raw_ip.c revision 195699
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 195699 2009-07-14 22:48:30Z 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#ifdef IPSEC 72#include <netipsec/ipsec.h> 73#endif /*IPSEC*/ 74 75#include <security/mac/mac_framework.h> 76 77VNET_DEFINE(struct inpcbhead, ripcb); 78VNET_DEFINE(struct inpcbinfo, ripcbinfo); 79 80#define V_ripcb VNET_GET(ripcb) 81#define V_ripcbinfo VNET_GET(ripcbinfo) 82 83/* 84 * Control and data hooks for ipfw and dummynet. 85 * The data hooks are not used here but it is convenient 86 * to keep them all in one place. 87 */ 88int (*ip_fw_ctl_ptr)(struct sockopt *) = NULL; 89int (*ip_dn_ctl_ptr)(struct sockopt *) = NULL; 90int (*ip_fw_chk_ptr)(struct ip_fw_args *args) = NULL; 91int (*ip_dn_io_ptr)(struct mbuf **m, int dir, struct ip_fw_args *fwa) = NULL; 92 93/* 94 * Hooks for multicast routing. They all default to NULL, so leave them not 95 * initialized and rely on BSS being set to 0. 96 */ 97 98/* 99 * The socket used to communicate with the multicast routing daemon. 100 */ 101VNET_DEFINE(struct socket *, ip_mrouter); 102 103/* 104 * The various mrouter and rsvp functions. 105 */ 106int (*ip_mrouter_set)(struct socket *, struct sockopt *); 107int (*ip_mrouter_get)(struct socket *, struct sockopt *); 108int (*ip_mrouter_done)(void); 109int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 110 struct ip_moptions *); 111int (*mrt_ioctl)(u_long, caddr_t, int); 112int (*legal_vif_num)(int); 113u_long (*ip_mcast_src)(int); 114 115void (*rsvp_input_p)(struct mbuf *m, int off); 116int (*ip_rsvp_vif)(struct socket *, struct sockopt *); 117void (*ip_rsvp_force_done)(struct socket *); 118 119/* 120 * Hash functions 121 */ 122 123#define INP_PCBHASH_RAW_SIZE 256 124#define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \ 125 (((proto) + (laddr) + (faddr)) % (mask) + 1) 126 127static void 128rip_inshash(struct inpcb *inp) 129{ 130 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; 131 struct inpcbhead *pcbhash; 132 int hash; 133 134 INP_INFO_WLOCK_ASSERT(pcbinfo); 135 INP_WLOCK_ASSERT(inp); 136 137 if (inp->inp_ip_p != 0 && 138 inp->inp_laddr.s_addr != INADDR_ANY && 139 inp->inp_faddr.s_addr != INADDR_ANY) { 140 hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr, 141 inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask); 142 } else 143 hash = 0; 144 pcbhash = &pcbinfo->ipi_hashbase[hash]; 145 LIST_INSERT_HEAD(pcbhash, inp, inp_hash); 146} 147 148static void 149rip_delhash(struct inpcb *inp) 150{ 151 152 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 153 INP_WLOCK_ASSERT(inp); 154 155 LIST_REMOVE(inp, inp_hash); 156} 157 158/* 159 * Raw interface to IP protocol. 160 */ 161 162/* 163 * Initialize raw connection block q. 164 */ 165static void 166rip_zone_change(void *tag) 167{ 168 169 uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets); 170} 171 172static int 173rip_inpcb_init(void *mem, int size, int flags) 174{ 175 struct inpcb *inp = mem; 176 177 INP_LOCK_INIT(inp, "inp", "rawinp"); 178 return (0); 179} 180 181void 182rip_init(void) 183{ 184 185 INP_INFO_LOCK_INIT(&V_ripcbinfo, "rip"); 186 LIST_INIT(&V_ripcb); 187#ifdef VIMAGE 188 V_ripcbinfo.ipi_vnet = curvnet; 189#endif 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} 201 202#ifdef VIMAGE 203void 204rip_destroy(void) 205{ 206 207 hashdestroy(V_ripcbinfo.ipi_hashbase, M_PCB, 208 V_ripcbinfo.ipi_hashmask); 209 hashdestroy(V_ripcbinfo.ipi_porthashbase, M_PCB, 210 V_ripcbinfo.ipi_porthashmask); 211} 212#endif 213 214static int 215rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n, 216 struct sockaddr_in *ripsrc) 217{ 218 int policyfail = 0; 219 220 INP_RLOCK_ASSERT(last); 221 222#ifdef IPSEC 223 /* check AH/ESP integrity. */ 224 if (ipsec4_in_reject(n, last)) { 225 policyfail = 1; 226 } 227#endif /* IPSEC */ 228#ifdef MAC 229 if (!policyfail && mac_inpcb_check_deliver(last, n) != 0) 230 policyfail = 1; 231#endif 232 /* Check the minimum TTL for socket. */ 233 if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl) 234 policyfail = 1; 235 if (!policyfail) { 236 struct mbuf *opts = NULL; 237 struct socket *so; 238 239 so = last->inp_socket; 240 if ((last->inp_flags & INP_CONTROLOPTS) || 241 (so->so_options & (SO_TIMESTAMP | SO_BINTIME))) 242 ip_savecontrol(last, &opts, ip, n); 243 SOCKBUF_LOCK(&so->so_rcv); 244 if (sbappendaddr_locked(&so->so_rcv, 245 (struct sockaddr *)ripsrc, n, opts) == 0) { 246 /* should notify about lost packet */ 247 m_freem(n); 248 if (opts) 249 m_freem(opts); 250 SOCKBUF_UNLOCK(&so->so_rcv); 251 } else 252 sorwakeup_locked(so); 253 } else 254 m_freem(n); 255 return (policyfail); 256} 257 258/* 259 * Setup generic address and protocol structures for raw_input routine, then 260 * pass them along with mbuf chain. 261 */ 262void 263rip_input(struct mbuf *m, int off) 264{ 265 struct ifnet *ifp; 266 struct ip *ip = mtod(m, struct ip *); 267 int proto = ip->ip_p; 268 struct inpcb *inp, *last; 269 struct sockaddr_in ripsrc; 270 int hash; 271 272 bzero(&ripsrc, sizeof(ripsrc)); 273 ripsrc.sin_len = sizeof(ripsrc); 274 ripsrc.sin_family = AF_INET; 275 ripsrc.sin_addr = ip->ip_src; 276 last = NULL; 277 278 ifp = m->m_pkthdr.rcvif; 279 280 hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr, 281 ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask); 282 INP_INFO_RLOCK(&V_ripcbinfo); 283 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) { 284 if (inp->inp_ip_p != proto) 285 continue; 286#ifdef INET6 287 /* XXX inp locking */ 288 if ((inp->inp_vflag & INP_IPV4) == 0) 289 continue; 290#endif 291 if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr) 292 continue; 293 if (inp->inp_faddr.s_addr != ip->ip_src.s_addr) 294 continue; 295 if (jailed(inp->inp_cred)) { 296 /* 297 * XXX: If faddr was bound to multicast group, 298 * jailed raw socket will drop datagram. 299 */ 300 if (prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) 301 continue; 302 } 303 if (last != NULL) { 304 struct mbuf *n; 305 306 n = m_copy(m, 0, (int)M_COPYALL); 307 if (n != NULL) 308 (void) rip_append(last, ip, n, &ripsrc); 309 /* XXX count dropped packet */ 310 INP_RUNLOCK(last); 311 } 312 INP_RLOCK(inp); 313 last = inp; 314 } 315 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) { 316 if (inp->inp_ip_p && inp->inp_ip_p != proto) 317 continue; 318#ifdef INET6 319 /* XXX inp locking */ 320 if ((inp->inp_vflag & INP_IPV4) == 0) 321 continue; 322#endif 323 if (!in_nullhost(inp->inp_laddr) && 324 !in_hosteq(inp->inp_laddr, ip->ip_dst)) 325 continue; 326 if (!in_nullhost(inp->inp_faddr) && 327 !in_hosteq(inp->inp_faddr, ip->ip_src)) 328 continue; 329 if (jailed(inp->inp_cred)) { 330 /* 331 * Allow raw socket in jail to receive multicast; 332 * assume process had PRIV_NETINET_RAW at attach, 333 * and fall through into normal filter path if so. 334 */ 335 if (!IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) && 336 prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0) 337 continue; 338 } 339 /* 340 * If this raw socket has multicast state, and we 341 * have received a multicast, check if this socket 342 * should receive it, as multicast filtering is now 343 * the responsibility of the transport layer. 344 */ 345 if (inp->inp_moptions != NULL && 346 IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 347 struct sockaddr_in group; 348 int blocked; 349 350 bzero(&group, sizeof(struct sockaddr_in)); 351 group.sin_len = sizeof(struct sockaddr_in); 352 group.sin_family = AF_INET; 353 group.sin_addr = ip->ip_dst; 354 355 blocked = imo_multi_filter(inp->inp_moptions, ifp, 356 (struct sockaddr *)&group, 357 (struct sockaddr *)&ripsrc); 358 if (blocked != MCAST_PASS) { 359 IPSTAT_INC(ips_notmember); 360 continue; 361 } 362 } 363 if (last != NULL) { 364 struct mbuf *n; 365 366 n = m_copy(m, 0, (int)M_COPYALL); 367 if (n != NULL) 368 (void) rip_append(last, ip, n, &ripsrc); 369 /* XXX count dropped packet */ 370 INP_RUNLOCK(last); 371 } 372 INP_RLOCK(inp); 373 last = inp; 374 } 375 INP_INFO_RUNLOCK(&V_ripcbinfo); 376 if (last != NULL) { 377 if (rip_append(last, ip, m, &ripsrc) != 0) 378 IPSTAT_INC(ips_delivered); 379 INP_RUNLOCK(last); 380 } else { 381 m_freem(m); 382 IPSTAT_INC(ips_noproto); 383 IPSTAT_DEC(ips_delivered); 384 } 385} 386 387/* 388 * Generate IP header and pass packet to ip_output. Tack on options user may 389 * have setup with control call. 390 */ 391int 392rip_output(struct mbuf *m, struct socket *so, u_long dst) 393{ 394 struct ip *ip; 395 int error; 396 struct inpcb *inp = sotoinpcb(so); 397 int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) | 398 IP_ALLOWBROADCAST; 399 400 /* 401 * If the user handed us a complete IP packet, use it. Otherwise, 402 * allocate an mbuf for a header and fill it in. 403 */ 404 if ((inp->inp_flags & INP_HDRINCL) == 0) { 405 if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) { 406 m_freem(m); 407 return(EMSGSIZE); 408 } 409 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT); 410 if (m == NULL) 411 return(ENOBUFS); 412 413 INP_RLOCK(inp); 414 ip = mtod(m, struct ip *); 415 ip->ip_tos = inp->inp_ip_tos; 416 if (inp->inp_flags & INP_DONTFRAG) 417 ip->ip_off = IP_DF; 418 else 419 ip->ip_off = 0; 420 ip->ip_p = inp->inp_ip_p; 421 ip->ip_len = m->m_pkthdr.len; 422 ip->ip_src = inp->inp_laddr; 423 error = prison_get_ip4(inp->inp_cred, &ip->ip_src); 424 if (error != 0) { 425 INP_RUNLOCK(inp); 426 m_freem(m); 427 return (error); 428 } 429 ip->ip_dst.s_addr = dst; 430 ip->ip_ttl = inp->inp_ip_ttl; 431 } else { 432 if (m->m_pkthdr.len > IP_MAXPACKET) { 433 m_freem(m); 434 return(EMSGSIZE); 435 } 436 INP_RLOCK(inp); 437 ip = mtod(m, struct ip *); 438 error = prison_check_ip4(inp->inp_cred, &ip->ip_src); 439 if (error != 0) { 440 INP_RUNLOCK(inp); 441 m_freem(m); 442 return (error); 443 } 444 445 /* 446 * Don't allow both user specified and setsockopt options, 447 * and don't allow packet length sizes that will crash. 