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_inet6.h" 37 #include "opt_ipsec.h" 38 39 #include <sys/param.h> 40 #include <sys/jail.h> 41 #include <sys/kernel.h> 42 #include <sys/lock.h> 43 #include <sys/malloc.h> 44 #include <sys/mbuf.h> 45 #include <sys/priv.h> 46 #include <sys/proc.h> 47 #include <sys/protosw.h> 48 #include <sys/rwlock.h> 49 #include <sys/signalvar.h> 50 #include <sys/socket.h> 51 #include <sys/socketvar.h> 52 #include <sys/sx.h> 53 #include <sys/sysctl.h> 54 #include <sys/systm.h> 55 56 #include <vm/uma.h> 57 58 #include <net/if.h> 59 #include <net/route.h> 60 #include <net/vnet.h> 61 62 #include <netinet/in.h> 63 #include <netinet/in_systm.h> 64 #include <netinet/in_pcb.h> 65 #include <netinet/in_var.h> 66 #include <netinet/ip.h> 67 #include <netinet/ip_var.h> 68 #include <netinet/ip_mroute.h> 69 70 #ifdef IPSEC 71 #include <netipsec/ipsec.h> 72 #endif /*IPSEC*/ 73 74 #include <security/mac/mac_framework.h> 75 76 VNET_DEFINE(struct inpcbhead, ripcb); 77 VNET_DEFINE(struct inpcbinfo, ripcbinfo); 78 79 #define V_ripcb VNET(ripcb) 80 #define V_ripcbinfo VNET(ripcbinfo) 81 82 /* 83 * Control and data hooks for ipfw, dummynet, divert and so on. 84 * The data hooks are not used here but it is convenient 85 * to keep them all in one place. 86 */ 87 VNET_DEFINE(ip_fw_chk_ptr_t, ip_fw_chk_ptr) = NULL; 88 VNET_DEFINE(ip_fw_ctl_ptr_t, ip_fw_ctl_ptr) = NULL; 89 90 int (*ip_dn_ctl_ptr)(struct sockopt *); 91 int (*ip_dn_io_ptr)(struct mbuf **, int, struct ip_fw_args *); 92 void (*ip_divert_ptr)(struct mbuf *, int); 93 int (*ng_ipfw_input_p)(struct mbuf **, int, 94 struct ip_fw_args *, int); 95 96 /* 97 * Hooks for multicast routing. They all default to NULL, so leave them not 98 * initialized and rely on BSS being set to 0. 99 */ 100 101 /* 102 * The socket used to communicate with the multicast routing daemon. 103 */ 104 VNET_DEFINE(struct socket *, ip_mrouter); 105 106 /* 107 * The various mrouter and rsvp functions. 108 */ 109 int (*ip_mrouter_set)(struct socket *, struct sockopt *); 110 int (*ip_mrouter_get)(struct socket *, struct sockopt *); 111 int (*ip_mrouter_done)(void); 112 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 113 struct ip_moptions *); 114 int (*mrt_ioctl)(u_long, caddr_t, int); 115 int (*legal_vif_num)(int); 116 u_long (*ip_mcast_src)(int); 117 118 void (*rsvp_input_p)(struct mbuf *m, int off); 119 int (*ip_rsvp_vif)(struct socket *, struct sockopt *); 120 void (*ip_rsvp_force_done)(struct socket *); 121 122 /* 123 * Hash functions 124 */ 125 126 #define INP_PCBHASH_RAW_SIZE 256 127 #define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \ 128 (((proto) + (laddr) + (faddr)) % (mask) + 1) 129 130 static void 131 rip_inshash(struct inpcb *inp) 132 { 133 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 134 struct inpcbhead *pcbhash; 135 int hash; 136 137 INP_INFO_WLOCK_ASSERT(pcbinfo); 138 INP_WLOCK_ASSERT(inp); 139 140 if (inp->inp_ip_p != 0 && 141 inp->inp_laddr.s_addr != INADDR_ANY && 142 inp->inp_faddr.s_addr != INADDR_ANY) { 143 hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr, 144 inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask); 145 } else 146 hash = 0; 147 pcbhash = &pcbinfo->ipi_hashbase[hash]; 148 LIST_INSERT_HEAD(pcbhash, inp, inp_hash); 149 } 150 151 static void 152 rip_delhash(struct inpcb *inp) 153 { 154 155 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 156 INP_WLOCK_ASSERT(inp); 157 158 LIST_REMOVE(inp, inp_hash); 159 } 160 161 /* 162 * Raw interface to IP protocol. 