1 /*- 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. 4 * All rights reserved. 5 * 6 * Redistribution and use in source and binary forms, with or without 7 * modification, are permitted provided that the following conditions 8 * are met: 9 * 1. Redistributions of source code must retain the above copyright 10 * notice, this list of conditions and the following disclaimer. 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in the 13 * documentation and/or other materials provided with the distribution. 14 * 4. Neither the name of the University nor the names of its contributors 15 * may be used to endorse or promote products derived from this software 16 * without specific prior written permission. 17 * 18 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 21 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 28 * SUCH DAMAGE. 29 * 30 * @(#)raw_ip.c 8.7 (Berkeley) 5/15/95 31 */ 32 33 #include <sys/cdefs.h> 34 __FBSDID("$FreeBSD$"); 35 36 #include "opt_inet.h" 37 #include "opt_inet6.h" 38 #include "opt_ipsec.h" 39 40 #include <sys/param.h> 41 #include <sys/jail.h> 42 #include <sys/kernel.h> 43 #include <sys/lock.h> 44 #include <sys/malloc.h> 45 #include <sys/mbuf.h> 46 #include <sys/priv.h> 47 #include <sys/proc.h> 48 #include <sys/protosw.h> 49 #include <sys/rwlock.h> 50 #include <sys/signalvar.h> 51 #include <sys/socket.h> 52 #include <sys/socketvar.h> 53 #include <sys/sx.h> 54 #include <sys/sysctl.h> 55 #include <sys/systm.h> 56 57 #include <vm/uma.h> 58 59 #include <net/if.h> 60 #include <net/if_var.h> 61 #include <net/route.h> 62 #include <net/vnet.h> 63 64 #include <netinet/in.h> 65 #include <netinet/in_systm.h> 66 #include <netinet/in_pcb.h> 67 #include <netinet/in_var.h> 68 #include <netinet/if_ether.h> 69 #include <netinet/ip.h> 70 #include <netinet/ip_var.h> 71 #include <netinet/ip_mroute.h> 72 73 #ifdef IPSEC 74 #include <netipsec/ipsec.h> 75 #endif /*IPSEC*/ 76 77 #include <security/mac/mac_framework.h> 78 79 VNET_DEFINE(int, ip_defttl) = IPDEFTTL; 80 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 81 &VNET_NAME(ip_defttl), 0, 82 "Maximum TTL on IP packets"); 83 84 VNET_DEFINE(struct inpcbhead, ripcb); 85 VNET_DEFINE(struct inpcbinfo, ripcbinfo); 86 87 #define V_ripcb VNET(ripcb) 88 #define V_ripcbinfo VNET(ripcbinfo) 89 90 /* 91 * Control and data hooks for ipfw, dummynet, divert and so on. 92 * The data hooks are not used here but it is convenient 93 * to keep them all in one place. 94 */ 95 VNET_DEFINE(ip_fw_chk_ptr_t, ip_fw_chk_ptr) = NULL; 96 VNET_DEFINE(ip_fw_ctl_ptr_t, ip_fw_ctl_ptr) = NULL; 97 98 int (*ip_dn_ctl_ptr)(struct sockopt *); 99 int (*ip_dn_io_ptr)(struct mbuf **, int, struct ip_fw_args *); 100 void (*ip_divert_ptr)(struct mbuf *, int); 101 int (*ng_ipfw_input_p)(struct mbuf **, int, 102 struct ip_fw_args *, int); 103 104 #ifdef INET 105 /* 106 * Hooks for multicast routing. They all default to NULL, so leave them not 107 * initialized and rely on BSS being set to 0. 108 */ 109 110 /* 111 * The socket used to communicate with the multicast routing daemon. 112 */ 113 VNET_DEFINE(struct socket *, ip_mrouter); 114 115 /* 116 * The various mrouter and rsvp functions. 