1 /*- 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)raw_ip.c 8.7 (Berkeley) 5/15/95 30 * $FreeBSD$ 31 */ 32 33 #include "opt_inet6.h" 34 #include "opt_ipsec.h" 35 #include "opt_mac.h" 36 37 #include <sys/param.h> 38 #include <sys/jail.h> 39 #include <sys/kernel.h> 40 #include <sys/lock.h> 41 #include <sys/malloc.h> 42 #include <sys/mbuf.h> 43 #include <sys/priv.h> 44 #include <sys/proc.h> 45 #include <sys/protosw.h> 46 #include <sys/signalvar.h> 47 #include <sys/socket.h> 48 #include <sys/socketvar.h> 49 #include <sys/sx.h> 50 #include <sys/sysctl.h> 51 #include <sys/systm.h> 52 53 #include <vm/uma.h> 54 55 #include <net/if.h> 56 #include <net/route.h> 57 58 #include <netinet/in.h> 59 #include <netinet/in_systm.h> 60 #include <netinet/in_pcb.h> 61 #include <netinet/in_var.h> 62 #include <netinet/ip.h> 63 #include <netinet/ip_var.h> 64 #include <netinet/ip_mroute.h> 65 66 #include <netinet/ip_fw.h> 67 #include <netinet/ip_dummynet.h> 68 69 #ifdef FAST_IPSEC 70 #include <netipsec/ipsec.h> 71 #endif /*FAST_IPSEC*/ 72 73 #ifdef IPSEC 74 #include <netinet6/ipsec.h> 75 #endif /*IPSEC*/ 76 77 #include <security/mac/mac_framework.h> 78 79 struct inpcbhead ripcb; 80 struct inpcbinfo ripcbinfo; 81 82 /* control hooks for ipfw and dummynet */ 83 ip_fw_ctl_t *ip_fw_ctl_ptr = NULL; 84 ip_dn_ctl_t *ip_dn_ctl_ptr = NULL; 85 86 /* 87 * hooks for multicast routing. They all default to NULL, 88 * so leave them not initialized and rely on BSS being set to 0. 89 */ 90 91 /* The socket used to communicate with the multicast routing daemon. */ 92 struct socket *ip_mrouter; 93 94 /* The various mrouter and rsvp functions */ 95 int (*ip_mrouter_set)(struct socket *, struct sockopt *); 96 int (*ip_mrouter_get)(struct socket *, struct sockopt *); 97 int (*ip_mrouter_done)(void); 98 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 99 struct ip_moptions *); 100 int (*mrt_ioctl)(int, caddr_t); 101 int (*legal_vif_num)(int); 102 u_long (*ip_mcast_src)(int); 103 104 void (*rsvp_input_p)(struct mbuf *m, int off); 105 int (*ip_rsvp_vif)(struct socket *, struct sockopt *); 106 void (*ip_rsvp_force_done)(struct socket *); 107 108 /* 109 * Nominal space allocated to a raw ip socket. 110 */ 111 #define RIPSNDQ 8192 112 #define RIPRCVQ 8192 113 114 /* 115 * Raw interface to IP protocol. 116 */ 117 118 /* 119 * Initialize raw connection block q. 120 */ 121 static void 122 rip_zone_change(void *tag) 123 { 124 125 uma_zone_set_max(ripcbinfo.ipi_zone, maxsockets); 126 } 127 128 static int 129 rip_inpcb_init(void *mem, int size, int flags) 130 { 131 struct inpcb *inp = (struct inpcb *) mem; 132 INP_LOCK_INIT(inp, "inp", "rawinp"); 133 return (0); 134 } 135 136 void 137 rip_init() 138 { 139 INP_INFO_LOCK_INIT(&ripcbinfo, "rip"); 140 LIST_INIT(&ripcb); 141 ripcbinfo.listhead = &ripcb; 142 /* 143 * XXX We don't use the hash list for raw IP, but it's easier 144 * to allocate a one entry hash list than it is to check all 145 * over the place for hashbase == NULL. 146 */ 147 