raw_ip.c revision 190951
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 190951 2009-04-11 23:35:20Z rwatson $"); 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 V_ripcbinfo.ipi_listhead = &V_ripcb; 191 V_ripcbinfo.ipi_hashbase = 192 hashinit(INP_PCBHASH_RAW_SIZE, M_PCB, &V_ripcbinfo.ipi_hashmask); 193 V_ripcbinfo.ipi_porthashbase = 194 hashinit(1, M_PCB, &V_ripcbinfo.ipi_porthashmask); 195 V_ripcbinfo.ipi_zone = uma_zcreate("ripcb", sizeof(struct inpcb), 196 NULL, NULL, rip_inpcb_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 197 uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets); 198 EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL, 199 EVENTHANDLER_PRI_ANY); 200} 201 202static int 203rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n, 204 struct sockaddr_in *ripsrc) 205{ 206 int policyfail = 0; 207 208 INP_RLOCK_ASSERT(last); 209 210#ifdef IPSEC 211 /* check AH/ESP integrity. */ 212 if (ipsec4_in_reject(n, last)) { 213 policyfail = 1; 214 } 215#endif /* IPSEC */ 216#ifdef MAC 217 if (!policyfail && mac_inpcb_check_deliver(last, n) != 0) 218 policyfail = 1; 219#endif 220 /* Check the minimum TTL for socket. */ 221 if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl) 222 policyfail = 1; 223 if (!policyfail) { 224 struct mbuf *opts = NULL; 225 struct socket *so; 226 227 so = last->inp_socket; 228 if ((last->inp_flags & INP_CONTROLOPTS) || 229 (so->so_options & (SO_TIMESTAMP | SO_BINTIME))) 230 ip_savecontrol(last, &opts, ip, n); 231 SOCKBUF_LOCK(&so->so_rcv); 232 if (sbappendaddr_locked(&so->so_rcv, 233 (struct sockaddr *)ripsrc, n, opts) == 0) { 234 /* should notify about lost packet */ 235 m_freem(n); 236 if (opts) 237 m_freem(opts); 238 SOCKBUF_UNLOCK(&so->so_rcv); 239 } else 240 sorwakeup_locked(so); 241 } else 242 m_freem(n); 243 return (policyfail); 244} 245 246/* 247 * Setup generic address and protocol structures for raw_input routine, then 248 * pass them along with mbuf chain. 249 */ 250void 251rip_input(struct mbuf *m, int off) 252{ 253 INIT_VNET_INET(curvnet); 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(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(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 struct sockaddr_in group; 337 int blocked; 338 339 bzero(&group, sizeof(struct sockaddr_in)); 340 group.sin_len = sizeof(struct sockaddr_in); 341 group.sin_family = AF_INET; 342 group.sin_addr = ip->ip_dst; 343 344 blocked = imo_multi_filter(inp->inp_moptions, ifp, 345 (struct sockaddr *)&group, 346 (struct sockaddr *)&ripsrc); 347 if (blocked != MCAST_PASS) { 348 IPSTAT_INC(ips_notmember); 349 continue; 350 } 351 } 352 if (last != NULL) { 353 struct mbuf *n; 354 355 n = m_copy(m, 0, (int)M_COPYALL); 356 if (n != NULL) 357 (void) rip_append(last, ip, n, &ripsrc); 358 /* XXX count dropped packet */ 359 INP_RUNLOCK(last); 360 } 361 INP_RLOCK(inp); 362 last = inp; 363 } 364 INP_INFO_RUNLOCK(&V_ripcbinfo); 365 if (last != NULL) { 366 if (rip_append(last, ip, m, &ripsrc) != 0) 367 IPSTAT_INC(ips_delivered); 368 INP_RUNLOCK(last); 369 } else { 370 m_freem(m); 371 IPSTAT_INC(ips_noproto); 372 IPSTAT_DEC(ips_delivered); 373 } 374} 375 376/* 377 * Generate IP header and pass packet to ip_output. Tack on options user may 378 * have setup with control call. 379 */ 380int 381rip_output(struct mbuf *m, struct socket *so, u_long dst) 382{ 383 INIT_VNET_INET(so->so_vnet); 384 struct ip *ip; 385 int error; 386 struct inpcb *inp = sotoinpcb(so); 387 int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) | 388 IP_ALLOWBROADCAST; 389 390 /* 391 * If the user handed us a complete IP packet, use it. Otherwise, 392 * allocate an mbuf for a header and fill it in. 393 */ 394 if ((inp->inp_flags & INP_HDRINCL) == 0) { 395 if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) { 396 m_freem(m); 397 return(EMSGSIZE); 398 } 399 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT); 400 if (m == NULL) 401 return(ENOBUFS); 402 403 INP_RLOCK(inp); 404 ip = mtod(m, struct ip *); 405 ip->ip_tos = inp->inp_ip_tos; 406 if (inp->inp_flags & INP_DONTFRAG) 407 ip->ip_off = IP_DF; 408 else 409 ip->ip_off = 0; 410 ip->ip_p = inp->inp_ip_p; 411 ip->ip_len = m->m_pkthdr.len; 412 ip->ip_src = inp->inp_laddr; 413 error = prison_get_ip4(inp->inp_cred, &ip->ip_src); 414 if (error != 0) { 415 INP_RUNLOCK(inp); 416 m_freem(m); 417 return (error); 418 } 419 ip->ip_dst.s_addr = dst; 420 ip->ip_ttl = inp->inp_ip_ttl; 421 } else { 422 if (m->m_pkthdr.len > IP_MAXPACKET) { 423 m_freem(m); 424 return(EMSGSIZE); 425 } 426 INP_RLOCK(inp); 427 ip = mtod(m, struct ip *); 428 error = prison_check_ip4(inp->inp_cred, &ip->ip_src); 429 if (error != 0) { 430 INP_RUNLOCK(inp); 431 m_freem(m); 432 return (error); 433 } 434 435 /* 436 * Don't allow both user specified and setsockopt options, 437 * and don't allow packet length sizes that will crash. 438 */ 439 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) 440 || (ip->ip_len > m->m_pkthdr.len) 441 || (ip->ip_len < (ip->ip_hl << 2))) { 442 INP_RUNLOCK(inp); 443 m_freem(m); 444 return (EINVAL); 445 } 446 if (ip->ip_id == 0) 447 ip->ip_id = ip_newid(); 448 449 /* 450 * XXX prevent ip_output from overwriting header fields. 451 */ 452 flags |= IP_RAWOUTPUT; 453 IPSTAT_INC(ips_rawout); 454 } 455 456 if (inp->inp_flags & INP_ONESBCAST) 457 flags |= IP_SENDONES; 458 459#ifdef MAC 460 mac_inpcb_create_mbuf(inp, m); 461#endif 462 463 error = ip_output(m, inp->inp_options, NULL, flags, 464 inp->inp_moptions, inp); 465 INP_RUNLOCK(inp); 466 return (error); 467} 468 469/* 470 * Raw IP socket option processing. 