raw_ip.c revision 222748
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: head/sys/netinet/raw_ip.c 222748 2011-06-06 12:55:02Z rwatson $"); 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/route.h> 61#include <net/vnet.h> 62 63#include <netinet/in.h> 64#include <netinet/in_systm.h> 65#include <netinet/in_pcb.h> 66#include <netinet/in_var.h> 67#include <netinet/if_ether.h> 68#include <netinet/ip.h> 69#include <netinet/ip_var.h> 70#include <netinet/ip_mroute.h> 71 72#ifdef IPSEC 73#include <netipsec/ipsec.h> 74#endif /*IPSEC*/ 75 76#include <security/mac/mac_framework.h> 77 78VNET_DEFINE(int, ip_defttl) = IPDEFTTL; 79SYSCTL_VNET_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_RW, 80 &VNET_NAME(ip_defttl), 0, 81 "Maximum TTL on IP packets"); 82 83VNET_DEFINE(struct inpcbhead, ripcb); 84VNET_DEFINE(struct inpcbinfo, ripcbinfo); 85 86#define V_ripcb VNET(ripcb) 87#define V_ripcbinfo VNET(ripcbinfo) 88 89/* 90 * Control and data hooks for ipfw, dummynet, divert and so on. 91 * The data hooks are not used here but it is convenient 92 * to keep them all in one place. 93 */ 94VNET_DEFINE(ip_fw_chk_ptr_t, ip_fw_chk_ptr) = NULL; 95VNET_DEFINE(ip_fw_ctl_ptr_t, ip_fw_ctl_ptr) = NULL; 96 97int (*ip_dn_ctl_ptr)(struct sockopt *); 98int (*ip_dn_io_ptr)(struct mbuf **, int, struct ip_fw_args *); 99void (*ip_divert_ptr)(struct mbuf *, int); 100int (*ng_ipfw_input_p)(struct mbuf **, int, 101 struct ip_fw_args *, int); 102 103#ifdef INET 104/* 105 * Hooks for multicast routing. They all default to NULL, so leave them not 106 * initialized and rely on BSS being set to 0. 107 */ 108 109/* 110 * The socket used to communicate with the multicast routing daemon. 111 */ 112VNET_DEFINE(struct socket *, ip_mrouter); 113 114/* 115 * The various mrouter and rsvp functions. 116 */ 117int (*ip_mrouter_set)(struct socket *, struct sockopt *); 118int (*ip_mrouter_get)(struct socket *, struct sockopt *); 119int (*ip_mrouter_done)(void); 120int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 121 struct ip_moptions *); 122int (*mrt_ioctl)(u_long, caddr_t, int); 123int (*legal_vif_num)(int); 124u_long (*ip_mcast_src)(int); 125 126void (*rsvp_input_p)(struct mbuf *m, int off); 127int (*ip_rsvp_vif)(struct socket *, struct sockopt *); 128void (*ip_rsvp_force_done)(struct socket *); 129#endif /* INET */ 130 131u_long rip_sendspace = 9216; 132SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW, 133 &rip_sendspace, 0, "Maximum outgoing raw IP datagram size"); 134 135u_long rip_recvspace = 9216; 136SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW, 137 &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams"); 138 139/* 140 * Hash functions 141 */ 142 143#define INP_PCBHASH_RAW_SIZE 256 144#define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \ 145 (((proto) + (laddr) + (faddr)) % (mask) + 1) 146 147#ifdef INET 148static void 149rip_inshash(struct inpcb *inp) 150{ 151 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 152 struct inpcbhead *pcbhash; 153 int hash; 154 155 INP_INFO_WLOCK_ASSERT(pcbinfo); 156 INP_WLOCK_ASSERT(inp); 157 158 if (inp->inp_ip_p != 0 && 159 inp->inp_laddr.s_addr != INADDR_ANY && 160 inp->inp_faddr.s_addr != INADDR_ANY) { 161 hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr, 162 inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask); 163 } else 164 hash = 0; 165 pcbhash = &pcbinfo->ipi_hashbase[hash]; 166 LIST_INSERT_HEAD(pcbhash, inp, inp_hash); 167} 168 169static void 170rip_delhash(struct inpcb *inp) 171{ 172 173 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 174 INP_WLOCK_ASSERT(inp); 175 176 LIST_REMOVE(inp, inp_hash); 177} 178#endif /* INET */ 179 180/* 181 * Raw interface to IP protocol. 