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. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * $FreeBSD: src/sys/netinet/ip_divert.c,v 1.42.2.6 2003/01/23 21:06:45 sam Exp $
35 #include "opt_ipdivert.h"
36 #include "opt_ipsec.h"
39 #error "IPDIVERT requires INET."
42 #include <sys/param.h>
43 #include <sys/kernel.h>
44 #include <sys/malloc.h>
46 #include <sys/socket.h>
47 #include <sys/protosw.h>
48 #include <sys/socketvar.h>
49 #include <sys/socketvar2.h>
50 #include <sys/sysctl.h>
51 #include <sys/systm.h>
54 #include <sys/in_cksum.h>
56 #include <sys/msgport.h>
59 #include <net/route.h>
61 #include <net/netmsg2.h>
62 #include <net/netisr2.h>
63 #include <sys/thread2.h>
64 #include <sys/mplock2.h>
66 #include <netinet/in.h>
67 #include <netinet/in_systm.h>
68 #include <netinet/ip.h>
69 #include <netinet/in_pcb.h>
70 #include <netinet/in_var.h>
71 #include <netinet/ip_var.h>
72 #include <netinet/ip_divert.h>
79 * Allocate enough space to hold a full IP packet
81 #define DIVSNDQ (65536 + 100)
82 #define DIVRCVQ (65536 + 100)
84 #define DIV_IS_OUTPUT(sin) ((sin) == NULL || (sin)->sin_addr.s_addr == 0)
86 #define DIV_OUTPUT 0x10000
87 #define DIV_INPUT 0x20000
90 * Divert sockets work in conjunction with ipfw, see the divert(4)
91 * manpage for features.
92 * Internally, packets selected by ipfw in ip_input() or ip_output(),
93 * and never diverted before, are passed to the input queue of the
94 * divert socket with a given 'divert_port' number (as specified in
95 * the matching ipfw rule), and they are tagged with a 16 bit cookie
96 * (representing the rule number of the matching ipfw rule), which
97 * is passed to process reading from the socket.
99 * Packets written to the divert socket are again tagged with a cookie
100 * (usually the same as above) and a destination address.
101 * If the destination address is INADDR_ANY then the packet is
102 * treated as outgoing and sent to ip_output(), otherwise it is
103 * treated as incoming and sent to ip_input().
104 * In both cases, the packet is tagged with the cookie.
106 * On reinjection, processing in ip_input() and ip_output()
107 * will be exactly the same as for the original packet, except that
108 * ipfw processing will start at the rule number after the one
109 * written in the cookie (so, tagging a packet with a cookie of 0
110 * will cause it to be effectively considered as a standard packet).
113 /* Internal variables */
114 static struct inpcbinfo divcbinfo
;
115 static struct inpcbportinfo divcbportinfo
;
117 static u_long div_sendspace
= DIVSNDQ
; /* XXX sysctl ? */
118 static u_long div_recvspace
= DIVRCVQ
; /* XXX sysctl ? */
120 static struct mbuf
*ip_divert(struct mbuf
*, int, int);
123 * Initialize divert connection block queue.
128 in_pcbinfo_init(&divcbinfo
, 0, FALSE
);
129 in_pcbportinfo_init(&divcbportinfo
, 1, 0);
131 * XXX We don't use the hash list for divert IP, but it's easier
132 * to allocate a one entry hash list than it is to check all
133 * over the place for hashbase == NULL.
135 divcbinfo
.hashbase
= hashinit(1, M_PCB
, &divcbinfo
.hashmask
);
136 divcbinfo
.portinfo
= &divcbportinfo
;
137 divcbinfo
.wildcardhashbase
= hashinit(1, M_PCB
,
138 &divcbinfo
.wildcardhashmask
);
139 divcbinfo
.ipi_size
= sizeof(struct inpcb
);
140 ip_divert_p
= ip_divert
;
144 * IPPROTO_DIVERT is not a real IP protocol; don't allow any packets
145 * with that protocol number to enter the system from the outside.
