1 /* $NetBSD: ip_carp.c,v 1.53 2014/04/04 12:53:04 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 #include "opt_mbuftrace.h" 32 33 #include <sys/cdefs.h> 34 __KERNEL_RCSID(0, "$NetBSD: ip_carp.c,v 1.53 2014/04/04 12:53:04 bouyer 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 s = splsoftnet(); 359 IFADDR_FOREACH(ifa, &sc->sc_if) { 360 switch (ifa->ifa_addr->sa_family) { 361 case AF_INET: { 362 int count = 0; 363 struct rtentry *rt; 364 int hr_otherif, nr_ourif; 365 366 /* 367 * Avoid screwing with the routes if there are other 368 * carp interfaces which are master and have the same 369 * address. 370 */ 371 if (sc->sc_carpdev != NULL && 372 sc->sc_carpdev->if_carp != NULL) { 373 count = carp_addrcount( 374 (struct carp_if *)sc->sc_carpdev->if_carp, 375 ifatoia(ifa), CARP_COUNT_MASTER); 376 if ((cmd == RTM_ADD && count != 1) || 377 (cmd == RTM_DELETE && count != 0)) 378 continue; 379 } 380 381 /* Remove the existing host route, if any */ 382 rtrequest(RTM_DELETE, ifa->ifa_addr, 383 ifa->ifa_addr, ifa->ifa_netmask, 384 RTF_HOST, NULL); 385 386 rt = NULL; 387 (void)rtrequest(RTM_GET, ifa->ifa_addr, ifa->ifa_addr, 388 ifa->ifa_netmask, RTF_HOST, &rt); 389 hr_otherif = (rt && rt->rt_ifp != &sc->sc_if && 390 rt->rt_flags & (RTF_CLONING|RTF_CLONED)); 391 if (rt != NULL) { 392 RTFREE(rt); 393 rt = NULL; 394 } 395 396 /* Check for a network route on our interface */ 397 398 rt = NULL; 399 (void)rtrequest(RTM_GET, ifa->ifa_addr, ifa->ifa_addr, 400 ifa->ifa_netmask, 0, &rt); 401 nr_ourif = (rt && rt->rt_ifp == &sc->sc_if); 402 403 switch (cmd) { 404 case RTM_ADD: 405 if (hr_otherif) { 406 ifa->ifa_rtrequest = NULL; 407 ifa->ifa_flags &= ~RTF_CLONING; 408 409 rtrequest(RTM_ADD, ifa->ifa_addr, 410 ifa->ifa_addr, ifa->ifa_netmask, 411 RTF_UP | RTF_HOST, NULL); 412 } 413 if (!hr_otherif || nr_ourif || !rt) { 414 if (nr_ourif && !(rt->rt_flags & 415 RTF_CLONING)) 416 rtrequest(RTM_DELETE, 417 ifa->ifa_addr, 418 ifa->ifa_addr, 419 ifa->ifa_netmask, 0, NULL); 420 421 ifa->ifa_rtrequest = arp_rtrequest; 422 ifa->ifa_flags |= RTF_CLONING; 423 424 if (rtrequest(RTM_ADD, ifa->ifa_addr, 425 ifa->ifa_addr, ifa->ifa_netmask, 0, 426 NULL) == 0) 427 ifa->ifa_flags |= IFA_ROUTE; 428 } 429 break; 430 case RTM_DELETE: 431 break; 432 default: 433 break; 434 } 435 if (rt != NULL) { 436 RTFREE(rt); 437 rt = NULL; 438 } 439 break; 440 } 441 442 #ifdef INET6 443 case AF_INET6: 444 if (cmd == RTM_ADD) 445 in6_ifaddloop(ifa); 446 else 447 in6_ifremloop(ifa); 448 break; 449 #endif /* INET6 */ 450 default: 451 break; 452 } 453 } 454 splx(s); 455 } 456 457 /* 458 * process input packet. 459 * we have rearranged checks order compared to the rfc, 460 * but it seems more efficient this way or not possible otherwise. 461 */ 462 void 463 carp_proto_input(struct mbuf *m, ...) 464 { 465 struct ip *ip = mtod(m, struct ip *); 466 struct carp_softc *sc = NULL; 467 struct carp_header *ch; 468 int iplen, len; 469 va_list ap; 470 471 va_start(ap, m); 472 va_end(ap); 473 474 CARP_STATINC(CARP_STAT_IPACKETS); 475 MCLAIM(m, &carp_proto_mowner_rx); 476 477 if (!carp_opts[CARPCTL_ALLOW]) { 478 m_freem(m); 479 return; 480 } 481 482 /* check if received on a valid carp interface */ 483 if (m->m_pkthdr.rcvif->if_type != IFT_CARP) { 484 CARP_STATINC(CARP_STAT_BADIF); 485 CARP_LOG(sc, ("packet received on non-carp interface: %s", 486 m->m_pkthdr.rcvif->if_xname)); 487 m_freem(m); 488 return; 489 } 490 491 /* verify that the IP TTL is 255. */ 492 if (ip->ip_ttl != CARP_DFLTTL) { 493 CARP_STATINC(CARP_STAT_BADTTL); 494 CARP_LOG(sc, ("received ttl %d != %d on %s", ip->ip_ttl, 495 CARP_DFLTTL, m->m_pkthdr.rcvif->if_xname)); 496 m_freem(m); 497 return; 498 } 499 500 /* 501 * verify that the received packet length is 502 * equal to the CARP header 503 */ 504 iplen = ip->ip_hl << 2; 505 len = iplen + sizeof(*ch); 506 if (len > m->m_pkthdr.len) { 507 CARP_STATINC(CARP_STAT_BADLEN); 508 CARP_LOG(sc, ("packet too short %d on %s", m->m_pkthdr.len, 509 m->m_pkthdr.rcvif->if_xname)); 510 m_freem(m); 511 return; 512 } 513 514 if ((m = m_pullup(m, len)) == NULL) { 515 CARP_STATINC(CARP_STAT_HDROPS); 516 return; 517 } 518 ip = mtod(m, struct ip *); 519 ch = (struct carp_header *)((char *)ip + iplen); 520 /* verify the CARP checksum */ 521 m->m_data += iplen; 522 if (carp_cksum(m, len - iplen)) { 523 CARP_STATINC(CARP_STAT_BADSUM); 524 CARP_LOG(sc, ("checksum failed on %s", 525 m->m_pkthdr.rcvif->if_xname)); 526 m_freem(m); 527 return; 528 } 529 m->m_data -= iplen; 530 531 carp_proto_input_c(m, ch, AF_INET); 532 } 533 534 #ifdef INET6 535 int 536 carp6_proto_input(struct mbuf **mp, int *offp, int proto) 537 { 538 struct mbuf *m = *mp; 539 struct carp_softc *sc = NULL; 540 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 541 struct carp_header *ch; 542 u_int len; 543 544 CARP_STATINC(CARP_STAT_IPACKETS6); 545 MCLAIM(m, &carp_proto6_mowner_rx); 546 547 if (!carp_opts[CARPCTL_ALLOW]) { 548 m_freem(m); 549 return (IPPROTO_DONE); 550 } 551 552 /* check if received on a valid carp interface */ 553 if (m->m_pkthdr.rcvif->if_type != IFT_CARP) { 554 CARP_STATINC(CARP_STAT_BADIF); 555 CARP_LOG(sc, ("packet received on non-carp interface: %s", 556 m->m_pkthdr.rcvif->if_xname)); 557 m_freem(m); 558 return (IPPROTO_DONE); 559 } 560 561 /* verify that the IP TTL is 255 */ 562 if (ip6->ip6_hlim != CARP_DFLTTL) { 563 CARP_STATINC(CARP_STAT_BADTTL); 564 CARP_LOG(sc, ("received ttl %d != %d on %s", ip6->ip6_hlim, 565 CARP_DFLTTL, m->m_pkthdr.rcvif->if_xname)); 566 m_freem(m); 567 return (IPPROTO_DONE); 568 } 569 570 /* verify that we have a complete carp packet */ 571 len = m->m_len; 572 IP6_EXTHDR_GET(ch, struct carp_header *, m, *offp, sizeof(*ch)); 573 if (ch == NULL) { 574 CARP_STATINC(CARP_STAT_BADLEN); 575 CARP_LOG(sc, ("packet size %u too small", len)); 576 return (IPPROTO_DONE); 577 } 578 579 580 /* verify the CARP checksum */ 581 m->m_data += *offp; 582 if (carp_cksum(m, sizeof(*ch))) { 583 CARP_STATINC(CARP_STAT_BADSUM); 584 CARP_LOG(sc, ("checksum failed, on %s", 585 m->m_pkthdr.rcvif->if_xname)); 586 m_freem(m); 587 return (IPPROTO_DONE); 588 } 589 m->m_data -= *offp; 590 591 carp_proto_input_c(m, ch, AF_INET6); 592 return (IPPROTO_DONE); 593 } 594 #endif /* INET6 */ 595 596 void 597 carp_proto_input_c(struct mbuf *m, struct carp_header *ch, sa_family_t af) 598 { 599 struct carp_softc *sc; 600 u_int64_t tmp_counter; 601 struct timeval sc_tv, ch_tv; 602 603 TAILQ_FOREACH(sc, &((struct carp_if *) 604 m->m_pkthdr.rcvif->if_carpdev->if_carp)->vhif_vrs, sc_list) 605 if (sc->sc_vhid == ch->carp_vhid) 606 break; 607 608 if (!sc || (sc->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) != 609 (IFF_UP|IFF_RUNNING)) { 610 CARP_STATINC(CARP_STAT_BADVHID); 611 m_freem(m); 612 return; 613 } 614 615 /* 616 * Check if our own advertisement was duplicated 617 * from a non simplex interface. 