raw_ip.c revision 199287
1124771Sgrehan/*- 2124771Sgrehan * Copyright (c) 1982, 1986, 1988, 1993 3124771Sgrehan * The Regents of the University of California. 4124771Sgrehan * All rights reserved. 5124771Sgrehan * 6124771Sgrehan * Redistribution and use in source and binary forms, with or without 7124771Sgrehan * modification, are permitted provided that the following conditions 8124771Sgrehan * are met: 9124771Sgrehan * 1. Redistributions of source code must retain the above copyright 10124771Sgrehan * notice, this list of conditions and the following disclaimer. 11124771Sgrehan * 2. Redistributions in binary form must reproduce the above copyright 12124771Sgrehan * notice, this list of conditions and the following disclaimer in the 13124771Sgrehan * documentation and/or other materials provided with the distribution. 14124771Sgrehan * 4. Neither the name of the University nor the names of its contributors 15124771Sgrehan * may be used to endorse or promote products derived from this software 16124771Sgrehan * without specific prior written permission. 17124771Sgrehan * 18124771Sgrehan * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 19124771Sgrehan * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 20124771Sgrehan * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 21124771Sgrehan * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 22124771Sgrehan * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 23124771Sgrehan * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 24124771Sgrehan * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 25124771Sgrehan * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 26124771Sgrehan * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 27124771Sgrehan * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 28124771Sgrehan * SUCH DAMAGE. 29124771Sgrehan * 30124771Sgrehan * @(#)raw_ip.c 8.7 (Berkeley) 5/15/95 31124771Sgrehan */ 32124771Sgrehan 33124771Sgrehan#include <sys/cdefs.h> 34124771Sgrehan__FBSDID("$FreeBSD: head/sys/netinet/raw_ip.c 199287 2009-11-15 11:07:22Z bms $"); 35124771Sgrehan 36124771Sgrehan#include "opt_inet6.h" 37124771Sgrehan#include "opt_ipsec.h" 38124771Sgrehan 39124771Sgrehan#include <sys/param.h> 40124771Sgrehan#include <sys/jail.h> 41124771Sgrehan#include <sys/kernel.h> 42124771Sgrehan#include <sys/lock.h> 43124771Sgrehan#include <sys/malloc.h> 44124771Sgrehan#include <sys/mbuf.h> 45124771Sgrehan#include <sys/priv.h> 46124771Sgrehan#include <sys/proc.h> 47124771Sgrehan#include <sys/protosw.h> 48124771Sgrehan#include <sys/rwlock.h> 49124771Sgrehan#include <sys/signalvar.h> 50124771Sgrehan#include <sys/socket.h> 51124771Sgrehan#include <sys/socketvar.h> 52124771Sgrehan#include <sys/sx.h> 53124771Sgrehan#include <sys/sysctl.h> 54124771Sgrehan#include <sys/systm.h> 55124771Sgrehan 56124771Sgrehan#include <vm/uma.h> 57124771Sgrehan 58124771Sgrehan#include <net/if.h> 59124771Sgrehan#include <net/route.h> 60124771Sgrehan#include <net/vnet.h> 61124771Sgrehan 62124771Sgrehan#include <netinet/in.h> 63124771Sgrehan#include <netinet/in_systm.h> 64124771Sgrehan#include <netinet/in_pcb.h> 65124771Sgrehan#include <netinet/in_var.h> 66124771Sgrehan#include <netinet/ip.h> 67124771Sgrehan#include <netinet/ip_var.h> 68124771Sgrehan#include <netinet/ip_mroute.h> 69124771Sgrehan 70124771Sgrehan#ifdef IPSEC 71124771Sgrehan#include <netipsec/ipsec.h> 72124771Sgrehan#endif /*IPSEC*/ 73124771Sgrehan 74124771Sgrehan#include <security/mac/mac_framework.h> 75124771Sgrehan 76124771SgrehanVNET_DEFINE(struct inpcbhead, ripcb); 77124771SgrehanVNET_DEFINE(struct inpcbinfo, ripcbinfo); 78124771Sgrehan 79124771Sgrehan#define V_ripcb VNET(ripcb) 80124771Sgrehan#define V_ripcbinfo VNET(ripcbinfo) 81124771Sgrehan 82124771Sgrehan/* 83124771Sgrehan * Control and data hooks for ipfw and dummynet. 