1 /* $NetBSD: ip_carp.c,v 1.103 2019/06/01 19:13:41 joerg 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 #ifdef _KERNEL_OPT 31 #include "opt_inet.h" 32 #include "opt_mbuftrace.h" 33 #endif 34 35 #include <sys/cdefs.h> 36 __KERNEL_RCSID(0, "$NetBSD: ip_carp.c,v 1.103 2019/06/01 19:13:41 joerg Exp $"); 37 38 /* 39 * TODO: 40 * - iface reconfigure 41 * - support for hardware checksum calculations; 42 * 43 */ 44 45 #include <sys/param.h> 46 #include <sys/proc.h> 47 #include <sys/mbuf.h> 48 #include <sys/socket.h> 49 #include <sys/socketvar.h> 50 #include <sys/callout.h> 51 #include <sys/ioctl.h> 52 #include <sys/errno.h> 53 #include <sys/device.h> 54 #include <sys/time.h> 55 #include <sys/kernel.h> 56 #include <sys/kauth.h> 57 #include <sys/sysctl.h> 58 #include <sys/ucred.h> 59 #include <sys/syslog.h> 60 #include <sys/acct.h> 61 #include <sys/cprng.h> 62 #include <sys/cpu.h> 63 #include <sys/pserialize.h> 64 #include <sys/psref.h> 65 66 #include <net/if.h> 67 #include <net/pfil.h> 68 #include <net/if_types.h> 69 #include <net/if_ether.h> 70 #include <net/route.h> 71 #include <net/netisr.h> 72 #include <net/net_stats.h> 73 #include <netinet/if_inarp.h> 74 #include <netinet/wqinput.h> 75 76 #if NFDDI > 0 77 #include <net/if_fddi.h> 78 #endif 79 #if NTOKEN > 0 80 #include <net/if_token.h> 81 #endif 82 83 #ifdef INET 84 #include <netinet/in.h> 85 #include <netinet/in_systm.h> 86 #include <netinet/in_var.h> 87 #include <netinet/ip.h> 88 #include <netinet/ip_var.h> 89 90 #include <net/if_dl.h> 91 #endif 92 93 #ifdef INET6 94 #include <netinet/icmp6.h> 95 #include <netinet/ip6.h> 96 #include <netinet6/ip6_var.h> 97 #include <netinet6/nd6.h> 98 #include <netinet6/scope6_var.h> 99 #include <netinet6/in6_var.h> 100 #endif 101 102 #include <net/bpf.h> 103 104 #include <sys/sha1.h> 105 106 #include <netinet/ip_carp.h> 107 108 #include "ioconf.h" 109 110 struct carp_mc_entry { 111 LIST_ENTRY(carp_mc_entry) mc_entries; 112 union { 113 struct ether_multi *mcu_enm; 114 } mc_u; 115 struct sockaddr_storage mc_addr; 116 }; 117 #define mc_enm mc_u.mcu_enm 118 119 struct carp_softc { 120 struct ethercom sc_ac; 121 #define sc_if sc_ac.ec_if 122 #define sc_carpdev sc_ac.ec_if.if_carpdev 123 int ah_cookie; 124 int lh_cookie; 125 struct ip_moptions sc_imo; 126 #ifdef INET6 127 struct ip6_moptions sc_im6o; 128 #endif /* INET6 */ 129 TAILQ_ENTRY(carp_softc) sc_list; 130 131 enum { INIT = 0, BACKUP, MASTER } sc_state; 132 133 int sc_suppress; 134 int sc_bow_out; 135 136 int sc_sendad_errors; 137 #define CARP_SENDAD_MAX_ERRORS 3 138 int sc_sendad_success; 139 #define CARP_SENDAD_MIN_SUCCESS 3 140 141 int sc_vhid; 142 int sc_advskew; 143 int sc_naddrs; 144 int sc_naddrs6; 145 int sc_advbase; /* seconds */ 146 int sc_init_counter; 147 u_int64_t sc_counter; 148 149 /* authentication */ 150 #define CARP_HMAC_PAD 64 151 unsigned char sc_key[CARP_KEY_LEN]; 152 unsigned char sc_pad[CARP_HMAC_PAD]; 153 SHA1_CTX sc_sha1; 154 u_int32_t sc_hashkey[2]; 155 156 struct callout sc_ad_tmo; /* advertisement timeout */ 157 struct callout sc_md_tmo; /* master down timeout */ 158 struct callout sc_md6_tmo; /* master down timeout */ 159 160 LIST_HEAD(__carp_mchead, carp_mc_entry) carp_mc_listhead; 161 }; 162 163 int carp_suppress_preempt = 0; 164 static int carp_opts[CARPCTL_MAXID] = { 0, 1, 0, 0, 0 }; /* XXX for now */ 165 166 static percpu_t *carpstat_percpu; 167 168 #define CARP_STATINC(x) _NET_STATINC(carpstat_percpu, x) 169 170 #ifdef MBUFTRACE 171 static struct mowner carp_proto_mowner_rx = MOWNER_INIT("carp", "rx"); 172 static struct mowner carp_proto_mowner_tx = MOWNER_INIT("carp", "tx"); 173 static struct mowner carp_proto6_mowner_rx = MOWNER_INIT("carp6", "rx"); 174 static struct mowner carp_proto6_mowner_tx = MOWNER_INIT("carp6", "tx"); 175 #endif 176 177 struct carp_if { 178 TAILQ_HEAD(, carp_softc) vhif_vrs; 179 int vhif_nvrs; 180 181 struct ifnet *vhif_ifp; 182 }; 183 184 #define CARP_LOG(sc, s) \ 185 if (carp_opts[CARPCTL_LOG]) { \ 186 if (sc) \ 187 log(LOG_INFO, "%s: ", \ 188 (sc)->sc_if.if_xname); \ 189 else \ 190 log(LOG_INFO, "carp: "); \ 191 addlog s; \ 192 addlog("\n"); \ 193 } 194 195 static void carp_hmac_prepare(struct carp_softc *); 196 static void carp_hmac_generate(struct carp_softc *, u_int32_t *, 197 unsigned char *); 198 static int carp_hmac_verify(struct carp_softc *, u_int32_t *, 199 unsigned char *); 200 static void carp_setroute(struct carp_softc *, int); 201 static void carp_proto_input_c(struct mbuf *, struct carp_header *, 202 sa_family_t); 203 static void carpdetach(struct carp_softc *); 204 static void carp_prepare_ad(struct mbuf *, struct carp_softc *, 205 struct carp_header *); 206 static void carp_send_ad_all(void); 207 static void carp_send_ad(void *); 208 static void carp_send_arp(struct carp_softc *); 209 static void carp_master_down(void *); 210 static int carp_ioctl(struct ifnet *, u_long, void *); 211 static void carp_start(struct ifnet *); 212 static void carp_setrun(struct carp_softc *, sa_family_t); 213 static void carp_set_state(struct carp_softc *, int); 214 static int carp_addrcount(struct carp_if *, struct in_ifaddr *, int); 215 enum { CARP_COUNT_MASTER, CARP_COUNT_RUNNING }; 216 217 static void carp_multicast_cleanup(struct carp_softc *); 218 static int carp_set_ifp(struct carp_softc *, struct ifnet *); 219 static void carp_set_enaddr(struct carp_softc *); 220 #if 0 221 static void carp_addr_updated(void *); 222 #endif 223 static u_int32_t carp_hash(struct carp_softc *, u_char *); 224 static int carp_set_addr(struct carp_softc *, struct sockaddr_in *); 225 static int carp_join_multicast(struct carp_softc *); 226 #ifdef INET6 227 static void carp_send_na(struct carp_softc *); 228 static int carp_set_addr6(struct carp_softc *, struct sockaddr_in6 *); 229 static int carp_join_multicast6(struct carp_softc *); 230 #endif 231 static int carp_clone_create(struct if_clone *, int); 232 static int carp_clone_destroy(struct ifnet *); 233 static int carp_ether_addmulti(struct carp_softc *, struct ifreq *); 234 static int carp_ether_delmulti(struct carp_softc *, struct ifreq *); 235 static void carp_ether_purgemulti(struct carp_softc *); 236 237 static void sysctl_net_inet_carp_setup(struct sysctllog **); 238 239 /* workqueue-based pr_input */ 240 static struct wqinput *carp_wqinput; 241 static void _carp_proto_input(struct mbuf *, int, int); 242 #ifdef INET6 243 static struct wqinput *carp6_wqinput; 244 static void _carp6_proto_input(struct mbuf *, int, int); 245 #endif 246 247 struct if_clone carp_cloner = 248 IF_CLONE_INITIALIZER("carp", carp_clone_create, carp_clone_destroy); 249 250 static __inline u_int16_t 251 carp_cksum(struct mbuf *m, int len) 252 { 253 return (in_cksum(m, len)); 254 } 255 256 #ifdef INET6 257 static __inline u_int16_t 258 carp6_cksum(struct mbuf *m, uint32_t off, uint32_t len) 259 { 260 return (in6_cksum(m, IPPROTO_CARP, off, len)); 261 } 262 #endif 263 264 static void 265 carp_hmac_prepare(struct carp_softc *sc) 266 { 267 u_int8_t carp_version = CARP_VERSION, type = CARP_ADVERTISEMENT; 268 u_int8_t vhid = sc->sc_vhid & 0xff; 269 SHA1_CTX sha1ctx; 270 u_int32_t kmd[5]; 271 struct ifaddr *ifa; 272 int i, found; 273 struct in_addr last, cur, in; 274 #ifdef INET6 275 struct in6_addr last6, cur6, in6; 276 #endif /* INET6 */ 277 278 /* compute ipad from key */ 279 memset(sc->sc_pad, 0, sizeof(sc->sc_pad)); 280 memcpy(sc->sc_pad, sc->sc_key, sizeof(sc->sc_key)); 281 for (i = 0; i < sizeof(sc->sc_pad); i++) 282 sc->sc_pad[i] ^= 0x36; 283 284 /* precompute first part of inner hash */ 285 SHA1Init(&sc->sc_sha1); 286 SHA1Update(&sc->sc_sha1, sc->sc_pad, sizeof(sc->sc_pad)); 287 SHA1Update(&sc->sc_sha1, (void *)&carp_version, sizeof(carp_version)); 288 SHA1Update(&sc->sc_sha1, (void *)&type, sizeof(type)); 289 290 /* generate a key for the arpbalance hash, before the vhid is hashed */ 291 memcpy(&sha1ctx, &sc->sc_sha1, sizeof(sha1ctx)); 292 SHA1Final((unsigned char *)kmd, &sha1ctx); 293 sc->sc_hashkey[0] = kmd[0] ^ kmd[1]; 294 sc->sc_hashkey[1] = kmd[2] ^ kmd[3]; 295 296 /* the rest of the precomputation */ 297 SHA1Update(&sc->sc_sha1, (void *)&vhid, sizeof(vhid)); 298 299 /* Hash the addresses from smallest to largest, not interface order */ 300 #ifdef INET 301 cur.s_addr = 0; 302 do { 