182 */ 183 184/* 185 * Initialize raw connection block q. 186 */ 187static void 188rip_zone_change(void *tag) 189{ 190 191 uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets); 192} 193 194static int 195rip_inpcb_init(void *mem, int size, int flags) 196{ 197 struct inpcb *inp = mem; 198 199 INP_LOCK_INIT(inp, "inp", "rawinp"); 200 return (0); 201} 202 203void 204rip_init(void) 205{ 206 207 in_pcbinfo_init(&V_ripcbinfo, "rip", &V_ripcb, INP_PCBHASH_RAW_SIZE, 208 1, "ripcb", rip_inpcb_init, NULL, UMA_ZONE_NOFREE, 209 IPI_HASHFIELDS_NONE); 210 EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL, 211 EVENTHANDLER_PRI_ANY); 212} 213 214#ifdef VIMAGE 215void 216rip_destroy(void) 217{ 218 219 in_pcbinfo_destroy(&V_ripcbinfo); 220} 221#endif 222 223#ifdef INET 224static int 225rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n, 226 struct sockaddr_in *ripsrc) 227{ 228 int policyfail = 0; 229 230 INP_LOCK_ASSERT(last); 231 232#ifdef IPSEC 233 /* check AH/ESP integrity. */ 234 if (ipsec4_in_reject(n, last)) { 235 policyfail = 1; 236 } 237#endif /* IPSEC */ 238#ifdef MAC 239 if (!policyfail && mac_inpcb_check_deliver(last, n) != 0) 240 policyfail = 1; 241#endif 242 /* Check the minimum TTL for socket. */ 243 if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl) 244 policyfail = 1; 245 if (!policyfail) { 246 struct mbuf *opts = NULL; 247 struct socket *so; 248 249 so = last->inp_socket; 250 if ((last->inp_flags & INP_CONTROLOPTS) || 251 (so->so_options & (SO_TIMESTAMP | SO_BINTIME))) 252 ip_savecontrol(last, &opts, ip, n); 253 SOCKBUF_LOCK(&so->so_rcv); 254 if (sbappendaddr_locked(&so->so_rcv, 255 (struct sockaddr *)ripsrc, n, opts) == 0) { 256 /* should notify about lost packet */ 257 m_freem(n); 258 if (opts) 259 m_freem(opts); 260 SOCKBUF_UNLOCK(&so->so_rcv); 261 } else 262 sorwakeup_locked(so); 263 } else 264 m_freem(n); 265 return (policyfail); 266} 267 268/* 269 * Setup generic address and protocol structures for raw_input routine, then 270 * pass them along with mbuf chain. 271 */ 272void 273rip_input(struct mbuf *m, int off) 274{ 275 struct ifnet *ifp; 276 struct ip *ip = mtod(m, struct ip *); 277 int proto = ip->ip_p; 278 struct inpcb *inp, *last; 279 struct sockaddr_in ripsrc; 280 int hash; 281 282 bzero(&ripsrc, sizeof(ripsrc)); 283 ripsrc.sin_len = sizeof(ripsrc); 284 ripsrc.sin_family = AF_INET; 285 ripsrc.sin_addr = ip->ip_src; 286 last = NULL; 287 288 ifp = m->m_pkthdr.rcvif; 289 290 hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr, 291 ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask); 292 INP_INFO_RLOCK(&V_ripcbinfo); 293 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) { 294 if (inp->inp_ip_p != proto) 295 continue; 296#ifdef INET6 297 /* XXX inp locking */ 298 if ((inp->inp_vflag & INP_IPV4) == 0) 299 continue; 300#endif 301 if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr) 302 continue; 303 if (inp->inp_faddr.s_addr != ip->ip_src.s_addr) 304 continue; 305 if (jailed_without_vnet(inp->inp_cred)) { 306 /* 307 * XXX: If faddr was bound to multicast group, 308 * jailed raw socket will drop datagram. 