Fix LINT build.
[dragonfly.git] / sys / netinet / ip_divert.c
CommitLineData
984263bc
MD
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: src/sys/netinet/ip_divert.c,v 1.42.2.6 2003/01/23 21:06:45 sam Exp $
87b66be9 34 * $DragonFly: src/sys/netinet/ip_divert.c,v 1.40 2008/10/21 13:51:01 sephe Exp $
984263bc
MD
35 */
36
0030d31a
SZ
37#define _IP_VHL
38
984263bc
MD
39#include "opt_inet.h"
40#include "opt_ipfw.h"
41#include "opt_ipdivert.h"
42#include "opt_ipsec.h"
43
44#ifndef INET
45#error "IPDIVERT requires INET."
46#endif
47
48#include <sys/param.h>
49#include <sys/kernel.h>
50#include <sys/malloc.h>
51#include <sys/mbuf.h>
52#include <sys/socket.h>
53#include <sys/protosw.h>
54#include <sys/socketvar.h>
93bffeca 55#include <sys/socketvar2.h>
984263bc
MD
56#include <sys/sysctl.h>
57#include <sys/systm.h>
58#include <sys/proc.h>
895c1f85 59#include <sys/priv.h>
0030d31a 60#include <sys/in_cksum.h>
87b66be9 61#include <sys/lock.h>
6f3f81a8
SZ
62#ifdef SMP
63#include <sys/msgport.h>
64#endif
984263bc 65
984263bc
MD
66#include <net/if.h>
67#include <net/route.h>
684a93c4 68
6f3f81a8
SZ
69#ifdef SMP
70#include <net/netmsg2.h>
71#endif
684a93c4
MD
72#include <sys/thread2.h>
73#include <sys/mplock2.h>
984263bc
MD
74
75#include <netinet/in.h>
76#include <netinet/in_systm.h>
77#include <netinet/ip.h>
78#include <netinet/in_pcb.h>
79#include <netinet/in_var.h>
80#include <netinet/ip_var.h>
68edaf54 81#include <netinet/ip_divert.h>
984263bc
MD
82
83/*
84 * Divert sockets
85 */
86
87/*
88 * Allocate enough space to hold a full IP packet
89 */
90#define DIVSNDQ (65536 + 100)
91#define DIVRCVQ (65536 + 100)
92
934c6849
SZ
93#define DIV_IS_OUTPUT(sin) ((sin) == NULL || (sin)->sin_addr.s_addr == 0)
94
6f3f81a8
SZ
95#define DIV_OUTPUT 0x10000
96#define DIV_INPUT 0x20000
97
984263bc
MD
98/*
99 * Divert sockets work in conjunction with ipfw, see the divert(4)
100 * manpage for features.
101 * Internally, packets selected by ipfw in ip_input() or ip_output(),
102 * and never diverted before, are passed to the input queue of the
103 * divert socket with a given 'divert_port' number (as specified in
104 * the matching ipfw rule), and they are tagged with a 16 bit cookie
105 * (representing the rule number of the matching ipfw rule), which
106 * is passed to process reading from the socket.
107 *
108 * Packets written to the divert socket are again tagged with a cookie
109 * (usually the same as above) and a destination address.
110 * If the destination address is INADDR_ANY then the packet is
111 * treated as outgoing and sent to ip_output(), otherwise it is
112 * treated as incoming and sent to ip_input().
113 * In both cases, the packet is tagged with the cookie.
114 *
115 * On reinjection, processing in ip_input() and ip_output()
116 * will be exactly the same as for the original packet, except that
117 * ipfw processing will start at the rule number after the one
118 * written in the cookie (so, tagging a packet with a cookie of 0
119 * will cause it to be effectively considered as a standard packet).
120 */
121
122/* Internal variables */
984263bc
MD
123static struct inpcbinfo divcbinfo;
124
125static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */
126static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */
127
0030d31a
SZ
128static struct mbuf *ip_divert(struct mbuf *, int, int);
129
6cef7136
MD
130static struct lwkt_token div_token = LWKT_TOKEN_MP_INITIALIZER(div_token);
131
984263bc
MD
132/*
133 * Initialize divert connection block queue.
134 */
135void
136div_init(void)
137{
d2e9e54c 138 in_pcbinfo_init(&divcbinfo);
984263bc
MD
139 /*
140 * XXX We don't use the hash list for divert IP, but it's easier
141 * to allocate a one entry hash list than it is to check all
142 * over the place for hashbase == NULL.
