raw_ip.c revision 178888
1214501Srpaulo/*- 2214501Srpaulo * Copyright (c) 1982, 1986, 1988, 1993 3252726Srpaulo * The Regents of the University of California. All rights reserved. 4214501Srpaulo * 5252726Srpaulo * Redistribution and use in source and binary forms, with or without 6252726Srpaulo * modification, are permitted provided that the following conditions 7214501Srpaulo * are met: 8214501Srpaulo * 1. Redistributions of source code must retain the above copyright 9214501Srpaulo * notice, this list of conditions and the following disclaimer. 10214501Srpaulo * 2. Redistributions in binary form must reproduce the above copyright 11214501Srpaulo * notice, this list of conditions and the following disclaimer in the 12214501Srpaulo * documentation and/or other materials provided with the distribution. 13214501Srpaulo * 4. Neither the name of the University nor the names of its contributors 14214501Srpaulo * may be used to endorse or promote products derived from this software 15214501Srpaulo * without specific prior written permission. 16214501Srpaulo * 17214501Srpaulo * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18214501Srpaulo * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19214501Srpaulo * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20214501Srpaulo * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21214501Srpaulo * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22214501Srpaulo * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23214501Srpaulo * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24214501Srpaulo * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25214501Srpaulo * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26214501Srpaulo * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27214501Srpaulo * SUCH DAMAGE. 28214501Srpaulo * 29214501Srpaulo * @(#)raw_ip.c 8.7 (Berkeley) 5/15/95 30214501Srpaulo */ 31214501Srpaulo 32214501Srpaulo#include <sys/cdefs.h> 33214501Srpaulo__FBSDID("$FreeBSD: head/sys/netinet/raw_ip.c 178888 2008-05-09 23:03:00Z julian $"); 34214501Srpaulo 35214501Srpaulo#include "opt_inet6.h" 36214501Srpaulo#include "opt_ipsec.h" 37214501Srpaulo#include "opt_mac.h" 38214501Srpaulo 39214501Srpaulo#include <sys/param.h> 40214501Srpaulo#include <sys/jail.h> 41214501Srpaulo#include <sys/kernel.h> 42214501Srpaulo#include <sys/lock.h> 43252726Srpaulo#include <sys/malloc.h> 44214501Srpaulo#include <sys/mbuf.h> 45214501Srpaulo#include <sys/priv.h> 46214501Srpaulo#include <sys/proc.h> 47214501Srpaulo#include <sys/protosw.h> 48214501Srpaulo#include <sys/signalvar.h> 49214501Srpaulo#include <sys/socket.h> 50214501Srpaulo#include <sys/socketvar.h> 51214501Srpaulo#include <sys/sx.h> 52214501Srpaulo#include <sys/sysctl.h> 53214501Srpaulo#include <sys/systm.h> 54214501Srpaulo 55214501Srpaulo#include <vm/uma.h> 56214501Srpaulo 57214501Srpaulo#include <net/if.h> 58214501Srpaulo#include <net/route.h> 59214501Srpaulo 60214501Srpaulo#include <netinet/in.h> 61214501Srpaulo#include <netinet/in_systm.h> 62214501Srpaulo#include <netinet/in_pcb.h> 