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/malloc.h> 48 #include <sys/mbuf.h> 49 #include <sys/proc.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 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 * A 16 bit cookie is passed to and from the user process. 80 * The user process can send it back to help the caller know 81 * something about where the packet originally came from. 82 * 83 * In the case of ipfw, then the cookie is the rule that sent 84 * us here. On reinjection is is the rule after which processing 85 * should continue. Leaving it the same will make processing start 86 * at the rule number after that which sent it here. Setting it to 87 * 0 will restart processing at the beginning. 88 * 89 * For divert_packet(), ip_divert_cookie is an input value only. 90 * For div_output(), ip_divert_cookie is an output value only. 91 */ 92 u_int16_t ip_divert_cookie; 93 94 /* Internal variables */ 95 static struct inpcbhead divcb; 96 static struct inpcbinfo divcbinfo; 97 98 static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */ 99 static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */ 100 101 /* Optimization: have this preinitialized */ 102 static struct sockaddr_in divsrc = { sizeof(divsrc), AF_INET }; 103 104 /* Internal functions */ 105 static int div_output(struct socket *so, 106 struct mbuf *m, struct sockaddr *addr, struct mbuf *control); 107 108 /* 109 * Initialize divert connection block queue. 110 */ 111 void 112 div_init(void) 113 { 114 LIST_INIT(&divcb); 115 divcbinfo.listhead = &divcb; 116 /* 117 * XXX We don't use the hash list for divert IP, but it's easier 118 * to allocate a one entry hash list than it is to check all 119 * over the place for hashbase == NULL. 120 */ 121 divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask); 122 divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask); 123 divcbinfo.ipi_zone = zinit("divcb", sizeof(struct inpcb), 124 maxsockets, ZONE_INTERRUPT, 0); 125 } 126 127 /* 128 * IPPROTO_DIVERT is not a real IP protocol; don't allow any packets 129 * with that protocol number to enter the system from the outside. 130 */ 131 void 132 div_input(struct mbuf *m, int off) 133 { 134 ipstat.ips_noproto++; 135 m_freem(m); 136 } 137 138 /* 139 * Divert a packet by passing it up to the divert socket at port 'port'. 140 * 141 * Setup generic address and protocol structures for div_input routine, 142 * then pass them along with mbuf chain. 143 */ 144 void 145 divert_packet(struct mbuf *m, int incoming, int port) 146 { 147 struct ip *ip; 148 struct inpcb *inp; 149 struct socket *sa; 150 u_int16_t nport; 151 152 /* Sanity check */ 153 KASSERT(port != 0, ("%s: port=0", __FUNCTION__)); 154 155 /* Record and reset divert cookie */ 156 divsrc.sin_port = ip_divert_cookie; 157 ip_divert_cookie = 0; 158 159 /* Assure header */ 160 if (m->m_len < sizeof(struct ip) && 161 (m = m_pullup(m, sizeof(struct ip))) == 0) { 162 return; 163 } 164 ip = mtod(m, struct ip *); 165 166 /* 167 * Record receive interface address, if any. 168 * But only for incoming packets. 169 */ 170 divsrc.sin_addr.s_addr = 0; 171 if (incoming) { 172 struct ifaddr *ifa; 173 174 /* Sanity check */ 175 KASSERT((m->m_flags & M_PKTHDR), ("%s: !PKTHDR", __FUNCTION__)); 176 177 /* Find IP address for receive interface */ 178 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) { 179 if (ifa->ifa_addr == NULL) 180 continue; 181 if (ifa->ifa_addr->sa_family != AF_INET) 182 continue; 183 divsrc.sin_addr = 184 ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr; 185 break; 186 } 187 } 188 /* 189 * Record the incoming interface name whenever we have one. 190 */ 191 bzero(&divsrc.sin_zero, sizeof(divsrc.sin_zero)); 192 if (m->m_pkthdr.rcvif) { 193 /* 194 * Hide the actual interface name in there in the 195 * sin_zero array. XXX This needs to be moved to a 196 * different sockaddr type for divert, e.g. 197 * sockaddr_div with multiple fields like 198 * sockaddr_dl. Presently we have only 7 bytes 199 * but that will do for now as most interfaces 200 * are 4 or less + 2 or less bytes for unit. 201 * There is probably a faster way of doing this, 202 * possibly taking it from the sockaddr_dl on the iface. 