309 */ 310 if (prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) 311 continue; 312 } 313 if (last != NULL) { 314 struct mbuf *n; 315 316 n = m_copy(m, 0, (int)M_COPYALL); 317 if (n != NULL) 318 (void) rip_append(last, ip, n, &ripsrc); 319 /* XXX count dropped packet */ 320 INP_RUNLOCK(last); 321 } 322 INP_RLOCK(inp); 323 last = inp; 324 } 325 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) { 326 if (inp->inp_ip_p && inp->inp_ip_p != proto) 327 continue; 328#ifdef INET6 329 /* XXX inp locking */ 330 if ((inp->inp_vflag & INP_IPV4) == 0) 331 continue; 332#endif 333 if (!in_nullhost(inp->inp_laddr) && 334 !in_hosteq(inp->inp_laddr, ip->ip_dst)) 335 continue; 336 if (!in_nullhost(inp->inp_faddr) && 337 !in_hosteq(inp->inp_faddr, ip->ip_src)) 338 continue; 339 if (jailed_without_vnet(inp->inp_cred)) { 340 /* 341 * Allow raw socket in jail to receive multicast; 342 * assume process had PRIV_NETINET_RAW at attach, 343 * and fall through into normal filter path if so. 344 */ 345 if (!IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 346 prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) 347 continue; 348 } 349 /* 350 * If this raw socket has multicast state, and we 351 * have received a multicast, check if this socket 352 * should receive it, as multicast filtering is now 353 * the responsibility of the transport layer. 354 */ 355 if (inp->inp_moptions != NULL && 356 IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 357 /* 358 * If the incoming datagram is for IGMP, allow it 359 * through unconditionally to the raw socket. 360 * 361 * In the case of IGMPv2, we may not have explicitly 362 * joined the group, and may have set IFF_ALLMULTI 363 * on the interface. imo_multi_filter() may discard 364 * control traffic we actually need to see. 365 * 366 * Userland multicast routing daemons should continue 367 * filter the control traffic appropriately. 368 */ 369 int blocked; 370 371 blocked = MCAST_PASS; 372 if (proto != IPPROTO_IGMP) { 373 struct sockaddr_in group; 374 375 bzero(&group, sizeof(struct sockaddr_in)); 376 group.sin_len = sizeof(struct sockaddr_in); 377 group.sin_family = AF_INET; 378 group.sin_addr = ip->ip_dst; 379 380 blocked = imo_multi_filter(inp->inp_moptions, 381 ifp, 382 (struct sockaddr *)&group, 383 (struct sockaddr *)&ripsrc); 384 } 385 386 if (blocked != MCAST_PASS) { 387 IPSTAT_INC(ips_notmember); 388 continue; 389 } 390 } 391 if (last != NULL) { 392 struct mbuf *n; 393 394 n = m_copy(m, 0, (int)M_COPYALL); 395 if (n != NULL) 396 (void) rip_append(last, ip, n, &ripsrc); 397 /* XXX count dropped packet */ 398 INP_RUNLOCK(last); 399 } 400 INP_RLOCK(inp); 401 last = inp; 402 } 403 INP_INFO_RUNLOCK(&V_ripcbinfo); 404 if (last != NULL) { 405 if (rip_append(last, ip, m, &ripsrc) != 0) 406 IPSTAT_INC(ips_delivered); 407 INP_RUNLOCK(last); 408 } else { 409 m_freem(m); 410 IPSTAT_INC(ips_noproto); 411 IPSTAT_DEC(ips_delivered); 412 } 413} 414 415/* 416 * Generate IP header and pass packet to ip_output. Tack on options user may 417 * have setup with control call. 418 */ 419int 420rip_output(struct mbuf *m, struct socket *so, u_long dst) 421{ 422 struct ip *ip; 423 int error; 424 struct inpcb *inp = sotoinpcb(so); 425 int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) | 426 IP_ALLOWBROADCAST; 427 428 /* 429 * If the user handed us a complete IP packet, use it. Otherwise, 430 * allocate an mbuf for a header and fill it in. 431 */ 432 if ((inp->inp_flags & INP_HDRINCL) == 0) { 433 if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) { 434 m_freem(m); 435 return(EMSGSIZE); 436 } 437 