raw_ip.c revision 193217
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 193217 2009-06-01 10:30:00Z pjd $"); 35 36#include "opt_inet6.h" 37#include "opt_ipsec.h" 38#include "opt_route.h" 39#include "opt_mac.h" 40 41#include <sys/param.h> 42#include <sys/jail.h> 43#include <sys/kernel.h> 44#include <sys/lock.h> 45#include <sys/malloc.h> 46#include <sys/mbuf.h> 47#include <sys/priv.h> 48#include <sys/proc.h> 49#include <sys/protosw.h> 50#include <sys/rwlock.h> 51#include <sys/signalvar.h> 52#include <sys/socket.h> 53#include <sys/socketvar.h> 54#include <sys/sx.h> 55#include <sys/sysctl.h> 56#include <sys/systm.h> 57#include <sys/vimage.h> 58 59#include <vm/uma.h> 60 61#include <net/if.h> 62#include <net/route.h> 63#include <net/vnet.h> 64 65#include <netinet/in.h> 66#include <netinet/in_systm.h> 67#include <netinet/in_pcb.h> 68#include <netinet/in_var.h> 69#include <netinet/ip.h> 70#include <netinet/ip_var.h> 71#include <netinet/ip_mroute.h> 72 73#include <netinet/ip_fw.h> 74#include <netinet/ip_dummynet.h> 75#include <netinet/vinet.h> 76 77#ifdef IPSEC 78#include <netipsec/ipsec.h> 79#endif /*IPSEC*/ 80 81#include <security/mac/mac_framework.h> 82 83#ifdef VIMAGE_GLOBALS 84struct inpcbhead ripcb; 85struct inpcbinfo ripcbinfo; 86#endif 87 88/* control hooks for ipfw and dummynet */ 89ip_fw_ctl_t *ip_fw_ctl_ptr = NULL; 90ip_dn_ctl_t *ip_dn_ctl_ptr = NULL; 91 92/* 93 * Hooks for multicast routing. They all default to NULL, so leave them not 94 * initialized and rely on BSS being set to 0. 95 */ 96 97/* 98 * The socket used to communicate with the multicast routing daemon. 99 */ 100#ifdef VIMAGE_GLOBALS 101struct socket *ip_mrouter; 102#endif 103 104/* 105 * The various mrouter and rsvp functions. 106 */ 107int (*ip_mrouter_set)(struct socket *, struct sockopt *); 108int (*ip_mrouter_get)(struct socket *, struct sockopt *); 109int (*ip_mrouter_done)(void); 110int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 111 struct ip_moptions *); 112int (*mrt_ioctl)(int, caddr_t, int); 113int (*legal_vif_num)(int); 114u_long (*ip_mcast_src)(int); 115 116void (*rsvp_input_p)(struct mbuf *m, int off); 117int (*ip_rsvp_vif)(struct socket *, struct sockopt *); 118void (*ip_rsvp_force_done)(struct socket *); 119 120/* 121 * Hash functions 122 */ 123 124#define INP_PCBHASH_RAW_SIZE 256 125#define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \ 126 (((proto) + (laddr) + (faddr)) % (mask) + 1) 127 128static void 129rip_inshash(struct inpcb *inp) 130{ 131 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 132 struct inpcbhead *pcbhash; 133 int hash; 134 135 INP_INFO_WLOCK_ASSERT(pcbinfo); 136 INP_WLOCK_ASSERT(inp); 137 138 if (inp->inp_ip_p != 0 && 139 inp->inp_laddr.s_addr != INADDR_ANY && 140 inp->inp_faddr.s_addr != INADDR_ANY) { 141 hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr, 142 inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask); 143 } else 144 hash = 0; 145 pcbhash = &pcbinfo->ipi_hashbase[hash]; 146 LIST_INSERT_HEAD(pcbhash, inp, inp_hash); 147} 148 149static void 150rip_delhash(struct inpcb *inp) 151{ 152 153 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 154 INP_WLOCK_ASSERT(inp); 155 156 LIST_REMOVE(inp, inp_hash); 157} 158 159/* 160 * Raw interface to IP protocol. 