2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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.
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
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 1.20 2005/02/08 22:56:19 hsu Exp $
39 #include "opt_ipdivert.h"
40 #include "opt_ipsec.h"
43 #error "IPDIVERT requires INET."
46 #include <sys/param.h>
47 #include <sys/kernel.h>
48 #include <sys/malloc.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>
57 #include <vm/vm_zone.h>
60 #include <net/route.h>
62 #include <netinet/in.h>
63 #include <netinet/in_systm.h>
64 #include <netinet/ip.h>
65 #include <netinet/in_pcb.h>
66 #include <netinet/in_var.h>
67 #include <netinet/ip_var.h>
74 * Allocate enough space to hold a full IP packet
76 #define DIVSNDQ (65536 + 100)
77 #define DIVRCVQ (65536 + 100)
80 * Divert sockets work in conjunction with ipfw, see the divert(4)
81 * manpage for features.
82 * Internally, packets selected by ipfw in ip_input() or ip_output(),
83 * and never diverted before, are passed to the input queue of the
84 * divert socket with a given 'divert_port' number (as specified in
85 * the matching ipfw rule), and they are tagged with a 16 bit cookie
86 * (representing the rule number of the matching ipfw rule), which
87 * is passed to process reading from the socket.
89 * Packets written to the divert socket are again tagged with a cookie
90 * (usually the same as above) and a destination address.
91 * If the destination address is INADDR_ANY then the packet is
92 * treated as outgoing and sent to ip_output(), otherwise it is
93 * treated as incoming and sent to ip_input().
94 * In both cases, the packet is tagged with the cookie.
96 * On reinjection, processing in ip_input() and ip_output()
97 * will be exactly the same as for the original packet, except that
98 * ipfw processing will start at the rule number after the one
99 * written in the cookie (so, tagging a packet with a cookie of 0
100 * will cause it to be effectively considered as a standard packet).
103 /* Internal variables */
104 static struct inpcbinfo divcbinfo;
106 static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */
107 static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */
110 * Initialize divert connection block queue.
115 in_pcbinfo_init(&divcbinfo);
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.
121 divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask);
122 divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask);
123 divcbinfo.wildcardhashbase = hashinit(1, M_PCB,
124 &divcbinfo.wildcardhashmask);
125 divcbinfo.ipi_zone = zinit("divcb", sizeof(struct inpcb),
126 maxsockets, ZONE_INTERRUPT, 0);
130 * IPPROTO_DIVERT is not a real IP protocol; don't allow any packets
131 * with that protocol number to enter the system from the outside.
134 div_input(struct mbuf *m, ...)
136 ipstat.ips_noproto++;
141 * Divert a packet by passing it up to the divert socket at port 'port'.
143 * Setup generic address and protocol structures for div_input routine,
144 * then pass them along with mbuf chain.
147 divert_packet(struct mbuf *m, int incoming, int port, int rule)
149 struct sockaddr_in divsrc = { sizeof divsrc, AF_INET };
156 KASSERT(port != 0, ("%s: port=0", __FUNCTION__));
158 divsrc.sin_port = rule; /* record matching rule */
161 if (m->m_len < sizeof(struct ip) &&
162 (m = m_pullup(m, sizeof(struct ip))) == NULL)
164 ip = mtod(m, struct ip *);
167 * Record receive interface address, if any.
168 * But only for incoming packets.
170 divsrc.sin_addr.s_addr = 0;
175 KASSERT((m->m_flags & M_PKTHDR), ("%s: !PKTHDR", __FUNCTION__));
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)
181 if (ifa->ifa_addr->sa_family != AF_INET)
184 ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr;
189 * Record the incoming interface name whenever we have one.
191 if (m->m_pkthdr.rcvif) {
193 * Hide the actual interface name in there in the
194 * sin_zero array. XXX This needs to be moved to a
195 * different sockaddr type for divert, e.g.
196 * sockaddr_div with multiple fields like
197 * sockaddr_dl. Presently we have only 7 bytes
198 * but that will do for now as most interfaces
199 * are 4 or less + 2 or less bytes for unit.
200 * There is probably a faster way of doing this,
201 * possibly taking it from the sockaddr_dl on the iface.
202 * This solves the problem of a P2P link and a LAN interface
203 * having the same address, which can result in the wrong
204 * interface being assigned to the packet when fed back
205 * into the divert socket. Theoretically if the daemon saves
206 * and re-uses the sockaddr_in as suggested in the man pages,
207 * this iface name will come along for the ride.
208 * (see div_output for the other half of this.)
210 snprintf(divsrc.sin_zero, sizeof divsrc.sin_zero,
211 m->m_pkthdr.rcvif->if_xname);
214 /* Put packet on socket queue, if any */
216 nport = htons((u_int16_t)port);
217 LIST_FOREACH(inp, &divcbinfo.pcblisthead, inp_list) {
218 if (inp->inp_flags & INP_PLACEMARKER)
220 if (inp->inp_lport == nport)
221 sa = inp->inp_socket;
224 if (sbappendaddr(&sa->so_rcv, (struct sockaddr *)&divsrc, m,
225 (struct mbuf *)NULL) == 0)
231 ipstat.ips_noproto++;
232 ipstat.ips_delivered--;
237 * Deliver packet back into the IP processing machinery.