ripcbinfo.hashbase = hashinit(1, M_PCB, &ripcbinfo.hashmask); 148 ripcbinfo.porthashbase = hashinit(1, M_PCB, &ripcbinfo.porthashmask); 149 ripcbinfo.ipi_zone = uma_zcreate("ripcb", sizeof(struct inpcb), 150 NULL, NULL, rip_inpcb_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 151 uma_zone_set_max(ripcbinfo.ipi_zone, maxsockets); 152 EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, 153 NULL, EVENTHANDLER_PRI_ANY); 154 } 155 156 static struct sockaddr_in ripsrc = { sizeof(ripsrc), AF_INET }; 157 158 static int 159 raw_append(struct inpcb *last, struct ip *ip, struct mbuf *n) 160 { 161 int policyfail = 0; 162 163 INP_LOCK_ASSERT(last); 164 165 #if defined(IPSEC) || defined(FAST_IPSEC) 166 /* check AH/ESP integrity. */ 167 if (ipsec4_in_reject(n, last)) { 168 policyfail = 1; 169 #ifdef IPSEC 170 ipsecstat.in_polvio++; 171 #endif /*IPSEC*/ 172 /* do not inject data to pcb */ 173 } 174 #endif /*IPSEC || FAST_IPSEC*/ 175 #ifdef MAC 176 if (!policyfail && mac_check_inpcb_deliver(last, n) != 0) 177 policyfail = 1; 178 #endif 179 /* Check the minimum TTL for socket. */ 180 if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl) 181 policyfail = 1; 182 if (!policyfail) { 183 struct mbuf *opts = NULL; 184 struct socket *so; 185 186 so = last->inp_socket; 187 if ((last->inp_flags & INP_CONTROLOPTS) || 188 (so->so_options & (SO_TIMESTAMP | SO_BINTIME))) 189 ip_savecontrol(last, &opts, ip, n); 190 SOCKBUF_LOCK(&so->so_rcv); 191 if (sbappendaddr_locked(&so->so_rcv, 192 (struct sockaddr *)&ripsrc, n, opts) == 0) { 193 /* should notify about lost packet */ 194 m_freem(n); 195 if (opts) 196 m_freem(opts); 197 SOCKBUF_UNLOCK(&so->so_rcv); 198 } else 199 sorwakeup_locked(so); 200 } else 201 m_freem(n); 202 return policyfail; 203 } 204 205 /* 206 * Setup generic address and protocol structures 207 * for raw_input routine, then pass them along with 208 * mbuf chain. 209 */ 210 void 211 rip_input(struct mbuf *m, int off) 212 { 213 struct ip *ip = mtod(m, struct ip *); 214 int proto = ip->ip_p; 215 struct inpcb *inp, *last; 216 217 INP_INFO_RLOCK(&ripcbinfo); 218 ripsrc.sin_addr = ip->ip_src; 219 last = NULL; 220 LIST_FOREACH(inp, &ripcb, inp_list) { 221 INP_LOCK(inp); 222 if (inp->inp_ip_p && inp->inp_ip_p != proto) { 223 docontinue: 224 INP_UNLOCK(inp); 225 continue; 226 } 227 #ifdef INET6 228 if ((inp->inp_vflag & INP_IPV4) == 0) 229 goto docontinue; 230 #endif 231 if (inp->inp_laddr.s_addr && 232 inp->inp_laddr.s_addr != ip->ip_dst.s_addr) 233 goto docontinue; 234 if (inp->inp_faddr.s_addr && 235 inp->inp_faddr.s_addr != ip->ip_src.s_addr) 236 goto docontinue; 237 if (jailed(inp->inp_socket->so_cred)) 238 if (htonl(prison_getip(inp->inp_socket->so_cred)) != 239 ip->ip_dst.s_addr) 240 goto docontinue; 241 if (last) { 242 struct mbuf *n; 243 244 n = m_copy(m, 0, (int)M_COPYALL); 245 if (n != NULL) 246 (void) raw_append(last, ip, n); 247 /* XXX count dropped packet */ 248 INP_UNLOCK(last); 249 } 250 last = inp; 251 } 252 if (last != NULL) { 253 if (raw_append(last, ip, m) != 0) 254 ipstat.ips_delivered--; 255 INP_UNLOCK(last); 256 } else { 257 m_freem(m); 258 ipstat.ips_noproto++; 259 ipstat.ips_delivered--; 260 } 261 INP_INFO_RUNLOCK(&ripcbinfo); 262 } 263 264 /* 265 * Generate IP header and pass packet to ip_output. 