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 1.38 2008/08/28 14:10:03 sephe 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>
56 #include <sys/thread2.h>
58 #include <sys/msgport.h>
61 #include <vm/vm_zone.h>
64 #include <net/route.h>
66 #include <net/netmsg2.h>
69 #include <netinet/in.h>
70 #include <netinet/in_systm.h>
71 #include <netinet/ip.h>
72 #include <netinet/in_pcb.h>
73 #include <netinet/in_var.h>
74 #include <netinet/ip_var.h>
75 #include <netinet/ip_divert.h>
82 * Allocate enough space to hold a full IP packet
84 #define DIVSNDQ (65536 + 100)
85 #define DIVRCVQ (65536 + 100)
87 #define DIV_IS_OUTPUT(sin) ((sin) == NULL || (sin)->sin_addr.s_addr == 0)
89 #define DIV_OUTPUT 0x10000
90 #define DIV_INPUT 0x20000
93 * Divert sockets work in conjunction with ipfw, see the divert(4)
94 * manpage for features.
95 * Internally, packets selected by ipfw in ip_input() or ip_output(),
96 * and never diverted before, are passed to the input queue of the
97 * divert socket with a given 'divert_port' number (as specified in
98 * the matching ipfw rule), and they are tagged with a 16 bit cookie
99 * (representing the rule number of the matching ipfw rule), which
100 * is passed to process reading from the socket.
102 * Packets written to the divert socket are again tagged with a cookie
103 * (usually the same as above) and a destination address.
104 * If the destination address is INADDR_ANY then the packet is
105 * treated as outgoing and sent to ip_output(), otherwise it is
106 * treated as incoming and sent to ip_input().
107 * In both cases, the packet is tagged with the cookie.
109 * On reinjection, processing in ip_input() and ip_output()
110 * will be exactly the same as for the original packet, except that
111 * ipfw processing will start at the rule number after the one
112 * written in the cookie (so, tagging a packet with a cookie of 0
113 * will cause it to be effectively considered as a standard packet).
116 /* Internal variables */
117 static struct inpcbinfo divcbinfo;
119 static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */
120 static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */
123 * Initialize divert connection block queue.
128 in_pcbinfo_init(&divcbinfo);
130 * XXX We don't use the hash list for divert IP, but it's easier
131 * to allocate a one entry hash list than it is to check all
132 * over the place for hashbase == NULL.
134 divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask);
135 divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask);
136 divcbinfo.wildcardhashbase = hashinit(1, M_PCB,
137 &divcbinfo.wildcardhashmask);
138 divcbinfo.ipi_zone = zinit("divcb", sizeof(struct inpcb),
139 maxsockets, ZONE_INTERRUPT, 0);
143 * IPPROTO_DIVERT is not a real IP protocol; don't allow any packets
144 * with that protocol number to enter the system from the outside.
147 div_input(struct mbuf *m, ...)
149 ipstat.ips_noproto++;
154 div_soport(struct socket *so, struct sockaddr *nam,
155 struct mbuf **mptr, int req)
157 struct sockaddr_in *sin;
161 /* Except for send(), everything happens on CPU0 */
163 return cpu0_soport(so, nam, mptr, req);
165 sin = (struct sockaddr_in *)nam;
169 m->m_pkthdr.rcvif = NULL;
170 dir = DIV_IS_OUTPUT(sin) ? IP_MPORT_OUT : IP_MPORT_IN;
176 * Try locating the interface, if we originally had one.
177 * This is done even for outgoing packets, since for a
178 * forwarded packet, there must be an interface attached.
180 * Find receive interface with the given name, stuffed
181 * (if it exists) in the sin_zero[] field.
182 * The name is user supplied data so don't trust its size
183 * or that it is zero terminated.
185 for (i = 0; sin->sin_zero[i] && i < sizeof(sin->sin_zero); i++)
187 if (i > 0 && i < sizeof(sin->sin_zero))
188 m->m_pkthdr.rcvif = ifunit(sin->sin_zero);
191 if (dir == IP_MPORT_IN && m->m_pkthdr.rcvif == NULL) {
193 * No luck with the name, check by IP address.
194 * Clear the port and the ifname to make sure
195 * there are no distractions for ifa_ifwithaddr.
197 * Be careful not to trash sin->sin_port; it will
198 * be used later in div_output().
