raw_ip.c revision 178285
1130803Smarcel/*- 2130803Smarcel * Copyright (c) 1982, 1986, 1988, 1993 3130803Smarcel * The Regents of the University of California. All rights reserved. 4130803Smarcel * 5130803Smarcel * Redistribution and use in source and binary forms, with or without 6130803Smarcel * modification, are permitted provided that the following conditions 7130803Smarcel * are met: 8130803Smarcel * 1. Redistributions of source code must retain the above copyright 9130803Smarcel * notice, this list of conditions and the following disclaimer. 10130803Smarcel * 2. Redistributions in binary form must reproduce the above copyright 11130803Smarcel * notice, this list of conditions and the following disclaimer in the 12130803Smarcel * documentation and/or other materials provided with the distribution. 13130803Smarcel * 4. Neither the name of the University nor the names of its contributors 14130803Smarcel * may be used to endorse or promote products derived from this software 15130803Smarcel * without specific prior written permission. 16130803Smarcel * 17130803Smarcel * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18130803Smarcel * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19130803Smarcel * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20130803Smarcel * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21130803Smarcel * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22130803Smarcel * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23130803Smarcel * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24130803Smarcel * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25130803Smarcel * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26130803Smarcel * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27130803Smarcel * SUCH DAMAGE. 28130803Smarcel * 29130803Smarcel * @(#)raw_ip.c 8.7 (Berkeley) 5/15/95 30130803Smarcel */ 31130803Smarcel 32130803Smarcel#include <sys/cdefs.h> 33130803Smarcel__FBSDID("$FreeBSD: head/sys/netinet/raw_ip.c 178285 2008-04-17 21:38:18Z rwatson $"); 34130803Smarcel 35130803Smarcel#include "opt_inet6.h" 36130803Smarcel#include "opt_ipsec.h" 37130803Smarcel#include "opt_mac.h" 38130803Smarcel 39130803Smarcel#include <sys/param.h> 40130803Smarcel#include <sys/jail.h> 41130803Smarcel#include <sys/kernel.h> 42130803Smarcel#include <sys/lock.h> 43130803Smarcel#include <sys/malloc.h> 44130803Smarcel#include <sys/mbuf.h> 45130803Smarcel#include <sys/priv.h> 46130803Smarcel#include <sys/proc.h> 47130803Smarcel#include <sys/protosw.h> 48130803Smarcel#include <sys/signalvar.h> 49130803Smarcel#include <sys/socket.h> 50130803Smarcel#include <sys/socketvar.h> 51130803Smarcel#include <sys/sx.h> 52130803Smarcel#include <sys/sysctl.h> 53130803Smarcel#include <sys/systm.h> 54130803Smarcel 55130803Smarcel#include <vm/uma.h> 56130803Smarcel 57130803Smarcel#include <net/if.h> 58130803Smarcel#include <net/route.h> 59130803Smarcel 60130803Smarcel#include <netinet/in.h> 61130803Smarcel#include <netinet/in_systm.h> 62130803Smarcel#include <netinet/in_pcb.h> 63130803Smarcel#include <netinet/in_var.h> 64130803Smarcel#include <netinet/ip.h> 65130803Smarcel#include <netinet/ip_var.h> 66130803Smarcel#include <netinet/ip_mroute.h> 67130803Smarcel 68130803Smarcel#include <netinet/ip_fw.h> 69130803Smarcel#include <netinet/ip_dummynet.h> 70130803Smarcel 71130803Smarcel#ifdef IPSEC 72130803Smarcel#include <netipsec/ipsec.h> 73130803Smarcel#endif /*IPSEC*/ 74130803Smarcel 75130803Smarcel#include <security/mac/mac_framework.h> 76130803Smarcel 77130803Smarcelstruct inpcbhead ripcb; 78130803Smarcelstruct inpcbinfo ripcbinfo; 79130803Smarcel 80130803Smarcel/* control hooks for ipfw and dummynet */ 81130803Smarcelip_fw_ctl_t *ip_fw_ctl_ptr = NULL; 82130803Smarcelip_dn_ctl_t *ip_dn_ctl_ptr = NULL; 83130803Smarcel 84130803Smarcel/* 85130803Smarcel * hooks for multicast routing. They all default to NULL, 86130803Smarcel * so leave them not initialized and rely on BSS being set to 0. 