63214501Srpaulo#include <netinet/in_var.h> 64214501Srpaulo#include <netinet/ip.h> 65214501Srpaulo#include <netinet/ip_var.h> 66214501Srpaulo#include <netinet/ip_mroute.h> 67214501Srpaulo 68214501Srpaulo#include <netinet/ip_fw.h> 69214501Srpaulo#include <netinet/ip_dummynet.h> 70214501Srpaulo 71214501Srpaulo#ifdef IPSEC 72214501Srpaulo#include <netipsec/ipsec.h> 73214501Srpaulo#endif /*IPSEC*/ 74214501Srpaulo 75214501Srpaulo#include <security/mac/mac_framework.h> 76214501Srpaulo 77214501Srpaulostruct inpcbhead ripcb; 78214501Srpaulostruct inpcbinfo ripcbinfo; 79214501Srpaulo 80214501Srpaulo/* control hooks for ipfw and dummynet */ 81214501Srpauloip_fw_ctl_t *ip_fw_ctl_ptr = NULL; 82214501Srpauloip_dn_ctl_t *ip_dn_ctl_ptr = NULL; 83214501Srpaulo 84214501Srpaulo/* 85214501Srpaulo * hooks for multicast routing. They all default to NULL, 86214501Srpaulo * so leave them not initialized and rely on BSS being set to 0. 87214501Srpaulo */ 88214501Srpaulo 89214501Srpaulo/* The socket used to communicate with the multicast routing daemon. */ 90214501Srpaulostruct socket *ip_mrouter; 91214501Srpaulo 92214501Srpaulo/* The various mrouter and rsvp functions */ 93214501Srpauloint (*ip_mrouter_set)(struct socket *, struct sockopt *); 94214501Srpauloint (*ip_mrouter_get)(struct socket *, struct sockopt *); 95214501Srpauloint (*ip_mrouter_done)(void); 96214501Srpauloint (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 97214501Srpaulo struct ip_moptions *); 98214501Srpauloint (*mrt_ioctl)(int, caddr_t, int); 99252726Srpauloint (*legal_vif_num)(int); 100252726Srpaulou_long (*ip_mcast_src)(int); 101214501Srpaulo 102214501Srpaulovoid (*rsvp_input_p)(struct mbuf *m, int off); 103214501Srpauloint (*ip_rsvp_vif)(struct socket *, struct sockopt *); 104214501Srpaulovoid (*ip_rsvp_force_done)(struct socket *); 105214501Srpaulo 106214501Srpaulo/* 107214501Srpaulo * Raw interface to IP protocol. 108214501Srpaulo */ 109214501Srpaulo 110214501Srpaulo/* 111214501Srpaulo * Initialize raw connection block q. 112214501Srpaulo */ 113214501Srpaulostatic void 114214501Srpaulorip_zone_change(void *tag) 115214501Srpaulo{ 116214501Srpaulo 117214501Srpaulo uma_zone_set_max(ripcbinfo.ipi_zone, maxsockets); 118214501Srpaulo} 119214501Srpaulo 120214501Srpaulostatic int 121214501Srpaulorip_inpcb_init(void *mem, int size, int flags) 122214501Srpaulo{ 123214501Srpaulo struct inpcb *inp = mem; 124214501Srpaulo 125214501Srpaulo INP_LOCK_INIT(inp, "inp", "rawinp"); 126214501Srpaulo return (0); 127214501Srpaulo} 128214501Srpaulo 129214501Srpaulovoid 130214501Srpaulorip_init(void) 131214501Srpaulo{ 132214501Srpaulo 133214501Srpaulo INP_INFO_LOCK_INIT(&ripcbinfo, "rip"); 134214501Srpaulo LIST_INIT(&ripcb); 135214501Srpaulo ripcbinfo.ipi_listhead = &ripcb; 136214501Srpaulo /* 137214501Srpaulo * XXX We don't use the hash list for raw IP, but it's easier 138252726Srpaulo * to allocate a one entry hash list than it is to check all 139252726Srpaulo * over the place for hashbase == NULL. 140252726Srpaulo */ 141214501Srpaulo