203 bzero(sin->sin_zero, sizeof(sin->sin_zero));
204 sin_port = sin->sin_port; /* save */
206 ifa = ifa_ifwithaddr((struct sockaddr *)sin);
212 sin->sin_port = sin_port; /* restore */
213 m->m_pkthdr.rcvif = ifa->ifa_ifp;
216 return ip_mport(mptr, dir);
220 * Divert a packet by passing it up to the divert socket at port 'port'.
222 * Setup generic address and protocol structures for div_input routine,
223 * then pass them along with mbuf chain.
226 div_packet(struct mbuf *m, int incoming, int port)
228 struct sockaddr_in divsrc = { sizeof divsrc, AF_INET };
232 struct divert_info *divinfo;
235 /* Locate the divert info */
236 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
237 divinfo = m_tag_data(mtag);
238 divsrc.sin_port = divinfo->skipto;
241 * Record receive interface address, if any.
242 * But only for incoming packets.
244 divsrc.sin_addr.s_addr = 0;
246 struct ifaddr_container *ifac;
248 /* Find IP address for receive interface */
249 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid],
251 struct ifaddr *ifa = ifac->ifa;
253 if (ifa->ifa_addr == NULL)
255 if (ifa->ifa_addr->sa_family != AF_INET)
258 ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr;
263 * Record the incoming interface name whenever we have one.
265 if (m->m_pkthdr.rcvif) {
267 * Hide the actual interface name in there in the
268 * sin_zero array. XXX This needs to be moved to a
269 * different sockaddr type for divert, e.g.
270 * sockaddr_div with multiple fields like
271 * sockaddr_dl. Presently we have only 7 bytes
272 * but that will do for now as most interfaces
273 * are 4 or less + 2 or less bytes for unit.
274 * There is probably a faster way of doing this,
275 * possibly taking it from the sockaddr_dl on the iface.
276 * This solves the problem of a P2P link and a LAN interface
277 * having the same address, which can result in the wrong
278 * interface being assigned to the packet when fed back
279 * into the divert socket. Theoretically if the daemon saves
280 * and re-uses the sockaddr_in as suggested in the man pages,
281 * this iface name will come along for the ride.
282 * (see div_output for the other half of this.)
284 ksnprintf(divsrc.sin_zero, sizeof divsrc.sin_zero,
285 m->m_pkthdr.rcvif->if_xname);
288 /* Put packet on socket queue, if any */
290 nport = htons((u_int16_t)port);
291 LIST_FOREACH(inp, &divcbinfo.pcblisthead, inp_list) {
292 if (inp->inp_flags & INP_PLACEMARKER)
294 if (inp->inp_lport == nport)
295 sa = inp->inp_socket;
298 if (ssb_appendaddr(&sa->so_rcv, (struct sockaddr *)&divsrc, m,
299 (struct mbuf *)NULL) == 0)
305 ipstat.ips_noproto++;
306 ipstat.ips_delivered--;
312 div_packet_handler(struct netmsg *nmsg)
314 struct netmsg_packet *nmp;
315 struct lwkt_msg *msg;
317 int port, incoming = 0;
319 nmp = (struct netmsg_packet *)nmsg;
322 msg = &nmsg->nm_lmsg;
323 port = msg->u.ms_result32 & 0xffff;
324 if (msg->u.ms_result32 & DIV_INPUT)
327 div_packet(m, incoming, port);
332 divert_packet(struct mbuf *m, int incoming)
335 struct divert_info *divinfo;
341 if (m->m_len < sizeof(struct ip) &&
342 (m = m_pullup(m, sizeof(struct ip))) == NULL)
345 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
346 KASSERT(mtag != NULL, ("%s no divert tag!", __func__));
347 divinfo = m_tag_data(mtag);
349 port = divinfo->port;
350 KASSERT(port != 0, ("%s: port=0", __func__));
354 struct netmsg_packet *nmp;
355 struct lwkt_msg *msg;
357 nmp = &m->m_hdr.mh_netmsg;
358 netmsg_init(&nmp->nm_netmsg, &netisr_apanic_rport, 0,
362 msg = &nmp->nm_netmsg.nm_lmsg;
363 msg->u.ms_result32 = port; /* port is 16bits */
365 msg->u.ms_result32 |= DIV_INPUT;
367 msg->u.ms_result32 |= DIV_OUTPUT;
369 lwkt_sendmsg(cpu_portfn(0), &nmp->nm_netmsg.nm_lmsg);
372 div_packet(m, incoming, port);
376 * Deliver packet back into the IP processing machinery.
