ip_divert.c revision 36903
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 * $Id: ip_divert.c,v 1.29 1998/06/06 20:45:25 julian Exp $ 34 */ 35 36#include "opt_inet.h" 37#include "opt_ipfw.h" 38 39#ifndef INET 40#error "IPDIVERT requires INET." 41#endif 42 43#include <sys/param.h> 44#include <sys/malloc.h> 45#include <sys/mbuf.h> 46#include <sys/socket.h> 47#include <sys/protosw.h> 48#include <sys/socketvar.h> 49#include <sys/systm.h> 50#include <sys/proc.h> 51 52#include <vm/vm_zone.h> 53 54#include <net/if.h> 55#include <net/route.h> 56 57#include <netinet/in.h> 58#include <netinet/in_systm.h> 59#include <netinet/ip.h> 60#include <netinet/in_pcb.h> 61#include <netinet/in_var.h> 62#include <netinet/ip_var.h> 63 64/* 65 * Divert sockets 66 */ 67 68/* 69 * Allocate enough space to hold a full IP packet 70 */ 71#define DIVSNDQ (65536 + 100) 72#define DIVRCVQ (65536 + 100) 73 74/* Global variables */ 75 76/* 77 * ip_input() and ip_output() set this secret value before calling us to 78 * let us know which divert port to divert a packet to; this is done so 79 * we can use the existing prototype for struct protosw's pr_input(). 80 * This is stored in host order. 81 */ 82u_short ip_divert_port; 83 84/* 85 * #ifdef IPFW_DIVERT_OLDRESTART 86 * We set this value to a non-zero port number when we want the call to 87 * ip_fw_chk() in ip_input() or ip_output() to ignore ``divert <port>'' 88 * chain entries. This is stored in host order. 89 * #else 90 * A 16 bit cookie is passed to the user process. 91 * The user process can send it back to help the caller know something 92 * about where the packet came from. 93 * 94 * If IPFW is the caller then the cookie is the rule that sent 95 * us here. On reinjection is is the rule after which processing 96 * should continue. Leaving it the same will make processing start 97 * at the rule number after that which sent it here. Setting it to 98 * 0 will restart processing at the beginning. 99 * #endif 100 */ 101u_int16_t ip_divert_cookie; 102 103/* Internal variables */ 104 105static struct inpcbhead divcb; 106static struct inpcbinfo divcbinfo; 107 108static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */ 109static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */ 110 111/* Optimization: have this preinitialized */ 112static struct sockaddr_in divsrc = { sizeof(divsrc), AF_INET }; 113 114/* Internal functions */ 115 116static int div_output(struct socket *so, 117 struct mbuf *m, struct sockaddr *addr, struct mbuf *control); 118 119/* 120 * Initialize divert connection block queue. 121 */ 122void 123div_init(void) 124{ 125 LIST_INIT(&divcb); 126 divcbinfo.listhead = &divcb; 127 /* 128 * XXX We don't use the hash list for divert IP, but it's easier 129 * to allocate a one entry hash list than it is to check all 130 * over the place for hashbase == NULL. 131 */ 132 divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask); 133 divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask); 134 divcbinfo.ipi_zone = zinit("divcb", sizeof(struct inpcb), 135 maxsockets, ZONE_INTERRUPT, 0); 136} 137 138/* 139 * Setup generic address and protocol structures 140 * for div_input routine, then pass them along with 141 * mbuf chain. ip->ip_len is assumed to have had 142 * the header length (hlen) subtracted out already. 143 * We tell whether the packet was incoming or outgoing 144 * by seeing if hlen == 0, which is a hack. 145 */ 146void 147div_input(struct mbuf *m, int hlen) 148{ 149 struct ip *ip; 150 struct inpcb *inp; 151 struct socket *sa; 152 153 /* Sanity check */ 154 if (ip_divert_port == 0) 155 panic("div_input: port is 0"); 156 157 /* Assure header */ 158 if (m->m_len < sizeof(struct ip) && 159 (m = m_pullup(m, sizeof(struct ip))) == 0) { 160 return; 161 } 162 ip = mtod(m, struct ip *); 163 164 /* Record divert port */ 165#ifdef IPFW_DIVERT_OLDRESTART 166 divsrc.sin_port = htons(ip_divert_cookie); 167#else 168 divsrc.sin_port = ip_divert_cookie; 169#endif /* IPFW_DIVERT_OLDRESTART */ 170 ip_divert_cookie = 0; 171 172 /* Restore packet header fields */ 173 ip->ip_len += hlen; 174 HTONS(ip->ip_len); 175 HTONS(ip->ip_off); 176 177 /* Record receive interface address, if any */ 178 divsrc.sin_addr.s_addr = 0; 179 if (hlen) { 180 struct ifaddr *ifa; 181 182#ifdef DIAGNOSTIC 183 /* Sanity check */ 184 if (!