87130803Smarcel */ 88130803Smarcel 89130803Smarcel/* The socket used to communicate with the multicast routing daemon. */ 90130803Smarcelstruct socket *ip_mrouter; 91130803Smarcel 92130803Smarcel/* The various mrouter and rsvp functions */ 93130803Smarcelint (*ip_mrouter_set)(struct socket *, struct sockopt *); 94130803Smarcelint (*ip_mrouter_get)(struct socket *, struct sockopt *); 95130803Smarcelint (*ip_mrouter_done)(void); 96130803Smarcelint (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 97130803Smarcel struct ip_moptions *); 98130803Smarcelint (*mrt_ioctl)(int, caddr_t); 99130803Smarcelint (*legal_vif_num)(int); 100130803Smarcelu_long (*ip_mcast_src)(int); 101130803Smarcel 102130803Smarcelvoid (*rsvp_input_p)(struct mbuf *m, int off); 103130803Smarcelint (*ip_rsvp_vif)(struct socket *, struct sockopt *); 104130803Smarcelvoid (*ip_rsvp_force_done)(struct socket *); 105130803Smarcel 106130803Smarcel/* 107130803Smarcel * Raw interface to IP protocol. 108130803Smarcel */ 109130803Smarcel 110130803Smarcel/* 111130803Smarcel * Initialize raw connection block q. 112130803Smarcel */ 113130803Smarcelstatic void 114130803Smarcelrip_zone_change(void *tag) 115130803Smarcel{ 116130803Smarcel 117130803Smarcel uma_zone_set_max(ripcbinfo.ipi_zone, maxsockets); 118130803Smarcel} 119130803Smarcel 120130803Smarcelstatic int 121130803Smarcelrip_inpcb_init(void *mem, int size, int flags) 122130803Smarcel{ 123130803Smarcel struct inpcb *inp = mem; 124130803Smarcel 125130803Smarcel INP_LOCK_INIT(inp, "inp", "rawinp"); 126130803Smarcel return (0); 127130803Smarcel} 128130803Smarcel 129130803Smarcelvoid 130130803Smarcelrip_init(void) 131130803Smarcel{ 132130803Smarcel 133130803Smarcel INP_INFO_LOCK_INIT(&ripcbinfo, "rip"); 134130803Smarcel LIST_INIT(&ripcb); 135130803Smarcel ripcbinfo.ipi_listhead = &ripcb; 136130803Smarcel /* 137130803Smarcel * XXX We don't use the hash list for raw IP, but it's easier 138130803Smarcel * to allocate a one entry hash list than it is to check all 139130803Smarcel * over the place for hashbase == NULL. 140130803Smarcel */ 141130803Smarcel ripcbinfo.ipi_hashbase = hashinit(1, M_PCB, &ripcbinfo.ipi_hashmask); 142130803Smarcel ripcbinfo.ipi_porthashbase = hashinit(1, M_PCB, 143130803Smarcel &ripcbinfo.ipi_porthashmask); 144130803Smarcel ripcbinfo.ipi_zone = uma_zcreate("ripcb", sizeof(struct inpcb), 145130803Smarcel NULL, NULL, rip_inpcb_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 146130803Smarcel uma_zone_set_max(ripcbinfo.ipi_zone, maxsockets); 147130803Smarcel EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, 148130803Smarcel NULL, EVENTHANDLER_PRI_ANY); 149130803Smarcel} 150130803Smarcel 151130803Smarcelstatic struct sockaddr_in ripsrc = { sizeof(ripsrc), AF_INET }; 152130803Smarcel 153130803Smarcelstatic int 154130803Smarcelraw_append(struct inpcb *last, struct ip *ip, struct mbuf *n) 155130803Smarcel{ 156130803Smarcel int policyfail = 0; 157130803Smarcel 158130803Smarcel INP_WLOCK_ASSERT(last); 159130803Smarcel 160130803Smarcel#ifdef IPSEC 161130803Smarcel /* check AH/ESP integrity. */ 162130803Smarcel if (ipsec4_in_reject(n, last)) { 163130803Smarcel policyfail = 1; 164130803Smarcel } 165130803Smarcel#endif /* IPSEC */ 166130803Smarcel#ifdef MAC 167130803Smarcel if (!policyfail && mac_inpcb_check_deliver(last, n) != 0) 168130803Smarcel policyfail = 1; 169130803Smarcel#endif 170130803Smarcel /* Check the minimum TTL for socket. */ 171130803Smarcel if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl) 172130803Smarcel policyfail = 1; 173130803Smarcel if (!policyfail) { 174130803Smarcel struct mbuf *opts = NULL; 175130803Smarcel struct socket *so; 176130803Smarcel 177130803Smarcel so = last->inp_socket; 178130803Smarcel if ((last->inp_flags & INP_CONTROLOPTS) || 179130803Smarcel (so->so_options & (SO_TIMESTAMP | SO_BINTIME))) 180130803Smarcel ip_savecontrol(last, &opts, ip, n); 181130803Smarcel SOCKBUF_LOCK(&so->so_rcv); 182130803Smarcel if (sbappendaddr_locked(&so->so_rcv, 183130803Smarcel (struct sockaddr *)&ripsrc, n, opts) == 0) { 184130803Smarcel /* should notify about lost packet */ 185130803Smarcel m_freem(n); 186130803Smarcel if (opts) 187130803Smarcel m_freem(opts); 188130803Smarcel SOCKBUF_UNLOCK(&so->so_rcv); 189130803Smarcel } else 190130803Smarcel sorwakeup_locked(so); 191130803Smarcel } else 192130803Smarcel m_freem(n); 193130803Smarcel return policyfail; 194130803Smarcel} 195130803Smarcel 196130803Smarcel/* 197130803Smarcel * Setup generic address and protocol structures 198130803Smarcel * for raw_input routine, then pass them along with 199130803Smarcel * mbuf chain. 200130803Smarcel */ 201130803Smarcelvoid 202130803Smarcelrip_input(struct mbuf *m, int off) 203130803Smarcel{ 204130803Smarcel struct ip *ip = mtod(m, struct ip *); 205130803Smarcel int proto = ip->ip_p; 206130803Smarcel struct inpcb *inp, *last; 207130803Smarcel 208130803Smarcel INP_INFO_RLOCK(&ripcbinfo); 209130803Smarcel ripsrc.sin_addr = ip->ip_src; 210130803Smarcel last = NULL; 211130803Smarcel LIST_FOREACH(inp, &ripcb, inp_list) { 212130803Smarcel INP_WLOCK(inp); 213130803Smarcel if (inp->inp_ip_p && inp->inp_ip_p != proto) { 214130803Smarcel docontinue: 215130803Smarcel INP_WUNLOCK(inp); 216130803Smarcel continue; 217130803Smarcel } 218130803Smarcel#ifdef INET6 219130803Smarcel if ((inp->inp_vflag & INP_IPV4) == 0) 220130803Smarcel goto docontinue; 221130803Smarcel#endif 222130803Smarcel if (inp->inp_laddr.s_addr && 223130803Smarcel inp->inp_laddr.s_addr != ip->ip_dst.s_addr) 224130803Smarcel goto docontinue; 225130803Smarcel if (inp->inp_faddr.s_addr && 226130803Smarcel inp->inp_faddr.s_addr != ip->ip_src.s_addr) 227130803Smarcel goto docontinue; 228130803Smarcel if (jailed(inp->inp_socket->so_cred)) 229130803Smarcel if (htonl(prison_getip(inp->inp_socket->so_cred)) != 230130803Smarcel ip->ip_dst.s_addr) 231130803Smarcel goto docontinue; 232130803Smarcel if (last) { 233130803Smarcel struct mbuf *n; 234130803Smarcel 235130803Smarcel n = m_copy(m, 0, (int)M_COPYALL); 236130803Smarcel if (n != NULL) 237130803Smarcel (void) raw_append(last, ip, n); 238130803Smarcel /* XXX count dropped packet */ 239130803Smarcel INP_WUNLOCK(last); 240130803Smarcel } 241130803Smarcel last = inp; 242130803Smarcel } 243130803Smarcel if (last != NULL) { 244130803Smarcel if (raw_append(last, ip, m) != 0) 245130803Smarcel ipstat.ips_delivered--; 246130803Smarcel INP_WUNLOCK(last); 247130803Smarcel } else { 248130803Smarcel m_freem(m); 249130803Smarcel ipstat.ips_noproto++; 250130803Smarcel ipstat.ips_delivered--; 251130803Smarcel } 252130803Smarcel INP_INFO_RUNLOCK(&ripcbinfo); 253130803Smarcel} 254130803Smarcel 255130803Smarcel/* 256130803Smarcel * Generate IP header and pass packet to ip_output. 