raw_ip.c revision 185895
1263970Sdes/*- 298675Sdes * Copyright (c) 1982, 1986, 1988, 1993 398675Sdes * The Regents of the University of California. 498675Sdes * All rights reserved. 598675Sdes * 698675Sdes * Redistribution and use in source and binary forms, with or without 798675Sdes * modification, are permitted provided that the following conditions 898675Sdes * are met: 998675Sdes * 1. Redistributions of source code must retain the above copyright 1098675Sdes * notice, this list of conditions and the following disclaimer. 1198675Sdes * 2. Redistributions in binary form must reproduce the above copyright 1298675Sdes * notice, this list of conditions and the following disclaimer in the 1398675Sdes * documentation and/or other materials provided with the distribution. 1498675Sdes * 4. Neither the name of the University nor the names of its contributors 1598675Sdes * may be used to endorse or promote products derived from this software 1698675Sdes * without specific prior written permission. 1798675Sdes * 1898675Sdes * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 1998675Sdes * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 2098675Sdes * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 2198675Sdes * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 2298675Sdes * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 2398675Sdes * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 2498675Sdes * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 2598675Sdes * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 2698675Sdes * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 2798675Sdes * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 28162852Sdes * SUCH DAMAGE. 29162852Sdes * 30162852Sdes * @(#)raw_ip.c 8.7 (Berkeley) 5/15/95 31162852Sdes */ 32162852Sdes 3398675Sdes#include <sys/cdefs.h> 3498675Sdes__FBSDID("$FreeBSD: head/sys/netinet/raw_ip.c 185895 2008-12-10 23:12:39Z zec $"); 3598675Sdes 36162852Sdes#include "opt_inet6.h" 37162852Sdes#include "opt_ipsec.h" 3898675Sdes#include "opt_mac.h" 39162852Sdes 40162852Sdes#include <sys/param.h> 41162852Sdes#include <sys/jail.h> 42162852Sdes#include <sys/kernel.h> 4398675Sdes#include <sys/lock.h> 4498675Sdes#include <sys/malloc.h> 4598675Sdes#include <sys/mbuf.h> 4698675Sdes#include <sys/priv.h> 4798675Sdes#include <sys/proc.h> 4898675Sdes#include <sys/protosw.h> 4998675Sdes#include <sys/rwlock.h> 5098675Sdes#include <sys/signalvar.h> 5198675Sdes#include <sys/socket.h> 52126274Sdes#include <sys/socketvar.h> 5398675Sdes#include <sys/sx.h> 5498675Sdes#include <sys/sysctl.h> 55126274Sdes#include <sys/systm.h> 56126274Sdes#include <sys/vimage.h> 5798675Sdes 58126274Sdes#include <vm/uma.h> 59126274Sdes 60126274Sdes#include <net/if.h> 61126274Sdes#include <net/route.h> 62263970Sdes#include <net/vnet.h> 63263970Sdes 64126274Sdes#include <netinet/in.h> 65126274Sdes#include <netinet/in_systm.h> 66126274Sdes#include <netinet/in_pcb.h> 6798675Sdes#include <netinet/in_var.h> 68124208Sdes#include <netinet/ip.h> 69146998Sdes#include <netinet/ip_var.h> 70146998Sdes#include <netinet/ip_mroute.h> 71263970Sdes 72263970Sdes#include <netinet/ip_fw.h> 