163 */ 164 165 /* 166 * Initialize raw connection block q. 167 */ 168 static void 169 rip_zone_change(void *tag) 170 { 171 172 uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets); 173 } 174 175 static int 176 rip_inpcb_init(void *mem, int size, int flags) 177 { 178 struct inpcb *inp = mem; 179 180 INP_LOCK_INIT(inp, "inp", "rawinp"); 181 return (0); 182 } 183 184 void 185 rip_init(void) 186 { 187 188 in_pcbinfo_init(&V_ripcbinfo, "rip", &V_ripcb, INP_PCBHASH_RAW_SIZE, 189 1, "ripcb", rip_inpcb_init, NULL, UMA_ZONE_NOFREE); 190 EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL, 191 EVENTHANDLER_PRI_ANY); 192 } 193 194 #ifdef VIMAGE 195 void 196 rip_destroy(void) 197 { 198 199 in_pcbinfo_destroy(&V_ripcbinfo); 200 } 201 #endif 202 203 static int 204 rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n, 205 struct sockaddr_in *ripsrc) 206 { 207 int policyfail = 0; 208 209 INP_RLOCK_ASSERT(last); 210 211 #ifdef IPSEC 212 /* check AH/ESP integrity. */ 213 if (ipsec4_in_reject(n, last)) { 214 policyfail = 1; 215 } 216 #endif /* IPSEC */ 217 #ifdef MAC 218 if (!policyfail && mac_inpcb_check_deliver(last, n) != 0) 219 policyfail = 1; 220 #endif 221 /* Check the minimum TTL for socket. */ 222 if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl) 223 policyfail = 1; 224 if (!policyfail) { 225 struct mbuf *opts = NULL; 226 struct socket *so; 227 228 so = last->inp_socket; 229 if ((last->inp_flags & INP_CONTROLOPTS) || 230 (so->so_options & (SO_TIMESTAMP | SO_BINTIME))) 231 ip_savecontrol(last, &opts, ip, n); 232 SOCKBUF_LOCK(&so->so_rcv); 233 if (sbappendaddr_locked(&so->so_rcv, 234 (struct sockaddr *)ripsrc, n, opts) == 0) { 235 /* should notify about lost packet */ 236 m_freem(n); 237 if (opts) 238 m_freem(opts); 239 SOCKBUF_UNLOCK(&so->so_rcv); 240 } else 241 sorwakeup_locked(so); 242 } else 243 m_freem(n); 244 return (policyfail); 245 } 246 247 /* 248 * Setup generic address and protocol structures for raw_input routine, then 249 * pass them along with mbuf chain. 250 */ 251 void 252 rip_input(struct mbuf *m, int off) 253 { 254 struct ifnet *ifp; 255 struct ip *ip = mtod(m, struct ip *); 256 int proto = ip->ip_p; 257 struct inpcb *inp, *last; 258 struct sockaddr_in ripsrc; 259 int hash; 260 261 bzero(&ripsrc, sizeof(ripsrc)); 262 ripsrc.sin_len = sizeof(ripsrc); 263 ripsrc.sin_family = AF_INET; 264 ripsrc.sin_addr = ip->ip_src; 265 last = NULL; 266 267 ifp = m->m_pkthdr.rcvif; 268 269 hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr, 270 ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask); 271 INP_INFO_RLOCK(&V_ripcbinfo); 272 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) { 273 if (inp->inp_ip_p != proto) 274 continue; 275 #ifdef INET6 276 /* XXX inp locking */ 277 if ((inp->inp_vflag & INP_IPV4) == 0) 278 continue; 279 #endif 280 if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr) 281 continue; 282 if (inp->inp_faddr.s_addr != ip->ip_src.s_addr) 283 continue; 284 if (jailed_without_vnet(inp->inp_cred)) { 285 /* 286 * XXX: If faddr was bound to multicast group, 287 * jailed raw socket will drop datagram. 