471 * 472 * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could 473 * only be created by a privileged process, and as such, socket option 474 * operations to manage system properties on any raw socket were allowed to 475 * take place without explicit additional access control checks. However, 476 * raw sockets can now also be created in jail(), and therefore explicit 477 * checks are now required. Likewise, raw sockets can be used by a process 478 * after it gives up privilege, so some caution is required. For options 479 * passed down to the IP layer via ip_ctloutput(), checks are assumed to be 480 * performed in ip_ctloutput() and therefore no check occurs here. 481 * Unilaterally checking priv_check() here breaks normal IP socket option 482 * operations on raw sockets. 483 * 484 * When adding new socket options here, make sure to add access control 485 * checks here as necessary. 486 */ 487int 488rip_ctloutput(struct socket *so, struct sockopt *sopt) 489{ 490 struct inpcb *inp = sotoinpcb(so); 491 int error, optval; 492 493 if (sopt->sopt_level != IPPROTO_IP) { 494 if ((sopt->sopt_level == SOL_SOCKET) && 495 (sopt->sopt_name == SO_SETFIB)) { 496 inp->inp_inc.inc_fibnum = so->so_fibnum; 497 return (0); 498 } 499 return (EINVAL); 500 } 501 502 error = 0; 503 switch (sopt->sopt_dir) { 504 case SOPT_GET: 505 switch (sopt->sopt_name) { 506 case IP_HDRINCL: 507 optval = inp->inp_flags & INP_HDRINCL; 508 error = sooptcopyout(sopt, &optval, sizeof optval); 509 break; 510 511 case IP_FW_ADD: /* ADD actually returns the body... */ 512 case IP_FW_GET: 513 case IP_FW_TABLE_GETSIZE: 514 case IP_FW_TABLE_LIST: 515 case IP_FW_NAT_GET_CONFIG: 516 case IP_FW_NAT_GET_LOG: 517 if (ip_fw_ctl_ptr != NULL) 518 error = ip_fw_ctl_ptr(sopt); 519 else 520 error = ENOPROTOOPT; 521 break; 522 523 case IP_DUMMYNET_GET: 524 if (ip_dn_ctl_ptr != NULL) 525 error = ip_dn_ctl_ptr(sopt); 526 else 527 error = ENOPROTOOPT; 528 break ; 529 530 case MRT_INIT: 531 case MRT_DONE: 532 case MRT_ADD_VIF: 533 case MRT_DEL_VIF: 534 case MRT_ADD_MFC: 535 case MRT_DEL_MFC: 536 case MRT_VERSION: 537 case MRT_ASSERT: 538 case MRT_API_SUPPORT: 539 case MRT_API_CONFIG: 540 case MRT_ADD_BW_UPCALL: 541 case MRT_DEL_BW_UPCALL: 542 error = priv_check(curthread, PRIV_NETINET_MROUTE); 543 if (error != 0) 544 return (error); 545 error = ip_mrouter_get ? ip_mrouter_get(so, sopt) : 546 EOPNOTSUPP; 547 break; 548 549 default: 550 error = ip_ctloutput(so, sopt); 551 break; 552 } 553 break; 554 555 case SOPT_SET: 556 switch (sopt->sopt_name) { 557 case IP_HDRINCL: 558 error = sooptcopyin(sopt, &optval, sizeof optval, 559 sizeof optval); 560 if (error) 561 break; 562 if (optval) 563 inp->inp_flags |= INP_HDRINCL; 564 else 565 inp->inp_flags &= ~INP_HDRINCL; 566 break; 567 568 case IP_FW_ADD: 569 case IP_FW_DEL: 570 case IP_FW_FLUSH: 571 case IP_FW_ZERO: 572 case IP_FW_RESETLOG: 573 case IP_FW_TABLE_ADD: 574 case IP_FW_TABLE_DEL: 575 case IP_FW_TABLE_FLUSH: 576 case IP_FW_NAT_CFG: 577 case IP_FW_NAT_DEL: 578 if (ip_fw_ctl_ptr != NULL) 579 error = ip_fw_ctl_ptr(sopt); 580 else 581 error = ENOPROTOOPT; 582 break; 583 584 case IP_DUMMYNET_CONFIGURE: 585 case IP_DUMMYNET_DEL: 586 case IP_DUMMYNET_FLUSH: 587 if (ip_dn_ctl_ptr != NULL) 