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 * 3. All advertising materials mentioning features or use of this software 14 * must display the following acknowledgement: 15 * This product includes software developed by the University of 16 * California, Berkeley and its contributors. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * $FreeBSD: src/sys/netinet/ip_divert.c,v 1.42.2.6 2003/01/23 21:06:45 sam Exp $ 34 */ 35 36 #include "opt_inet.h" 37 #include "opt_ipfw.h" 38 #include "opt_ipdivert.h" 39 #include "opt_ipsec.h" 40 41 #ifndef INET 42 #error "IPDIVERT requires INET." 43 #endif 44 45 #include <sys/param.h> 46 #include <sys/kernel.h> 47 #include <sys/malloc.h> 48 #include <sys/mbuf.h> 49 #include <sys/socket.h> 50 #include <sys/protosw.h> 51 #include <sys/socketvar.h> 52 #include <sys/sysctl.h> 53 #include <sys/systm.h> 54 #include <sys/proc.h> 55 56 #include <vm/vm_zone.h> 57 58 #include <net/if.h> 59 #include <net/route.h> 60 61 #include <netinet/in.h> 62 #include <netinet/in_systm.h> 63 #include <netinet/ip.h> 64 #include <netinet/in_pcb.h> 65 #include <netinet/in_var.h> 66 #include <netinet/ip_var.h> 67 68 /* 69 * Divert sockets 70 */ 71 72 /* 73 * Allocate enough space to hold a full IP packet 74 */ 75 #define DIVSNDQ (65536 + 100) 76 #define DIVRCVQ (65536 + 100) 77 78 /* 79 * Divert sockets work in conjunction with ipfw, see the divert(4) 80 * manpage for features. 81 * Internally, packets selected by ipfw in ip_input() or ip_output(), 82 * and never diverted before, are passed to the input queue of the 83 * divert socket with a given 'divert_port' number (as specified in 84 * the matching ipfw rule), and they are tagged with a 16 bit cookie 85 * (representing the rule number of the matching ipfw rule), which 86 * is passed to process reading from the socket. 87 * 88 * Packets written to the divert socket are again tagged with a cookie 89 * (usually the same as above) and a destination address. 90 * If the destination address is INADDR_ANY then the packet is 91 * treated as outgoing and sent to ip_output(), otherwise it is 92 * treated as incoming and sent to ip_input(). 93 * In both cases, the packet is tagged with the cookie. 94 * 95 * On reinjection, processing in ip_input() and ip_output() 96 * will be exactly the same as for the original packet, except that 97 * ipfw processing will start at the rule number after the one 98 * written in the cookie (so, tagging a packet with a cookie of 0 99 * will cause it to be effectively considered as a standard packet). 100 */ 101 102 /* Internal variables */ 103 static struct inpcbhead divcb; 104 static struct inpcbinfo divcbinfo; 105 106 static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */ 107 static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */ 108 109 /* Optimization: have this preinitialized */ 110 static struct sockaddr_in divsrc = { sizeof(divsrc), AF_INET }; 111 112 /* 113 * Initialize divert connection block queue. 114 */ 115 void 116 div_init(void) 117 { 118 LIST_INIT(&divcb); 119 divcbinfo.listhead = &divcb; 120 /* 121 * XXX We don't use the hash list for divert IP, but it's easier 122 * to allocate a one entry hash list than it is to check all 123 * over the place for hashbase == NULL. 124 */ 125 divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask); 126 divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask); 127 divcbinfo.ipi_zone = zinit("divcb", sizeof(struct inpcb), 128 maxsockets, ZONE_INTERRUPT, 0); 129 } 130 131 /* 132 * IPPROTO_DIVERT is not a real IP protocol; don't allow any packets 133 * with that protocol number to enter the system from the outside. 