raw_ip.c revision 194951
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 194951 2009-06-25 11:52:33Z rwatson $"); 35 36#include "opt_inet6.h" 37#include "opt_ipsec.h" 38 39#include <sys/param.h> 40#include <sys/jail.h> 41#include <sys/kernel.h> 42#include <sys/lock.h> 43#include <sys/malloc.h> 44#include <sys/mbuf.h> 45#include <sys/priv.h> 46#include <sys/proc.h> 47#include <sys/protosw.h> 48#include <sys/rwlock.h> 49#include <sys/signalvar.h> 50#include <sys/socket.h> 51#include <sys/socketvar.h> 52#include <sys/sx.h> 53#include <sys/sysctl.h> 54#include <sys/systm.h> 55#include <sys/vimage.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/ip.h> 68#include <netinet/ip_var.h> 69#include <netinet/ip_mroute.h> 70 71#include <netinet/vinet.h> 72 73#ifdef IPSEC 74#include <netipsec/ipsec.h> 75#endif /*IPSEC*/ 76 77#include <security/mac/mac_framework.h> 78 79#ifdef VIMAGE_GLOBALS 80struct inpcbhead ripcb; 81struct inpcbinfo ripcbinfo; 82#endif 83 84/* 85 * Control and data hooks for ipfw and dummynet. 86 * The data hooks are not used here but it is convenient 87 * to keep them all in one place. 88 */ 89int (*ip_fw_ctl_ptr)(struct sockopt *) = NULL; 90int (*ip_dn_ctl_ptr)(struct sockopt *) = NULL; 91int (*ip_fw_chk_ptr)(struct ip_fw_args *args) = NULL; 92int (*ip_dn_io_ptr)(struct mbuf **m, int dir, struct ip_fw_args *fwa) = NULL; 93 94/* 95 * Hooks for multicast routing. They all default to NULL, so leave them not 96 * initialized and rely on BSS being set to 0. 97 */ 98 99/* 100 * The socket used to communicate with the multicast routing daemon. 101 */ 102#ifdef VIMAGE_GLOBALS 103struct socket *ip_mrouter; 104#endif 105 106/* 107 * The various mrouter and rsvp functions. 108 */ 109int (*ip_mrouter_set)(struct socket *, struct sockopt *); 110int (*ip_mrouter_get)(struct socket *, struct sockopt *); 111int (*ip_mrouter_done)(void); 112int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 113 struct ip_moptions *); 114int (*mrt_ioctl)(u_long, caddr_t, int); 115int (*legal_vif_num)(int); 116u_long (*ip_mcast_src)(int); 117 118void (*rsvp_input_p)(struct mbuf *m, int off); 119int (*ip_rsvp_vif)(struct socket *, struct sockopt *); 120void (*ip_rsvp_force_done)(struct socket *); 121 122/* 123 * Hash functions 124 */ 125 126#define INP_PCBHASH_RAW_SIZE 256 127#define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \ 128 (((proto) + (laddr) + (faddr)) % (mask) + 1) 129 130static void 131rip_inshash(struct inpcb *inp) 132{ 133 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 134 struct inpcbhead *pcbhash; 135 int hash; 136 137 INP_INFO_WLOCK_ASSERT(pcbinfo); 138 INP_WLOCK_ASSERT(inp); 139 140 if (inp->inp_ip_p != 0 && 141 inp->inp_laddr.s_addr != INADDR_ANY && 142 inp->inp_faddr.s_addr != INADDR_ANY) { 143 hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr, 144 inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask); 145 } else 146 hash = 0; 147 pcbhash = &pcbinfo->ipi_hashbase[hash]; 148 LIST_INSERT_HEAD(pcbhash, inp, inp_hash); 149} 150 151static void 152rip_delhash(struct inpcb *inp) 153{ 154 155 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 156 INP_WLOCK_ASSERT(inp); 157 158 LIST_REMOVE(inp, inp_hash); 159} 160 161/* 162 * Raw interface to IP protocol. 