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 1.12 2004/04/13 00:14:01 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 ? */
109 /* Optimization: have this preinitialized */
110 static struct sockaddr_in divsrc = { sizeof(divsrc), AF_INET };
113 * Initialize divert connection block queue.
118 LIST_INIT(&divcbinfo.listhead);
120 * XXX We don't use the hash list for divert IP, but it's easier
121 * to allocate a one entry hash list than it is to check all
122 * over the place for hashbase == NULL.
124 divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask);
125 divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask);
126 divcbinfo.wildcardhashbase = hashinit(1, M_PCB,
127 &divcbinfo.wildcardhashmask);
128 divcbinfo.ipi_zone = zinit("divcb", sizeof(struct inpcb),
129 maxsockets, ZONE_INTERRUPT, 0);
133 * IPPROTO_DIVERT is not a real IP protocol; don't allow any packets
134 * with that protocol number to enter the system from the outside.
137 div_input(struct mbuf *m, int off, int proto)
139 ipstat.ips_noproto++;
144 * Divert a packet by passing it up to the divert socket at port 'port'.
146 * Setup generic address and protocol structures for div_input routine,
147 * then pass them along with mbuf chain.
150 divert_packet(struct mbuf *m, int incoming, int port, int rule)
158 KASSERT(port != 0, ("%s: port=0", __FUNCTION__));
160 divsrc.sin_port = rule; /* record matching rule */
163 if (m->m_len < sizeof(struct ip) &&
164 (m = m_pullup(m, sizeof(struct ip))) == 0)
166 ip = mtod(m, struct ip *);
169 * Record receive interface address, if any.
170 * But only for incoming packets.
172 divsrc.sin_addr.s_addr = 0;
177 KASSERT((m->m_flags & M_PKTHDR), ("%s: !PKTHDR", __FUNCTION__));
179 /* Find IP address for receive interface */
180 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
181 if (ifa->ifa_addr == NULL)
183 if (ifa->ifa_addr->sa_family != AF_INET)
186 ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr;
191 * Record the incoming interface name whenever we have one.
193 bzero(&divsrc.sin_zero, sizeof(divsrc.sin_zero));
194 if (m->m_pkthdr.rcvif) {
196 * Hide the actual interface name in there in the
197 * sin_zero array. XXX This needs to be moved to a
198 * different sockaddr type for divert, e.g.
199 * sockaddr_div with multiple fields like
200 * sockaddr_dl. Presently we have only 7 bytes
201 * but that will do for now as most interfaces
202 * are 4 or less + 2 or less bytes for unit.
203 * There is probably a faster way of doing this,
204 * possibly taking it from the sockaddr_dl on the iface.
205 * This solves the problem of a P2P link and a LAN interface
206 * having the same address, which can result in the wrong
207 * interface being assigned to the packet when fed back
208 * into the divert socket. Theoretically if the daemon saves
209 * and re-uses the sockaddr_in as suggested in the man pages,
210 * this iface name will come along for the ride.
211 * (see div_output for the other half of this.)
213 snprintf(divsrc.sin_zero, sizeof(divsrc.sin_zero),
214 m->m_pkthdr.rcvif->if_xname);
217 /* Put packet on socket queue, if any */
219 nport = htons((u_int16_t)port);
220 LIST_FOREACH(inp, &divcbinfo.listhead, inp_list) {
221 if (inp->inp_lport == nport)
222 sa = inp->inp_socket;
225 if (sbappendaddr(&sa->so_rcv, (struct sockaddr *)&divsrc,
226 m, (struct mbuf *)0) == 0)
232 ipstat.ips_noproto++;
233 ipstat.ips_delivered--;
238 * Deliver packet back into the IP processing machinery.
240 * If no address specified, or address is 0.0.0.0, send to ip_output();
241 * otherwise, send to ip_input() and mark as having been received on
242 * the interface with that address.
245 div_output(struct socket *so, struct mbuf *m,
246 struct sockaddr_in *sin, struct mbuf *control)
249 struct m_hdr divert_tag;
252 * Prepare the tag for divert info. Note that a packet
253 * with a 0 tag in mh_data is effectively untagged,
254 * so we could optimize that case.
256 divert_tag.mh_type = MT_TAG;
257 divert_tag.mh_flags = PACKET_TAG_DIVERT;
258 divert_tag.mh_next = m;
259 divert_tag.mh_data = 0; /* the matching rule # */
260 m->m_pkthdr.rcvif = NULL; /* XXX is it necessary ? */
263 m_freem(control); /* XXX */
265 /* Loopback avoidance and state recovery */
269 divert_tag.mh_data = (caddr_t)(int)sin->sin_port;
271 * Find receive interface with the given name, stuffed
272 * (if it exists) in the sin_zero[] field.
273 * The name is user supplied data so don't trust its size
274 * or that it is zero terminated.
