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 * $DragonFly: src/sys/netinet/ip_divert.c,v 1.14 2004/06/07 02:36:22 dillon Exp $ 35 */ 36 37 #include "opt_inet.h" 38 #include "opt_ipfw.h" 39 #include "opt_ipdivert.h" 40 #include "opt_ipsec.h" 41 42 #ifndef INET 43 #error "IPDIVERT requires INET." 44 #endif 45 46 #include <sys/param.h> 47 #include <sys/kernel.h> 48 #include <sys/malloc.h> 49 #include <sys/mbuf.h> 50 #include <sys/socket.h> 51 #include <sys/protosw.h> 52 #include <sys/socketvar.h> 53 #include <sys/sysctl.h> 54 #include <sys/systm.h> 55 #include <sys/proc.h> 56 57 #include <vm/vm_zone.h> 58 59 #include <net/if.h> 60 #include <net/route.h> 61 62 #include <netinet/in.h> 63 #include <netinet/in_systm.h> 64 #include <netinet/ip.h> 65 #include <netinet/in_pcb.h> 66 #include <netinet/in_var.h> 67 #include <netinet/ip_var.h> 68 69 /* 70 * Divert sockets 71 */ 72 73 /* 74 * Allocate enough space to hold a full IP packet 75 */ 76 #define DIVSNDQ (65536 + 100) 77 #define DIVRCVQ (65536 + 100) 78 79 /* 80 * Divert sockets work in conjunction with ipfw, see the divert(4) 81 * manpage for features. 82 * Internally, packets selected by ipfw in ip_input() or ip_output(), 83 * and never diverted before, are passed to the input queue of the 84 * divert socket with a given 'divert_port' number (as specified in 85 * the matching ipfw rule), and they are tagged with a 16 bit cookie 86 * (representing the rule number of the matching ipfw rule), which 87 * is passed to process reading from the socket. 88 * 89 * Packets written to the divert socket are again tagged with a cookie 90 * (usually the same as above) and a destination address. 91 * If the destination address is INADDR_ANY then the packet is 92 * treated as outgoing and sent to ip_output(), otherwise it is 93 * treated as incoming and sent to ip_input(). 94 * In both cases, the packet is tagged with the cookie. 95 * 96 * On reinjection, processing in ip_input() and ip_output() 97 * will be exactly the same as for the original packet, except that 98 * ipfw processing will start at the rule number after the one 99 * written in the cookie (so, tagging a packet with a cookie of 0 100 * will cause it to be effectively considered as a standard packet). 101 */ 102 103 /* Internal variables */ 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 in_pcbinfo_init(&divcbinfo); 119 /* 120 * XXX We don't use the hash list for divert IP, but it's easier 121 * to allocate a one entry hash list than it is to check all 122 * over the place for hashbase == NULL. 123 */ 124 divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask); 125 divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask); 126 divcbinfo.wildcardhashbase = hashinit(1, M_PCB, 127 &divcbinfo.wildcardhashmask); 128 divcbinfo.ipi_zone = zinit("divcb", sizeof(struct inpcb), 129 maxsockets, ZONE_INTERRUPT, 0); 130 } 131 132 /* 133 * IPPROTO_DIVERT is not a real IP protocol; don't allow any packets 134 * with that protocol number to enter the system from the outside. 135 */ 136 void 137 div_input(struct mbuf *m, ...) 138 { 139 ipstat.ips_noproto++; 140 m_freem(m); 141 } 142 143 /* 144 * Divert a packet by passing it up to the divert socket at port 'port'. 145 * 146 * Setup generic address and protocol structures for div_input routine, 147 * then pass them along with mbuf chain. 