(m->m_flags & M_PKTHDR)) 185 panic("div_input: no pkt hdr"); 186#endif 187 188 /* More fields affected by ip_input() */ 189 HTONS(ip->ip_id); 190 191 /* Find IP address for receive interface */ 192 for (ifa = m->m_pkthdr.rcvif->if_addrhead.tqh_first; 193 ifa != NULL; ifa = ifa->ifa_link.tqe_next) { 194 if (ifa->ifa_addr == NULL) 195 continue; 196 if (ifa->ifa_addr->sa_family != AF_INET) 197 continue; 198 divsrc.sin_addr = 199 ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr; 200 break; 201 } 202 } 203 if (m->m_pkthdr.rcvif) { 204 char name[32]; 205 /* 206 * Hide the actual interface name in there in the 207 * sin_zero array. XXX This needs to be moved to a 208 * different sockaddr type for divert, e.g. 209 * sockaddr_div with multiple fields like 210 * sockaddr_dl. Presently we have only 7 bytes 211 * but that will do for now as most interfaces 212 * are 4 or less + 2 or less bytes for unit. 213 * There is probably a faster way of doing this, 214 * possibly taking it from the sockaddr_dl on the iface. 215 * This solves the problem of a P2P link and a LAN interface 216 * having the same address, which can result in the wrong 217 * interface being assigned to the packet when fed back 218 * into the divert socket. Theoretically if the daemon saves 219 * and re-uses the sockaddr_in as suggested in the man pages, 220 * this iface name will come along for the ride. 221 * (see div_output for the other half of this.) 222 */ 223 sprintf(name, "%s%d", 224 m->m_pkthdr.rcvif->if_name, m->m_pkthdr.rcvif->if_unit); 225 strncpy(divsrc.sin_zero, name, 7); 226 } 227 228 /* Put packet on socket queue, if any */ 229 sa = NULL; 230 for (inp = divcb.lh_first; inp != NULL; inp = inp->inp_list.le_next) { 231 if (inp->inp_lport == htons(ip_divert_port)) 232 sa = inp->inp_socket; 233 } 234 if (sa) { 235 if (sbappendaddr(&sa->so_rcv, (struct sockaddr *)&divsrc, 236 m, (struct mbuf *)0) == 0) 237 m_freem(m); 238 else 239 sorwakeup(sa); 240 } else { 241 m_freem(m); 242 ipstat.ips_noproto++; 243 ipstat.ips_delivered--; 244 } 245} 246 247/* 248 * Deliver packet back into the IP processing machinery. 249 * 250 * If no address specified, or address is 0.0.0.0, send to ip_output(); 251 * otherwise, send to ip_input() and mark as having been received on 252 * the interface with that address. 253 * 254 * If no address specified, or dest port is 0, allow packet to divert 255 * back to this socket; otherwise, don't. 256 */ 257static int 258div_output(so, m, addr, control) 259 struct socket *so; 260 register struct mbuf *m; 261 struct sockaddr *addr; 262 struct mbuf *control; 263{ 264 register struct inpcb *const inp = sotoinpcb(so); 265 register struct ip *const ip = mtod(m, struct ip *); 266 struct sockaddr_in *sin = (struct sockaddr_in *)addr; 267 int error = 0; 268 269 if (control) 270 m_freem(control); /* XXX */ 271 272 /* Loopback avoidance and state recovery */ 273 if (sin) { 274 int len = 0; 275 char *c = sin->sin_zero; 276#ifdef IPFW_DIVERT_OLDRESTART 277 ip_divert_cookie = ntohs(sin->sin_port); 278#else 279 ip_divert_cookie = sin->sin_port; 280#endif /* IPFW_DIVERT_OLDRESTART */ 281 282 /* 283 * Find receive interface with the given name or IP address. 284 * The name is user supplied data so don't trust it's size or 285 * that it is zero terminated. The name has priority. 286 * We are presently assuming that the sockaddr_in 287 * has not been replaced by a sockaddr_div, so we limit it 288 * to 16 bytes in total. the name is stuffed (if it exists) 289 * in the sin_zero[] field. 