148 div_input(struct mbuf
**mp
, int *offp
, int proto
)
150 struct mbuf
*m
= *mp
;
152 ipstat
.ips_noproto
++;
154 return(IPPROTO_DONE
);
158 * Divert a packet by passing it up to the divert socket at port 'port'.
160 * Setup generic address and protocol structures for div_input routine,
161 * then pass them along with mbuf chain.
164 div_packet(struct mbuf
*m
, int incoming
, int port
)
166 struct sockaddr_in divsrc
= { sizeof divsrc
, AF_INET
};
170 struct divert_info
*divinfo
;
175 /* Locate the divert info */
176 mtag
= m_tag_find(m
, PACKET_TAG_IPFW_DIVERT
, NULL
);
177 divinfo
= m_tag_data(mtag
);
178 divsrc
.sin_port
= divinfo
->skipto
;
181 * Record receive interface address, if any.
182 * But only for incoming packets.
184 divsrc
.sin_addr
.s_addr
= 0;
186 struct ifaddr_container
*ifac
;
188 /* Find IP address for receive interface */
189 TAILQ_FOREACH(ifac
, &m
->m_pkthdr
.rcvif
->if_addrheads
[mycpuid
],
191 struct ifaddr
*ifa
= ifac
->ifa
;
193 if (ifa
->ifa_addr
== NULL
)
195 if (ifa
->ifa_addr
->sa_family
!= AF_INET
)
198 ((struct sockaddr_in
*) ifa
->ifa_addr
)->sin_addr
;
203 * Record the incoming interface name whenever we have one.
205 if (m
->m_pkthdr
.rcvif
) {
207 * Hide the actual interface name in there in the
208 * sin_zero array. XXX This needs to be moved to a
209 * different sockaddr type for divert, e.g.
210 * sockaddr_div with multiple fields like
211 * sockaddr_dl. Presently we have only 7 bytes
212 * but that will do for now as most interfaces
213 * are 4 or less + 2 or less bytes for unit.
214 * There is probably a faster way of doing this,
215 * possibly taking it from the sockaddr_dl on the iface.
216 * This solves the problem of a P2P link and a LAN interface
217 * having the same address, which can result in the wrong
218 * interface being assigned to the packet when fed back
219 * into the divert socket. Theoretically if the daemon saves
220 * and re-uses the sockaddr_in as suggested in the man pages,
221 * this iface name will come along for the ride.
222 * (see div_output for the other half of this.)
224 ksnprintf(divsrc
.sin_zero
, sizeof divsrc
.sin_zero
, "%s",
225 m
->m_pkthdr
.rcvif
->if_xname
);
228 /* Put packet on socket queue, if any */
230 nport
= htons((u_int16_t
)port
);
233 * Following loop to locate the inpcb is MPSAFE since the inpcb
234 * insertion/removal happens on the same CPU (CPU0).