148 */ 149 void 150 divert_packet(struct mbuf *m, int incoming, int port, int rule) 151 { 152 struct ip *ip; 153 struct inpcb *inp; 154 struct socket *sa; 155 u_int16_t nport; 156 157 /* Sanity check */ 158 KASSERT(port != 0, ("%s: port=0", __FUNCTION__)); 159 160 divsrc.sin_port = rule; /* record matching rule */ 161 162 /* Assure header */ 163 if (m->m_len < sizeof(struct ip) && 164 (m = m_pullup(m, sizeof(struct ip))) == 0) 165 return; 166 ip = mtod(m, struct ip *); 167 168 /* 169 * Record receive interface address, if any. 170 * But only for incoming packets. 171 */ 172 divsrc.sin_addr.s_addr = 0; 173 if (incoming) { 174 struct ifaddr *ifa; 175 176 /* Sanity check */ 177 KASSERT((m->m_flags & M_PKTHDR), ("%s: !PKTHDR", __FUNCTION__)); 178 179 /* Find IP address for receive interface */ 180 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 181 if (ifa->ifa_addr == NULL) 182 continue; 183 if (ifa->ifa_addr->sa_family != AF_INET) 184 continue; 185 divsrc.sin_addr = 186 ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr; 187 break; 188 } 189 } 190 /* 191 * Record the incoming interface name whenever we have one. 192 */ 193 bzero(&divsrc.sin_zero, sizeof(divsrc.sin_zero)); 194 if (m->m_pkthdr.rcvif) { 195 /* 196 * Hide the actual interface name in there in the 197 * sin_zero array. XXX This needs to be moved to a 198 * different sockaddr type for divert, e.g. 199 * sockaddr_div with multiple fields like 200 * sockaddr_dl. Presently we have only 7 bytes 201 * but that will do for now as most interfaces 202 * are 4 or less + 2 or less bytes for unit. 203 * There is probably a faster way of doing this, 204 * possibly taking it from the sockaddr_dl on the iface. 205 * This solves the problem of a P2P link and a LAN interface 206 * having the same address, which can result in the wrong 207 * interface being assigned to the packet when fed back 208 * into the divert socket. Theoretically if the daemon saves 209 * and re-uses the sockaddr_in as suggested in the man pages, 210 * this iface name will come along for the ride. 211 * (see div_output for the other half of this.) 212 */ 213 snprintf(divsrc.sin_zero, sizeof(divsrc.sin_zero), 214 m->m_pkthdr.rcvif->if_xname); 215 } 216 217 /* Put packet on socket queue, if any */ 218 sa = NULL; 219 nport = htons((u_int16_t)port); 220 LIST_FOREACH(inp, &divcbinfo.pcblisthead, inp_list) { 221 if (inp->inp_flags & INP_PLACEMARKER) 222 continue; 223 if (inp->inp_lport == nport) 224 sa = inp->inp_socket; 225 } 226 if (sa) { 227 if (sbappendaddr(&sa->so_rcv, (struct sockaddr *)&divsrc, 228 m, (struct mbuf *)0) == 0) 229 m_freem(m); 230 else 231 sorwakeup(sa); 232 } else { 233 m_freem(m); 234 ipstat.ips_noproto++; 235 ipstat.ips_delivered--; 236 } 237 } 238 239 /* 240 * Deliver packet back into the IP processing machinery. 241 * 242 * If no address specified, or address is 0.0.0.0, send to ip_output(); 243 * otherwise, send to ip_input() and mark as having been received on 244 * the interface with that address. 245 */ 246 static int 247 div_output(struct socket *so, struct mbuf *m, 248 struct sockaddr_in *sin, struct mbuf *control) 249 { 250 int error = 0; 251 struct m_hdr divert_tag; 252 253 /* 254 * Prepare the tag for divert info. Note that a packet 255 * with a 0 tag in mh_data is effectively untagged, 256 * so we could optimize that case. 257 */ 258 divert_tag.mh_type = MT_TAG; 259 divert_tag.mh_flags = PACKET_TAG_DIVERT; 260 divert_tag.mh_next = m; 261 divert_tag.mh_data = 0; /* the matching rule # */ 262 m->m_pkthdr.rcvif = NULL; /* XXX is it necessary ? */ 263 264 if (control) 265 m_freem(control); /* XXX */ 266 267 /* Loopback avoidance and state recovery */ 268 if (sin) { 269 int i; 270 271 divert_tag.mh_data = (caddr_t)(int)sin->sin_port; 272 /* 273 * Find receive interface with the given name, stuffed 274 * (if it exists) in the sin_zero[] field. 