618 * XXX If there is no address on our physical interface 619 * there is no way to distinguish our ads from the ones 620 * another carp host might have sent us. 621 */ 622 if ((sc->sc_carpdev->if_flags & IFF_SIMPLEX) == 0) { 623 struct sockaddr sa; 624 struct ifaddr *ifa; 625 626 memset(&sa, 0, sizeof(sa)); 627 sa.sa_family = af; 628 ifa = ifaof_ifpforaddr(&sa, sc->sc_carpdev); 629 630 if (ifa && af == AF_INET) { 631 struct ip *ip = mtod(m, struct ip *); 632 if (ip->ip_src.s_addr == 633 ifatoia(ifa)->ia_addr.sin_addr.s_addr) { 634 m_freem(m); 635 return; 636 } 637 } 638 #ifdef INET6 639 if (ifa && af == AF_INET6) { 640 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 641 struct in6_addr in6_src, in6_found; 642 643 in6_src = ip6->ip6_src; 644 in6_found = ifatoia6(ifa)->ia_addr.sin6_addr; 645 if (IN6_IS_ADDR_LINKLOCAL(&in6_src)) 646 in6_src.s6_addr16[1] = 0; 647 if (IN6_IS_ADDR_LINKLOCAL(&in6_found)) 648 in6_found.s6_addr16[1] = 0; 649 if (IN6_ARE_ADDR_EQUAL(&in6_src, &in6_found)) { 650 m_freem(m); 651 return; 652 } 653 } 654 #endif /* INET6 */ 655 } 656 657 nanotime(&sc->sc_if.if_lastchange); 658 sc->sc_if.if_ipackets++; 659 sc->sc_if.if_ibytes += m->m_pkthdr.len; 660 661 /* verify the CARP version. */ 662 if (ch->carp_version != CARP_VERSION) { 663 CARP_STATINC(CARP_STAT_BADVER); 664 sc->sc_if.if_ierrors++; 665 CARP_LOG(sc, ("invalid version %d != %d", 666 ch->carp_version, CARP_VERSION)); 667 m_freem(m); 668 return; 669 } 670 671 /* verify the hash */ 672 if (carp_hmac_verify(sc, ch->carp_counter, ch->carp_md)) { 673 CARP_STATINC(CARP_STAT_BADAUTH); 674 sc->sc_if.if_ierrors++; 675 CARP_LOG(sc, ("incorrect hash")); 676 m_freem(m); 677 return; 678 } 679 680 tmp_counter = ntohl(ch->carp_counter[0]); 681 tmp_counter = tmp_counter<<32; 682 tmp_counter += ntohl(ch->carp_counter[1]); 683 684 /* XXX Replay protection goes here */ 685 686 sc->sc_init_counter = 0; 687 sc->sc_counter = tmp_counter; 688 689 690 sc_tv.tv_sec = sc->sc_advbase; 691 if (carp_suppress_preempt && sc->sc_advskew < 240) 692 sc_tv.tv_usec = 240 * 1000000 / 256; 693 else 694 sc_tv.tv_usec = sc->sc_advskew * 1000000 / 256; 695 ch_tv.tv_sec = ch->carp_advbase; 696 ch_tv.tv_usec = ch->carp_advskew * 1000000 / 256; 697 698 switch (sc->sc_state) { 699 case INIT: 700 break; 701 case MASTER: 702 /* 703 * If we receive an advertisement from a backup who's going to 704 * be more frequent than us, go into BACKUP state. 705 */ 706 if (timercmp(&sc_tv, &ch_tv, >) || 707 timercmp(&sc_tv, &ch_tv, ==)) { 708 callout_stop(&sc->sc_ad_tmo); 709 CARP_LOG(sc, ("MASTER -> BACKUP (more frequent advertisement received)")); 710 carp_set_state(sc, BACKUP); 711 carp_setrun(sc, 0); 712 carp_setroute(sc, RTM_DELETE); 713 } 714 break; 715 case BACKUP: 716 /* 717 * If we're pre-empting masters who advertise slower than us, 718 * and this one claims to be slower, treat him as down. 719 */ 720 if (carp_opts[CARPCTL_PREEMPT] && timercmp(&sc_tv, &ch_tv, <)) { 721 CARP_LOG(sc, ("BACKUP -> MASTER (preempting a slower master)")); 722 carp_master_down(sc); 723 break; 724 } 725 726 /* 727 * If the master is going to advertise at such a low frequency 728 * that he's guaranteed to time out, we'd might as well just 729 * treat him as timed out now. 730 */ 731 sc_tv.tv_sec = sc->sc_advbase * 3; 732 if (timercmp(&sc_tv, &ch_tv, <)) { 733 CARP_LOG(sc, ("BACKUP -> MASTER (master timed out)")); 734 carp_master_down(sc); 735 break; 736 } 737 738 /* 739 * Otherwise, we reset the counter and wait for the next 740 * advertisement. 741 */ 742 carp_setrun(sc, af); 743 break; 744 } 745 746 m_freem(m); 747 return; 748 } 749 750 /* 751 * Interface side of the CARP implementation. 752 */ 753 754 /* ARGSUSED */ 755 void 756 carpattach(int n) 757 { 758 if_clone_attach(&carp_cloner); 759 760 carpstat_percpu = percpu_alloc(sizeof(uint64_t) * CARP_NSTATS); 761 } 762 763 int 764 carp_clone_create(struct if_clone *ifc, int unit) 765 { 766 extern int ifqmaxlen; 767 struct carp_softc *sc; 768 struct ifnet *ifp; 769 770 sc = malloc(sizeof(*sc), M_DEVBUF, M_NOWAIT|M_ZERO); 771 if (!sc) 772 return (ENOMEM); 773 774 sc->sc_suppress = 0; 775 sc->sc_advbase = CARP_DFLTINTV; 776 sc->sc_vhid = -1; /* required setting */ 777 sc->sc_advskew = 0; 778 sc->sc_init_counter = 1; 779 sc->sc_naddrs = sc->sc_naddrs6 = 0; 780 #ifdef INET6 781 sc->sc_im6o.im6o_multicast_hlim = CARP_DFLTTL; 782 #endif /* INET6 */ 783 784 callout_init(&sc->sc_ad_tmo, 0); 785 callout_init(&sc->sc_md_tmo, 0); 786 callout_init(&sc->sc_md6_tmo, 0); 787 788 callout_setfunc(&sc->sc_ad_tmo, carp_send_ad, sc); 789 callout_setfunc(&sc->sc_md_tmo, carp_master_down, sc); 790 callout_setfunc(&sc->sc_md6_tmo, carp_master_down, sc); 791 792 LIST_INIT(&sc->carp_mc_listhead); 793 ifp = &sc->sc_if; 794 ifp->if_softc = sc; 795 snprintf(ifp->if_xname, sizeof ifp->if_xname, "%s%d", ifc->ifc_name, 796 unit); 797 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; 798 ifp->if_ioctl = carp_ioctl; 799 ifp->if_start = carp_start; 800 ifp->if_output = carp_output; 801 ifp->if_type = IFT_CARP; 802 ifp->if_addrlen = ETHER_ADDR_LEN; 803 ifp->if_hdrlen = ETHER_HDR_LEN; 804 ifp->if_mtu = ETHERMTU; 805 IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen); 806 IFQ_SET_READY(&ifp->if_snd); 807 if_attach(ifp); 808 809 if_alloc_sadl(ifp); 810 ifp->if_broadcastaddr = etherbroadcastaddr; 811 carp_set_enaddr(sc); 812 LIST_INIT(&sc->sc_ac.ec_multiaddrs); 813 bpf_attach(ifp, DLT_EN10MB, ETHER_HDR_LEN); 814 #ifdef MBUFTRACE 815 strlcpy(sc->sc_ac.ec_tx_mowner.mo_name, ifp->if_xname, 816 sizeof(sc->sc_ac.ec_tx_mowner.mo_name)); 817 strlcpy(sc->sc_ac.ec_tx_mowner.mo_descr, "tx", 818 sizeof(sc->sc_ac.ec_tx_mowner.mo_descr)); 819 strlcpy(sc->sc_ac.ec_rx_mowner.mo_name, ifp->if_xname, 820 sizeof(sc->sc_ac.ec_rx_mowner.mo_name)); 821 strlcpy(sc->sc_ac.ec_rx_mowner.mo_descr, "rx", 822 sizeof(sc->sc_ac.ec_rx_mowner.mo_descr)); 823 MOWNER_ATTACH(&sc->sc_ac.ec_tx_mowner); 824 MOWNER_ATTACH(&sc->sc_ac.ec_rx_mowner); 825 ifp->if_mowner = &sc->sc_ac.ec_tx_mowner; 826 #endif 827 return (0); 828 } 829 830 int 831 carp_clone_destroy(struct ifnet *ifp) 832 { 833 struct carp_softc *sc = ifp->if_softc; 834 835 carpdetach(ifp->if_softc); 836 ether_ifdetach(ifp); 837 if_detach(ifp); 838 callout_destroy(&sc->sc_ad_tmo); 839 callout_destroy(&sc->sc_md_tmo); 840 callout_destroy(&sc->sc_md6_tmo); 841 free(ifp->if_softc, M_DEVBUF); 842 843 return (0); 844 } 845 846 void 847 carpdetach(struct carp_softc *sc) 848 { 849 struct carp_if *cif; 850 int s; 851 852 callout_stop(&sc->sc_ad_tmo); 853 callout_stop(&sc->sc_md_tmo); 854 callout_stop(&sc->sc_md6_tmo); 855 856 if (sc->sc_suppress) 857 carp_suppress_preempt--; 858 sc->sc_suppress = 0; 859 860 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) 861 carp_suppress_preempt--; 862 sc->sc_sendad_errors = 0; 863 864 carp_set_state(sc, INIT); 865 sc->sc_if.if_flags &= ~IFF_UP; 866 carp_setrun(sc, 0); 867 carp_multicast_cleanup(sc); 868 869 s = splnet(); 870 if (sc->sc_carpdev != NULL) { 871 /* XXX linkstatehook removal */ 872 cif = (struct carp_if *)sc->sc_carpdev->if_carp; 873 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list); 874 if (!