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 $
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/socketvar2.h>
54 #include <sys/sysctl.h>
55 #include <sys/systm.h>
58 #include <sys/in_cksum.h>
60 #include <sys/msgport.h>
63 #include <net/route.h>
65 #include <net/netmsg2.h>
66 #include <net/netisr2.h>
67 #include <sys/thread2.h>
68 #include <sys/mplock2.h>
70 #include <netinet/in.h>
71 #include <netinet/in_systm.h>
72 #include <netinet/ip.h>
73 #include <netinet/in_pcb.h>
74 #include <netinet/in_var.h>
75 #include <netinet/ip_var.h>
76 #include <netinet/ip_divert.h>
83 * Allocate enough space to hold a full IP packet
85 #define DIVSNDQ (65536 + 100)
86 #define DIVRCVQ (65536 + 100)
88 #define DIV_IS_OUTPUT(sin) ((sin) == NULL || (sin)->sin_addr.s_addr == 0)
90 #define DIV_OUTPUT 0x10000
91 #define DIV_INPUT 0x20000
94 * Divert sockets work in conjunction with ipfw, see the divert(4)
95 * manpage for features.
96 * Internally, packets selected by ipfw in ip_input() or ip_output(),
97 * and never diverted before, are passed to the input queue of the
98 * divert socket with a given 'divert_port' number (as specified in
99 * the matching ipfw rule), and they are tagged with a 16 bit cookie
100 * (representing the rule number of the matching ipfw rule), which
101 * is passed to process reading from the socket.
103 * Packets written to the divert socket are again tagged with a cookie
104 * (usually the same as above) and a destination address.
105 * If the destination address is INADDR_ANY then the packet is
106 * treated as outgoing and sent to ip_output(), otherwise it is
107 * treated as incoming and sent to ip_input().
108 * In both cases, the packet is tagged with the cookie.
110 * On reinjection, processing in ip_input() and ip_output()
111 * will be exactly the same as for the original packet, except that
112 * ipfw processing will start at the rule number after the one
113 * written in the cookie (so, tagging a packet with a cookie of 0
114 * will cause it to be effectively considered as a standard packet).
117 /* Internal variables */
118 static struct inpcbinfo divcbinfo;
120 static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */
121 static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */
123 static struct mbuf *ip_divert(struct mbuf *, int, int);
125 static struct lwkt_token div_token = LWKT_TOKEN_INITIALIZER(div_token);
128 * Initialize divert connection block queue.
133 in_pcbinfo_init(&divcbinfo);
135 * XXX We don't use the hash list for divert IP, but it's easier
136 * to allocate a one entry hash list than it is to check all
137 * over the place for hashbase == NULL.
139 divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask);
140 divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask);
141 divcbinfo.wildcardhashbase = hashinit(1, M_PCB,
142 &divcbinfo.wildcardhashmask);
143 divcbinfo.ipi_size = sizeof(struct inpcb);
144 ip_divert_p = ip_divert;
148 * IPPROTO_DIVERT is not a real IP protocol; don't allow any packets
149 * with that protocol number to enter the system from the outside.
152 div_input(struct mbuf **mp, int *offp, int proto)
154 struct mbuf *m = *mp;
156 ipstat.ips_noproto++;
158 return(IPPROTO_DONE);
162 * Divert a packet by passing it up to the divert socket at port 'port'.
164 * Setup generic address and protocol structures for div_input routine,
165 * then pass them along with mbuf chain.
168 div_packet(struct mbuf *m, int incoming, int port)
170 struct sockaddr_in divsrc = { sizeof divsrc, AF_INET };
174 struct divert_info *divinfo;
177 /* Locate the divert info */
178 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
179 divinfo = m_tag_data(mtag);
180 divsrc.sin_port = divinfo->skipto;
183 * Record receive interface address, if any.
184 * But only for incoming packets.
186 divsrc.sin_addr.s_addr = 0;
188 struct ifaddr_container *ifac;
190 /* Find IP address for receive interface */
191 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid],
193 struct ifaddr *ifa = ifac->ifa;
195 if (ifa->ifa_addr == NULL)
197 if (ifa->ifa_addr->sa_family != AF_INET)
200 ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr;
205 * Record the incoming interface name whenever we have one.