134 */ 135 void 136 div_input(struct mbuf *m, int off, int proto) 137 { 138 ipstat.ips_noproto++; 139 m_freem(m); 140 } 141 142 /* 143 * Divert a packet by passing it up to the divert socket at port 'port'. 144 * 145 * Setup generic address and protocol structures for div_input routine, 146 * then pass them along with mbuf chain. 147 */ 148 void 149 divert_packet(struct mbuf *m, int incoming, int port, int rule) 150 { 151 struct ip *ip; 152 struct inpcb *inp; 153 struct socket *sa; 154 u_int16_t nport; 155 156 /* Sanity check */ 157 KASSERT(port != 0, ("%s: port=0", __FUNCTION__)); 158 159 divsrc.sin_port = rule; /* record matching rule */ 160 161 /* Assure header */ 162 if (m->m_len < sizeof(struct ip) && 163 (m = m_pullup(m, sizeof(struct ip))) == 0) 164 return; 165 ip = mtod(m, struct ip *); 166 167 /* 168 * Record receive interface address, if any. 169 * But only for incoming packets. 170 */ 171 divsrc.sin_addr.s_addr = 0; 172 if (incoming) { 173 struct ifaddr *ifa; 174 175 /* Sanity check */ 176 KASSERT((m->m_flags & M_PKTHDR), ("%s: !PKTHDR", __FUNCTION__)); 177 178 /* Find IP address for receive interface */ 179 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 180 if (ifa->ifa_addr == NULL) 181 continue; 182 if (ifa->ifa_addr->sa_family != AF_INET) 183 continue; 184 divsrc.sin_addr = 185 ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr; 186 break; 187 } 188 } 189 /* 190 * Record the incoming interface name whenever we have one. 191 */ 192 bzero(&divsrc.sin_zero, sizeof(divsrc.sin_zero)); 193 if (m->m_pkthdr.rcvif) { 194 /* 195 * Hide the actual interface name in there in the 196 * sin_zero array. XXX This needs to be moved to a 197 * different sockaddr type for divert, e.g. 198 * sockaddr_div with multiple fields like 199 * sockaddr_dl. Presently we have only 7 bytes 200 * but that will do for now as most interfaces 201 * are 4 or less + 2 or less bytes for unit. 202 * There is probably a faster way of doing this, 203 * possibly taking it from the sockaddr_dl on the iface. 204 * This solves the problem of a P2P link and a LAN interface 205 * having the same address, which can result in the wrong 206 * interface being assigned to the packet when fed back 207 * into the divert socket. Theoretically if the daemon saves 208 * and re-uses the sockaddr_in as suggested in the man pages, 209 * this iface name will come along for the ride. 210 * (see div_output for the other half of this.) 211 */ 212 snprintf(divsrc.sin_zero, sizeof(divsrc.sin_zero), 213 "%s%d", m->m_pkthdr.rcvif->if_name, 214 m->m_pkthdr.rcvif->if_unit); 215 } 216 217 /* Put packet on socket queue, if any */ 218 sa = NULL; 219 nport = htons((u_int16_t)port); 220 LIST_FOREACH(inp, &divcb, inp_list) { 221 if (inp->inp_lport == nport) 222 sa = inp->inp_socket; 223 } 224 if (sa) { 225 if (sbappendaddr(&sa->so_rcv, (struct sockaddr *)&divsrc, 226 m, (struct mbuf *)0) == 0) 227 m_freem(m); 228 else 229 sorwakeup(sa); 230 } else { 231 m_freem(m); 232 ipstat.ips_noproto++; 233 ipstat.ips_delivered--; 234 } 235 } 236 237 /* 238 * Deliver packet back into the IP processing machinery. 239 * 240 * If no address specified, or address is 0.0.0.0, send to ip_output(); 241 * otherwise, send to ip_input() and mark as having been received on 242 * the interface with that address. 243 */ 244 static int 245 div_output(struct socket *so, struct mbuf *m, 246 struct sockaddr_in *sin, struct mbuf *control) 247 { 248 int error = 0; 249 struct m_hdr divert_tag; 250 251 /* 252 * Prepare the tag for divert info. Note that a packet 253 * with a 0 tag in mh_data is effectively untagged, 254 * so we could optimize that case. 255 */ 256 divert_tag.mh_type = MT_TAG; 257 divert_tag.mh_flags = PACKET_TAG_DIVERT; 258 divert_tag.mh_next = m; 259 divert_tag.mh_data = 0; /* the matching rule # */ 260 m->m_pkthdr.rcvif = NULL; /* XXX is it necessary ? */ 261 262 if (control) 263 m_freem(control); /* XXX */ 264 265 /* Loopback avoidance and state recovery */ 266 if (sin) { 267 int i; 268 269 divert_tag.mh_data = (caddr_t)(int)sin->sin_port; 270 /* 271 * Find receive interface with the given name, stuffed 272 * (if it exists) in the sin_zero[] field. 