143 */
144 divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask);
145 divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask);
2322f6ca
JH
146 divcbinfo.wildcardhashbase = hashinit(1, M_PCB,
147 &divcbinfo.wildcardhashmask);
9f42c129 148 divcbinfo.ipi_size = sizeof(struct inpcb);
0030d31a 149 ip_divert_p = ip_divert;
984263bc
MD
150}
151
152/*
153 * IPPROTO_DIVERT is not a real IP protocol; don't allow any packets
154 * with that protocol number to enter the system from the outside.
155 */
156void
a00138cb 157div_input(struct mbuf *m, ...)
984263bc
MD
158{
159 ipstat.ips_noproto++;
160 m_freem(m);
161}
162
934c6849 163struct lwkt_port *
48e7b118 164div_soport(struct socket *so, struct sockaddr *nam, struct mbuf **mptr)
934c6849
SZ
165{
166 struct sockaddr_in *sin;
167 struct mbuf *m;
168 int dir;
169
934c6849
SZ
170 sin = (struct sockaddr_in *)nam;
171 m = *mptr;
172 M_ASSERTPKTHDR(m);
173
174 m->m_pkthdr.rcvif = NULL;
175 dir = DIV_IS_OUTPUT(sin) ? IP_MPORT_OUT : IP_MPORT_IN;
176
177 if (sin != NULL) {
178 int i;
179
180 /*
181 * Try locating the interface, if we originally had one.
182 * This is done even for outgoing packets, since for a
183 * forwarded packet, there must be an interface attached.
184 *
185 * Find receive interface with the given name, stuffed
186 * (if it exists) in the sin_zero[] field.
187 * The name is user supplied data so don't trust its size
188 * or that it is zero terminated.
189 */
190 for (i = 0; sin->sin_zero[i] && i < sizeof(sin->sin_zero); i++)
191 ;
192 if (i > 0 && i < sizeof(sin->sin_zero))
193 m->m_pkthdr.rcvif = ifunit(sin->sin_zero);
194 }
195
196 if (dir == IP_MPORT_IN && m->m_pkthdr.rcvif == NULL) {
197 /*
198 * No luck with the name, check by IP address.
199 * Clear the port and the ifname to make sure
200 * there are no distractions for ifa_ifwithaddr.
201 *
202 * Be careful not to trash sin->sin_port; it will
203 * be used later in div_output().
204 */
205 struct ifaddr *ifa;
206 u_short sin_port;
207
208 bzero(sin->sin_zero, sizeof(sin->sin_zero));
209 sin_port = sin->sin_port; /* save */
210 sin->sin_port = 0;
211 ifa = ifa_ifwithaddr((struct sockaddr *)sin);
212 if (ifa == NULL) {
213 m_freem(m);
214 *mptr = NULL;
215 return NULL;
216 }
217 sin->sin_port = sin_port; /* restore */
218 m->m_pkthdr.rcvif = ifa->ifa_ifp;
219 }
220
c3c96e44
MD
221 /*
222 * Recalculate the protocol thread.
223 */
224 ip_cpufn(mptr, 0, dir);
225 m = *mptr;
226 if (m) {
93bffeca 227 KKASSERT(m->m_flags & M_HASH);
c3c96e44
MD
228 return(cpu_portfn(m->m_pkthdr.hash));
229 }
230 return(NULL);
934c6849
SZ
231}
232
984263bc
MD
233/*
234 * Divert a packet by passing it up to the divert socket at port 'port'.
235 *
236 * Setup generic address and protocol structures for div_input routine,
237 * then pass them along with mbuf chain.
238 */
6f3f81a8
SZ
239static void
240div_packet(struct mbuf *m, int incoming, int port)
984263bc 241{
668488bd 242 struct sockaddr_in divsrc = { sizeof divsrc, AF_INET };
984263bc
MD
243 struct inpcb *inp;
244 struct socket *sa;
e5ecc832 245 struct m_tag *mtag;
68edaf54 246 struct divert_info *divinfo;
984263bc
MD
247 u_int16_t nport;
248
68edaf54 249 /* Locate the divert info */
077c74e2 250 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
68edaf54
SZ
251 divinfo = m_tag_data(mtag);
252 divsrc.sin_port = divinfo->skipto;
077c74e2 253
984263bc
MD
254 /*
255 * Record receive interface address, if any.
256 * But only for incoming packets.