7398675Sdes#include <netinet/ip_dummynet.h> 7498675Sdes#include <netinet/vinet.h> 7598675Sdes 7698675Sdes#ifdef IPSEC 7798675Sdes#include <netipsec/ipsec.h> 7898675Sdes#endif /*IPSEC*/ 7998675Sdes 8098675Sdes#include <security/mac/mac_framework.h> 81 82#ifdef VIMAGE_GLOBALS 83struct inpcbhead ripcb; 84struct inpcbinfo ripcbinfo; 85#endif 86 87/* control hooks for ipfw and dummynet */ 88ip_fw_ctl_t *ip_fw_ctl_ptr = NULL; 89ip_dn_ctl_t *ip_dn_ctl_ptr = NULL; 90 91/* 92 * Hooks for multicast routing. They all default to NULL, so leave them not 93 * initialized and rely on BSS being set to 0. 94 */ 95 96/* 97 * The socket used to communicate with the multicast routing daemon. 98 */ 99#ifdef VIMAGE_GLOBALS 100struct socket *ip_mrouter; 101#endif 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)(int, 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 INIT_VNET_INET(curvnet); 169 170 uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets); 171} 172 173static int 174rip_inpcb_init(void *mem, int size, int flags) 175{ 176 struct inpcb *inp = mem; 177 178 INP_LOCK_INIT(inp, "inp", "rawinp"); 179 return (0); 180} 181 182void 183rip_init(void) 184{ 185 INIT_VNET_INET(curvnet); 186 187 INP_INFO_LOCK_INIT(&V_ripcbinfo, "rip"); 188 LIST_INIT(&V_ripcb); 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 201static int 202rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n, 203 struct sockaddr_in *ripsrc) 204{ 205 int policyfail = 0; 206 207 INP_RLOCK_ASSERT(last); 208 209#ifdef IPSEC 210 /* check AH/ESP integrity. */ 211 if (ipsec4_in_reject(n, last)) { 212 policyfail = 1; 213 } 214#endif /* IPSEC */ 215#ifdef MAC 216 if (!policyfail && mac_inpcb_check_deliver(last, n) != 0) 217 policyfail = 1; 218#endif 219 /* Check the minimum TTL for socket. */ 220 if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl) 221 policyfail = 1; 222 if (!policyfail) { 223 struct mbuf *opts = NULL; 224 struct socket *so; 225 226 so = last->inp_socket; 227 if ((last->inp_flags & INP_CONTROLOPTS) || 228 (so->so_options & (SO_TIMESTAMP | SO_BINTIME))) 229 ip_savecontrol(last, &opts, ip, n); 230 SOCKBUF_LOCK(&so->so_rcv); 231 if (sbappendaddr_locked(&so->so_rcv, 232 (struct sockaddr *)ripsrc, n, opts) == 0) { 233 /* should notify about lost packet */ 234 m_freem(n); 235 if (opts) 236 m_freem(opts); 237 SOCKBUF_UNLOCK(&so->so_rcv); 238 } else 239 sorwakeup_locked(so); 240 } else 241 m_freem(n); 242 return (policyfail); 243} 244 245/* 246 * Setup generic address and protocol structures for raw_input routine, then 247 * pass them along with mbuf chain. 248 */ 249void 250rip_input(struct mbuf *m, int off) 251{ 252 INIT_VNET_INET(curvnet); 253 struct ip *ip = mtod(m, struct ip *); 254 int proto = ip->ip_p; 255 struct inpcb *inp, *last; 256 struct sockaddr_in ripsrc; 257 int hash; 258 259 bzero(&ripsrc, sizeof(ripsrc)); 260 ripsrc.sin_len = sizeof(ripsrc); 261 ripsrc.sin_family = AF_INET; 262 ripsrc.sin_addr = ip->ip_src; 263 last = NULL; 264 hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr, 265 ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask); 266 INP_INFO_RLOCK(&V_ripcbinfo); 267 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) { 