161 */ 162 163/* 164 * Initialize raw connection block q. 165 */ 166static void 167rip_zone_change(void *tag) 168{ 169 INIT_VNET_INET(curvnet); 170 171 uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets); 172} 173 174static int 175rip_inpcb_init(void *mem, int size, int flags) 176{ 177 struct inpcb *inp = mem; 178 179 INP_LOCK_INIT(inp, "inp", "rawinp"); 180 return (0); 181} 182 183void 184rip_init(void) 185{ 186 INIT_VNET_INET(curvnet); 187 188 INP_INFO_LOCK_INIT(&V_ripcbinfo, "rip"); 189 LIST_INIT(&V_ripcb); 190#ifdef VIMAGE 191 V_ripcbinfo.ipi_vnet = curvnet; 192#endif 193 V_ripcbinfo.ipi_listhead = &V_ripcb; 194 V_ripcbinfo.ipi_hashbase = 195 hashinit(INP_PCBHASH_RAW_SIZE, M_PCB, &V_ripcbinfo.ipi_hashmask); 196 V_ripcbinfo.ipi_porthashbase = 197 hashinit(1, M_PCB, &V_ripcbinfo.ipi_porthashmask); 198 V_ripcbinfo.ipi_zone = uma_zcreate("ripcb", sizeof(struct inpcb), 199 NULL, NULL, rip_inpcb_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 200 uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets); 201 EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL, 202 EVENTHANDLER_PRI_ANY); 203} 204 205static int 206rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n, 207 struct sockaddr_in *ripsrc) 208{ 209 int policyfail = 0; 210 211 INP_RLOCK_ASSERT(last); 212 213#ifdef IPSEC 214 /* check AH/ESP integrity. */ 215 if (ipsec4_in_reject(n, last)) { 216 policyfail = 1; 217 } 218#endif /* IPSEC */ 219#ifdef MAC 220 if (!policyfail && mac_inpcb_check_deliver(last, n) != 0) 221 policyfail = 1; 222#endif 223 /* Check the minimum TTL for socket. */ 224 if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl) 225 policyfail = 1; 226 if (!policyfail) { 227 struct mbuf *opts = NULL; 228 struct socket *so; 229 230 so = last->inp_socket; 231 if ((last->inp_flags & INP_CONTROLOPTS) || 232 (so->so_options & (SO_TIMESTAMP | SO_BINTIME))) 233 ip_savecontrol(last, &opts, ip, n); 234 SOCKBUF_LOCK(&so->so_rcv); 235 if (sbappendaddr_locked(&so->so_rcv, 236 (struct sockaddr *)ripsrc, n, opts) == 0) { 237 /* should notify about lost packet */ 238 m_freem(n); 239 if (opts) 240 m_freem(opts); 241 SOCKBUF_UNLOCK(&so->so_rcv); 242 } else 243 sorwakeup_locked(so); 244 } else 245 m_freem(n); 246 return (policyfail); 247} 248 249/* 250 * Setup generic address and protocol structures for raw_input routine, then 251 * pass them along with mbuf chain. 252 */ 253void 254rip_input(struct mbuf *m, int off) 255{ 256 INIT_VNET_INET(curvnet); 257 struct ifnet *ifp; 258 struct ip *ip = mtod(m, struct ip *); 259 int proto = ip->ip_p; 260 struct inpcb *inp, *last; 261 struct sockaddr_in ripsrc; 262 int hash; 263 264 bzero(&ripsrc, sizeof(ripsrc)); 265 ripsrc.sin_len = sizeof(ripsrc); 266 ripsrc.sin_family = AF_INET; 267 ripsrc.sin_addr = ip->ip_src; 268 last = NULL; 269 270 ifp = m->m_pkthdr.rcvif; 271 272 hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr, 273 ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask); 274 INP_INFO_RLOCK(&V_ripcbinfo); 275 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) { 276 if (inp->inp_ip_p != proto) 277 continue; 278#ifdef INET6 279 /* XXX inp locking */ 280 if ((inp->inp_vflag & INP_IPV4) == 0) 281 continue; 282#endif 283 if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr) 284 continue; 285 if (inp->inp_faddr.s_addr != ip->ip_src.s_addr) 286 continue; 287 if (jailed(inp->inp_cred)) { 288 /* 289 * XXX: If faddr was bound to multicast group, 290 * jailed raw socket will drop datagram. 