--cif->vhif_nvrs) { 875 ifpromisc(sc->sc_carpdev, 0); 876 sc->sc_carpdev->if_carp = NULL; 877 free(cif, M_IFADDR); 878 } 879 } 880 sc->sc_carpdev = NULL; 881 splx(s); 882 } 883 884 /* Detach an interface from the carp. */ 885 void 886 carp_ifdetach(struct ifnet *ifp) 887 { 888 struct carp_softc *sc, *nextsc; 889 struct carp_if *cif = (struct carp_if *)ifp->if_carp; 890 891 for (sc = TAILQ_FIRST(&cif->vhif_vrs); sc; sc = nextsc) { 892 nextsc = TAILQ_NEXT(sc, sc_list); 893 carpdetach(sc); 894 } 895 } 896 897 int 898 carp_prepare_ad(struct mbuf *m, struct carp_softc *sc, 899 struct carp_header *ch) 900 { 901 if (sc->sc_init_counter) { 902 /* this could also be seconds since unix epoch */ 903 sc->sc_counter = cprng_fast64(); 904 } else 905 sc->sc_counter++; 906 907 ch->carp_counter[0] = htonl((sc->sc_counter>>32)&0xffffffff); 908 ch->carp_counter[1] = htonl(sc->sc_counter&0xffffffff); 909 910 carp_hmac_generate(sc, ch->carp_counter, ch->carp_md); 911 912 return (0); 913 } 914 915 void 916 carp_send_ad_all(void) 917 { 918 struct ifnet *ifp; 919 struct carp_if *cif; 920 struct carp_softc *vh; 921 922 TAILQ_FOREACH(ifp, &ifnet, if_list) { 923 if (ifp->if_carp == NULL || ifp->if_type == IFT_CARP) 924 continue; 925 926 cif = (struct carp_if *)ifp->if_carp; 927 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 928 if ((vh->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) == 929 (IFF_UP|IFF_RUNNING) && vh->sc_state == MASTER) 930 carp_send_ad(vh); 931 } 932 } 933 } 934 935 936 void 937 carp_send_ad(void *v) 938 { 939 struct carp_header ch; 940 struct timeval tv; 941 struct carp_softc *sc = v; 942 struct carp_header *ch_ptr; 943 struct mbuf *m; 944 int error, len, advbase, advskew, s; 945 struct ifaddr *ifa; 946 struct sockaddr sa; 947 948 s = splsoftnet(); 949 950 advbase = advskew = 0; /* Sssssh compiler */ 951 if (sc->sc_carpdev == NULL) { 952 sc->sc_if.if_oerrors++; 953 goto retry_later; 954 } 955 956 /* bow out if we've gone to backup (the carp interface is going down) */ 957 if (sc->sc_bow_out) { 958 sc->sc_bow_out = 0; 959 advbase = 255; 960 advskew = 255; 961 } else { 962 advbase = sc->sc_advbase; 963 if (!carp_suppress_preempt || sc->sc_advskew > 240) 964 advskew = sc->sc_advskew; 965 else 966 advskew = 240; 967 tv.tv_sec = advbase; 968 tv.tv_usec = advskew * 1000000 / 256; 969 } 970 971 ch.carp_version = CARP_VERSION; 972 ch.carp_type = CARP_ADVERTISEMENT; 973 ch.carp_vhid = sc->sc_vhid; 974 ch.carp_advbase = advbase; 975 ch.carp_advskew = advskew; 976 ch.carp_authlen = 7; /* XXX DEFINE */ 977 ch.carp_pad1 = 0; /* must be zero */ 978 ch.carp_cksum = 0; 979 980 981 #ifdef INET 982 if (sc->sc_naddrs) { 983 struct ip *ip; 984 985 MGETHDR(m, M_DONTWAIT, MT_HEADER); 986 if (m == NULL) { 987 sc->sc_if.if_oerrors++; 988 CARP_STATINC(CARP_STAT_ONOMEM); 989 /* XXX maybe less ? */ 990 goto retry_later; 991 } 992 MCLAIM(m, &carp_proto_mowner_tx); 993 len = sizeof(*ip) + sizeof(ch); 994 m->m_pkthdr.len = len; 995 m->m_pkthdr.rcvif = NULL; 996 m->m_len = len; 997 MH_ALIGN(m, m->m_len); 998 m->m_flags |= M_MCAST; 999 ip = mtod(m, struct ip *); 1000 ip->ip_v = IPVERSION; 1001 ip->ip_hl = sizeof(*ip) >> 2; 1002 ip->ip_tos = IPTOS_LOWDELAY; 1003 ip->ip_len = htons(len); 1004 ip->ip_id = 0; /* no need for id, we don't support fragments */ 1005 ip->ip_off = htons(IP_DF); 1006 ip->ip_ttl = CARP_DFLTTL; 1007 ip->ip_p = IPPROTO_CARP; 1008 ip->ip_sum = 0; 1009 1010 memset(&sa, 0, sizeof(sa)); 1011 sa.sa_family = AF_INET; 1012 ifa = ifaof_ifpforaddr(&sa, sc->sc_carpdev); 1013 if (ifa == NULL) 1014 ip->ip_src.s_addr = 0; 1015 else 1016 ip->ip_src.s_addr = 1017 ifatoia(ifa)->ia_addr.sin_addr.s_addr; 1018 ip->ip_dst.s_addr = INADDR_CARP_GROUP; 1019 1020 ch_ptr = (struct carp_header *)(&ip[1]); 1021 memcpy(ch_ptr, &ch, sizeof(ch)); 1022 if (carp_prepare_ad(m, sc, ch_ptr)) 1023 goto retry_later; 1024 1025 m->m_data += sizeof(*ip); 1026 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip)); 1027 m->m_data -= sizeof(*ip); 1028 1029 nanotime(&sc->sc_if.if_lastchange); 1030 sc->sc_if.if_opackets++; 1031 sc->sc_if.if_obytes += len; 1032 CARP_STATINC(CARP_STAT_OPACKETS); 1033 1034 error = ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, 1035 NULL); 1036 if (error) { 1037 if (error == ENOBUFS) 1038 CARP_STATINC(CARP_STAT_ONOMEM); 1039 else 1040 CARP_LOG(sc, ("ip_output failed: %d", error)); 1041 sc->sc_if.if_oerrors++; 1042 if (sc->sc_sendad_errors < INT_MAX) 1043 sc->sc_sendad_errors++; 1044 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) { 1045 carp_suppress_preempt++; 1046 if (carp_suppress_preempt == 1) 1047 carp_send_ad_all(); 1048 } 1049 sc->sc_sendad_success = 0; 1050 } else { 1051 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) { 1052 if (++sc->sc_sendad_success >= 1053 CARP_SENDAD_MIN_SUCCESS) { 1054 carp_suppress_preempt--; 1055 sc->sc_sendad_errors = 0; 1056 } 1057 } else 1058 sc->sc_sendad_errors = 0; 1059 } 1060 } 1061 #endif /* INET */ 1062 #ifdef INET6 1063 if (sc->sc_naddrs6) { 1064 struct ip6_hdr *ip6; 1065 1066 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1067 if (m == NULL) { 1068 sc->sc_if.if_oerrors++; 1069 CARP_STATINC(CARP_STAT_ONOMEM); 1070 /* XXX maybe less ? */ 1071 goto retry_later; 1072 } 1073 MCLAIM(m, &carp_proto6_mowner_tx); 1074 len = sizeof(*ip6) + sizeof(ch); 1075 m->m_pkthdr.len = len; 1076 m->m_pkthdr.rcvif = NULL; 1077 m->m_len = len; 1078 MH_ALIGN(m, m->m_len); 1079 m->m_flags |= M_MCAST; 1080 ip6 = mtod(m, struct ip6_hdr *); 1081 memset(ip6, 0, sizeof(*ip6)); 1082 ip6->ip6_vfc |= IPV6_VERSION; 1083 ip6->ip6_hlim = CARP_DFLTTL; 1084 ip6->ip6_nxt = IPPROTO_CARP; 1085 1086 /* set the source address */ 1087 memset(&sa, 0, sizeof(sa)); 1088 sa.sa_family = AF_INET6; 1089 ifa = ifaof_ifpforaddr(&sa, sc->sc_carpdev); 1090 if (ifa == NULL) /* This should never happen with IPv6 */ 1091 memset(&ip6->ip6_src, 0, sizeof(struct in6_addr)); 1092 else 1093 bcopy(ifatoia6(ifa)->ia_addr.sin6_addr.s6_addr, 1094 &ip6->ip6_src, sizeof(struct