266 * Tack on options user may have setup with control call. 267 */ 268 int 269 rip_output(struct mbuf *m, struct socket *so, u_long dst) 270 { 271 struct ip *ip; 272 int error; 273 struct inpcb *inp = sotoinpcb(so); 274 int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) | 275 IP_ALLOWBROADCAST; 276 277 /* 278 * If the user handed us a complete IP packet, use it. 279 * Otherwise, allocate an mbuf for a header and fill it in. 280 */ 281 if ((inp->inp_flags & INP_HDRINCL) == 0) { 282 if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) { 283 m_freem(m); 284 return(EMSGSIZE); 285 } 286 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT); 287 if (m == NULL) 288 return(ENOBUFS); 289 290 INP_LOCK(inp); 291 ip = mtod(m, struct ip *); 292 ip->ip_tos = inp->inp_ip_tos; 293 if (inp->inp_flags & INP_DONTFRAG) 294 ip->ip_off = IP_DF; 295 else 296 ip->ip_off = 0; 297 ip->ip_p = inp->inp_ip_p; 298 ip->ip_len = m->m_pkthdr.len; 299 if (jailed(inp->inp_socket->so_cred)) 300 ip->ip_src.s_addr = 301 htonl(prison_getip(inp->inp_socket->so_cred)); 302 else 303 ip->ip_src = inp->inp_laddr; 304 ip->ip_dst.s_addr = dst; 305 ip->ip_ttl = inp->inp_ip_ttl; 306 } else { 307 if (m->m_pkthdr.len > IP_MAXPACKET) { 308 m_freem(m); 309 return(EMSGSIZE); 310 } 311 INP_LOCK(inp); 312 ip = mtod(m, struct ip *); 313 if (jailed(inp->inp_socket->so_cred)) { 314 if (ip->ip_src.s_addr != 315 htonl(prison_getip(inp->inp_socket->so_cred))) { 316 INP_UNLOCK(inp); 317 m_freem(m); 318 return (EPERM); 319 } 320 } 321 /* don't allow both user specified and setsockopt options, 322 and don't allow packet length sizes that will crash */ 323 if (((ip->ip_hl != (sizeof (*ip) >> 2)) 324 && inp->inp_options) 325 || (ip->ip_len > m->m_pkthdr.len) 326 || (ip->ip_len < (ip->ip_hl << 2))) { 327 INP_UNLOCK(inp); 328 m_freem(m); 329 return EINVAL; 330 } 331 if (ip->ip_id == 0) 332 ip->ip_id = ip_newid(); 333 /* XXX prevent ip_output from overwriting header fields */ 334 flags |= IP_RAWOUTPUT; 335 ipstat.ips_rawout++; 336 } 337 338 if (inp->inp_flags & INP_ONESBCAST) 339 flags |= IP_SENDONES; 340 341 #ifdef MAC 342 mac_create_mbuf_from_inpcb(inp, m); 343 #endif 344 345 error = ip_output(m, inp->inp_options, NULL, flags, 346 inp->inp_moptions, inp); 347 INP_UNLOCK(inp); 348 return error; 349 } 350 351 /* 352 * Raw IP socket option processing. 