84124771Sgrehan * The data hooks are not used here but it is convenient 85124771Sgrehan * to keep them all in one place. 86124771Sgrehan */ 87124771SgrehanVNET_DEFINE(ip_fw_chk_ptr_t, ip_fw_chk_ptr) = NULL; 88124771SgrehanVNET_DEFINE(ip_fw_ctl_ptr_t, ip_fw_ctl_ptr) = NULL; 89124771Sgrehanint (*ip_dn_ctl_ptr)(struct sockopt *) = NULL; 90124771Sgrehanint (*ip_dn_io_ptr)(struct mbuf **m, int dir, struct ip_fw_args *fwa) = 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 */ 100VNET_DEFINE(struct socket *, ip_mrouter); 101 102/* 103 * The various mrouter and rsvp functions. 104 */ 105int (*ip_mrouter_set)(struct socket *, struct sockopt *); 106int (*ip_mrouter_get)(struct socket *, struct sockopt *); 107int (*ip_mrouter_done)(void); 108int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 109 struct ip_moptions *); 110int (*mrt_ioctl)(u_long, caddr_t, int); 111int (*legal_vif_num)(int); 112u_long (*ip_mcast_src)(int); 113 114void (*rsvp_input_p)(struct mbuf *m, int off); 115int (*ip_rsvp_vif)(struct socket *, struct sockopt *); 116void (*ip_rsvp_force_done)(struct socket *); 117 118/* 119 * Hash functions 120 */ 121 122#define INP_PCBHASH_RAW_SIZE 256 123#define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \ 124 (((proto) + (laddr) + (faddr)) % (mask) + 1) 125 126static void 127rip_inshash(struct inpcb *inp) 128{ 129 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 130 struct inpcbhead *pcbhash; 131 int hash; 132 133 INP_INFO_WLOCK_ASSERT(pcbinfo); 134 INP_WLOCK_ASSERT(inp); 135 136 if (inp->inp_ip_p != 0 && 137 inp->inp_laddr.s_addr != INADDR_ANY && 138 inp->inp_faddr.s_addr != INADDR_ANY) { 139 hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr, 140 inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask); 141 } else 142 hash = 0; 143 pcbhash = &pcbinfo->ipi_hashbase[hash]; 144 LIST_INSERT_HEAD(pcbhash, inp, inp_hash); 145} 146 147static void 148rip_delhash(struct inpcb *inp) 149{ 150 151 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 152 INP_WLOCK_ASSERT(inp); 153 154 LIST_REMOVE(inp, inp_hash); 155} 156 157/* 158 * Raw interface to IP protocol. 159 */ 160 161/* 162 * Initialize raw connection block q. 163 */ 164static void 165rip_zone_change(void *tag) 166{ 167 168 uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets); 169} 170 171static int 172rip_inpcb_init(void *mem, int size, int flags) 173{ 174 struct inpcb *inp = mem; 175 176 INP_LOCK_INIT(inp, "inp", "rawinp"); 177 return (0); 178} 179 180void 181rip_init(void) 182{ 183 184 INP_INFO_LOCK_INIT(&V_ripcbinfo, "rip"); 185 LIST_INIT(&V_ripcb); 186#ifdef VIMAGE 187 V_ripcbinfo.ipi_vnet = curvnet; 188#endif 189 V_ripcbinfo.ipi_listhead = &V_ripcb; 190 V_ripcbinfo.ipi_hashbase = 191 hashinit(INP_PCBHASH_RAW_SIZE, M_PCB, &V_ripcbinfo.ipi_hashmask); 192 V_ripcbinfo.ipi_porthashbase = 193 hashinit(1, M_PCB, &V_ripcbinfo.ipi_porthashmask); 194 V_ripcbinfo.ipi_zone = uma_zcreate("ripcb", sizeof(struct inpcb), 195 NULL, NULL, rip_inpcb_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 196 uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets); 197 EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL, 198 EVENTHANDLER_PRI_ANY); 199} 200 201#ifdef VIMAGE 202void 203rip_destroy(void) 204{ 205 206 hashdestroy(V_ripcbinfo.ipi_hashbase, M_PCB, 