kernel: Make SMP support default (and non-optional).
[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 $
34 */
35
0030d31a
SZ
36#define _IP_VHL
37
984263bc 38#include "opt_inet.h"
984263bc
MD
39#include "opt_ipdivert.h"
40#include "opt_ipsec.h"
41
42#ifndef INET
43#error "IPDIVERT requires INET."
44#endif
45
46#include <sys/param.h>
47#include <sys/kernel.h>
48#include <sys/malloc.h>
49#include <sys/mbuf.h>
50#include <sys/socket.h>
51#include <sys/protosw.h>
52#include <sys/socketvar.h>
93bffeca 53#include <sys/socketvar2.h>
984263bc
MD
54#include <sys/sysctl.h>
55#include <sys/systm.h>
56#include <sys/proc.h>
895c1f85 57#include <sys/priv.h>
0030d31a 58#include <sys/in_cksum.h>
87b66be9 59#include <sys/lock.h>
6f3f81a8 60#include <sys/msgport.h>
984263bc 61
984263bc
MD
62#include <net/if.h>
63#include <net/route.h>
684a93c4 64
6f3f81a8 65#include <net/netmsg2.h>
684a93c4
MD
66#include <sys/thread2.h>
67#include <sys/mplock2.h>
984263bc
MD
68
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>
68edaf54 75#include <netinet/ip_divert.h>
984263bc
MD
76
77/*
78 * Divert sockets
79 */
80
81/*
82 * Allocate enough space to hold a full IP packet
83 */
84#define DIVSNDQ (65536 + 100)
85#define DIVRCVQ (65536 + 100)
86
934c6849
SZ
87#define DIV_IS_OUTPUT(sin) ((sin) == NULL || (sin)->sin_addr.s_addr == 0)
88
6f3f81a8
SZ
89#define DIV_OUTPUT 0x10000
90#define DIV_INPUT 0x20000
91
984263bc
MD
92/*
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.
101 *
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.
108 *
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).
114 */
115
116/* Internal variables */
984263bc
MD
117static struct inpcbinfo divcbinfo;
118
119static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */
120static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */
121
0030d31a
SZ
122static struct mbuf *ip_divert(struct mbuf *, int, int);
123
a3c18566 124static struct lwkt_token div_token = LWKT_TOKEN_INITIALIZER(div_token);
6cef7136 125
984263bc
MD
126/*
127 * Initialize divert connection block queue.
128 */
129void
130div_init(void)
131{
d2e9e54c 132 in_pcbinfo_init(&divcbinfo);
984263bc
MD
133 /*
134 * XXX We don't use the hash list for divert IP, but it's easier
135 * to allocate a one entry hash list than it is to check all
136 * over the place for hashbase == NULL.
137 */
138 divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask);
139 divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask);
2322f6ca
JH
140 divcbinfo.wildcardhashbase = hashinit(1, M_PCB,
141 &divcbinfo.wildcardhashmask);
9f42c129 142 divcbinfo.ipi_size = sizeof(struct inpcb);
0030d31a 143 ip_divert_p = ip_divert;
984263bc
MD
144}
145
146/*
147 * IPPROTO_DIVERT is not a real IP protocol; don't allow any packets
148 * with that protocol number to enter the system from the outside.
149 */
002c1265
MD
150int
151div_input(struct mbuf **mp, int *offp, int proto)
984263bc 152{
002c1265
MD
153 struct mbuf *m = *mp;
154
984263bc
MD
155 ipstat.ips_noproto++;
156 m_freem(m);
002c1265 157 return(IPPROTO_DONE);
984263bc
MD
158}
159
160/*
161 * Divert a packet by passing it up to the divert socket at port 'port'.
162 *
163 * Setup generic address and protocol structures for div_input routine,
164 * then pass them along with mbuf chain.