117 */ 118 int (*ip_mrouter_set)(struct socket *, struct sockopt *); 119 int (*ip_mrouter_get)(struct socket *, struct sockopt *); 120 int (*ip_mrouter_done)(void); 121 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 122 struct ip_moptions *); 123 int (*mrt_ioctl)(u_long, caddr_t, int); 124 int (*legal_vif_num)(int); 125 u_long (*ip_mcast_src)(int); 126 127 int (*rsvp_input_p)(struct mbuf **, int *, int); 128 int (*ip_rsvp_vif)(struct socket *, struct sockopt *); 129 void (*ip_rsvp_force_done)(struct socket *); 130 #endif /* INET */ 131 132 u_long rip_sendspace = 9216; 133 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW, 134 &rip_sendspace, 0, "Maximum outgoing raw IP datagram size"); 135 136 u_long rip_recvspace = 9216; 137 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW, 138 &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams"); 139 140 /* 141 * Hash functions 142 */ 143 144 #define INP_PCBHASH_RAW_SIZE 256 145 #define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \ 146 (((proto) + (laddr) + (faddr)) % (mask) + 1) 147 148 #ifdef INET 149 static void 150 rip_inshash(struct inpcb *inp) 151 { 152 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 153 struct inpcbhead *pcbhash; 154 int hash; 155 156 INP_INFO_WLOCK_ASSERT(pcbinfo); 157 INP_WLOCK_ASSERT(inp); 158 159 if (inp->inp_ip_p != 0 && 160 inp->inp_laddr.s_addr != INADDR_ANY && 161 inp->inp_faddr.s_addr != INADDR_ANY) { 162 hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr, 163 inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask); 164 } else 165 hash = 0; 166 pcbhash = &pcbinfo->ipi_hashbase[hash]; 167 LIST_INSERT_HEAD(pcbhash, inp, inp_hash); 168 } 169 170 static void 171 rip_delhash(struct inpcb *inp) 172 { 173 174 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 175 INP_WLOCK_ASSERT(inp); 176 177 LIST_REMOVE(inp, inp_hash); 178 } 179 #endif /* INET */ 180 181 /* 182 * Raw interface to IP protocol. 183 */ 184 185 /* 186 * Initialize raw connection block q. 187 */ 188 static void 189 rip_zone_change(void *tag) 190 { 191 192 uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets); 193 } 194 195 static int 196 rip_inpcb_init(void *mem, int size, int flags) 197 { 198 struct inpcb *inp = mem; 199 200 INP_LOCK_INIT(inp, "inp", "rawinp"); 201 return (0); 202 } 203 204 void 205 rip_init(void) 206 { 207 208 in_pcbinfo_init(&V_ripcbinfo, "rip", &V_ripcb, INP_PCBHASH_RAW_SIZE, 209 1, "ripcb", rip_inpcb_init, NULL, UMA_ZONE_NOFREE, 210 IPI_HASHFIELDS_NONE); 211 EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL, 212 EVENTHANDLER_PRI_ANY); 213 } 214 215 #ifdef VIMAGE 216 void 217 rip_destroy(void) 218 { 219 220 in_pcbinfo_destroy(&V_ripcbinfo); 221 } 222 #endif 223 224 #ifdef INET 225 static int 226 rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n, 227 struct sockaddr_in *ripsrc) 228 { 229 int policyfail = 0; 230 231 INP_LOCK_ASSERT(last); 232 233 #ifdef IPSEC 234 /* check AH/ESP integrity. */ 235 if (ipsec4_in_reject(n, last)) { 236 policyfail = 1; 237 } 238 #endif /* IPSEC */ 239 #ifdef MAC 240 if (!policyfail && mac_inpcb_check_deliver(last, n) != 0) 241 policyfail = 1; 242 #endif 243 /* Check the minimum TTL for socket. */ 244 if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl) 245 policyfail = 1; 246 if (!policyfail) { 247 struct mbuf *opts = NULL; 248 struct socket *so; 249 250 so = last->inp_socket; 251 if ((last->inp_flags & INP_CONTROLOPTS) || 252 (so->so_options & (SO_TIMESTAMP | SO_BINTIME))) 253 ip_savecontrol(last, &opts, ip, n); 254 SOCKBUF_LOCK(&so->so_rcv); 255 if (sbappendaddr_locked(&so->so_rcv, 256 (struct sockaddr *)ripsrc, n, opts) == 0) { 257 /* should notify about lost packet */ 258 m_freem(n); 259 if (opts) 260 m_freem(opts); 261 SOCKBUF_UNLOCK(&so->so_rcv); 262 } else 263 sorwakeup_locked(so); 264 } else 265 m_freem(n); 266 return (policyfail); 267 } 268 269 /* 270 * Setup generic address and protocol structures for raw_input routine, then 271 * pass them along with mbuf chain. 