303 int s; 304 found = 0; 305 last = cur; 306 cur.s_addr = 0xffffffff; 307 s = pserialize_read_enter(); 308 IFADDR_READER_FOREACH(ifa, &sc->sc_if) { 309 in.s_addr = ifatoia(ifa)->ia_addr.sin_addr.s_addr; 310 if (ifa->ifa_addr->sa_family == AF_INET && 311 ntohl(in.s_addr) > ntohl(last.s_addr) && 312 ntohl(in.s_addr) < ntohl(cur.s_addr)) { 313 cur.s_addr = in.s_addr; 314 found++; 315 } 316 } 317 pserialize_read_exit(s); 318 if (found) 319 SHA1Update(&sc->sc_sha1, (void *)&cur, sizeof(cur)); 320 } while (found); 321 #endif /* INET */ 322 323 #ifdef INET6 324 memset(&cur6, 0x00, sizeof(cur6)); 325 do { 326 int s; 327 found = 0; 328 last6 = cur6; 329 memset(&cur6, 0xff, sizeof(cur6)); 330 s = pserialize_read_enter(); 331 IFADDR_READER_FOREACH(ifa, &sc->sc_if) { 332 in6 = ifatoia6(ifa)->ia_addr.sin6_addr; 333 if (IN6_IS_ADDR_LINKLOCAL(&in6)) 334 in6.s6_addr16[1] = 0; 335 if (ifa->ifa_addr->sa_family == AF_INET6 && 336 memcmp(&in6, &last6, sizeof(in6)) > 0 && 337 memcmp(&in6, &cur6, sizeof(in6)) < 0) { 338 cur6 = in6; 339 found++; 340 } 341 } 342 pserialize_read_exit(s); 343 if (found) 344 SHA1Update(&sc->sc_sha1, (void *)&cur6, sizeof(cur6)); 345 } while (found); 346 #endif /* INET6 */ 347 348 /* convert ipad to opad */ 349 for (i = 0; i < sizeof(sc->sc_pad); i++) 350 sc->sc_pad[i] ^= 0x36 ^ 0x5c; 351 } 352 353 static void 354 carp_hmac_generate(struct carp_softc *sc, u_int32_t counter[2], 355 unsigned char md[20]) 356 { 357 SHA1_CTX sha1ctx; 358 359 /* fetch first half of inner hash */ 360 memcpy(&sha1ctx, &sc->sc_sha1, sizeof(sha1ctx)); 361 362 SHA1Update(&sha1ctx, (void *)counter, sizeof(sc->sc_counter)); 363 SHA1Final(md, &sha1ctx); 364 365 /* outer hash */ 366 SHA1Init(&sha1ctx); 367 SHA1Update(&sha1ctx, sc->sc_pad, sizeof(sc->sc_pad)); 368 SHA1Update(&sha1ctx, md, 20); 369 SHA1Final(md, &sha1ctx); 370 } 371 372 static int 373 carp_hmac_verify(struct carp_softc *sc, u_int32_t counter[2], 374 unsigned char md[20]) 375 { 376 unsigned char md2[20]; 377 378 carp_hmac_generate(sc, counter, md2); 379 380 return (memcmp(md, md2, sizeof(md2))); 381 } 382 383 static void 384 carp_setroute(struct carp_softc *sc, int cmd) 385 { 386 struct ifaddr *ifa; 387 int s, bound; 388 389 KERNEL_LOCK(1, NULL); 390 bound = curlwp_bind(); 391 s = pserialize_read_enter(); 392 IFADDR_READER_FOREACH(ifa, &sc->sc_if) { 393 struct psref psref; 394 ifa_acquire(ifa, &psref); 395 pserialize_read_exit(s); 396 397 switch (ifa->ifa_addr->sa_family) { 398 case AF_INET: { 399 int count = 0; 400 struct rtentry *rt; 401 int hr_otherif, nr_ourif; 402 403 /* 404 * Avoid screwing with the routes if there are other 405 * carp interfaces which are master and have the same 406 * address. 407 */ 408 if (sc->sc_carpdev != NULL && 409 sc->sc_carpdev->if_carp != NULL) { 410 count = carp_addrcount( 411 (struct carp_if *)sc->sc_carpdev->if_carp, 412 ifatoia(ifa), CARP_COUNT_MASTER); 413 if ((cmd == RTM_ADD && count != 1) || 414 (cmd == RTM_DELETE && count != 0)) 415 goto next; 416 } 417 418 /* Remove the existing host route, if any */ 419 rtrequest(RTM_DELETE, ifa->ifa_addr, 420 ifa->ifa_addr, ifa->ifa_netmask, 421 RTF_HOST, NULL); 422 423 rt = NULL; 424 (void)rtrequest(RTM_GET, ifa->ifa_addr, ifa->ifa_addr, 425 ifa->ifa_netmask, RTF_HOST, &rt); 426 hr_otherif = (rt && rt->rt_ifp != &sc->sc_if && 427 (rt->rt_flags & RTF_CONNECTED)); 428 if (rt != NULL) { 429 rt_unref(rt); 430 rt = NULL; 431 } 432 433 /* Check for a network route on our interface */ 434 435 rt = NULL; 436 (void)rtrequest(RTM_GET, ifa->ifa_addr, ifa->ifa_addr, 437 ifa->ifa_netmask, 0, &rt); 438 nr_ourif = (rt && rt->rt_ifp == &sc->sc_if); 439 440 switch (cmd) { 441 case RTM_ADD: 442 if (hr_otherif) { 443 ifa->ifa_rtrequest = NULL; 444 ifa->ifa_flags &= ~RTF_CONNECTED; 445 446 rtrequest(RTM_ADD, ifa->ifa_addr, 447 ifa->ifa_addr, ifa->ifa_netmask, 448 RTF_UP | RTF_HOST, NULL); 449 } 450 if (!hr_otherif || nr_ourif || !rt) { 451 if (nr_ourif && 452 (rt->rt_flags & RTF_CONNECTED) == 0) 453 rtrequest(RTM_DELETE, 454 ifa->ifa_addr, 455 ifa->ifa_addr, 456 ifa->ifa_netmask, 0, NULL); 457 458 ifa->ifa_rtrequest = arp_rtrequest; 459 ifa->ifa_flags |= RTF_CONNECTED; 460 461 if (rtrequest(RTM_ADD, ifa->ifa_addr, 462 ifa->ifa_addr, ifa->ifa_netmask, 0, 463 NULL) == 0) 464 ifa->ifa_flags |= IFA_ROUTE; 465 } 466 break; 467 case RTM_DELETE: 468 break; 469 default: 470 break; 471 } 472 if (rt != NULL) { 473 rt_unref(rt); 474 rt = NULL; 475 } 476 break; 477 } 478 479 #ifdef INET6 480 case AF_INET6: 481 if (cmd == RTM_ADD) 482 in6_ifaddlocal(ifa); 483 else 484 in6_ifremlocal(ifa); 485 break; 486 #endif /* INET6 */ 487 default: 488 break; 489 } 490 next: 491 s = pserialize_read_enter(); 492 ifa_release(ifa, &psref); 493 } 494 pserialize_read_exit(s); 495 curlwp_bindx(bound); 496 KERNEL_UNLOCK_ONE(NULL); 497 } 498 499 /* 500 * process input packet. 501 * we have rearranged checks order compared to the rfc, 502 * but it seems more efficient this way or not possible otherwise. 503 */ 504 static void 505 _carp_proto_input(struct mbuf *m, int hlen, int proto) 506 { 507 struct ip *ip = mtod(m, struct ip *); 508 struct carp_softc *sc = NULL; 509 struct carp_header *ch; 510 int iplen, len; 511 struct ifnet *rcvif; 512 513 CARP_STATINC(CARP_STAT_IPACKETS); 514 MCLAIM(m, &carp_proto_mowner_rx); 515 516 if (!carp_opts[CARPCTL_ALLOW]) { 517 m_freem(m); 518 return; 519 } 520 521 rcvif = m_get_rcvif_NOMPSAFE(m); 522 /* check if received on a valid carp interface */ 523 if (rcvif->if_type != IFT_CARP) { 524 CARP_STATINC(CARP_STAT_BADIF); 525 CARP_LOG(sc, ("packet received on non-carp interface: %s", 526 rcvif->if_xname)); 527 m_freem(m); 528 return; 529 } 530 531 /* verify that the IP TTL is 255. */ 532 if (ip->ip_ttl != CARP_DFLTTL) { 533 CARP_STATINC(CARP_STAT_BADTTL); 534 CARP_LOG(sc, ("received ttl %d != %d on %s", ip->ip_ttl, 535 CARP_DFLTTL, rcvif->if_xname)); 536 m_freem(m); 537 return; 538 } 539 540 /* 541 * verify that the received packet length is 542 * equal to the CARP header 543 */ 544 iplen = ip->ip_hl << 2; 545 len = iplen + sizeof(*ch); 546 if (len > m->m_pkthdr.len) { 547 CARP_STATINC(CARP_STAT_BADLEN); 548 CARP_LOG(sc, ("packet too short %d on %s", m->m_pkthdr.len, 549 rcvif->if_xname)); 550 m_freem(m); 551 return; 552 } 553 554 if ((m = m_pullup(m, len)) == NULL) { 555 CARP_STATINC(CARP_STAT_HDROPS); 556 return; 557 } 558 ip = mtod(m, struct ip *); 559 ch = (struct carp_header *)((char *)ip + iplen); 560 /* verify the CARP checksum */ 561 m->m_data += iplen; 562 if (carp_cksum(m, len - iplen)) { 563 CARP_STATINC(CARP_STAT_BADSUM); 564 CARP_LOG(sc, ("checksum failed on %s", 565 rcvif->if_xname)); 566 m_freem(m); 567 return; 568 } 569 m->m_data -= iplen; 570 571 carp_proto_input_c(m, ch, AF_INET); 572 } 573 574 void 575 carp_proto_input(struct mbuf *m, int off, int proto) 576 { 577 578 wqinput_input(carp_wqinput, m, 0, 0); 579 } 580 581 #ifdef INET6 582 static void 583 _carp6_proto_input(struct mbuf *m, int off, int proto) 584 { 585 struct carp_softc *sc = NULL; 586 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 587 struct carp_header *ch; 588 u_int len; 589 struct ifnet *rcvif; 590 591 CARP_STATINC(CARP_STAT_IPACKETS6); 592 MCLAIM(m, &carp_proto6_mowner_rx); 593 594 if (!carp_opts[CARPCTL_ALLOW]) { 595 m_freem(m); 596 return; 597 } 598 599 rcvif = m_get_rcvif_NOMPSAFE(m); 600 601 /* check if received on a valid carp interface */ 602 if (rcvif->if_type != IFT_CARP) { 603 CARP_STATINC(CARP_STAT_BADIF); 604 CARP_LOG(sc, ("packet received on non-carp interface: %s", 605 rcvif->if_xname)); 606 m_freem(m); 607 return; 608 } 609 610 /* verify that the IP TTL is 255 */ 611 if (ip6->ip6_hlim != CARP_DFLTTL) { 612 CARP_STATINC(CARP_STAT_BADTTL); 613 CARP_LOG(sc, ("received ttl %d != %d on %s", ip6->ip6_hlim, 614 CARP_DFLTTL, rcvif->if_xname)); 615 m_freem(m); 616 return; 617 } 618 619 /* verify that we have a complete carp packet */ 620 len = m->m_len; 621 M_REGION_GET(ch, struct carp_header *, m, off, sizeof(*ch)); 622 if (ch == NULL) { 623 CARP_STATINC(CARP_STAT_BADLEN); 624 CARP_LOG(sc, ("packet size %u too small", len)); 625 return; 626 } 627 628 /* verify the CARP checksum */ 629 if (carp6_cksum(m, off, sizeof(*ch))) { 630 CARP_STATINC(CARP_STAT_BADSUM); 631 CARP_LOG(sc, ("checksum failed, on %s", rcvif->if_xname)); 632 m_freem(m); 633 return; 634 } 635 636 carp_proto_input_c(m, ch, AF_INET6); 637 return; 638 } 639 640 int 641 carp6_proto_input(struct mbuf **mp, int *offp, int proto) 642 { 643 644 wqinput_input(carp6_wqinput, *mp, *offp, proto); 645 646 return IPPROTO_DONE; 647 } 648 #endif /* INET6 */ 649 650 static void 651 carp_proto_input_c(struct mbuf *m, struct carp_header *ch, sa_family_t af) 652 { 653 struct carp_softc *sc; 654 u_int64_t tmp_counter; 655 struct timeval sc_tv, ch_tv; 656 657 TAILQ_FOREACH(sc, &((struct carp_if *) 658 m_get_rcvif_NOMPSAFE(m)->if_carpdev->if_carp)->vhif_vrs, sc_list) 659 if (sc->sc_vhid == ch->carp_vhid) 660 break; 661 662 if (!sc || (sc->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) != 663 (IFF_UP|IFF_RUNNING)) { 664 CARP_STATINC(CARP_STAT_BADVHID); 665 m_freem(m); 666 return; 667 } 668 669 /* 670 * Check if our own advertisement was duplicated 671 * from a non simplex interface. 