165 */
6f3f81a8
SZ
166static void
167div_packet(struct mbuf *m, int incoming, int port)
984263bc 168{
668488bd 169 struct sockaddr_in divsrc = { sizeof divsrc, AF_INET };
984263bc
MD
170 struct inpcb *inp;
171 struct socket *sa;
e5ecc832 172 struct m_tag *mtag;
68edaf54 173 struct divert_info *divinfo;
984263bc
MD
174 u_int16_t nport;
175
68edaf54 176 /* Locate the divert info */
077c74e2 177 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
68edaf54
SZ
178 divinfo = m_tag_data(mtag);
179 divsrc.sin_port = divinfo->skipto;
077c74e2 180
984263bc
MD
181 /*
182 * Record receive interface address, if any.
183 * But only for incoming packets.
184 */
185 divsrc.sin_addr.s_addr = 0;
186 if (incoming) {
b2632176 187 struct ifaddr_container *ifac;
984263bc 188
984263bc 189 /* Find IP address for receive interface */
b2632176
SZ
190 TAILQ_FOREACH(ifac, &m->m_pkthdr.rcvif->if_addrheads[mycpuid],
191 ifa_link) {
192 struct ifaddr *ifa = ifac->ifa;
193
984263bc
MD
194 if (ifa->ifa_addr == NULL)
195 continue;
196 if (ifa->ifa_addr->sa_family != AF_INET)
197 continue;
198 divsrc.sin_addr =
199 ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr;
200 break;
201 }
202 }
203 /*
204 * Record the incoming interface name whenever we have one.
205 */
984263bc
MD
206 if (m->m_pkthdr.rcvif) {
207 /*
f23061d4 208 * Hide the actual interface name in there in the
984263bc
MD
209 * sin_zero array. XXX This needs to be moved to a
210 * different sockaddr type for divert, e.g.
f23061d4 211 * sockaddr_div with multiple fields like
984263bc
MD
212 * sockaddr_dl. Presently we have only 7 bytes
213 * but that will do for now as most interfaces
214 * are 4 or less + 2 or less bytes for unit.
215 * There is probably a faster way of doing this,
216 * possibly taking it from the sockaddr_dl on the iface.
217 * This solves the problem of a P2P link and a LAN interface
218 * having the same address, which can result in the wrong
219 * interface being assigned to the packet when fed back
220 * into the divert socket. Theoretically if the daemon saves
221 * and re-uses the sockaddr_in as suggested in the man pages,
222 * this iface name will come along for the ride.
223 * (see div_output for the other half of this.)
f23061d4 224 */
a7a95662 225 ksnprintf(divsrc.sin_zero, sizeof divsrc.sin_zero, "%s",
87b66be9 226 m->m_pkthdr.rcvif->if_xname);
984263bc
MD
227 }
228
229 /* Put packet on socket queue, if any */
230 sa = NULL;
231 nport = htons((u_int16_t)port);
87b66be9
SZ
232
233 /*
234 * XXX
235 * Following loop to locate the inpcb is MPSAFE since the inpcb
236 * insertion/removal happens on the same CPU (CPU0), however,
237 * saving/testing the socket pointer is not MPSAFE. So we still
238 * need to hold BGL here.
239 */
6cef7136 240 lwkt_gettoken(&div_token);
d2e9e54c
MD
241 LIST_FOREACH(inp, &divcbinfo.pcblisthead, inp_list) {
242 if (inp->inp_flags & INP_PLACEMARKER)
243 continue;
984263bc
MD