ripcbinfo.ipi_hashbase = hashinit(1, M_PCB, &ripcbinfo.ipi_hashmask); 142214501Srpaulo ripcbinfo.ipi_porthashbase = hashinit(1, M_PCB, 143214501Srpaulo &ripcbinfo.ipi_porthashmask); 144214501Srpaulo ripcbinfo.ipi_zone = uma_zcreate("ripcb", sizeof(struct inpcb), 145214501Srpaulo NULL, NULL, rip_inpcb_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 146214501Srpaulo uma_zone_set_max(ripcbinfo.ipi_zone, maxsockets); 147214501Srpaulo EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, 148214501Srpaulo NULL, EVENTHANDLER_PRI_ANY); 149214501Srpaulo} 150214501Srpaulo 151214501Srpaulostatic struct sockaddr_in ripsrc = { sizeof(ripsrc), AF_INET }; 152214501Srpaulo 153214501Srpaulostatic int 154214501Srpauloraw_append(struct inpcb *last, struct ip *ip, struct mbuf *n) 155214501Srpaulo{ 156214501Srpaulo int policyfail = 0; 157214501Srpaulo 158214501Srpaulo INP_RLOCK_ASSERT(last); 159214501Srpaulo 160214501Srpaulo#ifdef IPSEC 161214501Srpaulo /* check AH/ESP integrity. */ 162214501Srpaulo if (ipsec4_in_reject(n, last)) { 163214501Srpaulo policyfail = 1; 164214501Srpaulo } 165214501Srpaulo#endif /* IPSEC */ 166214501Srpaulo#ifdef MAC 167214501Srpaulo if (!policyfail && mac_inpcb_check_deliver(last, n) != 0) 168252726Srpaulo policyfail = 1; 169252726Srpaulo#endif 170252726Srpaulo /* Check the minimum TTL for socket. */ 171252726Srpaulo if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl) 172252726Srpaulo policyfail = 1; 173214501Srpaulo if (!policyfail) { 174214501Srpaulo struct mbuf *opts = NULL; 175214501Srpaulo struct socket *so; 176214501Srpaulo 177214501Srpaulo so = last->inp_socket; 178214501Srpaulo if ((last->inp_flags & INP_CONTROLOPTS) || 179214501Srpaulo (so->so_options & (SO_TIMESTAMP | SO_BINTIME))) 180214501Srpaulo ip_savecontrol(last, &opts, ip, n); 181214501Srpaulo SOCKBUF_LOCK(&so->so_rcv); 182214501Srpaulo if (sbappendaddr_locked(&so->so_rcv, 183214501Srpaulo (struct sockaddr *)&ripsrc, n, opts) == 0) { 184214501Srpaulo /* should notify about lost packet */ 185214501Srpaulo m_freem(n); 186214501Srpaulo if (opts) 187214501Srpaulo m_freem(opts); 188214501Srpaulo SOCKBUF_UNLOCK(&so->so_rcv); 189214501Srpaulo } else 190214501Srpaulo sorwakeup_locked(so); 191214501Srpaulo } else 192214501Srpaulo m_freem(n); 193214501Srpaulo return policyfail; 194214501Srpaulo} 195214501Srpaulo 196214501Srpaulo/* 197214501Srpaulo * Setup generic address and protocol structures 198214501Srpaulo * for raw_input routine, then pass them along with 199214501Srpaulo * mbuf chain. 200214501Srpaulo */ 201214501Srpaulovoid 202214501Srpaulorip_input(struct mbuf *m, int off) 203214501Srpaulo{ 204214501Srpaulo struct ip *ip = mtod(m, struct ip *); 205214501Srpaulo int proto = ip->ip_p; 206214501Srpaulo struct inpcb *inp, *last; 207214501Srpaulo 208214501Srpaulo INP_INFO_RLOCK(&ripcbinfo); 209214501Srpaulo ripsrc.sin_addr = ip->ip_src; 210214501Srpaulo last = NULL; 211214501Srpaulo LIST_FOREACH(inp, &ripcb, inp_list) { 212252726Srpaulo INP_RLOCK(inp); 213252726Srpaulo if (inp->inp_ip_p && inp->inp_ip_p != proto) { 