in6_addr)); 1095 /* set the multicast destination */ 1096 1097 ip6->ip6_dst.s6_addr16[0] = htons(0xff02); 1098 ip6->ip6_dst.s6_addr8[15] = 0x12; 1099 if (in6_setscope(&ip6->ip6_dst, sc->sc_carpdev, NULL) != 0) { 1100 sc->sc_if.if_oerrors++; 1101 m_freem(m); 1102 CARP_LOG(sc, ("in6_setscope failed")); 1103 goto retry_later; 1104 } 1105 1106 ch_ptr = (struct carp_header *)(&ip6[1]); 1107 memcpy(ch_ptr, &ch, sizeof(ch)); 1108 if (carp_prepare_ad(m, sc, ch_ptr)) 1109 goto retry_later; 1110 1111 m->m_data += sizeof(*ip6); 1112 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip6)); 1113 m->m_data -= sizeof(*ip6); 1114 1115 nanotime(&sc->sc_if.if_lastchange); 1116 sc->sc_if.if_opackets++; 1117 sc->sc_if.if_obytes += len; 1118 CARP_STATINC(CARP_STAT_OPACKETS6); 1119 1120 error = ip6_output(m, NULL, NULL, 0, &sc->sc_im6o, NULL, NULL); 1121 if (error) { 1122 if (error == ENOBUFS) 1123 CARP_STATINC(CARP_STAT_ONOMEM); 1124 else 1125 CARP_LOG(sc, ("ip6_output failed: %d", error)); 1126 sc->sc_if.if_oerrors++; 1127 if (sc->sc_sendad_errors < INT_MAX) 1128 sc->sc_sendad_errors++; 1129 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) { 1130 carp_suppress_preempt++; 1131 if (carp_suppress_preempt == 1) 1132 carp_send_ad_all(); 1133 } 1134 sc->sc_sendad_success = 0; 1135 } else { 1136 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) { 1137 if (++sc->sc_sendad_success >= 1138 CARP_SENDAD_MIN_SUCCESS) { 1139 carp_suppress_preempt--; 1140 sc->sc_sendad_errors = 0; 1141 } 1142 } else 1143 sc->sc_sendad_errors = 0; 1144 } 1145 } 1146 #endif /* INET6 */ 1147 1148 retry_later: 1149 splx(s); 1150 if (advbase != 255 || advskew != 255) 1151 callout_schedule(&sc->sc_ad_tmo, tvtohz(&tv)); 1152 } 1153 1154 /* 1155 * Broadcast a gratuitous ARP request containing 1156 * the virtual router MAC address for each IP address 1157 * associated with the virtual router. 1158 */ 1159 void 1160 carp_send_arp(struct carp_softc *sc) 1161 { 1162 struct ifaddr *ifa; 1163 struct in_addr *in; 1164 int s = splsoftnet(); 1165 1166 IFADDR_FOREACH(ifa, &sc->sc_if) { 1167 1168 if (ifa->ifa_addr->sa_family != AF_INET) 1169 continue; 1170 1171 in = &ifatoia(ifa)->ia_addr.sin_addr; 1172 arprequest(sc->sc_carpdev, in, in, CLLADDR(sc->sc_if.if_sadl)); 1173 } 1174 splx(s); 1175 } 1176 1177 #ifdef INET6 1178 void 1179 carp_send_na(struct carp_softc *sc) 1180 { 1181 struct ifaddr *ifa; 1182 struct in6_addr *in6; 1183 static struct in6_addr mcast = IN6ADDR_LINKLOCAL_ALLNODES_INIT; 1184 int s = splsoftnet(); 1185 1186 IFADDR_FOREACH(ifa, &sc->sc_if) { 1187 1188 if (ifa->ifa_addr->sa_family != AF_INET6) 1189 continue; 1190 1191 in6 = &ifatoia6(ifa)->ia_addr.sin6_addr; 1192 nd6_na_output(sc->sc_carpdev, &mcast, in6, 1193 ND_NA_FLAG_OVERRIDE, 1, NULL); 1194 } 1195 splx(s); 1196 } 1197 #endif /* INET6 */ 1198 1199 /* 1200 * Based on bridge_hash() in if_bridge.c 1201 */ 1202 #define mix(a,b,c) \ 1203 do { \ 1204 a -= b; a -= c; a ^= (c >> 13); \ 1205 b -= c; b -= a; b ^= (a << 8); \ 1206 c -= a; c -= b; c ^= (b >> 13); \ 1207 a -= b; a -= c; a ^= (c >> 12); \ 1208 b -= c; b -= a; b ^= (a << 16); \ 1209 c -= a; c -= b; c ^= (b >> 5); \ 1210 a -= b; a -= c; a ^= (c >> 3); \ 1211 b -= c; b -= a; b ^= (a << 10); \ 1212 c -= a; c -= b; c ^= (b >> 15); \ 1213 } while (0) 1214 1215 u_int32_t 1216 carp_hash(struct carp_softc *sc, u_char *src) 1217 { 1218 u_int32_t a = 0x9e3779b9, b = sc->sc_hashkey[0], c = sc->sc_hashkey[1]; 1219 1220 c += sc->sc_key[3] << 24; 1221 c += sc->sc_key[2] << 16; 1222 c += sc->sc_key[1] << 8; 1223 c += sc->sc_key[0]; 1224 b += src[5] << 8; 1225 b += src[4]; 1226 a += src[3] << 24; 1227 a += src[2] << 16; 1228 a += src[1] << 8; 1229 a += src[0]; 1230 1231 mix(a, b, c); 1232 return (c); 1233 } 1234 1235 int 1236 carp_addrcount(struct carp_if *cif, struct in_ifaddr *ia, int type) 1237 { 1238 struct carp_softc *vh; 1239 struct ifaddr *ifa; 1240 int count = 0; 1241 1242 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1243 if ((type == CARP_COUNT_RUNNING && 1244 (vh->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) == 1245 (IFF_UP|IFF_RUNNING)) || 1246 (type == CARP_COUNT_MASTER && vh->sc_state == MASTER)) { 1247 IFADDR_FOREACH(ifa, &vh->sc_if) { 1248 if (ifa->ifa_addr->sa_family == AF_INET && 1249 ia->ia_addr.sin_addr.s_addr == 1250 ifatoia(ifa)->ia_addr.sin_addr.s_addr) 1251 count++; 1252 } 1253 } 1254 } 1255 return (count); 1256 } 1257 1258 int 1259 carp_iamatch(struct in_ifaddr *ia, u_char *src, 1260 u_int32_t *count, u_int32_t index) 1261 { 1262 struct carp_softc *sc = ia->ia_ifp->if_softc; 1263 1264 if (carp_opts[CARPCTL_ARPBALANCE]) { 1265 /* 1266 * We use the source ip to decide which virtual host should 1267 * handle the request. If we're master of that virtual host, 1268 * then we respond, otherwise, just drop the arp packet on 1269 * the floor. 1270 */ 1271 1272 /* Count the elegible carp interfaces with this address */ 1273 if (*count == 0) 1274 *count = carp_addrcount( 1275 (struct carp_if *)ia->ia_ifp->if_carpdev->if_carp, 1276 ia, CARP_COUNT_RUNNING); 1277 1278 /* This should never happen, but... */ 1279 if (*count == 0) 1280 return (0); 1281 1282 if (carp_hash(sc, src) % *count == index - 1 && 1283 sc->sc_state == MASTER) { 1284 return (1); 1285 } 1286 } else { 1287 if (sc->sc_state == MASTER) 1288 return (1); 1289 } 1290 1291 return (0); 1292 } 1293 1294 #ifdef INET6 1295 struct ifaddr * 1296 carp_iamatch6(void *v, struct in6_addr *taddr) 1297 { 1298 struct carp_if *cif = v; 1299 struct carp_softc *vh; 1300 struct ifaddr *ifa; 1301 1302 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1303 IFADDR_FOREACH(ifa, &vh->sc_if) { 1304 if (IN6_ARE_ADDR_EQUAL(taddr, 1305 &ifatoia6(ifa)->ia_addr.sin6_addr) && 1306 ((vh->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) == 1307 (IFF_UP|IFF_RUNNING)) && vh->sc_state == MASTER) 1308 return (ifa); 1309 } 1310 } 1311 1312 return (NULL); 1313 } 1314 #endif /* INET6 */ 1315 1316 struct ifnet * 1317 carp_ourether(void *v, struct ether_header *eh, u_char iftype, int src) 1318 { 1319 struct carp_if *cif = (struct carp_if *)v; 1320 struct carp_softc *vh; 1321 u_int8_t *ena; 1322 1323 if (src) 1324 ena = (u_int8_t *)&eh->ether_shost; 1325 else 1326 ena = (u_int8_t *)&eh->ether_dhost; 1327 1328 switch (iftype) { 1329 case IFT_ETHER: 1330 case IFT_FDDI: 1331 if (ena[0] || ena[1] || ena[2] != 0x5e || ena[3] || ena[4] != 1) 1332 return (NULL); 1333 break; 1334 case IFT_ISO88025: 1335 if (ena[0] != 3 || ena[1] || ena[4] || ena[5]) 1336 return (NULL); 1337 break; 1338 default: 1339 return (NULL); 1340 break; 1341 } 1342 1343 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) 1344 if ((vh->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) == 1345 (IFF_UP|IFF_RUNNING) && vh->sc_state == MASTER && 1346 !memcmp(ena, CLLADDR(vh->sc_if.if_sadl), 1347 ETHER_ADDR_LEN)) { 1348 return (&vh->sc_if); 1349 } 1350 1351 return (NULL); 1352 } 1353 1354 int 1355 carp_input(struct mbuf *m, u_int8_t *shost, u_int8_t *dhost, u_int16_t etype) 1356 { 1357 struct ether_header eh; 1358 struct carp_if *cif = (struct carp_if *)m->m_pkthdr.rcvif->if_carp; 1359 struct ifnet *ifp; 1360 1361 memcpy(&eh.ether_shost, shost, sizeof(eh.ether_shost)); 1362 memcpy(&eh.ether_dhost, dhost, sizeof(eh.ether_dhost)); 1363 eh.ether_type = etype; 1364 1365 if (m->m_flags & (M_BCAST|M_MCAST)) { 1366 struct carp_softc *vh; 1367 struct mbuf *m0; 1368 1369 /* 1370 * XXX Should really check the list of multicast addresses 1371 * for each CARP interface _before_ copying. 