcmp(&in6, &cur6, sizeof(in6)) < 0) { 308 cur6 = in6; 309 found++; 310 } 311 } 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 } 321 322 void 323 carp_hmac_generate(struct carp_softc *sc, u_int32_t counter[2], 324 unsigned char md[20]) 325 { 326 SHA1_CTX sha1ctx; 327 328 /* fetch first half of inner hash */ 329 memcpy(&sha1ctx, &sc->sc_sha1, sizeof(sha1ctx)); 330 331 SHA1Update(&sha1ctx, (void *)counter, sizeof(sc->sc_counter)); 332 SHA1Final(md, &sha1ctx); 333 334 /* outer hash */ 335 SHA1Init(&sha1ctx); 336 SHA1Update(&sha1ctx, sc->sc_pad, sizeof(sc->sc_pad)); 337 SHA1Update(&sha1ctx, md, 20); 338 SHA1Final(md, &sha1ctx); 339 } 340 341 int 342 carp_hmac_verify(struct carp_softc *sc, u_int32_t counter[2], 343 unsigned char md[20]) 344 { 345 unsigned char md2[20]; 346 347 carp_hmac_generate(sc, counter, md2); 348 349 return (memcmp(md, md2, sizeof(md2))); 350 } 351 352 void 353 carp_setroute(struct carp_softc *sc, int cmd) 354 { 355 struct ifaddr *ifa; 356 int s; 357 358 KERNEL_LOCK(1, NULL); 359 s = splsoftnet(); 360 IFADDR_FOREACH(ifa, &sc->sc_if) { 361 switch (ifa->ifa_addr->sa_family) { 362 case AF_INET: { 363 int count = 0; 364 struct rtentry *rt; 365 int hr_otherif, nr_ourif; 366 367 /* 368 * Avoid screwing with the routes if there are other 369 * carp interfaces which are master and have the same 370 * address. 371 */ 372 if (sc->sc_carpdev != NULL && 373 sc->sc_carpdev->if_carp != NULL) { 374 count = carp_addrcount( 375 (struct carp_if *)sc->sc_carpdev->if_carp, 376 ifatoia(ifa), CARP_COUNT_MASTER); 377 if ((cmd == RTM_ADD && count != 1) || 378 (cmd == RTM_DELETE && count != 0)) 379 continue; 380 } 381 382 /* Remove the existing host route, if any */ 383 rtrequest(RTM_DELETE, ifa->ifa_addr, 384 ifa->ifa_addr, ifa->ifa_netmask, 385 RTF_HOST, NULL); 386 387 rt = NULL; 388 (void)rtrequest(RTM_GET, ifa->ifa_addr, ifa->ifa_addr, 389 ifa->ifa_netmask, RTF_HOST, &rt); 390 hr_otherif = (rt && rt->rt_ifp != &sc->sc_if && 391 rt->rt_flags & (RTF_CLONING|RTF_CLONED)); 392 if (rt != NULL) { 393 rtfree(rt); 394 rt = NULL; 395 } 396 397 /* Check for a network route on our interface */ 398 399 rt = NULL; 400 (void)rtrequest(RTM_GET, ifa->ifa_addr, ifa->ifa_addr, 401 ifa->ifa_netmask, 0, &rt); 402 nr_ourif = (rt && rt->rt_ifp == &sc->sc_if); 403 404 switch (cmd) { 405 case RTM_ADD: 406 if (hr_otherif) { 407 ifa->ifa_rtrequest = NULL; 408 ifa->ifa_flags &= ~RTF_CLONING; 409 410 rtrequest(RTM_ADD, ifa->ifa_addr, 411 ifa->ifa_addr, ifa->ifa_netmask, 412 RTF_UP | RTF_HOST, NULL); 413 } 414 if (!hr_otherif || nr_ourif || !rt) { 415 if (nr_ourif && !(rt->rt_flags & 416 RTF_CLONING)) 417 rtrequest(RTM_DELETE, 418 ifa->ifa_addr, 419 ifa->ifa_addr, 420 ifa->ifa_netmask, 0, NULL); 421 422 ifa->ifa_rtrequest = arp_rtrequest; 423 ifa->ifa_flags |= RTF_CLONING; 424 425 if (rtrequest(RTM_ADD, ifa->ifa_addr, 426 ifa->ifa_addr, ifa->ifa_netmask, 0, 427 NULL) == 0) 428 ifa->ifa_flags |= IFA_ROUTE; 429 } 430 break; 431 case RTM_DELETE: 432 break; 433 default: 434 break; 435 } 436 if (rt != NULL) { 437 rtfree(rt); 438 rt = NULL; 439 } 440 break; 441 } 442 443 #ifdef INET6 444 case AF_INET6: 445 if (cmd == RTM_ADD) 446 in6_ifaddloop(ifa); 447 else 448 in6_ifremloop(ifa); 449 break; 450 #endif /* INET6 */ 451 default: 452 break; 453 } 454 } 455 splx(s); 456 KERNEL_UNLOCK_ONE(NULL); 457 } 458 459 /* 460 * process input packet. 461 * we have rearranged checks order compared to the rfc, 462 * but it seems more efficient this way or not possible otherwise. 463 */ 464 void 465 carp_proto_input(struct mbuf *m, ...) 466 { 467 struct ip *ip = mtod(m, struct ip *); 468 struct carp_softc *sc = NULL; 469 struct carp_header *ch; 470 int iplen, len; 471 va_list ap; 472 473 va_start(ap, m); 474 va_end(ap); 475 476 CARP_STATINC(CARP_STAT_IPACKETS); 477 MCLAIM(m, &carp_proto_mowner_rx); 478 479 if (!carp_opts[CARPCTL_ALLOW]) { 480 m_freem(m); 481 return; 482 } 483 484 /* check if received on a valid carp interface */ 485 if (m->m_pkthdr.rcvif->if_type != IFT_CARP) { 486 CARP_STATINC(CARP_STAT_BADIF); 487 CARP_LOG(sc, ("packet received on non-carp interface: %s", 488 m->m_pkthdr.rcvif->if_xname)); 489 m_freem(m); 490 return; 491 } 492 493 /* verify that the IP TTL is 255. */ 494 if (ip->ip_ttl != CARP_DFLTTL) { 495 CARP_STATINC(CARP_STAT_BADTTL); 496 CARP_LOG(sc, ("received ttl %d != %d on %s", ip->ip_ttl, 497 CARP_DFLTTL, m->m_pkthdr.rcvif->if_xname)); 498 m_freem(m); 499 return; 500 } 501 502 /* 503 * verify that the received packet length is 504 * equal to the CARP header 505 */ 506 iplen = ip->ip_hl << 2; 507 len = iplen + sizeof(*ch); 508 if (len > m->m_pkthdr.len) { 509 CARP_STATINC(CARP_STAT_BADLEN); 510 CARP_LOG(sc, ("packet too short %d on %s", m->m_pkthdr.len, 511 m->m_pkthdr.rcvif->if_xname)); 512 m_freem(m); 513 return; 514 } 515 516 if ((m = m_pullup(m, len)) == NULL) { 517 CARP_STATINC(CARP_STAT_HDROPS); 518 return; 519 } 520 ip = mtod(m, struct ip *); 521 ch = (struct carp_header *)((char *)ip + iplen); 522 /* verify the CARP checksum */ 523 m->m_data += iplen; 524 if (carp_cksum(m, len - iplen)) { 525 CARP_STATINC(CARP_STAT_BADSUM); 526 CARP_LOG(sc, ("checksum failed on %s", 527 m->m_pkthdr.rcvif->if_xname)); 528 m_freem(m); 529 return; 530 } 531 m->m_data -= iplen; 532 533 carp_proto_input_c(m, ch, AF_INET); 534 } 535 536 #ifdef INET6 537 int 538 carp6_proto_input(struct mbuf **mp, int *offp, int proto) 539 { 540 struct mbuf *m = *mp; 541 struct carp_softc *sc = NULL; 542 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 543 struct carp_header *ch; 544 u_int len; 545 546 CARP_STATINC(CARP_STAT_IPACKETS6); 547 MCLAIM(m, &carp_proto6_mowner_rx); 548 549 if (!carp_opts[CARPCTL_ALLOW]) { 550 m_freem(m); 551 return (IPPROTO_DONE); 552 } 553 554 /* check if received on a valid carp interface */ 555 if (m->m_pkthdr.rcvif->if_type != IFT_CARP) { 556 CARP_STATINC(CARP_STAT_BADIF); 557 CARP_LOG(sc, ("packet received on non-carp interface: %s", 558 m->m_pkthdr.rcvif->if_xname)); 559 m_freem(m); 560 return (IPPROTO_DONE); 561 } 562 563 /* verify that the IP TTL is 255 */ 564 if (ip6->ip6_hlim != CARP_DFLTTL) { 565 CARP_STATINC(CARP_STAT_BADTTL); 566 CARP_LOG(sc, ("received ttl %d != %d on %s", ip6->ip6_hlim, 567 CARP_DFLTTL, m->m_pkthdr.rcvif->if_xname)); 568 m_freem(m); 569 return (IPPROTO_DONE); 570 } 571 572 /* verify that we have a complete carp packet */ 573 len = m->m_len; 574 IP6_EXTHDR_GET(ch, struct carp_header *, m, *offp, sizeof(*ch)); 575 if (ch == NULL) { 576 CARP_STATINC(CARP_STAT_BADLEN); 577 CARP_LOG(sc, ("packet size %u too small", len)); 578 return (IPPROTO_DONE); 579 } 580 581 582 /* verify the CARP checksum */ 583 m->m_data += *offp; 584 if (carp_cksum(m, sizeof(*ch))) { 585 CARP_STATINC(CARP_STAT_BADSUM); 586 CARP_LOG(sc, ("checksum failed, on %s", 587 m->m_pkthdr.rcvif->if_xname)); 588 m_freem(m); 589 return (IPPROTO_DONE); 590 } 591 m->m_data -= *offp; 592 593 carp_proto_input_c(m, ch, AF_INET6); 594 return (IPPROTO_DONE); 595 } 596 #endif /* INET6 */ 597 598 void 599 carp_proto_input_c(struct mbuf *m, struct carp_header *ch, sa_family_t af) 600 { 601 struct carp_softc *sc; 602 u_int64_t tmp_counter; 603 struct timeval sc_tv, ch_tv; 604 605 TAILQ_FOREACH(sc, &((struct carp_if *) 606 m->m_pkthdr.rcvif->if_carpdev->if_carp)->vhif_vrs, sc_list) 607 if (sc->sc_vhid == ch->carp_vhid) 608 break; 609 610 if (!sc || (sc->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) != 611 (IFF_UP|IFF_RUNNING)) { 612 CARP_STATINC(CARP_STAT_BADVHID); 613 m_freem(m); 614 return; 615 } 616 617 /* 618 * Check if our own advertisement was duplicated 619 * from a non simplex interface. 