163 */ 164 165/* 166 * Initialize raw connection block q. 167 */ 168static void 169rip_zone_change(void *tag) 170{ 171 INIT_VNET_INET(curvnet); 172 173 uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets); 174} 175 176static int 177rip_inpcb_init(void *mem, int size, int flags) 178{ 179 struct inpcb *inp = mem; 180 181 INP_LOCK_INIT(inp, "inp", "rawinp"); 182 return (0); 183} 184 185void 186rip_init(void) 187{ 188 INIT_VNET_INET(curvnet); 189 190 INP_INFO_LOCK_INIT(&V_ripcbinfo, "rip"); 191 LIST_INIT(&V_ripcb); 192#ifdef VIMAGE 193 V_ripcbinfo.ipi_vnet = curvnet; 194#endif 195 V_ripcbinfo.ipi_listhead = &V_ripcb; 196 V_ripcbinfo.ipi_hashbase = 197 hashinit(INP_PCBHASH_RAW_SIZE, M_PCB, &V_ripcbinfo.ipi_hashmask); 198 V_ripcbinfo.ipi_porthashbase = 199 hashinit(1, M_PCB, &V_ripcbinfo.ipi_porthashmask); 200 V_ripcbinfo.ipi_zone = uma_zcreate("ripcb", sizeof(struct inpcb), 201 NULL, NULL, rip_inpcb_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 202 uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets); 203 EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL, 204 EVENTHANDLER_PRI_ANY); 205} 206 207#ifdef VIMAGE 208void 209rip_destroy(void) 210{ 211 INIT_VNET_INET(curvnet); 212 213 hashdestroy(V_ripcbinfo.ipi_hashbase, M_PCB, 214 V_ripcbinfo.ipi_hashmask); 215 hashdestroy(V_ripcbinfo.ipi_porthashbase, M_PCB, 216 V_ripcbinfo.ipi_porthashmask); 217} 218#endif 219 220static int 221rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n, 222 struct sockaddr_in *ripsrc) 223{ 224 int policyfail = 0; 225 226 INP_RLOCK_ASSERT(last); 227 228#ifdef IPSEC 229 /* check AH/ESP integrity. */ 230 if (ipsec4_in_reject(n, last)) { 231 policyfail = 1; 232 } 233#endif /* IPSEC */ 234#ifdef MAC 235 if (!policyfail && mac_inpcb_check_deliver(last, n) != 0) 236 policyfail = 1; 237#endif 238 /* Check the minimum TTL for socket. */ 239 if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl) 240 policyfail = 1; 241 if (!policyfail) { 242 struct mbuf *opts = NULL; 243 struct socket *so; 244 245 so = last->inp_socket; 246 if ((last->inp_flags & INP_CONTROLOPTS) || 247 (so->so_options & (SO_TIMESTAMP | SO_BINTIME))) 248 ip_savecontrol(last, &opts, ip, n); 249 SOCKBUF_LOCK(&so->so_rcv); 250 if (sbappendaddr_locked(&so->so_rcv, 251 (struct sockaddr *)ripsrc, n, opts) == 0) { 252 /* should notify about lost packet */ 253 m_freem(n); 254 if (opts) 255 m_freem(opts); 256 SOCKBUF_UNLOCK(&so->so_rcv); 257 } else 258 sorwakeup_locked(so); 259 } else 260 m_freem(n); 261 return (policyfail); 262} 263 264/* 265 * Setup generic address and protocol structures for raw_input routine, then 266 * pass them along with mbuf chain. 267 */ 268void 269rip_input(struct mbuf *m, int off) 270{ 271 INIT_VNET_INET(curvnet); 272 struct ifnet *ifp; 273 struct ip *ip = mtod(m, struct ip *); 274 int proto = ip->ip_p; 275 struct inpcb *inp, *last; 276 struct sockaddr_in ripsrc; 277 int hash; 278 279 bzero(&ripsrc, sizeof(ripsrc)); 280 ripsrc.sin_len = sizeof(ripsrc); 281 ripsrc.sin_family = AF_INET; 282 ripsrc.sin_addr = ip->ip_src; 283 last = NULL; 284 285 ifp = m->m_pkthdr.rcvif; 286 287 hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr, 288 ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask); 289 INP_INFO_RLOCK(&V_ripcbinfo); 290 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) { 291 if (inp->inp_ip_p != proto) 292 continue; 293#ifdef INET6 294 /* XXX inp locking */ 295 if ((inp->inp_vflag & INP_IPV4) == 0) 296 continue; 297#endif 298 if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr) 299 continue; 300 if (inp->inp_faddr.s_addr != ip->ip_src.s_addr) 301 continue; 302 if (jailed(inp->inp_cred)) { 303 /* 304 * XXX: If faddr was bound to multicast group, 305 * jailed raw socket will drop datagram. 