207 if (m->m_pkthdr.rcvif) {
209 * Hide the actual interface name in there in the
210 * sin_zero array. XXX This needs to be moved to a
211 * different sockaddr type for divert, e.g.
212 * sockaddr_div with multiple fields like
213 * sockaddr_dl. Presently we have only 7 bytes
214 * but that will do for now as most interfaces
215 * are 4 or less + 2 or less bytes for unit.
216 * There is probably a faster way of doing this,
217 * possibly taking it from the sockaddr_dl on the iface.
218 * This solves the problem of a P2P link and a LAN interface
219 * having the same address, which can result in the wrong
220 * interface being assigned to the packet when fed back
221 * into the divert socket. Theoretically if the daemon saves
222 * and re-uses the sockaddr_in as suggested in the man pages,
223 * this iface name will come along for the ride.
224 * (see div_output for the other half of this.)
226 ksnprintf(divsrc.sin_zero, sizeof divsrc.sin_zero, "%s",
227 m->m_pkthdr.rcvif->if_xname);
230 /* Put packet on socket queue, if any */
232 nport = htons((u_int16_t)port);
236 * Following loop to locate the inpcb is MPSAFE since the inpcb
237 * insertion/removal happens on the same CPU (CPU0), however,
238 * saving/testing the socket pointer is not MPSAFE. So we still
239 * need to hold BGL here.
241 lwkt_gettoken(&div_token);
242 LIST_FOREACH(inp, &divcbinfo.pcblisthead, inp_list) {
243 if (inp->inp_flags & INP_PLACEMARKER)
245 if (inp->inp_lport == nport)
246 sa = inp->inp_socket;
249 lwkt_gettoken(&sa->so_rcv.ssb_token);
250 if (ssb_appendaddr(&sa->so_rcv, (struct sockaddr *)&divsrc, m, NULL) == 0)
254 lwkt_reltoken(&sa->so_rcv.ssb_token);
257 ipstat.ips_noproto++;
258 ipstat.ips_delivered--;
260 lwkt_reltoken(&div_token);
264 div_packet_handler(netmsg_t msg)
267 int port, incoming = 0;
269 m = msg->packet.nm_packet;
271 port = msg->lmsg.u.ms_result32 & 0xffff;
272 if (msg->lmsg.u.ms_result32 & DIV_INPUT)
274 div_packet(m, incoming, port);
275 /* no reply, msg embedded in mbuf */
279 divert_packet(struct mbuf *m, int incoming)
282 struct divert_info *divinfo;
288 if (m->m_len < sizeof(struct ip) &&
289 (m = m_pullup(m, sizeof(struct ip))) == NULL)
292 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
293 KASSERT(mtag != NULL, ("%s no divert tag!", __func__));
294 divinfo = m_tag_data(mtag);
296 port = divinfo->port;
297 KASSERT(port != 0, ("%s: port=0", __func__));
300 struct netmsg_packet *nmp;
302 nmp = &m->m_hdr.mh_netmsg;
303 netmsg_init(&nmp->base, NULL, &netisr_apanic_rport,
304 0, div_packet_handler);
307 nmp->base.lmsg.u.ms_result32 = port; /* port is 16bits */
309 nmp->base.lmsg.u.ms_result32 |= DIV_INPUT;
311 nmp->base.lmsg.u.ms_result32 |= DIV_OUTPUT;
313 lwkt_sendmsg(netisr_cpuport(0), &nmp->base.lmsg);
315 div_packet(m, incoming, port);
320 * Deliver packet back into the IP processing machinery.
322 * If no address specified, or address is 0.0.0.0, send to ip_output();
323 * otherwise, send to ip_input() and mark as having been received on
324 * the interface with that address.
327 div_output(struct socket *so, struct mbuf *m,
328 struct sockaddr_in *sin, struct mbuf *control)
332 struct divert_info *divinfo;
335 m_freem(control); /* XXX */
338 * Prepare the tag for divert info. Note that a packet
339 * with a 0 tag in mh_data is effectively untagged,
340 * so we could optimize that case.