448 */ 449 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) 450 || (ip->ip_len > m->m_pkthdr.len) 451 || (ip->ip_len < (ip->ip_hl << 2))) { 452 INP_RUNLOCK(inp); 453 m_freem(m); 454 return (EINVAL); 455 } 456 if (ip->ip_id == 0) 457 ip->ip_id = ip_newid(); 458 459 /* 460 * XXX prevent ip_output from overwriting header fields. 461 */ 462 flags |= IP_RAWOUTPUT; 463 IPSTAT_INC(ips_rawout); 464 } 465 466 if (inp->inp_flags & INP_ONESBCAST) 467 flags |= IP_SENDONES; 468 469#ifdef MAC 470 mac_inpcb_create_mbuf(inp, m); 471#endif 472 473 error = ip_output(m, inp->inp_options, NULL, flags, 474 inp->inp_moptions, inp); 475 INP_RUNLOCK(inp); 476 return (error); 477} 478 479/* 480 * Raw IP socket option processing. 481 * 482 * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could 483 * only be created by a privileged process, and as such, socket option 484 * operations to manage system properties on any raw socket were allowed to 485 * take place without explicit additional access control checks. However, 486 * raw sockets can now also be created in jail(), and therefore explicit 487 * checks are now required. Likewise, raw sockets can be used by a process 488 * after it gives up privilege, so some caution is required. For options 489 * passed down to the IP layer via ip_ctloutput(), checks are assumed to be 490 * performed in ip_ctloutput() and therefore no check occurs here. 491 * Unilaterally checking priv_check() here breaks normal IP socket option 492 * operations on raw sockets. 493 * 494 * When adding new socket options here, make sure to add access control 495 * checks here as necessary. 496 */ 497int 498rip_ctloutput(struct socket *so, struct sockopt *sopt) 499{ 500 struct inpcb *inp = sotoinpcb(so); 501 int error, optval; 502 503 if (sopt->sopt_level != IPPROTO_IP) { 504 if ((sopt->sopt_level == SOL_SOCKET) && 505 (sopt->sopt_name == SO_SETFIB)) { 506 inp->inp_inc.inc_fibnum = so->so_fibnum; 507 return (0); 508 } 509 return (EINVAL); 510 } 511 512 error = 0; 513 switch (sopt->sopt_dir) { 514 case SOPT_GET: 515 switch (sopt->sopt_name) { 516 case IP_HDRINCL: 517 optval = inp->inp_flags & INP_HDRINCL; 518 error = sooptcopyout(sopt, &optval, sizeof optval); 519 break; 520 521 case IP_FW_ADD: /* ADD actually returns the body... */ 522 case IP_FW_GET: 523 case IP_FW_TABLE_GETSIZE: 524 case IP_FW_TABLE_LIST: 525 case IP_FW_NAT_GET_CONFIG: 526 case IP_FW_NAT_GET_LOG: 527 if (ip_fw_ctl_ptr != NULL) 528 error = ip_fw_ctl_ptr(sopt); 529 else 530 error = ENOPROTOOPT; 531 break; 532 533 case IP_DUMMYNET_GET: 534 if (ip_dn_ctl_ptr != NULL) 535 error = ip_dn_ctl_ptr(sopt); 536 else 537 error = ENOPROTOOPT; 538 break ; 539 540 case MRT_INIT: 541 case MRT_DONE: 542 case MRT_ADD_VIF: 543 case MRT_DEL_VIF: 544 case MRT_ADD_MFC: 545 case MRT_DEL_MFC: 546 case MRT_VERSION: 547 case MRT_ASSERT: 548 case