672 * XXX If there is no address on our physical interface 673 * there is no way to distinguish our ads from the ones 674 * another carp host might have sent us. 675 */ 676 if ((sc->sc_carpdev->if_flags & IFF_SIMPLEX) == 0) { 677 struct sockaddr sa; 678 struct ifaddr *ifa; 679 int s; 680 681 memset(&sa, 0, sizeof(sa)); 682 sa.sa_family = af; 683 s = pserialize_read_enter(); 684 ifa = ifaof_ifpforaddr(&sa, sc->sc_carpdev); 685 686 if (ifa && af == AF_INET) { 687 struct ip *ip = mtod(m, struct ip *); 688 if (ip->ip_src.s_addr == 689 ifatoia(ifa)->ia_addr.sin_addr.s_addr) { 690 pserialize_read_exit(s); 691 m_freem(m); 692 return; 693 } 694 } 695 #ifdef INET6 696 if (ifa && af == AF_INET6) { 697 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 698 struct in6_addr in6_src, in6_found; 699 700 in6_src = ip6->ip6_src; 701 in6_found = ifatoia6(ifa)->ia_addr.sin6_addr; 702 if (IN6_IS_ADDR_LINKLOCAL(&in6_src)) 703 in6_src.s6_addr16[1] = 0; 704 if (IN6_IS_ADDR_LINKLOCAL(&in6_found)) 705 in6_found.s6_addr16[1] = 0; 706 if (IN6_ARE_ADDR_EQUAL(&in6_src, &in6_found)) { 707 pserialize_read_exit(s); 708 m_freem(m); 709 return; 710 } 711 } 712 #endif /* INET6 */ 713 pserialize_read_exit(s); 714 } 715 716 nanotime(&sc->sc_if.if_lastchange); 717 sc->sc_if.if_ipackets++; 718 sc->sc_if.if_ibytes += m->m_pkthdr.len; 719 720 /* verify the CARP version. */ 721 if (ch->carp_version != CARP_VERSION) { 722 CARP_STATINC(CARP_STAT_BADVER); 723 sc->sc_if.if_ierrors++; 724 CARP_LOG(sc, ("invalid version %d != %d", 725 ch->carp_version, CARP_VERSION)); 726 m_freem(m); 727 return; 728 } 729 730 /* verify the hash */ 731 if (carp_hmac_verify(sc, ch->carp_counter, ch->carp_md)) { 732 struct ip *ip; 733 char ipbuf[INET_ADDRSTRLEN]; 734 #ifdef INET6 735 struct ip6_hdr *ip6; 736 char ip6buf[INET6_ADDRSTRLEN]; 737 #endif 738 739 CARP_STATINC(CARP_STAT_BADAUTH); 740 sc->sc_if.if_ierrors++; 741 742 switch(af) { 743 case AF_INET: 744 ip = mtod(m, struct ip *); 745 CARP_LOG(sc, ("incorrect hash from %s", 746 IN_PRINT(ipbuf, &ip->ip_src))); 747 break; 748 749 #ifdef INET6 750 case AF_INET6: 751 ip6 = mtod(m, struct ip6_hdr *); 752 CARP_LOG(sc, ("incorrect hash from %s", 753 IN6_PRINT(ip6buf, &ip6->ip6_src))); 754 break; 755 #endif 756 757 default: CARP_LOG(sc, ("incorrect hash")); 758 break; 759 } 760 m_freem(m); 761 return; 762 } 763 764 tmp_counter = ntohl(ch->carp_counter[0]); 765 tmp_counter = tmp_counter<<32; 766 tmp_counter += ntohl(ch->carp_counter[1]); 767 768 /* XXX Replay protection goes here */ 769 770 sc->sc_init_counter = 0; 771 sc->sc_counter = tmp_counter; 772 773 774 sc_tv.tv_sec = sc->sc_advbase; 775 if (carp_suppress_preempt && sc->sc_advskew < 240) 776 sc_tv.tv_usec = 240 * 1000000 / 256; 777 else 778 sc_tv.tv_usec = sc->sc_advskew * 1000000 / 256; 779 ch_tv.tv_sec = ch->carp_advbase; 780 ch_tv.tv_usec = ch->carp_advskew * 1000000 / 256; 781 782 switch (sc->sc_state) { 783 case INIT: 784 break; 785 case MASTER: 786 /* 787 * If we receive an advertisement from a backup who's going to 788 * be more frequent than us, go into BACKUP state. 789 */ 790 if (timercmp(&sc_tv, &ch_tv, >) || 791 timercmp(&sc_tv, &ch_tv, ==)) { 792 callout_stop(&sc->sc_ad_tmo); 793 CARP_LOG(sc, ("MASTER -> BACKUP (more frequent advertisement received)")); 794 carp_set_state(sc, BACKUP); 795 carp_setrun(sc, 0); 796 carp_setroute(sc, RTM_DELETE); 797 } 798 break; 799 case BACKUP: 800 /* 801 * If we're pre-empting masters who advertise slower than us, 802 * and this one claims to be slower, treat him as down. 803 */ 804 if (carp_opts[CARPCTL_PREEMPT] && timercmp(&sc_tv, &ch_tv, <)) { 805 CARP_LOG(sc, ("BACKUP -> MASTER (preempting a slower master)")); 806 carp_master_down(sc); 807 break; 808 } 809 810 /* 811 * If the master is going to advertise at such a low frequency 812 * that he's guaranteed to time out, we'd might as well just 813 * treat him as timed out now. 814 */ 815 sc_tv.tv_sec = sc->sc_advbase * 3; 816 if (timercmp(&sc_tv, &ch_tv, <)) { 817 CARP_LOG(sc, ("BACKUP -> MASTER (master timed out)")); 818 carp_master_down(sc); 819 break; 820 } 821 822 /* 823 * Otherwise, we reset the counter and wait for the next 824 * advertisement. 825 */ 826 carp_setrun(sc, af); 827 break; 828 } 829 830 m_freem(m); 831 return; 832 } 833 834 /* 835 * Interface side of the CARP implementation. 836 */ 837 838 /* ARGSUSED */ 839 void 840 carpattach(int n) 841 { 842 if_clone_attach(&carp_cloner); 843 844 carpstat_percpu = percpu_alloc(sizeof(uint64_t) * CARP_NSTATS); 845 } 846 847 static int 848 carp_clone_create(struct if_clone *ifc, int unit) 849 { 850 extern int ifqmaxlen; 851 struct carp_softc *sc; 852 struct ifnet *ifp; 853 int rv; 854 855 sc = malloc(sizeof(*sc), M_DEVBUF, M_NOWAIT|M_ZERO); 856 if (!sc) 857 return (ENOMEM); 858 859 sc->sc_suppress = 0; 860 sc->sc_advbase = CARP_DFLTINTV; 861 sc->sc_vhid = -1; /* required setting */ 862 sc->sc_advskew = 0; 863 sc->sc_init_counter = 1; 864 sc->sc_naddrs = sc->sc_naddrs6 = 0; 865 #ifdef INET6 866 sc->sc_im6o.im6o_multicast_hlim = CARP_DFLTTL; 867 #endif /* INET6 */ 868 869 callout_init(&sc->sc_ad_tmo, 0); 870 callout_init(&sc->sc_md_tmo, 0); 871 callout_init(&sc->sc_md6_tmo, 0); 872 873 callout_setfunc(&sc->sc_ad_tmo, carp_send_ad, sc); 874 callout_setfunc(&sc->sc_md_tmo, carp_master_down, sc); 875 callout_setfunc(&sc->sc_md6_tmo, carp_master_down, sc); 876 877 LIST_INIT(&sc->carp_mc_listhead); 878 ifp = &sc->sc_if; 879 ifp->if_softc = sc; 880 snprintf(ifp->if_xname, sizeof ifp->if_xname, "%s%d", ifc->ifc_name, 881 unit); 882 ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST; 883 ifp->if_ioctl = carp_ioctl; 884 ifp->if_start = carp_start; 885 IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen); 886 IFQ_SET_READY(&ifp->if_snd); 887 rv = if_initialize(ifp); 888 if (rv != 0) { 889 callout_destroy(&sc->sc_ad_tmo); 890 callout_destroy(&sc->sc_md_tmo); 891 callout_destroy(&sc->sc_md6_tmo); 892 free(ifp->if_softc, M_DEVBUF); 893 894 return rv; 895 } 896 ether_ifattach(ifp, NULL); 897 carp_set_enaddr(sc); 898 /* Overwrite ethernet defaults */ 899 ifp->if_type = IFT_CARP; 900 ifp->if_output = carp_output; 901 if_register(ifp); 902 903 return (0); 904 } 905 906 static int 907 carp_clone_destroy(struct ifnet *ifp) 908 { 909 struct carp_softc *sc = ifp->if_softc; 910 911 carpdetach(ifp->if_softc); 912 ether_ifdetach(ifp); 913 if_detach(ifp); 914 callout_destroy(&sc->sc_ad_tmo); 915 callout_destroy(&sc->sc_md_tmo); 916 callout_destroy(&sc->sc_md6_tmo); 917 free(ifp->if_softc, M_DEVBUF); 918 919 return (0); 920 } 921 922 static void 923 carpdetach(struct carp_softc *sc) 924 { 925 struct carp_if *cif; 926 int s; 927 928 callout_stop(&sc->sc_ad_tmo); 929 callout_stop(&sc->sc_md_tmo); 930 callout_stop(&sc->sc_md6_tmo); 931 932 if (sc->sc_suppress) 933 carp_suppress_preempt--; 934 sc->sc_suppress = 0; 935 936 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) 937 carp_suppress_preempt--; 938 sc->sc_sendad_errors = 0; 939 940 carp_set_state(sc, INIT); 941 sc->sc_if.if_flags &= ~IFF_UP; 942 carp_setrun(sc, 0); 943 carp_multicast_cleanup(sc); 944 945 KERNEL_LOCK(1, NULL); 946 s = splnet(); 947 if (sc->sc_carpdev != NULL) { 948 /* XXX linkstatehook removal */ 949 cif = (struct carp_if *)sc->sc_carpdev->if_carp; 950 TAILQ_REMOVE(&cif->vhif_vrs, sc, sc_list); 951 if (!