291 */ 292 if (prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) 293 continue; 294 } 295 if (last != NULL) { 296 struct mbuf *n; 297 298 n = m_copy(m, 0, (int)M_COPYALL); 299 if (n != NULL) 300 (void) rip_append(last, ip, n, &ripsrc); 301 /* XXX count dropped packet */ 302 INP_RUNLOCK(last); 303 } 304 INP_RLOCK(inp); 305 last = inp; 306 } 307 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) { 308 if (inp->inp_ip_p && inp->inp_ip_p != proto) 309 continue; 310#ifdef INET6 311 /* XXX inp locking */ 312 if ((inp->inp_vflag & INP_IPV4) == 0) 313 continue; 314#endif 315 if (!in_nullhost(inp->inp_laddr) && 316 !in_hosteq(inp->inp_laddr, ip->ip_dst)) 317 continue; 318 if (!in_nullhost(inp->inp_faddr) && 319 !in_hosteq(inp->inp_faddr, ip->ip_src)) 320 continue; 321 if (jailed(inp->inp_cred)) { 322 /* 323 * Allow raw socket in jail to receive multicast; 324 * assume process had PRIV_NETINET_RAW at attach, 325 * and fall through into normal filter path if so. 326 */ 327 if (!IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 328 prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) 329 continue; 330 } 331 /* 332 * If this raw socket has multicast state, and we 333 * have received a multicast, check if this socket 334 * should receive it, as multicast filtering is now 335 * the responsibility of the transport layer. 336 */ 337 if (inp->inp_moptions != NULL && 338 IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 339 struct sockaddr_in group; 340 int blocked; 341 342 bzero(&group, sizeof(struct sockaddr_in)); 343 group.sin_len = sizeof(struct sockaddr_in); 344 group.sin_family = AF_INET; 345 group.sin_addr = ip->ip_dst; 346 347 blocked = imo_multi_filter(inp->inp_moptions, ifp, 348 (struct sockaddr *)&group, 349 (struct sockaddr *)&ripsrc); 350 if (blocked != MCAST_PASS) { 351 IPSTAT_INC(ips_notmember); 352 continue; 353 } 354 } 355 if (last != NULL) { 356 struct mbuf *n; 357 358 n = m_copy(m, 0, (int)M_COPYALL); 359 if (n != NULL) 360 (void) rip_append(last, ip, n, &ripsrc); 361 /* XXX count dropped packet */ 362 INP_RUNLOCK(last); 363 } 364 INP_RLOCK(inp); 365 last = inp; 366 } 367 INP_INFO_RUNLOCK(&V_ripcbinfo); 368 if (last != NULL) { 369 if (rip_append(last, ip, m, &ripsrc) != 0) 370 IPSTAT_INC(ips_delivered); 371 INP_RUNLOCK(last); 372 } else { 373 m_freem(m); 374 IPSTAT_INC(ips_noproto); 375 IPSTAT_DEC(ips_delivered); 376 } 377} 378 379/* 380 * Generate IP header and pass packet to ip_output. Tack on options user may 381 * have setup with control call. 382 */ 383int 384rip_output(struct mbuf *m, struct socket *so, u_long dst) 385{ 386 INIT_VNET_INET(so->so_vnet); 387 struct ip *ip; 388 int error; 389 struct inpcb *inp = sotoinpcb(so); 390 int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) | 391 IP_ALLOWBROADCAST; 392 393 /* 394 * If the user handed us a complete IP packet, use it. Otherwise, 395 * allocate an mbuf for a header and fill it in. 396 */ 397 if ((inp->inp_flags & INP_HDRINCL) == 0) { 398 if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) { 399 m_freem(m); 400 return(EMSGSIZE); 401 } 402 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT); 403 if (m == NULL) 404 return(ENOBUFS); 405 406 INP_RLOCK(inp); 407 ip = mtod(m, struct