MRT_API_SUPPORT: 549 case MRT_API_CONFIG: 550 case MRT_ADD_BW_UPCALL: 551 case MRT_DEL_BW_UPCALL: 552 error = priv_check(curthread, PRIV_NETINET_MROUTE); 553 if (error != 0) 554 return (error); 555 error = ip_mrouter_get ? ip_mrouter_get(so, sopt) : 556 EOPNOTSUPP; 557 break; 558 559 default: 560 error = ip_ctloutput(so, sopt); 561 break; 562 } 563 break; 564 565 case SOPT_SET: 566 switch (sopt->sopt_name) { 567 case IP_HDRINCL: 568 error = sooptcopyin(sopt, &optval, sizeof optval, 569 sizeof optval); 570 if (error) 571 break; 572 if (optval) 573 inp->inp_flags |= INP_HDRINCL; 574 else 575 inp->inp_flags &= ~INP_HDRINCL; 576 break; 577 578 case IP_FW_ADD: 579 case IP_FW_DEL: 580 case IP_FW_FLUSH: 581 case IP_FW_ZERO: 582 case IP_FW_RESETLOG: 583 case IP_FW_TABLE_ADD: 584 case IP_FW_TABLE_DEL: 585 case IP_FW_TABLE_FLUSH: 586 case IP_FW_NAT_CFG: 587 case IP_FW_NAT_DEL: 588 if (ip_fw_ctl_ptr != NULL) 589 error = ip_fw_ctl_ptr(sopt); 590 else 591 error = ENOPROTOOPT; 592 break; 593 594 case IP_DUMMYNET_CONFIGURE: 595 case IP_DUMMYNET_DEL: 596 case IP_DUMMYNET_FLUSH: 597 if (ip_dn_ctl_ptr != NULL) 598 error = ip_dn_ctl_ptr(sopt); 599 else 600 error = ENOPROTOOPT ; 601 break ; 602 603 case IP_RSVP_ON: 604 error = priv_check(curthread, PRIV_NETINET_MROUTE); 605 if (error != 0) 606 return (error); 607 error = ip_rsvp_init(so); 608 break; 609 610 case IP_RSVP_OFF: 611 error = priv_check(curthread, PRIV_NETINET_MROUTE); 612 if (error != 0) 613 return (error); 614 error = ip_rsvp_done(); 615 break; 616 617 case IP_RSVP_VIF_ON: 618 case IP_RSVP_VIF_OFF: 619 error = priv_check(curthread, PRIV_NETINET_MROUTE); 620 if (error != 0) 621 return (error); 622 error = ip_rsvp_vif ? 623 ip_rsvp_vif(so, sopt) : EINVAL; 624 break; 625 626 case MRT_INIT: 627 case MRT_DONE: 628 case MRT_ADD_VIF: 629 case MRT_DEL_VIF: 630 case MRT_ADD_MFC: 631 case MRT_DEL_MFC: 632 case MRT_VERSION: 633 case MRT_ASSERT: 634 case MRT_API_SUPPORT: 635 case MRT_API_CONFIG: 636 case MRT_ADD_BW_UPCALL: 637 case MRT_DEL_BW_UPCALL: 638 error = priv_check(curthread, PRIV_NETINET_MROUTE); 639 if (error != 0) 640 return (error); 641 error = ip_mrouter_set ? ip_mrouter_set(so, sopt) : 642 EOPNOTSUPP; 643 break; 644 645 default: 646 error = ip_ctloutput(so, sopt); 647 break; 648 } 649 break; 650 } 651 652 return (error); 653} 654 655/* 656 * This function exists solely to receive the PRC_IFDOWN messages which are 657 * sent by if_down(). It looks for an ifaddr whose ifa_addr is sa, and calls 658 * in_ifadown() to remove all routes corresponding to that address. It also 659 * receives the PRC_IFUP messages from if_up() and reinstalls the interface 660 * routes. 661 */ 662void 663rip_ctlinput(int cmd, struct sockaddr *sa, void *vip) 664{ 665 struct in_ifaddr *ia; 666 struct ifnet *ifp; 667 int err; 668 int flags; 669 670 switch (cmd) { 671 case PRC_IFDOWN: 672 IN_IFADDR_RLOCK(); 673 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 674 if (ia->ia_ifa.ifa_addr == sa 675 && (ia->ia_flags & IFA_ROUTE)) { 676 ifa_ref(&ia->ia_ifa); 677 IN_IFADDR_RUNLOCK(); 678 /* 679 * in_ifscrub kills the interface route. 