288 */ 289 if (prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) 290 continue; 291 } 292 if (last != NULL) { 293 struct mbuf *n; 294 295 n = m_copy(m, 0, (int)M_COPYALL); 296 if (n != NULL) 297 (void) rip_append(last, ip, n, &ripsrc); 298 /* XXX count dropped packet */ 299 INP_RUNLOCK(last); 300 } 301 INP_RLOCK(inp); 302 last = inp; 303 } 304 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) { 305 if (inp->inp_ip_p && inp->inp_ip_p != proto) 306 continue; 307 #ifdef INET6 308 /* XXX inp locking */ 309 if ((inp->inp_vflag & INP_IPV4) == 0) 310 continue; 311 #endif 312 if (!in_nullhost(inp->inp_laddr) && 313 !in_hosteq(inp->inp_laddr, ip->ip_dst)) 314 continue; 315 if (!in_nullhost(inp->inp_faddr) && 316 !in_hosteq(inp->inp_faddr, ip->ip_src)) 317 continue; 318 if (jailed_without_vnet(inp->inp_cred)) { 319 /* 320 * Allow raw socket in jail to receive multicast; 321 * assume process had PRIV_NETINET_RAW at attach, 322 * and fall through into normal filter path if so. 323 */ 324 if (!IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 325 prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) 326 continue; 327 } 328 /* 329 * If this raw socket has multicast state, and we 330 * have received a multicast, check if this socket 331 * should receive it, as multicast filtering is now 332 * the responsibility of the transport layer. 333 */ 334 if (inp->inp_moptions != NULL && 335 IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 336 /* 337 * If the incoming datagram is for IGMP, allow it 338 * through unconditionally to the raw socket. 339 * 340 * In the case of IGMPv2, we may not have explicitly 341 * joined the group, and may have set IFF_ALLMULTI 342 * on the interface. imo_multi_filter() may discard 343 * control traffic we actually need to see. 344 * 345 * Userland multicast routing daemons should continue 346 * filter the control traffic appropriately. 347 */ 348 int blocked; 349 350 blocked = MCAST_PASS; 351 if (proto != IPPROTO_IGMP) { 352 struct sockaddr_in group; 353 354 bzero(&group, sizeof(struct sockaddr_in)); 355 group.sin_len = sizeof(struct sockaddr_in); 356 group.sin_family = AF_INET; 357 group.sin_addr = ip->ip_dst; 358 359 blocked = imo_multi_filter(inp->inp_moptions, 360 ifp, 361 (struct sockaddr *)&group, 362 (struct sockaddr *)&ripsrc); 363 } 364 365 if (blocked != MCAST_PASS) { 366 IPSTAT_INC(ips_notmember); 367 continue; 368 } 369 } 370 if (last != NULL) { 371 struct mbuf *n; 372 373 n = m_copy(m, 0, (int)M_COPYALL); 374 if (n != NULL) 375 (void) rip_append(last, ip, n, &ripsrc); 376 /* XXX count dropped packet */ 377 INP_RUNLOCK(last); 378 } 379 INP_RLOCK(inp); 380 last = inp; 381 } 382 INP_INFO_RUNLOCK(&V_ripcbinfo); 383 if (last != NULL) { 384 if (rip_append(last, ip, m, &ripsrc) != 0) 385 IPSTAT_INC(ips_delivered); 386 INP_RUNLOCK(last); 387 } else { 388 m_freem(m); 389 IPSTAT_INC(ips_noproto); 390 IPSTAT_DEC(ips_delivered); 391 } 392 } 393 394 /* 395 * Generate IP header and pass packet to ip_output. Tack on options user may 396 * have setup with control call. 397 */ 398 int 399 rip_output(struct mbuf *m, struct socket *so, u_long dst) 400 { 401 struct ip *ip; 402 int error; 403 struct inpcb *inp = sotoinpcb(so); 404 int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) | 405 IP_ALLOWBROADCAST; 406 407 /* 408 * If the user handed us a complete IP packet, use it. Otherwise, 409 * allocate an mbuf for a header and fill it in. 410 */ 411 if ((inp->inp_flags & INP_HDRINCL) == 0) { 412 if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) { 413 m_freem(m); 414 return(EMSGSIZE); 415 } 416 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT); 417 if (m == NULL) 418 return(ENOBUFS); 419 420 INP_RLOCK(inp); 421 ip = mtod(m, struct ip *); 422 ip->ip_tos = inp->inp_ip_tos; 423 if (inp->inp_flags & INP_DONTFRAG) 424 ip->ip_off = IP_DF; 425 else 426 ip->ip_off = 0; 427 ip->ip_p = inp->inp_ip_p; 428 ip->ip_len = m->m_pkthdr.len; 429 ip->ip_src = inp->inp_laddr; 430 if (jailed(inp->inp_cred)) { 431 /* 432 * prison_local_ip4() would be good enough but would 433 * let a source of INADDR_ANY pass, which we do not 434 * want to see from jails. We do not go through the 435 * pain of in_pcbladdr() for raw sockets. 