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 1.10 2004/03/06 05:00:41 hsu 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/netisr.h> 61 #include <net/route.h> 62 63 #include <netinet/in.h> 64 #include <netinet/in_systm.h> 65 #include <netinet/ip.h> 66 #include <netinet/in_pcb.h> 67 #include <netinet/in_var.h> 68 #include <netinet/ip_var.h> 69 70 /* 71 * Divert sockets 72 */ 73 74 /* 75 * Allocate enough space to hold a full IP packet 76 */ 77 #define DIVSNDQ (65536 + 100) 78 #define DIVRCVQ (65536 + 100) 79 80 /* 81 * Divert sockets work in conjunction with ipfw, see the divert(4) 82 * manpage for features. 83 * Internally, packets selected by ipfw in ip_input() or ip_output(), 84 * and never diverted before, are passed to the input queue of the 85 * divert socket with a given 'divert_port' number (as specified in 86 * the matching ipfw rule), and they are tagged with a 16 bit cookie 87 * (representing the rule number of the matching ipfw rule), which 88 * is passed to process reading from the socket. 89 * 90 * Packets written to the divert socket are again tagged with a cookie 91 * (usually the same as above) and a destination address. 92 * If the destination address is INADDR_ANY then the packet is 93 * treated as outgoing and sent to ip_output(), otherwise it is 94 * treated as incoming and sent to ip_input(). 95 * In both cases, the packet is tagged with the cookie. 96 * 97 * On reinjection, processing in ip_input() and ip_output() 98 * will be exactly the same as for the original packet, except that 99 * ipfw processing will start at the rule number after the one 100 * written in the cookie (so, tagging a packet with a cookie of 0 101 * will cause it to be effectively considered as a standard packet). 102 */ 103 104 /* Internal variables */ 105 static struct inpcbinfo divcbinfo; 106 107 static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */ 108 static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */ 109 110 /* Optimization: have this preinitialized */ 111 static struct sockaddr_in divsrc = { sizeof(divsrc), AF_INET }; 112 113 /* 114 * Initialize divert connection block queue. 115 */ 116 void 117 div_init(void) 118 { 119 LIST_INIT(&divcbinfo.listhead); 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.bindhashbase = hashinit(1, M_PCB, &divcbinfo.bindhashmask); 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, int off, int proto) 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.listhead, 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 struct netmsg_packet msg; 310 311 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 /* Send packet to input processing */ 329 msg.nm_packet = (struct mbuf *)&divert_tag; 330 ip_input((struct netmsg *)&msg); 331 } 332 333 return error; 334 335 cantsend: 336 m_freem(m); 337 return error; 338 } 339 340 static int 341 div_attach(struct socket *so, int proto, struct pru_attach_info *ai) 342 { 343 struct inpcb *inp; 344 int error, s; 345 346 inp = sotoinpcb(so); 347 if (inp) 348 panic("div_attach"); 349 if ((error = suser_cred(ai->p_ucred, NULL_CRED_OKAY)) != 0) 350 return error; 351 352 error = soreserve(so, div_sendspace, div_recvspace, ai->sb_rlimit); 353 if (error) 354 return error; 355 s = splnet(); 356 error = in_pcballoc(so, &divcbinfo); 357 splx(s); 358 if (error) 359 return error; 360 inp = (struct inpcb *)so->so_pcb; 361 inp->inp_ip_p = proto; 362 inp->inp_vflag |= INP_IPV4; 363 inp->inp_flags |= INP_HDRINCL; 364 /* The socket is always "connected" because 365 we always know "where" to send the packet */ 366 so->so_state |= SS_ISCONNECTED; 367 return 0; 368 } 369 370 static int 371 div_detach(struct socket *so) 372 { 373 struct inpcb *inp; 374 375 inp = sotoinpcb(so); 376 if (inp == 0) 377 panic("div_detach"); 378 in_pcbdetach(inp); 379 return 0; 380 } 381 