raw_ip.c revision 222143
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 222143 2011-05-20 19:12:20Z qingli $"); 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 EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL, 210 EVENTHANDLER_PRI_ANY); 211} 212 213#ifdef VIMAGE 214void 215rip_destroy(void) 216{ 217 218 in_pcbinfo_destroy(&V_ripcbinfo); 219} 220#endif 221 222#ifdef INET 223static int 224rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n, 225 struct sockaddr_in *ripsrc) 226{ 227 int policyfail = 0; 228 229 INP_RLOCK_ASSERT(last); 230 231#ifdef IPSEC 232 /* check AH/ESP integrity. */ 233 if (ipsec4_in_reject(n, last)) { 234 policyfail = 1; 235 } 236#endif /* IPSEC */ 237#ifdef MAC 238 if (!policyfail && mac_inpcb_check_deliver(last, n) != 0) 239 policyfail = 1; 240#endif 241 /* Check the minimum TTL for socket. */ 242 if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl) 243 policyfail = 1; 244 if (!policyfail) { 245 struct mbuf *opts = NULL; 246 struct socket *so; 247 248 so = last->inp_socket; 249 if ((last->inp_flags & INP_CONTROLOPTS) || 250 (so->so_options & (SO_TIMESTAMP | SO_BINTIME))) 251 ip_savecontrol(last, &opts, ip, n); 252 SOCKBUF_LOCK(&so->so_rcv); 253 if (sbappendaddr_locked(&so->so_rcv, 254 (struct sockaddr *)ripsrc, n, opts) == 0) { 255 /* should notify about lost packet */ 256 m_freem(n); 257 if (opts) 258 m_freem(opts); 259 SOCKBUF_UNLOCK(&so->so_rcv); 260 } else 261 sorwakeup_locked(so); 262 } else 263 m_freem(n); 264 return (policyfail); 265} 266 267/* 268 * Setup generic address and protocol structures for raw_input routine, then 269 * pass them along with mbuf chain. 270 */ 271void 272rip_input(struct mbuf *m, int off) 273{ 274 struct ifnet *ifp; 275 struct ip *ip = mtod(m, struct ip *); 276 int proto = ip->ip_p; 277 struct inpcb *inp, *last; 278 struct sockaddr_in ripsrc; 279 int hash; 280 281 bzero(&ripsrc, sizeof(ripsrc)); 282 ripsrc.sin_len = sizeof(ripsrc); 283 ripsrc.sin_family = AF_INET; 284 ripsrc.sin_addr = ip->ip_src; 285 last = NULL; 286 287 ifp = m->m_pkthdr.rcvif; 288 289 hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr, 290 ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask); 291 INP_INFO_RLOCK(&V_ripcbinfo); 292 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) { 293 if (inp->inp_ip_p != proto) 294 continue; 295#ifdef INET6 296 /* XXX inp locking */ 297 if ((inp->inp_vflag & INP_IPV4) == 0) 298 continue; 299#endif 300 if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr) 301 continue; 302 if (inp->inp_faddr.s_addr != ip->ip_src.s_addr) 303 continue; 304 if (jailed_without_vnet(inp->inp_cred)) { 305 /* 306 * XXX: If faddr was bound to multicast group, 307 * jailed raw socket will drop datagram. 