353 * 354 * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could 355 * only be created by a privileged process, and as such, socket option 356 * operations to manage system properties on any raw socket were allowed to 357 * take place without explicit additional access control checks. However, 358 * raw sockets can now also be created in jail(), and therefore explicit 359 * checks are now required. Likewise, raw sockets can be used by a process 360 * after it gives up privilege, so some caution is required. For options 361 * passed down to the IP layer via ip_ctloutput(), checks are assumed to be 362 * performed in ip_ctloutput() and therefore no check occurs here. 363 * Unilaterally checking suser() here breaks normal IP socket option 364 * operations on raw sockets. 365 * 366 * When adding new socket options here, make sure to add access control 367 * checks here as necessary. 368 */ 369 int 370 rip_ctloutput(struct socket *so, struct sockopt *sopt) 371 { 372 struct inpcb *inp = sotoinpcb(so); 373 int error, optval; 374 375 if (sopt->sopt_level != IPPROTO_IP) 376 return (EINVAL); 377 378 error = 0; 379 switch (sopt->sopt_dir) { 380 case SOPT_GET: 381 switch (sopt->sopt_name) { 382 case IP_HDRINCL: 383 optval = inp->inp_flags & INP_HDRINCL; 384 error = sooptcopyout(sopt, &optval, sizeof optval); 385 break; 386 387 case IP_FW_ADD: /* ADD actually returns the body... */ 388 case IP_FW_GET: 389 case IP_FW_TABLE_GETSIZE: 390 case IP_FW_TABLE_LIST: 391 /* 392 * XXXRW: Isn't this checked one layer down? Yes, it 393 * is. 394 */ 395 error = priv_check(curthread, PRIV_NETINET_IPFW); 396 if (error != 0) 397 return (error); 398 if (ip_fw_ctl_ptr != NULL) 399 error = ip_fw_ctl_ptr(sopt); 400 else 401 error = ENOPROTOOPT; 402 break; 403 404 case IP_DUMMYNET_GET: 405 error = priv_check(curthread, PRIV_NETINET_DUMMYNET); 406 if (error != 0) 407 return (error); 408 if (ip_dn_ctl_ptr != NULL) 409 error = ip_dn_ctl_ptr(sopt); 410 else 411 error = ENOPROTOOPT; 412 break ; 413 414 case MRT_INIT: 415 case MRT_DONE: 416 case MRT_ADD_VIF: 417 case MRT_DEL_VIF: 418 case MRT_ADD_MFC: 419 case MRT_DEL_MFC: 420 case MRT_VERSION: 421 case MRT_ASSERT: 422 case MRT_API_SUPPORT: 423 case MRT_API_CONFIG: 424 case MRT_ADD_BW_UPCALL: 425 case MRT_DEL_BW_UPCALL: 426 error = priv_check(curthread, PRIV_NETINET_MROUTE); 427 if (error != 0) 428 return (error); 429 error = ip_mrouter_get ? ip_mrouter_get(so, sopt) : 430 EOPNOTSUPP; 431 break; 432 433 default: 434 error = ip_ctloutput(so, sopt); 435 break; 436 } 437 break; 438 439 case SOPT_SET: 440 switch (sopt->sopt_name) { 441 case IP_HDRINCL: 442 error = sooptcopyin(sopt, &optval, sizeof optval, 443 sizeof optval); 444 if (error) 445 break; 446 if (optval) 447 inp->inp_flags |= INP_HDRINCL; 448 else 449 inp->inp_flags &= ~INP_HDRINCL; 450 break; 451 452 case IP_FW_ADD: 453 case IP_FW_DEL: 454 case IP_FW_FLUSH: 455 case IP_FW_ZERO: 456 case IP_FW_RESETLOG: 457 case IP_FW_TABLE_ADD: 458 case IP_FW_TABLE_DEL: 459 case IP_FW_TABLE_FLUSH: 460 /* 461 * XXXRW: Isn't this checked one layer down? 462 */ 463 error = priv_check(curthread, PRIV_NETINET_IPFW); 464 if (error != 0) 465 return (error); 466 if (ip_fw_ctl_ptr != NULL) 467 error = ip_fw_ctl_ptr(sopt); 468 else 469 error = ENOPROTOOPT; 470 break; 471 472 case IP_DUMMYNET_CONFIGURE: 473 case IP_DUMMYNET_DEL: 474 case IP_DUMMYNET_FLUSH: 475 error = priv_check(curthread, PRIV_NETINET_DUMMYNET); 476 if (error != 0) 477 return (error); 478 if (ip_dn_ctl_ptr != NULL) 479 error = ip_dn_ctl_ptr(sopt); 480 else 481 error = ENOPROTOOPT ; 482 break ; 483 484 case IP_RSVP_ON: 485 error = priv_check(curthread, PRIV_NETINET_MROUTE); 486 if (error != 0) 487 return (error); 488 error = ip_rsvp_init(so); 489 break; 490 491 case IP_RSVP_OFF: 492 error = priv_check(curthread, PRIV_NETINET_MROUTE); 493 if (error != 0) 494 return (error); 495 error = ip_rsvp_done(); 496 break; 497 498 case IP_RSVP_VIF_ON: 499 case IP_RSVP_VIF_OFF: 500 error = priv_check(curthread, PRIV_NETINET_MROUTE); 501 if (error != 0) 502 return (error); 503 error = ip_rsvp_vif ? 504 ip_rsvp_vif(so, sopt) : EINVAL; 505 break; 506 507 case MRT_INIT: 508 case MRT_DONE: 509 case MRT_ADD_VIF: 510 case MRT_DEL_VIF: 511 case MRT_ADD_MFC: 512 case MRT_DEL_MFC: 513 case MRT_VERSION: 514 case MRT_ASSERT: 515 case MRT_API_SUPPORT: 516 case MRT_API_CONFIG: 517 case MRT_ADD_BW_UPCALL: 518 case MRT_DEL_BW_UPCALL: 519 error = priv_check(curthread, PRIV_NETINET_MROUTE); 520 if (error != 0) 521 return (error); 522 error = ip_mrouter_set ? ip_mrouter_set(so, sopt) : 523 EOPNOTSUPP; 524 break; 525 526 default: 527 error = ip_ctloutput(so, sopt); 528 break; 529 } 530 break; 531 } 532 533 return (error); 534 } 535 536 /* 537 * This function exists solely to receive the PRC_IFDOWN messages which 538 * are sent by if_down(). It looks for an ifaddr whose ifa_addr is sa, 539 * and calls in_ifadown() to remove all routes corresponding to that address. 540 * It also receives the PRC_IFUP messages from if_up() and reinstalls the 541 * interface routes. 542 */ 543 void 544 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip) 545 { 546 struct in_ifaddr *ia; 547 struct ifnet *ifp; 548 int err; 549 int flags; 550 551 switch (cmd) { 552 case PRC_IFDOWN: 553 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) { 554 if (ia->ia_ifa.ifa_addr == sa 555 && (ia->ia_flags & IFA_ROUTE)) { 556 /* 557 * in_ifscrub kills the interface route. 558 */ 559 in_ifscrub(ia->ia_ifp, ia); 560 /* 561 * in_ifadown gets rid of all the rest of 562 * the routes. This is not quite the right 563 * thing to do, but at least if we are running 564 * a routing process they will come back. 565 */ 566 in_ifadown(&ia->ia_ifa, 0); 567 break; 568 } 569 } 570 break; 571 572 case PRC_IFUP: 573 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) { 574 if (ia->ia_ifa.ifa_addr == sa) 575 break; 576 } 577 if (ia == 0 || (ia->ia_flags & IFA_ROUTE)) 578 return; 579 flags = RTF_UP; 580 ifp = ia->ia_ifa.ifa_ifp; 581 582 if ((ifp->if_flags & IFF_LOOPBACK) 583 || (ifp->if_flags & IFF_POINTOPOINT)) 584 flags |= RTF_HOST; 585 586 err = rtinit(&ia->ia_ifa, RTM_ADD, flags); 587 if (err == 0) 588 ia->ia_flags |= IFA_ROUTE; 589 break; 590 } 591 } 592 593 u_long rip_sendspace = RIPSNDQ; 594 u_long rip_recvspace = RIPRCVQ; 595 596 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW, 597 &rip_sendspace, 0, "Maximum