--cif->vhif_nvrs) { 952 ifpromisc(sc->sc_carpdev, 0); 953 sc->sc_carpdev->if_carp = NULL; 954 free(cif, M_IFADDR); 955 } 956 } 957 sc->sc_carpdev = NULL; 958 splx(s); 959 KERNEL_UNLOCK_ONE(NULL); 960 } 961 962 /* Detach an interface from the carp. */ 963 void 964 carp_ifdetach(struct ifnet *ifp) 965 { 966 struct carp_softc *sc, *nextsc; 967 struct carp_if *cif = (struct carp_if *)ifp->if_carp; 968 969 for (sc = TAILQ_FIRST(&cif->vhif_vrs); sc; sc = nextsc) { 970 nextsc = TAILQ_NEXT(sc, sc_list); 971 carpdetach(sc); 972 } 973 } 974 975 static void 976 carp_prepare_ad(struct mbuf *m, struct carp_softc *sc, 977 struct carp_header *ch) 978 { 979 if (sc->sc_init_counter) { 980 /* this could also be seconds since unix epoch */ 981 sc->sc_counter = cprng_fast64(); 982 } else 983 sc->sc_counter++; 984 985 ch->carp_counter[0] = htonl((sc->sc_counter>>32)&0xffffffff); 986 ch->carp_counter[1] = htonl(sc->sc_counter&0xffffffff); 987 988 carp_hmac_generate(sc, ch->carp_counter, ch->carp_md); 989 } 990 991 static void 992 carp_send_ad_all(void) 993 { 994 struct ifnet *ifp; 995 struct carp_if *cif; 996 struct carp_softc *vh; 997 int s; 998 int bound = curlwp_bind(); 999 1000 s = pserialize_read_enter(); 1001 IFNET_READER_FOREACH(ifp) { 1002 struct psref psref; 1003 if (ifp->if_carp == NULL || ifp->if_type == IFT_CARP) 1004 continue; 1005 1006 if_acquire(ifp, &psref); 1007 pserialize_read_exit(s); 1008 1009 cif = (struct carp_if *)ifp->if_carp; 1010 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1011 if ((vh->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) == 1012 (IFF_UP|IFF_RUNNING) && vh->sc_state == MASTER) 1013 carp_send_ad(vh); 1014 } 1015 1016 s = pserialize_read_enter(); 1017 if_release(ifp, &psref); 1018 } 1019 pserialize_read_exit(s); 1020 curlwp_bindx(bound); 1021 } 1022 1023 1024 static void 1025 carp_send_ad(void *v) 1026 { 1027 struct carp_header ch; 1028 struct timeval tv; 1029 struct carp_softc *sc = v; 1030 struct carp_header *ch_ptr; 1031 struct mbuf *m; 1032 int error, len, advbase, advskew, s; 1033 struct sockaddr sa; 1034 1035 KERNEL_LOCK(1, NULL); 1036 s = splsoftnet(); 1037 1038 advbase = advskew = 0; /* Sssssh compiler */ 1039 if (sc->sc_carpdev == NULL) { 1040 sc->sc_if.if_oerrors++; 1041 goto retry_later; 1042 } 1043 1044 /* bow out if we've gone to backup (the carp interface is going down) */ 1045 if (sc->sc_bow_out) { 1046 sc->sc_bow_out = 0; 1047 advbase = 255; 1048 advskew = 255; 1049 } else { 1050 advbase = sc->sc_advbase; 1051 if (!carp_suppress_preempt || sc->sc_advskew > 240) 1052 advskew = sc->sc_advskew; 1053 else 1054 advskew = 240; 1055 tv.tv_sec = advbase; 1056 tv.tv_usec = advskew * 1000000 / 256; 1057 } 1058 1059 ch.carp_version = CARP_VERSION; 1060 ch.carp_type = CARP_ADVERTISEMENT; 1061 ch.carp_vhid = sc->sc_vhid; 1062 ch.carp_advbase = advbase; 1063 ch.carp_advskew = advskew; 1064 ch.carp_authlen = 7; /* XXX DEFINE */ 1065 ch.carp_pad1 = 0; /* must be zero */ 1066 ch.carp_cksum = 0; 1067 1068 1069 #ifdef INET 1070 if (sc->sc_naddrs) { 1071 struct ip *ip; 1072 struct ifaddr *ifa; 1073 int _s; 1074 1075 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1076 if (m == NULL) { 1077 sc->sc_if.if_oerrors++; 1078 CARP_STATINC(CARP_STAT_ONOMEM); 1079 /* XXX maybe less ? */ 1080 goto retry_later; 1081 } 1082 MCLAIM(m, &carp_proto_mowner_tx); 1083 len = sizeof(*ip) + sizeof(ch); 1084 m->m_pkthdr.len = len; 1085 m_reset_rcvif(m); 1086 m->m_len = len; 1087 m_align(m, m->m_len); 1088 m->m_flags |= M_MCAST; 1089 ip = mtod(m, struct ip *); 1090 ip->ip_v = IPVERSION; 1091 ip->ip_hl = sizeof(*ip) >> 2; 1092 ip->ip_tos = IPTOS_LOWDELAY; 1093 ip->ip_len = htons(len); 1094 ip->ip_id = 0; /* no need for id, we don't support fragments */ 1095 ip->ip_off = htons(IP_DF); 1096 ip->ip_ttl = CARP_DFLTTL; 1097 ip->ip_p = IPPROTO_CARP; 1098 ip->ip_sum = 0; 1099 1100 memset(&sa, 0, sizeof(sa)); 1101 sa.sa_family = AF_INET; 1102 _s = pserialize_read_enter(); 1103 ifa = ifaof_ifpforaddr(&sa, sc->sc_carpdev); 1104 if (ifa == NULL) 1105 ip->ip_src.s_addr = 0; 1106 else 1107 ip->ip_src.s_addr = 1108 ifatoia(ifa)->ia_addr.sin_addr.s_addr; 1109 pserialize_read_exit(_s); 1110 ip->ip_dst.s_addr = INADDR_CARP_GROUP; 1111 1112 ch_ptr = (struct carp_header *)(&ip[1]); 1113 memcpy(ch_ptr, &ch, sizeof(ch)); 1114 carp_prepare_ad(m, sc, ch_ptr); 1115 1116 m->m_data += sizeof(*ip); 1117 ch_ptr->carp_cksum = carp_cksum(m, len - sizeof(*ip)); 1118 m->m_data -= sizeof(*ip); 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_OPACKETS); 1124 1125 error = ip_output(m, NULL, NULL, IP_RAWOUTPUT, &sc->sc_imo, 1126 NULL); 1127 if (error) { 1128 if (error == ENOBUFS) 1129 CARP_STATINC(CARP_STAT_ONOMEM); 1130 else 1131 CARP_LOG(sc, ("ip_output failed: %d", error)); 1132 sc->sc_if.if_oerrors++; 1133 if (sc->sc_sendad_errors < INT_MAX) 1134 sc->sc_sendad_errors++; 1135 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) { 1136 carp_suppress_preempt++; 1137 if (carp_suppress_preempt == 1) 1138 carp_send_ad_all(); 1139 } 1140 sc->sc_sendad_success = 0; 1141 } else { 1142 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) { 1143 if (++sc->sc_sendad_success >= 1144 CARP_SENDAD_MIN_SUCCESS) { 1145 carp_suppress_preempt--; 1146 sc->sc_sendad_errors = 0; 1147 } 1148 } else 1149 sc->sc_sendad_errors = 0; 1150 } 1151 } 1152 #endif /* INET */ 1153 #ifdef INET6 1154 if (sc->sc_naddrs6) { 1155 struct ip6_hdr *ip6; 1156 struct ifaddr *ifa; 1157 int _s; 1158 1159 MGETHDR(m, M_DONTWAIT, MT_HEADER); 1160 if (m == NULL) { 1161 sc->sc_if.if_oerrors++; 1162 CARP_STATINC(CARP_STAT_ONOMEM); 1163 /* XXX maybe less ? */ 1164 goto retry_later; 1165 } 1166 MCLAIM(m, &carp_proto6_mowner_tx); 1167 len = sizeof(*ip6) + sizeof(ch); 1168 m->m_pkthdr.len = len; 1169 m_reset_rcvif(m); 1170 m->m_len = len; 1171 m_align(m, m->m_len); 1172 m->m_flags |= M_MCAST; 1173 ip6 = mtod(m, struct ip6_hdr *); 1174 memset(ip6, 0, sizeof(*ip6)); 1175 ip6->ip6_vfc |= IPV6_VERSION; 1176 ip6->ip6_hlim = CARP_DFLTTL; 1177 ip6->ip6_nxt = IPPROTO_CARP; 1178 1179 /* set the source address */ 1180 memset(&sa, 0, sizeof(sa)); 1181 sa.sa_family = AF_INET6; 1182 _s = pserialize_read_enter(); 1183 ifa = ifaof_ifpforaddr(&sa, sc->sc_carpdev); 1184 if (ifa == NULL) /* This should never happen with IPv6 */ 1185 memset(&ip6->ip6_src, 0, sizeof(struct in6_addr)); 1186 else 1187 bcopy(ifatoia6(ifa)->ia_addr.sin6_addr.s6_addr, 1188 &ip6->ip6_src, sizeof(struct in6_addr)); 1189 pserialize_read_exit(_s); 1190 /* set the multicast destination */ 1191 1192 ip6->ip6_dst.s6_addr16[0] = htons(0xff02); 1193 ip6->ip6_dst.s6_addr8[15] = 0x12; 1194 if (in6_setscope(&ip6->ip6_dst, &sc->sc_if, NULL) != 0) { 1195 sc->sc_if.if_oerrors++; 1196 m_freem(m); 1197 CARP_LOG(sc, ("in6_setscope failed")); 1198 goto retry_later; 1199 } 1200 1201 ch_ptr = (struct carp_header *)(&ip6[1]); 1202 memcpy(ch_ptr, &ch, sizeof(ch)); 1203 carp_prepare_ad(m, sc, ch_ptr); 1204 1205 ch_ptr->carp_cksum = carp6_cksum(m, sizeof(*ip6), 1206 len - sizeof(*ip6)); 1207 1208 nanotime(&sc->sc_if.if_lastchange); 1209 sc->sc_if.if_opackets++; 1210 sc->sc_if.if_obytes += len; 1211 CARP_STATINC(CARP_STAT_OPACKETS6); 1212 1213 error = ip6_output(m, NULL, NULL, 0, &sc->sc_im6o, NULL, NULL); 1214 if (error) { 1215 if (error == ENOBUFS) 1216 CARP_STATINC(CARP_STAT_ONOMEM); 1217 else 1218 CARP_LOG(sc, ("ip6_output failed: %d", error)); 1219 sc->sc_if.if_oerrors++; 1220 if (sc->sc_sendad_errors < INT_MAX) 1221 sc->sc_sendad_errors++; 1222 if (sc->sc_sendad_errors == CARP_SENDAD_MAX_ERRORS) { 1223 carp_suppress_preempt++; 1224 if (carp_suppress_preempt == 1) 1225 carp_send_ad_all(); 1226 } 1227 sc->sc_sendad_success = 0; 1228 } else { 1229 if (sc->sc_sendad_errors >= CARP_SENDAD_MAX_ERRORS) { 1230 if (++sc->sc_sendad_success >= 1231 CARP_SENDAD_MIN_SUCCESS) { 1232 carp_suppress_preempt--; 1233 sc->sc_sendad_errors = 0; 1234 } 1235 } else 1236 sc->sc_sendad_errors = 0; 1237 } 1238 } 1239 #endif /* INET6 */ 1240 1241 retry_later: 1242 splx(s); 1243 KERNEL_UNLOCK_ONE(NULL); 1244 if (advbase != 255 || advskew != 255) 1245 callout_schedule(&sc->sc_ad_tmo, tvtohz(&tv)); 1246 } 1247 1248 /* 1249 * Broadcast a gratuitous ARP request containing 1250 * the virtual router MAC address for each IP address 1251 * associated with the virtual router. 