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.
244 div_output(struct socket *so, struct mbuf *m,
245 struct sockaddr_in *sin, struct mbuf *control)
248 struct m_hdr divert_tag;
251 * Prepare the tag for divert info. Note that a packet
252 * with a 0 tag in mh_data is effectively untagged,
253 * so we could optimize that case.
255 divert_tag.mh_type = MT_TAG;
256 divert_tag.mh_flags = PACKET_TAG_DIVERT;
257 divert_tag.mh_next = m;
258 divert_tag.mh_data = 0; /* the matching rule # */
259 m->m_pkthdr.rcvif = NULL; /* XXX is it necessary ? */
262 m_freem(control); /* XXX */
264 /* Loopback avoidance and state recovery */
268 divert_tag.mh_data = (caddr_t)(int)sin->sin_port;
270 * Find receive interface with the given name, stuffed
271 * (if it exists) in the sin_zero[] field.
272 * The name is user supplied data so don't trust its size
273 * or that it is zero terminated.
275 for (i = 0; sin->sin_zero[i] && i < sizeof sin->sin_zero; i++)
277 if ( i > 0 && i < sizeof sin->sin_zero)
278 m->m_pkthdr.rcvif = ifunit(sin->sin_zero);
281 /* Reinject packet into the system as incoming or outgoing */
282 if (!sin || sin->sin_addr.s_addr == 0) {
283 struct inpcb *const inp = so->so_pcb;
284 struct ip *const ip = mtod(m, struct ip *);
287 * Don't allow both user specified and setsockopt options,
288 * and don't allow packet length sizes that will crash
290 if (((ip->ip_hl != (sizeof *ip) >> 2) && inp->inp_options) ||
291 ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) {
296 /* Convert fields to host order for ip_output() */
297 ip->ip_len = ntohs(ip->ip_len);
298 ip->ip_off = ntohs(ip->ip_off);
300 /* Send packet to output processing */
301 ipstat.ips_rawout++; /* XXX */
302 error = ip_output((struct mbuf *)&divert_tag,
303 inp->inp_options, &inp->inp_route,
304 (so->so_options & SO_DONTROUTE) |
305 IP_ALLOWBROADCAST | IP_RAWOUTPUT,
306 inp->inp_moptions, NULL);
308 if (m->m_pkthdr.rcvif == NULL) {
310 * No luck with the name, check by IP address.
311 * Clear the port and the ifname to make sure
312 * there are no distractions for ifa_ifwithaddr.
316 bzero(sin->sin_zero, sizeof sin->sin_zero);
318 ifa = ifa_ifwithaddr((struct sockaddr *) sin);
320 error = EADDRNOTAVAIL;
323 m->m_pkthdr.rcvif = ifa->ifa_ifp;
325 ip_input((struct mbuf *)&divert_tag);
336 div_attach(struct socket *so, int proto, struct pru_attach_info *ai)
344 if ((error = suser_cred(ai->p_ucred, NULL_CRED_OKAY)) != 0)
347 error = soreserve(so, div_sendspace, div_recvspace, ai->sb_rlimit);
351 error = in_pcballoc(so, &divcbinfo);
355 inp = (struct inpcb *)so->so_pcb;
356 inp->inp_ip_p = proto;
357 inp->inp_vflag |= INP_IPV4;
358 inp->inp_flags |= INP_HDRINCL;
360 * The socket is always "connected" because
361 * we always know "where" to send the packet.
363 so->so_state |= SS_ISCONNECTED;
368 div_detach(struct socket *so)
380 div_abort(struct socket *so)
382 soisdisconnected(so);
383 return div_detach(so);
387 div_disconnect(struct socket *so)
389 if (!(so->so_state & SS_ISCONNECTED))
391 return div_abort(so);
395 div_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
403 /* in_pcbbind assumes that nam is a sockaddr_in
404 * and in_pcbbind requires a valid address. Since divert
405 * sockets don't we need to make sure the address is
406 * filled in properly.
407 * XXX -- divert should not be abusing in_pcbind
408 * and should probably have its own family.
410 if (nam->sa_family != AF_INET)
411 error = EAFNOSUPPORT;
413 ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY;
414 error = in_pcbbind(inp, nam, td);
421 div_shutdown(struct socket *so)
428 div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
429 struct mbuf *control, struct thread *td)
431 /* Packet must have a header (but that's about it) */
432 if (m->m_len < sizeof(struct ip) &&
433 (m = m_pullup(m, sizeof(struct ip))) == NULL) {
434 ipstat.ips_toosmall++;
440 return div_output(so, m, (struct sockaddr_in *)nam, control);
443 SYSCTL_DECL(_net_inet_divert);
444 SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLFLAG_RD, &divcbinfo, 0,
445 in_pcblist_global, "S,xinpcb", "List of active divert sockets");
447 struct pr_usrreqs div_usrreqs = {
448 div_abort, pru_accept_notsupp, div_attach, div_bind,
449 pru_connect_notsupp, pru_connect2_notsupp, in_control, div_detach,
450 div_disconnect, pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp,
451 pru_rcvoob_notsupp, div_send, pru_sense_null, div_shutdown,
452 in_setsockaddr, sosend, soreceive, sopoll