ip *); 408 ip->ip_tos = inp->inp_ip_tos; 409 if (inp->inp_flags & INP_DONTFRAG) 410 ip->ip_off = IP_DF; 411 else 412 ip->ip_off = 0; 413 ip->ip_p = inp->inp_ip_p; 414 ip->ip_len = m->m_pkthdr.len; 415 ip->ip_src = inp->inp_laddr; 416 error = prison_get_ip4(inp->inp_cred, &ip->ip_src); 417 if (error != 0) { 418 INP_RUNLOCK(inp); 419 m_freem(m); 420 return (error); 421 } 422 ip->ip_dst.s_addr = dst; 423 ip->ip_ttl = inp->inp_ip_ttl; 424 } else { 425 if (m->m_pkthdr.len > IP_MAXPACKET) { 426 m_freem(m); 427 return(EMSGSIZE); 428 } 429 INP_RLOCK(inp); 430 ip = mtod(m, struct ip *); 431 error = prison_check_ip4(inp->inp_cred, &ip->ip_src); 432 if (error != 0) { 433 INP_RUNLOCK(inp); 434 m_freem(m); 435 return (error); 436 } 437 438 /* 439 * Don't allow both user specified and setsockopt options, 440 * and don't allow packet length sizes that will crash. 441 */ 442 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) 443 || (ip->ip_len > m->m_pkthdr.len) 444 || (ip->ip_len < (ip->ip_hl << 2))) { 445 INP_RUNLOCK(inp); 446 m_freem(m); 447 return (EINVAL); 448 } 449 if (ip->ip_id == 0) 450 ip->ip_id = ip_newid(); 451 452 /* 453 * XXX prevent ip_output from overwriting header fields. 454 */ 455 flags |= IP_RAWOUTPUT; 456 IPSTAT_INC(ips_rawout); 457 } 458 459 if (inp->inp_flags & INP_ONESBCAST) 460 flags |= IP_SENDONES; 461 462#ifdef MAC 463 mac_inpcb_create_mbuf(inp, m); 464#endif 465 466 error = ip_output(m, inp->inp_options, NULL, flags, 467 inp->inp_moptions, inp); 468 INP_RUNLOCK(inp); 469 return (error); 470} 471 472/* 473 * Raw IP socket option processing. 474 * 475 * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could 476 * only be created by a privileged process, and as such, socket option 477 * operations to manage system properties on any raw socket were allowed to 478 * take place without explicit additional access control checks. However, 479 * raw sockets can now also be created in jail(), and therefore explicit 480 * checks are now required. Likewise, raw sockets can be used by a process 481 * after it gives up privilege, so some caution is required. For options 482 * passed down to the IP layer via ip_ctloutput(), checks are assumed to be 483 * performed in ip_ctloutput() and therefore no check occurs here. 484 * Unilaterally checking priv_check() here breaks normal IP socket option 485 * operations on raw sockets. 486 * 487 * When adding new socket options here, make sure to add access control 488 * checks here as necessary. 489 */ 490int 491rip_ctloutput(struct socket *so, struct sockopt *sopt) 492{ 493 struct inpcb *inp = sotoinpcb(so); 494 int error, optval; 495 496 if (sopt->sopt_level != IPPROTO_IP) { 497 if ((sopt->sopt_level == SOL_SOCKET) && 498 (sopt->sopt_name == SO_SETFIB)) { 499 inp->inp_inc.inc_fibnum = so->so_fibnum; 500 return (0); 501 } 502 return (EINVAL); 503 } 504 505 error = 0; 506 switch (sopt->sopt_dir) { 507 case SOPT_GET: 508 switch (sopt->sopt_name) { 509 case IP_HDRINCL: 510 optval = inp->inp_flags & INP_HDRINCL; 511 error = sooptcopyout(sopt, &optval, sizeof optval); 512 break; 513 514 case IP_FW_ADD: /* ADD actually returns the body... */ 515 case IP_FW_GET: 516 case IP_FW_TABLE_GETSIZE: 517 case IP_FW_TABLE_LIST: 518 case IP_FW_NAT_GET_CONFIG: 519 case IP_FW_NAT_GET_LOG: 520 if (ip_fw_ctl_ptr != NULL) 521 error = ip_fw_ctl_ptr(sopt); 522 else 523 error = ENOPROTOOPT; 524 break; 525 526 case IP_DUMMYNET_GET: 527 if (ip_dn_ctl_ptr != NULL) 528 error = ip_dn_ctl_ptr(sopt); 529 else 530 error = ENOPROTOOPT; 531 break ; 532 533 case MRT_INIT: 534 case MRT_DONE: 535 case MRT_ADD_VIF: 536 case MRT_DEL_VIF: 537 case MRT_ADD_MFC: 538 case MRT_DEL_MFC: 539 case MRT_VERSION: 540 case MRT_ASSERT: 541 case MRT_API_SUPPORT: 542 case MRT_API_CONFIG: 543 case MRT_ADD_BW_UPCALL: 544 case MRT_DEL_BW_UPCALL: 545 error = priv_check(curthread, PRIV_NETINET_MROUTE); 546 if (error != 0) 547 return (error); 548 error = ip_mrouter_get ? ip_mrouter_get(so, sopt) : 549 EOPNOTSUPP; 550 break; 551 552 default: 553 error = ip_ctloutput(so, sopt); 554 break; 555 } 556 break; 557 558 case SOPT_SET: 559 switch (sopt->sopt_name) { 560 case IP_HDRINCL: 561 error = sooptcopyin(sopt, &optval, sizeof optval, 562 sizeof optval); 563 if (error) 564 break; 565 if (optval) 566 inp->inp_flags |= INP_HDRINCL; 567 else 568 inp->inp_flags &= ~INP_HDRINCL; 569 break; 570 571 case IP_FW_ADD: 572 case IP_FW_DEL: 573 case IP_FW_FLUSH: 574 case IP_FW_ZERO: 575 case IP_FW_RESETLOG: 576 case IP_FW_TABLE_ADD: 577 case IP_FW_TABLE_DEL: 578 case IP_FW_TABLE_FLUSH: 579 case IP_FW_NAT_CFG: 580 case IP_FW_NAT_DEL: 581 if (ip_fw_ctl_ptr != NULL) 582 error = ip_fw_ctl_ptr(sopt); 583 else 584 error = ENOPROTOOPT; 585 break; 586 587 case IP_DUMMYNET_CONFIGURE: 588 case IP_DUMMYNET_DEL: 589 case IP_DUMMYNET_FLUSH: 590 if (ip_dn_ctl_ptr != NULL) 591 error = ip_dn_ctl_ptr(sopt); 592 else 593 error = ENOPROTOOPT ; 594 break ; 595 596 case IP_RSVP_ON: 597 error = priv_check(curthread, PRIV_NETINET_MROUTE); 598 if (error != 0) 599 return (error); 600 error = ip_rsvp_init(so); 601 break; 602 603 case IP_RSVP_OFF: 604 error = priv_check(curthread, PRIV_NETINET_MROUTE); 605 if (error != 0) 606 return (error); 607 error = ip_rsvp_done(); 608 break; 609 610 case IP_RSVP_VIF_ON: 611 case IP_RSVP_VIF_OFF: 612 error = priv_check(curthread, PRIV_NETINET_MROUTE); 613 if (error != 0) 614 return (error); 615 error = ip_rsvp_vif ? 616 ip_rsvp_vif(so, sopt) : EINVAL; 617 break; 618 619 case MRT_INIT: 620 case MRT_DONE: 621 case MRT_ADD_VIF: 622 case MRT_DEL_VIF: 623 case MRT_ADD_MFC: 624 case MRT_DEL_MFC: 625 case MRT_VERSION: 626 case MRT_ASSERT: 627 case MRT_API_SUPPORT: 628 case MRT_API_CONFIG: 629 case MRT_ADD_BW_UPCALL: 630 case MRT_DEL_BW_UPCALL: 631 error = priv_check(curthread, PRIV_NETINET_MROUTE); 632 if (error != 0) 633 return (error); 634 error = ip_mrouter_set ? ip_mrouter_set(so, sopt) : 635 EOPNOTSUPP; 636 break; 637 638 default: 639 error = ip_ctloutput(so, sopt); 640 break; 641 } 642 break; 643 } 644 645 return (error); 646} 647 648/* 649 * This function exists solely to receive the PRC_IFDOWN messages which are 650 * sent by if_down(). It looks for an ifaddr whose ifa_addr is sa, and calls 651 * in_ifadown() to remove all routes corresponding to that address. It also 652 * receives the PRC_IFUP messages from if_up() and reinstalls the interface 653 * routes. 654 */ 655void 656rip_ctlinput(int cmd, struct sockaddr *sa, void *vip) 657{ 658 INIT_VNET_INET(curvnet); 659 struct in_ifaddr *ia; 660 struct ifnet *ifp; 661 int err; 662 int flags; 663 664 switch (cmd) { 665 case PRC_IFDOWN: 666 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 667 if (ia->ia_ifa.ifa_addr == sa 668 && (ia->ia_flags & IFA_ROUTE)) { 669 /* 670 * in_ifscrub kills the interface route. 