382 static int 383 div_abort(struct socket *so) 384 { 385 soisdisconnected(so); 386 return div_detach(so); 387 } 388 389 static int 390 div_disconnect(struct socket *so) 391 { 392 if ((so->so_state & SS_ISCONNECTED) == 0) 393 return ENOTCONN; 394 return div_abort(so); 395 } 396 397 static int 398 div_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 399 { 400 struct inpcb *inp; 401 int s; 402 int error; 403 404 s = splnet(); 405 inp = sotoinpcb(so); 406 /* in_pcbbind assumes that nam is a sockaddr_in 407 * and in_pcbbind requires a valid address. Since divert 408 * sockets don't we need to make sure the address is 409 * filled in properly. 410 * XXX -- divert should not be abusing in_pcbind 411 * and should probably have its own family. 412 */ 413 if (nam->sa_family != AF_INET) 414 error = EAFNOSUPPORT; 415 else { 416 ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY; 417 error = in_pcbbind(inp, nam, td); 418 } 419 splx(s); 420 return error; 421 } 422 423 static int 424 div_shutdown(struct socket *so) 425 { 426 socantsendmore(so); 427 return 0; 428 } 429 430 static int 431 div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 432 struct mbuf *control, struct thread *td) 433 { 434 /* Packet must have a header (but that's about it) */ 435 if (m->m_len < sizeof (struct ip) && 436 (m = m_pullup(m, sizeof (struct ip))) == 0) { 437 ipstat.ips_toosmall++; 438 m_freem(m); 439 return EINVAL; 440 } 441 442 /* Send packet */ 443 return div_output(so, m, (struct sockaddr_in *)nam, control); 444 } 445 446 static int 447 div_pcblist(SYSCTL_HANDLER_ARGS) 448 { 449 int error, i, n, s; 450 struct inpcb *inp, **inp_list; 451 inp_gen_t gencnt; 452 struct xinpgen xig; 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 s = splnet(); 472 gencnt = divcbinfo.ipi_gencnt; 473 n = divcbinfo.ipi_count; 474 splx(s); 475 476 xig.xig_len = sizeof xig; 477 xig.xig_count = n; 478 xig.xig_gen = gencnt; 479 xig.xig_sogen = so_gencnt; 480 error = SYSCTL_OUT(req, &xig, sizeof xig); 481 if (error) 482 return error; 483 484 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 485 if (inp_list == 0) 486 return ENOMEM; 487 488 s = splnet(); 489 for (inp = LIST_FIRST(&divcbinfo.listhead), i = 0; inp && i < n; 490 inp = LIST_NEXT(inp, inp_list)) { 491 if (inp->inp_gencnt <= gencnt && !prison_xinpcb(req->td, inp)) 492 inp_list[i++] = inp; 493 } 494 splx(s); 495 n = i; 496 497 error = 0; 498 for (i = 0; i < n; i++) { 499 inp = inp_list[i]; 500 if (inp->inp_gencnt <= gencnt) { 501 struct xinpcb xi; 502 xi.xi_len = sizeof xi; 503 /* XXX should avoid extra copy */ 504 bcopy(inp, &xi.xi_inp, sizeof *inp); 505 if (inp->inp_socket) 506 sotoxsocket(inp->inp_socket, &xi.xi_socket); 507 error = SYSCTL_OUT(req, &xi, sizeof xi); 508 } 509 } 510 if (!error) { 511 /* 512 * Give the user an updated idea of our state. 513 * If the generation differs from what we told 514 * her before, she knows that something happened 515 * while we were processing this request, and it 516 * might be necessary to retry. 517 */ 518 s = splnet(); 519 xig.xig_gen = divcbinfo.ipi_gencnt; 520 xig.xig_sogen = so_gencnt; 521 xig.xig_count = divcbinfo.ipi_count; 522 splx(s); 523 error = SYSCTL_OUT(req, &xig, sizeof xig); 524 } 525 free(inp_list, M_TEMP); 526 return error; 527 } 528 529 SYSCTL_DECL(_net_inet_divert); 530 SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLFLAG_RD, 0, 0, 531 div_pcblist, "S,xinpcb", "List of active divert sockets"); 532 533 struct pr_usrreqs div_usrreqs = { 534 div_abort, pru_accept_notsupp, div_attach, div_bind, 535 pru_connect_notsupp, pru_connect2_notsupp, in_control, div_detach, 536 div_disconnect, pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp, 537 pru_rcvoob_notsupp, div_send, pru_sense_null, div_shutdown, 538 in_setsockaddr, sosend, soreceive, sopoll 539 }; 540