308 */ 309 if (prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) 310 continue; 311 } 312 if (last != NULL) { 313 struct mbuf *n; 314 315 n = m_copy(m, 0, (int)M_COPYALL); 316 if (n != NULL) 317 (void) rip_append(last, ip, n, &ripsrc); 318 /* XXX count dropped packet */ 319 INP_RUNLOCK(last); 320 } 321 INP_RLOCK(inp); 322 last = inp; 323 } 324 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) { 325 if (inp->inp_ip_p && inp->inp_ip_p != proto) 326 continue; 327#ifdef INET6 328 /* XXX inp locking */ 329 if ((inp->inp_vflag & INP_IPV4) == 0) 330 continue; 331#endif 332 if (!in_nullhost(inp->inp_laddr) && 333 !in_hosteq(inp->inp_laddr, ip->ip_dst)) 334 continue; 335 if (!in_nullhost(inp->inp_faddr) && 336 !in_hosteq(inp->inp_faddr, ip->ip_src)) 337 continue; 338 if (jailed_without_vnet(inp->inp_cred)) { 339 /* 340 * Allow raw socket in jail to receive multicast; 341 * assume process had PRIV_NETINET_RAW at attach, 342 * and fall through into normal filter path if so. 343 */ 344 if (!IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 345 prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) 346 continue; 347 } 348 /* 349 * If this raw socket has multicast state, and we 350 * have received a multicast, check if this socket 351 * should receive it, as multicast filtering is now 352 * the responsibility of the transport layer. 353 */ 354 if (inp->inp_moptions != NULL && 355 IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 356 /* 357 * If the incoming datagram is for IGMP, allow it 358 * through unconditionally to the raw socket. 359 * 360 * In the case of IGMPv2, we may not have explicitly 361 * joined the group, and may have set IFF_ALLMULTI 362 * on the interface. imo_multi_filter() may discard 363 * control traffic we actually need to see. 364 * 365 * Userland multicast routing daemons should continue 366 * filter the control traffic appropriately. 367 */ 368 int blocked; 369 370 blocked = MCAST_PASS; 371 if (proto != IPPROTO_IGMP) { 372 struct sockaddr_in group; 373 374 bzero(&group, sizeof(struct sockaddr_in)); 375 group.sin_len = sizeof(struct sockaddr_in); 376 group.sin_family = AF_INET; 377 group.sin_addr = ip->ip_dst; 378 379 blocked = imo_multi_filter(inp->inp_moptions, 380 ifp, 381 (struct sockaddr *)&group, 382 (struct sockaddr *)&ripsrc); 383 } 384 385 if (blocked != MCAST_PASS) { 386 IPSTAT_INC(ips_notmember); 387 continue; 388 } 389 } 390 if (last != NULL) { 391 struct mbuf *n; 392 393 n = m_copy(m, 0, (int)M_COPYALL); 394 if (n != NULL) 395 (void) rip_append(last, ip, n, &ripsrc); 396 /* XXX count dropped packet */ 397 INP_RUNLOCK(last); 398 } 399 INP_RLOCK(inp); 400 last = inp; 401 } 402 INP_INFO_RUNLOCK(&V_ripcbinfo); 403 if (last != NULL) { 404 if (rip_append(last, ip, m, &ripsrc) != 0) 405 IPSTAT_INC(ips_delivered); 406 INP_RUNLOCK(last); 407 } else { 408 m_freem(m); 409 IPSTAT_INC(ips_noproto); 410 IPSTAT_DEC(ips_delivered); 411 } 412} 413 414/* 415 * Generate IP header and pass packet to ip_output. Tack on options user may 416 * have setup with control call. 417 */ 418int 419rip_output(struct mbuf *m, struct socket *so, u_long dst) 420{ 421 struct ip *ip; 422 int error; 423 struct inpcb *inp = sotoinpcb(so); 