620 * XXX If there is no address on our physical interface 621 * there is no way to distinguish our ads from the ones 622 * another carp host might have sent us. 623 */ 624 if ((sc->sc_carpdev->if_flags & IFF_SIMPLEX) == 0) { 625 struct sockaddr sa; 626 struct ifaddr *ifa; 627 628 memset(&sa, 0, sizeof(sa)); 629 sa.sa_family = af; 630 ifa = ifaof_ifpforaddr(&sa, sc->sc_carpdev); 631 632 if (ifa && af == AF_INET) { 633 struct ip *ip = mtod(m, struct ip *); 634 if (ip->ip_src.s_addr == 635 ifatoia(ifa)->ia_addr.sin_addr.s_addr) { 636 m_freem(m); 637 return; 638 } 639 } 640 #ifdef INET6 641 if (ifa && af == AF_INET6) { 642 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 643 struct in6_addr in6_src, in6_found; 644 645 in6_src = ip6->ip6_src; 646 in6_found = ifatoia6(ifa)->ia_addr.sin6_addr; 647 if (IN6_IS_ADDR_LINKLOCAL(&in6_src)) 648 in6_src.s6_addr16[1] = 0; 649 if (IN6_IS_ADDR_LINKLOCAL(&in6_found)) 650 in6_found.s6_addr16[1] = 0; 651 if (IN6_ARE_ADDR_EQUAL(&in6_src, &in6_found)) { 652 m_freem(m); 653 return; 654 } 655 } 656 #endif /* INET6 */ 657 } 658 659 nanotime(&sc->sc_if.if_lastchange); 660 sc->sc_if.if_ipackets++; 661 sc->sc_if.if_ibytes += m->m_pkthdr.len; 662 663 /* verify the CARP version. */ 664 if (ch->carp_version != CARP_VERSION) { 665 CARP_STATINC(CARP_STAT_BADVER); 666 sc->sc_if.if_ierrors++; 667 CARP_LOG(sc, ("invalid version %d != %d", 668 ch->carp_version, CARP_VERSION)); 669 m_freem(m); 670 return; 671 } 672 673 /* verify the hash */ 674 if (carp_hmac_verify(sc, ch->carp_counter, ch->carp_md)) { 675 CARP_STATINC(CARP_STAT_BADAUTH); 676 sc->sc_if.if_ierrors++; 677 CARP_LOG(sc, ("incorrect hash")); 678 m_freem(m); 679 return; 680 } 681 682 tmp_counter = ntohl(ch->carp_counter[0]); 683 tmp_counter = tmp_counter<<32; 684 tmp_counter += ntohl(ch->carp_counter[1]); 685 686 /* XXX Replay protection goes here */ 687 688 sc->sc_init_counter = 0; 689 sc->sc_counter = tmp_counter; 690 691 692 sc_tv.tv_sec = sc->sc_advbase; 693 if (carp_suppress_preempt && sc->sc_advskew < 240) 694 sc_tv.tv_usec = 240 * 1000000 / 256; 695 else 696 sc_tv.tv_usec = sc->sc_advskew * 1000000 / 256; 697 ch_tv.tv_sec = ch->carp_advbase; 698 ch_tv.tv_usec = ch->carp_advskew * 1000000 / 256; 699 700 switch (sc->sc_state) { 701 case INIT: 702 break; 703 case MASTER: 704 /* 705 * If we receive an advertisement from a backup who's going to 706 * be more frequent than us, go into BACKUP state. 707 */ 708 if (timercmp(&sc_tv, &ch_tv, >) || 709 timercmp(&sc_tv, &ch_tv, ==)) { 710 callout_stop(&sc->sc_ad_tmo); 711 CARP_LOG(sc, ("MASTER -> BACKUP (more frequent advertisement received)")); 712 carp_set_state(sc, BACKUP); 713 carp_setrun(sc, 0); 714 carp_setroute(sc, RTM_DELETE); 715 } 716 break; 717 case BACKUP: 718 /* 719 * If we're pre-empting masters who advertise slower than us, 720 * and this one claims to be slower, treat him as down. 721 */ 722 if (carp_opts[CARPCTL_PREEMPT] && timercmp(&sc_tv, &ch_tv, <)) { 723 CARP_LOG(sc, ("BACKUP -> MASTER (preempting a slower master)")); 724 carp_master_down(sc); 725 break; 726 } 727 728 /* 729 * If the master is going to advertise at such a low frequency 730 * that he's guaranteed to time out, we'd might as well just 731 * treat him as timed out now. 732 */ 733 sc_tv.tv_sec = sc->sc_advbase * 3; 734 if (timercmp(&sc_tv, &ch_tv, <)) { 735 CARP_LOG(sc, ("BACKUP -> MASTER (master timed out)")); 736 carp_master_down(sc); 737 break; 738 } 739 740 /* 741 * Otherwise, we reset the counter and wait for the next 742 * advertisement. 743 */ 744 carp_setrun(sc, af); 745 break; 746 } 747 748 m_freem(m); 749 return; 750 } 751 752 /* 753 * Interface side of the CARP implementation. 754 */ 755 756 /* ARGSUSED */ 757 void 758 carpattach(int n) 759 { 760 if_clone_attach(&carp_cloner); 761 762 carpstat_percpu = percpu_alloc(sizeof(uint64_t) * CARP_NSTATS); 763 } 764 765 int 766 carp_clone_create(struct if_clone *ifc, int unit) 767 { 768 extern int ifqmaxlen; 769 struct carp_softc *sc; 770 struct ifnet *ifp; 771 772 sc = malloc(sizeof(*sc), M_DEVBUF, M_NOWAIT|M_ZERO); 773 if (!sc) 774 return (ENOMEM); 775 776 sc->sc_suppress = 0; 777 sc->sc_advbase = CARP_DFLTINTV; 778 sc->sc_vhid = -1; /* required setting */ 779 sc->sc_advskew = 0; 780 sc->sc_init_counter = 1; 781 sc->sc_naddrs = sc->sc_naddrs6 = 0; 782 #ifdef INET6 783 sc->sc_im6o.im6o_multicast_hlim = CARP_DFLTTL; 784 #endif /* INET6 */ 785 786 callout_init(&sc->sc_ad_tmo, 0); 787 callout_init(&sc->sc_md_tmo, 0); 788 callout_init(&sc->sc_md6_tmo, 0); 789 790 callout_setfunc(&sc->sc_ad_tmo, carp_send_ad, sc); 791 callout_setfunc(&sc->sc_md_tmo, carp_master_down, sc); 792 callout_setfunc(&sc->sc_md6_tmo, carp_master_down, sc); 793 794 LIST_INIT(&sc->carp_mc_listhead); 795 ifp = &sc->sc_if; 796 ifp->if_softc = sc; 797 snprintf(ifp->if_xname, sizeof ifp->if_xname, "%s%d", ifc->ifc_name, 798 unit); 799 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; 800 ifp->if_ioctl = carp_ioctl; 801 ifp->if_start = carp_start; 802 ifp->if_output = carp_output; 803 ifp->if_type = IFT_CARP; 804 ifp->if_addrlen = ETHER_ADDR_LEN; 805 ifp->if_hdrlen = ETHER_HDR_LEN; 806 ifp->if_mtu = ETHERMTU; 807 IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen); 808 IFQ_SET_READY(&ifp->if_snd); 809 if_attach(ifp); 810 811 if_alloc_sadl(ifp); 812 ifp->if_broadcastaddr = etherbroadcastaddr; 813 carp_set_enaddr(sc); 814 LIST_INIT(&sc->sc_ac.ec_multiaddrs); 815 bpf_attach(ifp, DLT_EN10MB, ETHER_HDR_LEN); 816 #ifdef MBUFTRACE 817 strlcpy(sc->sc_ac.ec_tx_mowner.mo_name, ifp->if_xname, 818 sizeof(sc->sc_ac.ec_tx_mowner.mo_name)); 819 strlcpy(sc->sc_ac.ec_tx_mowner.mo_descr, "tx", 820 sizeof(sc->sc_ac.ec_tx_mowner.mo_descr)); 821 strlcpy(sc->sc_ac.ec_rx_mowner.mo_name, ifp->if_xname, 822 sizeof(sc->sc_ac.ec_rx_mowner.mo_name)); 823 strlcpy(sc->sc_ac.ec_rx_mowner.mo_descr, "rx", 824 sizeof(sc->sc_ac.ec_rx_mowner.mo_descr)); 825 MOWNER_ATTACH(&sc->sc_ac.ec_tx_mowner); 826 MOWNER_ATTACH(&sc->sc_ac.ec_rx_mowner); 