436 */ 437 if (ip->ip_src.s_addr == INADDR_ANY) 438 error = prison_get_ip4(inp->inp_cred, 439 &ip->ip_src); 440 else 441 error = prison_local_ip4(inp->inp_cred, 442 &ip->ip_src); 443 if (error != 0) { 444 INP_RUNLOCK(inp); 445 m_freem(m); 446 return (error); 447 } 448 } 449 ip->ip_dst.s_addr = dst; 450 ip->ip_ttl = inp->inp_ip_ttl; 451 } else { 452 if (m->m_pkthdr.len > IP_MAXPACKET) { 453 m_freem(m); 454 return(EMSGSIZE); 455 } 456 INP_RLOCK(inp); 457 ip = mtod(m, struct ip *); 458 error = prison_check_ip4(inp->inp_cred, &ip->ip_src); 459 if (error != 0) { 460 INP_RUNLOCK(inp); 461 m_freem(m); 462 return (error); 463 } 464 465 /* 466 * Don't allow both user specified and setsockopt options, 467 * and don't allow packet length sizes that will crash. 468 */ 469 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) 470 || (ip->ip_len > m->m_pkthdr.len) 471 || (ip->ip_len < (ip->ip_hl << 2))) { 472 INP_RUNLOCK(inp); 473 m_freem(m); 474 return (EINVAL); 475 } 476 if (ip->ip_id == 0) 477 ip->ip_id = ip_newid(); 478 479 /* 480 * XXX prevent ip_output from overwriting header fields. 481 */ 482 flags |= IP_RAWOUTPUT; 483 IPSTAT_INC(ips_rawout); 484 } 485 486 if (inp->inp_flags & INP_ONESBCAST) 487 flags |= IP_SENDONES; 488 489 #ifdef MAC 490 mac_inpcb_create_mbuf(inp, m); 491 #endif 492 493 error = ip_output(m, inp->inp_options, NULL, flags, 494 inp->inp_moptions, inp); 495 INP_RUNLOCK(inp); 496 return (error); 497 } 498 499 /* 500 * Raw IP socket option processing. 501 * 502 * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could 503 * only be created by a privileged process, and as such, socket option 504 * operations to manage system properties on any raw socket were allowed to 505 * take place without explicit additional access control checks. However, 506 * raw sockets can now also be created in jail(), and therefore explicit 507 * checks are now required. Likewise, raw sockets can be used by a process 508 * after it gives up privilege, so some caution is required. For options 509 * passed down to the IP layer via ip_ctloutput(), checks are assumed to be 510 * performed in ip_ctloutput() and therefore no check occurs here. 511 * Unilaterally checking priv_check() here breaks normal IP socket option 512 * operations on raw sockets. 513 * 514 * When adding new socket options here, make sure to add access control 515 * checks here as necessary. 516 */ 517 int 518 rip_ctloutput(struct socket *so, struct sockopt *sopt) 519 { 520 struct inpcb *inp = sotoinpcb(so); 521 int error, optval; 522 523 if (sopt->sopt_level != IPPROTO_IP) { 524 if ((sopt->sopt_level == SOL_SOCKET) && 525 (sopt->sopt_name == SO_SETFIB)) { 526 inp->inp_inc.inc_fibnum = so->so_fibnum; 527 return (0); 528 } 529 return (EINVAL); 530 } 531 532 error = 0; 533 switch (sopt->sopt_dir) { 534 case SOPT_GET: 535 switch (sopt->sopt_name) { 536 case IP_HDRINCL: 537 optval = inp->inp_flags & INP_HDRINCL; 538 error = sooptcopyout(sopt, &optval, sizeof optval); 539 break; 540 541 case IP_FW3: /* generic ipfw v.3 functions */ 542 case IP_FW_ADD: /* ADD actually returns the body... */ 543 case IP_FW_GET: 544 case IP_FW_TABLE_GETSIZE: 545 case IP_FW_TABLE_LIST: 