outgoing raw IP datagram size"); 598 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW, 599 &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams"); 600 601 static int 602 rip_attach(struct socket *so, int proto, struct thread *td) 603 { 604 struct inpcb *inp; 605 int error; 606 607 inp = sotoinpcb(so); 608 KASSERT(inp == NULL, ("rip_attach: inp != NULL")); 609 /* 610 * XXXRW: Centralize privilege decision in kern_jail.c. 611 */ 612 if (jailed(td->td_ucred) && !jail_allow_raw_sockets) 613 return (EPERM); 614 error = priv_check_cred(td->td_ucred, PRIV_NETINET_RAW, 615 SUSER_ALLOWJAIL); 616 if (error) 617 return error; 618 if (proto >= IPPROTO_MAX || proto < 0) 619 return EPROTONOSUPPORT; 620 error = soreserve(so, rip_sendspace, rip_recvspace); 621 if (error) 622 return error; 623 INP_INFO_WLOCK(&ripcbinfo); 624 error = in_pcballoc(so, &ripcbinfo); 625 if (error) { 626 INP_INFO_WUNLOCK(&ripcbinfo); 627 return error; 628 } 629 inp = (struct inpcb *)so->so_pcb; 630 INP_INFO_WUNLOCK(&ripcbinfo); 631 inp->inp_vflag |= INP_IPV4; 632 inp->inp_ip_p = proto; 633 inp->inp_ip_ttl = ip_defttl; 634 INP_UNLOCK(inp); 635 return 0; 636 } 637 638 static void 639 rip_detach(struct socket *so) 640 { 641 struct inpcb *inp; 642 643 inp = sotoinpcb(so); 644 KASSERT(inp != NULL, ("rip_detach: inp == NULL")); 645 KASSERT(inp->inp_faddr.s_addr == INADDR_ANY, 646 ("rip_detach: not closed")); 647 648 INP_INFO_WLOCK(&ripcbinfo); 649 INP_LOCK(inp); 650 if (so == ip_mrouter && ip_mrouter_done) 651 ip_mrouter_done(); 652 if (ip_rsvp_force_done) 653 ip_rsvp_force_done(so); 654 if (so == ip_rsvpd) 655 ip_rsvp_done(); 656 in_pcbdetach(inp); 657 in_pcbfree(inp); 658 INP_INFO_WUNLOCK(&ripcbinfo); 659 } 660 661 static void 662 rip_dodisconnect(struct socket *so, struct inpcb *inp) 663 { 664 665 INP_LOCK_ASSERT(inp); 666 667 inp->inp_faddr.s_addr = INADDR_ANY; 668 SOCK_LOCK(so); 669 so->so_state &= ~SS_ISCONNECTED; 670 SOCK_UNLOCK(so); 671 } 672 673 static void 674 rip_abort(struct socket *so) 675 { 676 struct inpcb *inp; 677 678 inp = sotoinpcb(so); 679 KASSERT(inp != NULL, ("rip_abort: inp == NULL")); 680 681 INP_INFO_WLOCK(&ripcbinfo); 682 INP_LOCK(inp); 683 rip_dodisconnect(so, inp); 684 INP_UNLOCK(inp); 685 INP_INFO_WUNLOCK(&ripcbinfo); 686 } 687 688 static void 689 rip_close(struct socket *so) 690 { 691 struct inpcb *inp; 692 693 inp = sotoinpcb(so); 694 KASSERT(inp != NULL, ("rip_close: inp == NULL")); 695 696 INP_INFO_WLOCK(&ripcbinfo); 697 INP_LOCK(inp); 698 rip_dodisconnect(so, inp); 699 INP_UNLOCK(inp); 700 INP_INFO_WUNLOCK(&ripcbinfo); 701 } 702 703 static int 704 rip_disconnect(struct socket *so) 705 { 706 struct inpcb *inp; 707 708 if ((so->so_state & SS_ISCONNECTED) == 0) 709 return ENOTCONN; 710 711 inp = sotoinpcb(so); 712 KASSERT(inp != NULL, ("rip_disconnect: inp == NULL")); 713 INP_INFO_WLOCK(&ripcbinfo); 714 INP_LOCK(inp); 715 rip_dodisconnect(so, inp); 716 INP_UNLOCK(inp); 717 INP_INFO_WUNLOCK(&ripcbinfo); 718 return (0); 719 } 720 721 static int 722 