244 if (inp->inp_lport == nport)
245 sa = inp->inp_socket;
246 }
247 if (sa) {
6cef7136
MD
248 lwkt_gettoken(&sa->so_rcv.ssb_token);
249 if (ssb_appendaddr(&sa->so_rcv, (struct sockaddr *)&divsrc, m, NULL) == 0)
984263bc
MD
250 m_freem(m);
251 else
252 sorwakeup(sa);
6cef7136 253 lwkt_reltoken(&sa->so_rcv.ssb_token);
984263bc
MD
254 } else {
255 m_freem(m);
256 ipstat.ips_noproto++;
257 ipstat.ips_delivered--;
f23061d4 258 }
6cef7136 259 lwkt_reltoken(&div_token);
984263bc
MD
260}
261
6f3f81a8 262static void
002c1265 263div_packet_handler(netmsg_t msg)
6f3f81a8 264{
6f3f81a8
SZ
265 struct mbuf *m;
266 int port, incoming = 0;
267
002c1265 268 m = msg->packet.nm_packet;
6f3f81a8 269
002c1265
MD
270 port = msg->lmsg.u.ms_result32 & 0xffff;
271 if (msg->lmsg.u.ms_result32 & DIV_INPUT)
6f3f81a8 272 incoming = 1;
6f3f81a8 273 div_packet(m, incoming, port);
002c1265 274 /* no reply, msg embedded in mbuf */
6f3f81a8 275}
002c1265 276
0030d31a 277static void
b0cb7c2c 278divert_packet(struct mbuf *m, int incoming)
6f3f81a8 279{
b0cb7c2c
SZ
280 struct m_tag *mtag;
281 struct divert_info *divinfo;
282 int port;
283
6f3f81a8
SZ
284 M_ASSERTPKTHDR(m);
285
286 /* Assure header */
287 if (m->m_len < sizeof(struct ip) &&
288 (m = m_pullup(m, sizeof(struct ip))) == NULL)
289 return;
290
b0cb7c2c
SZ
291 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
292 KASSERT(mtag != NULL, ("%s no divert tag!", __func__));
293 divinfo = m_tag_data(mtag);
294
295 port = divinfo->port;
296 KASSERT(port != 0, ("%s: port=0", __func__));
6f3f81a8 297
6f3f81a8
SZ
298 if (mycpuid != 0) {
299 struct netmsg_packet *nmp;
6f3f81a8
SZ
300
301 nmp = &m->m_hdr.mh_netmsg;
002c1265 302 netmsg_init(&nmp->base, NULL, &netisr_apanic_rport,
c3c96e44 303 0, div_packet_handler);
6f3f81a8
SZ
304 nmp->nm_packet = m;
305
002c1265 306 nmp->base.lmsg.u.ms_result32 = port; /* port is 16bits */
6f3f81a8 307 if (incoming)
002c1265 308 nmp->base.lmsg.u.ms_result32 |= DIV_INPUT;
6f3f81a8 309 else
002c1265 310 nmp->base.lmsg.u.ms_result32 |= DIV_OUTPUT;
6f3f81a8 311
3abced87 312 lwkt_sendmsg(netisr_portfn(0), &nmp->base.lmsg);
67bf99c4
MD
313 } else {
314 div_packet(m, incoming, port);
315 }
6f3f81a8
SZ
316}
317
984263bc
MD
318/*
319 * Deliver packet back into the IP processing machinery.
320 *
321 * If no address specified, or address is 0.0.0.0, send to ip_output();
322 * otherwise, send to ip_input() and mark as having been received on
323 * the interface with that address.
324 */
325static int
326div_output(struct socket *so, struct mbuf *m,
327 struct sockaddr_in *sin, struct mbuf *control)
328{
329 int error = 0;
e5ecc832 330 struct m_tag *mtag;
68edaf54 331 struct divert_info *divinfo;
984263bc 332
934c6849
SZ
333 if (control)
334 m_freem(control); /* XXX */
335
984263bc
MD
336 /*
337 * Prepare the tag for divert info. Note that a packet
338 * with a 0 tag in mh_data is effectively untagged,
339 * so we could optimize that case.