306 */ 307 if (prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) 308 continue; 309 } 310 if (last != NULL) { 311 struct mbuf *n; 312 313 n = m_copy(m, 0, (int)M_COPYALL); 314 if (n != NULL) 315 (void) rip_append(last, ip, n, &ripsrc); 316 /* XXX count dropped packet */ 317 INP_RUNLOCK(last); 318 } 319 INP_RLOCK(inp); 320 last = inp; 321 } 322 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) { 323 if (inp->inp_ip_p && inp->inp_ip_p != proto) 324 continue; 325#ifdef INET6 326 /* XXX inp locking */ 327 if ((inp->inp_vflag & INP_IPV4) == 0) 328 continue; 329#endif 330 if (!in_nullhost(inp->inp_laddr) && 331 !in_hosteq(inp->inp_laddr, ip->ip_dst)) 332 continue; 333 if (!in_nullhost(inp->inp_faddr) && 334 !in_hosteq(inp->inp_faddr, ip->ip_src)) 335 continue; 336 if (jailed(inp->inp_cred)) { 337 /* 338 * Allow raw socket in jail to receive multicast; 339 * assume process had PRIV_NETINET_RAW at attach, 340 * and fall through into normal filter path if so. 341 */ 342 if (!IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 343 prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) 344 continue; 345 } 346 /* 347 * If this raw socket has multicast state, and we 348 * have received a multicast, check if this socket 349 * should receive it, as multicast filtering is now 350 * the responsibility of the transport layer. 351 */ 352 if (inp->inp_moptions != NULL && 353 IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 354 struct sockaddr_in group; 355 int blocked; 356 357 bzero(&group, sizeof(struct sockaddr_in)); 358 group.sin_len = sizeof(struct sockaddr_in); 359 group.sin_family = AF_INET; 360 group.sin_addr = ip->ip_dst; 361 362 blocked = imo_multi_filter(inp->inp_moptions, ifp, 363 (struct sockaddr *)&group, 364 (struct sockaddr *)&ripsrc); 365 if (blocked != MCAST_PASS) { 366 IPSTAT_INC(ips_notmember); 367 continue; 368 } 369 } 370 if (last != NULL) { 371 struct mbuf *n; 372 373 n = m_copy(m, 0, (int)M_COPYALL); 374 if (n != NULL) 375 (void) rip_append(last, ip, n, &ripsrc); 376 /* XXX count dropped packet */ 377 INP_RUNLOCK(last); 378 } 379 INP_RLOCK(inp); 380 last = inp; 381 } 382 INP_INFO_RUNLOCK(&V_ripcbinfo); 383 if (last != NULL) { 384 if (rip_append(last, ip, m, &ripsrc) != 0) 385 IPSTAT_INC(ips_delivered); 386 INP_RUNLOCK(last); 387 } else { 388 m_freem(m); 389 IPSTAT_INC(ips_noproto); 390 IPSTAT_DEC(ips_delivered); 391 } 392} 393 394/* 395 * Generate IP header and pass packet to ip_output. Tack on options user may 396 * have setup with control call. 397 */ 398int 399rip_output(struct mbuf *m, struct socket *so, u_long dst) 400{ 401 INIT_VNET_INET(so->so_vnet); 402 struct ip *ip; 403 int error; 404 struct inpcb *inp = sotoinpcb(so); 405 int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) | 406 IP_ALLOWBROADCAST; 407 408 /* 409 * If the user handed us a complete IP packet, use it. Otherwise, 410 * allocate an mbuf for a header and fill it in. 411 */ 412 if ((inp->inp_flags & INP_HDRINCL) == 0) { 413 if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) { 414 m_freem(m); 415 return(EMSGSIZE); 416 } 417 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT); 418 if (m == NULL) 419 return(ENOBUFS); 420 421 