1372 */ 1373 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1374 m0 = m_copym(m, 0, M_COPYALL, M_DONTWAIT); 1375 if (m0 == NULL) 1376 continue; 1377 m0->m_pkthdr.rcvif = &vh->sc_if; 1378 ether_input(&vh->sc_if, m0); 1379 } 1380 return (1); 1381 } 1382 1383 ifp = carp_ourether(cif, &eh, m->m_pkthdr.rcvif->if_type, 0); 1384 if (ifp == NULL) { 1385 return (1); 1386 } 1387 1388 m->m_pkthdr.rcvif = ifp; 1389 1390 bpf_mtap(ifp, m); 1391 ifp->if_ipackets++; 1392 ether_input(ifp, m); 1393 return (0); 1394 } 1395 1396 void 1397 carp_master_down(void *v) 1398 { 1399 struct carp_softc *sc = v; 1400 1401 switch (sc->sc_state) { 1402 case INIT: 1403 printf("%s: master_down event in INIT state\n", 1404 sc->sc_if.if_xname); 1405 break; 1406 case MASTER: 1407 break; 1408 case BACKUP: 1409 CARP_LOG(sc, ("INIT -> MASTER (preempting)")); 1410 carp_set_state(sc, MASTER); 1411 carp_send_ad(sc); 1412 carp_send_arp(sc); 1413 #ifdef INET6 1414 carp_send_na(sc); 1415 #endif /* INET6 */ 1416 carp_setrun(sc, 0); 1417 carp_setroute(sc, RTM_ADD); 1418 break; 1419 } 1420 } 1421 1422 /* 1423 * When in backup state, af indicates whether to reset the master down timer 1424 * for v4 or v6. If it's set to zero, reset the ones which are already pending. 1425 */ 1426 void 1427 carp_setrun(struct carp_softc *sc, sa_family_t af) 1428 { 1429 struct timeval tv; 1430 1431 if (sc->sc_carpdev == NULL) { 1432 sc->sc_if.if_flags &= ~IFF_RUNNING; 1433 carp_set_state(sc, INIT); 1434 return; 1435 } 1436 1437 if (sc->sc_if.if_flags & IFF_UP && sc->sc_vhid > 0 && 1438 (sc->sc_naddrs || sc->sc_naddrs6) && !sc->sc_suppress) { 1439 sc->sc_if.if_flags |= IFF_RUNNING; 1440 } else { 1441 sc->sc_if.if_flags &= ~IFF_RUNNING; 1442 carp_setroute(sc, RTM_DELETE); 1443 return; 1444 } 1445 1446 switch (sc->sc_state) { 1447 case INIT: 1448 carp_set_state(sc, BACKUP); 1449 carp_setroute(sc, RTM_DELETE); 1450 carp_setrun(sc, 0); 1451 break; 1452 case BACKUP: 1453 callout_stop(&sc->sc_ad_tmo); 1454 tv.tv_sec = 3 * sc->sc_advbase; 1455 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1456 switch (af) { 1457 #ifdef INET 1458 case AF_INET: 1459 callout_schedule(&sc->sc_md_tmo, tvtohz(&tv)); 1460 break; 1461 #endif /* INET */ 1462 #ifdef INET6 1463 case AF_INET6: 1464 callout_schedule(&sc->sc_md6_tmo, tvtohz(&tv)); 1465 break; 1466 #endif /* INET6 */ 1467 default: 1468 if (sc->sc_naddrs) 1469 callout_schedule(&sc->sc_md_tmo, tvtohz(&tv)); 1470 if (sc->sc_naddrs6) 1471 callout_schedule(&sc->sc_md6_tmo, tvtohz(&tv)); 1472 break; 1473 } 1474 break; 1475 case MASTER: 1476 tv.tv_sec = sc->sc_advbase; 1477 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1478 callout_schedule(&sc->sc_ad_tmo, tvtohz(&tv)); 1479 break; 1480 } 1481 } 1482 1483 void 1484 carp_multicast_cleanup(struct carp_softc *sc) 1485 { 1486 struct ip_moptions *imo = &sc->sc_imo; 1487 #ifdef INET6 1488 struct ip6_moptions *im6o = &sc->sc_im6o; 1489 #endif 1490 u_int16_t n = imo->imo_num_memberships; 1491 1492 /* Clean up our own multicast memberships */ 1493 while (n-- > 0) { 1494 if (imo->imo_membership[n] != NULL) { 1495 in_delmulti(imo->imo_membership[n]); 1496 imo->imo_membership[n] = NULL; 1497 } 1498 } 1499 imo->imo_num_memberships = 0; 1500 imo->imo_multicast_ifp = NULL; 1501 1502 #ifdef INET6 1503 while (!LIST_EMPTY(&im6o->im6o_memberships)) { 1504 struct in6_multi_mship *imm = 1505 LIST_FIRST(&im6o->im6o_memberships); 1506 1507 LIST_REMOVE(imm, i6mm_chain); 1508 in6_leavegroup(imm); 1509 } 1510 im6o->im6o_multicast_ifp = NULL; 1511 #endif 1512 1513 /* And any other multicast memberships */ 1514 carp_ether_purgemulti(sc); 1515 } 1516 1517 int 1518 carp_set_ifp(struct carp_softc *sc, struct ifnet *ifp) 1519 { 1520 struct carp_if *cif, *ncif = NULL; 1521 struct carp_softc *vr, *after = NULL; 1522 int myself = 0, error = 0; 1523 int s; 1524 1525 if (ifp == sc->sc_carpdev) 1526 return (0); 1527 1528 if (ifp != NULL) { 1529 if ((ifp->if_flags & IFF_MULTICAST) == 0) 1530 return (EADDRNOTAVAIL); 1531 1532 if (ifp->if_type == IFT_CARP) 1533 return (EINVAL); 1534 1535 if (ifp->if_carp == NULL) { 1536 ncif = malloc(sizeof(*cif), M_IFADDR, M_NOWAIT); 1537 if (ncif == NULL) 1538 return (ENOBUFS); 1539 if ((error = ifpromisc(ifp, 1))) { 1540 free(ncif, M_IFADDR); 1541 return (error); 1542 } 1543 1544 ncif->vhif_ifp = ifp; 1545 TAILQ_INIT(&ncif->vhif_vrs); 1546 } else { 1547 cif = (struct carp_if *)ifp->if_carp; 1548 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) 1549 if (vr != sc && vr->sc_vhid == sc->sc_vhid) 1550 return (EINVAL); 1551 } 1552 1553 /* detach from old interface */ 1554 if (sc->sc_carpdev != NULL) 1555 carpdetach(sc); 1556 1557 /* join multicast groups */ 1558 if (sc->sc_naddrs < 0 && 1559 (error = carp_join_multicast(sc)) != 0) { 1560 if (ncif != NULL) 1561 free(ncif, M_IFADDR); 1562 return (error); 1563 } 1564 1565 #ifdef INET6 1566 if (sc->sc_naddrs6 < 0 && 1567 (error = carp_join_multicast6(sc)) != 0) { 1568 if (ncif != NULL) 1569 free(ncif, M_IFADDR); 1570 carp_multicast_cleanup(sc); 1571 return (error); 1572 } 1573 #endif 1574 1575 /* attach carp interface to physical interface */ 1576 if (ncif != NULL) 1577 ifp->if_carp = (void *)ncif; 1578 sc->sc_carpdev = ifp; 1579 sc->sc_if.if_capabilities = ifp->if_capabilities & 1580 (IFCAP_TSOv4 | IFCAP_TSOv6 | 1581 IFCAP_CSUM_IPv4_Tx|IFCAP_CSUM_IPv4_Rx| 1582 IFCAP_CSUM_TCPv4_Tx|IFCAP_CSUM_TCPv4_Rx| 1583 IFCAP_CSUM_UDPv4_Tx|IFCAP_CSUM_UDPv4_Rx| 1584 IFCAP_CSUM_TCPv6_Tx|IFCAP_CSUM_TCPv6_Rx| 1585 IFCAP_CSUM_UDPv6_Tx|IFCAP_CSUM_UDPv6_Rx); 1586 1587 cif = (struct carp_if *)ifp->if_carp; 1588 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) { 1589 if (vr == sc) 1590 myself = 1; 1591 if (vr->sc_vhid < sc->sc_vhid) 1592 after = vr; 1593 } 1594 1595 if (!myself) { 1596 /* We're trying to keep things in order */ 1597 if (after == NULL) { 1598 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list); 1599 } else { 1600 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, 1601 sc, sc_list); 1602 } 1603 cif->vhif_nvrs++; 1604 } 1605 if (sc->sc_naddrs || sc->sc_naddrs6) 1606 sc->sc_if.if_flags |= IFF_UP; 1607 carp_set_enaddr(sc); 1608 s = splnet(); 1609 /* XXX linkstatehooks establish */ 1610 carp_carpdev_state(ifp); 1611 splx(s); 1612 } else { 1613 carpdetach(sc); 1614 sc->sc_if.if_flags &= ~(IFF_UP|IFF_RUNNING); 1615 } 1616 return (0); 1617 } 1618 1619 void 1620 carp_set_enaddr(struct carp_softc *sc) 1621 { 1622 uint8_t enaddr[ETHER_ADDR_LEN]; 1623 if (sc->sc_carpdev && sc->sc_carpdev->if_type == IFT_ISO88025) { 1624 enaddr[0] = 3; 1625 enaddr[1] = 0; 1626 enaddr[2] = 0x40 >> (sc->sc_vhid - 1); 1627 enaddr[3] = 0x40000 >> (sc->sc_vhid - 1); 1628 enaddr[4] = 0; 1629 enaddr[5] = 0; 1630 } else { 1631 enaddr[0] = 0; 1632 enaddr[1] = 0; 1633 enaddr[2] = 0x5e; 1634 enaddr[3] = 0; 1635 enaddr[4] = 1; 1636 enaddr[5] = sc->sc_vhid; 1637 } 1638 if_set_sadl(&sc->sc_if, enaddr, sizeof(enaddr), false); 1639 } 1640 1641 void 1642 carp_addr_updated(void *v) 1643 { 1644 struct carp_softc *sc = (struct carp_softc *) v; 1645 struct ifaddr *ifa; 1646 int new_naddrs = 0, new_naddrs6 = 0; 1647 1648 IFADDR_FOREACH(ifa, &sc->sc_if) { 1649 if (ifa->ifa_addr->sa_family == AF_INET) 1650 new_naddrs++; 1651 else if (ifa->ifa_addr->sa_family == AF_INET6) 1652 new_naddrs6++; 1653 } 1654 1655 /* Handle a callback after SIOCDIFADDR */ 1656 if (new_naddrs < sc->sc_naddrs || new_naddrs6 < sc->sc_naddrs6) { 1657 struct in_addr mc_addr; 1658 struct in_multi *inm; 1659 1660 sc->sc_naddrs = new_naddrs; 1661 sc->sc_naddrs6 = new_naddrs6; 1662 1663 /* Re-establish multicast membership removed by in_control */ 1664 mc_addr.s_addr = INADDR_CARP_GROUP; 1665 IN_LOOKUP_MULTI(mc_addr, &sc->sc_if, inm); 1666 if (inm == NULL) { 1667 memset(&sc->sc_imo, 0, sizeof(sc->sc_imo)); 1668 1669 if (sc->sc_carpdev != NULL && sc->sc_naddrs > 0) 1670 carp_join_multicast(sc); 1671 } 1672 1673 if (sc->sc_naddrs == 0 && sc->sc_naddrs6 == 0) { 1674 sc->sc_if.if_flags &= ~IFF_UP; 1675 carp_set_state(sc, INIT); 1676 } else 1677 carp_hmac_prepare(sc); 1678 } 1679 1680 carp_setrun(sc, 0); 1681 } 1682 1683 int 1684 carp_set_addr(struct carp_softc *sc, struct sockaddr_in *sin) 1685 { 1686 struct ifnet *ifp = sc->sc_carpdev; 1687 struct in_ifaddr *ia, *ia_if; 1688 int error = 0; 1689 1690 if (sin->sin_addr.s_addr == 0) { 1691 if (!(sc->sc_if.if_flags & IFF_UP)) 1692 carp_set_state(sc, INIT); 1693 if (sc->sc_naddrs) 1694 sc->sc_if.if_flags |= IFF_UP; 1695 carp_setrun(sc, 0); 1696 return (0); 1697 } 1698 1699 /* we have to do this by hand to ensure we don't match on ourselves */ 1700 ia_if = NULL; 1701 for (ia = TAILQ_FIRST(&in_ifaddrhead); ia; 1702 ia = TAILQ_NEXT(ia, ia_list)) { 1703 1704 /* and, yeah, we need a multicast-capable iface too */ 1705 if (ia->ia_ifp != &sc->sc_if && 1706 ia->ia_ifp->if_type != IFT_CARP && 1707 (ia->ia_ifp->if_flags & IFF_MULTICAST) && 1708 (sin->sin_addr.s_addr & ia->ia_subnetmask) == 1709 ia->ia_subnet) { 1710 if (!ia_if) 1711 ia_if = ia; 1712 } 1713 } 1714 1715 if (ia_if) { 1716 ia = ia_if; 1717 if (ifp) { 1718 if (ifp != ia->ia_ifp) 1719 return (EADDRNOTAVAIL); 1720 } else { 1721 ifp = ia->ia_ifp; 1722 } 1723 } 1724 1725 if ((error = carp_set_ifp(sc, ifp))) 1726 return (error); 1727 1728 if (sc->sc_carpdev == NULL) 1729 return (EADDRNOTAVAIL); 1730 1731 if (sc->sc_naddrs == 0 && (error = carp_join_multicast(sc)) != 0) 1732 return (error); 1733 1734 sc->sc_naddrs++; 1735 if (sc->sc_carpdev != NULL) 1736 sc->sc_if.if_flags |= IFF_UP; 1737 1738 carp_set_state(sc, INIT); 1739 carp_setrun(sc, 0); 1740 1741 /* 1742 * Hook if_addrhooks so that we get a callback after in_ifinit has run, 1743 * to correct any inappropriate routes that it inserted. 1744 */ 1745 if (sc->ah_cookie == 0) { 1746 /* XXX link address hook */ 1747 } 1748 1749 return (0); 1750 } 1751 1752 int 1753 carp_join_multicast(struct carp_softc *sc) 1754 { 1755 struct ip_moptions *imo = &sc->sc_imo, tmpimo; 1756 struct in_addr addr; 1757 1758 memset(&tmpimo, 0, sizeof(tmpimo)); 1759 addr.s_addr = INADDR_CARP_GROUP; 1760 if ((tmpimo.imo_membership[0] = 1761 in_addmulti(&addr, &sc->sc_if)) == NULL) { 1762 return (ENOBUFS); 1763 } 1764 1765 imo->imo_membership[0] = tmpimo.imo_membership[0]; 1766 imo->imo_num_memberships = 1; 1767 imo->imo_multicast_ifp = &sc->sc_if; 1768 imo->imo_multicast_ttl = CARP_DFLTTL; 1769 imo->imo_multicast_loop = 0; 1770 return (0); 1771 } 1772 1773 1774 #ifdef INET6 1775 int 1776 carp_set_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6) 1777 { 1778 struct ifnet *ifp = sc->sc_carpdev; 1779 struct in6_ifaddr *ia, *ia_if; 1780 int error = 0; 1781 1782 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1783 if (!(sc->sc_if.if_flags & IFF_UP)) 1784 carp_set_state(sc, INIT); 1785 if (sc->sc_naddrs6) 1786 sc->sc_if.if_flags |= IFF_UP; 1787 carp_setrun(sc, 0); 1788 return (0); 1789 } 1790 1791 /* we have to do this by hand to ensure we don't match on ourselves */ 1792 ia_if = NULL; 1793 for (ia = in6_ifaddr; ia; ia = ia->ia_next) { 1794 int i; 1795 1796 for (i = 0; i < 4; i++) { 1797 if ((sin6->sin6_addr.s6_addr32[i] & 1798 ia->ia_prefixmask.sin6_addr.s6_addr32[i]) != 1799 (ia->ia_addr.sin6_addr.s6_addr32[i] & 1800 ia->ia_prefixmask.sin6_addr.s6_addr32[i])) 1801 break; 1802 } 1803 /* and, yeah, we need a multicast-capable iface too */ 1804 if (ia->ia_ifp != &sc->sc_if && 1805 ia->ia_ifp->if_type != IFT_CARP && 1806 (ia->ia_ifp->if_flags & IFF_MULTICAST) && 1807 (i == 4)) { 1808 if (!ia_if) 1809 ia_if = ia; 1810 } 1811 } 1812 1813 if (ia_if) { 1814 ia = ia_if; 1815 if (sc->sc_carpdev) { 1816 if (sc->sc_carpdev != ia->ia_ifp) 1817 return (EADDRNOTAVAIL); 1818 } else { 1819 ifp = ia->ia_ifp; 1820 } 1821 } 1822 1823 if ((error = carp_set_ifp(sc, ifp))) 1824 return (error); 1825 1826 if (sc->sc_carpdev == NULL) 1827 return (EADDRNOTAVAIL); 1828 1829 if (sc->sc_naddrs6 == 0 && (error = carp_join_multicast6(sc)) != 0) 1830 return (error); 1831 1832 sc->sc_naddrs6++; 1833 if (sc->sc_carpdev != NULL) 1834 sc->sc_if.if_flags |= IFF_UP; 1835 carp_set_state(sc, INIT); 1836 carp_setrun(sc, 0); 1837 1838 return (0); 1839 } 1840 1841 int 1842 carp_join_multicast6(struct carp_softc *sc) 1843 { 1844 struct in6_multi_mship *imm, *imm2; 1845 struct ip6_moptions *im6o = &sc->sc_im6o; 1846 struct sockaddr_in6 addr6; 1847 int error; 1848 1849 /* Join IPv6 CARP multicast group */ 1850 memset(&addr6, 0, sizeof(addr6)); 1851 addr6.sin6_family = AF_INET6; 1852 addr6.sin6_len = sizeof(addr6); 1853 addr6.sin6_addr.s6_addr16[0] = htons(0xff02); 