671 */ 672 in_ifscrub(ia->ia_ifp, ia); 673 /* 674 * in_ifadown gets rid of all the rest of the 675 * routes. This is not quite the right thing 676 * to do, but at least if we are running a 677 * routing process they will come back. 678 */ 679 in_ifadown(&ia->ia_ifa, 0); 680 break; 681 } 682 } 683 break; 684 685 case PRC_IFUP: 686 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 687 if (ia->ia_ifa.ifa_addr == sa) 688 break; 689 } 690 if (ia == 0 || (ia->ia_flags & IFA_ROUTE)) 691 return; 692 flags = RTF_UP; 693 ifp = ia->ia_ifa.ifa_ifp; 694 695 if ((ifp->if_flags & IFF_LOOPBACK) 696 || (ifp->if_flags & IFF_POINTOPOINT)) 697 flags |= RTF_HOST; 698 699 err = rtinit(&ia->ia_ifa, RTM_ADD, flags); 700 if (err == 0) 701 ia->ia_flags |= IFA_ROUTE; 702 break; 703 } 704} 705 706u_long rip_sendspace = 9216; 707u_long rip_recvspace = 9216; 708 709SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW, 710 &rip_sendspace, 0, "Maximum outgoing raw IP datagram size"); 711SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW, 712 &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams"); 713 714static int 715rip_attach(struct socket *so, int proto, struct thread *td) 716{ 717 INIT_VNET_INET(so->so_vnet); 718 struct inpcb *inp; 719 int error; 720 721 inp = sotoinpcb(so); 722 KASSERT(inp == NULL, ("rip_attach: inp != NULL")); 723 724 error = priv_check(td, PRIV_NETINET_RAW); 725 if (error) 726 return (error); 727 if (proto >= IPPROTO_MAX || proto < 0) 728 return EPROTONOSUPPORT; 729 error = soreserve(so, rip_sendspace, rip_recvspace); 730 if (error) 731 return (error); 732 INP_INFO_WLOCK(&V_ripcbinfo); 733 error = in_pcballoc(so, &V_ripcbinfo); 734 if (error) { 735 INP_INFO_WUNLOCK(&V_ripcbinfo); 736 return (error); 737 } 738 inp = (struct inpcb *)so->so_pcb; 739 inp->inp_vflag |= INP_IPV4; 740 inp->inp_ip_p = proto; 741 inp->inp_ip_ttl = V_ip_defttl; 742 rip_inshash(inp); 743 INP_INFO_WUNLOCK(&V_ripcbinfo); 744 INP_WUNLOCK(inp); 745 return (0); 746} 747 748static void 749rip_detach(struct socket *so) 750{ 751 INIT_VNET_INET(so->so_vnet); 752 struct inpcb *inp; 753 754 inp = sotoinpcb(so); 755 KASSERT(inp != NULL, ("rip_detach: inp == NULL")); 756 KASSERT(inp->inp_faddr.s_addr == INADDR_ANY, 757 ("rip_detach: not closed")); 758 759 INP_INFO_WLOCK(&V_ripcbinfo); 760 INP_WLOCK(inp); 761 rip_delhash(inp); 762 if (so == V_ip_mrouter && ip_mrouter_done) 763 ip_mrouter_done(); 764 if (ip_rsvp_force_done) 765 ip_rsvp_force_done(so); 766 if (so == V_ip_rsvpd) 767 ip_rsvp_done(); 768 in_pcbdetach(inp); 769 in_pcbfree(inp); 770 INP_INFO_WUNLOCK(&V_ripcbinfo); 771} 772 773static void 774rip_dodisconnect(struct socket *so, struct inpcb *inp) 775{ 776 777 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 778 INP_WLOCK_ASSERT(inp); 779 780 rip_delhash(inp); 781 inp->inp_faddr.s_addr = INADDR_ANY; 782 rip_inshash(inp); 783 SOCK_LOCK(so); 784 so->so_state &= ~SS_ISCONNECTED; 785 SOCK_UNLOCK(so); 786} 787 788static void 789rip_abort(struct socket *so) 790{ 791 INIT_VNET_INET(so->so_vnet); 792 struct inpcb *inp; 793 794 inp = sotoinpcb(so); 795 KASSERT(inp != NULL, ("rip_abort: inp == NULL")); 796 797 INP_INFO_WLOCK(&V_ripcbinfo); 798 INP_WLOCK(inp); 799 