272 */ 273 int 274 rip_input(struct mbuf **mp, int *offp, int proto) 275 { 276 struct ifnet *ifp; 277 struct mbuf *m = *mp; 278 struct ip *ip = mtod(m, struct ip *); 279 struct inpcb *inp, *last; 280 struct sockaddr_in ripsrc; 281 int hash; 282 283 *mp = NULL; 284 285 bzero(&ripsrc, sizeof(ripsrc)); 286 ripsrc.sin_len = sizeof(ripsrc); 287 ripsrc.sin_family = AF_INET; 288 ripsrc.sin_addr = ip->ip_src; 289 last = NULL; 290 291 ifp = m->m_pkthdr.rcvif; 292 /* 293 * Applications on raw sockets expect host byte order. 294 */ 295 ip->ip_len = ntohs(ip->ip_len); 296 ip->ip_off = ntohs(ip->ip_off); 297 298 hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr, 299 ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask); 300 INP_INFO_RLOCK(&V_ripcbinfo); 301 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) { 302 if (inp->inp_ip_p != proto) 303 continue; 304 #ifdef INET6 305 /* XXX inp locking */ 306 if ((inp->inp_vflag & INP_IPV4) == 0) 307 continue; 308 #endif 309 if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr) 310 continue; 311 if (inp->inp_faddr.s_addr != ip->ip_src.s_addr) 312 continue; 313 if (jailed_without_vnet(inp->inp_cred)) { 314 /* 315 * XXX: If faddr was bound to multicast group, 316 * jailed raw socket will drop datagram. 317 */ 318 if (prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) 319 continue; 320 } 321 if (last != NULL) { 322 struct mbuf *n; 323 324 n = m_copy(m, 0, (int)M_COPYALL); 325 if (n != NULL) 326 (void) rip_append(last, ip, n, &ripsrc); 327 /* XXX count dropped packet */ 328 INP_RUNLOCK(last); 329 } 330 INP_RLOCK(inp); 331 last = inp; 332 } 333 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) { 334 if (inp->inp_ip_p && inp->inp_ip_p != proto) 335 continue; 336 #ifdef INET6 337 /* XXX inp locking */ 338 if ((inp->inp_vflag & INP_IPV4) == 0) 339 continue; 340 #endif 341 if (!in_nullhost(inp->inp_laddr) && 342 !in_hosteq(inp->inp_laddr, ip->ip_dst)) 343 continue; 344 if (!in_nullhost(inp->inp_faddr) && 345 !in_hosteq(inp->inp_faddr, ip->ip_src)) 346 continue; 347 if (jailed_without_vnet(inp->inp_cred)) { 348 /* 349 * Allow raw socket in jail to receive multicast; 350 * assume process had PRIV_NETINET_RAW at attach, 351 * and fall through into normal filter path if so. 352 */ 353 if (!IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 354 prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) 355 continue; 356 } 357 /* 358 * If this raw socket has multicast state, and we 359 * have received a multicast, check if this socket 360 * should receive it, as multicast filtering is now 361 * the responsibility of the transport layer. 362 */ 363 if (inp->inp_moptions != NULL && 364 IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 365 /* 366 * If the incoming datagram is for IGMP, allow it 367 * through unconditionally to the raw socket. 368 * 369 * In the case of IGMPv2, we may not have explicitly 370 * joined the group, and may have set IFF_ALLMULTI 371 * on the interface. imo_multi_filter() may discard 372 * control traffic we actually need to see. 373 * 374 * Userland multicast routing daemons should continue 375 * filter the control traffic appropriately. 376 */ 377 int blocked; 378 379 blocked = MCAST_PASS; 380 if (proto != IPPROTO_IGMP) { 381 struct sockaddr_in group; 382 383 bzero(&group, sizeof(struct sockaddr_in)); 384 group.sin_len = sizeof(struct sockaddr_in); 385 group.sin_family = AF_INET; 386 group.sin_addr = ip->ip_dst; 387 388 blocked = imo_multi_filter(inp->inp_moptions, 389 ifp, 390 (struct sockaddr *)&group, 391 (struct sockaddr *)&ripsrc); 392 } 393 394 if (blocked != MCAST_PASS) { 395 IPSTAT_INC(ips_notmember); 396 continue; 397 } 398 } 399 if (last != NULL) { 400 struct mbuf *n; 401 402 n = m_copy(m, 0, (int)M_COPYALL); 403 if (n != NULL) 404 (void) rip_append(last, ip, n, &ripsrc); 405 /* XXX count dropped packet */ 406 INP_RUNLOCK(last); 407 } 408 INP_RLOCK(inp); 409 last = inp; 410 } 411 INP_INFO_RUNLOCK(&V_ripcbinfo); 412 if (last != NULL) { 413 if (rip_append(last, ip, m, &ripsrc) != 0) 414 IPSTAT_INC(ips_delivered); 415 INP_RUNLOCK(last); 416 } else { 417 m_freem(m); 418 IPSTAT_INC(ips_noproto); 419 IPSTAT_DEC(ips_delivered); 420 } 421 return (IPPROTO_DONE); 422 } 423 424 /* 425 * Generate IP header and pass packet to ip_output. Tack on options user may 426 * have setup with control call. 427 */ 428 int 429 rip_output(struct mbuf *m, struct socket *so, u_long dst) 430 { 431 struct ip *ip; 432 int error; 433 struct inpcb *inp = sotoinpcb(so); 434 int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) | 435 IP_ALLOWBROADCAST; 436 437 /* 438 * If the user handed us a complete IP packet, use it. Otherwise, 439 * allocate an mbuf for a header and fill it in. 440 */ 441 if ((inp->inp_flags & INP_HDRINCL) == 0) { 442 if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) { 443 m_freem(m); 444 return(EMSGSIZE); 445 } 446 M_PREPEND(m, sizeof(struct ip), M_NOWAIT); 447 if (m == NULL) 448 return(ENOBUFS); 449 450 INP_RLOCK(inp); 451 ip = mtod(m, struct ip *); 452 ip->ip_tos = inp->inp_ip_tos; 453 if (inp->inp_flags & INP_DONTFRAG) 454 ip->ip_off = htons(IP_DF); 455 else 456 ip->ip_off = htons(0); 457 ip->ip_p = inp->inp_ip_p; 458 ip->ip_len = htons(m->m_pkthdr.len); 459 ip->ip_src = inp->inp_laddr; 460 ip->ip_dst.s_addr = dst; 461 if (jailed(inp->inp_cred)) { 462 /* 463 * prison_local_ip4() would be good enough but would 464 * let a source of INADDR_ANY pass, which we do not 465 * want to see from jails. 466 */ 467 if (ip->ip_src.s_addr == INADDR_ANY) { 468 error = in_pcbladdr(inp, &ip->ip_dst, &ip->ip_src, 469 inp->inp_cred); 470 } else { 471 error = prison_local_ip4(inp->inp_cred, 472 &ip->ip_src); 473 } 474 if (error != 0) { 475 INP_RUNLOCK(inp); 476 m_freem(m); 477 return (error); 478 } 479 } 480 ip->ip_ttl = inp->inp_ip_ttl; 481 } else { 482 if (m->m_pkthdr.len > IP_MAXPACKET) { 483 m_freem(m); 484 return(EMSGSIZE); 485 } 486 INP_RLOCK(inp); 487 ip = mtod(m, struct ip *); 488 error = prison_check_ip4(inp->inp_cred, &ip->ip_src); 489 if (error != 0) { 490 INP_RUNLOCK(inp); 491 m_freem(m); 492 return (error); 493 } 494 495 /* 496 * Don't allow both user specified and setsockopt options, 497 * and don't allow packet length sizes that will crash. 498 */ 499 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) 500 || (ip->ip_len > m->m_pkthdr.len) 501 || (ip->ip_len < (ip->ip_hl << 2))) { 502 INP_RUNLOCK(inp); 503 m_freem(m); 504 return (EINVAL); 505 } 506 if (ip->ip_id == 0) 507 ip->ip_id = ip_newid(); 508 509 /* 510 * Applications on raw sockets pass us packets 511 * in host byte order. 512 */ 513 ip->ip_len = htons(ip->ip_len); 514 ip->ip_off = htons(ip->ip_off); 515 516 /* 517 * XXX prevent ip_output from overwriting header fields. 518 */ 519 flags |= IP_RAWOUTPUT; 520 IPSTAT_INC(ips_rawout); 521 } 522 523 if (inp->inp_flags & INP_ONESBCAST) 524 flags |= IP_SENDONES; 525 526 #ifdef MAC 527 mac_inpcb_create_mbuf(inp, m); 528 #endif 529 530 error = ip_output(m, inp->inp_options, NULL, flags, 531 inp->inp_moptions, inp); 532 INP_RUNLOCK(inp); 533 return (error); 534 } 535 536 /* 537 * Raw IP socket option processing. 538 * 539 * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could 540 * only be created by a privileged process, and as such, socket option 541 * operations to manage system properties on any raw socket were allowed to 542 * take place without explicit additional access control checks. However, 543 * raw sockets can now also be created in jail(), and therefore explicit 544 * checks are now required. Likewise, raw sockets can be used by a process 545 * after it gives up privilege, so some caution is required. For options 546 * passed down to the IP layer via ip_ctloutput(), checks are assumed to be 547 * performed in ip_ctloutput() and therefore no check occurs here. 548 * Unilaterally checking priv_check() here breaks normal IP socket option 549 * operations on raw sockets. 