257 */
258 divsrc.sin_addr.s_addr = 0;
259 if (incoming) {
b2632176 260 struct ifaddr_container *ifac;
984263bc 261
984263bc 262 /* Find IP address for receive interface */
b2632176
SZ
263 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid],
264 ifa_link) {
265 struct ifaddr *ifa = ifac->ifa;
266
984263bc
MD
267 if (ifa->ifa_addr == NULL)
268 continue;
269 if (ifa->ifa_addr->sa_family != AF_INET)
270 continue;
271 divsrc.sin_addr =
272 ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr;
273 break;
274 }
275 }
276 /*
277 * Record the incoming interface name whenever we have one.
278 */
984263bc
MD
279 if (m->m_pkthdr.rcvif) {
280 /*
f23061d4 281 * Hide the actual interface name in there in the
984263bc
MD
282 * sin_zero array. XXX This needs to be moved to a
283 * different sockaddr type for divert, e.g.
f23061d4 284 * sockaddr_div with multiple fields like
984263bc
MD
285 * sockaddr_dl. Presently we have only 7 bytes
286 * but that will do for now as most interfaces
287 * are 4 or less + 2 or less bytes for unit.
288 * There is probably a faster way of doing this,
289 * possibly taking it from the sockaddr_dl on the iface.
290 * This solves the problem of a P2P link and a LAN interface
291 * having the same address, which can result in the wrong
292 * interface being assigned to the packet when fed back
293 * into the divert socket. Theoretically if the daemon saves
294 * and re-uses the sockaddr_in as suggested in the man pages,
295 * this iface name will come along for the ride.
296 * (see div_output for the other half of this.)
f23061d4 297 */
f8c7a42d 298 ksnprintf(divsrc.sin_zero, sizeof divsrc.sin_zero,
87b66be9 299 m->m_pkthdr.rcvif->if_xname);
984263bc
MD
300 }
301
302 /* Put packet on socket queue, if any */
303 sa = NULL;
304 nport = htons((u_int16_t)port);
87b66be9
SZ
305
306 /*
307 * XXX
308 * Following loop to locate the inpcb is MPSAFE since the inpcb
309 * insertion/removal happens on the same CPU (CPU0), however,
310 * saving/testing the socket pointer is not MPSAFE. So we still
311 * need to hold BGL here.
312 */
6cef7136 313 lwkt_gettoken(&div_token);
d2e9e54c
MD
314 LIST_FOREACH(inp, &divcbinfo.pcblisthead, inp_list) {
315 if (inp->inp_flags & INP_PLACEMARKER)
316 continue;
984263bc
MD
317 if (inp->inp_lport == nport)
318 sa = inp->inp_socket;
319 }
320 if (sa) {
6cef7136
MD
321 lwkt_gettoken(&sa->so_rcv.ssb_token);
322 if (ssb_appendaddr(&sa->so_rcv, (struct sockaddr *)&divsrc, m, NULL) == 0)
984263bc
MD
323 m_freem(m);
324 else
325 sorwakeup(sa);
6cef7136 326 lwkt_reltoken(&sa->so_rcv.ssb_token);
984263bc
MD
327 } else {
328 m_freem(m);
329 ipstat.ips_noproto++;
330 ipstat.ips_delivered--;
f23061d4 331 }
6cef7136 332 lwkt_reltoken(&div_token);
984263bc
MD
333}
334
6f3f81a8
SZ
335#ifdef SMP
336static void
337div_packet_handler(struct netmsg *nmsg)
338{
339 struct netmsg_packet *nmp;
340 struct lwkt_msg *msg;
341 struct mbuf *m;
342 int port, incoming = 0;
343
344 nmp = (struct netmsg_packet *)nmsg;
345 m = nmp->nm_packet;
346
347 msg = &nmsg->nm_lmsg;
348 port = msg->u.ms_result32 & 0xffff;
349 if (msg->u.ms_result32 & DIV_INPUT)
350 incoming = 1;
351
352 div_packet(m, incoming, port);
353}
354#endif /* SMP */
355
0030d31a 356static void
b0cb7c2c 357divert_packet(struct mbuf *m, int incoming)
6f3f81a8 358{
b0cb7c2c
SZ
359 struct m_tag *mtag;
360 struct divert_info *divinfo;
361 int port;
362
6f3f81a8
SZ
363 M_ASSERTPKTHDR(m);
364
365 /* Assure header */
366 if (m->m_len < sizeof(struct ip) &&
367 (m = m_pullup(m, sizeof(struct ip))) == NULL)
368 return;
369
b0cb7c2c
SZ
370 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
371 KASSERT(mtag != NULL, ("%s no divert tag!", __func__));
372 divinfo = m_tag_data(mtag);
373
374 port = divinfo->port;
375 KASSERT(port != 0, ("%s: port=0", __func__));
6f3f81a8
SZ
376
377#ifdef SMP
378 if (mycpuid != 0) {
379 struct netmsg_packet *nmp;
380 struct lwkt_msg *msg;
381
382 nmp = &m->m_hdr.mh_netmsg;
91887230 383 netmsg_init(&nmp->nm_netmsg, NULL, &netisr_apanic_rport,
c3c96e44 384 0, div_packet_handler);
6f3f81a8
SZ
385 nmp->nm_packet = m;
386
387 msg = &nmp->nm_netmsg.nm_lmsg;
388 msg->u.ms_result32 = port; /* port is 16bits */
389 if (incoming)
390 msg->u.ms_result32 |= DIV_INPUT;
391 else
392 msg->u.ms_result32 |= DIV_OUTPUT;
393
394 lwkt_sendmsg(cpu_portfn(0), &nmp->nm_netmsg.nm_lmsg);
395 } else
396#endif
397 div_packet(m, incoming, port);
398}
399
984263bc
MD
400/*
401 * Deliver packet back into the IP processing machinery.