273 * The name is user supplied data so don't trust its size 274 * or that it is zero terminated. 275 */ 276 for (i = 0; sin->sin_zero[i] && i < sizeof(sin->sin_zero); i++) 277 ; 278 if ( i > 0 && i < sizeof(sin->sin_zero)) 279 m->m_pkthdr.rcvif = ifunit(sin->sin_zero); 280 } 281 282 /* Reinject packet into the system as incoming or outgoing */ 283 if (!sin || sin->sin_addr.s_addr == 0) { 284 struct inpcb *const inp = sotoinpcb(so); 285 struct ip *const ip = mtod(m, struct ip *); 286 287 /* 288 * Don't allow both user specified and setsockopt options, 289 * and don't allow packet length sizes that will crash 290 */ 291 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) || 292 ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) { 293 error = EINVAL; 294 goto cantsend; 295 } 296 297 /* Convert fields to host order for ip_output() */ 298 ip->ip_len = ntohs(ip->ip_len); 299 ip->ip_off = ntohs(ip->ip_off); 300 301 /* Send packet to output processing */ 302 ipstat.ips_rawout++; /* XXX */ 303 error = ip_output((struct mbuf *)&divert_tag, 304 inp->inp_options, &inp->inp_route, 305 (so->so_options & SO_DONTROUTE) | 306 IP_ALLOWBROADCAST | IP_RAWOUTPUT, 307 inp->inp_moptions, NULL); 308 } else { 309 if (m->m_pkthdr.rcvif == NULL) { 310 /* 311 * No luck with the name, check by IP address. 312 * Clear the port and the ifname to make sure 313 * there are no distractions for ifa_ifwithaddr. 314 */ 315 struct ifaddr *ifa; 316 317 bzero(sin->sin_zero, sizeof(sin->sin_zero)); 318 sin->sin_port = 0; 319 ifa = ifa_ifwithaddr((struct sockaddr *) sin); 320 if (ifa == NULL) { 321 error = EADDRNOTAVAIL; 322 goto cantsend; 323 } 324 m->m_pkthdr.rcvif = ifa->ifa_ifp; 325 } 326 /* Send packet to input processing */ 327 ip_input((struct mbuf *)&divert_tag); 328 } 329 330 return error; 331 332 cantsend: 333 m_freem(m); 334 return error; 335 } 336 337 static int 338 div_attach(struct socket *so, int proto, struct proc *p) 339 { 340 struct inpcb *inp; 341 int error, s; 342 343 inp = sotoinpcb(so); 344 if (inp) 345 panic("div_attach"); 346 if (p && (error = suser(p)) != 0) 347 return error; 348 349 error = soreserve(so, div_sendspace, div_recvspace); 350 if (error) 351 return error; 352 s = splnet(); 353 error = in_pcballoc(so, &divcbinfo, p); 354 splx(s); 355 if (error) 356 return error; 357 inp = (struct inpcb *)so->so_pcb; 358 inp->inp_ip_p = proto; 359 inp->inp_vflag |= INP_IPV4; 360 inp->inp_flags |= INP_HDRINCL; 361 /* The socket is always "connected" because 362 we always know "where" to send the packet */ 363 so->so_state |= SS_ISCONNECTED; 364 return 0; 365 } 366 367 static int 368 div_detach(struct socket *so) 369 { 370 struct inpcb *inp; 371 372 inp = sotoinpcb(so); 373 if (inp == 0) 374 panic("div_detach"); 375 in_pcbdetach(inp); 376 return 0; 377 } 378 379 static int 380 div_abort(struct socket *so) 381 { 382 soisdisconnected(so); 383 return div_detach(so); 384 } 385 386 static int 387 div_disconnect(struct socket *so) 388 { 389 if ((so->so_state & SS_ISCONNECTED) == 0) 390 return ENOTCONN; 391 return div_abort(so); 392 } 393 394 static int 395 div_bind(struct socket *so, struct sockaddr *nam, struct proc *p) 396 { 397 struct inpcb *inp; 398 int s; 399 int error; 400 401 s = splnet(); 402 inp = sotoinpcb(so); 403 /* in_pcbbind assumes that nam is a sockaddr_in 404 * and in_pcbbind requires a valid address. Since divert 405 * sockets don't we need to make sure the address is 406 * filled in properly. 