207 V_ripcbinfo.ipi_hashmask); 208 hashdestroy(V_ripcbinfo.ipi_porthashbase, M_PCB, 209 V_ripcbinfo.ipi_porthashmask); 210} 211#endif 212 213static int 214rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n, 215 struct sockaddr_in *ripsrc) 216{ 217 int policyfail = 0; 218 219 INP_RLOCK_ASSERT(last); 220 221#ifdef IPSEC 222 /* check AH/ESP integrity. */ 223 if (ipsec4_in_reject(n, last)) { 224 policyfail = 1; 225 } 226#endif /* IPSEC */ 227#ifdef MAC 228 if (!policyfail && mac_inpcb_check_deliver(last, n) != 0) 229 policyfail = 1; 230#endif 231 /* Check the minimum TTL for socket. */ 232 if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl) 233 policyfail = 1; 234 if (!policyfail) { 235 struct mbuf *opts = NULL; 236 struct socket *so; 237 238 so = last->inp_socket; 239 if ((last->inp_flags & INP_CONTROLOPTS) || 240 (so->so_options & (SO_TIMESTAMP | SO_BINTIME))) 241 ip_savecontrol(last, &opts, ip, n); 242 SOCKBUF_LOCK(&so->so_rcv); 243 if (sbappendaddr_locked(&so->so_rcv, 244 (struct sockaddr *)ripsrc, n, opts) == 0) { 245 /* should notify about lost packet */ 246 m_freem(n); 247 if (opts) 248 m_freem(opts); 249 SOCKBUF_UNLOCK(&so->so_rcv); 250 } else 251 sorwakeup_locked(so); 252 } else 253 m_freem(n); 254 return (policyfail); 255} 256 257/* 258 * Setup generic address and protocol structures for raw_input routine, then 259 * pass them along with mbuf chain. 260 */ 261void 262rip_input(struct mbuf *m, int off) 263{ 264 struct ifnet *ifp; 265 struct ip *ip = mtod(m, struct ip *); 266 int proto = ip->ip_p; 267 struct inpcb *inp, *last; 268 struct sockaddr_in ripsrc; 269 int hash; 270 271 bzero(&ripsrc, sizeof(ripsrc)); 272 ripsrc.sin_len = sizeof(ripsrc); 273 ripsrc.sin_family = AF_INET; 274 ripsrc.sin_addr = ip->ip_src; 275 last = NULL; 276 277 ifp = m->m_pkthdr.rcvif; 278 279 hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr, 280 ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask); 281 INP_INFO_RLOCK(&V_ripcbinfo); 282 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) { 283 if (inp->inp_ip_p != proto) 284 continue; 285#ifdef INET6 286 /* XXX inp locking */ 287 if ((inp->inp_vflag & INP_IPV4) == 0) 288 continue; 289#endif 290 if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr) 291 continue; 292 if (inp->inp_faddr.s_addr != ip->ip_src.s_addr) 293 continue; 294 if (jailed(inp->inp_cred)) { 295 /* 296 * XXX: If faddr was bound to multicast group, 297 * jailed raw socket will drop datagram. 298 */ 299 if (prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) 300 continue; 301 } 302 if (last != NULL) { 303 struct mbuf *n; 304 305 n = m_copy(m, 0, (int)M_COPYALL); 306 if (n != NULL) 307 (void) rip_append(last, ip, n, &ripsrc); 308 /* XXX count dropped packet */ 309 INP_RUNLOCK(last); 310 } 311 INP_RLOCK(inp); 312 last = inp; 313 } 314 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) { 315 if (inp->inp_ip_p && inp->inp_ip_p != proto) 316 continue; 317#ifdef INET6 318 /* XXX inp locking */ 319 if ((inp->inp_vflag & INP_IPV4) == 0) 320 continue; 321#endif 322 if (!in_nullhost(inp->inp_laddr) && 323 !in_hosteq(inp->inp_laddr, ip->ip_dst)) 324 continue; 325 if (!in_nullhost(inp->inp_faddr) && 326 !in_hosteq(inp->inp_faddr, ip->ip_src)) 327 continue; 328 if (jailed(inp->inp_cred)) { 329 /* 330 * Allow raw socket in jail to receive multicast; 331 * assume process had PRIV_NETINET_RAW at attach, 332 * and fall through into normal filter path if so. 333 */ 334 if (!IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 335 prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) 336 continue; 337 } 338 /* 339 * If this raw socket has multicast state, and we 340 * have received a multicast, check if this socket 341 * should receive it, as multicast filtering is now 342 * the responsibility of the transport layer. 