340 */
68edaf54 341 mtag = m_tag_get(PACKET_TAG_IPFW_DIVERT, sizeof(*divinfo), MB_DONTWAIT);
e5ecc832
JS
342 if (mtag == NULL) {
343 error = ENOBUFS;
344 goto cantsend;
345 }
346 m_tag_prepend(m, mtag);
984263bc
MD
347
348 /* Loopback avoidance and state recovery */
68edaf54 349 divinfo = m_tag_data(mtag);
934c6849 350 if (sin)
68edaf54 351 divinfo->skipto = sin->sin_port;
934c6849 352 else
68edaf54 353 divinfo->skipto = 0;
984263bc
MD
354
355 /* Reinject packet into the system as incoming or outgoing */
934c6849 356 if (DIV_IS_OUTPUT(sin)) {
984263bc
MD
357 struct ip *const ip = mtod(m, struct ip *);
358
b171561b
SZ
359 /* Don't allow packet length sizes that will crash */
360 if ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len) {
984263bc
MD
361 error = EINVAL;
362 goto cantsend;
363 }
364
365 /* Convert fields to host order for ip_output() */
366 ip->ip_len = ntohs(ip->ip_len);
367 ip->ip_off = ntohs(ip->ip_off);
368
369 /* Send packet to output processing */
370 ipstat.ips_rawout++; /* XXX */
b171561b 371 error = ip_output(m, NULL, NULL,
984263bc
MD
372 (so->so_options & SO_DONTROUTE) |
373 IP_ALLOWBROADCAST | IP_RAWOUTPUT,
b171561b 374 NULL, NULL);
984263bc 375 } else {
e5ecc832 376 ip_input(m);
984263bc 377 }
984263bc
MD
378 return error;
379
380cantsend:
381 m_freem(m);
382 return error;
383}
384
002c1265
MD
385static void
386div_attach(netmsg_t msg)
984263bc 387{
002c1265
MD
388 struct socket *so = msg->attach.base.nm_so;
389 int proto = msg->attach.nm_proto;
390 struct pru_attach_info *ai = msg->attach.nm_ai;
984263bc 391 struct inpcb *inp;
1cae611f 392 int error;
984263bc 393
ed894f8c 394 inp = so->so_pcb;
984263bc
MD
395 if (inp)
396 panic("div_attach");
002c1265
MD
397 error = priv_check_cred(ai->p_ucred, PRIV_ROOT, NULL_CRED_OKAY);
398 if (error)
399 goto out;
984263bc 400
e4700d00 401 error = soreserve(so, div_sendspace, div_recvspace, ai->sb_rlimit);
984263bc 402 if (error)
002c1265 403 goto out;
6cef7136 404 lwkt_gettoken(&div_token);
3abced87 405 sosetport(so, netisr_portfn(0));
e4700d00 406 error = in_pcballoc(so, &divcbinfo);
6cef7136
MD
407 if (error) {
408 lwkt_reltoken(&div_token);
002c1265 409 goto out;
6cef7136 410 }
984263bc
MD
411 inp = (struct inpcb *)so->so_pcb;
412 inp->inp_ip_p = proto;
413 inp->inp_vflag |= INP_IPV4;
414 inp->inp_flags |= INP_HDRINCL;
5fe66e68
JH
415 /*
416 * The socket is always "connected" because
417 * we always know "where" to send the packet.
418 */
6cef7136
MD
419 sosetstate(so, SS_ISCONNECTED);
420 lwkt_reltoken(&div_token);
002c1265
MD
421 error = 0;
422out:
423 lwkt_replymsg(&msg->attach.base.lmsg, error);
984263bc
MD
424}
425
002c1265
MD
426static void
427div_detach(netmsg_t msg)
984263bc 428{
002c1265 429 struct socket *so = msg->detach.base.nm_so;
984263bc
MD
430 struct inpcb *inp;
431
ed894f8c 432 inp = so->so_pcb;
5fe66e68 433 if (inp == NULL)
984263bc
MD
434 panic("div_detach");
435 in_pcbdetach(inp);
002c1265 436 lwkt_replymsg(&msg->detach.base.lmsg, 0);
984263bc
MD
437}
438
6cef7136
MD
439/*
440 * NOTE: (so) is referenced from soabort*() and netmsg_pru_abort()
441 * will sofree() it when we return.
442 */
002c1265
MD
443static void
444div_abort(netmsg_t msg)
984263bc 445{
002c1265 446 struct socket *so = msg->abort.base.nm_so;
6cef7136 447
984263bc 448 soisdisconnected(so);
002c1265
MD
449 div_detach(msg);
450 /* msg invalid now */
984263bc
MD
451}
452
002c1265
MD
453static void
454div_disconnect(netmsg_t msg)
984263bc 455{
002c1265 456 struct socket *so = msg->disconnect.base.nm_so;
6cef7136
MD
457 int error;
458
002c1265
MD
459 if (so->so_state & SS_ISCONNECTED) {
460 soreference(so);
461 div_abort(msg);
462 /* msg invalid now */
463 sofree(so);
464 return;
465 }
466 error = ENOTCONN;
467 lwkt_replymsg(&msg->disconnect.base.lmsg, error);
984263bc
MD
468}
469
002c1265
MD
470static void
471div_bind(netmsg_t msg)
984263bc 472{
002c1265
MD
473 struct socket *so = msg->bind.base.nm_so;
474 struct sockaddr *nam = msg->bind.nm_nam;
984263bc
MD
475 int error;
476
65b1ea7e
SZ
477 /*
478 * in_pcbbind assumes that nam is a sockaddr_in
984263bc
MD
479 * and in_pcbbind requires a valid address. Since divert
480 * sockets don't we need to make sure the address is
481 * filled in properly.