203 * This solves the problem of a P2P link and a LAN interface 204 * having the same address, which can result in the wrong 205 * interface being assigned to the packet when fed back 206 * into the divert socket. Theoretically if the daemon saves 207 * and re-uses the sockaddr_in as suggested in the man pages, 208 * this iface name will come along for the ride. 209 * (see div_output for the other half of this.) 210 */ 211 snprintf(divsrc.sin_zero, sizeof(divsrc.sin_zero), 212 "%s%d", m->m_pkthdr.rcvif->if_name, 213 m->m_pkthdr.rcvif->if_unit); 214 } 215 216 /* Put packet on socket queue, if any */ 217 sa = NULL; 218 nport = htons((u_int16_t)port); 219 LIST_FOREACH(inp, &divcb, inp_list) { 220 if (inp->inp_lport == nport) 221 sa = inp->inp_socket; 222 } 223 if (sa) { 224 if (sbappendaddr(&sa->so_rcv, (struct sockaddr *)&divsrc, 225 m, (struct mbuf *)0) == 0) 226 m_freem(m); 227 else 228 sorwakeup(sa); 229 } else { 230 m_freem(m); 231 ipstat.ips_noproto++; 232 ipstat.ips_delivered--; 233 } 234 } 235 236 /* 237 * Deliver packet back into the IP processing machinery. 238 * 239 * If no address specified, or address is 0.0.0.0, send to ip_output(); 240 * otherwise, send to ip_input() and mark as having been received on 241 * the interface with that address. 242 */ 243 static int 244 div_output(so, m, addr, control) 245 struct socket *so; 246 register struct mbuf *m; 247 struct sockaddr *addr; 248 struct mbuf *control; 249 { 250 register struct inpcb *const inp = sotoinpcb(so); 251 register struct ip *const ip = mtod(m, struct ip *); 252 struct sockaddr_in *sin = (struct sockaddr_in *)addr; 253 int error = 0; 254 255 if (control) 256 m_freem(control); /* XXX */ 257 258 /* Loopback avoidance and state recovery */ 259 if (sin) { 260 int len = 0; 261 char *c = sin->sin_zero; 262 263 ip_divert_cookie = sin->sin_port; 264 265 /* 266 * Find receive interface with the given name or IP address. 267 * The name is user supplied data so don't trust it's size or 268 * that it is zero terminated. The name has priority. 269 * We are presently assuming that the sockaddr_in 270 * has not been replaced by a sockaddr_div, so we limit it 271 * to 16 bytes in total. the name is stuffed (if it exists) 272 * in the sin_zero[] field. 273 */ 274 while (*c++ && (len++ < sizeof(sin->sin_zero))); 275 if ((len > 0) && (len < sizeof(sin->sin_zero))) 276 m->m_pkthdr.rcvif = ifunit(sin->sin_zero); 277 } else { 278 ip_divert_cookie = 0; 279 } 280 281 /* Reinject packet into the system as incoming or outgoing */ 282 if (!sin || sin->sin_addr.s_addr == 0) { 283 /* 284 * Don't allow both user specified and setsockopt options, 285 * and don't allow packet length sizes that will crash 286 */ 287 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) || 288 ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) { 289 error = EINVAL; 290 goto cantsend; 291 } 292 293 /* Convert fields to host order for ip_output() */ 294 NTOHS(ip->ip_len); 295 NTOHS(ip->ip_off); 296 297 /* Send packet to output processing */ 298 ipstat.ips_rawout++; /* XXX */ 299 error = ip_output(m, inp->inp_options, &inp->inp_route, 300 (so->so_options & SO_DONTROUTE) | 301 IP_ALLOWBROADCAST | IP_RAWOUTPUT, 302 inp->inp_moptions); 303 } else { 304 struct ifaddr *ifa; 305 306 /* If no luck with the name above. check by IP address. */ 307 if (m->m_pkthdr.rcvif == NULL) { 308 /* 309 * Make sure there are no distractions 310 * for ifa_ifwithaddr. Clear the port and the ifname. 311 * Maybe zap all 8 bytes at once using a 64bit write? 