rip_dodisconnect(so, inp); 800 INP_WUNLOCK(inp); 801 INP_INFO_WUNLOCK(&V_ripcbinfo); 802} 803 804static void 805rip_close(struct socket *so) 806{ 807 INIT_VNET_INET(so->so_vnet); 808 struct inpcb *inp; 809 810 inp = sotoinpcb(so); 811 KASSERT(inp != NULL, ("rip_close: inp == NULL")); 812 813 INP_INFO_WLOCK(&V_ripcbinfo); 814 INP_WLOCK(inp); 815 rip_dodisconnect(so, inp); 816 INP_WUNLOCK(inp); 817 INP_INFO_WUNLOCK(&V_ripcbinfo); 818} 819 820static int 821rip_disconnect(struct socket *so) 822{ 823 INIT_VNET_INET(so->so_vnet); 824 struct inpcb *inp; 825 826 if ((so->so_state & SS_ISCONNECTED) == 0) 827 return (ENOTCONN); 828 829 inp = sotoinpcb(so); 830 KASSERT(inp != NULL, ("rip_disconnect: inp == NULL")); 831 832 INP_INFO_WLOCK(&V_ripcbinfo); 833 INP_WLOCK(inp); 834 rip_dodisconnect(so, inp); 835 INP_WUNLOCK(inp); 836 INP_INFO_WUNLOCK(&V_ripcbinfo); 837 return (0); 838} 839 840static int 841rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 842{ 843 INIT_VNET_NET(so->so_vnet); 844 INIT_VNET_INET(so->so_vnet); 845 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 846 struct inpcb *inp; 847 int error; 848 849 if (nam->sa_len != sizeof(*addr)) 850 return (EINVAL); 851 852 error = prison_check_ip4(td->td_ucred, &addr->sin_addr); 853 if (error != 0) 854 return (error); 855 856 inp = sotoinpcb(so); 857 KASSERT(inp != NULL, ("rip_bind: inp == NULL")); 858 859 if (TAILQ_EMPTY(&V_ifnet) || 860 (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) || 861 (addr->sin_addr.s_addr && 862 (inp->inp_flags & INP_BINDANY) == 0 && 863 ifa_ifwithaddr((struct sockaddr *)addr) == NULL)) 864 return (EADDRNOTAVAIL); 865 866 INP_INFO_WLOCK(&V_ripcbinfo); 867 INP_WLOCK(inp); 868 rip_delhash(inp); 869 inp->inp_laddr = addr->sin_addr; 870 rip_inshash(inp); 871 INP_WUNLOCK(inp); 872 INP_INFO_WUNLOCK(&V_ripcbinfo); 873 return (0); 874} 875 876static int 877rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 878{ 879 INIT_VNET_NET(so->so_vnet); 880 INIT_VNET_INET(so->so_vnet); 881 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 882 struct inpcb *inp; 883 884 if (nam->sa_len != sizeof(*addr)) 885 return (EINVAL); 886 if (TAILQ_EMPTY(&V_ifnet)) 887 return (EADDRNOTAVAIL); 888 if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) 889 return (EAFNOSUPPORT); 890 891 inp = sotoinpcb(so); 892 KASSERT(inp != NULL, ("rip_connect: inp == NULL")); 893 894 INP_INFO_WLOCK(&V_ripcbinfo); 895 INP_WLOCK(inp); 896 rip_delhash(inp); 897 inp->inp_faddr = addr->sin_addr; 898 rip_inshash(inp); 899 soisconnected(so); 900 INP_WUNLOCK(inp); 901 INP_INFO_WUNLOCK(&V_ripcbinfo); 902 return (0); 903} 904 905static int 906rip_shutdown(struct socket *so) 907{ 908 struct inpcb *inp; 909 910 inp = sotoinpcb(so); 911 KASSERT(inp != NULL, ("rip_shutdown: inp == NULL")); 912 913 INP_WLOCK(inp); 914 socantsendmore(so); 915 INP_WUNLOCK(inp); 916 return (0); 917} 918 919static int 920rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 921 struct mbuf *control, struct thread *td) 922{ 923 struct inpcb *inp; 924 u_long dst; 925 926 inp = sotoinpcb(so); 927 KASSERT(inp != NULL, ("rip_send: inp == NULL")); 928 929 /* 930 * Note: 'dst' reads below are unlocked. 