268 if (inp->inp_ip_p != proto) 269 continue; 270#ifdef INET6 271 /* XXX inp locking */ 272 if ((inp->inp_vflag & INP_IPV4) == 0) 273 continue; 274#endif 275 if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr) 276 continue; 277 if (inp->inp_faddr.s_addr != ip->ip_src.s_addr) 278 continue; 279 if (jailed(inp->inp_cred)) { 280 if (!prison_check_ip4(inp->inp_cred, &ip->ip_dst)) 281 continue; 282 } 283 if (last) { 284 struct mbuf *n; 285 286 n = m_copy(m, 0, (int)M_COPYALL); 287 if (n != NULL) 288 (void) rip_append(last, ip, n, &ripsrc); 289 /* XXX count dropped packet */ 290 INP_RUNLOCK(last); 291 } 292 INP_RLOCK(inp); 293 last = inp; 294 } 295 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) { 296 if (inp->inp_ip_p && inp->inp_ip_p != proto) 297 continue; 298#ifdef INET6 299 /* XXX inp locking */ 300 if ((inp->inp_vflag & INP_IPV4) == 0) 301 continue; 302#endif 303 if (inp->inp_laddr.s_addr && 304 inp->inp_laddr.s_addr != ip->ip_dst.s_addr) 305 continue; 306 if (inp->inp_faddr.s_addr && 307 inp->inp_faddr.s_addr != ip->ip_src.s_addr) 308 continue; 309 if (jailed(inp->inp_cred)) { 310 if (!prison_check_ip4(inp->inp_cred, &ip->ip_dst)) 311 continue; 312 } 313 if (last) { 314 struct mbuf *n; 315 316 n = m_copy(m, 0, (int)M_COPYALL); 317 if (n != NULL) 318 (void) rip_append(last, ip, n, &ripsrc); 319 /* XXX count dropped packet */ 320 INP_RUNLOCK(last); 321 } 322 INP_RLOCK(inp); 323 last = inp; 324 } 325 INP_INFO_RUNLOCK(&V_ripcbinfo); 326 if (last != NULL) { 327 if (rip_append(last, ip, m, &ripsrc) != 0) 328 V_ipstat.ips_delivered--; 329 INP_RUNLOCK(last); 330 } else { 331 m_freem(m); 332 V_ipstat.ips_noproto++; 333 V_ipstat.ips_delivered--; 334 } 335} 336 337/* 338 * Generate IP header and pass packet to ip_output. Tack on options user may 339 * have setup with control call. 340 */ 341int 342rip_output(struct mbuf *m, struct socket *so, u_long dst) 343{ 344 INIT_VNET_INET(so->so_vnet); 345 struct ip *ip; 346 int error; 347 struct inpcb *inp = sotoinpcb(so); 348 int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) | 349 IP_ALLOWBROADCAST; 350 351 /* 352 * If the user handed us a complete IP packet, use it. Otherwise, 353 * allocate an mbuf for a header and fill it in. 354 */ 355 if ((inp->inp_flags & INP_HDRINCL) == 0) { 356 if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) { 357 m_freem(m); 358 return(EMSGSIZE); 359 } 360 M_PREPEND(m, sizeof(struct ip), M_DONTWAIT); 361 if (m == NULL) 362 return(ENOBUFS); 363 364 INP_RLOCK(inp); 365 ip = mtod(m, struct ip *); 366 ip->ip_tos = inp->inp_ip_tos; 367 if (inp->inp_flags & INP_DONTFRAG) 368 ip->ip_off = IP_DF; 369 else 370 ip->ip_off = 0; 371 ip->ip_p = inp->inp_ip_p; 372 ip->ip_len = m->m_pkthdr.len; 373 if (jailed(inp->inp_cred)) { 374 if (prison_getip4(inp->inp_cred, &ip->ip_src)) { 375 INP_RUNLOCK(inp); 376 m_freem(m); 377 return (EPERM); 378 } 379 } else { 380 ip->ip_src = inp->inp_laddr; 381 } 382 ip->ip_dst.s_addr = dst; 383 ip->ip_ttl = inp->inp_ip_ttl; 384 } else { 385 if (m->m_pkthdr.len > IP_MAXPACKET) { 386 m_freem(m); 387 return(EMSGSIZE); 388 } 389 INP_RLOCK(inp); 390 ip = mtod(m, struct ip *); 391 if (!prison_check_ip4(inp->inp_cred, &ip->ip_src)) { 392 INP_RUNLOCK(inp); 393 m_freem(m); 394 return (EPERM); 395 } 396 397 /* 398 * Don't allow both user specified and setsockopt options, 399 * and don't allow packet length sizes that will crash. 