ip_divert.c revision 54175
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: head/sys/netinet/ip_divert.c 54175 1999-12-06 00:43:07Z archie $ 34 */ 35 36#include "opt_inet.h" 37#include "opt_ipfw.h" 38#include "opt_ipdivert.h" 39 40#ifndef INET 41#error "IPDIVERT requires INET." 42#endif 43 44#include <sys/param.h> 45#include <sys/malloc.h> 46#include <sys/mbuf.h> 47#include <sys/socket.h> 48#include <sys/protosw.h> 49#include <sys/socketvar.h> 50#include <sys/systm.h> 51#include <sys/proc.h> 52 53#include <vm/vm_zone.h> 54 55#include <net/if.h> 56#include <net/route.h> 57 58#include <netinet/in.h> 59#include <netinet/in_systm.h> 60#include <netinet/ip.h> 61#include <netinet/in_pcb.h> 62#include <netinet/in_var.h> 63#include <netinet/ip_var.h> 64 65/* 66 * Divert sockets 67 */ 68 69/* 70 * Allocate enough space to hold a full IP packet 71 */ 72#define DIVSNDQ (65536 + 100) 73#define DIVRCVQ (65536 + 100) 74 75/* 76 * A 16 bit cookie is passed to and from the user process. 77 * The user process can send it back to help the caller know 78 * something about where the packet originally came from. 79 * 80 * In the case of ipfw, then the cookie is the rule that sent 81 * us here. On reinjection is is the rule after which processing 82 * should continue. Leaving it the same will make processing start 83 * at the rule number after that which sent it here. Setting it to 84 * 0 will restart processing at the beginning. 85 * 86 * For divert_packet(), ip_divert_cookie is an input value only. 87 * For div_output(), ip_divert_cookie is an output value only. 88 */ 89u_int16_t ip_divert_cookie; 90 91/* Internal variables */ 92static struct inpcbhead divcb; 93static struct inpcbinfo divcbinfo; 94 95static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */ 96static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */ 97 98/* Optimization: have this preinitialized */ 99static struct sockaddr_in divsrc = { sizeof(divsrc), AF_INET }; 100 101/* Internal functions */ 102static int div_output(struct socket *so, 103 struct mbuf *m, struct sockaddr *addr, struct mbuf *control); 104 105/* 106 * Initialize divert connection block queue. 107 */ 108void 109div_init(void) 110{ 111 LIST_INIT(&divcb); 112 divcbinfo.listhead = &divcb; 113 /* 114 * XXX We don't use the hash list for divert IP, but it's easier 115 * to allocate a one entry hash list than it is to check all 116 * over the place for hashbase == NULL. 117 */ 118 divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask); 119 divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask); 120 divcbinfo.ipi_zone = zinit("divcb", sizeof(struct inpcb), 121 maxsockets, ZONE_INTERRUPT, 0); 122} 123 124/* 125 * IPPROTO_DIVERT is not a real IP protocol; don't allow any packets 126 * with that protocol number to enter the system from the outside. 127 */ 128void 129div_input(struct mbuf *m, int hlen) 130{ 131 ipstat.ips_noproto++; 132 m_freem(m); 133} 134 135/* 136 * Divert a packet by passing it up to the divert socket at port 'port'. 137 * 138 * Setup generic address and protocol structures for div_input routine, 139 * then pass them along with mbuf chain. 140 */ 141void 142divert_packet(struct mbuf *m, int incoming, int port) 143{ 144 struct ip *ip; 145 struct inpcb *inp; 146 struct socket *sa; 147 u_int16_t nport; 148 149 /* Sanity check */ 150 KASSERT(port != 0, ("%s: port=0", __FUNCTION__)); 151 152 /* Record and reset divert cookie */ 153 divsrc.sin_port = ip_divert_cookie; 154 ip_divert_cookie = 0; 155 156 /* Assure header */ 157 if (m->m_len < sizeof(struct ip) && 158 (m = m_pullup(m, sizeof(struct ip))) == 0) { 159 return; 160 } 161 ip = mtod(m, struct ip *); 162 163 /* 164 * Record receive interface address, if any. 165 * But only for incoming packets. 