546 case IP_FW_NAT_GET_CONFIG: 547 case IP_FW_NAT_GET_LOG: 548 if (V_ip_fw_ctl_ptr != NULL) 549 error = V_ip_fw_ctl_ptr(sopt); 550 else 551 error = ENOPROTOOPT; 552 break; 553 554 case IP_DUMMYNET3: /* generic dummynet v.3 functions */ 555 case IP_DUMMYNET_GET: 556 if (ip_dn_ctl_ptr != NULL) 557 error = ip_dn_ctl_ptr(sopt); 558 else 559 error = ENOPROTOOPT; 560 break ; 561 562 case MRT_INIT: 563 case MRT_DONE: 564 case MRT_ADD_VIF: 565 case MRT_DEL_VIF: 566 case MRT_ADD_MFC: 567 case MRT_DEL_MFC: 568 case MRT_VERSION: 569 case MRT_ASSERT: 570 case MRT_API_SUPPORT: 571 case MRT_API_CONFIG: 572 case MRT_ADD_BW_UPCALL: 573 case MRT_DEL_BW_UPCALL: 574 error = priv_check(curthread, PRIV_NETINET_MROUTE); 575 if (error != 0) 576 return (error); 577 error = ip_mrouter_get ? ip_mrouter_get(so, sopt) : 578 EOPNOTSUPP; 579 break; 580 581 default: 582 error = ip_ctloutput(so, sopt); 583 break; 584 } 585 break; 586 587 case SOPT_SET: 588 switch (sopt->sopt_name) { 589 case IP_HDRINCL: 590 error = sooptcopyin(sopt, &optval, sizeof optval, 591 sizeof optval); 592 if (error) 593 break; 594 if (optval) 595 inp->inp_flags |= INP_HDRINCL; 596 else 597 inp->inp_flags &= ~INP_HDRINCL; 598 break; 599 600 case IP_FW3: /* generic ipfw v.3 functions */ 601 case IP_FW_ADD: 602 case IP_FW_DEL: 603 case IP_FW_FLUSH: 604 case IP_FW_ZERO: 605 case IP_FW_RESETLOG: 606 case IP_FW_TABLE_ADD: 607 case IP_FW_TABLE_DEL: 608 case IP_FW_TABLE_FLUSH: 609 case IP_FW_NAT_CFG: 610 case IP_FW_NAT_DEL: 611 if (V_ip_fw_ctl_ptr != NULL) 612 error = V_ip_fw_ctl_ptr(sopt); 613 else 614 error = ENOPROTOOPT; 615 break; 616 617 case IP_DUMMYNET3: /* generic dummynet v.3 functions */ 618 case IP_DUMMYNET_CONFIGURE: 619 case IP_DUMMYNET_DEL: 620 case IP_DUMMYNET_FLUSH: 621 if (ip_dn_ctl_ptr != NULL) 622 error = ip_dn_ctl_ptr(sopt); 623 else 624 error = ENOPROTOOPT ; 625 break ; 626 627 case IP_RSVP_ON: 628 error = priv_check(curthread, PRIV_NETINET_MROUTE); 629 if (error != 0) 630 return (error); 631 error = ip_rsvp_init(so); 632 break; 633 634 case IP_RSVP_OFF: 635 error = priv_check(curthread, PRIV_NETINET_MROUTE); 636 if (error != 0) 637 return (error); 638 error = ip_rsvp_done(); 639 break; 640 641 case IP_RSVP_VIF_ON: 642 case IP_RSVP_VIF_OFF: 643 error = priv_check(curthread, PRIV_NETINET_MROUTE); 644 if (error != 0) 645 return (error); 646 error = ip_rsvp_vif ? 647 ip_rsvp_vif(so, sopt) : EINVAL; 648 break; 649 650 case MRT_INIT: 651 case MRT_DONE: 652 case MRT_ADD_VIF: 653 case MRT_DEL_VIF: 654 case MRT_ADD_MFC: 655 case MRT_DEL_MFC: 656 case MRT_VERSION: 657 case MRT_ASSERT: 658 case MRT_API_SUPPORT: 659 case MRT_API_CONFIG: 660 case MRT_ADD_BW_UPCALL: 661 case MRT_DEL_BW_UPCALL: 662 error = priv_check(curthread, PRIV_NETINET_MROUTE); 663 if (error != 0) 664 return (error); 665 error = ip_mrouter_set ? ip_mrouter_set(so, sopt) : 666 EOPNOTSUPP; 667 break; 668 669 default: 670 error = ip_ctloutput(so, sopt); 671 break; 672 } 673 break; 674 } 675 676 return (error); 677 } 678 679 /* 680 * This function exists solely to receive the PRC_IFDOWN messages which are 681 * sent by if_down(). It looks for an ifaddr whose ifa_addr is sa, and calls 682 * in_ifadown() to remove all routes corresponding to that address. It also 683 * receives the PRC_IFUP messages from if_up() and reinstalls the interface 684 * routes. 