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT); 438 if (m == NULL) 439 return(ENOBUFS); 440 441 INP_RLOCK(inp); 442 ip = mtod(m, struct ip *); 443 ip->ip_tos = inp->inp_ip_tos; 444 if (inp->inp_flags & INP_DONTFRAG) 445 ip->ip_off = IP_DF; 446 else 447 ip->ip_off = 0; 448 ip->ip_p = inp->inp_ip_p; 449 ip->ip_len = m->m_pkthdr.len; 450 ip->ip_src = inp->inp_laddr; 451 if (jailed(inp->inp_cred)) { 452 /* 453 * prison_local_ip4() would be good enough but would 454 * let a source of INADDR_ANY pass, which we do not 455 * want to see from jails. We do not go through the 456 * pain of in_pcbladdr() for raw sockets. 457 */ 458 if (ip->ip_src.s_addr == INADDR_ANY) 459 error = prison_get_ip4(inp->inp_cred, 460 &ip->ip_src); 461 else 462 error = prison_local_ip4(inp->inp_cred, 463 &ip->ip_src); 464 if (error != 0) { 465 INP_RUNLOCK(inp); 466 m_freem(m); 467 return (error); 468 } 469 } 470 ip->ip_dst.s_addr = dst; 471 ip->ip_ttl = inp->inp_ip_ttl; 472 } else { 473 if (m->m_pkthdr.len > IP_MAXPACKET) { 474 m_freem(m); 475 return(EMSGSIZE); 476 } 477 INP_RLOCK(inp); 478 ip = mtod(m, struct ip *); 479 error = prison_check_ip4(inp->inp_cred, &ip->ip_src); 480 if (error != 0) { 481 INP_RUNLOCK(inp); 482 m_freem(m); 483 return (error); 484 } 485 486 /* 487 * Don't allow both user specified and setsockopt options, 488 * and don't allow packet length sizes that will crash. 489 */ 490 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) 491 || (ip->ip_len > m->m_pkthdr.len) 492 || (ip->ip_len < (ip->ip_hl << 2))) { 493 INP_RUNLOCK(inp); 494 m_freem(m); 495 return (EINVAL); 496 } 497 if (ip->ip_id == 0) 498 ip->ip_id = ip_newid(); 499 500 /* 501 * XXX prevent ip_output from overwriting header fields. 502 */ 503 flags |= IP_RAWOUTPUT; 504 IPSTAT_INC(ips_rawout); 505 } 506 507 if (inp->inp_flags & INP_ONESBCAST) 508 flags |= IP_SENDONES; 509 510#ifdef MAC 511 mac_inpcb_create_mbuf(inp, m); 512#endif 513 514 error = ip_output(m, inp->inp_options, NULL, flags, 515 inp->inp_moptions, inp); 516 INP_RUNLOCK(inp); 517 return (error); 518} 519 520/* 521 * Raw IP socket option processing. 522 * 523 * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could 524 * only be created by a privileged process, and as such, socket option 525 * operations to manage system properties on any raw socket were allowed to 526 * take place without explicit additional access control checks. However, 527 * raw sockets can now also be created in jail(), and therefore explicit 528 * checks are now required. Likewise, raw sockets can be used by a process 529 * after it gives up privilege, so some caution is required. For options 530 * passed down to the IP layer via ip_ctloutput(), checks are assumed to be 531 * performed in ip_ctloutput() and therefore no check occurs here. 532 * Unilaterally checking priv_check() here breaks normal IP socket option 533 * operations on raw sockets. 534 * 535 * When adding new socket options here, make sure to add access control 536 * checks here as necessary. 