275 * The name is user supplied data so don't trust its size 276 * or that it is zero terminated. 277 */ 278 for (i = 0; sin->sin_zero[i] && i < sizeof(sin->sin_zero); i++) 279 ; 280 if ( i > 0 && i < sizeof(sin->sin_zero)) 281 m->m_pkthdr.rcvif = ifunit(sin->sin_zero); 282 } 283 284 /* Reinject packet into the system as incoming or outgoing */ 285 if (!sin || sin->sin_addr.s_addr == 0) { 286 struct inpcb *const inp = sotoinpcb(so); 287 struct ip *const ip = mtod(m, struct ip *); 288 289 /* 290 * Don't allow both user specified and setsockopt options, 291 * and don't allow packet length sizes that will crash 292 */ 293 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) || 294 ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) { 295 error = EINVAL; 296 goto cantsend; 297 } 298 299 /* Convert fields to host order for ip_output() */ 300 ip->ip_len = ntohs(ip->ip_len); 301 ip->ip_off = ntohs(ip->ip_off); 302 303 /* Send packet to output processing */ 304 ipstat.ips_rawout++; /* XXX */ 305 error = ip_output((struct mbuf *)&divert_tag, 306 inp->inp_options, &inp->inp_route, 307 (so->so_options & SO_DONTROUTE) | 308 IP_ALLOWBROADCAST | IP_RAWOUTPUT, 309 inp->inp_moptions, NULL); 310 } else { 311 if (m->m_pkthdr.rcvif == NULL) { 312 /* 313 * No luck with the name, check by IP address. 314 * Clear the port and the ifname to make sure 315 * there are no distractions for ifa_ifwithaddr. 316 */ 317 struct ifaddr *ifa; 318 319 bzero(sin->sin_zero, sizeof(sin->sin_zero)); 320 sin->sin_port = 0; 321 ifa = ifa_ifwithaddr((struct sockaddr *) sin); 322 if (ifa == NULL) { 323 error = EADDRNOTAVAIL; 324 goto cantsend; 325 } 326 m->m_pkthdr.rcvif = ifa->ifa_ifp; 327 } 328 ip_input((struct mbuf *)&divert_tag); 329 } 330 331 return error; 332 333 cantsend: 334 m_freem(m); 335 return error; 336 } 337 338 static int 339 div_attach(struct socket *so, int proto, struct pru_attach_info *ai) 340 { 341 struct inpcb *inp; 342 int error, s; 343 344 inp = sotoinpcb(so); 345 if (inp) 346 panic("div_attach"); 347 if ((error = suser_cred(ai->p_ucred, NULL_CRED_OKAY)) != 0) 348 return error; 349 350 error = soreserve(so, div_sendspace, div_recvspace, ai->sb_rlimit); 351 if (error) 352 return error; 353 s = splnet(); 354 error = in_pcballoc(so, &divcbinfo); 355 splx(s); 356 if (error) 357 return error; 358 inp = (struct inpcb *)so->so_pcb; 359 inp->inp_ip_p = proto; 360 inp->inp_vflag |= INP_IPV4; 361 inp->inp_flags |= INP_HDRINCL; 362 /* The socket is always "connected" because 363 we always know "where" to send the packet */ 364 so->so_state |= SS_ISCONNECTED; 365 return 0; 366 } 367 368 static int 369 div_detach(struct socket *so) 370 { 371 struct inpcb *inp; 372 373 inp = sotoinpcb(so); 374 if (inp == 0) 375 panic("div_detach"); 376 in_pcbdetach(inp); 377 return 0; 378 } 379 380 static int 381 div_abort(struct socket *so) 382 { 383 soisdisconnected(so); 384 return div_detach(so); 385 } 386 387 static int 388 div_disconnect(struct socket *so) 389 { 390 if ((so->so_state & SS_ISCONNECTED) == 0) 391 return ENOTCONN; 392 return div_abort(so); 393 } 394 395 static int 396 div_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 397 { 398 struct inpcb *inp; 399 int s; 400 int error; 401 402 s = splnet(); 403 inp = sotoinpcb(so); 404 /* in_pcbbind assumes that nam is a sockaddr_in 405 * and in_pcbbind requires a valid address. Since divert 406 * sockets don't we need to make sure the address is 407 * filled in properly. 408 * XXX -- divert should not be abusing in_pcbind 409 * and should probably have its own family. 