342 mtag = m_tag_get(PACKET_TAG_IPFW_DIVERT, sizeof(*divinfo), MB_DONTWAIT);
347 m_tag_prepend(m, mtag);
349 /* Loopback avoidance and state recovery */
350 divinfo = m_tag_data(mtag);
352 divinfo->skipto = sin->sin_port;
356 /* Reinject packet into the system as incoming or outgoing */
357 if (DIV_IS_OUTPUT(sin)) {
358 struct ip *const ip = mtod(m, struct ip *);
360 /* Don't allow packet length sizes that will crash */
361 if ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len) {
366 /* Convert fields to host order for ip_output() */
367 ip->ip_len = ntohs(ip->ip_len);
368 ip->ip_off = ntohs(ip->ip_off);
370 /* Send packet to output processing */
371 ipstat.ips_rawout++; /* XXX */
372 error = ip_output(m, NULL, NULL,
373 (so->so_options & SO_DONTROUTE) |
374 IP_ALLOWBROADCAST | IP_RAWOUTPUT,
387 div_attach(netmsg_t msg)
389 struct socket *so = msg->attach.base.nm_so;
390 int proto = msg->attach.nm_proto;
391 struct pru_attach_info *ai = msg->attach.nm_ai;
398 error = priv_check_cred(ai->p_ucred, PRIV_ROOT, NULL_CRED_OKAY);
402 error = soreserve(so, div_sendspace, div_recvspace, ai->sb_rlimit);
405 lwkt_gettoken(&div_token);
406 sosetport(so, netisr_cpuport(0));
407 error = in_pcballoc(so, &divcbinfo);
409 lwkt_reltoken(&div_token);
412 inp = (struct inpcb *)so->so_pcb;
413 inp->inp_ip_p = proto;
414 inp->inp_vflag |= INP_IPV4;
415 inp->inp_flags |= INP_HDRINCL;
417 * The socket is always "connected" because
418 * we always know "where" to send the packet.
420 sosetstate(so, SS_ISCONNECTED);
421 lwkt_reltoken(&div_token);
424 lwkt_replymsg(&msg->attach.base.lmsg, error);
428 div_detach(netmsg_t msg)
430 struct socket *so = msg->detach.base.nm_so;
437 lwkt_replymsg(&msg->detach.base.lmsg, 0);
441 * NOTE: (so) is referenced from soabort*() and netmsg_pru_abort()
442 * will sofree() it when we return.
445 div_abort(netmsg_t msg)
447 struct socket *so = msg->abort.base.nm_so;
449 soisdisconnected(so);
451 /* msg invalid now */
455 div_disconnect(netmsg_t msg)
457 struct socket *so = msg->disconnect.base.nm_so;
460 if (so->so_state & SS_ISCONNECTED) {
463 /* msg invalid now */
468 lwkt_replymsg(&msg->disconnect.base.lmsg, error);
472 div_bind(netmsg_t msg)
474 struct socket *so = msg->bind.base.nm_so;
475 struct sockaddr *nam = msg->bind.nm_nam;
479 * in_pcbbind assumes that nam is a sockaddr_in
480 * and in_pcbbind requires a valid address. Since divert
481 * sockets don't we need to make sure the address is
482 * filled in properly.
483 * XXX -- divert should not be abusing in_pcbind
484 * and should probably have its own family.