343 */ 344 if (inp->inp_moptions != NULL && 345 IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 346 /* 347 * If the incoming datagram is for IGMP, allow it 348 * through unconditionally to the raw socket. 349 * 350 * In the case of IGMPv2, we may not have explicitly 351 * joined the group, and may have set IFF_ALLMULTI 352 * on the interface. imo_multi_filter() may discard 353 * control traffic we actually need to see. 354 * 355 * Userland multicast routing daemons should continue 356 * filter the control traffic appropriately. 357 */ 358 int blocked; 359 360 blocked = MCAST_PASS; 361 if (proto != IPPROTO_IGMP) { 362 struct sockaddr_in group; 363 364 bzero(&group, sizeof(struct sockaddr_in)); 365 group.sin_len = sizeof(struct sockaddr_in); 366 group.sin_family = AF_INET; 367 group.sin_addr = ip->ip_dst; 368 369 blocked = imo_multi_filter(inp->inp_moptions, 370 ifp, 371 (struct sockaddr *)&group, 372 (struct sockaddr *)&ripsrc); 373 } 374 375 if (blocked != MCAST_PASS) { 376 IPSTAT_INC(ips_notmember); 377 continue; 378 } 379 } 380 if (last != NULL) { 381 struct mbuf *n; 382 383 n = m_copy(m, 0, (int)M_COPYALL); 384 if (n != NULL) 385 (void) rip_append(last, ip, n, &ripsrc); 386 /* XXX count dropped packet */ 387 INP_RUNLOCK(last); 388 } 389 INP_RLOCK(inp); 390 last = inp; 391 } 392 INP_INFO_RUNLOCK(&V_ripcbinfo); 393 if (last != NULL) { 394 if (rip_append(last, ip, m, &ripsrc) != 0) 395 IPSTAT_INC(ips_delivered); 396 INP_RUNLOCK(last); 397 } else { 398 m_freem(m); 399 IPSTAT_INC(ips_noproto); 400 IPSTAT_DEC(ips_delivered); 401 } 402} 403 404/* 405 * Generate IP header and pass packet to ip_output. Tack on options user may 406 * have setup with control call. 407 */ 408int 409rip_output(struct mbuf *m, struct socket *so, u_long dst) 410{ 411 struct ip *ip; 412 int error; 413 struct inpcb *inp = sotoinpcb(so); 414 int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) | 415 IP_ALLOWBROADCAST; 416 417 /* 418 * If the user handed us a complete IP packet, use it. Otherwise, 419 * allocate an mbuf for a header and fill it in. 420 */ 421 if ((inp->inp_flags & INP_HDRINCL) == 0) { 422 if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) { 423 m_freem(m); 424 return(EMSGSIZE); 425 } 426 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT); 427 if (m == NULL) 428 return(ENOBUFS); 429 430 INP_RLOCK(inp); 431 ip = mtod(m, struct ip *); 432 ip->ip_tos = inp->inp_ip_tos; 433 if (inp->inp_flags & INP_DONTFRAG) 434 ip->ip_off = IP_DF; 435 else 436 ip->ip_off = 0; 437 ip->ip_p = inp->inp_ip_p; 438 ip->ip_len = m->m_pkthdr.len; 439 ip->ip_src = inp->inp_laddr; 440 error = prison_get_ip4(inp->inp_cred, &ip->ip_src); 441 if (error != 0) { 442 INP_RUNLOCK(inp); 443 m_freem(m); 444 return (error); 445 } 446 ip->ip_dst.s_addr = dst; 447 ip->ip_ttl = inp->inp_ip_ttl; 448 } else { 449 if (m->m_pkthdr.len > IP_MAXPACKET) { 450 m_freem(m); 451 return(EMSGSIZE); 452 } 453 INP_RLOCK(inp); 454 ip = mtod(m, struct ip *); 455 error = prison_check_ip4(inp->inp_cred, &ip->ip_src); 456 if (error != 0) { 457 INP_RUNLOCK(inp); 458 m_freem(m); 459 return (error); 460 } 461 462 /* 463 * Don't allow both user specified and setsockopt options, 464 * and don't allow packet length sizes that will crash. 