550 * 551 * When adding new socket options here, make sure to add access control 552 * checks here as necessary. 553 * 554 * XXX-BZ inp locking? 555 */ 556 int 557 rip_ctloutput(struct socket *so, struct sockopt *sopt) 558 { 559 struct inpcb *inp = sotoinpcb(so); 560 int error, optval; 561 562 if (sopt->sopt_level != IPPROTO_IP) { 563 if ((sopt->sopt_level == SOL_SOCKET) && 564 (sopt->sopt_name == SO_SETFIB)) { 565 inp->inp_inc.inc_fibnum = so->so_fibnum; 566 return (0); 567 } 568 return (EINVAL); 569 } 570 571 error = 0; 572 switch (sopt->sopt_dir) { 573 case SOPT_GET: 574 switch (sopt->sopt_name) { 575 case IP_HDRINCL: 576 optval = inp->inp_flags & INP_HDRINCL; 577 error = sooptcopyout(sopt, &optval, sizeof optval); 578 break; 579 580 case IP_FW3: /* generic ipfw v.3 functions */ 581 case IP_FW_ADD: /* ADD actually returns the body... */ 582 case IP_FW_GET: 583 case IP_FW_TABLE_GETSIZE: 584 case IP_FW_TABLE_LIST: 585 case IP_FW_NAT_GET_CONFIG: 586 case IP_FW_NAT_GET_LOG: 587 if (V_ip_fw_ctl_ptr != NULL) 588 error = V_ip_fw_ctl_ptr(sopt); 589 else 590 error = ENOPROTOOPT; 591 break; 592 593 case IP_DUMMYNET3: /* generic dummynet v.3 functions */ 594 case IP_DUMMYNET_GET: 595 if (ip_dn_ctl_ptr != NULL) 596 error = ip_dn_ctl_ptr(sopt); 597 else 598 error = ENOPROTOOPT; 599 break ; 600 601 case MRT_INIT: 602 case MRT_DONE: 603 case MRT_ADD_VIF: 604 case MRT_DEL_VIF: 605 case MRT_ADD_MFC: 606 case MRT_DEL_MFC: 607 case MRT_VERSION: 608 case MRT_ASSERT: 609 case MRT_API_SUPPORT: 610 case MRT_API_CONFIG: 611 case MRT_ADD_BW_UPCALL: 612 case MRT_DEL_BW_UPCALL: 613 error = priv_check(curthread, PRIV_NETINET_MROUTE); 614 if (error != 0) 615 return (error); 616 error = ip_mrouter_get ? ip_mrouter_get(so, sopt) : 617 EOPNOTSUPP; 618 break; 619 620 default: 621 error = ip_ctloutput(so, sopt); 622 break; 623 } 624 break; 625 626 case SOPT_SET: 627 switch (sopt->sopt_name) { 628 case IP_HDRINCL: 629 error = sooptcopyin(sopt, &optval, sizeof optval, 630 sizeof optval); 631 if (error) 632 break; 633 if (optval) 634 inp->inp_flags |= INP_HDRINCL; 635 else 636 inp->inp_flags &= ~INP_HDRINCL; 637 break; 638 639 case IP_FW3: /* generic ipfw v.3 functions */ 640 case IP_FW_ADD: 641 case IP_FW_DEL: 642 case IP_FW_FLUSH: 643 case IP_FW_ZERO: 644 case IP_FW_RESETLOG: 645 case IP_FW_TABLE_ADD: 646 case IP_FW_TABLE_DEL: 647 case IP_FW_TABLE_FLUSH: 648 case IP_FW_NAT_CFG: 649 case IP_FW_NAT_DEL: 650 if (V_ip_fw_ctl_ptr != NULL) 651 error = V_ip_fw_ctl_ptr(sopt); 652 else 653 error = ENOPROTOOPT; 654 break; 655 656 case IP_DUMMYNET3: /* generic dummynet v.3 functions */ 657 case IP_DUMMYNET_CONFIGURE: 658 case IP_DUMMYNET_DEL: 659 case IP_DUMMYNET_FLUSH: 660 if (ip_dn_ctl_ptr != NULL) 661 error = ip_dn_ctl_ptr(sopt); 662 else 663 error = ENOPROTOOPT ; 664 break ; 665 666 case IP_RSVP_ON: 667 error = priv_check(curthread, PRIV_NETINET_MROUTE); 668 if (error != 0) 669 return (error); 670 error = ip_rsvp_init(so); 671 break; 672 673 case IP_RSVP_OFF: 674 error = priv_check(curthread, PRIV_NETINET_MROUTE); 675 if (error != 0) 676 return (error); 677 error = ip_rsvp_done(); 678 break; 679 680 case IP_RSVP_VIF_ON: 681 case IP_RSVP_VIF_OFF: 682 error = priv_check(curthread, PRIV_NETINET_MROUTE); 683 if (error != 0) 684 return (error); 685 error = ip_rsvp_vif ? 