257130803Smarcel * Tack on options user may have setup with control call. 258130803Smarcel */ 259130803Smarcelint 260130803Smarcelrip_output(struct mbuf *m, struct socket *so, u_long dst) 261130803Smarcel{ 262130803Smarcel struct ip *ip; 263130803Smarcel int error; 264130803Smarcel struct inpcb *inp = sotoinpcb(so); 265130803Smarcel int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) | 266130803Smarcel IP_ALLOWBROADCAST; 267130803Smarcel 268130803Smarcel /* 269130803Smarcel * If the user handed us a complete IP packet, use it. 270130803Smarcel * Otherwise, allocate an mbuf for a header and fill it in. 271130803Smarcel */ 272130803Smarcel if ((inp->inp_flags & INP_HDRINCL) == 0) { 273130803Smarcel if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) { 274130803Smarcel m_freem(m); 275130803Smarcel return(EMSGSIZE); 276130803Smarcel } 277130803Smarcel M_PREPEND(m, sizeof(struct ip), M_DONTWAIT); 278130803Smarcel if (m == NULL) 279130803Smarcel return(ENOBUFS); 280130803Smarcel 281130803Smarcel INP_WLOCK(inp); 282130803Smarcel ip = mtod(m, struct ip *); 283130803Smarcel ip->ip_tos = inp->inp_ip_tos; 284130803Smarcel if (inp->inp_flags & INP_DONTFRAG) 285130803Smarcel ip->ip_off = IP_DF; 286130803Smarcel else 287130803Smarcel ip->ip_off = 0; 288130803Smarcel ip->ip_p = inp->inp_ip_p; 289130803Smarcel ip->ip_len = m->m_pkthdr.len; 290130803Smarcel if (jailed(inp->inp_socket->so_cred)) 291130803Smarcel ip->ip_src.s_addr = 292130803Smarcel htonl(prison_getip(inp->inp_socket->so_cred)); 293130803Smarcel else 294130803Smarcel ip->ip_src = inp->inp_laddr; 295130803Smarcel ip->ip_dst.s_addr = dst; 296130803Smarcel ip->ip_ttl = inp->inp_ip_ttl; 297130803Smarcel } else { 298130803Smarcel if (m->m_pkthdr.len > IP_MAXPACKET) { 299130803Smarcel m_freem(m); 300130803Smarcel return(EMSGSIZE); 301130803Smarcel } 302130803Smarcel INP_WLOCK(inp); 303130803Smarcel ip = mtod(m, struct ip *); 304130803Smarcel if (jailed(inp->inp_socket->so_cred)) { 305130803Smarcel if (ip->ip_src.s_addr != 306130803Smarcel htonl(prison_getip(inp->inp_socket->so_cred))) { 307130803Smarcel INP_WUNLOCK(inp); 308130803Smarcel m_freem(m); 309130803Smarcel return (EPERM); 310130803Smarcel } 311130803Smarcel } 312130803Smarcel /* don't allow both user specified and setsockopt options, 313130803Smarcel and don't allow packet length sizes that will crash */ 314130803Smarcel if (((ip->ip_hl != (sizeof (*ip) >> 2)) 315130803Smarcel && inp->inp_options) 316130803Smarcel || (ip->ip_len > m->m_pkthdr.len) 317130803Smarcel || (ip->ip_len < (ip->ip_hl << 2))) { 318130803Smarcel INP_WUNLOCK(inp); 319130803Smarcel m_freem(m); 320130803Smarcel return EINVAL; 321130803Smarcel } 322130803Smarcel if (ip->ip_id == 0) 323130803Smarcel ip->ip_id = ip_newid(); 324130803Smarcel /* XXX prevent ip_output from overwriting header fields */ 325130803Smarcel flags |= IP_RAWOUTPUT; 326130803Smarcel ipstat.ips_rawout++; 327130803Smarcel } 328130803Smarcel 329130803Smarcel if (inp->inp_flags & INP_ONESBCAST) 330130803Smarcel flags |= IP_SENDONES; 331130803Smarcel 332130803Smarcel#ifdef MAC 333130803Smarcel mac_inpcb_create_mbuf(inp, m); 334130803Smarcel#endif 335130803Smarcel 336130803Smarcel error = ip_output(m, inp->inp_options, NULL, flags, 337130803Smarcel inp->inp_moptions, inp); 338130803Smarcel INP_WUNLOCK(inp); 339130803Smarcel return error; 340130803Smarcel} 341130803Smarcel 342130803Smarcel/* 343130803Smarcel * Raw IP socket option processing. 