465 */ 466 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) 467 || (ip->ip_len > m->m_pkthdr.len) 468 || (ip->ip_len < (ip->ip_hl << 2))) { 469 INP_RUNLOCK(inp); 470 m_freem(m); 471 return (EINVAL); 472 } 473 if (ip->ip_id == 0) 474 ip->ip_id = ip_newid(); 475 476 /* 477 * XXX prevent ip_output from overwriting header fields. 478 */ 479 flags |= IP_RAWOUTPUT; 480 IPSTAT_INC(ips_rawout); 481 } 482 483 if (inp->inp_flags & INP_ONESBCAST) 484 flags |= IP_SENDONES; 485 486#ifdef MAC 487 mac_inpcb_create_mbuf(inp, m); 488#endif 489 490 error = ip_output(m, inp->inp_options, NULL, flags, 491 inp->inp_moptions, inp); 492 INP_RUNLOCK(inp); 493 return (error); 494} 495 496/* 497 * Raw IP socket option processing. 498 * 499 * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could 500 * only be created by a privileged process, and as such, socket option 501 * operations to manage system properties on any raw socket were allowed to 502 * take place without explicit additional access control checks. However, 503 * raw sockets can now also be created in jail(), and therefore explicit 504 * checks are now required. Likewise, raw sockets can be used by a process 505 * after it gives up privilege, so some caution is required. For options 506 * passed down to the IP layer via ip_ctloutput(), checks are assumed to be 507 * performed in ip_ctloutput() and therefore no check occurs here. 508 * Unilaterally checking priv_check() here breaks normal IP socket option 509 * operations on raw sockets. 510 * 511 * When adding new socket options here, make sure to add access control 512 * checks here as necessary. 513 */ 514int 515rip_ctloutput(struct socket *so, struct sockopt *sopt) 516{ 517 struct inpcb *inp = sotoinpcb(so); 518 int error, optval; 519 520 if (sopt->sopt_level != IPPROTO_IP) { 521 if ((sopt->sopt_level == SOL_SOCKET) && 522 (sopt->sopt_name == SO_SETFIB)) { 523 inp->inp_inc.inc_fibnum = so->so_fibnum; 524 return (0); 525 } 526 return (EINVAL); 527 } 528 529 error = 0; 530 switch (sopt->sopt_dir) { 531 case SOPT_GET: 532 switch (sopt->sopt_name) { 533 case IP_HDRINCL: 534 optval = inp->inp_flags & INP_HDRINCL; 535 error = sooptcopyout(sopt, &optval, sizeof optval); 536 break; 537 538 case IP_FW_ADD: /* ADD actually returns the body... */ 539 case IP_FW_GET: 540 case IP_FW_TABLE_GETSIZE: 541 case IP_FW_TABLE_LIST: 542 case IP_FW_NAT_GET_CONFIG: 543 case IP_FW_NAT_GET_LOG: 544 if (V_ip_fw_ctl_ptr != NULL) 545 error = V_ip_fw_ctl_ptr(sopt); 546 else 547 error = ENOPROTOOPT; 548 break; 549 550 case IP_DUMMYNET_GET: 551 if (ip_dn_ctl_ptr != NULL) 552 error = ip_dn_ctl_ptr(sopt); 553 else 554 error = ENOPROTOOPT; 555 break ; 556 557 case MRT_INIT: 558 case MRT_DONE: 559 case MRT_ADD_VIF: 560 case MRT_DEL_VIF: 561 case MRT_ADD_MFC: 562 case MRT_DEL_MFC: 563 case MRT_VERSION: 564 case MRT_ASSERT: 565 case MRT_API_SUPPORT: 566 case MRT_API_CONFIG: 567 case MRT_ADD_BW_UPCALL: 568 case MRT_DEL_BW_UPCALL: 569 error = priv_check(curthread, PRIV_NETINET_MROUTE); 570 if (error != 0) 571 return (error); 572 error = ip_mrouter_get ? ip_mrouter_get(so, sopt) : 573 EOPNOTSUPP; 574 break; 575 576 default: 577 error = ip_ctloutput(so, sopt); 578 break; 579 } 580 break; 581 582 case SOPT_SET: 583 switch (sopt->sopt_name) { 584 case IP_HDRINCL: 585 error = sooptcopyin(sopt, &optval, sizeof optval, 586 sizeof optval); 587 if (error) 588 break; 589 if (optval) 590 inp->inp_flags |= INP_HDRINCL; 591 else 592 inp->inp_flags &= ~INP_HDRINCL; 593 break; 594 595 case IP_FW_ADD: 596 case IP_FW_DEL: 597 case IP_FW_FLUSH: 598 case IP_FW_ZERO: 599 case IP_FW_RESETLOG: 600 case