290 */ 291 while (*c++ && (len++ < sizeof(sin->sin_zero))); 292 if ((len > 0) && (len < sizeof(sin->sin_zero))) 293 m->m_pkthdr.rcvif = ifunit(sin->sin_zero); 294 } else { 295 ip_divert_cookie = 0; 296 } 297 298 /* Reinject packet into the system as incoming or outgoing */ 299 if (!sin || sin->sin_addr.s_addr == 0) { 300 /* 301 * Don't allow both user specified and setsockopt options, 302 * and don't allow packet length sizes that will crash 303 */ 304 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) || 305 ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) { 306 error = EINVAL; 307 goto cantsend; 308 } 309 310 /* Convert fields to host order for ip_output() */ 311 NTOHS(ip->ip_len); 312 NTOHS(ip->ip_off); 313 314 /* Send packet to output processing */ 315 ipstat.ips_rawout++; /* XXX */ 316 error = ip_output(m, inp->inp_options, &inp->inp_route, 317 (so->so_options & SO_DONTROUTE) | 318 IP_ALLOWBROADCAST | IP_RAWOUTPUT, inp->inp_moptions); 319 } else { 320 struct ifaddr *ifa; 321 322 /* If no luck with the name above. check by IP address. */ 323 if (m->m_pkthdr.rcvif == NULL) { 324 if (!(ifa = ifa_ifwithaddr((struct sockaddr *) sin))) { 325 error = EADDRNOTAVAIL; 326 goto cantsend; 327 } 328 m->m_pkthdr.rcvif = ifa->ifa_ifp; 329 } 330 331 /* Send packet to input processing */ 332 ip_input(m); 333 } 334 335 /* paranoid: Reset for next time (and other packets) */ 336 /* almost definitly already done in the ipfw filter but.. */ 337 ip_divert_cookie = 0; 338 return error; 339 340cantsend: 341 ip_divert_cookie = 0; 342 m_freem(m); 343 return error; 344} 345 346static int 347div_attach(struct socket *so, int proto, struct proc *p) 348{ 349 struct inpcb *inp; 350 int error, s; 351 352 inp = sotoinpcb(so); 353 if (inp) 354 panic("div_attach"); 355 if (p && (error = suser(p->p_ucred, &p->p_acflag)) != 0) 356 return error; 357 358 s = splnet(); 359 error = in_pcballoc(so, &divcbinfo, p); 360 splx(s); 361 if (error) 362 return error; 363 error = soreserve(so, div_sendspace, div_recvspace); 364 if (error) 365 return error; 366 inp = (struct inpcb *)so->so_pcb; 367 inp->inp_ip_p = proto; 368 inp->inp_flags |= INP_HDRINCL; 369 /* The socket is always "connected" because 370 we always know "where" to send the packet */ 371 so->so_state |= SS_ISCONNECTED; 372 return 0; 373} 374 375static int 376div_detach(struct socket *so) 377{ 378 struct inpcb *inp; 379 380 inp = sotoinpcb(so); 381 if (inp == 0) 382 panic("div_detach"); 383 in_pcbdetach(inp); 384 return 0; 385} 386 387static int 388div_abort(struct socket *so) 389{ 390 soisdisconnected(so); 391 return div_detach(so); 392} 393 394static int 395div_disconnect(struct socket *so) 396{ 397 if ((so->so_state & SS_ISCONNECTED) == 0) 398 return ENOTCONN; 399 return div_abort(so); 400} 401 402static int 403div_bind(struct socket *so, struct sockaddr *nam, struct proc *p) 404{ 405 struct inpcb *inp; 406 int s; 407 int error; 408 409 s = splnet(); 410 inp = sotoinpcb(so); 411 error = in_pcbbind(inp, nam, p); 412 splx(s); 413 return 0; 414} 415 416static int 417div_shutdown(struct socket *so) 418{ 419 socantsendmore(so); 420 return 0; 421} 422 423static int 424div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 425 struct mbuf *control, struct proc *p) 426{ 427 /* Packet must have a header (but that's about it) */ 428 if (m->m_len < sizeof (struct ip) || 429 (m = m_pullup(m, sizeof (struct ip))) == 0) { 430 ipstat.ips_toosmall++; 431 m_freem(m); 432 return EINVAL; 433 } 434 435 /* Send packet */ 436 return div_output(so, m, nam, control); 437} 438 439struct pr_usrreqs div_usrreqs = { 440 div_abort, pru_accept_notsupp, div_attach, div_bind, 441 pru_connect_notsupp, pru_connect2_notsupp, in_control, div_detach, 442 div_disconnect, pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp, 443 pru_rcvoob_notsupp, div_send, pru_sense_null, div_shutdown, 444 in_setsockaddr, sosend, soreceive, sopoll 445}; 446