344130803Smarcel * 345130803Smarcel * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could 346130803Smarcel * only be created by a privileged process, and as such, socket option 347130803Smarcel * operations to manage system properties on any raw socket were allowed to 348130803Smarcel * take place without explicit additional access control checks. However, 349130803Smarcel * raw sockets can now also be created in jail(), and therefore explicit 350130803Smarcel * checks are now required. Likewise, raw sockets can be used by a process 351130803Smarcel * after it gives up privilege, so some caution is required. For options 352 * passed down to the IP layer via ip_ctloutput(), checks are assumed to be 353 * performed in ip_ctloutput() and therefore no check occurs here. 354 * Unilaterally checking priv_check() here breaks normal IP socket option 355 * operations on raw sockets. 356 * 357 * When adding new socket options here, make sure to add access control 358 * checks here as necessary. 359 */ 360int 361rip_ctloutput(struct socket *so, struct sockopt *sopt) 362{ 363 struct inpcb *inp = sotoinpcb(so); 364 int error, optval; 365 366 if (sopt->sopt_level != IPPROTO_IP) 367 return (EINVAL); 368 369 error = 0; 370 switch (sopt->sopt_dir) { 371 case SOPT_GET: 372 switch (sopt->sopt_name) { 373 case IP_HDRINCL: 374 optval = inp->inp_flags & INP_HDRINCL; 375 error = sooptcopyout(sopt, &optval, sizeof optval); 376 break; 377 378 case IP_FW_ADD: /* ADD actually returns the body... */ 379 case IP_FW_GET: 380 case IP_FW_TABLE_GETSIZE: 381 case IP_FW_TABLE_LIST: 382 case IP_FW_NAT_GET_CONFIG: 383 case IP_FW_NAT_GET_LOG: 384 /* 385 * XXXRW: Isn't this checked one layer down? Yes, it 386 * is. 387 */ 388 error = priv_check(curthread, PRIV_NETINET_IPFW); 389 if (error != 0) 390 return (error); 391 if (ip_fw_ctl_ptr != NULL) 392 error = ip_fw_ctl_ptr(sopt); 393 else 394 error = ENOPROTOOPT; 395 break; 396 397 case IP_DUMMYNET_GET: 398 error = priv_check(curthread, PRIV_NETINET_DUMMYNET); 399 if (error != 0) 400 return (error); 401 if (ip_dn_ctl_ptr != NULL) 402 error = ip_dn_ctl_ptr(sopt); 403 else 404 error = ENOPROTOOPT; 405 break ; 406 407 case MRT_INIT: 408 case MRT_DONE: 409 case MRT_ADD_VIF: 410 case MRT_DEL_VIF: 411 case MRT_ADD_MFC: 412 case MRT_DEL_MFC: 413 case MRT_VERSION: 414 case MRT_ASSERT: 415 case MRT_API_SUPPORT: 416 case MRT_API_CONFIG: 417 case MRT_ADD_BW_UPCALL: 418 case MRT_DEL_BW_UPCALL: 419 error = priv_check(curthread, PRIV_NETINET_MROUTE); 420 if (error != 0) 421 return (error); 422 error = ip_mrouter_get ? ip_mrouter_get(so, sopt) : 423 EOPNOTSUPP; 424 break; 425 426 default: 427 error = ip_ctloutput(so, sopt); 428 break; 429 } 430 break; 431 432 case SOPT_SET: 433 switch (sopt->sopt_name) { 434 case IP_HDRINCL: 435 error = sooptcopyin(sopt, &optval, sizeof optval, 436 sizeof optval); 437 if (error) 438 break; 439 if (optval) 440 inp->inp_flags |= INP_HDRINCL; 441 else 442 inp->inp_flags &= ~INP_HDRINCL; 443 break; 444 445 case IP_FW_ADD: 446 case IP_FW_DEL: 447 case IP_FW_FLUSH: 448 case IP_FW_ZERO: 449 case IP_FW_RESETLOG: 450 case IP_FW_TABLE_ADD: 451 case IP_FW_TABLE_DEL: 452 case IP_FW_TABLE_FLUSH: 453 case IP_FW_NAT_CFG: 454 case IP_FW_NAT_DEL: 455 /* 456 * XXXRW: Isn't this checked one layer down? 457 */ 458 error = priv_check(curthread, PRIV_NETINET_IPFW); 459 if (error != 0) 460 return (error); 461 if (ip_fw_ctl_ptr != NULL) 462 error = ip_fw_ctl_ptr(sopt); 463 else 464 error = ENOPROTOOPT; 465 break; 466 467 case IP_DUMMYNET_CONFIGURE: 468 case IP_DUMMYNET_DEL: 469 case IP_DUMMYNET_FLUSH: 470 error = priv_check(curthread, PRIV_NETINET_DUMMYNET); 471 if (error != 0) 472 return (error); 473 if (ip_dn_ctl_ptr != NULL) 474 error = ip_dn_ctl_ptr(sopt); 475 else 476 error = ENOPROTOOPT ; 477 break ; 478 479 case IP_RSVP_ON: 480 error = priv_check(curthread, PRIV_NETINET_MROUTE); 481 if (error != 0) 482 return (error); 483 error = ip_rsvp_init(so); 484 break; 485 486 case IP_RSVP_OFF: 487 error = priv_check(curthread, PRIV_NETINET_MROUTE); 488 if (error != 0) 489 return (error); 490 error = ip_rsvp_done(); 491 break; 492 493 case IP_RSVP_VIF_ON: 494 case IP_RSVP_VIF_OFF: 495 error = priv_check(curthread, PRIV_NETINET_MROUTE); 496 if (error != 0) 497 return (error); 498 error = ip_rsvp_vif ? 