424 int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) | 425 IP_ALLOWBROADCAST; 426 427 /* 428 * If the user handed us a complete IP packet, use it. Otherwise, 429 * allocate an mbuf for a header and fill it in. 430 */ 431 if ((inp->inp_flags & INP_HDRINCL) == 0) { 432 if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) { 433 m_freem(m); 434 return(EMSGSIZE); 435 } 436 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT); 437 if (m == NULL) 438 return(ENOBUFS); 439 440 INP_RLOCK(inp); 441 ip = mtod(m, struct ip *); 442 ip->ip_tos = inp->inp_ip_tos; 443 if (inp->inp_flags & INP_DONTFRAG) 444 ip->ip_off = IP_DF; 445 else 446 ip->ip_off = 0; 447 ip->ip_p = inp->inp_ip_p; 448 ip->ip_len = m->m_pkthdr.len; 449 ip->ip_src = inp->inp_laddr; 450 if (jailed(inp->inp_cred)) { 451 /* 452 * prison_local_ip4() would be good enough but would 453 * let a source of INADDR_ANY pass, which we do not 454 * want to see from jails. We do not go through the 455 * pain of in_pcbladdr() for raw sockets. 456 */ 457 if (ip->ip_src.s_addr == INADDR_ANY) 458 error = prison_get_ip4(inp->inp_cred, 459 &ip->ip_src); 460 else 461 error = prison_local_ip4(inp->inp_cred, 462 &ip->ip_src); 463 if (error != 0) { 464 INP_RUNLOCK(inp); 465 m_freem(m); 466 return (error); 467 } 468 } 469 ip->ip_dst.s_addr = dst; 470 ip->ip_ttl = inp->inp_ip_ttl; 471 } else { 472 if (m->m_pkthdr.len > IP_MAXPACKET) { 473 m_freem(m); 474 return(EMSGSIZE); 475 } 476 INP_RLOCK(inp); 477 ip = mtod(m, struct ip *); 478 error = prison_check_ip4(inp->inp_cred, &ip->ip_src); 479 if (error != 0) { 480 INP_RUNLOCK(inp); 481 m_freem(m); 482 return (error); 483 } 484 485 /* 486 * Don't allow both user specified and setsockopt options, 487 * and don't allow packet length sizes that will crash. 488 */ 489 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) 490 || (ip->ip_len > m->m_pkthdr.len) 491 || (ip->ip_len < (ip->ip_hl << 2))) { 492 INP_RUNLOCK(inp); 493 m_freem(m); 494 return (EINVAL); 495 } 496 if (ip->ip_id == 0) 497 ip->ip_id = ip_newid(); 498 499 /* 500 * XXX prevent ip_output from overwriting header fields. 501 */ 502 flags |= IP_RAWOUTPUT; 503 IPSTAT_INC(ips_rawout); 504 } 505 506 if (inp->inp_flags & INP_ONESBCAST) 507 flags |= IP_SENDONES; 508 509#ifdef MAC 510 mac_inpcb_create_mbuf(inp, m); 511#endif 512 513 error = ip_output(m, inp->inp_options, NULL, flags, 514 inp->inp_moptions, inp); 515 INP_RUNLOCK(inp); 516 return (error); 517} 518 519/* 520 * Raw IP socket option processing. 521 * 522 * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could 523 * only be created by a privileged process, and as such, socket option 524 * operations to manage system properties on any raw socket were allowed to 525 * take place without explicit additional access control checks. However, 526 * raw sockets can now also be created in jail(), and therefore explicit 527 * checks are now required. Likewise, raw sockets can be used by a process 528 * after it gives up privilege, so some caution is required. For options 529 * passed down to the IP layer via ip_ctloutput(), checks are assumed to be 530 * performed in ip_ctloutput() and therefore no check occurs here. 531 * Unilaterally checking priv_check() here breaks normal IP socket option 532 * operations on raw sockets. 