1252 */ 1253 static void 1254 carp_send_arp(struct carp_softc *sc) 1255 { 1256 struct ifaddr *ifa; 1257 int s, bound; 1258 1259 KERNEL_LOCK(1, NULL); 1260 bound = curlwp_bind(); 1261 s = pserialize_read_enter(); 1262 IFADDR_READER_FOREACH(ifa, &sc->sc_if) { 1263 struct psref psref; 1264 1265 if (ifa->ifa_addr->sa_family != AF_INET) 1266 continue; 1267 1268 ifa_acquire(ifa, &psref); 1269 pserialize_read_exit(s); 1270 1271 arpannounce(sc->sc_carpdev, ifa, CLLADDR(sc->sc_if.if_sadl)); 1272 1273 s = pserialize_read_enter(); 1274 ifa_release(ifa, &psref); 1275 } 1276 pserialize_read_exit(s); 1277 curlwp_bindx(bound); 1278 KERNEL_UNLOCK_ONE(NULL); 1279 } 1280 1281 #ifdef INET6 1282 static void 1283 carp_send_na(struct carp_softc *sc) 1284 { 1285 struct ifaddr *ifa; 1286 struct in6_addr *in6; 1287 static struct in6_addr mcast = IN6ADDR_LINKLOCAL_ALLNODES_INIT; 1288 int s, bound; 1289 1290 KERNEL_LOCK(1, NULL); 1291 bound = curlwp_bind(); 1292 s = pserialize_read_enter(); 1293 IFADDR_READER_FOREACH(ifa, &sc->sc_if) { 1294 struct psref psref; 1295 1296 if (ifa->ifa_addr->sa_family != AF_INET6) 1297 continue; 1298 1299 ifa_acquire(ifa, &psref); 1300 pserialize_read_exit(s); 1301 1302 in6 = &ifatoia6(ifa)->ia_addr.sin6_addr; 1303 nd6_na_output(sc->sc_carpdev, &mcast, in6, 1304 ND_NA_FLAG_OVERRIDE, 1, NULL); 1305 1306 s = pserialize_read_enter(); 1307 ifa_release(ifa, &psref); 1308 } 1309 pserialize_read_exit(s); 1310 curlwp_bindx(bound); 1311 KERNEL_UNLOCK_ONE(NULL); 1312 } 1313 #endif /* INET6 */ 1314 1315 /* 1316 * Based on bridge_hash() in if_bridge.c 1317 */ 1318 #define mix(a,b,c) \ 1319 do { \ 1320 a -= b; a -= c; a ^= (c >> 13); \ 1321 b -= c; b -= a; b ^= (a << 8); \ 1322 c -= a; c -= b; c ^= (b >> 13); \ 1323 a -= b; a -= c; a ^= (c >> 12); \ 1324 b -= c; b -= a; b ^= (a << 16); \ 1325 c -= a; c -= b; c ^= (b >> 5); \ 1326 a -= b; a -= c; a ^= (c >> 3); \ 1327 b -= c; b -= a; b ^= (a << 10); \ 1328 c -= a; c -= b; c ^= (b >> 15); \ 1329 } while (0) 1330 1331 static u_int32_t 1332 carp_hash(struct carp_softc *sc, u_char *src) 1333 { 1334 u_int32_t a = 0x9e3779b9, b = sc->sc_hashkey[0], c = sc->sc_hashkey[1]; 1335 1336 c += sc->sc_key[3] << 24; 1337 c += sc->sc_key[2] << 16; 1338 c += sc->sc_key[1] << 8; 1339 c += sc->sc_key[0]; 1340 b += src[5] << 8; 1341 b += src[4]; 1342 a += src[3] << 24; 1343 a += src[2] << 16; 1344 a += src[1] << 8; 1345 a += src[0]; 1346 1347 mix(a, b, c); 1348 return (c); 1349 } 1350 1351 static int 1352 carp_addrcount(struct carp_if *cif, struct in_ifaddr *ia, int type) 1353 { 1354 struct carp_softc *vh; 1355 struct ifaddr *ifa; 1356 int count = 0; 1357 1358 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1359 if ((type == CARP_COUNT_RUNNING && 1360 (vh->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) == 1361 (IFF_UP|IFF_RUNNING)) || 1362 (type == CARP_COUNT_MASTER && vh->sc_state == MASTER)) { 1363 int s = pserialize_read_enter(); 1364 IFADDR_READER_FOREACH(ifa, &vh->sc_if) { 1365 if (ifa->ifa_addr->sa_family == AF_INET && 1366 ia->ia_addr.sin_addr.s_addr == 1367 ifatoia(ifa)->ia_addr.sin_addr.s_addr) 1368 count++; 1369 } 1370 pserialize_read_exit(s); 1371 } 1372 } 1373 return (count); 1374 } 1375 1376 int 1377 carp_iamatch(struct in_ifaddr *ia, u_char *src, 1378 u_int32_t *count, u_int32_t index) 1379 { 1380 struct carp_softc *sc = ia->ia_ifp->if_softc; 1381 1382 if (carp_opts[CARPCTL_ARPBALANCE]) { 1383 /* 1384 * We use the source ip to decide which virtual host should 1385 * handle the request. If we're master of that virtual host, 1386 * then we respond, otherwise, just drop the arp packet on 1387 * the floor. 1388 */ 1389 1390 /* Count the elegible carp interfaces with this address */ 1391 if (*count == 0) 1392 *count = carp_addrcount( 1393 (struct carp_if *)ia->ia_ifp->if_carpdev->if_carp, 1394 ia, CARP_COUNT_RUNNING); 1395 1396 /* This should never happen, but... */ 1397 if (*count == 0) 1398 return (0); 1399 1400 if (carp_hash(sc, src) % *count == index - 1 && 1401 sc->sc_state == MASTER) { 1402 return (1); 1403 } 1404 } else { 1405 if (sc->sc_state == MASTER) 1406 return (1); 1407 } 1408 1409 return (0); 1410 } 1411 1412 #ifdef INET6 1413 struct ifaddr * 1414 carp_iamatch6(void *v, struct in6_addr *taddr) 1415 { 1416 struct carp_if *cif = v; 1417 struct carp_softc *vh; 1418 struct ifaddr *ifa; 1419 1420 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1421 int s = pserialize_read_enter(); 1422 IFADDR_READER_FOREACH(ifa, &vh->sc_if) { 1423 if (IN6_ARE_ADDR_EQUAL(taddr, 1424 &ifatoia6(ifa)->ia_addr.sin6_addr) && 1425 ((vh->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) == 1426 (IFF_UP|IFF_RUNNING)) && vh->sc_state == MASTER) 1427 return (ifa); 1428 } 1429 pserialize_read_exit(s); 1430 } 1431 1432 return (NULL); 1433 } 1434 #endif /* INET6 */ 1435 1436 struct ifnet * 1437 carp_ourether(void *v, struct ether_header *eh, u_char iftype, int src) 1438 { 1439 struct carp_if *cif = (struct carp_if *)v; 1440 struct carp_softc *vh; 1441 u_int8_t *ena; 1442 1443 if (src) 1444 ena = (u_int8_t *)&eh->ether_shost; 1445 else 1446 ena = (u_int8_t *)&eh->ether_dhost; 1447 1448 switch (iftype) { 1449 case IFT_ETHER: 1450 case IFT_FDDI: 1451 if (ena[0] || ena[1] || ena[2] != 0x5e || ena[3] || ena[4] != 1) 1452 return (NULL); 1453 break; 1454 case IFT_ISO88025: 1455 if (ena[0] != 3 || ena[1] || ena[4] || ena[5]) 1456 return (NULL); 1457 break; 1458 default: 1459 return (NULL); 1460 break; 1461 } 1462 1463 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) 1464 if ((vh->sc_if.if_flags & (IFF_UP|IFF_RUNNING)) == 1465 (IFF_UP|IFF_RUNNING) && vh->sc_state == MASTER && 1466 !memcmp(ena, CLLADDR(vh->sc_if.if_sadl), 1467 ETHER_ADDR_LEN)) { 1468 return (&vh->sc_if); 1469 } 1470 1471 return (NULL); 1472 } 1473 1474 int 1475 carp_input(struct mbuf *m, u_int8_t *shost, u_int8_t *dhost, u_int16_t etype) 1476 { 1477 struct ether_header eh; 1478 struct carp_if *cif = (struct carp_if *)m_get_rcvif_NOMPSAFE(m)->if_carp; 1479 struct ifnet *ifp; 1480 1481 memcpy(&eh.ether_shost, shost, sizeof(eh.ether_shost)); 1482 memcpy(&eh.ether_dhost, dhost, sizeof(eh.ether_dhost)); 1483 eh.ether_type = etype; 1484 1485 if (m->m_flags & (M_BCAST|M_MCAST)) { 1486 struct carp_softc *vh; 1487 struct mbuf *m0; 1488 1489 /* 1490 * XXX Should really check the list of multicast addresses 1491 * for each CARP interface _before_ copying. 1492 */ 1493 TAILQ_FOREACH(vh, &cif->vhif_vrs, sc_list) { 1494 m0 = m_copym(m, 0, M_COPYALL, M_DONTWAIT); 1495 if (m0 == NULL) 1496 continue; 1497 m_set_rcvif(m0, &vh->sc_if); 1498 ether_input(&vh->sc_if, m0); 1499 } 1500 return (1); 1501 } 1502 1503 ifp = carp_ourether(cif, &eh, m_get_rcvif_NOMPSAFE(m)->if_type, 0); 1504 if (ifp == NULL) { 1505 return (1); 1506 } 1507 1508 m_set_rcvif(m, ifp); 1509 1510 bpf_mtap(ifp, m, BPF_D_IN); 1511 ifp->if_ipackets++; 1512 ether_input(ifp, m); 1513 return (0); 1514 } 1515 1516 static void 1517 carp_master_down(void *v) 1518 { 1519 struct carp_softc *sc = v; 1520 1521 switch (sc->sc_state) { 1522 case INIT: 1523 printf("%s: master_down event in INIT state\n", 1524 sc->sc_if.if_xname); 1525 break; 1526 case MASTER: 1527 break; 1528 case BACKUP: 1529 CARP_LOG(sc, ("INIT -> MASTER (preempting)")); 1530 carp_set_state(sc, MASTER); 1531 carp_send_ad(sc); 1532 carp_send_arp(sc); 1533 #ifdef INET6 1534 carp_send_na(sc); 1535 #endif /* INET6 */ 1536 carp_setrun(sc, 0); 1537 carp_setroute(sc, RTM_ADD); 1538 break; 1539 } 1540 } 1541 1542 /* 1543 * When in backup state, af indicates whether to reset the master down timer 1544 * for v4 or v6. If it's set to zero, reset the ones which are already pending. 