499 ip_rsvp_vif(so, sopt) : EINVAL; 500 break; 501 502 case MRT_INIT: 503 case MRT_DONE: 504 case MRT_ADD_VIF: 505 case MRT_DEL_VIF: 506 case MRT_ADD_MFC: 507 case MRT_DEL_MFC: 508 case MRT_VERSION: 509 case MRT_ASSERT: 510 case MRT_API_SUPPORT: 511 case MRT_API_CONFIG: 512 case MRT_ADD_BW_UPCALL: 513 case MRT_DEL_BW_UPCALL: 514 error = priv_check(curthread, PRIV_NETINET_MROUTE); 515 if (error != 0) 516 return (error); 517 error = ip_mrouter_set ? ip_mrouter_set(so, sopt) : 518 EOPNOTSUPP; 519 break; 520 521 default: 522 error = ip_ctloutput(so, sopt); 523 break; 524 } 525 break; 526 } 527 528 return (error); 529} 530 531/* 532 * This function exists solely to receive the PRC_IFDOWN messages which 533 * are sent by if_down(). It looks for an ifaddr whose ifa_addr is sa, 534 * and calls in_ifadown() to remove all routes corresponding to that address. 535 * It also receives the PRC_IFUP messages from if_up() and reinstalls the 536 * interface routes. 537 */ 538void 539rip_ctlinput(int cmd, struct sockaddr *sa, void *vip) 540{ 541 struct in_ifaddr *ia; 542 struct ifnet *ifp; 543 int err; 544 int flags; 545 546 switch (cmd) { 547 case PRC_IFDOWN: 548 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) { 549 if (ia->ia_ifa.ifa_addr == sa 550 && (ia->ia_flags & IFA_ROUTE)) { 551 /* 552 * in_ifscrub kills the interface route. 553 */ 554 in_ifscrub(ia->ia_ifp, ia); 555 /* 556 * in_ifadown gets rid of all the rest of 557 * the routes. This is not quite the right 558 * thing to do, but at least if we are running 559 * a routing process they will come back. 560 */ 561 in_ifadown(&ia->ia_ifa, 0); 562 break; 563 } 564 } 565 break; 566 567 case PRC_IFUP: 568 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link) { 569 if (ia->ia_ifa.ifa_addr == sa) 570 break; 571 } 572 if (ia == 0 || (ia->ia_flags & IFA_ROUTE)) 573 return; 574 flags = RTF_UP; 575 ifp = ia->ia_ifa.ifa_ifp; 576 577 if ((ifp->if_flags & IFF_LOOPBACK) 578 || (ifp->if_flags & IFF_POINTOPOINT)) 579 flags |= RTF_HOST; 580 581 err = rtinit(&ia->ia_ifa, RTM_ADD, flags); 582 if (err == 0) 583 ia->ia_flags |= IFA_ROUTE; 584 break; 585 } 586} 587 588u_long rip_sendspace = 9216; 589u_long rip_recvspace = 9216; 590 591SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW, 592 &rip_sendspace, 0, "Maximum outgoing raw IP datagram size"); 593SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW, 594 &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams"); 595 596static int 597rip_attach(struct socket *so, int proto, struct thread *td) 598{ 599 struct inpcb *inp; 600 int error; 601 602 inp = sotoinpcb(so); 603 KASSERT(inp == NULL, ("rip_attach: inp != NULL")); 604 605 error = priv_check(td, PRIV_NETINET_RAW); 606 if (error) 607 return error; 608 if (proto >= IPPROTO_MAX || proto < 0) 609 return EPROTONOSUPPORT; 610 error = soreserve(so, rip_sendspace, rip_recvspace); 611 if (error) 612 return error; 613 INP_INFO_WLOCK(&ripcbinfo); 614 error = in_pcballoc(so, &ripcbinfo); 615 if (error) { 616 INP_INFO_WUNLOCK(&ripcbinfo); 617 return error; 618 } 619 inp = (struct inpcb *)so->so_pcb; 620 INP_INFO_WUNLOCK(&ripcbinfo); 621 inp->inp_vflag |= INP_IPV4; 622 inp->inp_ip_p = proto; 623 inp->inp_ip_ttl = ip_defttl; 624 INP_WUNLOCK(inp); 625 return 0; 626} 627 628static void 629rip_detach(struct socket *so) 630{ 