402 *
403 * If no address specified, or address is 0.0.0.0, send to ip_output();
404 * otherwise, send to ip_input() and mark as having been received on
405 * the interface with that address.
406 */
407static int
408div_output(struct socket *so, struct mbuf *m,
409 struct sockaddr_in *sin, struct mbuf *control)
410{
411 int error = 0;
e5ecc832 412 struct m_tag *mtag;
68edaf54 413 struct divert_info *divinfo;
984263bc 414
934c6849
SZ
415 if (control)
416 m_freem(control); /* XXX */
417
984263bc
MD
418 /*
419 * Prepare the tag for divert info. Note that a packet
420 * with a 0 tag in mh_data is effectively untagged,
421 * so we could optimize that case.
422 */
68edaf54 423 mtag = m_tag_get(PACKET_TAG_IPFW_DIVERT, sizeof(*divinfo), MB_DONTWAIT);
e5ecc832
JS
424 if (mtag == NULL) {
425 error = ENOBUFS;
426 goto cantsend;
427 }
428 m_tag_prepend(m, mtag);
984263bc
MD
429
430 /* Loopback avoidance and state recovery */
68edaf54 431 divinfo = m_tag_data(mtag);
934c6849 432 if (sin)
68edaf54 433 divinfo->skipto = sin->sin_port;
934c6849 434 else
68edaf54 435 divinfo->skipto = 0;
984263bc
MD
436
437 /* Reinject packet into the system as incoming or outgoing */
934c6849 438 if (DIV_IS_OUTPUT(sin)) {
984263bc
MD
439 struct ip *const ip = mtod(m, struct ip *);
440
b171561b
SZ
441 /* Don't allow packet length sizes that will crash */
442 if ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len) {
984263bc
MD
443 error = EINVAL;
444 goto cantsend;
445 }
446
447 /* Convert fields to host order for ip_output() */
448 ip->ip_len = ntohs(ip->ip_len);
449 ip->ip_off = ntohs(ip->ip_off);
450
451 /* Send packet to output processing */
452 ipstat.ips_rawout++; /* XXX */
b171561b 453 error = ip_output(m, NULL, NULL,
984263bc
MD
454 (so->so_options & SO_DONTROUTE) |
455 IP_ALLOWBROADCAST | IP_RAWOUTPUT,
b171561b 456 NULL, NULL);
984263bc 457 } else {
e5ecc832 458 ip_input(m);
984263bc 459 }
984263bc
MD
460 return error;
461
462cantsend:
463 m_freem(m);
464 return error;
465}
466
467static int
e4700d00 468div_attach(struct socket *so, int proto, struct pru_attach_info *ai)
984263bc
MD
469{
470 struct inpcb *inp;
1cae611f 471 int error;
984263bc 472
ed894f8c 473 inp = so->so_pcb;
984263bc
MD
474 if (inp)
475 panic("div_attach");
895c1f85 476 if ((error = priv_check_cred(ai->p_ucred, PRIV_ROOT, NULL_CRED_OKAY)) != 0)
984263bc
MD
477 return error;
478
e4700d00 479 error = soreserve(so, div_sendspace, div_recvspace, ai->sb_rlimit);
984263bc
MD
480 if (error)
481 return error;
6cef7136 482 lwkt_gettoken(&div_token);
e4700d00 483 error = in_pcballoc(so, &divcbinfo);
6cef7136
MD
484 if (error) {
485 lwkt_reltoken(&div_token);
984263bc 486 return error;
6cef7136 487 }
984263bc
MD
488 inp = (struct inpcb *)so->so_pcb;
489 inp->inp_ip_p = proto;
490 inp->inp_vflag |= INP_IPV4;
491 inp->inp_flags |= INP_HDRINCL;
5fe66e68
JH
492 /*
493 * The socket is always "connected" because
494 * we always know "where" to send the packet.