680 */ 681 in_ifscrub(ia->ia_ifp, ia); 682 /* 683 * in_ifadown gets rid of all the rest of the 684 * routes. This is not quite the right thing 685 * to do, but at least if we are running a 686 * routing process they will come back. 687 */ 688 in_ifadown(&ia->ia_ifa, 0); 689 ifa_free(&ia->ia_ifa); 690 break; 691 } 692 } 693 if (ia == NULL) /* If ia matched, already unlocked. */ 694 IN_IFADDR_RUNLOCK(); 695 break; 696 697 case PRC_IFUP: 698 IN_IFADDR_RLOCK(); 699 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 700 if (ia->ia_ifa.ifa_addr == sa) 701 break; 702 } 703 if (ia == NULL || (ia->ia_flags & IFA_ROUTE)) { 704 IN_IFADDR_RUNLOCK(); 705 return; 706 } 707 ifa_ref(&ia->ia_ifa); 708 IN_IFADDR_RUNLOCK(); 709 flags = RTF_UP; 710 ifp = ia->ia_ifa.ifa_ifp; 711 712 if ((ifp->if_flags & IFF_LOOPBACK) 713 || (ifp->if_flags & IFF_POINTOPOINT)) 714 flags |= RTF_HOST; 715 716 err = rtinit(&ia->ia_ifa, RTM_ADD, flags); 717 if (err == 0) 718 ia->ia_flags |= IFA_ROUTE; 719 ifa_free(&ia->ia_ifa); 720 break; 721 } 722} 723 724u_long rip_sendspace = 9216; 725u_long rip_recvspace = 9216; 726 727SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW, 728 &rip_sendspace, 0, "Maximum outgoing raw IP datagram size"); 729SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW, 730 &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams"); 731 732static int 733rip_attach(struct socket *so, int proto, struct thread *td) 734{ 735 struct inpcb *inp; 736 int error; 737 738 inp = sotoinpcb(so); 739 KASSERT(inp == NULL, ("rip_attach: inp != NULL")); 740 741 error = priv_check(td, PRIV_NETINET_RAW); 742 if (error) 743 return (error); 744 if (proto >= IPPROTO_MAX || proto < 0) 745 return EPROTONOSUPPORT; 746 error = soreserve(so, rip_sendspace, rip_recvspace); 747 if (error) 748 return (error); 749 INP_INFO_WLOCK(&V_ripcbinfo); 750 error = in_pcballoc(so, &V_ripcbinfo); 751 if (error) { 752 INP_INFO_WUNLOCK(&V_ripcbinfo); 753 return (error); 754 } 755 inp = (struct inpcb *)so->so_pcb; 756 inp->inp_vflag |= INP_IPV4; 757 inp->inp_ip_p = proto; 758 inp->inp_ip_ttl = V_ip_defttl; 759 rip_inshash(inp); 760 INP_INFO_WUNLOCK(&V_ripcbinfo); 761 INP_WUNLOCK(inp); 762 return (0); 763} 764 765static void 766rip_detach(struct socket *so) 767{ 768 struct inpcb *inp; 769 770 inp = sotoinpcb(so); 771 KASSERT(inp != NULL, ("rip_detach: inp == NULL")); 772 KASSERT(inp->inp_faddr.s_addr == INADDR_ANY, 773 ("rip_detach: not closed")); 774 775 INP_INFO_WLOCK(&V_ripcbinfo); 776 INP_WLOCK(inp); 777 rip_delhash(inp); 778 if (so == V_ip_mrouter && ip_mrouter_done) 779 ip_mrouter_done(); 780 if (ip_rsvp_force_done) 781 ip_rsvp_force_done(so); 782 if (so == V_ip_rsvpd) 783 ip_rsvp_done(); 784 in_pcbdetach(inp); 785 in_pcbfree(inp); 786 INP_INFO_WUNLOCK(&V_ripcbinfo); 787} 788 789static void 790rip_dodisconnect(struct socket *so, struct inpcb *inp) 791{ 792 793 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 794 INP_WLOCK_ASSERT(inp); 795 796 rip_delhash(inp); 797 inp->inp_faddr.s_addr = INADDR_ANY; 798 rip_inshash(inp); 799 SOCK_LOCK(so); 800 so->so_state &= ~SS_ISCONNECTED; 801 SOCK_UNLOCK(so); 802} 803 804static void 805rip_abort(struct socket *so) 806{ 807 struct inpcb *inp; 808 809 inp = sotoinpcb(so); 810 KASSERT(inp != NULL, ("rip_abort: inp == NULL")); 