1854 addr6.sin6_addr.s6_addr16[1] = htons(sc->sc_if.if_index); 1855 addr6.sin6_addr.s6_addr8[15] = 0x12; 1856 if ((imm = in6_joingroup(&sc->sc_if, 1857 &addr6.sin6_addr, &error, 0)) == NULL) { 1858 return (error); 1859 } 1860 /* join solicited multicast address */ 1861 memset(&addr6.sin6_addr, 0, sizeof(addr6.sin6_addr)); 1862 addr6.sin6_addr.s6_addr16[0] = htons(0xff02); 1863 addr6.sin6_addr.s6_addr16[1] = htons(sc->sc_if.if_index); 1864 addr6.sin6_addr.s6_addr32[1] = 0; 1865 addr6.sin6_addr.s6_addr32[2] = htonl(1); 1866 addr6.sin6_addr.s6_addr32[3] = 0; 1867 addr6.sin6_addr.s6_addr8[12] = 0xff; 1868 if ((imm2 = in6_joingroup(&sc->sc_if, 1869 &addr6.sin6_addr, &error, 0)) == NULL) { 1870 in6_leavegroup(imm); 1871 return (error); 1872 } 1873 1874 /* apply v6 multicast membership */ 1875 im6o->im6o_multicast_ifp = &sc->sc_if; 1876 if (imm) 1877 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, 1878 i6mm_chain); 1879 if (imm2) 1880 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm2, 1881 i6mm_chain); 1882 1883 return (0); 1884 } 1885 1886 #endif /* INET6 */ 1887 1888 int 1889 carp_ioctl(struct ifnet *ifp, u_long cmd, void *data) 1890 { 1891 struct lwp *l = curlwp; /* XXX */ 1892 struct carp_softc *sc = ifp->if_softc, *vr; 1893 struct carpreq carpr; 1894 struct ifaddr *ifa; 1895 struct ifreq *ifr; 1896 struct ifnet *cdev = NULL; 1897 int error = 0; 1898 1899 ifa = (struct ifaddr *)data; 1900 ifr = (struct ifreq *)data; 1901 1902 switch (cmd) { 1903 case SIOCINITIFADDR: 1904 switch (ifa->ifa_addr->sa_family) { 1905 #ifdef INET 1906 case AF_INET: 1907 sc->sc_if.if_flags |= IFF_UP; 1908 memcpy(ifa->ifa_dstaddr, ifa->ifa_addr, 1909 sizeof(struct sockaddr)); 1910 error = carp_set_addr(sc, satosin(ifa->ifa_addr)); 1911 break; 1912 #endif /* INET */ 1913 #ifdef INET6 1914 case AF_INET6: 1915 sc->sc_if.if_flags|= IFF_UP; 1916 error = carp_set_addr6(sc, satosin6(ifa->ifa_addr)); 1917 break; 1918 #endif /* INET6 */ 1919 default: 1920 error = EAFNOSUPPORT; 1921 break; 1922 } 1923 break; 1924 1925 case SIOCSIFFLAGS: 1926 if ((error = ifioctl_common(ifp, cmd, data)) != 0) 1927 break; 1928 if (sc->sc_state != INIT && !(ifr->ifr_flags & IFF_UP)) { 1929 callout_stop(&sc->sc_ad_tmo); 1930 callout_stop(&sc->sc_md_tmo); 1931 callout_stop(&sc->sc_md6_tmo); 1932 if (sc->sc_state == MASTER) { 1933 /* we need the interface up to bow out */ 1934 sc->sc_if.if_flags |= IFF_UP; 1935 sc->sc_bow_out = 1; 1936 carp_send_ad(sc); 1937 } 1938 sc->sc_if.if_flags &= ~IFF_UP; 1939 carp_set_state(sc, INIT); 1940 carp_setrun(sc, 0); 1941 } else if (sc->sc_state == INIT && (ifr->ifr_flags & IFF_UP)) { 1942 sc->sc_if.if_flags |= IFF_UP; 1943 carp_setrun(sc, 0); 1944 } 1945 break; 1946 1947 case SIOCSVH: 1948 if (l == NULL) 1949 break; 1950 if ((error = kauth_authorize_network(l->l_cred, 1951 KAUTH_NETWORK_INTERFACE, 1952 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 1953 NULL)) != 0) 1954 break; 1955 if ((error = copyin(ifr->ifr_data, &carpr, sizeof carpr))) 1956 break; 1957 error = 1; 1958 if (carpr.carpr_carpdev[0] != '\0' && 1959 (cdev = ifunit(carpr.carpr_carpdev)) == NULL) 1960 return (EINVAL); 1961 if ((error = carp_set_ifp(sc, cdev))) 1962 return (error); 1963 if (sc->sc_state != INIT && carpr.carpr_state != sc->sc_state) { 1964 switch (carpr.carpr_state) { 1965 case BACKUP: 1966 callout_stop(&sc->sc_ad_tmo); 1967 carp_set_state(sc, BACKUP); 1968 carp_setrun(sc, 0); 1969 carp_setroute(sc, RTM_DELETE); 1970 break; 1971 case MASTER: 1972 carp_master_down(sc); 1973 break; 1974 default: 1975 break; 1976 } 1977 } 1978 if (carpr.carpr_vhid > 0) { 1979 if (carpr.carpr_vhid > 255) { 1980 error = EINVAL; 1981 break; 1982 } 1983 if (sc->sc_carpdev) { 1984 struct carp_if *cif; 1985 cif = (struct carp_if *)sc->sc_carpdev->if_carp; 1986 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) 1987 if (vr != sc && 1988 vr->sc_vhid == carpr.carpr_vhid) 1989 return (EINVAL); 1990 } 1991 sc->sc_vhid = carpr.carpr_vhid; 1992 carp_set_enaddr(sc); 1993 carp_set_state(sc, INIT); 1994 error--; 1995 } 1996 if (carpr.carpr_advbase > 0 || carpr.carpr_advskew > 0) { 1997 if (carpr.carpr_advskew > 254) { 1998 error = EINVAL; 1999 break; 2000 } 2001 if (carpr.carpr_advbase > 255) { 2002 error = EINVAL; 2003 break; 2004 } 2005 sc->sc_advbase = carpr.carpr_advbase; 2006 sc->sc_advskew = carpr.carpr_advskew; 2007 error--; 2008 } 2009 memcpy(sc->sc_key, carpr.carpr_key, sizeof(sc->sc_key)); 2010 if (error > 0) 2011 error = EINVAL; 2012 else { 2013 error = 0; 2014 carp_setrun(sc, 0); 2015 } 2016 break; 2017 2018 case SIOCGVH: 2019 memset(&carpr, 0, sizeof(carpr)); 2020 if (sc->sc_carpdev != NULL) 2021 strlcpy(carpr.carpr_carpdev, sc->sc_carpdev->if_xname, 2022 IFNAMSIZ); 2023 carpr.carpr_state = sc->sc_state; 2024 carpr.carpr_vhid = sc->sc_vhid; 2025 carpr.carpr_advbase = sc->sc_advbase; 2026 carpr.carpr_advskew = sc->sc_advskew; 2027 2028 if ((l != NULL) && (error = kauth_authorize_network(l->l_cred, 2029 KAUTH_NETWORK_INTERFACE, 2030 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 2031 NULL)) == 0) 2032 memcpy(carpr.carpr_key, sc->sc_key, 2033 sizeof(carpr.carpr_key)); 2034 error = copyout(&carpr, ifr->ifr_data, sizeof(carpr)); 2035 break; 2036 2037 case SIOCADDMULTI: 2038 error = carp_ether_addmulti(sc, ifr); 2039 break; 2040 2041 case SIOCDELMULTI: 2042 error = carp_ether_delmulti(sc, ifr); 2043 break; 2044 2045 case SIOCSIFCAP: 2046 if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET) 2047 error = 0; 2048 break; 2049 2050 default: 2051 error = ether_ioctl(ifp, cmd, data); 2052 } 2053 2054 carp_hmac_prepare(sc); 2055 return (error); 2056 } 2057 2058 2059 /* 2060 * Start output on carp interface. This function should never be called. 2061 */ 2062 void 2063 carp_start(struct ifnet *ifp) 2064 { 2065 #ifdef DEBUG 2066 printf("%s: start called\n", ifp->if_xname); 2067 #endif 2068 } 2069 2070 int 2071 carp_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa, 2072 struct rtentry *rt) 2073 { 2074 struct carp_softc *sc = ((struct carp_softc *)ifp->if_softc); 2075 2076 if (sc->sc_carpdev != NULL && sc->sc_state == MASTER) { 2077 return (sc->sc_carpdev->if_output(ifp, m, sa, rt)); 2078 } else { 2079 m_freem(m); 2080 return (ENETUNREACH); 2081 } 2082 } 2083 2084 void 2085 carp_set_state(struct carp_softc *sc, int state) 2086 { 2087 static const char *carp_states[] = { CARP_STATES }; 2088 if (sc->sc_state == state) 2089 return; 2090 2091 CARP_LOG(sc, ("state transition from: %s -> to: %s", carp_states[sc->sc_state], carp_states[state])); 2092 2093 sc->sc_state = state; 2094 switch (state) { 2095 case BACKUP: 2096 sc->sc_if.if_link_state = LINK_STATE_DOWN; 2097 break; 2098 case MASTER: 2099 sc->sc_if.if_link_state = LINK_STATE_UP; 2100 break; 2101 default: 2102 sc->sc_if.if_link_state = LINK_STATE_UNKNOWN; 2103 break; 2104 } 2105 rt_ifmsg(&sc->sc_if); 2106 } 2107 2108 void 2109 carp_carpdev_state(void *v) 2110 { 2111 struct carp_if *cif; 2112 struct carp_softc *sc; 2113 struct ifnet *ifp = v; 2114 2115 if (ifp->if_type == IFT_CARP) 2116 return; 2117 2118 cif = (struct carp_if *)ifp->if_carp; 2119 2120 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list) { 2121 int suppressed = sc->sc_suppress; 2122 2123 if (sc->sc_carpdev->if_link_state == LINK_STATE_DOWN || 2124 !