495 */
48e7b118 496 so->so_port = cpu0_soport(so, NULL, NULL);
6cef7136
MD
497 sosetstate(so, SS_ISCONNECTED);
498 lwkt_reltoken(&div_token);
984263bc
MD
499 return 0;
500}
501
502static int
503div_detach(struct socket *so)
504{
505 struct inpcb *inp;
506
ed894f8c 507 inp = so->so_pcb;
5fe66e68 508 if (inp == NULL)
984263bc
MD
509 panic("div_detach");
510 in_pcbdetach(inp);
511 return 0;
512}
513
6cef7136
MD
514/*
515 * NOTE: (so) is referenced from soabort*() and netmsg_pru_abort()
516 * will sofree() it when we return.
517 */
984263bc
MD
518static int
519div_abort(struct socket *so)
520{
6cef7136
MD
521 int error;
522
984263bc 523 soisdisconnected(so);
6cef7136
MD
524 error = div_detach(so);
525
526 return error;
984263bc
MD
527}
528
529static int
530div_disconnect(struct socket *so)
531{
6cef7136
MD
532 int error;
533
5fe66e68 534 if (!(so->so_state & SS_ISCONNECTED))
984263bc 535 return ENOTCONN;
6cef7136
MD
536 soreference(so);
537 error = div_abort(so);
538 sofree(so);
539
540 return error;
984263bc
MD
541}
542
543static int
dadab5e9 544div_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
984263bc 545{
984263bc
MD
546 int error;
547
65b1ea7e
SZ
548 /*
549 * in_pcbbind assumes that nam is a sockaddr_in
984263bc
MD
550 * and in_pcbbind requires a valid address. Since divert
551 * sockets don't we need to make sure the address is
552 * filled in properly.
553 * XXX -- divert should not be abusing in_pcbind
554 * and should probably have its own family.
555 */
65b1ea7e 556 if (nam->sa_family != AF_INET) {
984263bc 557 error = EAFNOSUPPORT;
65b1ea7e 558 } else {
984263bc 559 ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY;
65b1ea7e 560 error = in_pcbbind(so->so_pcb, nam, td);
984263bc 561 }
984263bc
MD
562 return error;
563}
564
565static int
566div_shutdown(struct socket *so)
567{
568 socantsendmore(so);
569 return 0;
570}
571
572static int
573div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
dadab5e9 574 struct mbuf *control, struct thread *td)
984263bc 575{
c3c96e44 576 /* Length check already done in ip_cpufn() */
934c6849 577 KASSERT(m->m_len >= sizeof(struct ip), ("IP header not in one mbuf"));
984263bc
MD
578
579 /* Send packet */
580 return div_output(so, m, (struct sockaddr_in *)nam, control);
581}
582
984263bc 583SYSCTL_DECL(_net_inet_divert);
48a2763f
JS
584SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLFLAG_RD, &divcbinfo, 0,
585 in_pcblist_global, "S,xinpcb", "List of active divert sockets");
984263bc
MD
586
587struct pr_usrreqs div_usrreqs = {
fa5e758c
MD
588 .pru_abort = div_abort,
589 .pru_accept = pru_accept_notsupp,
590 .pru_attach = div_attach,
591 .pru_bind = div_bind,
592 .pru_connect = pru_connect_notsupp,
593 .pru_connect2 = pru_connect2_notsupp,
594 .pru_control = in_control,
595 .pru_detach = div_detach,
596 .pru_disconnect = div_disconnect,
597 .pru_listen = pru_listen_notsupp,
598 .pru_peeraddr = in_setpeeraddr,
599 .pru_rcvd = pru_rcvd_notsupp,
600 .pru_rcvoob = pru_rcvoob_notsupp,
601 .pru_send = div_send,
602 .pru_sense = pru_sense_null,
603 .pru_shutdown = div_shutdown,
604 .pru_sockaddr = in_setsockaddr,
605 .pru_sosend = sosend,
8b5c39bb 606 .pru_soreceive = soreceive
984263bc 607};
0030d31a
SZ
608
609static struct mbuf *
610ip_divert_out(struct mbuf *m, int tee)
611{
612 struct mbuf *clone = NULL;
613 struct ip *ip = mtod(m, struct ip *);
614
615 /* Clone packet if we're doing a 'tee' */
616 if (tee)
617 clone = m_dup(m, MB_DONTWAIT);
618
619 /*
620 * XXX
621 * delayed checksums are not currently compatible
622 * with divert sockets.