214214501Srpaulo docontinue: 215214501Srpaulo INP_RUNLOCK(inp); 216214501Srpaulo continue; 217214501Srpaulo } 218214501Srpaulo#ifdef INET6 219214501Srpaulo if ((inp->inp_vflag & INP_IPV4) == 0) 220214501Srpaulo goto docontinue; 221214501Srpaulo#endif 222214501Srpaulo if (inp->inp_laddr.s_addr && 223214501Srpaulo inp->inp_laddr.s_addr != ip->ip_dst.s_addr) 224214501Srpaulo goto docontinue; 225214501Srpaulo if (inp->inp_faddr.s_addr && 226214501Srpaulo inp->inp_faddr.s_addr != ip->ip_src.s_addr) 227214501Srpaulo goto docontinue; 228214501Srpaulo if (jailed(inp->inp_socket->so_cred)) 229214501Srpaulo if (htonl(prison_getip(inp->inp_socket->so_cred)) != 230214501Srpaulo ip->ip_dst.s_addr) 231214501Srpaulo goto docontinue; 232214501Srpaulo if (last) { 233214501Srpaulo struct mbuf *n; 234214501Srpaulo 235214501Srpaulo n = m_copy(m, 0, (int)M_COPYALL); 236214501Srpaulo if (n != NULL) 237214501Srpaulo (void) raw_append(last, ip, n); 238214501Srpaulo /* XXX count dropped packet */ 239214501Srpaulo INP_RUNLOCK(last); 240214501Srpaulo } 241214501Srpaulo last = inp; 242214501Srpaulo } 243214501Srpaulo if (last != NULL) { 244214501Srpaulo if (raw_append(last, ip, m) != 0) 245214501Srpaulo ipstat.ips_delivered--; 246214501Srpaulo INP_RUNLOCK(last); 247214501Srpaulo } else { 248214501Srpaulo m_freem(m); 249214501Srpaulo ipstat.ips_noproto++; 250214501Srpaulo ipstat.ips_delivered--; 251214501Srpaulo } 252214501Srpaulo INP_INFO_RUNLOCK(&ripcbinfo); 253214501Srpaulo} 254214501Srpaulo 255214501Srpaulo/* 256214501Srpaulo * Generate IP header and pass packet to ip_output. 257214501Srpaulo * Tack on options user may have setup with control call. 258214501Srpaulo */ 259214501Srpauloint 260214501Srpaulorip_output(struct mbuf *m, struct socket *so, u_long dst) 261214501Srpaulo{ 262214501Srpaulo struct ip *ip; 263214501Srpaulo int error; 264214501Srpaulo struct inpcb *inp = sotoinpcb(so); 265214501Srpaulo int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) | 266214501Srpaulo IP_ALLOWBROADCAST; 267214501Srpaulo 268214501Srpaulo /* 269214501Srpaulo * If the user handed us a complete IP packet, use it. 270214501Srpaulo * Otherwise, allocate an mbuf for a header and fill it in. 271214501Srpaulo */ 272214501Srpaulo if ((inp->inp_flags & INP_HDRINCL) == 0) { 273214501Srpaulo if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) { 274214501Srpaulo m_freem(m); 275214501Srpaulo return(EMSGSIZE); 276214501Srpaulo } 277214501Srpaulo M_PREPEND(m, sizeof(struct ip), M_DONTWAIT); 278214501Srpaulo if (m == NULL) 279214501Srpaulo return(ENOBUFS); 280214501Srpaulo 281214501Srpaulo INP_RLOCK(inp); 282214501Srpaulo ip = mtod(m, struct ip *); 283214501Srpaulo ip->ip_tos = inp->inp_ip_tos; 284214501Srpaulo if (inp->inp_flags & INP_DONTFRAG) 285214501Srpaulo ip->ip_off = IP_DF; 286214501Srpaulo else 287214501Srpaulo ip->ip_off = 0; 288214501Srpaulo ip->ip_p = inp->inp_ip_p; 289214501Srpaulo ip->ip_len = m->m_pkthdr.len; 290214501Srpaulo if (jailed(inp->inp_socket->so_cred)) 291214501Srpaulo ip->ip_src.s_addr = 292214501Srpaulo