482 * XXX -- divert should not be abusing in_pcbind
483 * and should probably have its own family.
484 */
65b1ea7e 485 if (nam->sa_family != AF_INET) {
984263bc 486 error = EAFNOSUPPORT;
65b1ea7e 487 } else {
984263bc 488 ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY;
002c1265 489 error = in_pcbbind(so->so_pcb, nam, msg->bind.nm_td);
984263bc 490 }
002c1265 491 lwkt_replymsg(&msg->bind.base.lmsg, error);
984263bc
MD
492}
493
002c1265
MD
494static void
495div_shutdown(netmsg_t msg)
984263bc 496{
002c1265
MD
497 struct socket *so = msg->shutdown.base.nm_so;
498
984263bc 499 socantsendmore(so);
002c1265
MD
500
501 lwkt_replymsg(&msg->shutdown.base.lmsg, 0);
984263bc
MD
502}
503
002c1265
MD
504static void
505div_send(netmsg_t msg)
984263bc 506{
002c1265
MD
507 struct socket *so = msg->send.base.nm_so;
508 struct mbuf *m = msg->send.nm_m;
509 struct sockaddr *nam = msg->send.nm_addr;
510 struct mbuf *control = msg->send.nm_control;
511 int error;
512
c3c96e44 513 /* Length check already done in ip_cpufn() */
934c6849 514 KASSERT(m->m_len >= sizeof(struct ip), ("IP header not in one mbuf"));
984263bc
MD
515
516 /* Send packet */
002c1265
MD
517 error = div_output(so, m, (struct sockaddr_in *)nam, control);
518 lwkt_replymsg(&msg->send.base.lmsg, error);
984263bc
MD
519}
520
984263bc 521SYSCTL_DECL(_net_inet_divert);
48a2763f
JS
522SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLFLAG_RD, &divcbinfo, 0,
523 in_pcblist_global, "S,xinpcb", "List of active divert sockets");
984263bc
MD
524
525struct pr_usrreqs div_usrreqs = {
fa5e758c 526 .pru_abort = div_abort,
002c1265 527 .pru_accept = pr_generic_notsupp,
fa5e758c
MD
528 .pru_attach = div_attach,
529 .pru_bind = div_bind,
002c1265
MD
530 .pru_connect = pr_generic_notsupp,
531 .pru_connect2 = pr_generic_notsupp,
532 .pru_control = in_control_dispatch,
fa5e758c
MD
533 .pru_detach = div_detach,
534 .pru_disconnect = div_disconnect,
002c1265
MD
535 .pru_listen = pr_generic_notsupp,
536 .pru_peeraddr = in_setpeeraddr_dispatch,
537 .pru_rcvd = pr_generic_notsupp,
538 .pru_rcvoob = pr_generic_notsupp,
fa5e758c
MD
539 .pru_send = div_send,
540 .pru_sense = pru_sense_null,
541 .pru_shutdown = div_shutdown,
002c1265 542 .pru_sockaddr = in_setsockaddr_dispatch,
fa5e758c 543 .pru_sosend = sosend,
8b5c39bb 544 .pru_soreceive = soreceive
984263bc 545};
0030d31a
SZ
546
547static struct mbuf *
548ip_divert_out(struct mbuf *m, int tee)
549{
550 struct mbuf *clone = NULL;
551 struct ip *ip = mtod(m, struct ip *);
552
553 /* Clone packet if we're doing a 'tee' */
554 if (tee)
555 clone = m_dup(m, MB_DONTWAIT);
556
557 /*
558 * XXX
559 * delayed checksums are not currently compatible
560 * with divert sockets.