827 ifp->if_mowner = &sc->sc_ac.ec_tx_mowner; 828 #endif 829 return (0); 830 } 831 832 int 833 carp_clone_destroy(struct ifnet *ifp) 834 { 835 struct carp_softc *sc = ifp->if_softc; 836 837 carpdetach(ifp->if_softc); 838 ether_ifdetach(ifp); 839 if_detach(ifp); 840 callout_destroy(&sc->sc_ad_tmo); 841 callout_destroy(&sc->sc_md_tmo); 842 callout_destroy(&sc->sc_md6_tmo); 843 free(ifp->if_softc, M_DEVBUF); 844 845 return (0); 846 } 847 848 void 849 carpdetach(struct carp_softc *sc) 850 { 851 struct carp_if *cif; 852 int s; 853 854 callout_stop(&sc->sc_ad_tmo); 855 callout_stop(&sc->sc_md_tmo); 856 callout_stop(&sc->sc_md6_tmo); 857 858 if (sc->sc_suppress) 859 carp_suppress_preempt--; 860 sc->sc_suppress = 0; 861 862 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) 863 carp_suppress_preempt--; 864 sc->sc_sendad_errors = 0; 865 866 carp_set_state(sc, INIT); 867 sc->sc_if.if_flags &= ~IFF_UP; 868 carp_setrun(sc, 0); 869 carp_multicast_cleanup(sc); 870 871 KERNEL_LOCK(1, NULL); 872 s = splnet(); 873 if (sc->sc_carpdev != NULL) { 874 /* XXX linkstatehook removal */ 875 cif = (struct carp_if *)sc->sc_carpdev->if_carp; 876 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list); 877 if (!--cif->vhif_nvrs) { 878 ifpromisc(sc->sc_carpdev, 0); 879 sc->sc_carpdev->if_carp = NULL; 880 free(cif, M_IFADDR); 881 } 882 } 883 sc->sc_carpdev = NULL; 884 splx(s); 885 KERNEL_UNLOCK_ONE(NULL); 886 } 887 888 /* Detach an interface from the carp. */ 889 void 890 carp_ifdetach(struct ifnet *ifp) 891 { 892 struct carp_softc *sc, *nextsc; 893 struct carp_if *cif = (struct carp_if *)ifp->if_carp; 894 895 for (sc = TAILQ_FIRST(&cif->vhif_vrs); sc; sc = nextsc) { 896 nextsc = TAILQ_NEXT(sc, sc_list); 897 carpdetach(sc); 898 } 899 } 900 901 int 902 carp_prepare_ad(struct mbuf *m, struct carp_softc *sc, 903 struct carp_header *ch) 904 { 905 if (sc->sc_init_counter) { 906 /* this could also be seconds since unix epoch */ 907 sc->sc_counter = cprng_fast64(); 908 } else 909 sc->sc_counter++; 910 911 ch->carp_counter[0] = htonl((sc->sc_counter>>32)&0xffffffff); 912 ch->carp_counter[1] = htonl(sc->sc_counter&0xffffffff); 913 914 carp_hmac_generate(sc, ch->carp_counter, ch->carp_md); 915 916 return (0); 917 } 918 919 void 920 carp_send_ad_all(void) 921 { 922 struct ifnet *ifp; 923 struct carp_if *cif; 924 struct carp_softc *vh; 925 926 IFNET_FOREACH(ifp) { 927 if (ifp->if_carp == NULL || ifp->if_type == IFT_CARP) 928 continue; 929 930 cif = (struct carp_if *)ifp->if_carp; 931 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 932 if ((vh->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) == 933 (IFF_UP|IFF_RUNNING) && vh->sc_state == MASTER) 934 carp_send_ad(vh); 935 } 936 } 937 } 938 939 940 void 941 carp_send_ad(void *v) 942 { 943 struct carp_header ch; 944 struct timeval tv; 945 struct carp_softc *sc = v; 946 struct carp_header *ch_ptr; 947 struct mbuf *m; 948 int error, len, advbase, advskew, s; 949 struct ifaddr *ifa; 950 struct sockaddr sa; 951 952 KERNEL_LOCK(1, NULL); 953 s = splsoftnet(); 954 955 advbase = advskew = 0; /* Sssssh compiler */ 956 if (sc->sc_carpdev == NULL) { 957 sc->sc_if.if_oerrors++; 958 goto retry_later; 959 } 960 961 /* bow out if we've gone to backup (the carp interface is going down) */ 962 if (sc->sc_bow_out) { 963 sc->sc_bow_out = 0; 964 advbase = 255; 965 advskew = 255; 966 } else { 967 advbase = sc->sc_advbase; 968 if (!carp_suppress_preempt || sc->sc_advskew > 240) 969 advskew = sc->sc_advskew; 970 else 971 advskew = 240; 972 tv.tv_sec = advbase; 973 tv.tv_usec = advskew * 1000000 / 256; 974 } 975 976 ch.carp_version = CARP_VERSION; 977 ch.carp_type = CARP_ADVERTISEMENT; 978 ch.carp_vhid = sc->sc_vhid; 979 ch.carp_advbase = advbase; 980 ch.carp_advskew = advskew; 981 ch.carp_authlen = 7; /* XXX DEFINE */ 982 ch.carp_pad1 = 0; /* must be zero */ 983 ch.carp_cksum = 0; 984 985 986 #ifdef INET 987 if (sc->sc_naddrs) { 988 struct ip *ip; 989 990 MGETHDR(m, M_DONTWAIT, MT_HEADER); 991 if (m == NULL) { 992 sc->sc_if.if_oerrors++; 993 CARP_STATINC(CARP_STAT_ONOMEM); 994 /* XXX maybe less ? */ 995 goto retry_later; 996 } 997 MCLAIM(m, &carp_proto_mowner_tx); 998 len = sizeof(*ip) + sizeof(ch); 999 m->m_pkthdr.len = len; 1000 m->m_pkthdr.rcvif = NULL; 1001 m->m_len = len; 1002 MH_ALIGN(m, m->m_len); 1003 m->m_flags |= M_MCAST; 1004 ip = mtod(m, struct ip *); 1005 ip->ip_v = IPVERSION; 1006 ip->ip_hl = sizeof(*ip) >> 2; 1007 ip->ip_tos = IPTOS_LOWDELAY; 1008 ip->ip_len = htons(len); 1009 ip->ip_id = 0; /* no need for id, we don't support fragments */ 1010 ip->ip_off = htons(IP_DF); 1011 ip->ip_ttl = CARP_DFLTTL; 1012 ip->ip_p = IPPROTO_CARP; 1013 ip->ip_sum = 0; 1014 1015 memset(&sa, 0, sizeof(sa)); 1016 sa.sa_family = AF_INET; 1017 ifa = ifaof_ifpforaddr(&sa, sc->sc_carpdev); 1018 if (ifa == NULL) 1019 ip->ip_src.s_addr = 0; 1020 else 1021 ip->ip_src.s_addr = 1022 ifatoia(ifa)->ia_addr.sin_addr.s_addr; 1023 ip->ip_dst.s_addr = INADDR_CARP_GROUP; 1024 1025 ch_ptr = (struct carp_header *)(&ip[1]); 1026 memcpy(ch_ptr, &ch, sizeof(ch)); 1027 if (carp_prepare_ad(m, sc, ch_ptr)) 1028 goto retry_later; 1029 1030 m->m_data += sizeof(*ip); 1031 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip)); 1032 m->m_data -= sizeof(*ip); 1033 1034 nanotime(&sc->sc_if.if_lastchange); 1035 sc->sc_if.if_opackets++; 1036 sc->sc_if.if_obytes += len; 1037 CARP_STATINC(CARP_STAT_OPACKETS); 1038 1039 error = ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, 1040 NULL); 1041 if (error) { 1042 if (error == ENOBUFS) 1043 CARP_STATINC(CARP_STAT_ONOMEM); 1044 else 1045 CARP_LOG(sc, ("ip_output failed: %d", error)); 1046 sc->sc_if.if_oerrors++; 1047 if (sc->sc_sendad_errors < INT_MAX) 1048 sc->sc_sendad_errors++; 1049 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) { 1050 carp_suppress_preempt++; 1051 if (carp_suppress_preempt == 1) 1052 carp_send_ad_all(); 1053 } 1054 sc->sc_sendad_success = 0; 1055 } else { 1056 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) { 1057 if (++sc->sc_sendad_success >= 1058 CARP_SENDAD_MIN_SUCCESS) { 1059 carp_suppress_preempt--; 1060 sc->sc_sendad_errors = 0; 1061 } 1062 } else 1063 sc->sc_sendad_errors = 0; 1064 } 1065 } 1066 #endif /* INET */ 1067 #ifdef INET6 1068 if (sc->sc_naddrs6) { 1069 struct ip6_hdr *ip6; 1070 1071 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1072 if (m == NULL) { 1073 sc->sc_if.if_oerrors++; 1074 CARP_STATINC(CARP_STAT_ONOMEM); 1075 /* XXX maybe less ? */ 1076 goto retry_later; 1077 } 1078 MCLAIM(m, &carp_proto6_mowner_tx); 1079 len = sizeof(*ip6) + sizeof(ch); 1080 m->m_pkthdr.len = len; 1081 m->m_pkthdr.rcvif = NULL; 1082 m->m_len = len; 1083 MH_ALIGN(m, m->m_len); 1084 m->m_flags |= M_MCAST; 1085 ip6 = mtod(m, struct ip6_hdr *); 1086 memset(ip6, 0, sizeof(*ip6)); 1087 ip6->ip6_vfc |= IPV6_VERSION; 1088 ip6->ip6_hlim = CARP_DFLTTL; 1089 ip6->ip6_nxt = IPPROTO_CARP; 1090 1091 /* set the source address */ 1092 memset(&sa, 0, sizeof(sa)); 1093 sa.sa_family = AF_INET6; 1094 ifa = ifaof_ifpforaddr(&sa, sc->sc_carpdev); 1095 if (ifa == NULL) /* This should never happen with IPv6 */ 1096 memset(&ip6->ip6_src, 0, sizeof(struct in6_addr)); 1097 else 1098 bcopy(ifatoia6(ifa)->ia_addr.sin6_addr.s6_addr, 1099 &ip6->ip6_src, sizeof(struct in6_addr)); 1100 /* set the multicast destination */ 1101 1102 ip6->ip6_dst.s6_addr16[0] = htons(0xff02); 1103 ip6->ip6_dst.s6_addr8[15] = 0x12; 1104 if (in6_setscope(&ip6->ip6_dst, sc->sc_carpdev, NULL) != 0) { 1105 sc->sc_if.if_oerrors++; 1106 m_freem(m); 1107 CARP_LOG(sc, ("in6_setscope failed")); 1108 goto retry_later; 1109 } 1110 1111 ch_ptr = (struct carp_header *)(&ip6[1]); 1112 memcpy(ch_ptr, &ch, sizeof(ch)); 1113 if (carp_prepare_ad(m, sc, ch_ptr)) 1114 goto retry_later; 1115 1116 m->m_data += sizeof(*ip6); 1117 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip6)); 1118 m->m_data -= sizeof(*ip6); 1119 1120 nanotime(&sc->sc_if.if_lastchange); 1121 sc->sc_if.if_opackets++; 1122 sc->sc_if.if_obytes += len; 1123 CARP_STATINC(CARP_STAT_OPACKETS6); 1124 1125 error = ip6_output(m, NULL, NULL, 0, &sc->sc_im6o, NULL, NULL); 1126 if (error) { 1127 if (error == ENOBUFS) 1128 CARP_STATINC(CARP_STAT_ONOMEM); 1129 else 1130 CARP_LOG(sc, ("ip6_output failed: %d", error)); 1131 sc->sc_if.if_oerrors++; 1132 if (sc->sc_sendad_errors < INT_MAX) 1133 sc->sc_sendad_errors++; 1134 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) { 1135 carp_suppress_preempt++; 1136 if (carp_suppress_preempt == 1) 1137 carp_send_ad_all(); 1138 } 1139 sc->sc_sendad_success = 0; 1140 } else { 1141 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) { 1142 if (++sc->sc_sendad_success >= 1143 CARP_SENDAD_MIN_SUCCESS) { 1144 carp_suppress_preempt--; 1145 sc->sc_sendad_errors = 0; 1146 } 1147 } else 1148 sc->sc_sendad_errors = 0; 1149 } 1150 } 1151 #endif /* INET6 */ 1152 1153 retry_later: 1154 splx(s); 1155 KERNEL_UNLOCK_ONE(NULL); 1156 if (advbase != 255 || advskew != 255) 1157 callout_schedule(&sc->sc_ad_tmo, tvtohz(&tv)); 1158 } 1159 1160 /* 1161 * Broadcast a gratuitous ARP request containing 1162 * the virtual router MAC address for each IP address 1163 * associated with the virtual router. 1164 */ 1165 void 1166 carp_send_arp(struct carp_softc *sc) 1167 { 1168 struct ifaddr *ifa; 1169 struct in_addr *in; 1170 int s; 1171 1172 KERNEL_LOCK(1, NULL); 1173 s = splsoftnet(); 1174 IFADDR_FOREACH(ifa, &sc->sc_if) { 1175 1176 if (ifa->ifa_addr->sa_family != AF_INET) 1177 continue; 1178 1179 in = &ifatoia(ifa)->ia_addr.sin_addr; 1180 arprequest(sc->sc_carpdev, in, in, CLLADDR(sc->sc_if.if_sadl)); 1181 } 1182 splx(s); 1183 KERNEL_UNLOCK_ONE(NULL); 1184 } 1185 1186 #ifdef INET6 1187 void 1188 carp_send_na(struct carp_softc *sc) 1189 { 1190 struct ifaddr *ifa; 1191 struct in6_addr *in6; 1192 static struct in6_addr mcast = IN6ADDR_LINKLOCAL_ALLNODES_INIT; 1193 int s; 1194 1195 KERNEL_LOCK(1, NULL); 1196 s = splsoftnet(); 1197 1198 IFADDR_FOREACH(ifa, &sc->sc_if) { 1199 1200 if (ifa->ifa_addr->sa_family != AF_INET6) 1201 continue; 1202 1203 in6 = &ifatoia6(ifa)->ia_addr.sin6_addr; 1204 nd6_na_output(sc->sc_carpdev, &mcast, in6, 1205 ND_NA_FLAG_OVERRIDE, 1, NULL); 1206 } 1207 splx(s); 1208 KERNEL_UNLOCK_ONE(NULL); 1209 } 1210 #endif /* INET6 */ 1211 1212 /* 1213 * Based on bridge_hash() in if_bridge.c 1214 */ 1215 #define mix(a,b,c) \ 1216 do { \ 1217 a -= b; a -= c; a ^= (c >> 13); \ 1218 b -= c; b -= a; b ^= (a << 8); \ 1219 c -= a; c -= b; c ^= (b >> 13); \ 1220 a -= b; a -= c; a ^= (c >> 12); \ 1221 b -= c; b -= a; b ^= (a << 16); \ 1222 c -= a; c -= b; c ^= (b >> 5); \ 1223 a -= b; a -= c; a ^= (c >> 3); \ 1224 b -= c; b -= a; b ^= (a << 10); \ 1225 c -= a; c -= b; c ^= (b >> 15); \ 1226 } while (0) 1227 1228 u_int32_t 1229 carp_hash(struct carp_softc *sc, u_char *src) 1230 { 1231 u_int32_t a = 0x9e3779b9, b = sc->sc_hashkey[0], c = sc->sc_hashkey[1]; 1232 1233 c += sc->sc_key[3] << 24; 1234 c += sc->sc_key[2] << 16; 1235 c += sc->sc_key[1] << 8; 1236 c += sc->sc_key[0]; 1237 b += src[5] << 8; 1238 b += src[4]; 1239 a += src[3] << 24; 1240 a += src[2] << 16; 1241 a += src[1] << 8; 1242 a += src[0]; 1243 1244 mix(a, b, c); 1245 return (c); 1246 } 1247 1248 int 1249 carp_addrcount(struct carp_if *cif, struct in_ifaddr *ia, int type) 1250 { 1251 struct carp_softc *vh; 1252 struct ifaddr *ifa; 1253 int count = 0; 1254 1255 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1256 if ((type == CARP_COUNT_RUNNING && 1257 (vh->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) == 1258 (IFF_UP|IFF_RUNNING)) || 1259 (type == CARP_COUNT_MASTER && vh->sc_state == MASTER)) { 1260 IFADDR_FOREACH(ifa, &vh->sc_if) { 1261 if (ifa->ifa_addr->sa_family == AF_INET && 1262 ia->ia_addr.sin_addr.s_addr == 1263 ifatoia(ifa)->ia_addr.sin_addr.s_addr) 1264 count++; 1265 } 1266 } 1267 } 1268 return (count); 1269 } 1270 1271 int 1272 carp_iamatch(struct in_ifaddr *ia, u_char *src, 1273 u_int32_t *count, u_int32_t index) 1274 { 1275 struct carp_softc *sc = ia->ia_ifp->if_softc; 1276 1277 if (carp_opts[CARPCTL_ARPBALANCE]) { 1278 /* 1279 * We use the source ip to decide which virtual host should 1280 * handle the request. If we're master of that virtual host, 1281 * then we respond, otherwise, just drop the arp packet on 1282 * the floor. 1283 */ 1284 1285 /* Count the elegible carp interfaces with this address */ 1286 if (*count == 0) 1287 *count = carp_addrcount( 1288 (struct carp_if *)ia->ia_ifp->if_carpdev->if_carp, 1289 ia, CARP_COUNT_RUNNING); 1290 1291 /* This should never happen, but... */ 1292 if (*count == 0) 1293 return (0); 1294 1295 if (carp_hash(sc, src) % *count == index - 1 && 1296 sc->sc_state == MASTER) { 1297 return (1); 1298 } 1299 } else { 1300 if (sc->sc_state == MASTER) 1301 return (1); 1302 } 1303 1304 return (0); 1305 } 1306 1307 #ifdef INET6 1308 struct ifaddr * 1309 carp_iamatch6(void *v, struct in6_addr *taddr) 1310 { 1311 struct carp_if *cif = v; 1312 struct carp_softc *vh; 1313 struct ifaddr *ifa; 1314 1315 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1316 IFADDR_FOREACH(ifa, &vh->sc_if) { 1317 if (IN6_ARE_ADDR_EQUAL(taddr, 1318 &ifatoia6(ifa)->ia_addr.sin6_addr) && 1319 ((vh->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) == 1320 (IFF_UP|IFF_RUNNING)) && vh->sc_state == MASTER) 1321 return (ifa); 1322 } 1323 } 1324 1325 return (NULL); 1326 } 1327 #endif /* INET6 */ 1328 1329 struct ifnet * 1330 carp_ourether(void *v, struct ether_header *eh, u_char iftype, int src) 1331 { 1332 struct carp_if *cif = (struct carp_if *)v; 1333 struct carp_softc *vh; 1334 u_int8_t *ena; 1335 1336 if (src) 1337 ena = (u_int8_t *)&eh->ether_shost; 1338 else 1339 ena = (u_int8_t *)&eh->ether_dhost; 1340 1341 switch (iftype) { 1342 case IFT_ETHER: 1343 case IFT_FDDI: 1344 if (ena[0] || ena[1] || ena[2] != 0x5e || ena[3] || ena[4] != 1) 1345 return (NULL); 1346 break; 1347 case IFT_ISO88025: 1348 if (ena[0] != 3 || ena[1] || ena[4] || ena[5]) 1349 return (NULL); 1350 break; 1351 default: 1352 return (NULL); 1353 break; 1354 } 1355 1356 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) 1357 if ((vh->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) == 1358 (IFF_UP|IFF_RUNNING) && vh->sc_state == MASTER && 1359 !memcmp(ena, CLLADDR(vh->sc_if.if_sadl), 1360 ETHER_ADDR_LEN)) { 1361 return (&vh->sc_if); 1362 } 1363 1364 return (NULL); 1365 } 1366 1367 int 1368 carp_input(struct mbuf *m, u_int8_t *shost, u_int8_t *dhost, u_int16_t etype) 1369 { 1370 struct ether_header eh; 1371 struct carp_if *cif = (struct carp_if *)m->m_pkthdr.rcvif->if_carp; 1372 struct ifnet *ifp; 1373 1374 memcpy(&eh.ether_shost, shost, sizeof(eh.ether_shost)); 1375 memcpy(&eh.ether_dhost, dhost, sizeof(eh.ether_dhost)); 1376 eh.ether_type = etype; 1377 1378 if (m->m_flags & (M_BCAST|M_MCAST)) { 1379 struct carp_softc *vh; 1380 struct mbuf *m0; 1381 1382 /* 1383 * XXX Should really check the list of multicast addresses 1384 * for each CARP interface _before_ copying. 1385 */ 1386 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1387 m0 = m_copym(m, 0, M_COPYALL, M_DONTWAIT); 1388 if (m0 == NULL) 1389 continue; 1390 m0->m_pkthdr.rcvif = &vh->sc_if; 1391 ether_input(&vh->sc_if, m0); 1392 } 1393 return (1); 1394 } 1395 1396 ifp = carp_ourether(cif, &eh, m->m_pkthdr.rcvif->if_type, 0); 1397 if (ifp == NULL) { 1398 return (1); 1399 } 1400 1401 m->m_pkthdr.rcvif = ifp; 1402 1403 bpf_mtap(ifp, m); 1404 ifp->if_ipackets++; 1405 ether_input(ifp, m); 1406 return (0); 1407 } 1408 1409 void 1410 carp_master_down(void *v) 1411 { 1412 struct carp_softc *sc = v; 1413 1414 switch (sc->sc_state) { 1415 case INIT: 1416 printf("%s: master_down event in INIT state\n", 1417 sc->sc_if.if_xname); 1418 break; 1419 case MASTER: 1420 break; 1421 case BACKUP: 1422 CARP_LOG(sc, ("INIT -> MASTER (preempting)")); 1423 carp_set_state(sc, MASTER); 1424 carp_send_ad(sc); 1425 carp_send_arp(sc); 1426 #ifdef INET6 1427 carp_send_na(sc); 1428 #endif /* INET6 */ 1429 carp_setrun(sc, 0); 1430 carp_setroute(sc, RTM_ADD); 1431 break; 1432 } 1433 } 1434 1435 /* 1436 * When in backup state, af indicates whether to reset the master down timer 1437 * for v4 or v6. If it's set to zero, reset the ones which are already pending. 1438 */ 1439 void 1440 carp_setrun(struct carp_softc *sc, sa_family_t af) 1441 { 1442 struct timeval tv; 1443 1444 if (sc->sc_carpdev == NULL) { 1445 sc->sc_if.if_flags &= ~IFF_RUNNING; 1446 carp_set_state(sc, INIT); 1447 return; 1448 } 1449 1450 if (sc->sc_if.if_flags & IFF_UP && sc->sc_vhid > 0 && 1451 (sc->sc_naddrs || sc->sc_naddrs6) && !sc->sc_suppress) { 1452 sc->sc_if.if_flags |= IFF_RUNNING; 1453 } else { 1454 sc->sc_if.if_flags &= ~IFF_RUNNING; 1455 carp_setroute(sc, RTM_DELETE); 1456 return; 1457 } 1458 1459 switch (sc->sc_state) { 1460 case INIT: 1461 carp_set_state(sc, BACKUP); 1462 carp_setroute(sc, RTM_DELETE); 1463 carp_setrun(sc, 0); 1464 break; 1465 case BACKUP: 1466 callout_stop(&sc->sc_ad_tmo); 1467 tv.tv_sec = 3 * sc->sc_advbase; 1468 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1469 switch (af) { 1470 #ifdef INET 1471 case AF_INET: 1472 callout_schedule(&sc->sc_md_tmo, tvtohz(&tv)); 1473 break; 1474 #endif /* INET */ 1475 #ifdef INET6 1476 case AF_INET6: 1477 callout_schedule(&sc->sc_md6_tmo, tvtohz(&tv)); 1478 break; 1479 #endif /* INET6 */ 1480 default: 1481 if (sc->sc_naddrs) 1482 callout_schedule(&sc->sc_md_tmo, tvtohz(&tv)); 1483 if (sc->sc_naddrs6) 1484 callout_schedule(&sc->sc_md6_tmo, tvtohz(&tv)); 1485 break; 1486 } 1487 break; 1488 case MASTER: 1489 tv.tv_sec = sc->sc_advbase; 1490 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1491 callout_schedule(&sc->sc_ad_tmo, tvtohz(&tv)); 1492 break; 1493 } 1494 } 1495 1496 void 1497 carp_multicast_cleanup(struct carp_softc *sc) 1498 { 1499 struct ip_moptions *imo = &sc->sc_imo; 1500 #ifdef INET6 1501 struct ip6_moptions *im6o = &sc->sc_im6o; 1502 #endif 1503 u_int16_t n = imo->imo_num_memberships; 1504 1505 /* Clean up our own multicast memberships */ 1506 while (n-- > 0) { 1507 if (imo->imo_membership[n] != NULL) { 1508 in_delmulti(imo->imo_membership[n]); 1509 imo->imo_membership[n] = NULL; 1510 } 1511 } 1512 imo->imo_num_memberships = 0; 1513 imo->imo_multicast_ifp = NULL; 1514 1515 #ifdef INET6 1516 while (!LIST_EMPTY(&im6o->im6o_memberships)) { 1517 struct in6_multi_mship *imm = 1518 LIST_FIRST(&im6o->im6o_memberships); 1519 1520 LIST_REMOVE(imm, i6mm_chain); 1521 in6_leavegroup(imm); 1522 } 1523 im6o->im6o_multicast_ifp = NULL; 1524 #endif 1525 1526 /* And any other multicast memberships */ 1527 carp_ether_purgemulti(sc); 1528 } 1529 1530 int 1531 carp_set_ifp(struct carp_softc *sc, struct ifnet *ifp) 1532 { 1533 struct carp_if *cif, *ncif = NULL; 1534 struct carp_softc *vr, *after = NULL; 1535 int myself = 0, error = 0; 1536 int s; 1537 1538 if (ifp == sc->sc_carpdev) 1539 return (0); 1540 1541 if (ifp != NULL) { 1542 if ((ifp->if_flags & IFF_MULTICAST) == 0) 1543 return (EADDRNOTAVAIL); 1544 1545 if (ifp->if_type == IFT_CARP) 1546 return (EINVAL); 1547 1548 if (ifp->if_carp == NULL) { 1549 ncif = malloc(sizeof(*cif), M_IFADDR, M_NOWAIT); 1550 if (ncif == NULL) 1551 return (ENOBUFS); 1552 if ((error = ifpromisc(ifp, 1))) { 1553 free(ncif, M_IFADDR); 1554 return (error); 1555 } 1556 1557 ncif->vhif_ifp = ifp; 1558 TAILQ_INIT(&ncif->vhif_vrs); 1559 } else { 1560 cif = (struct carp_if *)ifp->if_carp; 1561 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) 1562 if (vr != sc && vr->sc_vhid == sc->sc_vhid) 1563 return (EINVAL); 1564 } 1565 1566 /* detach from old interface */ 1567 if (sc->sc_carpdev != NULL) 1568 carpdetach(sc); 1569 1570 /* join multicast groups */ 1571 if (sc->sc_naddrs < 0 && 1572 (error = carp_join_multicast(sc)) != 0) { 1573 if (ncif != NULL) 1574 free(ncif, M_IFADDR); 1575 return (error); 1576 } 1577 1578 #ifdef INET6 1579 if (sc->sc_naddrs6 < 0 && 1580 (error = carp_join_multicast6(sc)) != 0) { 1581 if (ncif != NULL) 1582 free(ncif, M_IFADDR); 1583 carp_multicast_cleanup(sc); 1584 return (error); 1585 } 1586 #endif 1587 1588 /* attach carp interface to physical interface */ 1589 if (ncif != NULL) 1590 ifp->if_carp = (void *)ncif; 1591 sc->sc_carpdev = ifp; 1592 sc->sc_if.if_capabilities = ifp->if_capabilities & 1593 (IFCAP_TSOv4 | IFCAP_TSOv6 | 1594 IFCAP_CSUM_IPv4_Tx|IFCAP_CSUM_IPv4_Rx| 1595 IFCAP_CSUM_TCPv4_Tx|IFCAP_CSUM_TCPv4_Rx| 1596 IFCAP_CSUM_UDPv4_Tx|IFCAP_CSUM_UDPv4_Rx| 1597 