685 */ 686 void 687 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip) 688 { 689 struct in_ifaddr *ia; 690 struct ifnet *ifp; 691 int err; 692 int flags; 693 694 switch (cmd) { 695 case PRC_IFDOWN: 696 IN_IFADDR_RLOCK(); 697 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 698 if (ia->ia_ifa.ifa_addr == sa 699 && (ia->ia_flags & IFA_ROUTE)) { 700 ifa_ref(&ia->ia_ifa); 701 IN_IFADDR_RUNLOCK(); 702 /* 703 * in_ifscrub kills the interface route. 704 */ 705 in_ifscrub(ia->ia_ifp, ia); 706 /* 707 * in_ifadown gets rid of all the rest of the 708 * routes. This is not quite the right thing 709 * to do, but at least if we are running a 710 * routing process they will come back. 711 */ 712 in_ifadown(&ia->ia_ifa, 0); 713 ifa_free(&ia->ia_ifa); 714 break; 715 } 716 } 717 if (ia == NULL) /* If ia matched, already unlocked. */ 718 IN_IFADDR_RUNLOCK(); 719 break; 720 721 case PRC_IFUP: 722 IN_IFADDR_RLOCK(); 723 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 724 if (ia->ia_ifa.ifa_addr == sa) 725 break; 726 } 727 if (ia == NULL || (ia->ia_flags & IFA_ROUTE)) { 728 IN_IFADDR_RUNLOCK(); 729 return; 730 } 731 ifa_ref(&ia->ia_ifa); 732 IN_IFADDR_RUNLOCK(); 733 flags = RTF_UP; 734 ifp = ia->ia_ifa.ifa_ifp; 735 736 if ((ifp->if_flags & IFF_LOOPBACK) 737 || (ifp->if_flags & IFF_POINTOPOINT)) 738 flags |= RTF_HOST; 739 740 err = rtinit(&ia->ia_ifa, RTM_ADD, flags); 741 if (err == 0) 742 ia->ia_flags |= IFA_ROUTE; 743 err = ifa_add_loopback_route((struct ifaddr *)ia, sa); 744 if (err == 0) 745 ia->ia_flags |= IFA_RTSELF; 746 ifa_free(&ia->ia_ifa); 747 break; 748 } 749 } 750 751 u_long rip_sendspace = 9216; 752 u_long rip_recvspace = 9216; 753 754 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW, 755 &rip_sendspace, 0, "Maximum outgoing raw IP datagram size"); 756 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW, 757 &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams"); 758 759 static int 760 rip_attach(struct socket *so, int proto, struct thread *td) 761 { 762 struct inpcb *inp; 763 int error; 764 765 inp = sotoinpcb(so); 766 KASSERT(inp == NULL, ("rip_attach: inp != NULL")); 767 768 error = priv_check(td, PRIV_NETINET_RAW); 769 if (error) 770 return (error); 771 if (proto >= IPPROTO_MAX || proto < 0) 772 return EPROTONOSUPPORT; 773 error = soreserve(so, rip_sendspace, rip_recvspace); 774 if (error) 775 return (error); 776 INP_INFO_WLOCK(&V_ripcbinfo); 777 error = in_pcballoc(so, &V_ripcbinfo); 778 if (error) { 779 INP_INFO_WUNLOCK(&V_ripcbinfo); 780 return (error); 781 } 782 inp = (struct inpcb *)so->so_pcb; 783 inp->inp_vflag |= INP_IPV4; 784 inp->inp_ip_p = proto; 785 inp->inp_ip_ttl = V_ip_defttl; 786 rip_inshash(inp); 787 INP_INFO_WUNLOCK(&V_ripcbinfo); 788 INP_WUNLOCK(inp); 789 return (0); 790 } 791 792 static void 793 rip_detach(struct socket *so) 794 { 795 struct inpcb *inp; 796 797 inp = sotoinpcb(so); 798 KASSERT(inp != NULL, ("rip_detach: inp == NULL")); 799 KASSERT(inp->inp_faddr.s_addr == INADDR_ANY, 800 ("rip_detach: not closed")); 801 802 INP_INFO_WLOCK(&V_ripcbinfo); 803 INP_WLOCK(inp); 804 rip_delhash(inp); 805 if (so == V_ip_mrouter && ip_mrouter_done) 806 ip_mrouter_done(); 807 if (ip_rsvp_force_done) 808 ip_rsvp_force_done(so); 809 if (so == V_ip_rsvpd) 810 ip_rsvp_done(); 811 in_pcbdetach(inp); 812 in_pcbfree(inp); 813 INP_INFO_WUNLOCK(&V_ripcbinfo); 814 } 815 816 static void 817 rip_dodisconnect(struct socket *so, struct inpcb *inp) 818 { 819 820 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 821 INP_WLOCK_ASSERT(inp); 822 823 rip_delhash(inp); 824 inp->inp_faddr.s_addr = INADDR_ANY; 825 rip_inshash(inp); 826 SOCK_LOCK(so); 827 so->so_state &= ~SS_ISCONNECTED; 828 