686 ip_rsvp_vif(so, sopt) : EINVAL; 687 break; 688 689 case MRT_INIT: 690 case MRT_DONE: 691 case MRT_ADD_VIF: 692 case MRT_DEL_VIF: 693 case MRT_ADD_MFC: 694 case MRT_DEL_MFC: 695 case MRT_VERSION: 696 case MRT_ASSERT: 697 case MRT_API_SUPPORT: 698 case MRT_API_CONFIG: 699 case MRT_ADD_BW_UPCALL: 700 case MRT_DEL_BW_UPCALL: 701 error = priv_check(curthread, PRIV_NETINET_MROUTE); 702 if (error != 0) 703 return (error); 704 error = ip_mrouter_set ? ip_mrouter_set(so, sopt) : 705 EOPNOTSUPP; 706 break; 707 708 default: 709 error = ip_ctloutput(so, sopt); 710 break; 711 } 712 break; 713 } 714 715 return (error); 716 } 717 718 /* 719 * This function exists solely to receive the PRC_IFDOWN messages which are 720 * sent by if_down(). It looks for an ifaddr whose ifa_addr is sa, and calls 721 * in_ifadown() to remove all routes corresponding to that address. It also 722 * receives the PRC_IFUP messages from if_up() and reinstalls the interface 723 * routes. 724 */ 725 void 726 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip) 727 { 728 struct in_ifaddr *ia; 729 struct ifnet *ifp; 730 int err; 731 int flags; 732 733 switch (cmd) { 734 case PRC_IFDOWN: 735 IN_IFADDR_RLOCK(); 736 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 737 if (ia->ia_ifa.ifa_addr == sa 738 && (ia->ia_flags & IFA_ROUTE)) { 739 ifa_ref(&ia->ia_ifa); 740 IN_IFADDR_RUNLOCK(); 741 /* 742 * in_scrubprefix() kills the interface route. 743 */ 744 in_scrubprefix(ia, 0); 745 /* 746 * in_ifadown gets rid of all the rest of the 747 * routes. This is not quite the right thing 748 * to do, but at least if we are running a 749 * routing process they will come back. 750 */ 751 in_ifadown(&ia->ia_ifa, 0); 752 ifa_free(&ia->ia_ifa); 753 break; 754 } 755 } 756 if (ia == NULL) /* If ia matched, already unlocked. */ 757 IN_IFADDR_RUNLOCK(); 758 break; 759 760 case PRC_IFUP: 761 IN_IFADDR_RLOCK(); 762 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 763 if (ia->ia_ifa.ifa_addr == sa) 764 break; 765 } 766 if (ia == NULL || (ia->ia_flags & IFA_ROUTE)) { 767 IN_IFADDR_RUNLOCK(); 768 return; 769 } 770 ifa_ref(&ia->ia_ifa); 771 IN_IFADDR_RUNLOCK(); 772 flags = RTF_UP; 773 ifp = ia->ia_ifa.ifa_ifp; 774 775 if ((ifp->if_flags & IFF_LOOPBACK) 776 || (ifp->if_flags & IFF_POINTOPOINT)) 777 flags |= RTF_HOST; 778 779 err = ifa_del_loopback_route((struct ifaddr *)ia, sa); 780 781 err = rtinit(&ia->ia_ifa, RTM_ADD, flags); 782 if (err == 0) 783 ia->ia_flags |= IFA_ROUTE; 784 785 err = ifa_add_loopback_route((struct ifaddr *)ia, sa); 786 787 ifa_free(&ia->ia_ifa); 788 break; 789 } 790 } 791 792 static int 793 rip_attach(struct socket *so, int proto, struct thread *td) 794 { 795 struct inpcb *inp; 796 int error; 797 798 inp = sotoinpcb(so); 799 KASSERT(inp == NULL, ("rip_attach: inp != NULL")); 800 801 error = priv_check(td, PRIV_NETINET_RAW); 802 if (error) 803 return (error); 804 if (proto >= IPPROTO_MAX || proto < 0) 805 return EPROTONOSUPPORT; 806 error = soreserve(so, rip_sendspace, rip_recvspace); 807 if (error) 808 return (error); 809 INP_INFO_WLOCK(&V_ripcbinfo); 810 error = in_pcballoc(so, &V_ripcbinfo); 811 if (error) { 812 INP_INFO_WUNLOCK(&V_ripcbinfo); 813 return (error); 814 } 815 inp = (struct inpcb *)so->so_pcb; 816 inp->inp_vflag |= INP_IPV4; 817 inp->inp_ip_p = proto; 818 inp->inp_ip_ttl = V_ip_defttl; 819 rip_inshash(inp); 820 INP_INFO_WUNLOCK(&V_ripcbinfo); 821 INP_WUNLOCK(inp); 822 return (0); 823 } 824 825 static void 826 rip_detach(struct socket *so) 827 { 828 struct inpcb *inp; 829 830 inp = sotoinpcb(so); 831 KASSERT(inp != NULL, ("rip_detach: inp == NULL")); 832 KASSERT(inp->inp_faddr.s_addr == INADDR_ANY, 833 ("rip_detach: not closed")); 834 835 INP_INFO_WLOCK(&V_ripcbinfo); 836 INP_WLOCK(inp); 837 rip_delhash(inp); 838 if (so == V_ip_mrouter && ip_mrouter_done) 