1545 */ 1546 static void 1547 carp_setrun(struct carp_softc *sc, sa_family_t af) 1548 { 1549 struct timeval tv; 1550 1551 if (sc->sc_carpdev == NULL) { 1552 sc->sc_if.if_flags &= ~IFF_RUNNING; 1553 carp_set_state(sc, INIT); 1554 return; 1555 } 1556 1557 if (sc->sc_if.if_flags & IFF_UP && sc->sc_vhid > 0 && 1558 (sc->sc_naddrs || sc->sc_naddrs6) && !sc->sc_suppress) { 1559 sc->sc_if.if_flags |= IFF_RUNNING; 1560 } else { 1561 sc->sc_if.if_flags &= ~IFF_RUNNING; 1562 carp_setroute(sc, RTM_DELETE); 1563 return; 1564 } 1565 1566 switch (sc->sc_state) { 1567 case INIT: 1568 carp_set_state(sc, BACKUP); 1569 carp_setroute(sc, RTM_DELETE); 1570 carp_setrun(sc, 0); 1571 break; 1572 case BACKUP: 1573 callout_stop(&sc->sc_ad_tmo); 1574 tv.tv_sec = 3 * sc->sc_advbase; 1575 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1576 switch (af) { 1577 #ifdef INET 1578 case AF_INET: 1579 callout_schedule(&sc->sc_md_tmo, tvtohz(&tv)); 1580 break; 1581 #endif /* INET */ 1582 #ifdef INET6 1583 case AF_INET6: 1584 callout_schedule(&sc->sc_md6_tmo, tvtohz(&tv)); 1585 break; 1586 #endif /* INET6 */ 1587 default: 1588 if (sc->sc_naddrs) 1589 callout_schedule(&sc->sc_md_tmo, tvtohz(&tv)); 1590 #ifdef INET6 1591 if (sc->sc_naddrs6) 1592 callout_schedule(&sc->sc_md6_tmo, tvtohz(&tv)); 1593 #endif /* INET6 */ 1594 break; 1595 } 1596 break; 1597 case MASTER: 1598 tv.tv_sec = sc->sc_advbase; 1599 tv.tv_usec = sc->sc_advskew * 1000000 / 256; 1600 callout_schedule(&sc->sc_ad_tmo, tvtohz(&tv)); 1601 break; 1602 } 1603 } 1604 1605 static void 1606 carp_multicast_cleanup(struct carp_softc *sc) 1607 { 1608 struct ip_moptions *imo = &sc->sc_imo; 1609 #ifdef INET6 1610 struct ip6_moptions *im6o = &sc->sc_im6o; 1611 #endif 1612 u_int16_t n = imo->imo_num_memberships; 1613 1614 /* Clean up our own multicast memberships */ 1615 while (n-- > 0) { 1616 if (imo->imo_membership[n] != NULL) { 1617 in_delmulti(imo->imo_membership[n]); 1618 imo->imo_membership[n] = NULL; 1619 } 1620 } 1621 imo->imo_num_memberships = 0; 1622 imo->imo_multicast_if_index = 0; 1623 1624 #ifdef INET6 1625 while (!LIST_EMPTY(&im6o->im6o_memberships)) { 1626 struct in6_multi_mship *imm = 1627 LIST_FIRST(&im6o->im6o_memberships); 1628 1629 LIST_REMOVE(imm, i6mm_chain); 1630 in6_leavegroup(imm); 1631 } 1632 im6o->im6o_multicast_if_index = 0; 1633 #endif 1634 1635 /* And any other multicast memberships */ 1636 carp_ether_purgemulti(sc); 1637 } 1638 1639 static int 1640 carp_set_ifp(struct carp_softc *sc, struct ifnet *ifp) 1641 { 1642 struct carp_if *cif, *ncif = NULL; 1643 struct carp_softc *vr, *after = NULL; 1644 int myself = 0, error = 0; 1645 int s; 1646 1647 if (ifp == sc->sc_carpdev) 1648 return (0); 1649 1650 if (ifp != NULL) { 1651 if ((ifp->if_flags & IFF_MULTICAST) == 0) 1652 return (EADDRNOTAVAIL); 1653 1654 if (ifp->if_type == IFT_CARP) 1655 return (EINVAL); 1656 1657 if (ifp->if_carp == NULL) { 1658 ncif = malloc(sizeof(*cif), M_IFADDR, M_NOWAIT); 1659 if (ncif == NULL) 1660 return (ENOBUFS); 1661 if ((error = ifpromisc(ifp, 1))) { 1662 free(ncif, M_IFADDR); 1663 return (error); 1664 } 1665 1666 ncif->vhif_ifp = ifp; 1667 TAILQ_INIT(&ncif->vhif_vrs); 1668 } else { 1669 cif = (struct carp_if *)ifp->if_carp; 1670 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) 1671 if (vr != sc && vr->sc_vhid == sc->sc_vhid) 1672 return (EINVAL); 1673 } 1674 1675 /* detach from old interface */ 1676 if (sc->sc_carpdev != NULL) 1677 carpdetach(sc); 1678 1679 /* join multicast groups */ 1680 if (sc->sc_naddrs < 0 && 1681 (error = carp_join_multicast(sc)) != 0) { 1682 if (ncif != NULL) 1683 free(ncif, M_IFADDR); 1684 return (error); 1685 } 1686 1687 #ifdef INET6 1688 if (sc->sc_naddrs6 < 0 && 1689 (error = carp_join_multicast6(sc)) != 0) { 1690 if (ncif != NULL) 1691 free(ncif, M_IFADDR); 1692 carp_multicast_cleanup(sc); 1693 return (error); 1694 } 1695 #endif 1696 1697 /* attach carp interface to physical interface */ 1698 if (ncif != NULL) 1699 ifp->if_carp = (void *)ncif; 1700 sc->sc_carpdev = ifp; 1701 sc->sc_if.if_capabilities = ifp->if_capabilities & 1702 (IFCAP_TSOv4 | IFCAP_TSOv6 | 1703 IFCAP_CSUM_IPv4_Tx|IFCAP_CSUM_IPv4_Rx| 1704 IFCAP_CSUM_TCPv4_Tx|IFCAP_CSUM_TCPv4_Rx| 1705 IFCAP_CSUM_UDPv4_Tx|IFCAP_CSUM_UDPv4_Rx| 1706 IFCAP_CSUM_TCPv6_Tx|IFCAP_CSUM_TCPv6_Rx| 1707 IFCAP_CSUM_UDPv6_Tx|IFCAP_CSUM_UDPv6_Rx); 1708 1709 cif = (struct carp_if *)ifp->if_carp; 1710 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) { 1711 if (vr == sc) 1712 myself = 1; 1713 if (vr->sc_vhid < sc->sc_vhid) 1714 after = vr; 1715 } 1716 1717 if (!myself) { 1718 /* We're trying to keep things in order */ 1719 if (after == NULL) { 1720 TAILQ_INSERT_TAIL(&cif->vhif_vrs, sc, sc_list); 1721 } else { 1722 TAILQ_INSERT_AFTER(&cif->vhif_vrs, after, 1723 sc, sc_list); 1724 } 1725 cif->vhif_nvrs++; 1726 } 1727 if (sc->sc_naddrs || sc->sc_naddrs6) 1728 sc->sc_if.if_flags |= IFF_UP; 1729 carp_set_enaddr(sc); 1730 KERNEL_LOCK(1, NULL); 1731 s = splnet(); 1732 /* XXX linkstatehooks establish */ 1733 carp_carpdev_state(ifp); 1734 splx(s); 1735 KERNEL_UNLOCK_ONE(NULL); 1736 } else { 1737 carpdetach(sc); 1738 sc->sc_if.if_flags &= ~(IFF_UP|IFF_RUNNING); 1739 } 1740 return (0); 1741 } 1742 1743 static void 1744 carp_set_enaddr(struct carp_softc *sc) 1745 { 1746 uint8_t enaddr[ETHER_ADDR_LEN]; 1747 if (sc->sc_carpdev && sc->sc_carpdev->if_type == IFT_ISO88025) { 1748 enaddr[0] = 3; 1749 enaddr[1] = 0; 1750 enaddr[2] = 0x40 >> (sc->sc_vhid - 1); 1751 enaddr[3] = 0x40000 >> (sc->sc_vhid - 1); 1752 enaddr[4] = 0; 1753 enaddr[5] = 0; 1754 } else { 1755 enaddr[0] = 0; 1756 enaddr[1] = 0; 1757 enaddr[2] = 0x5e; 1758 enaddr[3] = 0; 1759 enaddr[4] = 1; 1760 enaddr[5] = sc->sc_vhid; 1761 } 1762 if_set_sadl(&sc->sc_if, enaddr, sizeof(enaddr), false); 1763 } 1764 1765 #if 0 1766 static void 1767 carp_addr_updated(void *v) 1768 { 1769 struct carp_softc *sc = (struct carp_softc *) v; 1770 struct ifaddr *ifa; 1771 int new_naddrs = 0, new_naddrs6 = 0; 1772 1773 IFADDR_READER_FOREACH(ifa, &sc->sc_if) { 1774 if (ifa->ifa_addr->sa_family == AF_INET) 1775 new_naddrs++; 1776 else if (ifa->ifa_addr->sa_family == AF_INET6) 1777 new_naddrs6++; 1778 } 1779 1780 /* Handle a callback after SIOCDIFADDR */ 1781 if (new_naddrs < sc->sc_naddrs || new_naddrs6 < sc->sc_naddrs6) { 1782 struct in_addr mc_addr; 1783 1784 sc->sc_naddrs = new_naddrs; 1785 sc->sc_naddrs6 = new_naddrs6; 1786 1787 /* Re-establish multicast membership removed by in_control */ 1788 mc_addr.s_addr = INADDR_CARP_GROUP; 1789 if (!in_multi_group(mc_addr, &sc->sc_if, 0)) { 1790 memset(&sc->sc_imo, 0, sizeof(sc->sc_imo)); 1791 1792 if (sc->sc_carpdev != NULL && sc->sc_naddrs > 0) 1793 carp_join_multicast(sc); 1794 } 1795 1796 if (sc->sc_naddrs == 0 && sc->sc_naddrs6 == 0) { 1797 sc->sc_if.if_flags &= ~IFF_UP; 1798 carp_set_state(sc, INIT); 1799 } else 1800 carp_hmac_prepare(sc); 1801 } 1802 1803 carp_setrun(sc, 0); 1804 } 1805 #endif 1806 1807 static int 1808 carp_set_addr(struct carp_softc *sc, struct sockaddr_in *sin) 1809 { 1810 struct ifnet *ifp = sc->sc_carpdev; 1811 struct in_ifaddr *ia, *ia_if; 1812 int error = 0; 1813 int s; 1814 1815 if (sin->sin_addr.s_addr == 0) { 1816 if (!(sc->sc_if.if_flags & IFF_UP)) 1817 carp_set_state(sc, INIT); 1818 if (sc->sc_naddrs) 1819 sc->sc_if.if_flags |= IFF_UP; 1820 carp_setrun(sc, 0); 1821 return (0); 1822 } 1823 1824 /* we have to do this by hand to ensure we don't match on ourselves */ 1825 ia_if = NULL; 1826 s = pserialize_read_enter(); 1827 IN_ADDRLIST_READER_FOREACH(ia) { 1828 /* and, yeah, we need a multicast-capable iface too */ 1829 if (ia->ia_ifp != &sc->sc_if && 1830 ia->ia_ifp->if_type != IFT_CARP && 1831 (ia->ia_ifp->if_flags & IFF_MULTICAST) && 1832 (sin->sin_addr.s_addr & ia->ia_subnetmask) == 1833 ia->ia_subnet) { 1834 if (!ia_if) 1835 ia_if = ia; 1836 } 1837 } 1838 1839 if (ia_if) { 1840 ia = ia_if; 1841 if (ifp) { 1842 if (ifp != ia->ia_ifp) 1843 return (EADDRNOTAVAIL); 1844 } else { 1845 /* FIXME NOMPSAFE */ 1846 ifp = ia->ia_ifp; 1847 } 1848 } 1849 pserialize_read_exit(s); 1850 1851 if ((error = carp_set_ifp(sc, ifp))) 1852 return (error); 1853 1854 if (sc->sc_carpdev == NULL) 1855 return (EADDRNOTAVAIL); 1856 1857 if (sc->sc_naddrs == 0 && (error = carp_join_multicast(sc)) != 0) 1858 return (error); 1859 1860 sc->sc_naddrs++; 1861 if (sc->sc_carpdev != NULL) 1862 sc->sc_if.if_flags |= IFF_UP; 1863 1864 carp_set_state(sc, INIT); 1865 carp_setrun(sc, 0); 1866 1867 /* 1868 * Hook if_addrhooks so that we get a callback after in_ifinit has run, 1869 * to correct any inappropriate routes that it inserted. 