236 LIST_FOREACH(inp
, &divcbinfo
.pcblisthead
, inp_list
) {
237 if (inp
->inp_flags
& INP_PLACEMARKER
)
239 if (inp
->inp_lport
== nport
)
240 sa
= inp
->inp_socket
;
243 lwkt_gettoken(&sa
->so_rcv
.ssb_token
);
244 if (ssb_appendaddr(&sa
->so_rcv
, (struct sockaddr
*)&divsrc
, m
, NULL
) == 0)
248 lwkt_reltoken(&sa
->so_rcv
.ssb_token
);
251 ipstat
.ips_noproto
++;
252 ipstat
.ips_delivered
--;
257 div_packet_handler(netmsg_t msg
)
260 int port
, incoming
= 0;
262 m
= msg
->packet
.nm_packet
;
264 port
= msg
->lmsg
.u
.ms_result32
& 0xffff;
265 if (msg
->lmsg
.u
.ms_result32
& DIV_INPUT
)
267 div_packet(m
, incoming
, port
);
268 /* no reply, msg embedded in mbuf */
272 divert_packet(struct mbuf
*m
, int incoming
)
275 struct divert_info
*divinfo
;
281 if (m
->m_len
< sizeof(struct ip
) &&
282 (m
= m_pullup(m
, sizeof(struct ip
))) == NULL
)
285 mtag
= m_tag_find(m
, PACKET_TAG_IPFW_DIVERT
, NULL
);
286 KASSERT(mtag
!= NULL
, ("%s no divert tag!", __func__
));
287 divinfo
= m_tag_data(mtag
);
289 port
= divinfo
->port
;
290 KASSERT(port
!= 0, ("%s: port=0", __func__
));
293 struct netmsg_packet
*nmp
;
295 nmp
= &m
->m_hdr
.mh_netmsg
;
296 netmsg_init(&nmp
->base
, NULL
, &netisr_apanic_rport
,
297 0, div_packet_handler
);
300 nmp
->base
.lmsg
.u
.ms_result32
= port
; /* port is 16bits */
302 nmp
->base
.lmsg
.u
.ms_result32
|= DIV_INPUT
;
304 nmp
->base
.lmsg
.u
.ms_result32
|= DIV_OUTPUT
;
306 lwkt_sendmsg(netisr_cpuport(0), &nmp
->base
.lmsg
);
308 div_packet(m
, incoming
, port
);
313 * Deliver packet back into the IP processing machinery.
315 * If no address specified, or address is 0.0.0.0, send to ip_output();
316 * otherwise, send to ip_input() and mark as having been received on
317 * the interface with that address.
320 div_output(struct socket
*so
, struct mbuf
*m
,
321 struct sockaddr_in
*sin
, struct mbuf
*control
)
325 struct divert_info
*divinfo
;
330 m_freem(control
); /* XXX */
333 * Prepare the tag for divert info. Note that a packet
334 * with a 0 tag in mh_data is effectively untagged,
335 * so we could optimize that case.
337 mtag
= m_tag_get(PACKET_TAG_IPFW_DIVERT
, sizeof(*divinfo
), M_NOWAIT
);
342 m_tag_prepend(m
, mtag
);
344 /* Loopback avoidance and state recovery */
345 divinfo
= m_tag_data(mtag
);
347 divinfo
->skipto
= sin
->sin_port
;
351 /* Reinject packet into the system as incoming or outgoing */
352 if (DIV_IS_OUTPUT(sin
)) {
353 struct ip
*const ip
= mtod(m
, struct ip
*);
355 /* Don't allow packet length sizes that will crash */
356 if ((u_short
)ntohs(ip
->ip_len
) > m
->m_pkthdr
.len
) {
361 /* Convert fields to host order for ip_output() */
362 ip
->ip_len
= ntohs(ip
->ip_len
);
363 ip
->ip_off
= ntohs(ip
->ip_off
);
365 /* Send packet to output processing */
366 ipstat
.ips_rawout
++; /* XXX */
367 error
= ip_output(m
, NULL
, NULL
,
368 (so
->so_options
& SO_DONTROUTE
) |
369 IP_ALLOWBROADCAST
| IP_RAWOUTPUT
,
382 div_attach(netmsg_t msg
)
384 struct socket
*so
= msg
->attach
.base
.nm_so
;
385 int proto
= msg
->attach
.nm_proto
;
386 struct pru_attach_info
*ai
= msg
->attach
.nm_ai
;
395 error
= priv_check_cred(ai
->p_ucred
, PRIV_ROOT
, NULL_CRED_OKAY
);
399 error
= soreserve(so
, div_sendspace
, div_recvspace
, ai
->sb_rlimit
);
402 error
= in_pcballoc(so
, &divcbinfo
);
405 inp
= (struct inpcb
*)so
->so_pcb
;
406 inp
->inp_ip_p
= proto
;
407 inp
->inp_flags
|= INP_HDRINCL
;
409 * The socket is always "connected" because
410 * we always know "where" to send the packet.