588 error = ip_dn_ctl_ptr(sopt); 589 else 590 error = ENOPROTOOPT ; 591 break ; 592 593 case IP_RSVP_ON: 594 error = priv_check(curthread, PRIV_NETINET_MROUTE); 595 if (error != 0) 596 return (error); 597 error = ip_rsvp_init(so); 598 break; 599 600 case IP_RSVP_OFF: 601 error = priv_check(curthread, PRIV_NETINET_MROUTE); 602 if (error != 0) 603 return (error); 604 error = ip_rsvp_done(); 605 break; 606 607 case IP_RSVP_VIF_ON: 608 case IP_RSVP_VIF_OFF: 609 error = priv_check(curthread, PRIV_NETINET_MROUTE); 610 if (error != 0) 611 return (error); 612 error = ip_rsvp_vif ? 613 ip_rsvp_vif(so, sopt) : EINVAL; 614 break; 615 616 case MRT_INIT: 617 case MRT_DONE: 618 case MRT_ADD_VIF: 619 case MRT_DEL_VIF: 620 case MRT_ADD_MFC: 621 case MRT_DEL_MFC: 622 case MRT_VERSION: 623 case MRT_ASSERT: 624 case MRT_API_SUPPORT: 625 case MRT_API_CONFIG: 626 case MRT_ADD_BW_UPCALL: 627 case MRT_DEL_BW_UPCALL: 628 error = priv_check(curthread, PRIV_NETINET_MROUTE); 629 if (error != 0) 630 return (error); 631 error = ip_mrouter_set ? ip_mrouter_set(so, sopt) : 632 EOPNOTSUPP; 633 break; 634 635 default: 636 error = ip_ctloutput(so, sopt); 637 break; 638 } 639 break; 640 } 641 642 return (error); 643} 644 645/* 646 * This function exists solely to receive the PRC_IFDOWN messages which are 647 * sent by if_down(). It looks for an ifaddr whose ifa_addr is sa, and calls 648 * in_ifadown() to remove all routes corresponding to that address. It also 649 * receives the PRC_IFUP messages from if_up() and reinstalls the interface 650 * routes. 651 */ 652void 653rip_ctlinput(int cmd, struct sockaddr *sa, void *vip) 654{ 655 INIT_VNET_INET(curvnet); 656 struct in_ifaddr *ia; 657 struct ifnet *ifp; 658 int err; 659 int flags; 660 661 switch (cmd) { 662 case PRC_IFDOWN: 663 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 664 if (ia->ia_ifa.ifa_addr == sa 665 && (ia->ia_flags & IFA_ROUTE)) { 666 /* 667 * in_ifscrub kills the interface route. 668 */ 669 in_ifscrub(ia->ia_ifp, ia); 670 /* 671 * in_ifadown gets rid of all the rest of the 672 * routes. This is not quite the right thing 673 * to do, but at least if we are running a 674 * routing process they will come back. 675 */ 676 in_ifadown(&ia->ia_ifa, 0); 677 break; 678 } 679 } 680 break; 681 682 case PRC_IFUP: 683 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 684 if (ia->ia_ifa.ifa_addr == sa) 685 break; 686 } 687 if (ia == 0 || (ia->ia_flags & IFA_ROUTE)) 688 return; 689 flags = RTF_UP; 690 ifp = ia->ia_ifa.ifa_ifp; 691 692 if ((ifp->if_flags & IFF_LOOPBACK) 693 || (ifp->if_flags & IFF_POINTOPOINT)) 694 flags |= RTF_HOST; 695 696 err = rtinit(&ia->ia_ifa, RTM_ADD, flags); 697 if (err == 0) 698 ia->ia_flags |= IFA_ROUTE; 699 break; 700 } 701} 702 703u_long rip_sendspace = 9216; 704u_long rip_recvspace = 9216; 705 706SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW, 707 &rip_sendspace, 0, "Maximum outgoing raw IP datagram size"); 708SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW, 709 &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams"); 710 711static int 712rip_attach(struct socket *so, int proto, struct thread *td) 713{ 714 INIT_VNET_INET(so->so_vnet); 715 struct inpcb *inp; 716 int error; 717 718 inp = sotoinpcb(so); 719 KASSERT(inp == NULL, ("rip_attach: inp != NULL")); 720 721 error = priv_check(td, PRIV_NETINET_RAW); 722 if (error) 723 return (error); 724 if (proto >= IPPROTO_MAX || proto < 0) 725 return EPROTONOSUPPORT; 726 error = soreserve(so, rip_sendspace, rip_recvspace); 727 if (error) 728 return (error); 729 INP_INFO_WLOCK(&V_ripcbinfo); 730 error = in_pcballoc(so, &V_ripcbinfo); 731 if (error) { 732 INP_INFO_WUNLOCK(&V_ripcbinfo); 733 return (error); 734 } 735 inp = (struct inpcb *)so->so_pcb; 736 inp->inp_vflag |= INP_IPV4; 737 inp->inp_ip_p = proto; 738 inp->inp_ip_ttl = V_ip_defttl; 739 rip_inshash(inp); 740 INP_INFO_WUNLOCK(&V_ripcbinfo); 741 INP_WUNLOCK(inp); 742 return (0); 743} 744 745static void 746rip_detach(struct socket *so) 747{ 748 INIT_VNET_INET(so->so_vnet); 749 struct inpcb *inp; 750 751 inp = sotoinpcb(so); 752 KASSERT(inp != NULL, ("rip_detach: inp == NULL")); 753 KASSERT(inp->inp_faddr.s_addr == INADDR_ANY, 754 ("rip_detach: not closed")); 755 756 INP_INFO_WLOCK(&V_ripcbinfo); 757 INP_WLOCK(inp); 758 rip_delhash(inp); 759 if (so == V_ip_mrouter && ip_mrouter_done) 760 ip_mrouter_done(); 761 if (ip_rsvp_force_done) 762 ip_rsvp_force_done(so); 763 if (so == V_ip_rsvpd) 764 ip_rsvp_done(); 765 in_pcbdetach(inp); 766 in_pcbfree(inp); 767 INP_INFO_WUNLOCK(&V_ripcbinfo); 768} 769 770static void 771rip_dodisconnect(struct socket *so, struct inpcb *inp) 772{ 773 774 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 775 INP_WLOCK_ASSERT(inp); 776 777 rip_delhash(inp); 778 inp->inp_faddr.s_addr = INADDR_ANY; 779 rip_inshash(inp); 780 SOCK_LOCK(so); 781 so->so_state &= ~SS_ISCONNECTED; 782 SOCK_UNLOCK(so); 783} 784 785static void 786rip_abort(struct socket *so) 787{ 788 INIT_VNET_INET(so->so_vnet); 789 struct inpcb *inp; 790 791 inp = sotoinpcb(so); 792 KASSERT(inp != NULL, ("rip_abort: inp == NULL")); 793 794 INP_INFO_WLOCK(&V_ripcbinfo); 795 INP_WLOCK(inp); 796 rip_dodisconnect(so, inp); 797 INP_WUNLOCK(inp); 798 INP_INFO_WUNLOCK(&V_ripcbinfo); 799} 800 801static void 802rip_close(struct socket *so) 803{ 804 INIT_VNET_INET(so->so_vnet); 805 struct inpcb *inp; 806 807 inp = sotoinpcb(so); 808 KASSERT(inp != NULL, ("rip_close: inp == NULL")); 809 810 INP_INFO_WLOCK(&V_ripcbinfo); 811 INP_WLOCK(inp); 812 rip_dodisconnect(so, inp); 813 INP_WUNLOCK(inp); 814 INP_INFO_WUNLOCK(&V_ripcbinfo); 815} 816 817static int 818rip_disconnect(struct socket *so) 819{ 820 INIT_VNET_INET(so->so_vnet); 821 struct inpcb *inp; 822 823 if ((so->so_state & SS_ISCONNECTED) == 0) 824 return (ENOTCONN); 825 826 inp = sotoinpcb(so); 827 KASSERT(inp != NULL, ("rip_disconnect: inp == NULL")); 828 829 