839 ip_mrouter_done(); 840 if (ip_rsvp_force_done) 841 ip_rsvp_force_done(so); 842 if (so == V_ip_rsvpd) 843 ip_rsvp_done(); 844 in_pcbdetach(inp); 845 in_pcbfree(inp); 846 INP_INFO_WUNLOCK(&V_ripcbinfo); 847 } 848 849 static void 850 rip_dodisconnect(struct socket *so, struct inpcb *inp) 851 { 852 struct inpcbinfo *pcbinfo; 853 854 pcbinfo = inp->inp_pcbinfo; 855 INP_INFO_WLOCK(pcbinfo); 856 INP_WLOCK(inp); 857 rip_delhash(inp); 858 inp->inp_faddr.s_addr = INADDR_ANY; 859 rip_inshash(inp); 860 SOCK_LOCK(so); 861 so->so_state &= ~SS_ISCONNECTED; 862 SOCK_UNLOCK(so); 863 INP_WUNLOCK(inp); 864 INP_INFO_WUNLOCK(pcbinfo); 865 } 866 867 static void 868 rip_abort(struct socket *so) 869 { 870 struct inpcb *inp; 871 872 inp = sotoinpcb(so); 873 KASSERT(inp != NULL, ("rip_abort: inp == NULL")); 874 875 rip_dodisconnect(so, inp); 876 } 877 878 static void 879 rip_close(struct socket *so) 880 { 881 struct inpcb *inp; 882 883 inp = sotoinpcb(so); 884 KASSERT(inp != NULL, ("rip_close: inp == NULL")); 885 886 rip_dodisconnect(so, inp); 887 } 888 889 static int 890 rip_disconnect(struct socket *so) 891 { 892 struct inpcb *inp; 893 894 if ((so->so_state & SS_ISCONNECTED) == 0) 895 return (ENOTCONN); 896 897 inp = sotoinpcb(so); 898 KASSERT(inp != NULL, ("rip_disconnect: inp == NULL")); 899 900 rip_dodisconnect(so, inp); 901 return (0); 902 } 903 904 static int 905 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 906 { 907 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 908 struct inpcb *inp; 909 int error; 910 911 if (nam->sa_len != sizeof(*addr)) 912 return (EINVAL); 913 914 error = prison_check_ip4(td->td_ucred, &addr->sin_addr); 915 if (error != 0) 916 return (error); 917 918 inp = sotoinpcb(so); 919 KASSERT(inp != NULL, ("rip_bind: inp == NULL")); 920 921 if (TAILQ_EMPTY(&V_ifnet) || 922 (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) || 923 (addr->sin_addr.s_addr && 924 (inp->inp_flags & INP_BINDANY) == 0 && 925 ifa_ifwithaddr_check((struct sockaddr *)addr) == 0)) 926 return (EADDRNOTAVAIL); 927 928 INP_INFO_WLOCK(&V_ripcbinfo); 929 INP_WLOCK(inp); 930 rip_delhash(inp); 931 inp->inp_laddr = addr->sin_addr; 932 rip_inshash(inp); 933 INP_WUNLOCK(inp); 934 INP_INFO_WUNLOCK(&V_ripcbinfo); 935 return (0); 936 } 937 938 static int 939 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 940 { 941 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 942 struct inpcb *inp; 943 944 if (nam->sa_len != sizeof(*addr)) 945 return (EINVAL); 946 if (TAILQ_EMPTY(&V_ifnet)) 947 return (EADDRNOTAVAIL); 948 if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) 949 return (EAFNOSUPPORT); 950 951 inp = sotoinpcb(so); 952 KASSERT(inp != NULL, ("rip_connect: inp == NULL")); 953 954 INP_INFO_WLOCK(&V_ripcbinfo); 955 INP_WLOCK(inp); 956 rip_delhash(inp); 957 inp->inp_faddr = addr->sin_addr; 958 rip_inshash(inp); 959 soisconnected(so); 960 INP_WUNLOCK(inp); 961 INP_INFO_WUNLOCK(&V_ripcbinfo); 962 return (0); 963 } 964 965 static int 966 rip_shutdown(struct socket *so) 967 { 968 struct inpcb *inp; 969 970 inp = sotoinpcb(so); 971 KASSERT(inp != NULL, ("rip_shutdown: inp == NULL")); 972 973 INP_WLOCK(inp); 974 socantsendmore(so); 975 INP_WUNLOCK(inp); 976 return (0); 977 } 978 979 static int 980 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 981 struct mbuf *control, struct thread *td) 982 { 983 struct inpcb *inp; 984 u_long dst; 985 986 inp = sotoinpcb(so); 987 KASSERT(inp != NULL, ("rip_send: inp == NULL")); 988 989 /* 990 * Note: 'dst' reads below are unlocked. 