400 */ 401 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) 402 || (ip->ip_len > m->m_pkthdr.len) 403 || (ip->ip_len < (ip->ip_hl << 2))) { 404 INP_RUNLOCK(inp); 405 m_freem(m); 406 return (EINVAL); 407 } 408 if (ip->ip_id == 0) 409 ip->ip_id = ip_newid(); 410 411 /* 412 * XXX prevent ip_output from overwriting header fields. 413 */ 414 flags |= IP_RAWOUTPUT; 415 V_ipstat.ips_rawout++; 416 } 417 418 if (inp->inp_flags & INP_ONESBCAST) 419 flags |= IP_SENDONES; 420 421#ifdef MAC 422 mac_inpcb_create_mbuf(inp, m); 423#endif 424 425 error = ip_output(m, inp->inp_options, NULL, flags, 426 inp->inp_moptions, inp); 427 INP_RUNLOCK(inp); 428 return (error); 429} 430 431/* 432 * Raw IP socket option processing. 433 * 434 * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could 435 * only be created by a privileged process, and as such, socket option 436 * operations to manage system properties on any raw socket were allowed to 437 * take place without explicit additional access control checks. However, 438 * raw sockets can now also be created in jail(), and therefore explicit 439 * checks are now required. Likewise, raw sockets can be used by a process 440 * after it gives up privilege, so some caution is required. For options 441 * passed down to the IP layer via ip_ctloutput(), checks are assumed to be 442 * performed in ip_ctloutput() and therefore no check occurs here. 443 * Unilaterally checking priv_check() here breaks normal IP socket option 444 * operations on raw sockets. 445 * 446 * When adding new socket options here, make sure to add access control 447 * checks here as necessary. 448 */ 449int 450rip_ctloutput(struct socket *so, struct sockopt *sopt) 451{ 452 struct inpcb *inp = sotoinpcb(so); 453 int error, optval; 454 455 if (sopt->sopt_level != IPPROTO_IP) { 456 if ((sopt->sopt_level == SOL_SOCKET) && 457 (sopt->sopt_name == SO_SETFIB)) { 458 inp->inp_inc.inc_fibnum = so->so_fibnum; 459 return (0); 460 } 461 return (EINVAL); 462 } 463 464 error = 0; 465 switch (sopt->sopt_dir) { 466 case SOPT_GET: 467 switch (sopt->sopt_name) { 468 case IP_HDRINCL: 469 optval = inp->inp_flags & INP_HDRINCL; 470 error = sooptcopyout(sopt, &optval, sizeof optval); 471 break; 472 473 case IP_FW_ADD: /* ADD actually returns the body... */ 474 case IP_FW_GET: 475 case IP_FW_TABLE_GETSIZE: 476 case IP_FW_TABLE_LIST: 477 case IP_FW_NAT_GET_CONFIG: 478 case IP_FW_NAT_GET_LOG: 479 if (ip_fw_ctl_ptr != NULL) 480 error = ip_fw_ctl_ptr(sopt); 481 else 482 error = ENOPROTOOPT; 483 break; 484 485 case IP_DUMMYNET_GET: 486 if (ip_dn_ctl_ptr != NULL) 487 error = ip_dn_ctl_ptr(sopt); 488 else 489 error = ENOPROTOOPT; 490 break ; 491 492 case MRT_INIT: 493 case MRT_DONE: 494 case MRT_ADD_VIF: 495 case MRT_DEL_VIF: 496 case MRT_ADD_MFC: 497 case MRT_DEL_MFC: 498 case MRT_VERSION: 499 case MRT_ASSERT: 500 case MRT_API_SUPPORT: 501 case MRT_API_CONFIG: 502 case MRT_ADD_BW_UPCALL: 503 case MRT_DEL_BW_UPCALL: 504 