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 * $FreeBSD$ 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/lock.h> 48 #include <sys/malloc.h> 49 #include <sys/mbuf.h> 50 #include <sys/proc.h> 51 #include <sys/protosw.h> 52 #include <sys/signalvar.h> 53 #include <sys/socket.h> 54 #include <sys/socketvar.h> 55 #include <sys/sx.h> 56 #include <sys/sysctl.h> 57 #include <sys/systm.h> 58 59 #include <vm/uma.h> 60 61 #include <net/if.h> 62 #include <net/route.h> 63 64 #include <netinet/in.h> 65 #include <netinet/in_pcb.h> 66 #include <netinet/in_systm.h> 67 #include <netinet/in_var.h> 68 #include <netinet/ip.h> 69 #include <netinet/ip_var.h> 70 71 /* 72 * Divert sockets 73 */ 74 75 /* 76 * Allocate enough space to hold a full IP packet 77 */ 78 #define DIVSNDQ (65536 + 100) 79 #define DIVRCVQ (65536 + 100) 80 81 /* 82 * Divert sockets work in conjunction with ipfw, see the divert(4) 83 * manpage for features. 84 * Internally, packets selected by ipfw in ip_input() or ip_output(), 85 * and never diverted before, are passed to the input queue of the 86 * divert socket with a given 'divert_port' number (as specified in 87 * the matching ipfw rule), and they are tagged with a 16 bit cookie 88 * (representing the rule number of the matching ipfw rule), which 89 * is passed to process reading from the socket. 90 * 91 * Packets written to the divert socket are again tagged with a cookie 92 * (usually the same as above) and a destination address. 93 * If the destination address is INADDR_ANY then the packet is 94 * treated as outgoing and sent to ip_output(), otherwise it is 95 * treated as incoming and sent to ip_input(). 96 * In both cases, the packet is tagged with the cookie. 97 * 98 * On reinjection, processing in ip_input() and ip_output() 99 * will be exactly the same as for the original packet, except that 100 * ipfw processing will start at the rule number after the one 101 * written in the cookie (so, tagging a packet with a cookie of 0 102 * will cause it to be effectively considered as a standard packet). 103 */ 104 105 /* Internal variables */ 106 static struct inpcbhead divcb; 107 static struct inpcbinfo divcbinfo; 108 109 static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */ 110 static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */ 111 112 /* Optimization: have this preinitialized */ 113 static struct sockaddr_in divsrc = { sizeof(divsrc), AF_INET }; 114 115 /* 116 * Initialize divert connection block queue. 117 */ 118 void 119 div_init(void) 120 { 121 INP_INFO_LOCK_INIT(&divcbinfo, "div"); 122 LIST_INIT(&divcb); 123 divcbinfo.listhead = &divcb; 124 /* 125 * XXX We don't use the hash list for divert IP, but it's easier 126 * to allocate a one entry hash list than it is to check all 127 * over the place for hashbase == NULL. 128 */ 129 divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask); 130 divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask); 131 divcbinfo.ipi_zone = uma_zcreate("divcb", sizeof(struct inpcb), 132 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 133 uma_zone_set_max(divcbinfo.ipi_zone, maxsockets); 134 } 135 136 /* 137 * IPPROTO_DIVERT is not a real IP protocol; don't allow any packets 138 * with that protocol number to enter the system from the outside. 139 */ 140 void 141 div_input(struct mbuf *m, int off) 142 { 143 ipstat.ips_noproto++; 144 m_freem(m); 145 } 146 147 /* 148 * Divert a packet by passing it up to the divert socket at port 'port'. 149 * 150 * Setup generic address and protocol structures for div_input routine, 151 * then pass them along with mbuf chain. 