623 */
624 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
625 in_delayed_cksum(m);
626 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
627 }
628
629 /* Restore packet header fields to original values */
630 ip->ip_len = htons(ip->ip_len);
631 ip->ip_off = htons(ip->ip_off);
632
633 /* Deliver packet to divert input routine */
634 divert_packet(m, 0);
635
636 /* If 'tee', continue with original packet */
637 return clone;
638}
639
640static struct mbuf *
641ip_divert_in(struct mbuf *m, int tee)
642{
643 struct mbuf *clone = NULL;
644 struct ip *ip = mtod(m, struct ip *);
645 struct m_tag *mtag;
646
647 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
648 const struct divert_info *divinfo;
649 u_short frag_off;
650 int hlen;
651
652 /*
653 * Only trust divert info in the fragment
654 * at offset 0.
655 */
656 frag_off = ip->ip_off << 3;
657 if (frag_off != 0) {
658 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
659 m_tag_delete(m, mtag);
660 }
661
662 /*
663 * Attempt reassembly; if it succeeds, proceed.
664 * ip_reass() will return a different mbuf.
665 */
666 m = ip_reass(m);
667 if (m == NULL)
668 return NULL;
669 ip = mtod(m, struct ip *);
670
671 /* Caller need to redispatch the packet, if it is for us */
672 m->m_pkthdr.fw_flags |= FW_MBUF_REDISPATCH;
673
674 /*
675 * Get the header length of the reassembled
676 * packet
677 */
678 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
679
680 /*
681 * Restore original checksum before diverting
682 * packet
683 */
684 ip->ip_len += hlen;
685 ip->ip_len = htons(ip->ip_len);
686 ip->ip_off = htons(ip->ip_off);
687 ip->ip_sum = 0;
688 if (hlen == sizeof(struct ip))
689 ip->ip_sum = in_cksum_hdr(ip);
690 else
691 ip->ip_sum = in_cksum(m, hlen);
692 ip->ip_off = ntohs(ip->ip_off);
693 ip->ip_len = ntohs(ip->ip_len);
694
695 /*
696 * Only use the saved divert info
697 */
698 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
699 if (mtag == NULL) {
700 /* Wrongly configured ipfw */
701 kprintf("ip_input no divert info\n");
702 m_freem(m);
703 return NULL;
704 }
705 divinfo = m_tag_data(mtag);
706 tee = divinfo->tee;
707 }
708
709 /*
710 * Divert or tee packet to the divert protocol if
711 * required.
712 */
713
714 /* Clone packet if we're doing a 'tee' */
715 if (tee)
716 clone = m_dup(m, MB_DONTWAIT);
717
718 /*
719 * Restore packet header fields to original
720 * values
721 */
722 ip->ip_len = htons(ip->ip_len);
723 ip->ip_off = htons(ip->ip_off);
724
725 /* Deliver packet to divert input routine */
726 divert_packet(m, 1);
727
728 /* Catch invalid reference */
729 m = NULL;
730 ip = NULL;
731
732 ipstat.ips_delivered++;
733
734 /* If 'tee', continue with original packet */
735 if (clone != NULL) {
736 /*
737 * Complete processing of the packet.
738 * XXX Better safe than sorry, remove the DIVERT tag.
739 */
740 mtag = m_tag_find(clone, PACKET_TAG_IPFW_DIVERT, NULL);
741 KKASSERT(mtag != NULL);
742 m_tag_delete(clone, mtag);
743 }
744 return clone;
745}
746
747static struct mbuf *
748ip_divert(struct mbuf *m, int tee, int incoming)
749{
750 struct mbuf *ret;
751
752 if (incoming)
753 ret = ip_divert_in(m, tee);
754 else
755 ret = ip_divert_out(m, tee);
756 return ret;
757}