error = priv_check(curthread, PRIV_NETINET_MROUTE); 505 if (error != 0) 506 return (error); 507 error = ip_mrouter_get ? ip_mrouter_get(so, sopt) : 508 EOPNOTSUPP; 509 break; 510 511 default: 512 error = ip_ctloutput(so, sopt); 513 break; 514 } 515 break; 516 517 case SOPT_SET: 518 switch (sopt->sopt_name) { 519 case IP_HDRINCL: 520 error = sooptcopyin(sopt, &optval, sizeof optval, 521 sizeof optval); 522 if (error) 523 break; 524 if (optval) 525 inp->inp_flags |= INP_HDRINCL; 526 else 527 inp->inp_flags &= ~INP_HDRINCL; 528 break; 529 530 case IP_FW_ADD: 531 case IP_FW_DEL: 532 case IP_FW_FLUSH: 533 case IP_FW_ZERO: 534 case IP_FW_RESETLOG: 535 case IP_FW_TABLE_ADD: 536 case IP_FW_TABLE_DEL: 537 case IP_FW_TABLE_FLUSH: 538 case IP_FW_NAT_CFG: 539 case IP_FW_NAT_DEL: 540 if (ip_fw_ctl_ptr != NULL) 541 error = ip_fw_ctl_ptr(sopt); 542 else 543 error = ENOPROTOOPT; 544 break; 545 546 case IP_DUMMYNET_CONFIGURE: 547 case IP_DUMMYNET_DEL: 548 case IP_DUMMYNET_FLUSH: 549 if (ip_dn_ctl_ptr != NULL) 550 error = ip_dn_ctl_ptr(sopt); 551 else 552 error = ENOPROTOOPT ; 553 break ; 554 555 case IP_RSVP_ON: 556 error = priv_check(curthread, PRIV_NETINET_MROUTE); 557 if (error != 0) 558 return (error); 559 error = ip_rsvp_init(so); 560 break; 561 562 case IP_RSVP_OFF: 563 error = priv_check(curthread, PRIV_NETINET_MROUTE); 564 if (error != 0) 565 return (error); 566 error = ip_rsvp_done(); 567 break; 568 569 case IP_RSVP_VIF_ON: 570 case IP_RSVP_VIF_OFF: 571 error = priv_check(curthread, PRIV_NETINET_MROUTE); 572 if (error != 0) 573 return (error); 574 error = ip_rsvp_vif ? 575 ip_rsvp_vif(so, sopt) : EINVAL; 576 break; 577 578 case MRT_INIT: 579 case MRT_DONE: 580 case MRT_ADD_VIF: 581 case MRT_DEL_VIF: 582 case MRT_ADD_MFC: 583 case MRT_DEL_MFC: 584 case MRT_VERSION: 585 case MRT_ASSERT: 586 case MRT_API_SUPPORT: 587 case MRT_API_CONFIG: 588 case MRT_ADD_BW_UPCALL: 589 case MRT_DEL_BW_UPCALL: 590 error = priv_check(curthread, PRIV_NETINET_MROUTE); 591 if (error != 0) 592 return (error); 593 error = ip_mrouter_set ? ip_mrouter_set(so, sopt) : 594 EOPNOTSUPP; 595 break; 596 597 default: 598 error = ip_ctloutput(so, sopt); 599 break; 600 } 601 break; 602 } 603 604 return (error); 605} 606 607/* 608 * This function exists solely to receive the PRC_IFDOWN messages which are 609 * sent by if_down(). It looks for an ifaddr whose ifa_addr is sa, and calls 610 * in_ifadown() to remove all routes corresponding to that address. It also 611 * receives the PRC_IFUP messages from if_up() and reinstalls the interface 612 * routes. 613 */ 614void 615rip_ctlinput(int cmd, struct sockaddr *sa, void *vip) 616{ 617 INIT_VNET_INET(curvnet); 618 struct in_ifaddr *ia; 619 struct ifnet *ifp; 620 int err; 621 int flags; 622 623 switch (cmd) { 624 case PRC_IFDOWN: 625 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 626 if (ia->ia_ifa.ifa_addr == sa 627 && (ia->ia_flags & IFA_ROUTE)) { 628 /* 629 * in_ifscrub kills the interface route. 