537 */ 538int 539rip_ctloutput(struct socket *so, struct sockopt *sopt) 540{ 541 struct inpcb *inp = sotoinpcb(so); 542 int error, optval; 543 544 if (sopt->sopt_level != IPPROTO_IP) { 545 if ((sopt->sopt_level == SOL_SOCKET) && 546 (sopt->sopt_name == SO_SETFIB)) { 547 inp->inp_inc.inc_fibnum = so->so_fibnum; 548 return (0); 549 } 550 return (EINVAL); 551 } 552 553 error = 0; 554 switch (sopt->sopt_dir) { 555 case SOPT_GET: 556 switch (sopt->sopt_name) { 557 case IP_HDRINCL: 558 optval = inp->inp_flags & INP_HDRINCL; 559 error = sooptcopyout(sopt, &optval, sizeof optval); 560 break; 561 562 case IP_FW3: /* generic ipfw v.3 functions */ 563 case IP_FW_ADD: /* ADD actually returns the body... */ 564 case IP_FW_GET: 565 case IP_FW_TABLE_GETSIZE: 566 case IP_FW_TABLE_LIST: 567 case IP_FW_NAT_GET_CONFIG: 568 case IP_FW_NAT_GET_LOG: 569 if (V_ip_fw_ctl_ptr != NULL) 570 error = V_ip_fw_ctl_ptr(sopt); 571 else 572 error = ENOPROTOOPT; 573 break; 574 575 case IP_DUMMYNET3: /* generic dummynet v.3 functions */ 576 case IP_DUMMYNET_GET: 577 if (ip_dn_ctl_ptr != NULL) 578 error = ip_dn_ctl_ptr(sopt); 579 else 580 error = ENOPROTOOPT; 581 break ; 582 583 case MRT_INIT: 584 case MRT_DONE: 585 case MRT_ADD_VIF: 586 case MRT_DEL_VIF: 587 case MRT_ADD_MFC: 588 case MRT_DEL_MFC: 589 case MRT_VERSION: 590 case MRT_ASSERT: 591 case MRT_API_SUPPORT: 592 case MRT_API_CONFIG: 593 case MRT_ADD_BW_UPCALL: 594 case MRT_DEL_BW_UPCALL: 595 error = priv_check(curthread, PRIV_NETINET_MROUTE); 596 if (error != 0) 597 return (error); 598 error = ip_mrouter_get ? ip_mrouter_get(so, sopt) : 599 EOPNOTSUPP; 600 break; 601 602 default: 603 error = ip_ctloutput(so, sopt); 604 break; 605 } 606 break; 607 608 case SOPT_SET: 609 switch (sopt->sopt_name) { 610 case IP_HDRINCL: 611 error = sooptcopyin(sopt, &optval, sizeof optval, 612 sizeof optval); 613 if (error) 614 break; 615 if (optval) 616 inp->inp_flags |= INP_HDRINCL; 617 else 618 inp->inp_flags &= ~INP_HDRINCL; 619 break; 620 621 case IP_FW3: /* generic ipfw v.3 functions */ 622 case IP_FW_ADD: 623 case IP_FW_DEL: 624 case IP_FW_FLUSH: 625 case IP_FW_ZERO: 626 case IP_FW_RESETLOG: 627 case IP_FW_TABLE_ADD: 628 case IP_FW_TABLE_DEL: 629 case IP_FW_TABLE_FLUSH: 630 case IP_FW_NAT_CFG: 631 case IP_FW_NAT_DEL: 632 if (V_ip_fw_ctl_ptr != NULL) 633 error = V_ip_fw_ctl_ptr(sopt); 634 else 635 error = ENOPROTOOPT; 636 break; 637 638 case IP_DUMMYNET3: /* generic dummynet v.3 functions */ 639 case IP_DUMMYNET_CONFIGURE: 640 case IP_DUMMYNET_DEL: 641 case IP_DUMMYNET_FLUSH: 642 if (ip_dn_ctl_ptr != NULL) 643 error = ip_dn_ctl_ptr(sopt); 644 else 645 error = ENOPROTOOPT ; 646 break ; 647 648 case IP_RSVP_ON: 649 error = priv_check(curthread, PRIV_NETINET_MROUTE); 650 if (error != 0) 651 return (error); 652 error = ip_rsvp_init(so); 653 break; 654 655 case IP_RSVP_OFF: 656 error = priv_check(curthread, PRIV_NETINET_MROUTE); 657 if (error != 0) 658 return (error); 659 error = ip_rsvp_done(); 660 break; 661 662 case IP_RSVP_VIF_ON: 663 case IP_RSVP_VIF_OFF: 664 error = priv_check(curthread, PRIV_NETINET_MROUTE); 665 if (error != 0) 666 return (error); 667 error = ip_rsvp_vif ? 668 ip_rsvp_vif(so, sopt) : EINVAL; 669 break; 670 671 case MRT_INIT: 672 case MRT_DONE: 673 case MRT_ADD_VIF: 674 case MRT_DEL_VIF: 675 case MRT_ADD_MFC: 676 case MRT_DEL_MFC: 677 case MRT_VERSION: 678 case MRT_ASSERT: 679 case MRT_API_SUPPORT: 680 case MRT_API_CONFIG: 681 case MRT_ADD_BW_UPCALL: 682 case MRT_DEL_BW_UPCALL: 683 error = priv_check(curthread, PRIV_NETINET_MROUTE); 684 if (error != 0) 685 return (error); 686 error = ip_mrouter_set ? ip_mrouter_set(so, sopt) : 687 EOPNOTSUPP; 688 break; 689 690 default: 691 error = ip_ctloutput(so, sopt); 692 break; 693 } 694 break; 695 } 696 697 return (error); 698} 699 700/* 701 * This function exists solely to receive the PRC_IFDOWN messages which are 702 * sent by if_down(). It looks for an ifaddr whose ifa_addr is sa, and calls 703 * in_ifadown() to remove all routes corresponding to that address. It also 704 * receives the PRC_IFUP messages from if_up() and reinstalls the interface 705 * routes. 