410 */ 411 if (nam->sa_family != AF_INET) 412 error = EAFNOSUPPORT; 413 else { 414 ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY; 415 error = in_pcbbind(inp, nam, td); 416 } 417 splx(s); 418 return error; 419 } 420 421 static int 422 div_shutdown(struct socket *so) 423 { 424 socantsendmore(so); 425 return 0; 426 } 427 428 static int 429 div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 430 struct mbuf *control, struct thread *td) 431 { 432 /* Packet must have a header (but that's about it) */ 433 if (m->m_len < sizeof (struct ip) && 434 (m = m_pullup(m, sizeof (struct ip))) == 0) { 435 ipstat.ips_toosmall++; 436 m_freem(m); 437 return EINVAL; 438 } 439 440 /* Send packet */ 441 return div_output(so, m, (struct sockaddr_in *)nam, control); 442 } 443 444 static int 445 div_pcblist(SYSCTL_HANDLER_ARGS) 446 { 447 int error, i, n; 448 struct inpcb *inp; 449 struct inpcb *marker; 450 inp_gen_t gencnt; 451 struct xinpgen xig; 452 struct xinpcb xi; 453 454 /* 455 * The process of preparing the TCB list is too time-consuming and 456 * resource-intensive to repeat twice on every request. 457 */ 458 if (req->oldptr == 0) { 459 n = divcbinfo.ipi_count; 460 req->oldidx = 2 * (sizeof xig) 461 + (n + n/8) * sizeof(struct xinpcb); 462 return 0; 463 } 464 465 if (req->newptr != 0) 466 return EPERM; 467 468 /* 469 * OK, now we're committed to doing something. 470 */ 471 gencnt = divcbinfo.ipi_gencnt; 472 n = divcbinfo.ipi_count; 473 474 xig.xig_len = sizeof xig; 475 xig.xig_count = n; 476 xig.xig_gen = gencnt; 477 xig.xig_sogen = so_gencnt; 478 xig.xig_cpu = 0; 479 error = SYSCTL_OUT(req, &xig, sizeof xig); 480 if (error) 481 return error; 482 483 marker = malloc(sizeof(struct inpcb), M_TEMP, M_WAITOK|M_ZERO); 484 marker->inp_flags |= INP_PLACEMARKER; 485 486 LIST_INSERT_HEAD(&divcbinfo.pcblisthead, marker, inp_list); 487 i = 0; 488 while ((inp = LIST_NEXT(marker, inp_list)) != NULL && i < n) { 489 LIST_REMOVE(marker, inp_list); 490 LIST_INSERT_AFTER(inp, marker, inp_list); 491 492 if (inp->inp_flags & INP_PLACEMARKER) 493 continue; 494 if (inp->inp_gencnt > gencnt) 495 continue; 496 if (prison_xinpcb(req->td, inp)) 497 continue; 498 xi.xi_len = sizeof xi; 499 bcopy(inp, &xi.xi_inp, sizeof *inp); 500 if (inp->inp_socket) 501 sotoxsocket(inp->inp_socket, &xi.xi_socket); 502 if ((error = SYSCTL_OUT(req, &xi, sizeof xi)) != 0) 503 break; 504 ++i; 505 } 506 LIST_REMOVE(marker, inp_list); 507 if (error == 0 && i < n) { 508 bzero(&xi, sizeof(xi)); 509 xi.xi_len = sizeof(xi); 510 while (i < n) { 511 error = SYSCTL_OUT(req, &xi, sizeof(xi)); 512 ++i; 513 } 514 } 515 if (error == 0) { 516 /* 517 * Give the user an updated idea of our state. 518 * If the generation differs from what we told 519 * her before, she knows that something happened 520 * while we were processing this request, and it 521 * might be necessary to retry. 522 */ 523 xig.xig_gen = divcbinfo.ipi_gencnt; 524 xig.xig_sogen = so_gencnt; 525 xig.xig_count = divcbinfo.ipi_count; 526 error = SYSCTL_OUT(req, &xig, sizeof xig); 527 } 528 free(marker, M_TEMP); 529 return error; 530 } 531 532 SYSCTL_DECL(_net_inet_divert); 533 SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLFLAG_RD, 0, 0, 534 div_pcblist, "S,xinpcb", "List of active divert sockets"); 535 536 struct pr_usrreqs div_usrreqs = { 537 div_abort, pru_accept_notsupp, div_attach, div_bind, 538 pru_connect_notsupp, pru_connect2_notsupp, in_control, div_detach, 539 div_disconnect, pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp, 540 pru_rcvoob_notsupp, div_send, pru_sense_null, div_shutdown, 541 in_setsockaddr, sosend, soreceive, sopoll 542 }; 543