630 */ 631 in_ifscrub(ia->ia_ifp, ia); 632 /* 633 * in_ifadown gets rid of all the rest of the 634 * routes. This is not quite the right thing 635 * to do, but at least if we are running a 636 * routing process they will come back. 637 */ 638 in_ifadown(&ia->ia_ifa, 0); 639 break; 640 } 641 } 642 break; 643 644 case PRC_IFUP: 645 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { 646 if (ia->ia_ifa.ifa_addr == sa) 647 break; 648 } 649 if (ia == 0 || (ia->ia_flags & IFA_ROUTE)) 650 return; 651 flags = RTF_UP; 652 ifp = ia->ia_ifa.ifa_ifp; 653 654 if ((ifp->if_flags & IFF_LOOPBACK) 655 || (ifp->if_flags & IFF_POINTOPOINT)) 656 flags |= RTF_HOST; 657 658 err = rtinit(&ia->ia_ifa, RTM_ADD, flags); 659 if (err == 0) 660 ia->ia_flags |= IFA_ROUTE; 661 break; 662 } 663} 664 665u_long rip_sendspace = 9216; 666u_long rip_recvspace = 9216; 667 668SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW, 669 &rip_sendspace, 0, "Maximum outgoing raw IP datagram size"); 670SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW, 671 &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams"); 672 673static int 674rip_attach(struct socket *so, int proto, struct thread *td) 675{ 676 INIT_VNET_INET(so->so_vnet); 677 struct inpcb *inp; 678 int error; 679 680 inp = sotoinpcb(so); 681 KASSERT(inp == NULL, ("rip_attach: inp != NULL")); 682 683 error = priv_check(td, PRIV_NETINET_RAW); 684 if (error) 685 return (error); 686 if (proto >= IPPROTO_MAX || proto < 0) 687 return EPROTONOSUPPORT; 688 error = soreserve(so, rip_sendspace, rip_recvspace); 689 if (error) 690 return (error); 691 INP_INFO_WLOCK(&V_ripcbinfo); 692 error = in_pcballoc(so, &V_ripcbinfo); 693 if (error) { 694 INP_INFO_WUNLOCK(&V_ripcbinfo); 695 return (error); 696 } 697 inp = (struct inpcb *)so->so_pcb; 698 inp->inp_vflag |= INP_IPV4; 699 inp->inp_ip_p = proto; 700 inp->inp_ip_ttl = V_ip_defttl; 701 rip_inshash(inp); 702 INP_INFO_WUNLOCK(&V_ripcbinfo); 703 INP_WUNLOCK(inp); 704 return (0); 705} 706 707static void 708rip_detach(struct socket *so) 709{ 710 INIT_VNET_INET(so->so_vnet); 711 struct inpcb *inp; 712 713 inp = sotoinpcb(so); 714 KASSERT(inp != NULL, ("rip_detach: inp == NULL")); 715 KASSERT(inp->inp_faddr.s_addr == INADDR_ANY, 716 ("rip_detach: not closed")); 717 718 INP_INFO_WLOCK(&V_ripcbinfo); 719 INP_WLOCK(inp); 720 rip_delhash(inp); 721 if (so == V_ip_mrouter && ip_mrouter_done) 722 ip_mrouter_done(); 723 if (ip_rsvp_force_done) 724 ip_rsvp_force_done(so); 725 if (so == V_ip_rsvpd) 726 ip_rsvp_done(); 727 in_pcbdetach(inp); 728 in_pcbfree(inp); 729 INP_INFO_WUNLOCK(&V_ripcbinfo); 730} 731 732static void 733rip_dodisconnect(struct socket *so, struct inpcb *inp) 734{ 735 736 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo); 737 INP_WLOCK_ASSERT(inp); 738 739 rip_delhash(inp); 740 inp->inp_faddr.s_addr = INADDR_ANY; 741 rip_inshash(inp); 742 SOCK_LOCK(so); 743 so->so_state &= ~SS_ISCONNECTED; 744 SOCK_UNLOCK(so); 745} 746 747static void 748rip_abort(struct socket *so) 749{ 750 INIT_VNET_INET(so->so_vnet); 751 struct inpcb *inp; 752 753 inp = sotoinpcb(so); 754 KASSERT(inp != NULL, ("rip_abort: inp == NULL")); 755 756 INP_INFO_WLOCK(&V_ripcbinfo); 757 INP_WLOCK(inp); 758 rip_dodisconnect(so, inp); 759 INP_WUNLOCK(inp); 760 INP_INFO_WUNLOCK(&V_ripcbinfo); 761} 762 763static void 764rip_close(struct socket *so) 765{ 766 INIT_VNET_INET(so->so_vnet); 767 