706 */ 707void 708rip_ctlinput(int cmd, struct sockaddr *sa, void *vip) 709{ 710 struct in_ifaddr *ia; 711 struct ifnet *ifp; 712 int err; 713 int flags; 714 715 switch (cmd) { 716 case PRC_IFDOWN: 717 IN_IFADDR_RLOCK(); 718 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 719 if (ia->ia_ifa.ifa_addr == sa 720 && (ia->ia_flags & IFA_ROUTE)) { 721 ifa_ref(&ia->ia_ifa); 722 IN_IFADDR_RUNLOCK(); 723 /* 724 * in_ifscrub kills the interface route. 725 */ 726 in_ifscrub(ia->ia_ifp, ia, 0); 727 /* 728 * in_ifadown gets rid of all the rest of the 729 * routes. This is not quite the right thing 730 * to do, but at least if we are running a 731 * routing process they will come back. 732 */ 733 in_ifadown(&ia->ia_ifa, 0); 734 ifa_free(&ia->ia_ifa); 735 break; 736 } 737 } 738 if (ia == NULL) /* If ia matched, already unlocked. */ 739 IN_IFADDR_RUNLOCK(); 740 break; 741 742 case PRC_IFUP: 743 IN_IFADDR_RLOCK(); 744 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 745 if (ia->ia_ifa.ifa_addr == sa) 746 break; 747 } 748 if (ia == NULL || (ia->ia_flags & IFA_ROUTE)) { 749 IN_IFADDR_RUNLOCK(); 750 return; 751 } 752 ifa_ref(&ia->ia_ifa); 753 IN_IFADDR_RUNLOCK(); 754 flags = RTF_UP; 755 ifp = ia->ia_ifa.ifa_ifp; 756 757 if ((ifp->if_flags & IFF_LOOPBACK) 758 || (ifp->if_flags & IFF_POINTOPOINT)) 759 flags |= RTF_HOST; 760 761 err = ifa_del_loopback_route((struct ifaddr *)ia, sa); 762 if (err == 0) 763 ia->ia_flags &= ~IFA_RTSELF; 764 765 err = rtinit(&ia->ia_ifa, RTM_ADD, flags); 766 if (err == 0) 767 ia->ia_flags |= IFA_ROUTE; 768 769 err = ifa_add_loopback_route((struct ifaddr *)ia, sa); 770 if (err == 0) 771 ia->ia_flags |= IFA_RTSELF; 772 773 ifa_free(&ia->ia_ifa); 774 break; 775 } 776} 777 778static int 779rip_attach(struct socket *so, int proto, struct thread *td) 780{ 781 struct inpcb *inp; 782 int error; 783 784 inp = sotoinpcb(so); 785 KASSERT(inp == NULL, ("rip_attach: inp != NULL")); 786 787 error = priv_check(td, PRIV_NETINET_RAW); 788 if (error) 789 return (error); 790 if (proto >= IPPROTO_MAX || proto < 0) 791 return EPROTONOSUPPORT; 792 error = soreserve(so, rip_sendspace, rip_recvspace); 793 if (error) 794 return (error); 795 INP_INFO_WLOCK(&V_ripcbinfo); 796 error = in_pcballoc(so, &V_ripcbinfo); 797 if (error) { 798 INP_INFO_WUNLOCK(&V_ripcbinfo); 799 return (error); 800 } 801 inp = (struct inpcb *)so->so_pcb; 802 inp->inp_vflag |= INP_IPV4; 803 inp->inp_ip_p = proto; 804 inp->inp_ip_ttl = V_ip_defttl; 805 rip_inshash(inp); 806 INP_INFO_WUNLOCK(&V_ripcbinfo); 807 INP_WUNLOCK(inp); 808 return (0); 809} 810 811static void 812rip_detach(struct socket *so) 813{ 814 struct inpcb *inp; 815 816 inp = sotoinpcb(so); 817 KASSERT(inp != NULL, ("rip_detach: inp == NULL")); 818 KASSERT(inp->inp_faddr.s_addr == INADDR_ANY, 819 ("rip_detach: not closed")); 820 821 INP_INFO_WLOCK(&V_ripcbinfo); 822 INP_WLOCK(inp); 823 rip_delhash(inp); 824 if (so == V_ip_mrouter && ip_mrouter_done) 825 ip_mrouter_done(); 826 if (ip_rsvp_force_done) 827 ip_rsvp_force_done(so); 828 if (so == V_ip_rsvpd) 829 ip_rsvp_done(); 830 in_pcbdetach(inp); 831 in_pcbfree(inp); 832 INP_INFO_WUNLOCK(&V_ripcbinfo); 833} 834 835static void 836rip_dodisconnect(struct socket *so, struct inpcb *inp) 837{ 838 struct inpcbinfo *pcbinfo; 839 840 pcbinfo = inp->inp_pcbinfo; 841 INP_INFO_WLOCK(pcbinfo); 