166 */ 167 divsrc.sin_addr.s_addr = 0; 168 if (incoming) { 169 struct ifaddr *ifa; 170 171 /* Sanity check */ 172 KASSERT((m->m_flags & M_PKTHDR), ("%s: !PKTHDR", __FUNCTION__)); 173 174 /* Find IP address for receive interface */ 175 for (ifa = m->m_pkthdr.rcvif->if_addrhead.tqh_first; 176 ifa != NULL; ifa = ifa->ifa_link.tqe_next) { 177 if (ifa->ifa_addr == NULL) 178 continue; 179 if (ifa->ifa_addr->sa_family != AF_INET) 180 continue; 181 divsrc.sin_addr = 182 ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr; 183 break; 184 } 185 } 186 /* 187 * Record the incoming interface name whenever we have one. 188 */ 189 bzero(&divsrc.sin_zero, sizeof(divsrc.sin_zero)); 190 if (m->m_pkthdr.rcvif) { 191 /* 192 * Hide the actual interface name in there in the 193 * sin_zero array. XXX This needs to be moved to a 194 * different sockaddr type for divert, e.g. 195 * sockaddr_div with multiple fields like 196 * sockaddr_dl. Presently we have only 7 bytes 197 * but that will do for now as most interfaces 198 * are 4 or less + 2 or less bytes for unit. 199 * There is probably a faster way of doing this, 200 * possibly taking it from the sockaddr_dl on the iface. 201 * This solves the problem of a P2P link and a LAN interface 202 * having the same address, which can result in the wrong 203 * interface being assigned to the packet when fed back 204 * into the divert socket. Theoretically if the daemon saves 205 * and re-uses the sockaddr_in as suggested in the man pages, 206 * this iface name will come along for the ride. 207 * (see div_output for the other half of this.) 208 */ 209 snprintf(divsrc.sin_zero, sizeof(divsrc.sin_zero), 210 "%s%d", m->m_pkthdr.rcvif->if_name, 211 m->m_pkthdr.rcvif->if_unit); 212 } 213 214 /* Put packet on socket queue, if any */ 215 sa = NULL; 216 nport = htons((u_int16_t)port); 217 for (inp = divcb.lh_first; inp != NULL; inp = inp->inp_list.le_next) { 218 if (inp->inp_lport == nport) 219 sa = inp->inp_socket; 220 } 221 if (sa) { 222 if (sbappendaddr(&sa->so_rcv, (struct sockaddr *)&divsrc, 223 m, (struct mbuf *)0) == 0) 224 m_freem(m); 225 else 226 sorwakeup(sa); 227 } else { 228 m_freem(m); 229 ipstat.ips_noproto++; 230 ipstat.ips_delivered--; 231 } 232} 233 234/* 235 * Deliver packet back into the IP processing machinery. 236 * 237 * If no address specified, or address is 0.0.0.0, send to ip_output(); 238 * otherwise, send to ip_input() and mark as having been received on 239 * the interface with that address. 240 */ 241static int 242div_output(so, m, addr, control) 243 struct socket *so; 244 register struct mbuf *m; 245 struct sockaddr *addr; 246 struct mbuf *control; 247{ 248 register struct inpcb *const inp = sotoinpcb(so); 249 register struct ip *const ip = mtod(m, struct ip *); 250 struct sockaddr_in *sin = (struct sockaddr_in *)addr; 251 int error = 0; 252 253 if (control) 254 m_freem(control); /* XXX */ 255 256 /* Loopback avoidance and state recovery */ 257 if (sin) { 258 int len = 0; 259 char *c = sin->sin_zero; 260 261 ip_divert_cookie = sin->sin_port; 262 263 /* 264 * Find receive interface with the given name or IP address. 265 * The name is user supplied data so don't trust it's size or 266 * that it is zero terminated. The name has priority. 267 * We are presently assuming that the sockaddr_in 268 * has not been replaced by a sockaddr_div, so we limit it 269 * to 16 bytes in total. the name is stuffed (if it exists) 270 * in the sin_zero[] field. 