INP_INFO_WLOCK(&V_ripcbinfo); 830 INP_WLOCK(inp); 831 rip_dodisconnect(so, inp); 832 INP_WUNLOCK(inp); 833 INP_INFO_WUNLOCK(&V_ripcbinfo); 834 return (0); 835} 836 837static int 838rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 839{ 840 INIT_VNET_NET(so->so_vnet); 841 INIT_VNET_INET(so->so_vnet); 842 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 843 struct inpcb *inp; 844 int error; 845 846 if (nam->sa_len != sizeof(*addr)) 847 return (EINVAL); 848 849 error = prison_check_ip4(td->td_ucred, &addr->sin_addr); 850 if (error != 0) 851 return (error); 852 853 if (TAILQ_EMPTY(&V_ifnet) || 854 (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) || 855 (addr->sin_addr.s_addr && 856 ifa_ifwithaddr((struct sockaddr *)addr) == 0)) 857 return (EADDRNOTAVAIL); 858 859 inp = sotoinpcb(so); 860 KASSERT(inp != NULL, ("rip_bind: inp == NULL")); 861 862 INP_INFO_WLOCK(&V_ripcbinfo); 863 INP_WLOCK(inp); 864 rip_delhash(inp); 865 inp->inp_laddr = addr->sin_addr; 866 rip_inshash(inp); 867 INP_WUNLOCK(inp); 868 INP_INFO_WUNLOCK(&V_ripcbinfo); 869 return (0); 870} 871 872static int 873rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 874{ 875 INIT_VNET_NET(so->so_vnet); 876 INIT_VNET_INET(so->so_vnet); 877 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 878 struct inpcb *inp; 879 880 if (nam->sa_len != sizeof(*addr)) 881 return (EINVAL); 882 if (TAILQ_EMPTY(&V_ifnet)) 883 return (EADDRNOTAVAIL); 884 if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) 885 return (EAFNOSUPPORT); 886 887 inp = sotoinpcb(so); 888 KASSERT(inp != NULL, ("rip_connect: inp == NULL")); 889 890 INP_INFO_WLOCK(&V_ripcbinfo); 891 INP_WLOCK(inp); 892 rip_delhash(inp); 893 inp->inp_faddr = addr->sin_addr; 894 rip_inshash(inp); 895 soisconnected(so); 896 INP_WUNLOCK(inp); 897 INP_INFO_WUNLOCK(&V_ripcbinfo); 898 return (0); 899} 900 901static int 902rip_shutdown(struct socket *so) 903{ 904 struct inpcb *inp; 905 906 inp = sotoinpcb(so); 907 KASSERT(inp != NULL, ("rip_shutdown: inp == NULL")); 908 909 INP_WLOCK(inp); 910 socantsendmore(so); 911 INP_WUNLOCK(inp); 912 return (0); 913} 914 915static int 916rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 917 struct mbuf *control, struct thread *td) 918{ 919 struct inpcb *inp; 920 u_long dst; 921 922 inp = sotoinpcb(so); 923 KASSERT(inp != NULL, ("rip_send: inp == NULL")); 924 925 /* 926 * Note: 'dst' reads below are unlocked. 927 */ 928 if (so->so_state & SS_ISCONNECTED) { 929 if (nam) { 930 m_freem(m); 931 return (EISCONN); 932 } 933 dst = inp->inp_faddr.s_addr; /* Unlocked read. */ 934 } else { 935 if (nam == NULL) { 936 m_freem(m); 937 return (ENOTCONN); 938 } 939 dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr; 940 } 941 return (rip_output(m, so, dst)); 942} 943 944static int 945rip_pcblist(SYSCTL_HANDLER_ARGS) 946{ 947 INIT_VNET_INET(curvnet); 948 int error, i, n; 949 struct inpcb *inp, **inp_list; 950 inp_gen_t gencnt; 951 struct xinpgen xig; 952 953 /* 954 * The process of preparing the TCB list is too time-consuming and 955 * resource-intensive to repeat twice on every request. 