811 812 INP_INFO_WLOCK(&V_ripcbinfo); 813 INP_WLOCK(inp); 814 rip_dodisconnect(so, inp); 815 INP_WUNLOCK(inp); 816 INP_INFO_WUNLOCK(&V_ripcbinfo); 817} 818 819static void 820rip_close(struct socket *so) 821{ 822 struct inpcb *inp; 823 824 inp = sotoinpcb(so); 825 KASSERT(inp != NULL, ("rip_close: inp == NULL")); 826 827 INP_INFO_WLOCK(&V_ripcbinfo); 828 INP_WLOCK(inp); 829 rip_dodisconnect(so, inp); 830 INP_WUNLOCK(inp); 831 INP_INFO_WUNLOCK(&V_ripcbinfo); 832} 833 834static int 835rip_disconnect(struct socket *so) 836{ 837 struct inpcb *inp; 838 839 if ((so->so_state & SS_ISCONNECTED) == 0) 840 return (ENOTCONN); 841 842 inp = sotoinpcb(so); 843 KASSERT(inp != NULL, ("rip_disconnect: 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 return (0); 851} 852 853static int 854rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 855{ 856 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 857 struct inpcb *inp; 858 int error; 859 860 if (nam->sa_len != sizeof(*addr)) 861 return (EINVAL); 862 863 error = prison_check_ip4(td->td_ucred, &addr->sin_addr); 864 if (error != 0) 865 return (error); 866 867 inp = sotoinpcb(so); 868 KASSERT(inp != NULL, ("rip_bind: inp == NULL")); 869 870 if (TAILQ_EMPTY(&V_ifnet) || 871 (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) || 872 (addr->sin_addr.s_addr && 873 (inp->inp_flags & INP_BINDANY) == 0 && 874 ifa_ifwithaddr_check((struct sockaddr *)addr) == 0)) 875 return (EADDRNOTAVAIL); 876 877 INP_INFO_WLOCK(&V_ripcbinfo); 878 INP_WLOCK(inp); 879 rip_delhash(inp); 880 inp->inp_laddr = addr->sin_addr; 881 rip_inshash(inp); 882 INP_WUNLOCK(inp); 883 INP_INFO_WUNLOCK(&V_ripcbinfo); 884 return (0); 885} 886 887static int 888rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 889{ 890 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 891 struct inpcb *inp; 892 893 if (nam->sa_len != sizeof(*addr)) 894 return (EINVAL); 895 if (TAILQ_EMPTY(&V_ifnet)) 896 return (EADDRNOTAVAIL); 897 if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) 898 return (EAFNOSUPPORT); 899 900 inp = sotoinpcb(so); 901 KASSERT(inp != NULL, ("rip_connect: inp == NULL")); 902 903 INP_INFO_WLOCK(&V_ripcbinfo); 904 INP_WLOCK(inp); 905 rip_delhash(inp); 906 inp->inp_faddr = addr->sin_addr; 907 rip_inshash(inp); 908 soisconnected(so); 909 INP_WUNLOCK(inp); 910 INP_INFO_WUNLOCK(&V_ripcbinfo); 911 return (0); 912} 913 914static int 915rip_shutdown(struct socket *so) 916{ 917 struct inpcb *inp; 918 919 inp = sotoinpcb(so); 920 KASSERT(inp != NULL, ("rip_shutdown: inp == NULL")); 921 922 INP_WLOCK(inp); 923 socantsendmore(so); 924 INP_WUNLOCK(inp); 925 return (0); 926} 927 928static int 929rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 930 struct mbuf *control, struct thread *td) 931{ 932 struct inpcb *inp; 933 u_long dst; 934 935 inp = sotoinpcb(so); 936 KASSERT(inp != NULL, ("rip_send: inp == NULL")); 937 938 /* 939 * Note: 'dst' reads below are unlocked. 