IP_FW_TABLE_ADD: 601 case IP_FW_TABLE_DEL: 602 case IP_FW_TABLE_FLUSH: 603 case IP_FW_NAT_CFG: 604 case IP_FW_NAT_DEL: 605 if (V_ip_fw_ctl_ptr != NULL) 606 error = V_ip_fw_ctl_ptr(sopt); 607 else 608 error = ENOPROTOOPT; 609 break; 610 611 case IP_DUMMYNET_CONFIGURE: 612 case IP_DUMMYNET_DEL: 613 case IP_DUMMYNET_FLUSH: 614 if (ip_dn_ctl_ptr != NULL) 615 error = ip_dn_ctl_ptr(sopt); 616 else 617 error = ENOPROTOOPT ; 618 break ; 619 620 case IP_RSVP_ON: 621 error = priv_check(curthread, PRIV_NETINET_MROUTE); 622 if (error != 0) 623 return (error); 624 error = ip_rsvp_init(so); 625 break; 626 627 case IP_RSVP_OFF: 628 error = priv_check(curthread, PRIV_NETINET_MROUTE); 629 if (error != 0) 630 return (error); 631 error = ip_rsvp_done(); 632 break; 633 634 case IP_RSVP_VIF_ON: 635 case IP_RSVP_VIF_OFF: 636 error = priv_check(curthread, PRIV_NETINET_MROUTE); 637 if (error != 0) 638 return (error); 639 error = ip_rsvp_vif ? 640 ip_rsvp_vif(so, sopt) : EINVAL; 641 break; 642 643 case MRT_INIT: 644 case MRT_DONE: 645 case MRT_ADD_VIF: 646 case MRT_DEL_VIF: 647 case MRT_ADD_MFC: 648 case MRT_DEL_MFC: 649 case MRT_VERSION: 650 case MRT_ASSERT: 651 case MRT_API_SUPPORT: 652 case MRT_API_CONFIG: 653 case MRT_ADD_BW_UPCALL: 654 case MRT_DEL_BW_UPCALL: 655 error = priv_check(curthread, PRIV_NETINET_MROUTE); 656 if (error != 0) 657 return (error); 658 error = ip_mrouter_set ? ip_mrouter_set(so, sopt) : 659 EOPNOTSUPP; 660 break; 661 662 default: 663 error = ip_ctloutput(so, sopt); 664 break; 665 } 666 break; 667 } 668 669 return (error); 670} 671 672/* 673 * This function exists solely to receive the PRC_IFDOWN messages which are 674 * sent by if_down(). It looks for an ifaddr whose ifa_addr is sa, and calls 675 * in_ifadown() to remove all routes corresponding to that address. It also 676 * receives the PRC_IFUP messages from if_up() and reinstalls the interface 677 * routes. 678 */ 679void 680rip_ctlinput(int cmd, struct sockaddr *sa, void *vip) 681{ 682 struct in_ifaddr *ia; 683 struct ifnet *ifp; 684 int err; 685 int flags; 686 687 switch (cmd) { 688 case PRC_IFDOWN: 689 IN_IFADDR_RLOCK(); 690 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 691 if (ia->ia_ifa.ifa_addr == sa 692 && (ia->ia_flags & IFA_ROUTE)) { 693 ifa_ref(&ia->ia_ifa); 694 IN_IFADDR_RUNLOCK(); 695 /* 696 * in_ifscrub kills the interface route. 697 */ 698 in_ifscrub(ia->ia_ifp, ia); 699 /* 700 * in_ifadown gets rid of all the rest of the 701 * routes. This is not quite the right thing 702 * to do, but at least if we are running a 703 * routing process they will come back. 704 */ 705 in_ifadown(&ia->ia_ifa, 0); 706 ifa_free(&ia->ia_ifa); 707 break; 708 } 709 } 710 if (ia == NULL) /* If ia matched, already unlocked. */ 711 IN_IFADDR_RUNLOCK(); 712 break; 713 714 case PRC_IFUP: 715 IN_IFADDR_RLOCK(); 716 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 717 if (ia->ia_ifa.ifa_addr == sa) 718 break; 719 } 720 if (ia == NULL || (ia->ia_flags & IFA_ROUTE)) { 721 IN_IFADDR_RUNLOCK(); 722 return; 723 } 724 ifa_ref(&ia->ia_ifa); 725 IN_IFADDR_RUNLOCK(); 726 flags = RTF_UP; 727 ifp = ia->ia_ifa.ifa_ifp; 728 729 if ((ifp->if_flags & IFF_LOOPBACK) 730 || (ifp->if_flags & IFF_POINTOPOINT)) 731 flags |= RTF_HOST; 732 733 err = rtinit(&ia->ia_ifa, RTM_ADD, flags); 734 if (err == 0) 735 ia->ia_flags |= IFA_ROUTE; 736 err = ifa_add_loopback_route((struct ifaddr *)ia, sa); 737 ifa_free(&ia->ia_ifa); 738 break; 739 } 740} 741 742u_long rip_sendspace = 9216; 743u_long rip_recvspace = 9216; 744 745SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW, 