INP_RLOCK(inp); 422 ip = mtod(m, struct ip *); 423 ip->ip_tos = inp->inp_ip_tos; 424 if (inp->inp_flags & INP_DONTFRAG) 425 ip->ip_off = IP_DF; 426 else 427 ip->ip_off = 0; 428 ip->ip_p = inp->inp_ip_p; 429 ip->ip_len = m->m_pkthdr.len; 430 ip->ip_src = inp->inp_laddr; 431 error = prison_get_ip4(inp->inp_cred, &ip->ip_src); 432 if (error != 0) { 433 INP_RUNLOCK(inp); 434 m_freem(m); 435 return (error); 436 } 437 ip->ip_dst.s_addr = dst; 438 ip->ip_ttl = inp->inp_ip_ttl; 439 } else { 440 if (m->m_pkthdr.len > IP_MAXPACKET) { 441 m_freem(m); 442 return(EMSGSIZE); 443 } 444 INP_RLOCK(inp); 445 ip = mtod(m, struct ip *); 446 error = prison_check_ip4(inp->inp_cred, &ip->ip_src); 447 if (error != 0) { 448 INP_RUNLOCK(inp); 449 m_freem(m); 450 return (error); 451 } 452 453 /* 454 * Don't allow both user specified and setsockopt options, 455 * and don't allow packet length sizes that will crash. 456 */ 457 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) 458 || (ip->ip_len > m->m_pkthdr.len) 459 || (ip->ip_len < (ip->ip_hl << 2))) { 460 INP_RUNLOCK(inp); 461 m_freem(m); 462 return (EINVAL); 463 } 464 if (ip->ip_id == 0) 465 ip->ip_id = ip_newid(); 466 467 /* 468 * XXX prevent ip_output from overwriting header fields. 469 */ 470 flags |= IP_RAWOUTPUT; 471 IPSTAT_INC(ips_rawout); 472 } 473 474 if (inp->inp_flags & INP_ONESBCAST) 475 flags |= IP_SENDONES; 476 477#ifdef MAC 478 mac_inpcb_create_mbuf(inp, m); 479#endif 480 481 error = ip_output(m, inp->inp_options, NULL, flags, 482 inp->inp_moptions, inp); 483 INP_RUNLOCK(inp); 484 return (error); 485} 486 487/* 488 * Raw IP socket option processing. 489 * 490 * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could 491 * only be created by a privileged process, and as such, socket option 492 * operations to manage system properties on any raw socket were allowed to 493 * take place without explicit additional access control checks. However, 494 * raw sockets can now also be created in jail(), and therefore explicit 495 * checks are now required. Likewise, raw sockets can be used by a process 496 * after it gives up privilege, so some caution is required. For options 497 * passed down to the IP layer via ip_ctloutput(), checks are assumed to be 498 * performed in ip_ctloutput() and therefore no check occurs here. 499 * Unilaterally checking priv_check() here breaks normal IP socket option 500 * operations on raw sockets. 501 * 502 * When adding new socket options here, make sure to add access control 503 * checks here as necessary. 504 */ 505int 506rip_ctloutput(struct socket *so, struct sockopt *sopt) 507{ 508 struct inpcb *inp = sotoinpcb(so); 509 int error, optval; 510 511 if (sopt->sopt_level != IPPROTO_IP) { 512 if ((sopt->sopt_level == SOL_SOCKET) && 513 (sopt->sopt_name == SO_SETFIB)) { 514 inp->inp_inc.inc_fibnum = so->so_fibnum; 515 return (0); 516 } 517 return (EINVAL); 518 } 519 520 error = 0; 521 switch (sopt->sopt_dir) { 522 case SOPT_GET: 523 switch (sopt->sopt_name) { 524 case IP_HDRINCL: 525 optval = inp->inp_flags & INP_HDRINCL; 526 error = sooptcopyout(sopt, &optval, sizeof optval); 527 break; 528 529 case IP_FW_ADD: /* ADD actually returns the body... */ 530 case IP_FW_GET: 531 case IP_FW_TABLE_GETSIZE: 532 case IP_FW_TABLE_LIST: 533 case IP_FW_NAT_GET_CONFIG: 534 case IP_FW_NAT_GET_LOG: 535 if (ip_fw_ctl_ptr != NULL) 536 error = ip_fw_ctl_ptr(sopt); 537 else 538 error = ENOPROTOOPT; 539 break; 540 541 case IP_DUMMYNET_GET: 542 if (ip_dn_ctl_ptr != NULL) 543 error = ip_dn_ctl_ptr(sopt); 544 else 545 error = ENOPROTOOPT; 546 break ; 547 548 case MRT_INIT: 549 case MRT_DONE: 550 case MRT_ADD_VIF: 551 case MRT_DEL_VIF: 552 case MRT_ADD_MFC: 553 case MRT_DEL_MFC: 554 case MRT_VERSION: 555 case MRT_ASSERT: 556 case MRT_API_SUPPORT: 557 case MRT_API_CONFIG: 558 case MRT_ADD_BW_UPCALL: 559 case MRT_DEL_BW_UPCALL: 560 error = priv_check(curthread, PRIV_NETINET_MROUTE); 561 if (error != 0) 562 return (error); 563 error = ip_mrouter_get ? ip_mrouter_get(so, sopt) : 564 EOPNOTSUPP; 565 break; 566 567 default: 568 error = ip_ctloutput(so, sopt); 569 break; 570 } 571 break; 572 573 case SOPT_SET: 574 switch (sopt->sopt_name) { 575 case IP_HDRINCL: 576 error = sooptcopyin(sopt, &optval, sizeof optval, 577 sizeof optval); 578 if (error) 579 break; 580 if (optval) 581 inp->inp_flags |= INP_HDRINCL; 582 else 583 inp->inp_flags &= ~INP_HDRINCL; 584 break; 585 586 case IP_FW_ADD: 587 case IP_FW_DEL: 588 case IP_FW_FLUSH: 589 case IP_FW_ZERO: 590 case IP_FW_RESETLOG: 591 case IP_FW_TABLE_ADD: 592 case IP_FW_TABLE_DEL: 593 case IP_FW_TABLE_FLUSH: 594 case IP_FW_NAT_CFG: 595 case IP_FW_NAT_DEL: 596 if (ip_fw_ctl_ptr != NULL) 597 error = ip_fw_ctl_ptr(sopt); 598 else 599 error = ENOPROTOOPT; 600 break; 601 602 case IP_DUMMYNET_CONFIGURE: 603 case IP_DUMMYNET_DEL: 604 case IP_DUMMYNET_FLUSH: 605 if (ip_dn_ctl_ptr != NULL) 606 error = ip_dn_ctl_ptr(sopt); 607 else 608 error = ENOPROTOOPT ; 609 break ; 610 611 case IP_RSVP_ON: 612 error = priv_check(curthread, PRIV_NETINET_MROUTE); 613 if (error != 0) 614 return (error); 615 error = ip_rsvp_init(so); 616 break; 617 618 case IP_RSVP_OFF: 619 error = priv_check(curthread, PRIV_NETINET_MROUTE); 620 if (error != 0) 621 return (error); 622 error = ip_rsvp_done(); 623 break; 624 625 case IP_RSVP_VIF_ON: 626 case IP_RSVP_VIF_OFF: 627 error = priv_check(curthread, PRIV_NETINET_MROUTE); 628 if (error != 0) 629 return (error); 630 error = ip_rsvp_vif ? 631 ip_rsvp_vif(so, sopt) : EINVAL; 632 break; 633 634 case MRT_INIT: 635 case MRT_DONE: 636 case MRT_ADD_VIF: 637 case MRT_DEL_VIF: 638 case MRT_ADD_MFC: 639 case MRT_DEL_MFC: 640 case MRT_VERSION: 641 case MRT_ASSERT: 642 case MRT_API_SUPPORT: 643 case MRT_API_CONFIG: 644 case MRT_ADD_BW_UPCALL: 645 case MRT_DEL_BW_UPCALL: 646 error = priv_check(curthread, PRIV_NETINET_MROUTE); 647 if (error != 0) 648 return (error); 649 error = ip_mrouter_set ? ip_mrouter_set(so, sopt) : 650 EOPNOTSUPP; 651 break; 652 653 default: 654 error = ip_ctloutput(so, sopt); 655 break; 656 } 657 break; 658 } 659 660 return (error); 661} 662 663/* 664 * This function exists solely to receive the PRC_IFDOWN messages which are 665 * sent by if_down(). It looks for an ifaddr whose ifa_addr is sa, and calls 666 * in_ifadown() to remove all routes corresponding to that address. It also 667 * receives the PRC_IFUP messages from if_up() and reinstalls the interface 668 * routes. 669 */ 670void 671rip_ctlinput(int cmd, struct sockaddr *sa, void *vip) 672{ 673 INIT_VNET_INET(curvnet); 674 struct in_ifaddr *ia; 675 struct ifnet *ifp; 676 int err; 677 int flags; 678 679 switch (cmd) { 680 case PRC_IFDOWN: 681 IN_IFADDR_RLOCK(); 682 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 683 if (ia->ia_ifa.ifa_addr == sa 684 && (ia->ia_flags & IFA_ROUTE)) { 685 ifa_ref(&ia->ia_ifa); 686 IN_IFADDR_RUNLOCK(); 687 /* 688 * in_ifscrub kills the interface route. 