SOCK_UNLOCK(so); 829 } 830 831 static void 832 rip_abort(struct socket *so) 833 { 834 struct inpcb *inp; 835 836 inp = sotoinpcb(so); 837 KASSERT(inp != NULL, ("rip_abort: inp == NULL")); 838 839 INP_INFO_WLOCK(&V_ripcbinfo); 840 INP_WLOCK(inp); 841 rip_dodisconnect(so, inp); 842 INP_WUNLOCK(inp); 843 INP_INFO_WUNLOCK(&V_ripcbinfo); 844 } 845 846 static void 847 rip_close(struct socket *so) 848 { 849 struct inpcb *inp; 850 851 inp = sotoinpcb(so); 852 KASSERT(inp != NULL, ("rip_close: inp == NULL")); 853 854 INP_INFO_WLOCK(&V_ripcbinfo); 855 INP_WLOCK(inp); 856 rip_dodisconnect(so, inp); 857 INP_WUNLOCK(inp); 858 INP_INFO_WUNLOCK(&V_ripcbinfo); 859 } 860 861 static int 862 rip_disconnect(struct socket *so) 863 { 864 struct inpcb *inp; 865 866 if ((so->so_state & SS_ISCONNECTED) == 0) 867 return (ENOTCONN); 868 869 inp = sotoinpcb(so); 870 KASSERT(inp != NULL, ("rip_disconnect: inp == NULL")); 871 872 INP_INFO_WLOCK(&V_ripcbinfo); 873 INP_WLOCK(inp); 874 rip_dodisconnect(so, inp); 875 INP_WUNLOCK(inp); 876 INP_INFO_WUNLOCK(&V_ripcbinfo); 877 return (0); 878 } 879 880 static int 881 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 882 { 883 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 884 struct inpcb *inp; 885 int error; 886 887 if (nam->sa_len != sizeof(*addr)) 888 return (EINVAL); 889 890 error = prison_check_ip4(td->td_ucred, &addr->sin_addr); 891 if (error != 0) 892 return (error); 893 894 inp = sotoinpcb(so); 895 KASSERT(inp != NULL, ("rip_bind: inp == NULL")); 896 897 if (TAILQ_EMPTY(&V_ifnet) || 898 (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) || 899 (addr->sin_addr.s_addr && 900 (inp->inp_flags & INP_BINDANY) == 0 && 901 ifa_ifwithaddr_check((struct sockaddr *)addr) == 0)) 902 return (EADDRNOTAVAIL); 903 904 INP_INFO_WLOCK(&V_ripcbinfo); 905 INP_WLOCK(inp); 906 rip_delhash(inp); 907 inp->inp_laddr = addr->sin_addr; 908 rip_inshash(inp); 909 INP_WUNLOCK(inp); 910 INP_INFO_WUNLOCK(&V_ripcbinfo); 911 return (0); 912 } 913 914 static int 915 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 916 { 917 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 918 struct inpcb *inp; 919 920 if (nam->sa_len != sizeof(*addr)) 921 return (EINVAL); 922 if (TAILQ_EMPTY(&V_ifnet)) 923 return (EADDRNOTAVAIL); 924 if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) 925 return (EAFNOSUPPORT); 926 927 inp = sotoinpcb(so); 928 KASSERT(inp != NULL, ("rip_connect: inp == NULL")); 929 930 INP_INFO_WLOCK(&V_ripcbinfo); 931 INP_WLOCK(inp); 932 rip_delhash(inp); 933 inp->inp_faddr = addr->sin_addr; 934 rip_inshash(inp); 935 soisconnected(so); 936 INP_WUNLOCK(inp); 937 INP_INFO_WUNLOCK(&V_ripcbinfo); 938 return (0); 939 } 940 941 static int 942 rip_shutdown(struct socket *so) 943 { 944 struct inpcb *inp; 945 946 inp = sotoinpcb(so); 947 KASSERT(inp != NULL, ("rip_shutdown: inp == NULL")); 948 949 INP_WLOCK(inp); 950 socantsendmore(so); 951 INP_WUNLOCK(inp); 952 return (0); 953 } 954 955 static int 956 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 957 struct mbuf *control, struct thread *td) 958 { 959 struct inpcb *inp; 960 u_long dst; 961 962 inp = sotoinpcb(so); 963 KASSERT(inp != NULL, ("rip_send: inp == NULL")); 964 965 /* 966 * Note: 'dst' reads below are unlocked. 