533 * 534 * When adding new socket options here, make sure to add access control 535 * checks here as necessary. 536 */ 537int 538rip_ctloutput(struct socket *so, struct sockopt *sopt) 539{ 540 struct inpcb *inp = sotoinpcb(so); 541 int error, optval; 542 543 if (sopt->sopt_level != IPPROTO_IP) { 544 if ((sopt->sopt_level == SOL_SOCKET) && 545 (sopt->sopt_name == SO_SETFIB)) { 546 inp->inp_inc.inc_fibnum = so->so_fibnum; 547 return (0); 548 } 549 return (EINVAL); 550 } 551 552 error = 0; 553 switch (sopt->sopt_dir) { 554 case SOPT_GET: 555 switch (sopt->sopt_name) { 556 case IP_HDRINCL: 557 optval = inp->inp_flags & INP_HDRINCL; 558 error = sooptcopyout(sopt, &optval, sizeof optval); 559 break; 560 561 case IP_FW3: /* generic ipfw v.3 functions */ 562 case IP_FW_ADD: /* ADD actually returns the body... */ 563 case IP_FW_GET: 564 case IP_FW_TABLE_GETSIZE: 565 case IP_FW_TABLE_LIST: 566 case IP_FW_NAT_GET_CONFIG: 567 case IP_FW_NAT_GET_LOG: 568 if (V_ip_fw_ctl_ptr != NULL) 569 error = V_ip_fw_ctl_ptr(sopt); 570 else 571 error = ENOPROTOOPT; 572 break; 573 574 case IP_DUMMYNET3: /* generic dummynet v.3 functions */ 575 case IP_DUMMYNET_GET: 576 if (ip_dn_ctl_ptr != NULL) 577 error = ip_dn_ctl_ptr(sopt); 578 else 579 error = ENOPROTOOPT; 580 break ; 581 582 case MRT_INIT: 583 case MRT_DONE: 584 case MRT_ADD_VIF: 585 case MRT_DEL_VIF: 586 case MRT_ADD_MFC: 587 case MRT_DEL_MFC: 588 case MRT_VERSION: 589 case MRT_ASSERT: 590 case MRT_API_SUPPORT: 591 case MRT_API_CONFIG: 592 case MRT_ADD_BW_UPCALL: 593 case MRT_DEL_BW_UPCALL: 594 error = priv_check(curthread, PRIV_NETINET_MROUTE); 595 if (error != 0) 596 return (error); 597 error = ip_mrouter_get ? ip_mrouter_get(so, sopt) : 598 EOPNOTSUPP; 599 break; 600 601 default: 602 error = ip_ctloutput(so, sopt); 603 break; 604 } 605 break; 606 607 case SOPT_SET: 608 switch (sopt->sopt_name) { 609 case IP_HDRINCL: 610 error = sooptcopyin(sopt, &optval, sizeof optval, 611 sizeof optval); 612 if (error) 613 break; 614 if (optval) 615 inp->inp_flags |= INP_HDRINCL; 616 else 617 inp->inp_flags &= ~INP_HDRINCL; 618 break; 619 620 case IP_FW3: /* generic ipfw v.3 functions */ 621 case IP_FW_ADD: 622 case IP_FW_DEL: 623 case IP_FW_FLUSH: 624 case IP_FW_ZERO: 625 case IP_FW_RESETLOG: 626 case IP_FW_TABLE_ADD: 627 case IP_FW_TABLE_DEL: 628 case IP_FW_TABLE_FLUSH: 629 case IP_FW_NAT_CFG: 630 case IP_FW_NAT_DEL: 631 if (V_ip_fw_ctl_ptr != NULL) 632 error = V_ip_fw_ctl_ptr(sopt); 633 else 634 error = ENOPROTOOPT; 635 break; 636 637 case IP_DUMMYNET3: /* generic dummynet v.3 functions */ 638 case IP_DUMMYNET_CONFIGURE: 639 case IP_DUMMYNET_DEL: 640 case IP_DUMMYNET_FLUSH: 641 if (ip_dn_ctl_ptr != NULL) 642 error = ip_dn_ctl_ptr(sopt); 643 else 644 error = ENOPROTOOPT ; 645 break ; 646 647 case IP_RSVP_ON: 648 error = priv_check(curthread, PRIV_NETINET_MROUTE); 649 if (error != 0) 650 return (error); 651 error = ip_rsvp_init(so); 652 break; 653 654 case IP_RSVP_OFF: 655 error = priv_check(curthread, PRIV_NETINET_MROUTE); 656 if (error != 0) 657 return (error); 658 error = ip_rsvp_done(); 659 break; 660 661 case IP_RSVP_VIF_ON: 662 case IP_RSVP_VIF_OFF: 663 error = priv_check(curthread, PRIV_NETINET_MROUTE); 664 if (error != 0) 665 return (error); 666 error = ip_rsvp_vif ? 667 ip_rsvp_vif(so, sopt) : EINVAL; 668 break; 669 670 case MRT_INIT: 671 case MRT_DONE: 672 case MRT_ADD_VIF: 673 case MRT_DEL_VIF: 674 case MRT_ADD_MFC: 675 case MRT_DEL_MFC: 676 case MRT_VERSION: 677 case MRT_ASSERT: 678 case MRT_API_SUPPORT: 679 case MRT_API_CONFIG: 680 case MRT_ADD_BW_UPCALL: 681 case MRT_DEL_BW_UPCALL: 682 error = priv_check(curthread, PRIV_NETINET_MROUTE); 683 if (error != 0) 684 return (error); 685 error = ip_mrouter_set ? ip_mrouter_set(so, sopt) : 686 EOPNOTSUPP; 687 break; 688 689 default: 690 error = ip_ctloutput(so, sopt); 691 break; 692 } 693 break; 694 } 695 696 return (error); 697} 698 699/* 700 * This function exists solely to receive the PRC_IFDOWN messages which are 701 * sent by if_down(). It looks for an ifaddr whose ifa_addr is sa, and calls 702 * in_ifadown() to remove all routes corresponding to that address. It also 703 * receives the PRC_IFUP messages from if_up() and reinstalls the interface 704 * routes. 705 */ 706void 707rip_ctlinput(int cmd, struct sockaddr *sa, void *vip) 708{ 709 struct in_ifaddr *ia; 710 struct ifnet *ifp; 711 int err; 712 int flags; 713 714 switch (cmd) { 715 case PRC_IFDOWN: 716 IN_IFADDR_RLOCK(); 717 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 718 if (ia->ia_ifa.ifa_addr == sa 719 && (ia->ia_flags & IFA_ROUTE)) { 720 ifa_ref(&ia->ia_ifa); 721 IN_IFADDR_RUNLOCK(); 722 /* 723 * in_ifscrub kills the interface route. 724 */ 725 in_ifscrub(ia->ia_ifp, ia, 0); 726 /* 727 * in_ifadown gets rid of all the rest of the 728 * routes. This is not quite the right thing 729 * to do, but at least if we are running a 730 * routing process they will come back. 731 */ 732 in_ifadown(&ia->ia_ifa, 0); 733 ifa_free(&ia->ia_ifa); 734 break; 735 } 736 } 737 if (ia == NULL) /* If ia matched, already unlocked. */ 738 IN_IFADDR_RUNLOCK(); 739 break; 740 741 case PRC_IFUP: 742 IN_IFADDR_RLOCK(); 743 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 744 if (ia->ia_ifa.ifa_addr == sa) 745 break; 746 } 747 if (ia == NULL || (ia->ia_flags & IFA_ROUTE)) { 748 IN_IFADDR_RUNLOCK(); 749 return; 750 } 751 ifa_ref(&ia->ia_ifa); 752 IN_IFADDR_RUNLOCK(); 753 flags = RTF_UP; 754 ifp = ia->ia_ifa.ifa_ifp; 755 756 if ((ifp->if_flags & IFF_LOOPBACK) 757 || (ifp->if_flags & IFF_POINTOPOINT)) 758 flags |= RTF_HOST; 759 760 err = ifa_del_loopback_route((struct ifaddr *)ia, sa); 761 if (err == 0) 762 ia->ia_flags &= ~IFA_RTSELF; 763 764 err = rtinit(&ia->ia_ifa, RTM_ADD, flags); 765 if (err == 0) 766 ia->ia_flags |= IFA_ROUTE; 767 768 err = ifa_add_loopback_route((struct ifaddr *)ia, sa); 769 if (err == 0) 770 ia->ia_flags |= IFA_RTSELF; 771 772 ifa_free(&ia->ia_ifa); 773 break; 774 } 775} 776 777static int 778rip_attach(struct socket *so, int proto, struct thread *td) 779{ 780 struct inpcb *inp; 781 int error; 782 783 inp = sotoinpcb(so); 784 KASSERT(inp == NULL, ("rip_attach: inp != NULL")); 785 786 error = priv_check(td, PRIV_NETINET_RAW); 787 if (error) 788 return (error); 789 if (proto >= IPPROTO_MAX || proto < 0) 790 return EPROTONOSUPPORT; 791 error = soreserve(so, rip_sendspace, rip_recvspace); 792 if (error) 793 return (error); 794 INP_INFO_WLOCK(&V_ripcbinfo); 795 error = in_pcballoc(so, &V_ripcbinfo); 796 if (error) { 797 INP_INFO_WUNLOCK(&V_ripcbinfo); 798 return (error); 799 } 800 inp = (struct inpcb *)so->so_pcb; 801 inp->inp_vflag |= INP_IPV4; 802 inp->inp_ip_p = proto; 803 inp->inp_ip_ttl = V_ip_defttl; 804 rip_inshash(inp); 805 INP_INFO_WUNLOCK(&V_ripcbinfo); 806 INP_WUNLOCK(inp); 807 return (0); 808} 809 810static void 811rip_detach(struct socket *so) 812{ 813 struct inpcb *inp; 814 815 inp = sotoinpcb(so); 816 KASSERT(inp != NULL, ("rip_detach: inp == NULL")); 817 KASSERT(inp->inp_faddr.s_addr == INADDR_ANY, 818 ("rip_detach: not closed")); 819 820 INP_INFO_WLOCK(&V_ripcbinfo); 821 INP_WLOCK(inp); 822 rip_delhash(inp); 823 if (so == V_ip_mrouter && ip_mrouter_done) 824 ip_mrouter_done(); 825 if (ip_rsvp_force_done) 826 ip_rsvp_force_done(so); 827 if (so == V_ip_rsvpd) 828 ip_rsvp_done(); 829 in_pcbdetach(inp); 830 in_pcbfree(inp); 831 INP_INFO_WUNLOCK(&V_ripcbinfo); 832} 833 834static void 835rip_dodisconnect(struct socket *so, struct inpcb *inp) 836{ 837 838 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 839 INP_WLOCK_ASSERT(inp); 840 841 rip_delhash(inp); 842 inp->inp_faddr.s_addr = INADDR_ANY; 843 rip_inshash(inp); 844 SOCK_LOCK(so); 845 so->so_state &= ~SS_ISCONNECTED; 846 SOCK_UNLOCK(so); 847} 848 849static void 850rip_abort(struct socket *so) 851{ 852 struct inpcb *inp; 853 854 inp = sotoinpcb(so); 855 KASSERT(inp != NULL, ("rip_abort: inp == NULL")); 856 857 INP_INFO_WLOCK(&V_ripcbinfo); 858 INP_WLOCK(inp); 859 rip_dodisconnect(so, inp); 860 INP_WUNLOCK(inp); 861 INP_INFO_WUNLOCK(&V_ripcbinfo); 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 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} 878 879static int 880rip_disconnect(struct socket *so) 881{ 882 struct inpcb *inp; 883 884 if ((so->so_state & SS_ISCONNECTED) == 0) 885 return (ENOTCONN); 886 887 inp = sotoinpcb(so); 888 KASSERT(inp != NULL, ("rip_disconnect: inp == NULL")); 889 890 INP_INFO_WLOCK(&V_ripcbinfo); 891 INP_WLOCK(inp); 892 rip_dodisconnect(so, inp); 893 INP_WUNLOCK(inp); 894 INP_INFO_WUNLOCK(&V_ripcbinfo); 895 return (0); 896} 897 898static int 899rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 