412 sosetstate(so
, SS_ISCONNECTED
);
415 lwkt_replymsg(&msg
->attach
.base
.lmsg
, error
);
419 div_detach(netmsg_t msg
)
421 struct socket
*so
= msg
->detach
.base
.nm_so
;
430 lwkt_replymsg(&msg
->detach
.base
.lmsg
, 0);
434 div_abort(netmsg_t msg
)
437 * Divert socket does not support listen(2),
438 * so this should never be called.
440 panic("div_abort is called");
444 div_disconnect(netmsg_t msg
)
446 struct socket
*so
= msg
->disconnect
.base
.nm_so
;
451 if (so
->so_state
& SS_ISCONNECTED
) {
452 soisdisconnected(so
);
457 lwkt_replymsg(&msg
->disconnect
.base
.lmsg
, error
);
461 div_bind(netmsg_t msg
)
463 struct socket
*so
= msg
->bind
.base
.nm_so
;
464 struct sockaddr
*nam
= msg
->bind
.nm_nam
;
470 * in_pcbbind assumes that nam is a sockaddr_in
471 * and in_pcbbind requires a valid address. Since divert
472 * sockets don't we need to make sure the address is
473 * filled in properly.
474 * XXX -- divert should not be abusing in_pcbind
475 * and should probably have its own family.
477 if (nam
->sa_family
!= AF_INET
) {
478 error
= EAFNOSUPPORT
;
480 ((struct sockaddr_in
*)nam
)->sin_addr
.s_addr
= INADDR_ANY
;
481 error
= in_pcbbind(so
->so_pcb
, nam
, msg
->bind
.nm_td
);
483 lwkt_replymsg(&msg
->bind
.base
.lmsg
, error
);
487 div_shutdown(netmsg_t msg
)
489 struct socket
*so
= msg
->shutdown
.base
.nm_so
;
495 lwkt_replymsg(&msg
->shutdown
.base
.lmsg
, 0);
499 div_send(netmsg_t msg
)
501 struct socket
*so
= msg
->send
.base
.nm_so
;
502 struct mbuf
*m
= msg
->send
.nm_m
;
503 struct sockaddr
*nam
= msg
->send
.nm_addr
;
504 struct mbuf
*control
= msg
->send
.nm_control
;
507 /* Length check already done in ip_hashfn() */
508 KASSERT(m
->m_len
>= sizeof(struct ip
), ("IP header not in one mbuf"));
511 error
= div_output(so
, m
, (struct sockaddr_in
*)nam
, control
);
512 lwkt_replymsg(&msg
->send
.base
.lmsg
, error
);
515 SYSCTL_DECL(_net_inet_divert
);
516 SYSCTL_PROC(_net_inet_divert
, OID_AUTO
, pcblist
, CTLFLAG_RD
, &divcbinfo
, 1,
517 in_pcblist_global
, "S,xinpcb", "List of active divert sockets");
519 struct pr_usrreqs div_usrreqs
= {
520 .pru_abort
= div_abort
,
521 .pru_accept
= pr_generic_notsupp
,
522 .pru_attach
= div_attach
,
523 .pru_bind
= div_bind
,
524 .pru_connect
= pr_generic_notsupp
,
525 .pru_connect2
= pr_generic_notsupp
,
526 .pru_control
= in_control_dispatch
,
527 .pru_detach
= div_detach
,
528 .pru_disconnect
= div_disconnect
,
529 .pru_listen
= pr_generic_notsupp
,
530 .pru_peeraddr
= in_setpeeraddr_dispatch
,
531 .pru_rcvd
= pr_generic_notsupp
,
532 .pru_rcvoob
= pr_generic_notsupp
,
533 .pru_send
= div_send
,
534 .pru_sense
= pru_sense_null
,
535 .pru_shutdown
= div_shutdown
,
536 .pru_sockaddr
= in_setsockaddr_dispatch
,
537 .pru_sosend
= sosend
,
538 .pru_soreceive
= soreceive
542 ip_divert_out(struct mbuf
*m
, int tee
)
544 struct mbuf
*clone
= NULL
;
545 struct ip
*ip
= mtod(m
, struct ip
*);
547 /* Clone packet if we're doing a 'tee' */
549 clone
= m_dup(m
, M_NOWAIT
);
553 * delayed checksums are not currently compatible
554 * with divert sockets.