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 723 { 724 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 725 struct inpcb *inp; 726 727 if (nam->sa_len != sizeof(*addr)) 728 return EINVAL; 729 730 if (jailed(td->td_ucred)) { 731 if (addr->sin_addr.s_addr == INADDR_ANY) 732 addr->sin_addr.s_addr = 733 htonl(prison_getip(td->td_ucred)); 734 if (htonl(prison_getip(td->td_ucred)) != addr->sin_addr.s_addr) 735 return (EADDRNOTAVAIL); 736 } 737 738 if (TAILQ_EMPTY(&ifnet) || 739 (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) || 740 (addr->sin_addr.s_addr && 741 ifa_ifwithaddr((struct sockaddr *)addr) == 0)) 742 return EADDRNOTAVAIL; 743 744 inp = sotoinpcb(so); 745 KASSERT(inp != NULL, ("rip_bind: inp == NULL")); 746 INP_INFO_WLOCK(&ripcbinfo); 747 INP_LOCK(inp); 748 inp->inp_laddr = addr->sin_addr; 749 INP_UNLOCK(inp); 750 INP_INFO_WUNLOCK(&ripcbinfo); 751 return 0; 752 } 753 754 static int 755 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 756 { 757 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 758 struct inpcb *inp; 759 760 if (nam->sa_len != sizeof(*addr)) 761 return EINVAL; 762 if (TAILQ_EMPTY(&ifnet)) 763 return EADDRNOTAVAIL; 764 if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) 765 return EAFNOSUPPORT; 766 767 inp = sotoinpcb(so); 768 KASSERT(inp != NULL, ("rip_connect: inp == NULL")); 769 INP_INFO_WLOCK(&ripcbinfo); 770 INP_LOCK(inp); 771 inp->inp_faddr = addr->sin_addr; 772 soisconnected(so); 773 INP_UNLOCK(inp); 774 INP_INFO_WUNLOCK(&ripcbinfo); 775 return 0; 776 } 777 778 static int 779 rip_shutdown(struct socket *so) 780 { 781 struct inpcb *inp; 782 783 inp = sotoinpcb(so); 784 KASSERT(inp != NULL, ("rip_shutdown: inp == NULL")); 785 INP_LOCK(inp); 786 socantsendmore(so); 787 INP_UNLOCK(inp); 788 return 0; 789 } 790 791 static int 792 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 793 struct mbuf *control, struct thread *td) 794 { 795 struct inpcb *inp; 796 u_long dst; 797 798 inp = sotoinpcb(so); 799 KASSERT(inp != NULL, ("rip_send: inp == NULL")); 800 /* 801 * Note: 'dst' reads below are unlocked. 802 */ 803 if (so->so_state & SS_ISCONNECTED) { 804 if (nam) { 805 m_freem(m); 806 return EISCONN; 807 } 808 dst = inp->inp_faddr.s_addr; /* Unlocked read. */ 809 } else { 810 if (nam == NULL) { 811 m_freem(m); 812 return ENOTCONN; 813 } 814 dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr; 815 } 816 return rip_output(m, so, dst); 817 } 818 819 static int 820 rip_pcblist(SYSCTL_HANDLER_ARGS) 821 { 822 int error, i, n; 823 struct inpcb *inp, **inp_list; 824 inp_gen_t gencnt; 825 struct xinpgen xig; 826 827 /* 828 * The process of preparing the TCB list is too time-consuming and 829 * resource-intensive to repeat twice on every request. 830 */ 831 if (req->oldptr == 0) { 832 n = ripcbinfo.ipi_count; 833 req->oldidx = 2 * (sizeof xig) 834 + (n + n/8) * sizeof(struct xinpcb); 835 return 0; 836 } 837 838 if (req->newptr != 0) 839 return EPERM; 840 841 /* 842 * OK, now we're committed to doing something. 