900{ 901 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 902 struct inpcb *inp; 903 int error; 904 905 if (nam->sa_len != sizeof(*addr)) 906 return (EINVAL); 907 908 error = prison_check_ip4(td->td_ucred, &addr->sin_addr); 909 if (error != 0) 910 return (error); 911 912 inp = sotoinpcb(so); 913 KASSERT(inp != NULL, ("rip_bind: inp == NULL")); 914 915 if (TAILQ_EMPTY(&V_ifnet) || 916 (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) || 917 (addr->sin_addr.s_addr && 918 (inp->inp_flags & INP_BINDANY) == 0 && 919 ifa_ifwithaddr_check((struct sockaddr *)addr) == 0)) 920 return (EADDRNOTAVAIL); 921 922 INP_INFO_WLOCK(&V_ripcbinfo); 923 INP_WLOCK(inp); 924 rip_delhash(inp); 925 inp->inp_laddr = addr->sin_addr; 926 rip_inshash(inp); 927 INP_WUNLOCK(inp); 928 INP_INFO_WUNLOCK(&V_ripcbinfo); 929 return (0); 930} 931 932static int 933rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 934{ 935 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 936 struct inpcb *inp; 937 938 if (nam->sa_len != sizeof(*addr)) 939 return (EINVAL); 940 if (TAILQ_EMPTY(&V_ifnet)) 941 return (EADDRNOTAVAIL); 942 if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) 943 return (EAFNOSUPPORT); 944 945 inp = sotoinpcb(so); 946 KASSERT(inp != NULL, ("rip_connect: inp == NULL")); 947 948 INP_INFO_WLOCK(&V_ripcbinfo); 949 INP_WLOCK(inp); 950 rip_delhash(inp); 951 inp->inp_faddr = addr->sin_addr; 952 rip_inshash(inp); 953 soisconnected(so); 954 INP_WUNLOCK(inp); 955 INP_INFO_WUNLOCK(&V_ripcbinfo); 956 return (0); 957} 958 959static int 960rip_shutdown(struct socket *so) 961{ 962 struct inpcb *inp; 963 964 inp = sotoinpcb(so); 965 KASSERT(inp != NULL, ("rip_shutdown: inp == NULL")); 966 967 INP_WLOCK(inp); 968 socantsendmore(so); 969 INP_WUNLOCK(inp); 970 return (0); 971} 972 973static int 974rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 975 struct mbuf *control, struct thread *td) 976{ 977 struct inpcb *inp; 978 u_long dst; 979 980 inp = sotoinpcb(so); 981 KASSERT(inp != NULL, ("rip_send: inp == NULL")); 982 983 /* 984 * Note: 'dst' reads below are unlocked. 985 */ 986 if (so->so_state & SS_ISCONNECTED) { 987 if (nam) { 988 m_freem(m); 989 return (EISCONN); 990 } 991 dst = inp->inp_faddr.s_addr; /* Unlocked read. */ 992 } else { 993 if (nam == NULL) { 994 m_freem(m); 995 return (ENOTCONN); 996 } 997 dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr; 998 } 999 return (rip_output(m, so, dst)); 1000} 1001#endif /* INET */ 1002 1003static int 1004rip_pcblist(SYSCTL_HANDLER_ARGS) 1005{ 1006 int error, i, n; 1007 struct inpcb *inp, **inp_list; 1008 inp_gen_t gencnt; 1009 struct xinpgen xig; 1010 1011 /* 1012 * The process of preparing the TCB list is too time-consuming and 1013 * resource-intensive to repeat twice on every request. 1014 */ 1015 if (req->oldptr == 0) { 1016 n = V_ripcbinfo.ipi_count; 1017 n += imax(n / 8, 10); 1018 req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb); 1019 return (0); 1020 } 1021 1022 if (req->newptr != 0) 1023 return (EPERM); 1024 1025 /* 1026 * OK, now we're committed to doing something. 