556 if (m
->m_pkthdr
.csum_flags
& CSUM_DELAY_DATA
) {
558 m
->m_pkthdr
.csum_flags
&= ~CSUM_DELAY_DATA
;
561 /* Restore packet header fields to original values */
562 ip
->ip_len
= htons(ip
->ip_len
);
563 ip
->ip_off
= htons(ip
->ip_off
);
565 /* Deliver packet to divert input routine */
568 /* If 'tee', continue with original packet */
573 ip_divert_in(struct mbuf
*m
, int tee
)
575 struct mbuf
*clone
= NULL
;
576 struct ip
*ip
= mtod(m
, struct ip
*);
579 if (ip
->ip_off
& (IP_MF
| IP_OFFMASK
)) {
580 const struct divert_info
*divinfo
;
585 * Only trust divert info in the fragment
588 frag_off
= ip
->ip_off
<< 3;
590 mtag
= m_tag_find(m
, PACKET_TAG_IPFW_DIVERT
, NULL
);
591 m_tag_delete(m
, mtag
);
595 * Attempt reassembly; if it succeeds, proceed.
596 * ip_reass() will return a different mbuf.
601 ip
= mtod(m
, struct ip
*);
603 /* Caller need to redispatch the packet, if it is for us */
604 m
->m_pkthdr
.fw_flags
|= FW_MBUF_REDISPATCH
;
607 * Get the header length of the reassembled
610 hlen
= IP_VHL_HL(ip
->ip_vhl
) << 2;
613 * Restore original checksum before diverting
617 ip
->ip_len
= htons(ip
->ip_len
);
618 ip
->ip_off
= htons(ip
->ip_off
);
620 if (hlen
== sizeof(struct ip
))
621 ip
->ip_sum
= in_cksum_hdr(ip
);
623 ip
->ip_sum
= in_cksum(m
, hlen
);
624 ip
->ip_off
= ntohs(ip
->ip_off
);
625 ip
->ip_len
= ntohs(ip
->ip_len
);
628 * Only use the saved divert info
630 mtag
= m_tag_find(m
, PACKET_TAG_IPFW_DIVERT
, NULL
);
632 /* Wrongly configured ipfw */
633 kprintf("ip_input no divert info\n");
637 divinfo
= m_tag_data(mtag
);
642 * Divert or tee packet to the divert protocol if
646 /* Clone packet if we're doing a 'tee' */
648 clone
= m_dup(m
, M_NOWAIT
);
651 * Restore packet header fields to original
654 ip
->ip_len
= htons(ip
->ip_len
);
655 ip
->ip_off
= htons(ip
->ip_off
);
657 /* Deliver packet to divert input routine */
660 /* Catch invalid reference */
664 ipstat
.ips_delivered
++;
666 /* If 'tee', continue with original packet */
669 * Complete processing of the packet.
670 * XXX Better safe than sorry, remove the DIVERT tag.
672 mtag
= m_tag_find(clone
, PACKET_TAG_IPFW_DIVERT
, NULL
);
673 KKASSERT(mtag
!= NULL
);
674 m_tag_delete(clone
, mtag
);
680 ip_divert(struct mbuf
*m
, int tee
, int incoming
)
685 ret
= ip_divert_in(m
, tee
);
687 ret
= ip_divert_out(m
, tee
);