843 */ 844 INP_INFO_RLOCK(&ripcbinfo); 845 gencnt = ripcbinfo.ipi_gencnt; 846 n = ripcbinfo.ipi_count; 847 INP_INFO_RUNLOCK(&ripcbinfo); 848 849 xig.xig_len = sizeof xig; 850 xig.xig_count = n; 851 xig.xig_gen = gencnt; 852 xig.xig_sogen = so_gencnt; 853 error = SYSCTL_OUT(req, &xig, sizeof xig); 854 if (error) 855 return error; 856 857 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 858 if (inp_list == 0) 859 return ENOMEM; 860 861 INP_INFO_RLOCK(&ripcbinfo); 862 for (inp = LIST_FIRST(ripcbinfo.listhead), i = 0; inp && i < n; 863 inp = LIST_NEXT(inp, inp_list)) { 864 INP_LOCK(inp); 865 if (inp->inp_gencnt <= gencnt && 866 cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0) { 867 /* XXX held references? */ 868 inp_list[i++] = inp; 869 } 870 INP_UNLOCK(inp); 871 } 872 INP_INFO_RUNLOCK(&ripcbinfo); 873 n = i; 874 875 error = 0; 876 for (i = 0; i < n; i++) { 877 inp = inp_list[i]; 878 INP_LOCK(inp); 879 if (inp->inp_gencnt <= gencnt) { 880 struct xinpcb xi; 881 bzero(&xi, sizeof(xi)); 882 xi.xi_len = sizeof xi; 883 /* XXX should avoid extra copy */ 884 bcopy(inp, &xi.xi_inp, sizeof *inp); 885 if (inp->inp_socket) 886 sotoxsocket(inp->inp_socket, &xi.xi_socket); 887 INP_UNLOCK(inp); 888 error = SYSCTL_OUT(req, &xi, sizeof xi); 889 } else 890 INP_UNLOCK(inp); 891 } 892 if (!error) { 893 /* 894 * Give the user an updated idea of our state. 895 * If the generation differs from what we told 896 * her before, she knows that something happened 897 * while we were processing this request, and it 898 * might be necessary to retry. 899 */ 900 INP_INFO_RLOCK(&ripcbinfo); 901 xig.xig_gen = ripcbinfo.ipi_gencnt; 902 xig.xig_sogen = so_gencnt; 903 xig.xig_count = ripcbinfo.ipi_count; 904 INP_INFO_RUNLOCK(&ripcbinfo); 905 error = SYSCTL_OUT(req, &xig, sizeof xig); 906 } 907 free(inp_list, M_TEMP); 908 return error; 909 } 910 911 /* 912 * This is the wrapper function for in_setsockaddr. We just pass down 913 * the pcbinfo for in_setpeeraddr to lock. 914 */ 915 static int 916 rip_sockaddr(struct socket *so, struct sockaddr **nam) 917 { 918 return (in_setsockaddr(so, nam, &ripcbinfo)); 919 } 920 921 /* 922 * This is the wrapper function for in_setpeeraddr. We just pass down 923 * the pcbinfo for in_setpeeraddr to lock. 924 */ 925 static int 926 rip_peeraddr(struct socket *so, struct sockaddr **nam) 927 { 928 return (in_setpeeraddr(so, nam, &ripcbinfo)); 929 } 930 931 932 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0, 933 rip_pcblist, "S,xinpcb", "List of active raw IP sockets"); 934 935 struct pr_usrreqs rip_usrreqs = { 936 .pru_abort = rip_abort, 937 .pru_attach = rip_attach, 938 .pru_bind = rip_bind, 939 .pru_connect = rip_connect, 940 .pru_control = in_control, 941 .pru_detach = rip_detach, 942 .pru_disconnect = rip_disconnect, 943 .pru_peeraddr = rip_peeraddr, 944 .pru_send = rip_send, 945 .pru_shutdown = rip_shutdown, 946 .pru_sockaddr = rip_sockaddr, 947 .pru_sosetlabel = in_pcbsosetlabel, 948 .pru_close = rip_close, 949 }; 950