1870 */ 1871 if (sc->ah_cookie == 0) { 1872 /* XXX link address hook */ 1873 } 1874 1875 return (0); 1876 } 1877 1878 static int 1879 carp_join_multicast(struct carp_softc *sc) 1880 { 1881 struct ip_moptions *imo = &sc->sc_imo, tmpimo; 1882 struct in_addr addr; 1883 1884 memset(&tmpimo, 0, sizeof(tmpimo)); 1885 addr.s_addr = INADDR_CARP_GROUP; 1886 if ((tmpimo.imo_membership[0] = 1887 in_addmulti(&addr, &sc->sc_if)) == NULL) { 1888 return (ENOBUFS); 1889 } 1890 1891 imo->imo_membership[0] = tmpimo.imo_membership[0]; 1892 imo->imo_num_memberships = 1; 1893 imo->imo_multicast_if_index = sc->sc_if.if_index; 1894 imo->imo_multicast_ttl = CARP_DFLTTL; 1895 imo->imo_multicast_loop = 0; 1896 return (0); 1897 } 1898 1899 1900 #ifdef INET6 1901 static int 1902 carp_set_addr6(struct carp_softc *sc, struct sockaddr_in6 *sin6) 1903 { 1904 struct ifnet *ifp = sc->sc_carpdev; 1905 struct in6_ifaddr *ia, *ia_if; 1906 int error = 0; 1907 int s; 1908 1909 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) { 1910 if (!(sc->sc_if.if_flags & IFF_UP)) 1911 carp_set_state(sc, INIT); 1912 if (sc->sc_naddrs6) 1913 sc->sc_if.if_flags |= IFF_UP; 1914 carp_setrun(sc, 0); 1915 return (0); 1916 } 1917 1918 /* we have to do this by hand to ensure we don't match on ourselves */ 1919 ia_if = NULL; 1920 s = pserialize_read_enter(); 1921 IN6_ADDRLIST_READER_FOREACH(ia) { 1922 int i; 1923 1924 for (i = 0; i < 4; i++) { 1925 if ((sin6->sin6_addr.s6_addr32[i] & 1926 ia->ia_prefixmask.sin6_addr.s6_addr32[i]) != 1927 (ia->ia_addr.sin6_addr.s6_addr32[i] & 1928 ia->ia_prefixmask.sin6_addr.s6_addr32[i])) 1929 break; 1930 } 1931 /* and, yeah, we need a multicast-capable iface too */ 1932 if (ia->ia_ifp != &sc->sc_if && 1933 ia->ia_ifp->if_type != IFT_CARP && 1934 (ia->ia_ifp->if_flags & IFF_MULTICAST) && 1935 (i == 4)) { 1936 if (!ia_if) 1937 ia_if = ia; 1938 } 1939 } 1940 pserialize_read_exit(s); 1941 1942 if (ia_if) { 1943 ia = ia_if; 1944 if (sc->sc_carpdev) { 1945 if (sc->sc_carpdev != ia->ia_ifp) 1946 return (EADDRNOTAVAIL); 1947 } else { 1948 ifp = ia->ia_ifp; 1949 } 1950 } 1951 1952 if ((error = carp_set_ifp(sc, ifp))) 1953 return (error); 1954 1955 if (sc->sc_carpdev == NULL) 1956 return (EADDRNOTAVAIL); 1957 1958 if (sc->sc_naddrs6 == 0 && (error = carp_join_multicast6(sc)) != 0) 1959 return (error); 1960 1961 sc->sc_naddrs6++; 1962 if (sc->sc_carpdev != NULL) 1963 sc->sc_if.if_flags |= IFF_UP; 1964 carp_set_state(sc, INIT); 1965 carp_setrun(sc, 0); 1966 1967 return (0); 1968 } 1969 1970 static int 1971 carp_join_multicast6(struct carp_softc *sc) 1972 { 1973 struct in6_multi_mship *imm, *imm2; 1974 struct ip6_moptions *im6o = &sc->sc_im6o; 1975 struct sockaddr_in6 addr6; 1976 int error; 1977 1978 /* Join IPv6 CARP multicast group */ 1979 memset(&addr6, 0, sizeof(addr6)); 1980 addr6.sin6_family = AF_INET6; 1981 addr6.sin6_len = sizeof(addr6); 1982 addr6.sin6_addr.s6_addr16[0] = htons(0xff02); 1983 addr6.sin6_addr.s6_addr16[1] = htons(sc->sc_if.if_index); 1984 addr6.sin6_addr.s6_addr8[15] = 0x12; 1985 if ((imm = in6_joingroup(&sc->sc_if, 1986 &addr6.sin6_addr, &error, 0)) == NULL) { 1987 return (error); 1988 } 1989 /* join solicited multicast address */ 1990 memset(&addr6.sin6_addr, 0, sizeof(addr6.sin6_addr)); 1991 addr6.sin6_addr.s6_addr16[0] = htons(0xff02); 1992 addr6.sin6_addr.s6_addr16[1] = htons(sc->sc_if.if_index); 1993 addr6.sin6_addr.s6_addr32[1] = 0; 1994 addr6.sin6_addr.s6_addr32[2] = htonl(1); 1995 addr6.sin6_addr.s6_addr32[3] = 0; 1996 addr6.sin6_addr.s6_addr8[12] = 0xff; 1997 if ((imm2 = in6_joingroup(&sc->sc_if, 1998 &addr6.sin6_addr, &error, 0)) == NULL) { 1999 in6_leavegroup(imm); 2000 return (error); 2001 } 2002 2003 /* apply v6 multicast membership */ 2004 im6o->im6o_multicast_if_index = sc->sc_if.if_index; 2005 if (imm) 2006 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm, 2007 i6mm_chain); 2008 if (imm2) 2009 LIST_INSERT_HEAD(&im6o->im6o_memberships, imm2, 2010 i6mm_chain); 2011 2012 return (0); 2013 } 2014 2015 #endif /* INET6 */ 2016 2017 static int 2018 carp_ioctl(struct ifnet *ifp, u_long cmd, void *data) 2019 { 2020 struct lwp *l = curlwp; /* XXX */ 2021 struct carp_softc *sc = ifp->if_softc, *vr; 2022 struct carpreq carpr; 2023 struct ifaddr *ifa; 2024 struct ifreq *ifr; 2025 struct ifnet *cdev = NULL; 2026 int error = 0; 2027 2028 ifa = (struct ifaddr *)data; 2029 ifr = (struct ifreq *)data; 2030 2031 switch (cmd) { 2032 case SIOCINITIFADDR: 2033 switch (ifa->ifa_addr->sa_family) { 2034 #ifdef INET 2035 case AF_INET: 2036 sc->sc_if.if_flags |= IFF_UP; 2037 memcpy(ifa->ifa_dstaddr, ifa->ifa_addr, 2038 sizeof(struct sockaddr)); 2039 error = carp_set_addr(sc, satosin(ifa->ifa_addr)); 2040 break; 2041 #endif /* INET */ 2042 #ifdef INET6 2043 case AF_INET6: 2044 sc->sc_if.if_flags|= IFF_UP; 2045 error = carp_set_addr6(sc, satosin6(ifa->ifa_addr)); 2046 break; 2047 #endif /* INET6 */ 2048 default: 2049 error = EAFNOSUPPORT; 2050 break; 2051 } 2052 break; 2053 2054 case SIOCSIFFLAGS: 2055 if ((error = ifioctl_common(ifp, cmd, data)) != 0) 2056 break; 2057 if (sc->sc_state != INIT && !(ifr->ifr_flags & IFF_UP)) { 2058 callout_stop(&sc->sc_ad_tmo); 2059 callout_stop(&sc->sc_md_tmo); 2060 callout_stop(&sc->sc_md6_tmo); 2061 if (sc->sc_state == MASTER) { 2062 /* we need the interface up to bow out */ 2063 sc->sc_if.if_flags |= IFF_UP; 2064 sc->sc_bow_out = 1; 2065 carp_send_ad(sc); 2066 } 2067 sc->sc_if.if_flags &= ~IFF_UP; 2068 carp_set_state(sc, INIT); 2069 carp_setrun(sc, 0); 2070 } else if (sc->sc_state == INIT && (ifr->ifr_flags & IFF_UP)) { 2071 sc->sc_if.if_flags |= IFF_UP; 2072 carp_setrun(sc, 0); 2073 } 2074 break; 2075 2076 case SIOCSVH: 2077 if (l == NULL) 2078 break; 2079 if ((error = kauth_authorize_network(l->l_cred, 2080 KAUTH_NETWORK_INTERFACE, 2081 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 2082 NULL)) != 0) 2083 break; 2084 if ((error = copyin(ifr->ifr_data, &carpr, sizeof carpr))) 2085 break; 2086 error = 1; 2087 if (carpr.carpr_carpdev[0] != '\0' && 2088 (cdev = ifunit(carpr.carpr_carpdev)) == NULL) 2089 return (EINVAL); 2090 if ((error = carp_set_ifp(sc, cdev))) 2091 return (error); 2092 if (sc->sc_state != INIT && carpr.carpr_state != sc->sc_state) { 2093 switch (carpr.carpr_state) { 2094 case BACKUP: 2095 callout_stop(&sc->sc_ad_tmo); 2096 carp_set_state(sc, BACKUP); 2097 carp_setrun(sc, 0); 2098 carp_setroute(sc, RTM_DELETE); 2099 break; 2100 case MASTER: 2101 carp_master_down(sc); 2102 break; 2103 default: 2104 break; 2105 } 2106 } 2107 if (carpr.carpr_vhid > 0) { 2108 if (carpr.carpr_vhid > 255) { 2109 error = EINVAL; 2110 break; 2111 } 2112 if (sc->sc_carpdev) { 2113 struct carp_if *cif; 2114 cif = (struct carp_if *)sc->sc_carpdev->if_carp; 2115 TAILQ_FOREACH(vr, &cif->vhif_vrs, sc_list) 2116 if (vr != sc && 2117 vr->sc_vhid == carpr.carpr_vhid) 2118 return (EINVAL); 2119 } 2120 sc->sc_vhid = carpr.carpr_vhid; 2121 carp_set_enaddr(sc); 2122 carp_set_state(sc, INIT); 2123 error--; 2124 } 2125 if (carpr.carpr_advbase > 0 || carpr.carpr_advskew > 0) { 2126 if (carpr.carpr_advskew > 254) { 2127 error = EINVAL; 2128 break; 2129 } 2130 if (carpr.carpr_advbase > 255) { 2131 error = EINVAL; 2132 break; 2133 } 2134 sc->sc_advbase = carpr.carpr_advbase; 2135 sc->sc_advskew = carpr.carpr_advskew; 2136 error--; 2137 } 2138 memcpy(sc->sc_key, carpr.carpr_key, sizeof(sc->sc_key)); 2139 if (error > 0) 2140 error = EINVAL; 2141 else { 2142 error = 0; 2143 carp_setrun(sc, 0); 2144 } 2145 break; 2146 2147 case SIOCGVH: 2148 memset(&carpr, 0, sizeof(carpr)); 2149 if (sc->sc_carpdev != NULL) 2150 strlcpy(carpr.carpr_carpdev, sc->sc_carpdev->if_xname, 2151 IFNAMSIZ); 2152 carpr.carpr_state = sc->sc_state; 2153 carpr.carpr_vhid = sc->sc_vhid; 2154 carpr.carpr_advbase = sc->sc_advbase; 2155 carpr.carpr_advskew = sc->sc_advskew; 2156 2157 if ((l != NULL) && (error = kauth_authorize_network(l->l_cred, 2158 KAUTH_NETWORK_INTERFACE, 2159 KAUTH_REQ_NETWORK_INTERFACE_SETPRIV, ifp, (void *)cmd, 2160 NULL)) == 0) 2161 memcpy(carpr.carpr_key, sc->sc_key, 2162 sizeof(carpr.carpr_key)); 2163 error = copyout(&carpr, ifr->ifr_data, sizeof(carpr)); 2164 break; 2165 2166 case SIOCADDMULTI: 2167 error = carp_ether_addmulti(sc, ifr); 2168 break; 2169 2170 case SIOCDELMULTI: 2171 error = carp_ether_delmulti(sc, ifr); 2172 break; 2173 2174 case SIOCSIFCAP: 2175 if ((error = ifioctl_common(ifp, cmd, data)) == ENETRESET) 2176 error = 0; 2177 break; 2178 2179 default: 2180 error = ether_ioctl(ifp, cmd, data); 2181 } 2182 2183 carp_hmac_prepare(sc); 2184 return (error); 2185 } 2186 2187 2188 /* 2189 * Start output on carp interface. This function should never be called. 