struct inpcb *inp; 768 769 inp = sotoinpcb(so); 770 KASSERT(inp != NULL, ("rip_close: inp == NULL")); 771 772 INP_INFO_WLOCK(&V_ripcbinfo); 773 INP_WLOCK(inp); 774 rip_dodisconnect(so, inp); 775 INP_WUNLOCK(inp); 776 INP_INFO_WUNLOCK(&V_ripcbinfo); 777} 778 779static int 780rip_disconnect(struct socket *so) 781{ 782 INIT_VNET_INET(so->so_vnet); 783 struct inpcb *inp; 784 785 if ((so->so_state & SS_ISCONNECTED) == 0) 786 return (ENOTCONN); 787 788 inp = sotoinpcb(so); 789 KASSERT(inp != NULL, ("rip_disconnect: inp == NULL")); 790 791 INP_INFO_WLOCK(&V_ripcbinfo); 792 INP_WLOCK(inp); 793 rip_dodisconnect(so, inp); 794 INP_WUNLOCK(inp); 795 INP_INFO_WUNLOCK(&V_ripcbinfo); 796 return (0); 797} 798 799static int 800rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 801{ 802 INIT_VNET_NET(so->so_vnet); 803 INIT_VNET_INET(so->so_vnet); 804 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 805 struct inpcb *inp; 806 807 if (nam->sa_len != sizeof(*addr)) 808 return (EINVAL); 809 810 if (!prison_check_ip4(td->td_ucred, &addr->sin_addr)) 811 return (EADDRNOTAVAIL); 812 813 if (TAILQ_EMPTY(&V_ifnet) || 814 (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) || 815 (addr->sin_addr.s_addr && 816 ifa_ifwithaddr((struct sockaddr *)addr) == 0)) 817 return (EADDRNOTAVAIL); 818 819 inp = sotoinpcb(so); 820 KASSERT(inp != NULL, ("rip_bind: inp == NULL")); 821 822 INP_INFO_WLOCK(&V_ripcbinfo); 823 INP_WLOCK(inp); 824 rip_delhash(inp); 825 inp->inp_laddr = addr->sin_addr; 826 rip_inshash(inp); 827 INP_WUNLOCK(inp); 828 INP_INFO_WUNLOCK(&V_ripcbinfo); 829 return (0); 830} 831 832static int 833rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 834{ 835 INIT_VNET_NET(so->so_vnet); 836 INIT_VNET_INET(so->so_vnet); 837 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 838 struct inpcb *inp; 839 840 if (nam->sa_len != sizeof(*addr)) 841 return (EINVAL); 842 if (TAILQ_EMPTY(&V_ifnet)) 843 return (EADDRNOTAVAIL); 844 if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) 845 return (EAFNOSUPPORT); 846 847 inp = sotoinpcb(so); 848 KASSERT(inp != NULL, ("rip_connect: inp == NULL")); 849 850 INP_INFO_WLOCK(&V_ripcbinfo); 851 INP_WLOCK(inp); 852 rip_delhash(inp); 853 inp->inp_faddr = addr->sin_addr; 854 rip_inshash(inp); 855 soisconnected(so); 856 INP_WUNLOCK(inp); 857 INP_INFO_WUNLOCK(&V_ripcbinfo); 858 return (0); 859} 860 861static int 862rip_shutdown(struct socket *so) 863{ 864 struct inpcb *inp; 865 866 inp = sotoinpcb(so); 867 KASSERT(inp != NULL, ("rip_shutdown: inp == NULL")); 868 869 INP_WLOCK(inp); 870 socantsendmore(so); 871 INP_WUNLOCK(inp); 872 return (0); 873} 874 875static int 876rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 877 struct mbuf *control, struct thread *td) 878{ 879 struct inpcb *inp; 880 u_long dst; 881 882 inp = sotoinpcb(so); 883 KASSERT(inp != NULL, ("rip_send: inp == NULL")); 884 885 /* 886 * Note: 'dst' reads below are unlocked. 