956 */ 957 if (req->oldptr == 0) { 958 n = V_ripcbinfo.ipi_count; 959 req->oldidx = 2 * (sizeof xig) 960 + (n + n/8) * sizeof(struct xinpcb); 961 return (0); 962 } 963 964 if (req->newptr != 0) 965 return (EPERM); 966 967 /* 968 * OK, now we're committed to doing something. 969 */ 970 INP_INFO_RLOCK(&V_ripcbinfo); 971 gencnt = V_ripcbinfo.ipi_gencnt; 972 n = V_ripcbinfo.ipi_count; 973 INP_INFO_RUNLOCK(&V_ripcbinfo); 974 975 xig.xig_len = sizeof xig; 976 xig.xig_count = n; 977 xig.xig_gen = gencnt; 978 xig.xig_sogen = so_gencnt; 979 error = SYSCTL_OUT(req, &xig, sizeof xig); 980 if (error) 981 return (error); 982 983 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 984 if (inp_list == 0) 985 return (ENOMEM); 986 987 INP_INFO_RLOCK(&V_ripcbinfo); 988 for (inp = LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n; 989 inp = LIST_NEXT(inp, inp_list)) { 990 INP_RLOCK(inp); 991 if (inp->inp_gencnt <= gencnt && 992 cr_canseeinpcb(req->td->td_ucred, inp) == 0) { 993 /* XXX held references? */ 994 inp_list[i++] = inp; 995 } 996 INP_RUNLOCK(inp); 997 } 998 INP_INFO_RUNLOCK(&V_ripcbinfo); 999 n = i; 1000 1001 error = 0; 1002 for (i = 0; i < n; i++) { 1003 inp = inp_list[i]; 1004 INP_RLOCK(inp); 1005 if (inp->inp_gencnt <= gencnt) { 1006 struct xinpcb xi; 1007 1008 bzero(&xi, sizeof(xi)); 1009 xi.xi_len = sizeof xi; 1010 /* XXX should avoid extra copy */ 1011 bcopy(inp, &xi.xi_inp, sizeof *inp); 1012 if (inp->inp_socket) 1013 sotoxsocket(inp->inp_socket, &xi.xi_socket); 1014 INP_RUNLOCK(inp); 1015 error = SYSCTL_OUT(req, &xi, sizeof xi); 1016 } else 1017 INP_RUNLOCK(inp); 1018 } 1019 if (!error) { 1020 /* 1021 * Give the user an updated idea of our state. If the 1022 * generation differs from what we told her before, she knows 1023 * that something happened while we were processing this 1024 * request, and it might be necessary to retry. 1025 */ 1026 INP_INFO_RLOCK(&V_ripcbinfo); 1027 xig.xig_gen = V_ripcbinfo.ipi_gencnt; 1028 xig.xig_sogen = so_gencnt; 1029 xig.xig_count = V_ripcbinfo.ipi_count; 1030 INP_INFO_RUNLOCK(&V_ripcbinfo); 1031 error = SYSCTL_OUT(req, &xig, sizeof xig); 1032 } 1033 free(inp_list, M_TEMP); 1034 return (error); 1035} 1036 1037SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0, 1038 rip_pcblist, "S,xinpcb", "List of active raw IP sockets"); 1039 1040struct pr_usrreqs rip_usrreqs = { 1041 .pru_abort = rip_abort, 1042 .pru_attach = rip_attach, 1043 .pru_bind = rip_bind, 1044 .pru_connect = rip_connect, 1045 .pru_control = in_control, 1046 .pru_detach = rip_detach, 1047 .pru_disconnect = rip_disconnect, 1048 .pru_peeraddr = in_getpeeraddr, 1049 .pru_send = rip_send, 1050 .pru_shutdown = rip_shutdown, 1051 .pru_sockaddr = in_getsockaddr, 1052 .pru_sosetlabel = in_pcbsosetlabel, 1053 .pru_close = rip_close, 1054}; 1055