842 INP_WLOCK(inp); 843 rip_delhash(inp); 844 inp->inp_faddr.s_addr = INADDR_ANY; 845 rip_inshash(inp); 846 SOCK_LOCK(so); 847 so->so_state &= ~SS_ISCONNECTED; 848 SOCK_UNLOCK(so); 849 INP_WUNLOCK(inp); 850 INP_INFO_WUNLOCK(pcbinfo); 851} 852 853static void 854rip_abort(struct socket *so) 855{ 856 struct inpcb *inp; 857 858 inp = sotoinpcb(so); 859 KASSERT(inp != NULL, ("rip_abort: inp == NULL")); 860 861 rip_dodisconnect(so, inp); 862} 863 864static void 865rip_close(struct socket *so) 866{ 867 struct inpcb *inp; 868 869 inp = sotoinpcb(so); 870 KASSERT(inp != NULL, ("rip_close: inp == NULL")); 871 872 rip_dodisconnect(so, inp); 873} 874 875static int 876rip_disconnect(struct socket *so) 877{ 878 struct inpcb *inp; 879 880 if ((so->so_state & SS_ISCONNECTED) == 0) 881 return (ENOTCONN); 882 883 inp = sotoinpcb(so); 884 KASSERT(inp != NULL, ("rip_disconnect: inp == NULL")); 885 886 rip_dodisconnect(so, inp); 887 return (0); 888} 889 890static int 891rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 892{ 893 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 894 struct inpcb *inp; 895 int error; 896 897 if (nam->sa_len != sizeof(*addr)) 898 return (EINVAL); 899 900 error = prison_check_ip4(td->td_ucred, &addr->sin_addr); 901 if (error != 0) 902 return (error); 903 904 inp = sotoinpcb(so); 905 KASSERT(inp != NULL, ("rip_bind: inp == NULL")); 906 907 if (TAILQ_EMPTY(&V_ifnet) || 908 (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) || 909 (addr->sin_addr.s_addr && 910 (inp->inp_flags & INP_BINDANY) == 0 && 911 ifa_ifwithaddr_check((struct sockaddr *)addr) == 0)) 912 return (EADDRNOTAVAIL); 913 914 INP_INFO_WLOCK(&V_ripcbinfo); 915 INP_WLOCK(inp); 916 rip_delhash(inp); 917 inp->inp_laddr = addr->sin_addr; 918 rip_inshash(inp); 919 INP_WUNLOCK(inp); 920 INP_INFO_WUNLOCK(&V_ripcbinfo); 921 return (0); 922} 923 924static int 925rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 926{ 927 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 928 struct inpcb *inp; 929 930 if (nam->sa_len != sizeof(*addr)) 931 return (EINVAL); 932 if (TAILQ_EMPTY(&V_ifnet)) 933 return (EADDRNOTAVAIL); 934 if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) 935 return (EAFNOSUPPORT); 936 937 inp = sotoinpcb(so); 938 KASSERT(inp != NULL, ("rip_connect: inp == NULL")); 939 940 INP_INFO_WLOCK(&V_ripcbinfo); 941 INP_WLOCK(inp); 942 rip_delhash(inp); 943 inp->inp_faddr = addr->sin_addr; 944 rip_inshash(inp); 945 soisconnected(so); 946 INP_WUNLOCK(inp); 947 INP_INFO_WUNLOCK(&V_ripcbinfo); 948 return (0); 949} 950 951static int 952rip_shutdown(struct socket *so) 953{ 954 struct inpcb *inp; 955 956 inp = sotoinpcb(so); 957 KASSERT(inp != NULL, ("rip_shutdown: inp == NULL")); 958 959 INP_WLOCK(inp); 960 socantsendmore(so); 961 INP_WUNLOCK(inp); 962 return (0); 963} 964 965static int 966rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 967 struct mbuf *control, struct thread *td) 968{ 969 struct inpcb *inp; 970 u_long dst; 971 972 inp = sotoinpcb(so); 973 KASSERT(inp != NULL, ("rip_send: inp == NULL")); 974 975 /* 976 * Note: 'dst' reads below are unlocked. 977 */ 978 if (so->so_state & SS_ISCONNECTED) { 979 if (nam) { 980 m_freem(m); 981 return (EISCONN); 982 } 983 dst = inp->inp_faddr.s_addr; /* Unlocked read. */ 984 } else { 985 if (nam == NULL) { 986 m_freem(m); 987 return (ENOTCONN); 988 } 989 dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr; 990 } 991 return (rip_output(m, so, dst)); 992} 993#endif /* INET */ 994 995static int 996rip_pcblist(SYSCTL_HANDLER_ARGS) 997{ 998 int error, i, n; 999 struct inpcb *inp, **inp_list; 1000 inp_gen_t gencnt; 1001 struct xinpgen xig; 1002 1003 /* 1004 * The process of preparing the TCB list is too time-consuming and 1005 * resource-intensive to repeat twice on every request. 