htonl(prison_getip(inp->inp_socket->so_cred)); 293214501Srpaulo else 294214501Srpaulo ip->ip_src = inp->inp_laddr; 295214501Srpaulo ip->ip_dst.s_addr = dst; 296214501Srpaulo ip->ip_ttl = inp->inp_ip_ttl; 297214501Srpaulo } else { 298214501Srpaulo if (m->m_pkthdr.len > IP_MAXPACKET) { 299214501Srpaulo m_freem(m); 300214501Srpaulo return(EMSGSIZE); 301214501Srpaulo } 302214501Srpaulo INP_RLOCK(inp); 303214501Srpaulo ip = mtod(m, struct ip *); 304214501Srpaulo if (jailed(inp->inp_socket->so_cred)) { 305214501Srpaulo if (ip->ip_src.s_addr != 306214501Srpaulo htonl(prison_getip(inp->inp_socket->so_cred))) { 307214501Srpaulo INP_RUNLOCK(inp); 308214501Srpaulo m_freem(m); 309214501Srpaulo return (EPERM); 310214501Srpaulo } 311252726Srpaulo } 312252726Srpaulo /* don't allow both user specified and setsockopt options, 313214501Srpaulo and don't allow packet length sizes that will crash */ 314214501Srpaulo if (((ip->ip_hl != (sizeof (*ip) >> 2)) 315214501Srpaulo && inp->inp_options) 316214501Srpaulo || (ip->ip_len > m->m_pkthdr.len) 317214501Srpaulo || (ip->ip_len < (ip->ip_hl << 2))) { 318214501Srpaulo INP_RUNLOCK(inp); 319214501Srpaulo m_freem(m); 320214501Srpaulo return EINVAL; 321214501Srpaulo } 322214501Srpaulo if (ip->ip_id == 0) 323214501Srpaulo ip->ip_id = ip_newid(); 324214501Srpaulo /* XXX prevent ip_output from overwriting header fields */ 325214501Srpaulo flags |= IP_RAWOUTPUT; 326214501Srpaulo ipstat.ips_rawout++; 327214501Srpaulo } 328214501Srpaulo 329214501Srpaulo if (inp->inp_flags & INP_ONESBCAST) 330214501Srpaulo flags |= IP_SENDONES; 331214501Srpaulo 332214501Srpaulo#ifdef MAC 333214501Srpaulo mac_inpcb_create_mbuf(inp, m); 334214501Srpaulo#endif 335214501Srpaulo 336214501Srpaulo error = ip_output(m, inp->inp_options, NULL, flags, 337214501Srpaulo inp->inp_moptions, inp); 338214501Srpaulo INP_RUNLOCK(inp); 339214501Srpaulo return error; 340214501Srpaulo} 341214501Srpaulo 342214501Srpaulo/* 343214501Srpaulo * Raw IP socket option processing. 344214501Srpaulo * 345214501Srpaulo * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could 346214501Srpaulo * only be created by a privileged process, and as such, socket option 347214501Srpaulo * operations to manage system properties on any raw socket were allowed to 348214501Srpaulo * take place without explicit additional access control checks. However, 349214501Srpaulo * raw sockets can now also be created in jail(), and therefore explicit 350214501Srpaulo * checks are now required. Likewise, raw sockets can be used by a process 351214501Srpaulo * after it gives up privilege, so some caution is required. For options 352214501Srpaulo * passed down to the IP layer via ip_ctloutput(), checks are assumed to be 353214501Srpaulo * performed in ip_ctloutput() and therefore no check occurs here. 354214501Srpaulo * Unilaterally checking priv_check() here breaks normal IP socket option 355214501Srpaulo * operations on raw sockets. 356214501Srpaulo * 357214501Srpaulo * When adding new socket options here, make sure to add access control 358214501Srpaulo * checks here as necessary. 