561 */
562 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
563 in_delayed_cksum(m);
564 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
565 }
566
567 /* Restore packet header fields to original values */
568 ip->ip_len = htons(ip->ip_len);
569 ip->ip_off = htons(ip->ip_off);
570
571 /* Deliver packet to divert input routine */
572 divert_packet(m, 0);
573
574 /* If 'tee', continue with original packet */
575 return clone;
576}
577
578static struct mbuf *
579ip_divert_in(struct mbuf *m, int tee)
580{
581 struct mbuf *clone = NULL;
582 struct ip *ip = mtod(m, struct ip *);
583 struct m_tag *mtag;
584
585 if (ip->ip_off & (IP_MF | IP_OFFMASK)) {
586 const struct divert_info *divinfo;
587 u_short frag_off;
588 int hlen;
589
590 /*
591 * Only trust divert info in the fragment
592 * at offset 0.
593 */
594 frag_off = ip->ip_off << 3;
595 if (frag_off != 0) {
596 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
597 m_tag_delete(m, mtag);
598 }
599
600 /*
601 * Attempt reassembly; if it succeeds, proceed.
602 * ip_reass() will return a different mbuf.
603 */
604 m = ip_reass(m);
605 if (m == NULL)
606 return NULL;
607 ip = mtod(m, struct ip *);
608
609 /* Caller need to redispatch the packet, if it is for us */
610 m->m_pkthdr.fw_flags |= FW_MBUF_REDISPATCH;
611
612 /*
613 * Get the header length of the reassembled
614 * packet
615 */
616 hlen = IP_VHL_HL(ip->ip_vhl) << 2;
617
618 /*
619 * Restore original checksum before diverting
620 * packet
621 */
622 ip->ip_len += hlen;
623 ip->ip_len = htons(ip->ip_len);
624 ip->ip_off = htons(ip->ip_off);
625 ip->ip_sum = 0;
626 if (hlen == sizeof(struct ip))
627 ip->ip_sum = in_cksum_hdr(ip);
628 else
629 ip->ip_sum = in_cksum(m, hlen);
630 ip->ip_off = ntohs(ip->ip_off);
631 ip->ip_len = ntohs(ip->ip_len);
632
633 /*
634 * Only use the saved divert info
635 */
636 mtag = m_tag_find(m, PACKET_TAG_IPFW_DIVERT, NULL);
637 if (mtag == NULL) {
638 /* Wrongly configured ipfw */
639 kprintf("ip_input no divert info\n");
640 m_freem(m);
641 return NULL;
642 }
643 divinfo = m_tag_data(mtag);
644 tee = divinfo->tee;
645 }
646
647 /*
648 * Divert or tee packet to the divert protocol if
649 * required.
650 */
651
652 /* Clone packet if we're doing a 'tee' */
653 if (tee)
654 clone = m_dup(m, MB_DONTWAIT);
655
656 /*
657 * Restore packet header fields to original
658 * values
659 */
660 ip->ip_len = htons(ip->ip_len);
661 ip->ip_off = htons(ip->ip_off);
662
663 /* Deliver packet to divert input routine */
664 divert_packet(m, 1);
665
666 /* Catch invalid reference */
667 m = NULL;
668 ip = NULL;
669
670 ipstat.ips_delivered++;
671
672 /* If 'tee', continue with original packet */
673 if (clone != NULL) {
674 /*
675 * Complete processing of the packet.
676 * XXX Better safe than sorry, remove the DIVERT tag.
677 */
678 mtag = m_tag_find(clone, PACKET_TAG_IPFW_DIVERT, NULL);
679 KKASSERT(mtag != NULL);
680 m_tag_delete(clone, mtag);
681 }
682 return clone;
683}
684
685static struct mbuf *
686ip_divert(struct mbuf *m, int tee, int incoming)
687{
688 struct mbuf *ret;
689
690 if (incoming)
691 ret = ip_divert_in(m, tee);
692 else
693 ret = ip_divert_out(m, tee);
694 return ret;
695}