991 */ 992 if (so->so_state & SS_ISCONNECTED) { 993 if (nam) { 994 m_freem(m); 995 return (EISCONN); 996 } 997 dst = inp->inp_faddr.s_addr; /* Unlocked read. */ 998 } else { 999 if (nam == NULL) { 1000 m_freem(m); 1001 return (ENOTCONN); 1002 } 1003 dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr; 1004 } 1005 return (rip_output(m, so, dst)); 1006 } 1007 #endif /* INET */ 1008 1009 static int 1010 rip_pcblist(SYSCTL_HANDLER_ARGS) 1011 { 1012 int error, i, n; 1013 struct inpcb *inp, **inp_list; 1014 inp_gen_t gencnt; 1015 struct xinpgen xig; 1016 1017 /* 1018 * The process of preparing the TCB list is too time-consuming and 1019 * resource-intensive to repeat twice on every request. 1020 */ 1021 if (req->oldptr == 0) { 1022 n = V_ripcbinfo.ipi_count; 1023 n += imax(n / 8, 10); 1024 req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb); 1025 return (0); 1026 } 1027 1028 if (req->newptr != 0) 1029 return (EPERM); 1030 1031 /* 1032 * OK, now we're committed to doing something. 1033 */ 1034 INP_INFO_RLOCK(&V_ripcbinfo); 1035 gencnt = V_ripcbinfo.ipi_gencnt; 1036 n = V_ripcbinfo.ipi_count; 1037 INP_INFO_RUNLOCK(&V_ripcbinfo); 1038 1039 xig.xig_len = sizeof xig; 1040 xig.xig_count = n; 1041 xig.xig_gen = gencnt; 1042 xig.xig_sogen = so_gencnt; 1043 error = SYSCTL_OUT(req, &xig, sizeof xig); 1044 if (error) 1045 return (error); 1046 1047 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 1048 if (inp_list == 0) 1049 return (ENOMEM); 1050 1051 INP_INFO_RLOCK(&V_ripcbinfo); 1052 for (inp = LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n; 1053 inp = LIST_NEXT(inp, inp_list)) { 1054 INP_WLOCK(inp); 1055 if (inp->inp_gencnt <= gencnt && 1056 cr_canseeinpcb(req->td->td_ucred, inp) == 0) { 1057 in_pcbref(inp); 1058 inp_list[i++] = inp; 1059 } 1060 INP_WUNLOCK(inp); 1061 } 1062 INP_INFO_RUNLOCK(&V_ripcbinfo); 1063 n = i; 1064 1065 error = 0; 1066 for (i = 0; i < n; i++) { 1067 inp = inp_list[i]; 1068 INP_RLOCK(inp); 1069 if (inp->inp_gencnt <= gencnt) { 1070 struct xinpcb xi; 1071 1072 bzero(&xi, sizeof(xi)); 1073 xi.xi_len = sizeof xi; 1074 /* XXX should avoid extra copy */ 1075 bcopy(inp, &xi.xi_inp, sizeof *inp); 1076 if (inp->inp_socket) 1077 sotoxsocket(inp->inp_socket, &xi.xi_socket); 1078 INP_RUNLOCK(inp); 1079 error = SYSCTL_OUT(req, &xi, sizeof xi); 1080 } else 1081 INP_RUNLOCK(inp); 1082 } 1083 INP_INFO_WLOCK(&V_ripcbinfo); 1084 for (i = 0; i < n; i++) { 1085 inp = inp_list[i]; 1086 INP_RLOCK(inp); 1087 if (!in_pcbrele_rlocked(inp)) 1088 INP_RUNLOCK(inp); 1089 } 1090 INP_INFO_WUNLOCK(&V_ripcbinfo); 1091 1092 if (!error) { 1093 /* 1094 * Give the user an updated idea of our state. If the 1095 * generation differs from what we told her before, she knows 1096 * that something happened while we were processing this 1097 * request, and it might be necessary to retry. 1098 */ 1099 INP_INFO_RLOCK(&V_ripcbinfo); 1100 xig.xig_gen = V_ripcbinfo.ipi_gencnt; 1101 xig.xig_sogen = so_gencnt; 1102 xig.xig_count = V_ripcbinfo.ipi_count; 1103 INP_INFO_RUNLOCK(&V_ripcbinfo); 1104 error = SYSCTL_OUT(req, &xig, sizeof xig); 1105 } 1106 free(inp_list, M_TEMP); 1107 return (error); 1108 } 1109 1110 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, 1111 CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0, 1112 rip_pcblist, "S,xinpcb", "List of active raw IP sockets"); 1113 1114 #ifdef INET 1115 struct pr_usrreqs rip_usrreqs = { 1116 .pru_abort = rip_abort, 1117 .pru_attach = rip_attach, 1118 .pru_bind = rip_bind, 1119 .pru_connect = rip_connect, 1120 .pru_control = in_control, 1121 .pru_detach = rip_detach, 1122 .pru_disconnect = rip_disconnect, 1123 .pru_peeraddr = in_getpeeraddr, 1124 .pru_send = rip_send, 1125 .pru_shutdown = rip_shutdown, 1126 .pru_sockaddr = in_getsockaddr, 1127 .pru_sosetlabel = in_pcbsosetlabel, 1128 .pru_close = rip_close, 1129 }; 1130 #endif /* INET */ 1131