1006 */ 1007 if (req->oldptr == 0) { 1008 n = V_ripcbinfo.ipi_count; 1009 n += imax(n / 8, 10); 1010 req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb); 1011 return (0); 1012 } 1013 1014 if (req->newptr != 0) 1015 return (EPERM); 1016 1017 /* 1018 * OK, now we're committed to doing something. 1019 */ 1020 INP_INFO_RLOCK(&V_ripcbinfo); 1021 gencnt = V_ripcbinfo.ipi_gencnt; 1022 n = V_ripcbinfo.ipi_count; 1023 INP_INFO_RUNLOCK(&V_ripcbinfo); 1024 1025 xig.xig_len = sizeof xig; 1026 xig.xig_count = n; 1027 xig.xig_gen = gencnt; 1028 xig.xig_sogen = so_gencnt; 1029 error = SYSCTL_OUT(req, &xig, sizeof xig); 1030 if (error) 1031 return (error); 1032 1033 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 1034 if (inp_list == 0) 1035 return (ENOMEM); 1036 1037 INP_INFO_RLOCK(&V_ripcbinfo); 1038 for (inp = LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n; 1039 inp = LIST_NEXT(inp, inp_list)) { 1040 INP_WLOCK(inp); 1041 if (inp->inp_gencnt <= gencnt && 1042 cr_canseeinpcb(req->td->td_ucred, inp) == 0) { 1043 in_pcbref(inp); 1044 inp_list[i++] = inp; 1045 } 1046 INP_WUNLOCK(inp); 1047 } 1048 INP_INFO_RUNLOCK(&V_ripcbinfo); 1049 n = i; 1050 1051 error = 0; 1052 for (i = 0; i < n; i++) { 1053 inp = inp_list[i]; 1054 INP_RLOCK(inp); 1055 if (inp->inp_gencnt <= gencnt) { 1056 struct xinpcb xi; 1057 1058 bzero(&xi, sizeof(xi)); 1059 xi.xi_len = sizeof xi; 1060 /* XXX should avoid extra copy */ 1061 bcopy(inp, &xi.xi_inp, sizeof *inp); 1062 if (inp->inp_socket) 1063 sotoxsocket(inp->inp_socket, &xi.xi_socket); 1064 INP_RUNLOCK(inp); 1065 error = SYSCTL_OUT(req, &xi, sizeof xi); 1066 } else 1067 INP_RUNLOCK(inp); 1068 } 1069 INP_INFO_WLOCK(&V_ripcbinfo); 1070 for (i = 0; i < n; i++) { 1071 inp = inp_list[i]; 1072 INP_RLOCK(inp); 1073 if (!in_pcbrele_rlocked(inp)) 1074 INP_RUNLOCK(inp); 1075 } 1076 INP_INFO_WUNLOCK(&V_ripcbinfo); 1077 1078 if (!error) { 1079 /* 1080 * Give the user an updated idea of our state. If the 1081 * generation differs from what we told her before, she knows 1082 * that something happened while we were processing this 1083 * request, and it might be necessary to retry. 1084 */ 1085 INP_INFO_RLOCK(&V_ripcbinfo); 1086 xig.xig_gen = V_ripcbinfo.ipi_gencnt; 1087 xig.xig_sogen = so_gencnt; 1088 xig.xig_count = V_ripcbinfo.ipi_count; 1089 INP_INFO_RUNLOCK(&V_ripcbinfo); 1090 error = SYSCTL_OUT(req, &xig, sizeof xig); 1091 } 1092 free(inp_list, M_TEMP); 1093 return (error); 1094} 1095 1096SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, 1097 CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0, 1098 rip_pcblist, "S,xinpcb", "List of active raw IP sockets"); 1099 1100#ifdef INET 1101struct pr_usrreqs rip_usrreqs = { 1102 .pru_abort = rip_abort, 1103 .pru_attach = rip_attach, 1104 .pru_bind = rip_bind, 1105 .pru_connect = rip_connect, 1106 .pru_control = in_control, 1107 .pru_detach = rip_detach, 1108 .pru_disconnect = rip_disconnect, 1109 .pru_peeraddr = in_getpeeraddr, 1110 .pru_send = rip_send, 1111 .pru_shutdown = rip_shutdown, 1112 .pru_sockaddr = in_getsockaddr, 1113 .pru_sosetlabel = in_pcbsosetlabel, 1114 .pru_close = rip_close, 1115}; 1116#endif /* INET */ 1117