940 */ 941 if (so->so_state & SS_ISCONNECTED) { 942 if (nam) { 943 m_freem(m); 944 return (EISCONN); 945 } 946 dst = inp->inp_faddr.s_addr; /* Unlocked read. */ 947 } else { 948 if (nam == NULL) { 949 m_freem(m); 950 return (ENOTCONN); 951 } 952 dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr; 953 } 954 return (rip_output(m, so, dst)); 955} 956 957static int 958rip_pcblist(SYSCTL_HANDLER_ARGS) 959{ 960 int error, i, n; 961 struct inpcb *inp, **inp_list; 962 inp_gen_t gencnt; 963 struct xinpgen xig; 964 965 /* 966 * The process of preparing the TCB list is too time-consuming and 967 * resource-intensive to repeat twice on every request. 968 */ 969 if (req->oldptr == 0) { 970 n = V_ripcbinfo.ipi_count; 971 req->oldidx = 2 * (sizeof xig) 972 + (n + n/8) * sizeof(struct xinpcb); 973 return (0); 974 } 975 976 if (req->newptr != 0) 977 return (EPERM); 978 979 /* 980 * OK, now we're committed to doing something. 981 */ 982 INP_INFO_RLOCK(&V_ripcbinfo); 983 gencnt = V_ripcbinfo.ipi_gencnt; 984 n = V_ripcbinfo.ipi_count; 985 INP_INFO_RUNLOCK(&V_ripcbinfo); 986 987 xig.xig_len = sizeof xig; 988 xig.xig_count = n; 989 xig.xig_gen = gencnt; 990 xig.xig_sogen = so_gencnt; 991 error = SYSCTL_OUT(req, &xig, sizeof xig); 992 if (error) 993 return (error); 994 995 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 996 if (inp_list == 0) 997 return (ENOMEM); 998 999 INP_INFO_RLOCK(&V_ripcbinfo); 1000 for (inp = LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n; 1001 inp = LIST_NEXT(inp, inp_list)) { 1002 INP_RLOCK(inp); 1003 if (inp->inp_gencnt <= gencnt && 1004 cr_canseeinpcb(req->td->td_ucred, inp) == 0) { 1005 /* XXX held references? */ 1006 inp_list[i++] = inp; 1007 } 1008 INP_RUNLOCK(inp); 1009 } 1010 INP_INFO_RUNLOCK(&V_ripcbinfo); 1011 n = i; 1012 1013 error = 0; 1014 for (i = 0; i < n; i++) { 1015 inp = inp_list[i]; 1016 INP_RLOCK(inp); 1017 if (inp->inp_gencnt <= gencnt) { 1018 struct xinpcb xi; 1019 1020 bzero(&xi, sizeof(xi)); 1021 xi.xi_len = sizeof xi; 1022 /* XXX should avoid extra copy */ 1023 bcopy(inp, &xi.xi_inp, sizeof *inp); 1024 if (inp->inp_socket) 1025 sotoxsocket(inp->inp_socket, &xi.xi_socket); 1026 INP_RUNLOCK(inp); 1027 error = SYSCTL_OUT(req, &xi, sizeof xi); 1028 } else 1029 INP_RUNLOCK(inp); 1030 } 1031 if (!error) { 1032 /* 1033 * Give the user an updated idea of our state. If the 1034 * generation differs from what we told her before, she knows 1035 * that something happened while we were processing this 1036 * request, and it might be necessary to retry. 1037 */ 1038 INP_INFO_RLOCK(&V_ripcbinfo); 1039 xig.xig_gen = V_ripcbinfo.ipi_gencnt; 1040 xig.xig_sogen = so_gencnt; 1041 xig.xig_count = V_ripcbinfo.ipi_count; 1042 INP_INFO_RUNLOCK(&V_ripcbinfo); 1043 error = SYSCTL_OUT(req, &xig, sizeof xig); 1044 } 1045 free(inp_list, M_TEMP); 1046 return (error); 1047} 1048 1049SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0, 1050 rip_pcblist, "S,xinpcb", "List of active raw IP sockets"); 1051 1052struct pr_usrreqs rip_usrreqs = { 1053 .pru_abort = rip_abort, 1054 .pru_attach = rip_attach, 1055 .pru_bind = rip_bind, 1056 .pru_connect = rip_connect, 1057 .pru_control = in_control, 1058 .pru_detach = rip_detach, 1059 .pru_disconnect = rip_disconnect, 1060 .pru_peeraddr = in_getpeeraddr, 1061 .pru_send = rip_send, 1062 .pru_shutdown = rip_shutdown, 1063 .pru_sockaddr = in_getsockaddr, 1064 .pru_sosetlabel = in_pcbsosetlabel, 1065 .pru_close = rip_close, 1066}; 1067