486 if (nam->sa_family != AF_INET) {
487 error = EAFNOSUPPORT;
489 ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY;
490 error = in_pcbbind(so->so_pcb, nam, msg->bind.nm_td);
492 lwkt_replymsg(&msg->bind.base.lmsg, error);
496 div_shutdown(netmsg_t msg)
498 struct socket *so = msg->shutdown.base.nm_so;
502 lwkt_replymsg(&msg->shutdown.base.lmsg, 0);
506 div_send(netmsg_t msg)
508 struct socket *so = msg->send.base.nm_so;
509 struct mbuf *m = msg->send.nm_m;
510 struct sockaddr *nam = msg->send.nm_addr;
511 struct mbuf *control = msg->send.nm_control;
514 /* Length check already done in ip_hashfn() */
515 KASSERT(m->m_len >= sizeof(struct ip), ("IP header not in one mbuf"));
518 error = div_output(so, m, (struct sockaddr_in *)nam, control);
519 lwkt_replymsg(&msg->send.base.lmsg, error);
522 SYSCTL_DECL(_net_inet_divert);
523 SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLFLAG_RD, &divcbinfo, 0,
524 in_pcblist_global, "S,xinpcb", "List of active divert sockets");
526 struct pr_usrreqs div_usrreqs = {
527 .pru_abort = div_abort,
528 .pru_accept = pr_generic_notsupp,
529 .pru_attach = div_attach,
530 .pru_bind = div_bind,
531 .pru_connect = pr_generic_notsupp,
532 .pru_connect2 = pr_generic_notsupp,
533 .pru_control = in_control_dispatch,
534 .pru_detach = div_detach,
535 .pru_disconnect = div_disconnect,
536 .pru_listen = pr_generic_notsupp,
537 .pru_peeraddr = in_setpeeraddr_dispatch,
538 .pru_rcvd = pr_generic_notsupp,
539 .pru_rcvoob = pr_generic_notsupp,
540 .pru_send = div_send,
541 .pru_sense = pru_sense_null,
542 .pru_shutdown = div_shutdown,
543 .pru_sockaddr = in_setsockaddr_dispatch,
544 .pru_sosend = sosend,
545 .pru_soreceive = soreceive
549 ip_divert_out(struct mbuf *m, int tee)
551 struct mbuf *clone = NULL;
552 struct ip *ip = mtod(m, struct ip *);
554 /* Clone packet if we're doing a 'tee' */
556 clone = m_dup(m, MB_DONTWAIT);
560 * delayed checksums are not currently compatible
561 * with divert sockets.
563 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
565 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
568 /* Restore packet header fields to original values */
569 ip->ip_len = htons(ip->ip_len);
570 ip->ip_off = htons(ip->ip_off);
572 /* Deliver packet to divert input routine */
575 /* If 'tee', continue with original packet */
580 ip_divert_in(struct mbuf *m, int tee)
582 struct mbuf *clone = NULL;
583 struct ip *ip = mtod(m, struct ip *);
586 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
587 const struct divert_info *divinfo;
592 * Only trust divert info in the fragment
595 frag_off = ip->ip_off << 3;
597 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
598 m_tag_delete(m, mtag);
602 * Attempt reassembly; if it succeeds, proceed.
603 * ip_reass() will return a different mbuf.
608 ip = mtod(m, struct ip *);
610 /* Caller need to redispatch the packet, if it is for us */
611 m->m_pkthdr.fw_flags |= FW_MBUF_REDISPATCH;
614 * Get the header length of the reassembled
617 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
620 * Restore original checksum before diverting
624 ip->ip_len = htons(ip->ip_len);
625 ip->ip_off = htons(ip->ip_off);
627 if (hlen == sizeof(struct ip))
628 ip->ip_sum = in_cksum_hdr(ip);
630 ip->ip_sum = in_cksum(m, hlen);
631 ip->ip_off = ntohs(ip->ip_off);
632 ip->ip_len = ntohs(ip->ip_len);
635 * Only use the saved divert info
637 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
639 /* Wrongly configured ipfw */
640 kprintf("ip_input no divert info\n");
644 divinfo = m_tag_data(mtag);
649 * Divert or tee packet to the divert protocol if
653 /* Clone packet if we're doing a 'tee' */
655 clone = m_dup(m, MB_DONTWAIT);
658 * Restore packet header fields to original
661 ip->ip_len = htons(ip->ip_len);
662 ip->ip_off = htons(ip->ip_off);
664 /* Deliver packet to divert input routine */
667 /* Catch invalid reference */
671 ipstat.ips_delivered++;
673 /* If 'tee', continue with original packet */
676 * Complete processing of the packet.
677 * XXX Better safe than sorry, remove the DIVERT tag.
679 mtag = m_tag_find(clone, PACKET_TAG_IPFW_DIVERT, NULL);
680 KKASSERT(mtag != NULL);
681 m_tag_delete(clone, mtag);
687 ip_divert(struct mbuf *m, int tee, int incoming)
692 ret = ip_divert_in(m, tee);
694 ret = ip_divert_out(m, tee);