312 */ 313 bzero(sin->sin_zero, sizeof(sin->sin_zero)); 314 /* *((u_int64_t *)sin->sin_zero) = 0; */ /* XXX ?? */ 315 sin->sin_port = 0; 316 if (!(ifa = ifa_ifwithaddr((struct sockaddr *) sin))) { 317 error = EADDRNOTAVAIL; 318 goto cantsend; 319 } 320 m->m_pkthdr.rcvif = ifa->ifa_ifp; 321 } 322 323 /* Send packet to input processing */ 324 ip_input(m); 325 } 326 327 /* paranoid: Reset for next time (and other packets) */ 328 /* almost definitly already done in the ipfw filter but.. */ 329 ip_divert_cookie = 0; 330 return error; 331 332 cantsend: 333 m_freem(m); 334 ip_divert_cookie = 0; 335 return error; 336 } 337 338 static int 339 div_attach(struct socket *so, int proto, struct thread *td) 340 { 341 struct inpcb *inp; 342 int error, s; 343 344 inp = sotoinpcb(so); 345 if (inp) 346 panic("div_attach"); 347 if (td && (error = suser_td(td)) != 0) 348 return error; 349 350 error = soreserve(so, div_sendspace, div_recvspace); 351 if (error) 352 return error; 353 s = splnet(); 354 error = in_pcballoc(so, &divcbinfo, td); 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 the socket 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, nam, control); 442 } 443 444 static int 445 div_pcblist(SYSCTL_HANDLER_ARGS) 446 { 447 int error, i, n, s; 448 struct inpcb *inp, **inp_list; 449 inp_gen_t gencnt; 450 struct xinpgen xig; 451 452 /* 453 * The process of preparing the TCB list is too time-consuming and 454 * resource-intensive to repeat twice on every request. 455 */ 456 if (req->oldptr == 0) { 457 n = divcbinfo.ipi_count; 458 req->oldidx = 2 * (sizeof xig) 459 + (n + n/8) * sizeof(struct xinpcb); 460 return 0; 461 } 462 463 if (req->newptr != 0) 464 return EPERM; 465 466 /* 467 * OK, now we're committed to doing something. 468 */ 469 s = splnet(); 470 gencnt = divcbinfo.ipi_gencnt; 471 n = divcbinfo.ipi_count; 472 splx(s); 473 474 xig.xig_len = sizeof xig; 475 xig.xig_count = n; 476 xig.xig_gen = gencnt; 477 xig.xig_sogen = so_gencnt; 478 error = SYSCTL_OUT(req, &xig, sizeof xig); 479 if (error) 480 return error; 481 482 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 483 if (inp_list == 0) 484 return ENOMEM; 485 486 s = splnet(); 487 for (inp = LIST_FIRST(divcbinfo.listhead), i = 0; inp && i < n; 488 inp = LIST_NEXT(inp, inp_list)) { 489 if (inp->inp_gencnt <= gencnt && !prison_xinpcb( 490 req->td->td_proc, inp)) 491 inp_list[i++] = inp; 492 } 493 splx(s); 494 n = i; 495 496 error = 0; 497 for (i = 0; i < n; i++) { 498 inp = inp_list[i]; 499 if (inp->inp_gencnt <= gencnt) { 500 struct xinpcb xi; 501 xi.xi_len = sizeof xi; 502 /* XXX should avoid extra copy */ 503 bcopy(inp, &xi.xi_inp, sizeof *inp); 504 if (inp->inp_socket) 505 sotoxsocket(inp->inp_socket, &xi.xi_socket); 506 error = SYSCTL_OUT(req, &xi, sizeof xi); 507 } 508 } 509 if (!error) { 510 /* 511 * Give the user an updated idea of our state. 512 * If the generation differs from what we told 513 * her before, she knows that something happened 514 * while we were processing this request, and it 515 * might be necessary to retry. 516 */ 517 s = splnet(); 518 xig.xig_gen = divcbinfo.ipi_gencnt; 519 xig.xig_sogen = so_gencnt; 520 xig.xig_count = divcbinfo.ipi_count; 521 splx(s); 522 error = SYSCTL_OUT(req, &xig, sizeof xig); 523 } 524 free(inp_list, M_TEMP); 525 return error; 526 } 527 528 SYSCTL_DECL(_net_inet_divert); 529 SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLFLAG_RD, 0, 0, 530 div_pcblist, "S,xinpcb", "List of active divert sockets"); 531 532 struct pr_usrreqs div_usrreqs = { 533 div_abort, pru_accept_notsupp, div_attach, div_bind, 534 pru_connect_notsupp, pru_connect2_notsupp, in_control, div_detach, 535 div_disconnect, pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp, 536 pru_rcvoob_notsupp, div_send, pru_sense_null, div_shutdown, 537 in_setsockaddr, sosend, soreceive, sopoll 538 }; 539