(sc->sc_carpdev->if_flags & IFF_UP)) { 2125 sc->sc_if.if_flags &= ~IFF_RUNNING; 2126 callout_stop(&sc->sc_ad_tmo); 2127 callout_stop(&sc->sc_md_tmo); 2128 callout_stop(&sc->sc_md6_tmo); 2129 carp_set_state(sc, INIT); 2130 sc->sc_suppress = 1; 2131 carp_setrun(sc, 0); 2132 if (!suppressed) { 2133 carp_suppress_preempt++; 2134 if (carp_suppress_preempt == 1) 2135 carp_send_ad_all(); 2136 } 2137 } else { 2138 carp_set_state(sc, INIT); 2139 sc->sc_suppress = 0; 2140 carp_setrun(sc, 0); 2141 if (suppressed) 2142 carp_suppress_preempt--; 2143 } 2144 } 2145 } 2146 2147 int 2148 carp_ether_addmulti(struct carp_softc *sc, struct ifreq *ifr) 2149 { 2150 const struct sockaddr *sa = ifreq_getaddr(SIOCADDMULTI, ifr); 2151 struct ifnet *ifp; 2152 struct carp_mc_entry *mc; 2153 u_int8_t addrlo[ETHER_ADDR_LEN], addrhi[ETHER_ADDR_LEN]; 2154 int error; 2155 2156 ifp = sc->sc_carpdev; 2157 if (ifp == NULL) 2158 return (EINVAL); 2159 2160 error = ether_addmulti(sa, &sc->sc_ac); 2161 if (error != ENETRESET) 2162 return (error); 2163 2164 /* 2165 * This is new multicast address. We have to tell parent 2166 * about it. Also, remember this multicast address so that 2167 * we can delete them on unconfigure. 2168 */ 2169 mc = malloc(sizeof(struct carp_mc_entry), M_DEVBUF, M_NOWAIT); 2170 if (mc == NULL) { 2171 error = ENOMEM; 2172 goto alloc_failed; 2173 } 2174 2175 /* 2176 * As ether_addmulti() returns ENETRESET, following two 2177 * statement shouldn't fail. 2178 */ 2179 (void)ether_multiaddr(sa, addrlo, addrhi); 2180 ETHER_LOOKUP_MULTI(addrlo, addrhi, &sc->sc_ac, mc->mc_enm); 2181 memcpy(&mc->mc_addr, sa, sa->sa_len); 2182 LIST_INSERT_HEAD(&sc->carp_mc_listhead, mc, mc_entries); 2183 2184 error = if_mcast_op(ifp, SIOCADDMULTI, sa); 2185 if (error != 0) 2186 goto ioctl_failed; 2187 2188 return (error); 2189 2190 ioctl_failed: 2191 LIST_REMOVE(mc, mc_entries); 2192 free(mc, M_DEVBUF); 2193 alloc_failed: 2194 (void)ether_delmulti(sa, &sc->sc_ac); 2195 2196 return (error); 2197 } 2198 2199 int 2200 carp_ether_delmulti(struct carp_softc *sc, struct ifreq *ifr) 2201 { 2202 const struct sockaddr *sa = ifreq_getaddr(SIOCDELMULTI, ifr); 2203 struct ifnet *ifp; 2204 struct ether_multi *enm; 2205 struct carp_mc_entry *mc; 2206 u_int8_t addrlo[ETHER_ADDR_LEN], addrhi[ETHER_ADDR_LEN]; 2207 int error; 2208 2209 ifp = sc->sc_carpdev; 2210 if (ifp == NULL) 2211 return (EINVAL); 2212 2213 /* 2214 * Find a key to lookup carp_mc_entry. We have to do this 2215 * before calling ether_delmulti for obvious reason. 2216 */ 2217 if ((error = ether_multiaddr(sa, addrlo, addrhi)) != 0) 2218 return (error); 2219 ETHER_LOOKUP_MULTI(addrlo, addrhi, &sc->sc_ac, enm); 2220 if (enm == NULL) 2221 return (EINVAL); 2222 2223 LIST_FOREACH(mc, &sc->carp_mc_listhead, mc_entries) 2224 if (mc->mc_enm == enm) 2225 break; 2226 2227 /* We won't delete entries we didn't add */ 2228 if (mc == NULL) 2229 return (EINVAL); 2230 2231 error = ether_delmulti(sa, &sc->sc_ac); 2232 if (error != ENETRESET) 2233 return (error); 2234 2235 /* We no longer use this multicast address. Tell parent so. */ 2236 error = if_mcast_op(ifp, SIOCDELMULTI, sa); 2237 if (error == 0) { 2238 /* And forget about this address. */ 2239 LIST_REMOVE(mc, mc_entries); 2240 free(mc, M_DEVBUF); 2241 } else 2242 (void)ether_addmulti(sa, &sc->sc_ac); 2243 return (error); 2244 } 2245 2246 /* 2247 * Delete any multicast address we have asked to add from parent 2248 * interface. Called when the carp is being unconfigured. 2249 */ 2250 void 2251 carp_ether_purgemulti(struct carp_softc *sc) 2252 { 2253 struct ifnet *ifp = sc->sc_carpdev; /* Parent. */ 2254 struct carp_mc_entry *mc; 2255 2256 if (ifp == NULL) 2257 return; 2258 2259 while ((mc = LIST_FIRST(&sc->carp_mc_listhead)) != NULL) { 2260 (void)if_mcast_op(ifp, SIOCDELMULTI, sstosa(&mc->mc_addr)); 2261 LIST_REMOVE(mc, mc_entries); 2262 free(mc, M_DEVBUF); 2263 } 2264 } 2265 2266 static int 2267 sysctl_net_inet_carp_stats(SYSCTLFN_ARGS) 2268 { 2269 2270 return (NETSTAT_SYSCTL(carpstat_percpu, CARP_NSTATS)); 2271 } 2272 2273 void 2274 carp_init(void) 2275 { 2276 2277 sysctl_net_inet_carp_setup(NULL); 2278 #ifdef MBUFTRACE 2279 MOWNER_ATTACH(&carp_proto_mowner_rx); 2280 MOWNER_ATTACH(&carp_proto_mowner_tx); 2281 MOWNER_ATTACH(&carp_proto6_mowner_rx); 2282 MOWNER_ATTACH(&carp_proto6_mowner_tx); 2283 #endif 2284 } 2285 2286 static void 2287 sysctl_net_inet_carp_setup(struct sysctllog **clog) 2288 { 2289 2290 sysctl_createv(clog, 0, NULL, NULL, 2291 CTLFLAG_PERMANENT, 2292 CTLTYPE_NODE, "inet", NULL, 2293 NULL, 0, NULL, 0, 2294 CTL_NET, PF_INET, CTL_EOL); 2295 sysctl_createv(clog, 0, NULL, NULL, 2296 CTLFLAG_PERMANENT, 2297 CTLTYPE_NODE, "carp", 2298 SYSCTL_DESCR("CARP related settings"), 2299 NULL, 0, NULL, 0, 2300 CTL_NET, PF_INET, IPPROTO_CARP, CTL_EOL); 2301 2302 sysctl_createv(clog, 0, NULL, NULL, 2303 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2304 CTLTYPE_INT, "preempt", 2305 SYSCTL_DESCR("Enable CARP Preempt"), 2306 NULL, 0, &carp_opts[CARPCTL_PREEMPT], 0, 2307 CTL_NET, PF_INET, IPPROTO_CARP, 2308 CTL_CREATE, CTL_EOL); 2309 sysctl_createv(clog, 0, NULL, NULL, 2310 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2311 CTLTYPE_INT, "arpbalance", 2312 SYSCTL_DESCR("Enable ARP balancing"), 2313 NULL, 0, &carp_opts[CARPCTL_ARPBALANCE], 0, 2314 CTL_NET, PF_INET, IPPROTO_CARP, 2315 CTL_CREATE, CTL_EOL); 2316 sysctl_createv(clog, 0, NULL, NULL, 2317 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2318 CTLTYPE_INT, "allow", 2319 SYSCTL_DESCR("Enable CARP"), 2320 NULL, 0, &carp_opts[CARPCTL_ALLOW], 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, "log", 2326 SYSCTL_DESCR("CARP logging"), 2327 NULL, 0, &carp_opts[CARPCTL_LOG], 0, 2328 CTL_NET, PF_INET, IPPROTO_CARP, 2329 CTL_CREATE, CTL_EOL); 2330 sysctl_createv(clog, 0, NULL, NULL, 2331 CTLFLAG_PERMANENT, 2332 CTLTYPE_STRUCT, "stats", 2333 SYSCTL_DESCR("CARP statistics"), 2334 sysctl_net_inet_carp_stats, 0, NULL, 0, 2335 CTL_NET, PF_INET, IPPROTO_CARP, CARPCTL_STATS, 2336 CTL_EOL); 2337 } 2338