152 */ 153 void 154 divert_packet(struct mbuf *m, int incoming, int port, int rule) 155 { 156 struct ip *ip; 157 struct inpcb *inp; 158 struct socket *sa; 159 u_int16_t nport; 160 161 /* Sanity check */ 162 KASSERT(port != 0, ("%s: port=0", __func__)); 163 164 divsrc.sin_port = rule; /* record matching rule */ 165 166 /* Assure header */ 167 if (m->m_len < sizeof(struct ip) && 168 (m = m_pullup(m, sizeof(struct ip))) == 0) 169 return; 170 ip = mtod(m, struct ip *); 171 172 /* 173 * Record receive interface address, if any. 174 * But only for incoming packets. 175 */ 176 divsrc.sin_addr.s_addr = 0; 177 if (incoming) { 178 struct ifaddr *ifa; 179 180 /* Sanity check */ 181 KASSERT((m->m_flags & M_PKTHDR), ("%s: !PKTHDR", __func__)); 182 183 /* Find IP address for receive interface */ 184 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 185 if (ifa->ifa_addr == NULL) 186 continue; 187 if (ifa->ifa_addr->sa_family != AF_INET) 188 continue; 189 divsrc.sin_addr = 190 ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr; 191 break; 192 } 193 } 194 /* 195 * Record the incoming interface name whenever we have one. 196 */ 197 bzero(&divsrc.sin_zero, sizeof(divsrc.sin_zero)); 198 if (m->m_pkthdr.rcvif) { 199 /* 200 * Hide the actual interface name in there in the 201 * sin_zero array. XXX This needs to be moved to a 202 * different sockaddr type for divert, e.g. 203 * sockaddr_div with multiple fields like 204 * sockaddr_dl. Presently we have only 7 bytes 205 * but that will do for now as most interfaces 206 * are 4 or less + 2 or less bytes for unit. 207 * There is probably a faster way of doing this, 208 * possibly taking it from the sockaddr_dl on the iface. 209 * This solves the problem of a P2P link and a LAN interface 210 * having the same address, which can result in the wrong 211 * interface being assigned to the packet when fed back 212 * into the divert socket. Theoretically if the daemon saves 213 * and re-uses the sockaddr_in as suggested in the man pages, 214 * this iface name will come along for the ride. 215 * (see div_output for the other half of this.) 216 */ 217 snprintf(divsrc.sin_zero, sizeof(divsrc.sin_zero), 218 "%s%d", m->m_pkthdr.rcvif->if_name, 219 m->m_pkthdr.rcvif->if_unit); 220 } 221 222 /* Put packet on socket queue, if any */ 223 sa = NULL; 224 nport = htons((u_int16_t)port); 225 LIST_FOREACH(inp, &divcb, inp_list) { 226 if (inp->inp_lport == nport) 227 sa = inp->inp_socket; 228 } 229 if (sa) { 230 if (sbappendaddr(&sa->so_rcv, (struct sockaddr *)&divsrc, 231 m, (struct mbuf *)0) == 0) 232 m_freem(m); 233 else 234 sorwakeup(sa); 235 } else { 236 m_freem(m); 237 ipstat.ips_noproto++; 238 ipstat.ips_delivered--; 239 } 240 } 241 242 /* 243 * Deliver packet back into the IP processing machinery. 244 * 245 * If no address specified, or address is 0.0.0.0, send to ip_output(); 246 * otherwise, send to ip_input() and mark as having been received on 247 * the interface with that address. 248 */ 249 static int 250 div_output(struct socket *so, struct mbuf *m, 251 struct sockaddr_in *sin, struct mbuf *control) 252 { 253 int error = 0; 254 struct m_hdr divert_tag; 255 256 /* 257 * Prepare the tag for divert info. Note that a packet 258 * with a 0 tag in mh_data is effectively untagged, 259 * so we could optimize that case. 260 */ 261 divert_tag.mh_type = MT_TAG; 262 divert_tag.mh_flags = PACKET_TAG_DIVERT; 263 divert_tag.mh_next = m; 264 divert_tag.mh_data = 0; /* the matching rule # */ 265 m->m_pkthdr.rcvif = NULL; /* XXX is it necessary ? */ 266 267 if (control) 268 m_freem(control); /* XXX */ 269 270 /* Loopback avoidance and state recovery */ 271 if (sin) { 272 int i; 273 274 divert_tag.mh_data = (caddr_t)(int)sin->sin_port; 275 /* 276 * Find receive interface with the given name, stuffed 277 * (if it exists) in the sin_zero[] field. 