1027 */ 1028 INP_INFO_RLOCK(&V_ripcbinfo); 1029 gencnt = V_ripcbinfo.ipi_gencnt; 1030 n = V_ripcbinfo.ipi_count; 1031 INP_INFO_RUNLOCK(&V_ripcbinfo); 1032 1033 xig.xig_len = sizeof xig; 1034 xig.xig_count = n; 1035 xig.xig_gen = gencnt; 1036 xig.xig_sogen = so_gencnt; 1037 error = SYSCTL_OUT(req, &xig, sizeof xig); 1038 if (error) 1039 return (error); 1040 1041 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 1042 if (inp_list == 0) 1043 return (ENOMEM); 1044 1045 INP_INFO_RLOCK(&V_ripcbinfo); 1046 for (inp = LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n; 1047 inp = LIST_NEXT(inp, inp_list)) { 1048 INP_WLOCK(inp); 1049 if (inp->inp_gencnt <= gencnt && 1050 cr_canseeinpcb(req->td->td_ucred, inp) == 0) { 1051 in_pcbref(inp); 1052 inp_list[i++] = inp; 1053 } 1054 INP_WUNLOCK(inp); 1055 } 1056 INP_INFO_RUNLOCK(&V_ripcbinfo); 1057 n = i; 1058 1059 error = 0; 1060 for (i = 0; i < n; i++) { 1061 inp = inp_list[i]; 1062 INP_RLOCK(inp); 1063 if (inp->inp_gencnt <= gencnt) { 1064 struct xinpcb xi; 1065 1066 bzero(&xi, sizeof(xi)); 1067 xi.xi_len = sizeof xi; 1068 /* XXX should avoid extra copy */ 1069 bcopy(inp, &xi.xi_inp, sizeof *inp); 1070 if (inp->inp_socket) 1071 sotoxsocket(inp->inp_socket, &xi.xi_socket); 1072 INP_RUNLOCK(inp); 1073 error = SYSCTL_OUT(req, &xi, sizeof xi); 1074 } else 1075 INP_RUNLOCK(inp); 1076 } 1077 INP_INFO_WLOCK(&V_ripcbinfo); 1078 for (i = 0; i < n; i++) { 1079 inp = inp_list[i]; 1080 INP_WLOCK(inp); 1081 if (!in_pcbrele(inp)) 1082 INP_WUNLOCK(inp); 1083 } 1084 INP_INFO_WUNLOCK(&V_ripcbinfo); 1085 1086 if (!error) { 1087 /* 1088 * Give the user an updated idea of our state. If the 1089 * generation differs from what we told her before, she knows 1090 * that something happened while we were processing this 1091 * request, and it might be necessary to retry. 1092 */ 1093 INP_INFO_RLOCK(&V_ripcbinfo); 1094 xig.xig_gen = V_ripcbinfo.ipi_gencnt; 1095 xig.xig_sogen = so_gencnt; 1096 xig.xig_count = V_ripcbinfo.ipi_count; 1097 INP_INFO_RUNLOCK(&V_ripcbinfo); 1098 error = SYSCTL_OUT(req, &xig, sizeof xig); 1099 } 1100 free(inp_list, M_TEMP); 1101 return (error); 1102} 1103 1104SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, 1105 CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0, 1106 rip_pcblist, "S,xinpcb", "List of active raw IP sockets"); 1107 1108#ifdef INET 1109struct pr_usrreqs rip_usrreqs = { 1110 .pru_abort = rip_abort, 1111 .pru_attach = rip_attach, 1112 .pru_bind = rip_bind, 1113 .pru_connect = rip_connect, 1114 .pru_control = in_control, 1115 .pru_detach = rip_detach, 1116 .pru_disconnect = rip_disconnect, 1117 .pru_peeraddr = in_getpeeraddr, 1118 .pru_send = rip_send, 1119 .pru_shutdown = rip_shutdown, 1120 .pru_sockaddr = in_getsockaddr, 1121 .pru_sosetlabel = in_pcbsosetlabel, 1122 .pru_close = rip_close, 1123}; 1124#endif /* INET */ 1125