631 struct inpcb *inp; 632 633 inp = sotoinpcb(so); 634 KASSERT(inp != NULL, ("rip_detach: inp == NULL")); 635 KASSERT(inp->inp_faddr.s_addr == INADDR_ANY, 636 ("rip_detach: not closed")); 637 638 INP_INFO_WLOCK(&ripcbinfo); 639 INP_WLOCK(inp); 640 if (so == ip_mrouter && ip_mrouter_done) 641 ip_mrouter_done(); 642 if (ip_rsvp_force_done) 643 ip_rsvp_force_done(so); 644 if (so == ip_rsvpd) 645 ip_rsvp_done(); 646 in_pcbdetach(inp); 647 in_pcbfree(inp); 648 INP_INFO_WUNLOCK(&ripcbinfo); 649} 650 651static void 652rip_dodisconnect(struct socket *so, struct inpcb *inp) 653{ 654 655 INP_WLOCK_ASSERT(inp); 656 657 inp->inp_faddr.s_addr = INADDR_ANY; 658 SOCK_LOCK(so); 659 so->so_state &= ~SS_ISCONNECTED; 660 SOCK_UNLOCK(so); 661} 662 663static void 664rip_abort(struct socket *so) 665{ 666 struct inpcb *inp; 667 668 inp = sotoinpcb(so); 669 KASSERT(inp != NULL, ("rip_abort: inp == NULL")); 670 671 INP_INFO_WLOCK(&ripcbinfo); 672 INP_WLOCK(inp); 673 rip_dodisconnect(so, inp); 674 INP_WUNLOCK(inp); 675 INP_INFO_WUNLOCK(&ripcbinfo); 676} 677 678static void 679rip_close(struct socket *so) 680{ 681 struct inpcb *inp; 682 683 inp = sotoinpcb(so); 684 KASSERT(inp != NULL, ("rip_close: inp == NULL")); 685 686 INP_INFO_WLOCK(&ripcbinfo); 687 INP_WLOCK(inp); 688 rip_dodisconnect(so, inp); 689 INP_WUNLOCK(inp); 690 INP_INFO_WUNLOCK(&ripcbinfo); 691} 692 693static int 694rip_disconnect(struct socket *so) 695{ 696 struct inpcb *inp; 697 698 if ((so->so_state & SS_ISCONNECTED) == 0) 699 return ENOTCONN; 700 701 inp = sotoinpcb(so); 702 KASSERT(inp != NULL, ("rip_disconnect: inp == NULL")); 703 INP_INFO_WLOCK(&ripcbinfo); 704 INP_WLOCK(inp); 705 rip_dodisconnect(so, inp); 706 INP_WUNLOCK(inp); 707 INP_INFO_WUNLOCK(&ripcbinfo); 708 return (0); 709} 710 711static int 712rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 713{ 714 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 715 struct inpcb *inp; 716 717 if (nam->sa_len != sizeof(*addr)) 718 return EINVAL; 719 720 if (jailed(td->td_ucred)) { 721 if (addr->sin_addr.s_addr == INADDR_ANY) 722 addr->sin_addr.s_addr = 723 htonl(prison_getip(td->td_ucred)); 724 if (htonl(prison_getip(td->td_ucred)) != addr->sin_addr.s_addr) 725 return (EADDRNOTAVAIL); 726 } 727 728 if (TAILQ_EMPTY(&ifnet) || 729 (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) || 730 (addr->sin_addr.s_addr && 731 ifa_ifwithaddr((struct sockaddr *)addr) == 0)) 732 return EADDRNOTAVAIL; 733 734 inp = sotoinpcb(so); 735 KASSERT(inp != NULL, ("rip_bind: inp == NULL")); 736 INP_INFO_WLOCK(&ripcbinfo); 737 INP_WLOCK(inp); 738 inp->inp_laddr = addr->sin_addr; 739 INP_WUNLOCK(inp); 740 INP_INFO_WUNLOCK(&ripcbinfo); 741 return 0; 742} 743 744static int 745rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 746{ 747 struct sockaddr_in *addr = (struct sockaddr_in *)nam; 748 struct inpcb *inp; 749 750 if (nam->sa_len != sizeof(*addr)) 751 return EINVAL; 752 if (TAILQ_EMPTY(&ifnet)) 753 return EADDRNOTAVAIL; 754 if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) 755 return EAFNOSUPPORT; 756 757 inp = sotoinpcb(so); 758 KASSERT(inp != NULL, ("rip_connect: inp == NULL")); 759 INP_INFO_WLOCK(&ripcbinfo); 760 INP_WLOCK(inp); 761 inp->inp_faddr = addr->sin_addr; 762 soisconnected(so); 763 INP_WUNLOCK(inp); 764 INP_INFO_WUNLOCK(&ripcbinfo); 765 return 0; 766} 767 768static int 769rip_shutdown(struct socket *so) 770{ 771 struct inpcb *inp; 772 773 inp = sotoinpcb(so); 774 KASSERT(inp != NULL, ("rip_shutdown: inp == NULL")); 775 INP_WLOCK(inp); 776 socantsendmore(so); 777 INP_WUNLOCK(inp); 778 return 0; 779} 780 781static int 782rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 783 struct mbuf *control, struct thread *td) 784{ 785 struct inpcb *inp; 786 u_long dst; 787 788 inp = sotoinpcb(so); 789 KASSERT(inp != NULL, ("rip_send: inp == NULL")); 790 /* 791 * Note: 'dst' reads below are unlocked. 