278 * The name is user supplied data so don't trust its size 279 * or that it is zero terminated. 280 */ 281 for (i = 0; sin->sin_zero[i] && i < sizeof(sin->sin_zero); i++) 282 ; 283 if ( i > 0 && i < sizeof(sin->sin_zero)) 284 m->m_pkthdr.rcvif = ifunit(sin->sin_zero); 285 } 286 287 /* Reinject packet into the system as incoming or outgoing */ 288 if (!sin || sin->sin_addr.s_addr == 0) { 289 struct inpcb *const inp = sotoinpcb(so); 290 struct ip *const ip = mtod(m, struct ip *); 291 292 /* 293 * Don't allow both user specified and setsockopt options, 294 * and don't allow packet length sizes that will crash 295 */ 296 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) || 297 ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) { 298 error = EINVAL; 299 goto cantsend; 300 } 301 302 /* Convert fields to host order for ip_output() */ 303 ip->ip_len = ntohs(ip->ip_len); 304 ip->ip_off = ntohs(ip->ip_off); 305 306 /* Send packet to output processing */ 307 ipstat.ips_rawout++; /* XXX */ 308 error = ip_output((struct mbuf *)&divert_tag, 309 inp->inp_options, &inp->inp_route, 310 (so->so_options & SO_DONTROUTE) | 311 IP_ALLOWBROADCAST | IP_RAWOUTPUT, 312 inp->inp_moptions); 313 } else { 314 if (m->m_pkthdr.rcvif == NULL) { 315 /* 316 * No luck with the name, check by IP address. 317 * Clear the port and the ifname to make sure 318 * there are no distractions for ifa_ifwithaddr. 319 */ 320 struct ifaddr *ifa; 321 322 bzero(sin->sin_zero, sizeof(sin->sin_zero)); 323 sin->sin_port = 0; 324 ifa = ifa_ifwithaddr((struct sockaddr *) sin); 325 if (ifa == NULL) { 326 error = EADDRNOTAVAIL; 327 goto cantsend; 328 } 329 m->m_pkthdr.rcvif = ifa->ifa_ifp; 330 } 331 /* Send packet to input processing */ 332 ip_input((struct mbuf *)&divert_tag); 333 } 334 335 return error; 336 337 cantsend: 338 m_freem(m); 339 return error; 340 } 341 342 static int 343 div_attach(struct socket *so, int proto, struct thread *td) 344 { 345 struct inpcb *inp; 346 int error, s; 347 348 inp = sotoinpcb(so); 349 if (inp) 350 panic("div_attach"); 351 if (td && (error = suser(td)) != 0) 352 return error; 353 354 error = soreserve(so, div_sendspace, div_recvspace); 355 if (error) 356 return error; 357 s = splnet(); 358 error = in_pcballoc(so, &divcbinfo, td); 359 splx(s); 360 if (error) 361 return error; 362 inp = (struct inpcb *)so->so_pcb; 363 inp->inp_ip_p = proto; 364 inp->inp_vflag |= INP_IPV4; 365 inp->inp_flags |= INP_HDRINCL; 366 /* The socket is always "connected" because 367 we always know "where" to send the packet */ 368 so->so_state |= SS_ISCONNECTED; 369 return 0; 370 } 371 372 static int 373 div_detach(struct socket *so) 374 { 375 struct inpcb *inp; 376 377 inp = sotoinpcb(so); 378 if (inp == 0) 379 panic("div_detach"); 380 in_pcbdetach(inp); 381 return 0; 382 } 383 384 static int 385 div_abort(struct socket *so) 386 { 387 soisdisconnected(so); 388 return div_detach(so); 389 } 390 391 static int 392 div_disconnect(struct socket *so) 393 { 394 if ((so->so_state & SS_ISCONNECTED) == 0) 395 return ENOTCONN; 396 return div_abort(so); 397 } 398 399 static int 400 div_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 401 { 402 struct inpcb *inp; 403 int s; 404 int error; 405 406 s = splnet(); 407 inp = sotoinpcb(so); 408 /* in_pcbbind assumes that nam is a sockaddr_in 409 * and in_pcbbind requires a valid address. Since divert 410 * sockets don't we need to make sure the address is 411 * filled in properly. 412 * XXX -- divert should not be abusing in_pcbind 413 * and should probably have its own family. 