746 &rip_sendspace, 0, "Maximum outgoing raw IP datagram size"); 747SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW, 748 &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams"); 749 750static int 751rip_attach(struct socket *so, int proto, struct thread *td) 752{ 753 struct inpcb *inp; 754 int error; 755 756 inp = sotoinpcb(so); 757 KASSERT(inp == NULL, ("rip_attach: inp != NULL")); 758 759 error = priv_check(td, PRIV_NETINET_RAW); 760 if (error) 761 return (error); 762 if (proto >= IPPROTO_MAX || proto < 0) 763 return EPROTONOSUPPORT; 764 error = soreserve(so, rip_sendspace, rip_recvspace); 765 if (error) 766 return (error); 767 INP_INFO_WLOCK(&V_ripcbinfo); 768 error = in_pcballoc(so, &V_ripcbinfo); 769 if (error) { 770 INP_INFO_WUNLOCK(&V_ripcbinfo); 771 return (error); 772 } 773 inp = (struct inpcb *)so->so_pcb; 774 inp->inp_vflag |= INP_IPV4; 775 inp->inp_ip_p = proto; 776 inp->inp_ip_ttl = V_ip_defttl; 777 rip_inshash(inp); 778 INP_INFO_WUNLOCK(&V_ripcbinfo); 779 INP_WUNLOCK(inp); 780 return (0); 781} 782 783static void 784rip_detach(struct socket *so) 785{ 786 struct inpcb *inp; 787 788 inp = sotoinpcb(so); 789 KASSERT(inp != NULL, ("rip_detach: inp == NULL")); 790 KASSERT(inp->inp_faddr.s_addr == INADDR_ANY, 791 ("rip_detach: not closed")); 792 793 INP_INFO_WLOCK(&V_ripcbinfo); 794 INP_WLOCK(inp); 795 rip_delhash(inp); 796 if (so == V_ip_mrouter && ip_mrouter_done) 797 ip_mrouter_done(); 798 if (ip_rsvp_force_done) 799 ip_rsvp_force_done(so); 800 if (so == V_ip_rsvpd) 801 ip_rsvp_done(); 802 in_pcbdetach(inp); 803 in_pcbfree(inp); 804 INP_INFO_WUNLOCK(&V_ripcbinfo); 805} 806 807static void 808rip_dodisconnect(struct socket *so, struct inpcb *inp) 809{ 810 811 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 812 INP_WLOCK_ASSERT(inp); 813 814 rip_delhash(inp); 815 inp->inp_faddr.s_addr = INADDR_ANY; 816 rip_inshash(inp); 817 SOCK_LOCK(so); 818 so->so_state &= ~SS_ISCONNECTED; 819 SOCK_UNLOCK(so); 820} 821 822static void 823rip_abort(struct socket *so) 824{ 825 struct inpcb *inp; 826 827 inp = sotoinpcb(so); 828 KASSERT(inp != NULL, ("rip_abort: inp == NULL")); 829 830 INP_INFO_WLOCK(&V_ripcbinfo); 831 INP_WLOCK(inp); 832 rip_dodisconnect(so, inp); 833 INP_WUNLOCK(inp); 834 INP_INFO_WUNLOCK(&V_ripcbinfo); 835} 836 837static void 838rip_close(struct socket *so) 839{ 840 struct inpcb *inp; 841 842 inp = sotoinpcb(so); 843 KASSERT(inp != NULL, ("rip_close: inp == NULL")); 844 845 INP_INFO_WLOCK(&V_ripcbinfo); 846 INP_WLOCK(inp); 847 rip_dodisconnect(so, inp); 848 INP_WUNLOCK(inp); 849 INP_INFO_WUNLOCK(&V_ripcbinfo); 850} 851 852static int 853rip_disconnect(struct socket *so) 854{ 855 struct inpcb *inp; 856 857 if ((so->so_state & SS_ISCONNECTED) == 0) 858 return (ENOTCONN); 859 860 inp = sotoinpcb(so); 861 KASSERT(inp != NULL, ("rip_disconnect: inp == NULL")); 862 863 INP_INFO_WLOCK(&V_ripcbinfo); 864 INP_WLOCK(inp); 865 rip_dodisconnect(so, inp); 866 INP_WUNLOCK(inp); 867 INP_INFO_WUNLOCK(&V_ripcbinfo); 868 return (0); 869} 870 871static int 872rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 873{ 874 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 875 struct inpcb *inp; 876 int error; 877 878 if (nam->sa_len != sizeof(*addr)) 879 return (EINVAL); 880 881 error = prison_check_ip4(td->td_ucred, &addr->sin_addr); 882 if (error != 0) 883 return (error); 884 885 inp = sotoinpcb(so); 886 KASSERT(inp != NULL, ("rip_bind: inp == NULL")); 887 888 if (TAILQ_EMPTY(&V_ifnet) || 889 (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) || 890 (addr->sin_addr.s_addr && 891 (inp->inp_flags & INP_BINDANY) == 0 && 892 ifa_ifwithaddr_check((struct sockaddr *)addr) == 0)) 893 return (EADDRNOTAVAIL); 894 895 INP_INFO_WLOCK(&V_ripcbinfo); 896 INP_WLOCK(inp); 897 rip_delhash(inp); 898 inp->inp_laddr = addr->sin_addr; 899 rip_inshash(inp); 900 INP_WUNLOCK(inp); 901 INP_INFO_WUNLOCK(&V_ripcbinfo); 902 return (0); 903} 904 905static int 906rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 907{ 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 int error, i, n; 979 struct inpcb *inp, **inp_list; 980 inp_gen_t gencnt; 981 struct xinpgen xig; 982 983 /* 984 * The process of preparing the TCB list is too time-consuming and 985 * resource-intensive to repeat twice on every request. 986 */ 987 if (req->oldptr == 0) { 988 n = V_ripcbinfo.ipi_count; 989 req->oldidx = 2 * (sizeof xig) 990 + (n + n/8) * sizeof(struct xinpcb); 991 return (0); 992 } 993 994 if (req->newptr != 0) 995 return (EPERM); 996 997 /* 998 * OK, now we're committed to doing something. 999 */ 1000 INP_INFO_RLOCK(&V_ripcbinfo); 1001 gencnt = V_ripcbinfo.ipi_gencnt; 1002 n = V_ripcbinfo.ipi_count; 1003 INP_INFO_RUNLOCK(&V_ripcbinfo); 1004 1005 xig.xig_len = sizeof xig; 1006 xig.xig_count = n; 1007 xig.xig_gen = gencnt; 1008 xig.xig_sogen = so_gencnt; 1009 error = SYSCTL_OUT(req, &xig, sizeof xig); 1010 if (error) 1011 return (error); 1012 1013 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 1014 if (inp_list == 0) 1015 return (ENOMEM); 1016 1017 INP_INFO_RLOCK(&V_ripcbinfo); 1018 for (inp = LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n; 1019 inp = LIST_NEXT(inp, inp_list)) { 1020 INP_RLOCK(inp); 1021 if (inp->inp_gencnt <= gencnt && 1022 cr_canseeinpcb(req->td->td_ucred, inp) == 0) { 1023 /* XXX held references? */ 1024 inp_list[i++] = inp; 1025 } 1026 INP_RUNLOCK(inp); 1027 } 1028 INP_INFO_RUNLOCK(&V_ripcbinfo); 1029 n = i; 1030 1031 error = 0; 1032 for (i = 0; i < n; i++) { 1033 inp = inp_list[i]; 1034 INP_RLOCK(inp); 1035 if (inp->inp_gencnt <= gencnt) { 1036 struct xinpcb xi; 1037 1038 bzero(&xi, sizeof(xi)); 1039 xi.xi_len = sizeof xi; 1040 /* XXX should avoid extra copy */ 1041 bcopy(inp, &xi.xi_inp, sizeof *inp); 1042 if (inp->inp_socket) 1043 sotoxsocket(inp->inp_socket, &xi.xi_socket); 1044 INP_RUNLOCK(inp); 1045 error = SYSCTL_OUT(req, &xi, sizeof xi); 1046 } else 1047 INP_RUNLOCK(inp); 1048 } 1049 if (!error) { 1050 /* 1051 * Give the user an updated idea of our state. If the 1052 * generation differs from what we told her before, she knows 1053 * that something happened while we were processing this 1054 * request, and it might be necessary to retry. 1055 */ 1056 INP_INFO_RLOCK(&V_ripcbinfo); 1057 xig.xig_gen = V_ripcbinfo.ipi_gencnt; 1058 xig.xig_sogen = so_gencnt; 1059 xig.xig_count = V_ripcbinfo.ipi_count; 1060 INP_INFO_RUNLOCK(&V_ripcbinfo); 1061 error = SYSCTL_OUT(req, &xig, sizeof xig); 1062 } 1063 free(inp_list, M_TEMP); 1064 return (error); 1065} 1066 1067SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0, 1068 rip_pcblist, "S,xinpcb", "List of active raw IP sockets"); 1069 1070struct pr_usrreqs rip_usrreqs = { 1071 .pru_abort = rip_abort, 1072 .pru_attach = rip_attach, 1073 .pru_bind = rip_bind, 1074 .pru_connect = rip_connect, 1075 .pru_control = in_control, 1076 .pru_detach = rip_detach, 1077 .pru_disconnect = rip_disconnect, 1078 .pru_peeraddr = in_getpeeraddr, 1079 .pru_send = rip_send, 1080 .pru_shutdown = rip_shutdown, 1081 .pru_sockaddr = in_getsockaddr, 1082 .pru_sosetlabel = in_pcbsosetlabel, 1083 .pru_close = rip_close, 1084}; 1085