689 */ 690 in_ifscrub(ia->ia_ifp, ia); 691 /* 692 * in_ifadown gets rid of all the rest of the 693 * routes. This is not quite the right thing 694 * to do, but at least if we are running a 695 * routing process they will come back. 696 */ 697 in_ifadown(&ia->ia_ifa, 0); 698 ifa_free(&ia->ia_ifa); 699 break; 700 } 701 } 702 if (ia == NULL) /* If ia matched, already unlocked. */ 703 IN_IFADDR_RUNLOCK(); 704 break; 705 706 case PRC_IFUP: 707 IN_IFADDR_RLOCK(); 708 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 709 if (ia->ia_ifa.ifa_addr == sa) 710 break; 711 } 712 if (ia == NULL || (ia->ia_flags & IFA_ROUTE)) { 713 IN_IFADDR_RUNLOCK(); 714 return; 715 } 716 ifa_ref(&ia->ia_ifa); 717 IN_IFADDR_RUNLOCK(); 718 flags = RTF_UP; 719 ifp = ia->ia_ifa.ifa_ifp; 720 721 if ((ifp->if_flags & IFF_LOOPBACK) 722 || (ifp->if_flags & IFF_POINTOPOINT)) 723 flags |= RTF_HOST; 724 725 err = rtinit(&ia->ia_ifa, RTM_ADD, flags); 726 if (err == 0) 727 ia->ia_flags |= IFA_ROUTE; 728 ifa_free(&ia->ia_ifa); 729 break; 730 } 731} 732 733u_long rip_sendspace = 9216; 734u_long rip_recvspace = 9216; 735 736SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW, 737 &rip_sendspace, 0, "Maximum outgoing raw IP datagram size"); 738SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW, 739 &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams"); 740 741static int 742rip_attach(struct socket *so, int proto, struct thread *td) 743{ 744 INIT_VNET_INET(so->so_vnet); 745 struct inpcb *inp; 746 int error; 747 748 inp = sotoinpcb(so); 749 KASSERT(inp == NULL, ("rip_attach: inp != NULL")); 750 751 error = priv_check(td, PRIV_NETINET_RAW); 752 if (error) 753 return (error); 754 if (proto >= IPPROTO_MAX || proto < 0) 755 return EPROTONOSUPPORT; 756 error = soreserve(so, rip_sendspace, rip_recvspace); 757 if (error) 758 return (error); 759 INP_INFO_WLOCK(&V_ripcbinfo); 760 error = in_pcballoc(so, &V_ripcbinfo); 761 if (error) { 762 INP_INFO_WUNLOCK(&V_ripcbinfo); 763 return (error); 764 } 765 inp = (struct inpcb *)so->so_pcb; 766 inp->inp_vflag |= INP_IPV4; 767 inp->inp_ip_p = proto; 768 inp->inp_ip_ttl = V_ip_defttl; 769 rip_inshash(inp); 770 INP_INFO_WUNLOCK(&V_ripcbinfo); 771 INP_WUNLOCK(inp); 772 return (0); 773} 774 775static void 776rip_detach(struct socket *so) 777{ 778 INIT_VNET_INET(so->so_vnet); 779 struct inpcb *inp; 780 781 inp = sotoinpcb(so); 782 KASSERT(inp != NULL, ("rip_detach: inp == NULL")); 783 KASSERT(inp->inp_faddr.s_addr == INADDR_ANY, 784 ("rip_detach: not closed")); 785 786 INP_INFO_WLOCK(&V_ripcbinfo); 787 INP_WLOCK(inp); 788 rip_delhash(inp); 789 if (so == V_ip_mrouter && ip_mrouter_done) 790 ip_mrouter_done(); 791 if (ip_rsvp_force_done) 792 ip_rsvp_force_done(so); 793 if (so == V_ip_rsvpd) 794 ip_rsvp_done(); 795 in_pcbdetach(inp); 796 in_pcbfree(inp); 797 INP_INFO_WUNLOCK(&V_ripcbinfo); 798} 799 800static void 801rip_dodisconnect(struct socket *so, struct inpcb *inp) 802{ 803 804 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 805 INP_WLOCK_ASSERT(inp); 806 807 rip_delhash(inp); 808 inp->inp_faddr.s_addr = INADDR_ANY; 809 rip_inshash(inp); 810 SOCK_LOCK(so); 811 so->so_state &= ~SS_ISCONNECTED; 812 SOCK_UNLOCK(so); 813} 814 815static void 816rip_abort(struct socket *so) 817{ 818 INIT_VNET_INET(so->so_vnet); 819 struct inpcb *inp; 820 821 inp = sotoinpcb(so); 