276 for (i = 0; sin->sin_zero[i] && i < sizeof(sin->sin_zero); i++)
278 if ( i > 0 && i < sizeof(sin->sin_zero))
279 m->m_pkthdr.rcvif = ifunit(sin->sin_zero);
282 /* Reinject packet into the system as incoming or outgoing */
283 if (!sin || sin->sin_addr.s_addr == 0) {
284 struct inpcb *const inp = sotoinpcb(so);
285 struct ip *const ip = mtod(m, struct ip *);
288 * Don't allow both user specified and setsockopt options,
289 * and don't allow packet length sizes that will crash
291 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) ||
292 ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) {
297 /* Convert fields to host order for ip_output() */
298 ip->ip_len = ntohs(ip->ip_len);
299 ip->ip_off = ntohs(ip->ip_off);
301 /* Send packet to output processing */
302 ipstat.ips_rawout++; /* XXX */
303 error = ip_output((struct mbuf *)&divert_tag,
304 inp->inp_options, &inp->inp_route,
305 (so->so_options & SO_DONTROUTE) |
306 IP_ALLOWBROADCAST | IP_RAWOUTPUT,
307 inp->inp_moptions, NULL);
309 if (m->m_pkthdr.rcvif == NULL) {
311 * No luck with the name, check by IP address.
312 * Clear the port and the ifname to make sure
313 * there are no distractions for ifa_ifwithaddr.
317 bzero(sin->sin_zero, sizeof(sin->sin_zero));
319 ifa = ifa_ifwithaddr((struct sockaddr *) sin);
321 error = EADDRNOTAVAIL;
324 m->m_pkthdr.rcvif = ifa->ifa_ifp;
326 ip_input((struct mbuf *)&divert_tag);
337 div_attach(struct socket *so, int proto, struct pru_attach_info *ai)
345 if ((error = suser_cred(ai->p_ucred, NULL_CRED_OKAY)) != 0)
348 error = soreserve(so, div_sendspace, div_recvspace, ai->sb_rlimit);
352 error = in_pcballoc(so, &divcbinfo);
356 inp = (struct inpcb *)so->so_pcb;
357 inp->inp_ip_p = proto;
358 inp->inp_vflag |= INP_IPV4;
359 inp->inp_flags |= INP_HDRINCL;
360 /* The socket is always "connected" because
361 we always know "where" to send the packet */
362 so->so_state |= SS_ISCONNECTED;
367 div_detach(struct socket *so)
379 div_abort(struct socket *so)
381 soisdisconnected(so);
382 return div_detach(so);
386 div_disconnect(struct socket *so)
388 if ((so->so_state & SS_ISCONNECTED) == 0)
390 return div_abort(so);
394 div_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
402 /* in_pcbbind assumes that nam is a sockaddr_in
403 * and in_pcbbind requires a valid address. Since divert
404 * sockets don't we need to make sure the address is
405 * filled in properly.
406 * XXX -- divert should not be abusing in_pcbind
407 * and should probably have its own family.
409 if (nam->sa_family != AF_INET)
410 error = EAFNOSUPPORT;
412 ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY;
413 error = in_pcbbind(inp, nam, td);
420 div_shutdown(struct socket *so)
427 div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
428 struct mbuf *control, struct thread *td)
430 /* Packet must have a header (but that's about it) */
431 if (m->m_len < sizeof (struct ip) &&
432 (m = m_pullup(m, sizeof (struct ip))) == 0) {
433 ipstat.ips_toosmall++;
439 return div_output(so, m, (struct sockaddr_in *)nam, control);
443 div_pcblist(SYSCTL_HANDLER_ARGS)
446 struct inpcb *inp, **inp_list;
451 * The process of preparing the TCB list is too time-consuming and
452 * resource-intensive to repeat twice on every request.
454 if (req->oldptr == 0) {
455 n = divcbinfo.ipi_count;
456 req->oldidx = 2 * (sizeof xig)
457 + (n + n/8) * sizeof(struct xinpcb);
461 if (req->newptr != 0)
465 * OK, now we're committed to doing something.
468 gencnt = divcbinfo.ipi_gencnt;
469 n = divcbinfo.ipi_count;
472 xig.xig_len = sizeof xig;
474 xig.xig_gen = gencnt;
475 xig.xig_sogen = so_gencnt;
476 error = SYSCTL_OUT(req, &xig, sizeof xig);
480 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
485 for (inp = LIST_FIRST(&divcbinfo.listhead), i = 0; inp && i < n;
486 inp = LIST_NEXT(inp, inp_list)) {
487 if (inp->inp_gencnt <= gencnt && !prison_xinpcb(req->td, inp))
494 for (i = 0; i < n; i++) {
496 if (inp->inp_gencnt <= gencnt) {
498 xi.xi_len = sizeof xi;
499 /* XXX should avoid extra copy */
500 bcopy(inp, &xi.xi_inp, sizeof *inp);
502 sotoxsocket(inp->inp_socket, &xi.xi_socket);
503 error = SYSCTL_OUT(req, &xi, sizeof xi);
508 * Give the user an updated idea of our state.
509 * If the generation differs from what we told
510 * her before, she knows that something happened
511 * while we were processing this request, and it
512 * might be necessary to retry.
515 xig.xig_gen = divcbinfo.ipi_gencnt;
516 xig.xig_sogen = so_gencnt;
517 xig.xig_count = divcbinfo.ipi_count;
519 error = SYSCTL_OUT(req, &xig, sizeof xig);
521 free(inp_list, M_TEMP);
525 SYSCTL_DECL(_net_inet_divert);
526 SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLFLAG_RD, 0, 0,
527 div_pcblist, "S,xinpcb", "List of active divert sockets");
529 struct pr_usrreqs div_usrreqs = {
530 div_abort, pru_accept_notsupp, div_attach, div_bind,
531 pru_connect_notsupp, pru_connect2_notsupp, in_control, div_detach,
532 div_disconnect, pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp,
533 pru_rcvoob_notsupp, div_send, pru_sense_null, div_shutdown,
534 in_setsockaddr, sosend, soreceive, sopoll