IFCAP_CSUM_TCPv6_Tx|IFCAP_CSUM_TCPv6_Rx| 1598 IFCAP_CSUM_UDPv6_Tx|IFCAP_CSUM_UDPv6_Rx); 1599 1600 cif = (struct carp_if *)ifp->if_carp; 1601 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) { 1602 if (vr == sc) 1603 myself = 1; 1604 if (vr->sc_vhid < sc->sc_vhid) 1605 after = vr; 1606 } 1607 1608 if (!myself) { 1609 /* We're trying to keep things in order */ 1610 if (after == NULL) { 1611 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list); 1612 } else { 1613 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, 1614 sc, sc_list); 1615 } 1616 cif->vhif_nvrs++; 1617 } 1618 if (sc->sc_naddrs || sc->sc_naddrs6) 1619 sc->sc_if.if_flags |= IFF_UP; 1620 carp_set_enaddr(sc); 1621 KERNEL_LOCK(1, NULL); 1622 s = splnet(); 1623 /* XXX linkstatehooks establish */ 1624 carp_carpdev_state(ifp); 1625 splx(s); 1626 KERNEL_UNLOCK_ONE(NULL); 1627 } else { 1628 carpdetach(sc); 1629 sc->sc_if.if_flags &= ~(IFF_UP|IFF_RUNNING); 1630 } 1631 return (0); 1632 } 1633 1634 void 1635 carp_set_enaddr(struct carp_softc *sc) 1636 { 1637 uint8_t enaddr[ETHER_ADDR_LEN]; 1638 if (sc->sc_carpdev && sc->sc_carpdev->if_type == IFT_ISO88025) { 1639 enaddr[0] = 3; 1640 enaddr[1] = 0; 1641 enaddr[2] = 0x40 >> (sc->sc_vhid - 1); 1642 enaddr[3] = 0x40000 >> (sc->sc_vhid - 1); 1643 enaddr[4] = 0; 1644 enaddr[5] = 0; 1645 } else { 1646 enaddr[0] = 0; 1647 enaddr[1] = 0; 1648 enaddr[2] = 0x5e; 1649 enaddr[3] = 0; 1650 enaddr[4] = 1; 1651 enaddr[5] = sc->sc_vhid; 1652 } 1653 if_set_sadl(&sc->sc_if, enaddr, sizeof(enaddr), false); 1654 } 1655 1656 void 1657 carp_addr_updated(void *v) 1658 { 1659 struct carp_softc *sc = (struct carp_softc *) v; 1660 struct ifaddr *ifa; 1661 int new_naddrs = 0, new_naddrs6 = 0; 1662 1663 IFADDR_FOREACH(ifa, &sc->sc_if) { 1664 if (ifa->ifa_addr->sa_family == AF_INET) 1665 new_naddrs++; 1666 else if (ifa->ifa_addr->sa_family == AF_INET6) 1667 new_naddrs6++; 1668 } 1669 1670 /* Handle a callback after SIOCDIFADDR */ 1671 if (new_naddrs < sc->sc_naddrs || new_naddrs6 < sc->sc_naddrs6) { 1672 struct in_addr mc_addr; 1673 1674 sc->sc_naddrs = new_naddrs; 1675 sc->sc_naddrs6 = new_naddrs6; 1676 1677 /* Re-establish multicast membership removed by in_control */ 1678 mc_addr.s_addr = INADDR_CARP_GROUP; 1679 if (!in_multi_group(mc_addr, &sc->sc_if, 0)) { 1680 memset(&sc->sc_imo, 0, sizeof(sc->sc_imo)); 1681 1682 if (sc->sc_carpdev != NULL && sc->sc_naddrs > 0) 1683 carp_join_multicast(sc); 1684 } 1685 1686 if (sc->sc_naddrs == 0 && sc->sc_naddrs6 == 0) { 1687 sc->sc_if.if_flags &= ~IFF_UP; 1688 carp_set_state(sc, INIT); 1689 } else 1690 carp_hmac_prepare(sc); 1691 } 1692 1693 carp_setrun(sc, 0); 1694 } 1695 1696 int 1697 carp_set_addr(struct carp_softc *sc, struct sockaddr_in *sin) 1698 { 1699 struct ifnet *ifp = sc->sc_carpdev; 1700 struct in_ifaddr *ia, *ia_if; 1701 int error = 0; 1702 1703 if (sin->sin_addr.s_addr == 0) { 1704 if (!(sc->sc_if.if_flags & IFF_UP)) 1705 carp_set_state(sc, INIT); 1706 if (sc->sc_naddrs) 1707 sc->sc_if.if_flags |= IFF_UP; 1708 carp_setrun(sc, 0); 1709 return (0); 1710 } 1711 1712 /* we have to do this by hand to ensure we don't match on ourselves */ 1713 ia_if = NULL; 1714 for (ia = TAILQ_FIRST(&in_ifaddrhead); ia; 1715 ia = TAILQ_NEXT(ia, ia_list)) { 1716 1717 /* and, yeah, we need a multicast-capable iface too */ 1718 if (ia->ia_ifp != &sc->sc_if && 1719 ia->ia_ifp->if_type != IFT_CARP && 1720 (ia->ia_ifp->if_flags & IFF_MULTICAST) && 1721 (sin->sin_addr.s_addr & ia->ia_subnetmask) == 1722 ia->ia_subnet) { 1723 if (!ia_if) 1724 ia_if = ia; 1725 } 1726 } 1727 1728 if (ia_if) { 1729 ia = ia_if; 1730 if (ifp) { 1731 if (ifp != ia->ia_ifp) 1732 return (EADDRNOTAVAIL); 1733 } else { 1734 ifp = ia->ia_ifp; 1735 } 1736 } 1737 1738 if ((error = carp_set_ifp(sc, ifp))) 1739 return (error); 1740 1741 if (sc->sc_carpdev == NULL) 1742 return (EADDRNOTAVAIL); 1743 1744 if (sc->sc_naddrs == 0 && (error = carp_join_multicast(sc)) != 0) 1745 return (error); 1746 1747 sc->sc_naddrs++; 1748 if (sc->sc_carpdev != NULL) 1749 sc->sc_if.if_flags |= IFF_UP; 1750 1751 carp_set_state(sc, INIT); 1752 carp_setrun(sc, 0); 1753 1754 /* 1755 * Hook if_addrhooks so that we get a callback after in_ifinit has run, 1756 * to correct any inappropriate routes that it inserted. 1757 */ 1758 if (sc->ah_cookie == 0) { 1759 /* XXX link address hook */ 1760 } 1761 1762 return (0); 1763 } 1764 1765 int 1766 carp_join_multicast(struct carp_softc *sc) 1767 { 1768 struct ip_moptions *imo = &sc->sc_imo, tmpimo; 1769 struct in_addr addr; 1770 1771 memset(&tmpimo, 0, sizeof(tmpimo)); 1772 addr.s_addr = INADDR_CARP_GROUP; 1773 if ((tmpimo.imo_membership[0] = 1774 in_addmulti(&addr, &sc->sc_if)) == NULL) { 1775 return (ENOBUFS); 1776 } 1777 1778 imo->imo_membership[0] = tmpimo.imo_membership[0]; 1779 imo->imo_num_memberships = 1; 1780 imo->imo_multicast_ifp = &sc->sc_if; 1781 imo->imo_multicast_ttl = CARP_DFLTTL; 1782 imo->imo_multicast_loop = 0; 1783 return (0); 1784 } 1785 1786 1787 #ifdef INET6 1788 int 1789 carp_set_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6) 1790 { 1791 struct ifnet *ifp = sc->sc_carpdev; 1792 struct in6_ifaddr *ia, *ia_if; 1793 int error = 0; 1794 1795 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1796 if (!(sc->sc_if.if_flags & IFF_UP)) 1797 carp_set_state(sc, INIT); 1798 if (sc->sc_naddrs6) 1799 sc->sc_if.if_flags |= IFF_UP; 1800 carp_setrun(sc, 0); 1801 return (0); 1802 } 1803 1804 /* we have to do this by hand to ensure we don't match on ourselves */ 1805 ia_if = NULL; 1806 for (ia = in6_ifaddr; ia; ia = ia->ia_next) { 1807 int i; 1808 1809 for (i = 0; i < 4; i++) { 1810 if ((sin6->sin6_addr.s6_addr32[i] & 1811 ia->ia_prefixmask.sin6_addr.s6_addr32[i]) != 1812 (ia->ia_addr.sin6_addr.s6_addr32[i] & 1813 ia->ia_prefixmask.sin6_addr.s6_addr32[i])) 1814 break; 1815 } 1816 /* and, yeah, we need a multicast-capable iface too */ 1817 if (ia->ia_ifp != &sc->sc_if && 1818 ia->ia_ifp->if_type != IFT_CARP && 1819 (ia->ia_ifp->if_flags & IFF_MULTICAST) && 1820 (i == 4)) { 1821 if (!ia_if) 1822 ia_if = ia; 1823 } 1824 } 1825 1826 if (ia_if) { 1827 ia = ia_if; 1828 if (sc->sc_carpdev) { 1829 if (sc->sc_carpdev != ia->ia_ifp) 1830 return (EADDRNOTAVAIL); 1831 } else { 1832 ifp = ia->ia_ifp; 1833 } 1834 } 1835 1836 if ((error = carp_set_ifp(sc, ifp))) 1837 return (error); 1838 1839 if (sc->sc_carpdev == NULL) 1840 return (EADDRNOTAVAIL); 1841 1842 if (sc->sc_naddrs6 == 0 && (error = carp_join_multicast6(sc)) != 0) 1843 return (error); 1844 1845 sc->sc_naddrs6++; 1846 if (sc->sc_carpdev != NULL) 1847 sc->sc_if.if_flags |= IFF_UP; 1848 carp_set_state(sc, INIT); 1849 carp_setrun(sc, 0); 1850 1851 return (0); 1852 } 1853 1854 int 1855 carp_join_multicast6(struct carp_softc *sc) 1856 { 1857 struct in6_multi_mship *imm, *imm2; 1858 struct ip6_moptions *im6o = &sc->sc_im6o; 