887 */ 888 if (so->so_state & SS_ISCONNECTED) { 889 if (nam) { 890 m_freem(m); 891 return (EISCONN); 892 } 893 dst = inp->inp_faddr.s_addr; /* Unlocked read. */ 894 } else { 895 if (nam == NULL) { 896 m_freem(m); 897 return (ENOTCONN); 898 } 899 dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr; 900 } 901 return (rip_output(m, so, dst)); 902} 903 904static int 905rip_pcblist(SYSCTL_HANDLER_ARGS) 906{ 907 INIT_VNET_INET(curvnet); 908 int error, i, n; 909 struct inpcb *inp, **inp_list; 910 inp_gen_t gencnt; 911 struct xinpgen xig; 912 913 /* 914 * The process of preparing the TCB list is too time-consuming and 915 * resource-intensive to repeat twice on every request. 916 */ 917 if (req->oldptr == 0) { 918 n = V_ripcbinfo.ipi_count; 919 req->oldidx = 2 * (sizeof xig) 920 + (n + n/8) * sizeof(struct xinpcb); 921 return (0); 922 } 923 924 if (req->newptr != 0) 925 return (EPERM); 926 927 /* 928 * OK, now we're committed to doing something. 929 */ 930 INP_INFO_RLOCK(&V_ripcbinfo); 931 gencnt = V_ripcbinfo.ipi_gencnt; 932 n = V_ripcbinfo.ipi_count; 933 INP_INFO_RUNLOCK(&V_ripcbinfo); 934 935 xig.xig_len = sizeof xig; 936 xig.xig_count = n; 937 xig.xig_gen = gencnt; 938 xig.xig_sogen = so_gencnt; 939 error = SYSCTL_OUT(req, &xig, sizeof xig); 940 if (error) 941 return (error); 942 943 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 944 if (inp_list == 0) 945 return (ENOMEM); 946 947 INP_INFO_RLOCK(&V_ripcbinfo); 948 for (inp = LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n; 949 inp = LIST_NEXT(inp, inp_list)) { 950 INP_RLOCK(inp); 951 if (inp->inp_gencnt <= gencnt && 952 cr_canseeinpcb(req->td->td_ucred, inp) == 0) { 953 /* XXX held references? */ 954 inp_list[i++] = inp; 955 } 956 INP_RUNLOCK(inp); 957 } 958 INP_INFO_RUNLOCK(&V_ripcbinfo); 959 n = i; 960 961 error = 0; 962 for (i = 0; i < n; i++) { 963 inp = inp_list[i]; 964 INP_RLOCK(inp); 965 if (inp->inp_gencnt <= gencnt) { 966 struct xinpcb xi; 967 bzero(&xi, sizeof(xi)); 968 xi.xi_len = sizeof xi; 969 /* XXX should avoid extra copy */ 970 bcopy(inp, &xi.xi_inp, sizeof *inp); 971 if (inp->inp_socket) 972 sotoxsocket(inp->inp_socket, &xi.xi_socket); 973 INP_RUNLOCK(inp); 974 error = SYSCTL_OUT(req, &xi, sizeof xi); 975 } else 976 INP_RUNLOCK(inp); 977 } 978 if (!error) { 979 /* 980 * Give the user an updated idea of our state. If the 981 * generation differs from what we told her before, she knows 982 * that something happened while we were processing this 983 * request, and it might be necessary to retry. 984 */ 985 INP_INFO_RLOCK(&V_ripcbinfo); 986 xig.xig_gen = V_ripcbinfo.ipi_gencnt; 987 xig.xig_sogen = so_gencnt; 988 xig.xig_count = V_ripcbinfo.ipi_count; 989 INP_INFO_RUNLOCK(&V_ripcbinfo); 990 error = SYSCTL_OUT(req, &xig, sizeof xig); 991 } 992 free(inp_list, M_TEMP); 993 return (error); 994} 995 996SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0, 997 rip_pcblist, "S,xinpcb", "List of active raw IP sockets"); 998 999struct pr_usrreqs rip_usrreqs = { 1000 .pru_abort = rip_abort, 1001 .pru_attach = rip_attach, 1002 .pru_bind = rip_bind, 1003 .pru_connect = rip_connect, 1004 .pru_control = in_control, 1005 .pru_detach = rip_detach, 1006 .pru_disconnect = rip_disconnect, 1007 .pru_peeraddr = in_getpeeraddr, 1008 .pru_send = rip_send, 1009 .pru_shutdown = rip_shutdown, 1010 .pru_sockaddr = in_getsockaddr, 1011 .pru_sosetlabel = in_pcbsosetlabel, 1012 .pru_close = rip_close, 1013}; 1014