822 KASSERT(inp != NULL, ("rip_abort: inp == NULL")); 823 824 INP_INFO_WLOCK(&V_ripcbinfo); 825 INP_WLOCK(inp); 826 rip_dodisconnect(so, inp); 827 INP_WUNLOCK(inp); 828 INP_INFO_WUNLOCK(&V_ripcbinfo); 829} 830 831static void 832rip_close(struct socket *so) 833{ 834 INIT_VNET_INET(so->so_vnet); 835 struct inpcb *inp; 836 837 inp = sotoinpcb(so); 838 KASSERT(inp != NULL, ("rip_close: inp == NULL")); 839 840 INP_INFO_WLOCK(&V_ripcbinfo); 841 INP_WLOCK(inp); 842 rip_dodisconnect(so, inp); 843 INP_WUNLOCK(inp); 844 INP_INFO_WUNLOCK(&V_ripcbinfo); 845} 846 847static int 848rip_disconnect(struct socket *so) 849{ 850 INIT_VNET_INET(so->so_vnet); 851 struct inpcb *inp; 852 853 if ((so->so_state & SS_ISCONNECTED) == 0) 854 return (ENOTCONN); 855 856 inp = sotoinpcb(so); 857 KASSERT(inp != NULL, ("rip_disconnect: inp == NULL")); 858 859 INP_INFO_WLOCK(&V_ripcbinfo); 860 INP_WLOCK(inp); 861 rip_dodisconnect(so, inp); 862 INP_WUNLOCK(inp); 863 INP_INFO_WUNLOCK(&V_ripcbinfo); 864 return (0); 865} 866 867static int 868rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 869{ 870 INIT_VNET_NET(so->so_vnet); 871 INIT_VNET_INET(so->so_vnet); 872 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 873 struct inpcb *inp; 874 int error; 875 876 if (nam->sa_len != sizeof(*addr)) 877 return (EINVAL); 878 879 error = prison_check_ip4(td->td_ucred, &addr->sin_addr); 880 if (error != 0) 881 return (error); 882 883 inp = sotoinpcb(so); 884 KASSERT(inp != NULL, ("rip_bind: inp == NULL")); 885 886 if (TAILQ_EMPTY(&V_ifnet) || 887 (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) || 888 (addr->sin_addr.s_addr && 889 (inp->inp_flags & INP_BINDANY) == 0 && 890 ifa_ifwithaddr_check((struct sockaddr *)addr) == 0)) 891 return (EADDRNOTAVAIL); 892 893 INP_INFO_WLOCK(&V_ripcbinfo); 894 INP_WLOCK(inp); 895 rip_delhash(inp); 896 inp->inp_laddr = addr->sin_addr; 897 rip_inshash(inp); 898 INP_WUNLOCK(inp); 899 INP_INFO_WUNLOCK(&V_ripcbinfo); 900 return (0); 901} 902 903static int 904rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 905{ 906 INIT_VNET_NET(so->so_vnet); 907 INIT_VNET_INET(so->so_vnet); 908 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 909 struct inpcb *inp; 910 911 if (nam->sa_len != sizeof(*addr)) 912 return (EINVAL); 913 if (TAILQ_EMPTY(&V_ifnet)) 914 return (EADDRNOTAVAIL); 915 if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) 916 return (EAFNOSUPPORT); 917 918 inp = sotoinpcb(so); 919 KASSERT(inp != NULL, ("rip_connect: inp == NULL")); 920 921 INP_INFO_WLOCK(&V_ripcbinfo); 922 INP_WLOCK(inp); 923 rip_delhash(inp); 924 inp->inp_faddr = addr->sin_addr; 925 rip_inshash(inp); 926 soisconnected(so); 927 INP_WUNLOCK(inp); 928 INP_INFO_WUNLOCK(&V_ripcbinfo); 929 return (0); 930} 931 932static int 933rip_shutdown(struct socket *so) 934{ 935 struct inpcb *inp; 936 937 inp = sotoinpcb(so); 938 KASSERT(inp != NULL, ("rip_shutdown: inp == NULL")); 939 940 INP_WLOCK(inp); 941 socantsendmore(so); 942 INP_WUNLOCK(inp); 943 return (0); 944} 945 946static int 947rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 948 struct mbuf *control, struct thread *td) 949{ 950 struct inpcb *inp; 951 u_long dst; 952 953 inp = sotoinpcb(so); 954 KASSERT(inp != NULL, ("rip_send: inp == NULL")); 955 956 /* 957 * Note: 'dst' reads below are unlocked. 