378 * If no address specified, or address is 0.0.0.0, send to ip_output();
379 * otherwise, send to ip_input() and mark as having been received on
380 * the interface with that address.
383 div_output(struct socket *so, struct mbuf *m,
384 struct sockaddr_in *sin, struct mbuf *control)
388 struct divert_info *divinfo;
391 m_freem(control); /* XXX */
394 * Prepare the tag for divert info. Note that a packet
395 * with a 0 tag in mh_data is effectively untagged,
396 * so we could optimize that case.
398 mtag = m_tag_get(PACKET_TAG_IPFW_DIVERT, sizeof(*divinfo), MB_DONTWAIT);
403 m_tag_prepend(m, mtag);
405 /* Loopback avoidance and state recovery */
406 divinfo = m_tag_data(mtag);
408 divinfo->skipto = sin->sin_port;
412 /* Reinject packet into the system as incoming or outgoing */
413 if (DIV_IS_OUTPUT(sin)) {
414 struct ip *const ip = mtod(m, struct ip *);
416 /* Don't allow packet length sizes that will crash */
417 if ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len) {
422 /* Convert fields to host order for ip_output() */
423 ip->ip_len = ntohs(ip->ip_len);
424 ip->ip_off = ntohs(ip->ip_off);
426 /* Send packet to output processing */
427 ipstat.ips_rawout++; /* XXX */
428 error = ip_output(m, NULL, NULL,
429 (so->so_options & SO_DONTROUTE) |
430 IP_ALLOWBROADCAST | IP_RAWOUTPUT,
443 div_attach(struct socket *so, int proto, struct pru_attach_info *ai)
451 if ((error = suser_cred(ai->p_ucred, NULL_CRED_OKAY)) != 0)
454 error = soreserve(so, div_sendspace, div_recvspace, ai->sb_rlimit);
457 error = in_pcballoc(so, &divcbinfo);
460 inp = (struct inpcb *)so->so_pcb;
461 inp->inp_ip_p = proto;
462 inp->inp_vflag |= INP_IPV4;
463 inp->inp_flags |= INP_HDRINCL;
465 * The socket is always "connected" because
466 * we always know "where" to send the packet.
468 so->so_state |= SS_ISCONNECTED;
473 div_detach(struct socket *so)
485 div_abort(struct socket *so)
487 soisdisconnected(so);
488 return div_detach(so);
492 div_disconnect(struct socket *so)
494 if (!(so->so_state & SS_ISCONNECTED))
496 return div_abort(so);
500 div_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
505 * in_pcbbind assumes that nam is a sockaddr_in
506 * and in_pcbbind requires a valid address. Since divert
507 * sockets don't we need to make sure the address is
508 * filled in properly.
509 * XXX -- divert should not be abusing in_pcbind
510 * and should probably have its own family.
512 if (nam->sa_family != AF_INET) {
513 error = EAFNOSUPPORT;
515 ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY;
516 error = in_pcbbind(so->so_pcb, nam, td);
522 div_shutdown(struct socket *so)
529 div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
530 struct mbuf *control, struct thread *td)
532 /* Length check already done in ip_mport() */
533 KASSERT(m->m_len >= sizeof(struct ip), ("IP header not in one mbuf"));
536 return div_output(so, m, (struct sockaddr_in *)nam, control);
539 SYSCTL_DECL(_net_inet_divert);
540 SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLFLAG_RD, &divcbinfo, 0,
541 in_pcblist_global, "S,xinpcb", "List of active divert sockets");
543 struct pr_usrreqs div_usrreqs = {
544 .pru_abort = div_abort,
545 .pru_accept = pru_accept_notsupp,
546 .pru_attach = div_attach,
547 .pru_bind = div_bind,
548 .pru_connect = pru_connect_notsupp,
549 .pru_connect2 = pru_connect2_notsupp,
550 .pru_control = in_control,
551 .pru_detach = div_detach,
552 .pru_disconnect = div_disconnect,
553 .pru_listen = pru_listen_notsupp,
554 .pru_peeraddr = in_setpeeraddr,
555 .pru_rcvd = pru_rcvd_notsupp,
556 .pru_rcvoob = pru_rcvoob_notsupp,
557 .pru_send = div_send,
558 .pru_sense = pru_sense_null,
559 .pru_shutdown = div_shutdown,
560 .pru_sockaddr = in_setsockaddr,
561 .pru_sosend = sosend,
562 .pru_soreceive = soreceive,