271 */ 272 while (*c++ && (len++ < sizeof(sin->sin_zero))); 273 if ((len > 0) && (len < sizeof(sin->sin_zero))) 274 m->m_pkthdr.rcvif = ifunit(sin->sin_zero); 275 } else { 276 ip_divert_cookie = 0; 277 } 278 279 /* Reinject packet into the system as incoming or outgoing */ 280 if (!sin || sin->sin_addr.s_addr == 0) { 281 /* 282 * Don't allow both user specified and setsockopt options, 283 * and don't allow packet length sizes that will crash 284 */ 285 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) || 286 ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) { 287 error = EINVAL; 288 goto cantsend; 289 } 290 291 /* Convert fields to host order for ip_output() */ 292 NTOHS(ip->ip_len); 293 NTOHS(ip->ip_off); 294 295 /* Send packet to output processing */ 296 ipstat.ips_rawout++; /* XXX */ 297 error = ip_output(m, inp->inp_options, &inp->inp_route, 298 (so->so_options & SO_DONTROUTE) | 299 IP_ALLOWBROADCAST | IP_RAWOUTPUT, inp->inp_moptions); 300 } else { 301 struct ifaddr *ifa; 302 303 /* If no luck with the name above. check by IP address. */ 304 if (m->m_pkthdr.rcvif == NULL) { 305 /* 306 * Make sure there are no distractions 307 * for ifa_ifwithaddr. Clear the port and the ifname. 308 * Maybe zap all 8 bytes at once using a 64bit write? 309 */ 310 bzero(sin->sin_zero, sizeof(sin->sin_zero)); 311 /* *((u_int64_t *)sin->sin_zero) = 0; */ /* XXX ?? */ 312 sin->sin_port = 0; 313 if (!(ifa = ifa_ifwithaddr((struct sockaddr *) sin))) { 314 error = EADDRNOTAVAIL; 315 goto cantsend; 316 } 317 m->m_pkthdr.rcvif = ifa->ifa_ifp; 318 } 319 320 /* Send packet to input processing */ 321 ip_input(m); 322 } 323 324 /* paranoid: Reset for next time (and other packets) */ 325 /* almost definitly already done in the ipfw filter but.. */ 326 ip_divert_cookie = 0; 327 return error; 328 329cantsend: 330 m_freem(m); 331 ip_divert_cookie = 0; 332 return error; 333} 334 335static int 336div_attach(struct socket *so, int proto, struct proc *p) 337{ 338 struct inpcb *inp; 339 int error, s; 340 341 inp = sotoinpcb(so); 342 if (inp) 343 panic("div_attach"); 344 if (p && (error = suser(p)) != 0) 345 return error; 346 347 s = splnet(); 348 error = in_pcballoc(so, &divcbinfo, p); 349 splx(s); 350 if (error) 351 return error; 352 error = soreserve(so, div_sendspace, div_recvspace); 353 if (error) 354 return error; 355 inp = (struct inpcb *)so->so_pcb; 356 inp->inp_ip_p = proto; 357 inp->inp_flags |= INP_HDRINCL; 358 /* The socket is always "connected" because 359 we always know "where" to send the packet */ 360 so->so_state |= SS_ISCONNECTED; 361 return 0; 362} 363 364static int 365div_detach(struct socket *so) 366{ 367 struct inpcb *inp; 368 369 inp = sotoinpcb(so); 370 if (inp == 0) 371 panic("div_detach"); 372 in_pcbdetach(inp); 373 return 0; 374} 375 376static int 377div_abort(struct socket *so) 378{ 379 soisdisconnected(so); 380 return div_detach(so); 381} 382 383static int 384div_disconnect(struct socket *so) 385{ 386 if ((so->so_state & SS_ISCONNECTED) == 0) 387 return ENOTCONN; 388 return div_abort(so); 389} 390 391static int 392div_bind(struct socket *so, struct sockaddr *nam, struct proc *p) 393{ 394 struct inpcb *inp; 395 int s; 396 int error; 397 398 s = splnet(); 399 inp = sotoinpcb(so); 400 error = in_pcbbind(inp, nam, p); 401 splx(s); 402 return 0; 403} 404 405static int 406div_shutdown(struct socket *so) 407{ 408 socantsendmore(so); 409 return 0; 410} 411 412static int 413div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 414 struct mbuf *control, struct proc *p) 415{ 416 /* Packet must have a header (but that's about it) */ 417 if (m->m_len < sizeof (struct ip) || 418 (m = m_pullup(m, sizeof (struct ip))) == 0) { 419 ipstat.ips_toosmall++; 420 m_freem(m); 421 return EINVAL; 422 } 423 424 /* Send packet */ 425 return div_output(so, m, nam, control); 426} 427 428struct pr_usrreqs div_usrreqs = { 429 div_abort, pru_accept_notsupp, div_attach, div_bind, 430 pru_connect_notsupp, pru_connect2_notsupp, in_control, div_detach, 431 div_disconnect, pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp, 432 pru_rcvoob_notsupp, div_send, pru_sense_null, div_shutdown, 433 in_setsockaddr, sosend, soreceive, sopoll 434}; 435