931 */ 932 if (so->so_state & SS_ISCONNECTED) { 933 if (nam) { 934 m_freem(m); 935 return (EISCONN); 936 } 937 dst = inp->inp_faddr.s_addr; /* Unlocked read. */ 938 } else { 939 if (nam == NULL) { 940 m_freem(m); 941 return (ENOTCONN); 942 } 943 dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr; 944 } 945 return (rip_output(m, so, dst)); 946} 947 948static int 949rip_pcblist(SYSCTL_HANDLER_ARGS) 950{ 951 INIT_VNET_INET(curvnet); 952 int error, i, n; 953 struct inpcb *inp, **inp_list; 954 inp_gen_t gencnt; 955 struct xinpgen xig; 956 957 /* 958 * The process of preparing the TCB list is too time-consuming and 959 * resource-intensive to repeat twice on every request. 960 */ 961 if (req->oldptr == 0) { 962 n = V_ripcbinfo.ipi_count; 963 req->oldidx = 2 * (sizeof xig) 964 + (n + n/8) * sizeof(struct xinpcb); 965 return (0); 966 } 967 968 if (req->newptr != 0) 969 return (EPERM); 970 971 /* 972 * OK, now we're committed to doing something. 973 */ 974 INP_INFO_RLOCK(&V_ripcbinfo); 975 gencnt = V_ripcbinfo.ipi_gencnt; 976 n = V_ripcbinfo.ipi_count; 977 INP_INFO_RUNLOCK(&V_ripcbinfo); 978 979 xig.xig_len = sizeof xig; 980 xig.xig_count = n; 981 xig.xig_gen = gencnt; 982 xig.xig_sogen = so_gencnt; 983 error = SYSCTL_OUT(req, &xig, sizeof xig); 984 if (error) 985 return (error); 986 987 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 988 if (inp_list == 0) 989 return (ENOMEM); 990 991 INP_INFO_RLOCK(&V_ripcbinfo); 992 for (inp = LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n; 993 inp = LIST_NEXT(inp, inp_list)) { 994 INP_RLOCK(inp); 995 if (inp->inp_gencnt <= gencnt && 996 cr_canseeinpcb(req->td->td_ucred, inp) == 0) { 997 /* XXX held references? */ 998 inp_list[i++] = inp; 999 } 1000 INP_RUNLOCK(inp); 1001 } 1002 INP_INFO_RUNLOCK(&V_ripcbinfo); 1003 n = i; 1004 1005 error = 0; 1006 for (i = 0; i < n; i++) { 1007 inp = inp_list[i]; 1008 INP_RLOCK(inp); 1009 if (inp->inp_gencnt <= gencnt) { 1010 struct xinpcb xi; 1011 1012 bzero(&xi, sizeof(xi)); 1013 xi.xi_len = sizeof xi; 1014 /* XXX should avoid extra copy */ 1015 bcopy(inp, &xi.xi_inp, sizeof *inp); 1016 if (inp->inp_socket) 1017 sotoxsocket(inp->inp_socket, &xi.xi_socket); 1018 INP_RUNLOCK(inp); 1019 error = SYSCTL_OUT(req, &xi, sizeof xi); 1020 } else 1021 INP_RUNLOCK(inp); 1022 } 1023 if (!error) { 1024 /* 1025 * Give the user an updated idea of our state. If the 1026 * generation differs from what we told her before, she knows 1027 * that something happened while we were processing this 1028 * request, and it might be necessary to retry. 1029 */ 1030 INP_INFO_RLOCK(&V_ripcbinfo); 1031 xig.xig_gen = V_ripcbinfo.ipi_gencnt; 1032 xig.xig_sogen = so_gencnt; 1033 xig.xig_count = V_ripcbinfo.ipi_count; 1034 INP_INFO_RUNLOCK(&V_ripcbinfo); 1035 error = SYSCTL_OUT(req, &xig, sizeof xig); 1036 } 1037 free(inp_list, M_TEMP); 1038 return (error); 1039} 1040 1041SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0, 1042 rip_pcblist, "S,xinpcb", "List of active raw IP sockets"); 1043 1044struct pr_usrreqs rip_usrreqs = { 1045 .pru_abort = rip_abort, 1046 .pru_attach = rip_attach, 1047 .pru_bind = rip_bind, 1048 .pru_connect = rip_connect, 1049 .pru_control = in_control, 1050 .pru_detach = rip_detach, 1051 .pru_disconnect = rip_disconnect, 1052 .pru_peeraddr = in_getpeeraddr, 1053 .pru_send = rip_send, 1054 .pru_shutdown = rip_shutdown, 1055 .pru_sockaddr = in_getsockaddr, 1056 .pru_sosetlabel = in_pcbsosetlabel, 1057 .pru_close = rip_close, 1058}; 1059