2190 */ 2191 static void 2192 carp_start(struct ifnet *ifp) 2193 { 2194 #ifdef DEBUG 2195 printf("%s: start called\n", ifp->if_xname); 2196 #endif 2197 } 2198 2199 int 2200 carp_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa, 2201 const struct rtentry *rt) 2202 { 2203 struct carp_softc *sc = ((struct carp_softc *)ifp->if_softc); 2204 KASSERT(KERNEL_LOCKED_P()); 2205 2206 if (sc->sc_carpdev != NULL && sc->sc_state == MASTER) { 2207 return if_output_lock(sc->sc_carpdev, ifp, m, sa, rt); 2208 } else { 2209 m_freem(m); 2210 return (ENETUNREACH); 2211 } 2212 } 2213 2214 static void 2215 carp_set_state(struct carp_softc *sc, int state) 2216 { 2217 static const char *carp_states[] = { CARP_STATES }; 2218 int link_state; 2219 2220 if (sc->sc_state == state) 2221 return; 2222 2223 CARP_LOG(sc, ("state transition from: %s -> to: %s", carp_states[sc->sc_state], carp_states[state])); 2224 2225 sc->sc_state = state; 2226 switch (state) { 2227 case BACKUP: 2228 link_state = LINK_STATE_DOWN; 2229 break; 2230 case MASTER: 2231 link_state = LINK_STATE_UP; 2232 break; 2233 default: 2234 link_state = LINK_STATE_UNKNOWN; 2235 break; 2236 } 2237 /* 2238 * The lock is needed to serialize a call of 2239 * if_link_state_change_softint from here and a call from softint. 2240 */ 2241 KERNEL_LOCK(1, NULL); 2242 if_link_state_change_softint(&sc->sc_if, link_state); 2243 KERNEL_UNLOCK_ONE(NULL); 2244 } 2245 2246 void 2247 carp_carpdev_state(void *v) 2248 { 2249 struct carp_if *cif; 2250 struct carp_softc *sc; 2251 struct ifnet *ifp = v; 2252 2253 if (ifp->if_type == IFT_CARP) 2254 return; 2255 2256 cif = (struct carp_if *)ifp->if_carp; 2257 2258 TAILQ_FOREACH(sc, &cif->vhif_vrs, sc_list) { 2259 int suppressed = sc->sc_suppress; 2260 2261 if (sc->sc_carpdev->if_link_state == LINK_STATE_DOWN || 2262 !(sc->sc_carpdev->if_flags & IFF_UP)) { 2263 sc->sc_if.if_flags &= ~IFF_RUNNING; 2264 callout_stop(&sc->sc_ad_tmo); 2265 callout_stop(&sc->sc_md_tmo); 2266 callout_stop(&sc->sc_md6_tmo); 2267 carp_set_state(sc, INIT); 2268 sc->sc_suppress = 1; 2269 carp_setrun(sc, 0); 2270 if (!suppressed) { 2271 carp_suppress_preempt++; 2272 if (carp_suppress_preempt == 1) 2273 carp_send_ad_all(); 2274 } 2275 } else { 2276 carp_set_state(sc, INIT); 2277 sc->sc_suppress = 0; 2278 carp_setrun(sc, 0); 2279 if (suppressed) 2280 carp_suppress_preempt--; 2281 } 2282 } 2283 } 2284 2285 static int 2286 carp_ether_addmulti(struct carp_softc *sc, struct ifreq *ifr) 2287 { 2288 const struct sockaddr *sa = ifreq_getaddr(SIOCADDMULTI, ifr); 2289 struct ifnet *ifp; 2290 struct carp_mc_entry *mc; 2291 u_int8_t addrlo[ETHER_ADDR_LEN], addrhi[ETHER_ADDR_LEN]; 2292 int error; 2293 2294 ifp = sc->sc_carpdev; 2295 if (ifp == NULL) 2296 return (EINVAL); 2297 2298 error = ether_addmulti(sa, &sc->sc_ac); 2299 if (error != ENETRESET) 2300 return (error); 2301 2302 /* 2303 * This is new multicast address. We have to tell parent 2304 * about it. Also, remember this multicast address so that 2305 * we can delete them on unconfigure. 2306 */ 2307 mc = malloc(sizeof(struct carp_mc_entry), M_DEVBUF, M_NOWAIT); 2308 if (mc == NULL) { 2309 error = ENOMEM; 2310 goto alloc_failed; 2311 } 2312 2313 /* 2314 * As ether_addmulti() returns ENETRESET, following two 2315 * statement shouldn't fail. 2316 */ 2317 (void)ether_multiaddr(sa, addrlo, addrhi); 2318 2319 ETHER_LOCK(&sc->sc_ac); 2320 mc->mc_enm = ether_lookup_multi(addrlo, addrhi, &sc->sc_ac); 2321 ETHER_UNLOCK(&sc->sc_ac); 2322 2323 memcpy(&mc->mc_addr, sa, sa->sa_len); 2324 LIST_INSERT_HEAD(&sc->carp_mc_listhead, mc, mc_entries); 2325 2326 error = if_mcast_op(ifp, SIOCADDMULTI, sa); 2327 if (error != 0) 2328 goto ioctl_failed; 2329 2330 return (error); 2331 2332 ioctl_failed: 2333 LIST_REMOVE(mc, mc_entries); 2334 free(mc, M_DEVBUF); 2335 alloc_failed: 2336 (void)ether_delmulti(sa, &sc->sc_ac); 2337 2338 return (error); 2339 } 2340 2341 static int 2342 carp_ether_delmulti(struct carp_softc *sc, struct ifreq *ifr) 2343 { 2344 const struct sockaddr *sa = ifreq_getaddr(SIOCDELMULTI, ifr); 2345 struct ifnet *ifp; 2346 struct ether_multi *enm; 2347 struct carp_mc_entry *mc; 2348 u_int8_t addrlo[ETHER_ADDR_LEN], addrhi[ETHER_ADDR_LEN]; 2349 int error; 2350 2351 ifp = sc->sc_carpdev; 2352 if (ifp == NULL) 2353 return (EINVAL); 2354 2355 /* 2356 * Find a key to lookup carp_mc_entry. We have to do this 2357 * before calling ether_delmulti for obvious reason. 2358 */ 2359 if ((error = ether_multiaddr(sa, addrlo, addrhi)) != 0) 2360 return (error); 2361 2362 ETHER_LOCK(&sc->sc_ac); 2363 enm = ether_lookup_multi(addrlo, addrhi, &sc->sc_ac); 2364 ETHER_UNLOCK(&sc->sc_ac); 2365 if (enm == NULL) 2366 return (EINVAL); 2367 2368 LIST_FOREACH(mc, &sc->carp_mc_listhead, mc_entries) 2369 if (mc->mc_enm == enm) 2370 break; 2371 2372 /* We won't delete entries we didn't add */ 2373 if (mc == NULL) 2374 return (EINVAL); 2375 2376 error = ether_delmulti(sa, &sc->sc_ac); 2377 if (error != ENETRESET) 2378 return (error); 2379 2380 /* We no longer use this multicast address. Tell parent so. */ 2381 error = if_mcast_op(ifp, SIOCDELMULTI, sa); 2382 if (error == 0) { 2383 /* And forget about this address. */ 2384 LIST_REMOVE(mc, mc_entries); 2385 free(mc, M_DEVBUF); 2386 } else 2387 (void)ether_addmulti(sa, &sc->sc_ac); 2388 return (error); 2389 } 2390 2391 /* 2392 * Delete any multicast address we have asked to add from parent 2393 * interface. Called when the carp is being unconfigured. 2394 */ 2395 static void 2396 carp_ether_purgemulti(struct carp_softc *sc) 2397 { 2398 struct ifnet *ifp = sc->sc_carpdev; /* Parent. */ 2399 struct carp_mc_entry *mc; 2400 2401 if (ifp == NULL) 2402 return; 2403 2404 while ((mc = LIST_FIRST(&sc->carp_mc_listhead)) != NULL) { 2405 (void)if_mcast_op(ifp, SIOCDELMULTI, sstosa(&mc->mc_addr)); 2406 LIST_REMOVE(mc, mc_entries); 2407 free(mc, M_DEVBUF); 2408 } 2409 } 2410 2411 static int 2412 sysctl_net_inet_carp_stats(SYSCTLFN_ARGS) 2413 { 2414 2415 return (NETSTAT_SYSCTL(carpstat_percpu, CARP_NSTATS)); 2416 } 2417 2418 void 2419 carp_init(void) 2420 { 2421 2422 sysctl_net_inet_carp_setup(NULL); 2423 #ifdef MBUFTRACE 2424 MOWNER_ATTACH(&carp_proto_mowner_rx); 2425 MOWNER_ATTACH(&carp_proto_mowner_tx); 2426 MOWNER_ATTACH(&carp_proto6_mowner_rx); 2427 MOWNER_ATTACH(&carp_proto6_mowner_tx); 2428 #endif 2429 2430 carp_wqinput = wqinput_create("carp", _carp_proto_input); 2431 #ifdef INET6 2432 carp6_wqinput = wqinput_create("carp6", _carp6_proto_input); 2433 #endif 2434 } 2435 2436 static void 2437 sysctl_net_inet_carp_setup(struct sysctllog **clog) 2438 { 2439 2440 sysctl_createv(clog, 0, NULL, NULL, 2441 CTLFLAG_PERMANENT, 2442 CTLTYPE_NODE, "inet", NULL, 2443 NULL, 0, NULL, 0, 2444 CTL_NET, PF_INET, CTL_EOL); 2445 sysctl_createv(clog, 0, NULL, NULL, 2446 CTLFLAG_PERMANENT, 2447 CTLTYPE_NODE, "carp", 2448 SYSCTL_DESCR("CARP related settings"), 2449 NULL, 0, NULL, 0, 2450 CTL_NET, PF_INET, IPPROTO_CARP, CTL_EOL); 2451 2452 sysctl_createv(clog, 0, NULL, NULL, 2453 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2454 CTLTYPE_INT, "preempt", 2455 SYSCTL_DESCR("Enable CARP Preempt"), 2456 NULL, 0, &carp_opts[CARPCTL_PREEMPT], 0, 2457 CTL_NET, PF_INET, IPPROTO_CARP, 2458 CTL_CREATE, CTL_EOL); 2459 sysctl_createv(clog, 0, NULL, NULL, 2460 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2461 CTLTYPE_INT, "arpbalance", 2462 SYSCTL_DESCR("Enable ARP balancing"), 2463 NULL, 0, &carp_opts[CARPCTL_ARPBALANCE], 0, 2464 CTL_NET, PF_INET, IPPROTO_CARP, 2465 CTL_CREATE, CTL_EOL); 2466 sysctl_createv(clog, 0, NULL, NULL, 2467 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2468 CTLTYPE_INT, "allow", 2469 SYSCTL_DESCR("Enable CARP"), 2470 NULL, 0, &carp_opts[CARPCTL_ALLOW], 0, 2471 CTL_NET, PF_INET, IPPROTO_CARP, 2472 CTL_CREATE, CTL_EOL); 2473 sysctl_createv(clog, 0, NULL, NULL, 2474 CTLFLAG_PERMANENT|CTLFLAG_READWRITE, 2475 CTLTYPE_INT, "log", 2476 SYSCTL_DESCR("CARP logging"), 2477 NULL, 0, &carp_opts[CARPCTL_LOG], 0, 2478 CTL_NET, PF_INET, IPPROTO_CARP, 2479 CTL_CREATE, CTL_EOL); 2480 sysctl_createv(clog, 0, NULL, NULL, 2481 CTLFLAG_PERMANENT, 2482 CTLTYPE_STRUCT, "stats", 2483 SYSCTL_DESCR("CARP statistics"), 2484 sysctl_net_inet_carp_stats, 0, NULL, 0, 2485 CTL_NET, PF_INET, IPPROTO_CARP, CARPCTL_STATS, 2486 CTL_EOL); 2487 } 2488