359214501Srpaulo */ 360214501Srpauloint 361214501Srpaulorip_ctloutput(struct socket *so, struct sockopt *sopt) 362214501Srpaulo{ 363214501Srpaulo struct inpcb *inp = sotoinpcb(so); 364214501Srpaulo int error, optval; 365214501Srpaulo 366214501Srpaulo if (sopt->sopt_level != IPPROTO_IP) 367214501Srpaulo return (EINVAL); 368214501Srpaulo 369214501Srpaulo error = 0; 370214501Srpaulo switch (sopt->sopt_dir) { 371214501Srpaulo case SOPT_GET: 372214501Srpaulo switch (sopt->sopt_name) { 373214501Srpaulo case IP_HDRINCL: 374214501Srpaulo optval = inp->inp_flags & INP_HDRINCL; 375214501Srpaulo error = sooptcopyout(sopt, &optval, sizeof optval); 376214501Srpaulo break; 377214501Srpaulo 378214501Srpaulo case IP_FW_ADD: /* ADD actually returns the body... */ 379214501Srpaulo case IP_FW_GET: 380214501Srpaulo case IP_FW_TABLE_GETSIZE: 381214501Srpaulo case IP_FW_TABLE_LIST: 382214501Srpaulo case IP_FW_NAT_GET_CONFIG: 383214501Srpaulo case IP_FW_NAT_GET_LOG: 384214501Srpaulo /* 385214501Srpaulo * XXXRW: Isn't this checked one layer down? Yes, it 386214501Srpaulo * is. 387214501Srpaulo */ 388214501Srpaulo error = priv_check(curthread, PRIV_NETINET_IPFW); 389214501Srpaulo if (error != 0) 390214501Srpaulo return (error); 391214501Srpaulo if (ip_fw_ctl_ptr != NULL) 392214501Srpaulo error = ip_fw_ctl_ptr(sopt); 393214501Srpaulo else 394214501Srpaulo error = ENOPROTOOPT; 395214501Srpaulo break; 396214501Srpaulo 397214501Srpaulo case IP_DUMMYNET_GET: 398214501Srpaulo error = priv_check(curthread, PRIV_NETINET_DUMMYNET); 399214501Srpaulo if (error != 0) 400214501Srpaulo return (error); 401214501Srpaulo if (ip_dn_ctl_ptr != NULL) 402214501Srpaulo error = ip_dn_ctl_ptr(sopt); 403214501Srpaulo else 404214501Srpaulo error = ENOPROTOOPT; 405214501Srpaulo break ; 406214501Srpaulo 407214501Srpaulo case MRT_INIT: 408214501Srpaulo case MRT_DONE: 409214501Srpaulo case MRT_ADD_VIF: 410214501Srpaulo case MRT_DEL_VIF: 411214501Srpaulo case MRT_ADD_MFC: 412214501Srpaulo case MRT_DEL_MFC: 413214501Srpaulo case MRT_VERSION: 414214501Srpaulo case MRT_ASSERT: 415214501Srpaulo case MRT_API_SUPPORT: 416214501Srpaulo case MRT_API_CONFIG: 417214501Srpaulo case MRT_ADD_BW_UPCALL: 418214501Srpaulo case MRT_DEL_BW_UPCALL: 419214501Srpaulo error = priv_check(curthread, PRIV_NETINET_MROUTE); 420214501Srpaulo if (error != 0) 421214501Srpaulo return (error); 422214501Srpaulo error = ip_mrouter_get ? ip_mrouter_get(so, sopt) : 423214501Srpaulo EOPNOTSUPP; 424214501Srpaulo break; 425214501Srpaulo 426214501Srpaulo default: 427214501Srpaulo error = ip_ctloutput(so, sopt); 428214501Srpaulo break; 429214501Srpaulo } 430214501Srpaulo 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_RLOCK(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_RUNLOCK(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_RLOCK(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_RUNLOCK(inp); 878 error = SYSCTL_OUT(req, &xi, sizeof xi); 879 } else 880 INP_RUNLOCK(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