958 */ 959 if (so->so_state & SS_ISCONNECTED) { 960 if (nam) { 961 m_freem(m); 962 return (EISCONN); 963 } 964 dst = inp->inp_faddr.s_addr; /* Unlocked read. */ 965 } else { 966 if (nam == NULL) { 967 m_freem(m); 968 return (ENOTCONN); 969 } 970 dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr; 971 } 972 return (rip_output(m, so, dst)); 973} 974 975static int 976rip_pcblist(SYSCTL_HANDLER_ARGS) 977{ 978 INIT_VNET_INET(curvnet); 979 int error, i, n; 980 struct inpcb *inp, **inp_list; 981 inp_gen_t gencnt; 982 struct xinpgen xig; 983 984 /* 985 * The process of preparing the TCB list is too time-consuming and 986 * resource-intensive to repeat twice on every request. 987 */ 988 if (req->oldptr == 0) { 989 n = V_ripcbinfo.ipi_count; 990 req->oldidx = 2 * (sizeof xig) 991 + (n + n/8) * sizeof(struct xinpcb); 992 return (0); 993 } 994 995 if (req->newptr != 0) 996 return (EPERM); 997 998 /* 999 * OK, now we're committed to doing something. 1000 */ 1001 INP_INFO_RLOCK(&V_ripcbinfo); 1002 gencnt = V_ripcbinfo.ipi_gencnt; 1003 n = V_ripcbinfo.ipi_count; 1004 INP_INFO_RUNLOCK(&V_ripcbinfo); 1005 1006 xig.xig_len = sizeof xig; 1007 xig.xig_count = n; 1008 xig.xig_gen = gencnt; 1009 xig.xig_sogen = so_gencnt; 1010 error = SYSCTL_OUT(req, &xig, sizeof xig); 1011 if (error) 1012 return (error); 1013 1014 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 1015 if (inp_list == 0) 1016 return (ENOMEM); 1017 1018 INP_INFO_RLOCK(&V_ripcbinfo); 1019 for (inp = LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n; 1020 inp = LIST_NEXT(inp, inp_list)) { 1021 INP_RLOCK(inp); 1022 if (inp->inp_gencnt <= gencnt && 1023 cr_canseeinpcb(req->td->td_ucred, inp) == 0) { 1024 /* XXX held references? */ 1025 inp_list[i++] = inp; 1026 } 1027 INP_RUNLOCK(inp); 1028 } 1029 INP_INFO_RUNLOCK(&V_ripcbinfo); 1030 n = i; 1031 1032 error = 0; 1033 for (i = 0; i < n; i++) { 1034 inp = inp_list[i]; 1035 INP_RLOCK(inp); 1036 if (inp->inp_gencnt <= gencnt) { 1037 struct xinpcb xi; 1038 1039 bzero(&xi, sizeof(xi)); 1040 xi.xi_len = sizeof xi; 1041 /* XXX should avoid extra copy */ 1042 bcopy(inp, &xi.xi_inp, sizeof *inp); 1043 if (inp->inp_socket) 1044 sotoxsocket(inp->inp_socket, &xi.xi_socket); 1045 INP_RUNLOCK(inp); 1046 error = SYSCTL_OUT(req, &xi, sizeof xi); 1047 } else 1048 INP_RUNLOCK(inp); 1049 } 1050 if (!error) { 1051 /* 1052 * Give the user an updated idea of our state. If the 1053 * generation differs from what we told her before, she knows 1054 * that something happened while we were processing this 1055 * request, and it might be necessary to retry. 1056 */ 1057 INP_INFO_RLOCK(&V_ripcbinfo); 1058 xig.xig_gen = V_ripcbinfo.ipi_gencnt; 1059 xig.xig_sogen = so_gencnt; 1060 xig.xig_count = V_ripcbinfo.ipi_count; 1061 INP_INFO_RUNLOCK(&V_ripcbinfo); 1062 error = SYSCTL_OUT(req, &xig, sizeof xig); 1063 } 1064 free(inp_list, M_TEMP); 1065 return (error); 1066} 1067 1068SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0, 1069 rip_pcblist, "S,xinpcb", "List of active raw IP sockets"); 1070 1071struct pr_usrreqs rip_usrreqs = { 1072 .pru_abort = rip_abort, 1073 .pru_attach = rip_attach, 1074 .pru_bind = rip_bind, 1075 .pru_connect = rip_connect, 1076 .pru_control = in_control, 1077 .pru_detach = rip_detach, 1078 .pru_disconnect = rip_disconnect, 1079 .pru_peeraddr = in_getpeeraddr, 1080 .pru_send = rip_send, 1081 .pru_shutdown = rip_shutdown, 1082 .pru_sockaddr = in_getsockaddr, 1083 .pru_sosetlabel = in_pcbsosetlabel, 1084 .pru_close = rip_close, 1085}; 1086