967 */ 968 if (so->so_state & SS_ISCONNECTED) { 969 if (nam) { 970 m_freem(m); 971 return (EISCONN); 972 } 973 dst = inp->inp_faddr.s_addr; /* Unlocked read. */ 974 } else { 975 if (nam == NULL) { 976 m_freem(m); 977 return (ENOTCONN); 978 } 979 dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr; 980 } 981 return (rip_output(m, so, dst)); 982 } 983 984 static int 985 rip_pcblist(SYSCTL_HANDLER_ARGS) 986 { 987 int error, i, n; 988 struct inpcb *inp, **inp_list; 989 inp_gen_t gencnt; 990 struct xinpgen xig; 991 992 /* 993 * The process of preparing the TCB list is too time-consuming and 994 * resource-intensive to repeat twice on every request. 995 */ 996 if (req->oldptr == 0) { 997 n = V_ripcbinfo.ipi_count; 998 n += imax(n / 8, 10); 999 req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb); 1000 return (0); 1001 } 1002 1003 if (req->newptr != 0) 1004 return (EPERM); 1005 1006 /* 1007 * OK, now we're committed to doing something. 1008 */ 1009 INP_INFO_RLOCK(&V_ripcbinfo); 1010 gencnt = V_ripcbinfo.ipi_gencnt; 1011 n = V_ripcbinfo.ipi_count; 1012 INP_INFO_RUNLOCK(&V_ripcbinfo); 1013 1014 xig.xig_len = sizeof xig; 1015 xig.xig_count = n; 1016 xig.xig_gen = gencnt; 1017 xig.xig_sogen = so_gencnt; 1018 error = SYSCTL_OUT(req, &xig, sizeof xig); 1019 if (error) 1020 return (error); 1021 1022 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 1023 if (inp_list == 0) 1024 return (ENOMEM); 1025 1026 INP_INFO_RLOCK(&V_ripcbinfo); 1027 for (inp = LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n; 1028 inp = LIST_NEXT(inp, inp_list)) { 1029 INP_WLOCK(inp); 1030 if (inp->inp_gencnt <= gencnt && 1031 cr_canseeinpcb(req->td->td_ucred, inp) == 0) { 1032 in_pcbref(inp); 1033 inp_list[i++] = inp; 1034 } 1035 INP_WUNLOCK(inp); 1036 } 1037 INP_INFO_RUNLOCK(&V_ripcbinfo); 1038 n = i; 1039 1040 error = 0; 1041 for (i = 0; i < n; i++) { 1042 inp = inp_list[i]; 1043 INP_RLOCK(inp); 1044 if (inp->inp_gencnt <= gencnt) { 1045 struct xinpcb xi; 1046 1047 bzero(&xi, sizeof(xi)); 1048 xi.xi_len = sizeof xi; 1049 /* XXX should avoid extra copy */ 1050 bcopy(inp, &xi.xi_inp, sizeof *inp); 1051 if (inp->inp_socket) 1052 sotoxsocket(inp->inp_socket, &xi.xi_socket); 1053 INP_RUNLOCK(inp); 1054 error = SYSCTL_OUT(req, &xi, sizeof xi); 1055 } else 1056 INP_RUNLOCK(inp); 1057 } 1058 INP_INFO_WLOCK(&V_ripcbinfo); 1059 for (i = 0; i < n; i++) { 1060 inp = inp_list[i]; 1061 INP_WLOCK(inp); 1062 if (!in_pcbrele(inp)) 1063 INP_WUNLOCK(inp); 1064 } 1065 INP_INFO_WUNLOCK(&V_ripcbinfo); 1066 1067 if (!error) { 1068 /* 1069 * Give the user an updated idea of our state. If the 1070 * generation differs from what we told her before, she knows 1071 * that something happened while we were processing this 1072 * request, and it might be necessary to retry. 1073 */ 1074 INP_INFO_RLOCK(&V_ripcbinfo); 1075 xig.xig_gen = V_ripcbinfo.ipi_gencnt; 1076 xig.xig_sogen = so_gencnt; 1077 xig.xig_count = V_ripcbinfo.ipi_count; 1078 INP_INFO_RUNLOCK(&V_ripcbinfo); 1079 error = SYSCTL_OUT(req, &xig, sizeof xig); 1080 } 1081 free(inp_list, M_TEMP); 1082 return (error); 1083 } 1084 1085 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, 1086 CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0, 1087 rip_pcblist, "S,xinpcb", "List of active raw IP sockets"); 1088 1089 struct pr_usrreqs rip_usrreqs = { 1090 .pru_abort = rip_abort, 1091 .pru_attach = rip_attach, 1092 .pru_bind = rip_bind, 1093 .pru_connect = rip_connect, 1094 .pru_control = in_control, 1095 .pru_detach = rip_detach, 1096 .pru_disconnect = rip_disconnect, 1097 .pru_peeraddr = in_getpeeraddr, 1098 .pru_send = rip_send, 1099 .pru_shutdown = rip_shutdown, 1100 .pru_sockaddr = in_getsockaddr, 1101 .pru_sosetlabel = in_pcbsosetlabel, 1102 .pru_close = rip_close, 1103 }; 1104