407 * XXX -- divert should not be abusing in_pcbind 408 * and should probably have its own family. 409 */ 410 if (nam->sa_family != AF_INET) 411 error = EAFNOSUPPORT; 412 else { 413 ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY; 414 error = in_pcbbind(inp, nam, p); 415 } 416 splx(s); 417 return error; 418 } 419 420 static int 421 div_shutdown(struct socket *so) 422 { 423 socantsendmore(so); 424 return 0; 425 } 426 427 static int 428 div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 429 struct mbuf *control, struct proc *p) 430 { 431 /* Packet must have a header (but that's about it) */ 432 if (m->m_len < sizeof (struct ip) && 433 (m = m_pullup(m, sizeof (struct ip))) == 0) { 434 ipstat.ips_toosmall++; 435 m_freem(m); 436 return EINVAL; 437 } 438 439 /* Send packet */ 440 return div_output(so, m, (struct sockaddr_in *)nam, control); 441 } 442 443 static int 444 div_pcblist(SYSCTL_HANDLER_ARGS) 445 { 446 int error, i, n, s; 447 struct inpcb *inp, **inp_list; 448 inp_gen_t gencnt; 449 struct xinpgen xig; 450 451 /* 452 * The process of preparing the TCB list is too time-consuming and 453 * resource-intensive to repeat twice on every request. 454 */ 455 if (req->oldptr == 0) { 456 n = divcbinfo.ipi_count; 457 req->oldidx = 2 * (sizeof xig) 458 + (n + n/8) * sizeof(struct xinpcb); 459 return 0; 460 } 461 462 if (req->newptr != 0) 463 return EPERM; 464 465 /* 466 * OK, now we're committed to doing something. 467 */ 468 s = splnet(); 469 gencnt = divcbinfo.ipi_gencnt; 470 n = divcbinfo.ipi_count; 471 splx(s); 472 473 xig.xig_len = sizeof xig; 474 xig.xig_count = n; 475 xig.xig_gen = gencnt; 476 xig.xig_sogen = so_gencnt; 477 error = SYSCTL_OUT(req, &xig, sizeof xig); 478 if (error) 479 return error; 480 481 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 482 if (inp_list == 0) 483 return ENOMEM; 484 485 s = splnet(); 486 for (inp = LIST_FIRST(divcbinfo.listhead), i = 0; inp && i < n; 487 inp = LIST_NEXT(inp, inp_list)) { 488 if (inp->inp_gencnt <= gencnt && !prison_xinpcb(req->p, inp)) 489 inp_list[i++] = inp; 490 } 491 splx(s); 492 n = i; 493 494 error = 0; 495 for (i = 0; i < n; i++) { 496 inp = inp_list[i]; 497 if (inp->inp_gencnt <= gencnt) { 498 struct xinpcb xi; 499 xi.xi_len = sizeof xi; 500 /* XXX should avoid extra copy */ 501 bcopy(inp, &xi.xi_inp, sizeof *inp); 502 if (inp->inp_socket) 503 sotoxsocket(inp->inp_socket, &xi.xi_socket); 504 error = SYSCTL_OUT(req, &xi, sizeof xi); 505 } 506 } 507 if (!error) { 508 /* 509 * Give the user an updated idea of our state. 510 * If the generation differs from what we told 511 * her before, she knows that something happened 512 * while we were processing this request, and it 513 * might be necessary to retry. 514 */ 515 s = splnet(); 516 xig.xig_gen = divcbinfo.ipi_gencnt; 517 xig.xig_sogen = so_gencnt; 518 xig.xig_count = divcbinfo.ipi_count; 519 splx(s); 520 error = SYSCTL_OUT(req, &xig, sizeof xig); 521 } 522 free(inp_list, M_TEMP); 523 return error; 524 } 525 526 SYSCTL_DECL(_net_inet_divert); 527 SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLFLAG_RD, 0, 0, 528 div_pcblist, "S,xinpcb", "List of active divert sockets"); 529 530 struct pr_usrreqs div_usrreqs = { 531 div_abort, pru_accept_notsupp, div_attach, div_bind, 532 pru_connect_notsupp, pru_connect2_notsupp, in_control, div_detach, 533 div_disconnect, pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp, 534 pru_rcvoob_notsupp, div_send, pru_sense_null, div_shutdown, 535 in_setsockaddr, sosend, soreceive, sopoll 536 }; 537