792 */ 793 if (so->so_state & SS_ISCONNECTED) { 794 if (nam) { 795 m_freem(m); 796 return EISCONN; 797 } 798 dst = inp->inp_faddr.s_addr; /* Unlocked read. */ 799 } else { 800 if (nam == NULL) { 801 m_freem(m); 802 return ENOTCONN; 803 } 804 dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr; 805 } 806 return rip_output(m, so, dst); 807} 808 809static int 810rip_pcblist(SYSCTL_HANDLER_ARGS) 811{ 812 int error, i, n; 813 struct inpcb *inp, **inp_list; 814 inp_gen_t gencnt; 815 struct xinpgen xig; 816 817 /* 818 * The process of preparing the TCB list is too time-consuming and 819 * resource-intensive to repeat twice on every request. 820 */ 821 if (req->oldptr == 0) { 822 n = ripcbinfo.ipi_count; 823 req->oldidx = 2 * (sizeof xig) 824 + (n + n/8) * sizeof(struct xinpcb); 825 return 0; 826 } 827 828 if (req->newptr != 0) 829 return EPERM; 830 831 /* 832 * OK, now we're committed to doing something. 833 */ 834 INP_INFO_RLOCK(&ripcbinfo); 835 gencnt = ripcbinfo.ipi_gencnt; 836 n = ripcbinfo.ipi_count; 837 INP_INFO_RUNLOCK(&ripcbinfo); 838 839 xig.xig_len = sizeof xig; 840 xig.xig_count = n; 841 xig.xig_gen = gencnt; 842 xig.xig_sogen = so_gencnt; 843 error = SYSCTL_OUT(req, &xig, sizeof xig); 844 if (error) 845 return error; 846 847 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 848 if (inp_list == 0) 849 return ENOMEM; 850 851 INP_INFO_RLOCK(&ripcbinfo); 852 for (inp = LIST_FIRST(ripcbinfo.ipi_listhead), i = 0; inp && i < n; 853 inp = LIST_NEXT(inp, inp_list)) { 854 INP_WLOCK(inp); 855 if (inp->inp_gencnt <= gencnt && 856 cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0) { 857 /* XXX held references? */ 858 inp_list[i++] = inp; 859 } 860 INP_WUNLOCK(inp); 861 } 862 INP_INFO_RUNLOCK(&ripcbinfo); 863 n = i; 864 865 error = 0; 866 for (i = 0; i < n; i++) { 867 inp = inp_list[i]; 868 INP_WLOCK(inp); 869 if (inp->inp_gencnt <= gencnt) { 870 struct xinpcb xi; 871 bzero(&xi, sizeof(xi)); 872 xi.xi_len = sizeof xi; 873 /* XXX should avoid extra copy */ 874 bcopy(inp, &xi.xi_inp, sizeof *inp); 875 if (inp->inp_socket) 876 sotoxsocket(inp->inp_socket, &xi.xi_socket); 877 INP_WUNLOCK(inp); 878 error = SYSCTL_OUT(req, &xi, sizeof xi); 879 } else 880 INP_WUNLOCK(inp); 881 } 882 if (!error) { 883 /* 884 * Give the user an updated idea of our state. 885 * If the generation differs from what we told 886 * her before, she knows that something happened 887 * while we were processing this request, and it 888 * might be necessary to retry. 889 */ 890 INP_INFO_RLOCK(&ripcbinfo); 891 xig.xig_gen = ripcbinfo.ipi_gencnt; 892 xig.xig_sogen = so_gencnt; 893 xig.xig_count = ripcbinfo.ipi_count; 894 INP_INFO_RUNLOCK(&ripcbinfo); 895 error = SYSCTL_OUT(req, &xig, sizeof xig); 896 } 897 free(inp_list, M_TEMP); 898 return error; 899} 900 901SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist, CTLFLAG_RD, 0, 0, 902 rip_pcblist, "S,xinpcb", "List of active raw IP sockets"); 903 904struct pr_usrreqs rip_usrreqs = { 905 .pru_abort = rip_abort, 906 .pru_attach = rip_attach, 907 .pru_bind = rip_bind, 908 .pru_connect = rip_connect, 909 .pru_control = in_control, 910 .pru_detach = rip_detach, 911 .pru_disconnect = rip_disconnect, 912 .pru_peeraddr = in_getpeeraddr, 913 .pru_send = rip_send, 914 .pru_shutdown = rip_shutdown, 915 .pru_sockaddr = in_getsockaddr, 916 .pru_sosetlabel = in_pcbsosetlabel, 917 .pru_close = rip_close, 918}; 919