414 */ 415 if (nam->sa_family != AF_INET) 416 error = EAFNOSUPPORT; 417 else { 418 ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY; 419 error = in_pcbbind(inp, nam, td); 420 } 421 splx(s); 422 return error; 423 } 424 425 static int 426 div_shutdown(struct socket *so) 427 { 428 socantsendmore(so); 429 return 0; 430 } 431 432 static int 433 div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 434 struct mbuf *control, struct thread *td) 435 { 436 /* Packet must have a header (but that's about it) */ 437 if (m->m_len < sizeof (struct ip) && 438 (m = m_pullup(m, sizeof (struct ip))) == 0) { 439 ipstat.ips_toosmall++; 440 m_freem(m); 441 return EINVAL; 442 } 443 444 /* Send packet */ 445 return div_output(so, m, (struct sockaddr_in *)nam, control); 446 } 447 448 static int 449 div_pcblist(SYSCTL_HANDLER_ARGS) 450 { 451 int error, i, n, s; 452 struct inpcb *inp, **inp_list; 453 inp_gen_t gencnt; 454 struct xinpgen xig; 455 456 /* 457 * The process of preparing the TCB list is too time-consuming and 458 * resource-intensive to repeat twice on every request. 459 */ 460 if (req->oldptr == 0) { 461 n = divcbinfo.ipi_count; 462 req->oldidx = 2 * (sizeof xig) 463 + (n + n/8) * sizeof(struct xinpcb); 464 return 0; 465 } 466 467 if (req->newptr != 0) 468 return EPERM; 469 470 /* 471 * OK, now we're committed to doing something. 472 */ 473 s = splnet(); 474 gencnt = divcbinfo.ipi_gencnt; 475 n = divcbinfo.ipi_count; 476 splx(s); 477 478 xig.xig_len = sizeof xig; 479 xig.xig_count = n; 480 xig.xig_gen = gencnt; 481 xig.xig_sogen = so_gencnt; 482 error = SYSCTL_OUT(req, &xig, sizeof xig); 483 if (error) 484 return error; 485 486 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 487 if (inp_list == 0) 488 return ENOMEM; 489 490 s = splnet(); 491 for (inp = LIST_FIRST(divcbinfo.listhead), i = 0; inp && i < n; 492 inp = LIST_NEXT(inp, inp_list)) { 493 if (inp->inp_gencnt <= gencnt && !prison_xinpcb(req->td, inp)) 494 inp_list[i++] = inp; 495 } 496 splx(s); 497 n = i; 498 499 error = 0; 500 for (i = 0; i < n; i++) { 501 inp = inp_list[i]; 502 if (inp->inp_gencnt <= gencnt) { 503 struct xinpcb xi; 504 xi.xi_len = sizeof xi; 505 /* XXX should avoid extra copy */ 506 bcopy(inp, &xi.xi_inp, sizeof *inp); 507 if (inp->inp_socket) 508 sotoxsocket(inp->inp_socket, &xi.xi_socket); 509 error = SYSCTL_OUT(req, &xi, sizeof xi); 510 } 511 } 512 if (!error) { 513 /* 514 * Give the user an updated idea of our state. 515 * If the generation differs from what we told 516 * her before, she knows that something happened 517 * while we were processing this request, and it 518 * might be necessary to retry. 519 */ 520 s = splnet(); 521 xig.xig_gen = divcbinfo.ipi_gencnt; 522 xig.xig_sogen = so_gencnt; 523 xig.xig_count = divcbinfo.ipi_count; 524 splx(s); 525 error = SYSCTL_OUT(req, &xig, sizeof xig); 526 } 527 free(inp_list, M_TEMP); 528 return error; 529 } 530 531 /* 532 * This is the wrapper function for in_setsockaddr. We just pass down 533 * the pcbinfo for in_setpeeraddr to lock. 534 */ 535 static int 536 div_sockaddr(struct socket *so, struct sockaddr **nam) 537 { 538 return (in_setsockaddr(so, nam, &divcbinfo)); 539 } 540 541 /* 542 * This is the wrapper function for in_setpeeraddr. We just pass down 543 * the pcbinfo for in_setpeeraddr to lock. 544 */ 545 static int 546 div_peeraddr(struct socket *so, struct sockaddr **nam) 547 { 548 return (in_setpeeraddr(so, nam, &divcbinfo)); 549 } 550 551 552 SYSCTL_DECL(_net_inet_divert); 553 SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLFLAG_RD, 0, 0, 554 div_pcblist, "S,xinpcb", "List of active divert sockets"); 555 556 struct pr_usrreqs div_usrreqs = { 557 div_abort, pru_accept_notsupp, div_attach, div_bind, 558 pru_connect_notsupp, pru_connect2_notsupp, in_control, div_detach, 559 div_disconnect, pru_listen_notsupp, div_peeraddr, pru_rcvd_notsupp, 560 pru_rcvoob_notsupp, div_send, pru_sense_null, div_shutdown, 561 div_sockaddr, sosend, soreceive, sopoll 562 }; 563