1859 struct sockaddr_in6 addr6; 1860 int error; 1861 1862 /* Join IPv6 CARP multicast group */ 1863 memset(&addr6, 0, sizeof(addr6)); 1864 addr6.sin6_family = AF_INET6; 1865 addr6.sin6_len = sizeof(addr6); 1866 addr6.sin6_addr.s6_addr16[0] = htons(0xff02); 1867 addr6.sin6_addr.s6_addr16[1] = htons(sc->sc_if.if_index); 1868 addr6.sin6_addr.s6_addr8[15] = 0x12; 1869 if ((imm = in6_joingroup(&sc->sc_if, 1870 &addr6.sin6_addr, &error, 0)) == NULL) { 1871 return (error); 1872 } 1873 /* join solicited multicast address */ 1874 memset(&addr6.sin6_addr, 0, sizeof(addr6.sin6_addr)); 1875 addr6.sin6_addr.s6_addr16[0] = htons(0xff02); 1876 addr6.sin6_addr.s6_addr16[1] = htons(sc->sc_if.if_index); 1877 addr6.sin6_addr.s6_addr32[1] = 0; 1878 addr6.sin6_addr.s6_addr32[2] = htonl(1); 1879 addr6.sin6_addr.s6_addr32[3] = 0; 1880 addr6.sin6_addr.s6_addr8[12] = 0xff; 1881 if ((imm2 = in6_joingroup(&sc->sc_if, 1882 &addr6.sin6_addr, &error, 0)) == NULL) { 1883 in6_leavegroup(imm); 1884 return (error); 1885 } 1886 1887 /* apply v6 multicast membership */ 1888 im6o->im6o_multicast_ifp = &sc->sc_if; 1889 if (imm) 1890 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, 1891 i6mm_chain); 1892 if (imm2) 1893 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm2, 1894 i6mm_chain); 1895 1896 return (0); 1897 } 1898 1899 #endif /* INET6 */ 1900 1901 int 1902 carp_ioctl(struct ifnet *ifp, u_long cmd, void *data) 1903 { 1904 struct lwp *l = curlwp; /* XXX */ 1905 struct carp_softc *sc = ifp->if_softc, *vr; 1906 struct carpreq carpr; 1907 struct ifaddr *ifa; 1908 struct ifreq *ifr; 1909 struct ifnet *cdev = NULL; 1910 int error = 0; 1911 1912 ifa = (struct ifaddr *)data; 1913 ifr = (struct ifreq *)data; 1914 1915 switch (cmd) { 1916 case SIOCINITIFADDR: 1917 switch (ifa->ifa_addr->sa_family) { 1918 #ifdef INET 1919 case AF_INET: 1920 sc->sc_if.if_flags |= IFF_UP; 1921 memcpy(ifa->ifa_dstaddr, ifa->ifa_addr, 1922 sizeof(struct sockaddr)); 1923 error = carp_set_addr(sc, satosin(ifa->ifa_addr)); 1924 break; 1925 #endif /* INET */ 1926 #ifdef INET6 1927 case AF_INET6: 1928 sc->sc_if.if_flags|= IFF_UP; 1929 error = carp_set_addr6(sc, satosin6(ifa->ifa_addr)); 1930 break; 1931 #endif /* INET6 */ 1932 default: 1933 error = EAFNOSUPPORT; 1934 break; 1935 } 1936 break; 1937 1938 case SIOCSIFFLAGS: 1939 if ((error = ifioctl_common(ifp, cmd, data)) != 0) 1940 break; 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 /* we need the interface up to bow out */ 1947 sc->sc_if.if_flags |= IFF_UP; 1948 sc->sc_bow_out = 1; 1949 carp_send_ad(sc); 1950 } 1951 sc->sc_if.if_flags &= ~IFF_UP; 1952 carp_set_state(sc, INIT); 1953 carp_setrun(sc, 0); 1954 } else if (sc->sc_state == INIT && (ifr->ifr_flags & IFF_UP)) { 1955 sc->sc_if.if_flags |= IFF_UP; 1956 carp_setrun(sc, 0); 1957 } 1958 break; 1959 1960 case SIOCSVH: 1961 if (l == NULL) 1962 break; 1963 if ((error = kauth_authorize_network(l->l_cred, 1964 KAUTH_NETWORK_INTERFACE, 1965 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 1966 NULL)) != 0) 1967 break; 1968 if ((error = copyin(ifr->ifr_data, &carpr, sizeof carpr))) 1969 break; 1970 error = 1; 1971 if (carpr.carpr_carpdev[0] != '\0' && 1972 (cdev = ifunit(carpr.carpr_carpdev)) == NULL) 1973 return (EINVAL); 1974 if ((error = carp_set_ifp(sc, cdev))) 1975 return (error); 1976 if (sc->sc_state != INIT && carpr.carpr_state != sc->sc_state) { 1977 switch (carpr.carpr_state) { 1978 case BACKUP: 1979 callout_stop(&sc->sc_ad_tmo); 1980 carp_set_state(sc, BACKUP); 1981 carp_setrun(sc, 0); 1982 carp_setroute(sc, RTM_DELETE); 1983 break; 1984 case MASTER: 1985 carp_master_down(sc); 1986 break; 1987 default: 1988 break; 1989 } 1990 } 1991 if (carpr.carpr_vhid > 0) { 1992 if (carpr.carpr_vhid > 255) { 1993 error = EINVAL; 1994 break; 1995 } 1996 if (sc->sc_carpdev) { 1997 struct carp_if *cif; 1998 cif = (struct carp_if *)sc->sc_carpdev->if_carp; 1999 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) 2000 if (vr != sc && 2001 vr->sc_vhid == carpr.carpr_vhid) 2002 return (EINVAL); 2003 } 2004 sc->sc_vhid = carpr.carpr_vhid; 2005 carp_set_enaddr(sc); 2006 carp_set_state(sc, INIT); 2007 error--; 2008 } 2009 if (carpr.carpr_advbase > 0 || carpr.carpr_advskew > 0) { 2010 if (carpr.carpr_advskew > 254) { 2011 error = EINVAL; 2012 break; 2013 } 2014 if (carpr.carpr_advbase > 255) { 2015 error = EINVAL; 2016 break; 2017 } 2018 sc->sc_advbase = carpr.carpr_advbase; 2019 sc->sc_advskew = carpr.carpr_advskew; 2020 error--; 2021 } 2022 memcpy(sc->sc_key, carpr.carpr_key, sizeof(sc->sc_key)); 2023 if (error > 0) 2024 error = EINVAL; 2025 else { 2026 error = 0; 2027 carp_setrun(sc, 0); 2028 } 2029 break; 2030 2031 case SIOCGVH: 2032 memset(&carpr, 0, sizeof(carpr)); 2033 if (sc->sc_carpdev != NULL) 2034 strlcpy(carpr.carpr_carpdev, sc->sc_carpdev->if_xname, 2035 IFNAMSIZ); 2036 carpr.carpr_state = sc->sc_state; 2037 carpr.carpr_vhid = sc->sc_vhid; 2038 carpr.carpr_advbase = sc->sc_advbase; 2039 carpr.carpr_advskew = sc->sc_advskew; 2040 2041 if ((l != NULL) && (error = kauth_authorize_network(l->l_cred, 2042 KAUTH_NETWORK_INTERFACE, 2043 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 2044 NULL)) == 0) 2045 memcpy(carpr.carpr_key, sc->sc_key, 2046 sizeof(carpr.carpr_key)); 2047 error = copyout(&carpr, ifr->ifr_data, sizeof(carpr)); 2048 break; 2049 2050 case SIOCADDMULTI: 2051 error = carp_ether_addmulti(sc, ifr); 2052 break; 2053 2054 case SIOCDELMULTI: 2055 error = carp_ether_delmulti(sc, ifr); 2056 break; 2057 2058 case SIOCSIFCAP: 2059 if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET) 2060 error = 0; 2061 break; 2062 2063 default: 2064 error = ether_ioctl(ifp, cmd, data); 2065 } 2066 2067 carp_hmac_prepare(sc); 2068 return (error); 2069 } 2070 2071 2072 /* 2073 * Start output on carp interface. This function should never be called. 2074 */ 2075 void 2076 carp_start(struct ifnet *ifp) 2077 { 2078 #ifdef DEBUG 2079 printf("%s: start called\n", ifp->if_xname); 2080 #endif 2081 } 2082 2083 int 2084 carp_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa, 2085 struct rtentry *rt) 2086 { 2087 struct carp_softc *sc = ((struct carp_softc *)ifp->if_softc); 2088 KASSERT(KERNEL_LOCKED_P()); 2089 2090 if (sc->sc_carpdev != NULL && sc->sc_state == MASTER) { 2091 return (sc->sc_carpdev->if_output(ifp, m, sa, rt)); 2092 } else { 2093 m_freem(m); 2094 return (ENETUNREACH); 2095 } 2096 } 2097 2098 void 2099 carp_set_state(struct carp_softc *sc, int state) 2100 { 2101 static const char *carp_states[] = { CARP_STATES }; 2102 if (sc->sc_state == state) 2103 return; 2104 2105 CARP_LOG(sc, ("state transition from: %s -> to: %s", carp_states[sc->sc_state], carp_states[state])); 2106 2107 sc->sc_state = state; 2108 switch (state) { 2109 case BACKUP: 2110 sc->sc_if.if_link_state = LINK_STATE_DOWN; 2111 break; 2112 case MASTER: 2113 sc->sc_if.if_link_state = LINK_STATE_UP; 2114 break; 2115 default: 2116 sc->sc_if.if_link_state = LINK_STATE_UNKNOWN; 2117 break; 2118 } 2119 rt_ifmsg(&sc->sc_if); 2120 } 2121 2122 void 2123 carp_carpdev_state(void *v) 2124 { 2125 struct carp_if *cif; 2126 struct carp_softc *sc; 2127 struct ifnet *ifp = v; 2128 2129 if (ifp->if_type == IFT_CARP) 2130 return; 2131 2132 cif = (struct carp_if *)ifp->if_carp; 2133 2134 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list) { 2135 int suppressed = sc->sc_suppress; 2136 2137 if (sc->sc_carpdev->if_link_state == LINK_STATE_DOWN || 2138 !(sc->sc_carpdev->if_flags & IFF_UP)) { 2139 sc->sc_if.if_flags &= ~IFF_RUNNING; 2140 callout_stop(&sc->sc_ad_tmo); 2141 callout_stop(&sc->sc_md_tmo); 2142 callout_stop(&sc->sc_md6_tmo); 2143 carp_set_state(sc, INIT); 2144 sc->sc_suppress = 1; 2145 carp_setrun(sc, 0); 2146 if (!suppressed) { 2147 carp_suppress_preempt++; 2148 if (carp_suppress_preempt == 1) 2149 carp_send_ad_all(); 2150 } 2151 } else { 2152 carp_set_state(sc, INIT); 2153 sc->sc_suppress = 0; 2154 carp_setrun(sc, 0); 2155 if (suppressed) 2156 carp_suppress_preempt--; 2157 } 2158 } 2159 } 2160 2161 int 2162 carp_ether_addmulti(struct carp_softc *sc, struct ifreq *ifr) 2163 { 2164 const struct sockaddr *sa = ifreq_getaddr(SIOCADDMULTI, ifr); 2165 struct ifnet *ifp; 2166 struct carp_mc_entry *mc; 2167 u_int8_t addrlo[ETHER_ADDR_LEN], addrhi[ETHER_ADDR_LEN]; 2168 int error; 2169 2170 ifp = sc->sc_carpdev; 2171 if (ifp == NULL) 2172 return (EINVAL); 2173 2174 error = ether_addmulti(sa, &sc->sc_ac); 2175 if (error != ENETRESET) 2176 return (error); 2177 2178 /* 2179 * This is new multicast address. We have to tell parent 2180 * about it. Also, remember this multicast address so that 2181 * we can delete them on unconfigure. 2182 */ 2183 mc = malloc(sizeof(struct carp_mc_entry), M_DEVBUF, M_NOWAIT); 2184 if (mc == NULL) { 2185 error = ENOMEM; 2186 goto alloc_failed; 2187 } 2188 2189 /* 2190 * As ether_addmulti() returns ENETRESET, following two 2191 * statement shouldn't fail. 2192 */ 2193 (void)ether_multiaddr(sa, addrlo, addrhi); 2194 ETHER_LOOKUP_MULTI(addrlo, addrhi, &sc->sc_ac, mc->mc_enm); 2195 memcpy(&mc->mc_addr, sa, sa->sa_len); 2196 LIST_INSERT_HEAD(&sc->carp_mc_listhead, mc, mc_entries); 2197 2198 error = if_mcast_op(ifp, SIOCADDMULTI, sa); 2199 if (error != 0) 2200 goto ioctl_failed; 2201 2202 return (error); 2203 2204 ioctl_failed: 2205 LIST_REMOVE(mc, mc_entries); 2206 free(mc, M_DEVBUF); 2207 alloc_failed: 2208 (void)ether_delmulti(sa, &sc->sc_ac); 2209 2210 return (error); 2211 } 2212 2213 int 2214 carp_ether_delmulti(struct carp_softc *sc, struct ifreq *ifr) 2215 { 2216 const struct sockaddr *sa = ifreq_getaddr(SIOCDELMULTI, ifr); 2217 struct ifnet *ifp; 2218 struct ether_multi *enm; 2219 struct carp_mc_entry *mc; 2220 u_int8_t addrlo[ETHER_ADDR_LEN], addrhi[ETHER_ADDR_LEN]; 2221 int error; 2222 2223 ifp = sc->sc_carpdev; 2224 if (ifp == NULL) 2225 return (EINVAL); 2226 2227 /* 2228 * Find a key to lookup carp_mc_entry. We have to do this 2229 * before calling ether_delmulti for obvious reason. 2230 */ 2231 if ((error = ether_multiaddr(sa, addrlo, addrhi)) != 0) 2232 return (error); 2233 ETHER_LOOKUP_MULTI(addrlo, addrhi, &sc->sc_ac, enm); 2234 if (enm == NULL) 2235 return (EINVAL); 2236 2237 LIST_FOREACH(mc, &sc->carp_mc_listhead, mc_entries) 2238 if (mc->mc_enm == enm) 2239 break; 2240 2241 /* We won't delete entries we didn't add */ 2242 if (mc == NULL) 2243 return (EINVAL); 2244 2245 error = ether_delmulti(sa, &sc->sc_ac); 2246 if (error != ENETRESET) 2247 return (error); 2248 2249 /* We no longer use this multicast address. Tell parent so. */ 2250 error = if_mcast_op(ifp, SIOCDELMULTI, sa); 2251 if (error == 0) { 2252 /* And forget about this address. */ 2253 LIST_REMOVE(mc, mc_entries); 2254 free(mc, M_DEVBUF); 2255 } else 2256 (void)ether_addmulti(sa, &sc->sc_ac); 2257 return (error); 2258 } 2259 2260 /* 2261 * Delete any multicast address we have asked to add from parent 2262 * interface. Called when the carp is being unconfigured. 2263 */ 2264 void 2265 carp_ether_purgemulti(struct carp_softc *sc) 2266 { 2267 struct ifnet *ifp = sc->sc_carpdev; /* Parent. */ 2268 struct carp_mc_entry *mc; 2269 2270 if (ifp == NULL) 2271 return; 2272 2273 while ((mc = LIST_FIRST(&sc->carp_mc_listhead)) != NULL) { 2274 (void)if_mcast_op(ifp, SIOCDELMULTI, sstosa(&mc->mc_addr)); 2275 LIST_REMOVE(mc, mc_entries); 2276 free(mc, M_DEVBUF); 2277 } 2278 } 2279 2280 static int 2281 sysctl_net_inet_carp_stats(SYSCTLFN_ARGS) 2282 { 2283 2284 return (NETSTAT_SYSCTL(carpstat_percpu, CARP_NSTATS)); 2285 } 2286 2287 void 2288 carp_init(void) 2289 { 2290 2291 sysctl_net_inet_carp_setup(NULL); 2292 #ifdef MBUFTRACE 2293 MOWNER_ATTACH(&carp_proto_mowner_rx); 2294 MOWNER_ATTACH(&carp_proto_mowner_tx); 2295 MOWNER_ATTACH(&carp_proto6_mowner_rx); 2296 MOWNER_ATTACH(&carp_proto6_mowner_tx); 2297 #endif 2298 } 2299 2300 static void 2301 sysctl_net_inet_carp_setup(struct sysctllog **clog) 2302 { 2303 2304 sysctl_createv(clog, 0, NULL, NULL, 2305 CTLFLAG_PERMANENT, 2306 CTLTYPE_NODE, "inet", NULL, 2307 NULL, 0, NULL, 0, 2308 CTL_NET, PF_INET, CTL_EOL); 2309 sysctl_createv(clog, 0, NULL, NULL, 2310 CTLFLAG_PERMANENT, 2311 CTLTYPE_NODE, "carp", 2312 SYSCTL_DESCR("CARP related settings"), 2313 NULL, 0, NULL, 0, 2314 CTL_NET, PF_INET, IPPROTO_CARP, CTL_EOL); 2315 2316 sysctl_createv(clog, 0, NULL, NULL, 2317 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2318 CTLTYPE_INT, "preempt", 2319 SYSCTL_DESCR("Enable CARP Preempt"), 2320 NULL, 0, &carp_opts[CARPCTL_PREEMPT], 0, 2321 CTL_NET, PF_INET, IPPROTO_CARP, 2322 CTL_CREATE, CTL_EOL); 2323 sysctl_createv(clog, 0, NULL, NULL, 2324 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2325 CTLTYPE_INT, "arpbalance", 2326 SYSCTL_DESCR("Enable ARP balancing"), 2327 NULL, 0, &carp_opts[CARPCTL_ARPBALANCE], 0, 2328 CTL_NET, PF_INET, IPPROTO_CARP, 2329 CTL_CREATE, CTL_EOL); 2330 sysctl_createv(clog, 0, NULL, NULL, 2331 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2332 CTLTYPE_INT, "allow", 2333 SYSCTL_DESCR("Enable CARP"), 2334 NULL, 0, &carp_opts[CARPCTL_ALLOW], 0, 2335 CTL_NET, PF_INET, IPPROTO_CARP, 2336 CTL_CREATE, CTL_EOL); 2337 sysctl_createv(clog, 0, NULL, NULL, 2338 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2339 CTLTYPE_INT, "log", 2340 SYSCTL_DESCR("CARP logging"), 2341 NULL, 0, &carp_opts[CARPCTL_LOG], 0, 2342 CTL_NET, PF_INET, IPPROTO_CARP, 2343 CTL_CREATE, CTL_EOL); 2344 sysctl_createv(clog, 0, NULL, NULL, 2345 CTLFLAG_PERMANENT, 2346 CTLTYPE_STRUCT, "stats", 2347 SYSCTL_DESCR("CARP statistics"), 2348 sysctl_net_inet_carp_stats, 0, NULL, 0, 2349 CTL_NET, PF_INET, IPPROTO_CARP, CARPCTL_STATS, 2350 CTL_EOL); 2351 } 2352