1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1982, 1986, 1988, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 */
31
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
34
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 #include "opt_sctp.h"
38 #ifndef INET
39 #error "IPDIVERT requires INET"
40 #endif
41
42 #include <sys/param.h>
43 #include <sys/eventhandler.h>
44 #include <sys/kernel.h>
45 #include <sys/lock.h>
46 #include <sys/malloc.h>
47 #include <sys/mbuf.h>
48 #include <sys/module.h>
49 #include <sys/kernel.h>
50 #include <sys/priv.h>
51 #include <sys/proc.h>
52 #include <sys/protosw.h>
53 #include <sys/socket.h>
54 #include <sys/socketvar.h>
55 #include <sys/sysctl.h>
56 #include <net/vnet.h>
57
58 #include <net/if.h>
59 #include <net/if_var.h>
60 #include <net/netisr.h>
61
62 #include <netinet/in.h>
63 #include <netinet/in_pcb.h>
64 #include <netinet/in_systm.h>
65 #include <netinet/in_var.h>
66 #include <netinet/ip.h>
67 #include <netinet/ip_var.h>
68 #ifdef INET6
69 #include <netinet/ip6.h>
70 #include <netinet6/ip6_var.h>
71 #endif
72 #if defined(SCTP) || defined(SCTP_SUPPORT)
73 #include <netinet/sctp_crc32.h>
74 #endif
75
76 #include <security/mac/mac_framework.h>
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
87 /*
88 * Divert sockets work in conjunction with ipfw or other packet filters,
89 * see the divert(4) manpage for features.
90 * Packets are selected by the packet filter and tagged with an
91 * MTAG_IPFW_RULE tag carrying the 'divert port' number (as set by
92 * the packet filter) and information on the matching filter rule for
93 * subsequent reinjection. The divert_port is used to put the packet
94 * on the corresponding divert socket, while the rule number is passed
95 * up (at least partially) as the sin_port in the struct sockaddr.
96 *
97 * Packets written to the divert socket carry in sin_addr a
98 * destination address, and in sin_port the number of the filter rule
99 * after which to continue processing.
100 * If the destination address is INADDR_ANY, the packet is treated as
101 * as outgoing and sent to ip_output(); otherwise it is treated as
102 * incoming and sent to ip_input().
103 * Further, sin_zero carries some information on the interface,
104 * which can be used in the reinject -- see comments in the code.
105 *
106 * On reinjection, processing in ip_input() and ip_output()
107 * will be exactly the same as for the original packet, except that
108 * packet filter processing will start at the rule number after the one
109 * written in the sin_port (ipfw does not allow a rule #0, so sin_port=0
110 * will apply the entire ruleset to the packet).
111 */
112
113 /* Internal variables. */
114 VNET_DEFINE_STATIC(struct inpcbhead, divcb);
115 VNET_DEFINE_STATIC(struct inpcbinfo, divcbinfo);
116
117 #define V_divcb VNET(divcb)
118 #define V_divcbinfo VNET(divcbinfo)
119
120 static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */
121 static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */
122
123 static eventhandler_tag ip_divert_event_tag;
124
125 static int div_output_inbound(int fmaily, struct socket *so, struct mbuf *m,
126 struct sockaddr_in *sin);
127 static int div_output_outbound(int family, struct socket *so, struct mbuf *m);
128
129 /*
130 * Initialize divert connection block queue.
131 */
132 static void
div_zone_change(void * tag)133 div_zone_change(void *tag)
134 {
135
136 uma_zone_set_max(V_divcbinfo.ipi_zone, maxsockets);
137 }
138
139 static int
div_inpcb_init(void * mem,int size,int flags)140 div_inpcb_init(void *mem, int size, int flags)
141 {
142 struct inpcb *inp = mem;
143
144 INP_LOCK_INIT(inp, "inp", "divinp");
145 return (0);
146 }
147
148 static void
div_init(void)149 div_init(void)
150 {
151
152 /*
153 * XXX We don't use the hash list for divert IP, but it's easier to
154 * allocate one-entry hash lists than it is to check all over the
155 * place for hashbase == NULL.
156 */
157 in_pcbinfo_init(&V_divcbinfo, "div", &V_divcb, 1, 1, "divcb",
158 div_inpcb_init, IPI_HASHFIELDS_NONE);
159 }
160
161 static void
div_destroy(void * unused __unused)162 div_destroy(void *unused __unused)
163 {
164
165 in_pcbinfo_destroy(&V_divcbinfo);
166 }
167 VNET_SYSUNINIT(divert, SI_SUB_PROTO_DOMAININIT, SI_ORDER_ANY,
168 div_destroy, NULL);
169
170 /*
171 * IPPROTO_DIVERT is not in the real IP protocol number space; this
172 * function should never be called. Just in case, drop any packets.
173 */
174 static int
div_input(struct mbuf ** mp,int * offp,int proto)175 div_input(struct mbuf **mp, int *offp, int proto)
176 {
177 struct mbuf *m = *mp;
178
179 KMOD_IPSTAT_INC(ips_noproto);
180 m_freem(m);
181 return (IPPROTO_DONE);
182 }
183
184 /*
185 * Divert a packet by passing it up to the divert socket at port 'port'.
186 *
187 * Setup generic address and protocol structures for div_input routine,
188 * then pass them along with mbuf chain.
189 */
190 static void
divert_packet(struct mbuf * m,bool incoming)191 divert_packet(struct mbuf *m, bool incoming)
192 {
193 struct ip *ip;
194 struct inpcb *inp;
195 struct socket *sa;
196 u_int16_t nport;
197 struct sockaddr_in divsrc;
198 struct m_tag *mtag;
199
200 NET_EPOCH_ASSERT();
201
202 mtag = m_tag_locate(m, MTAG_IPFW_RULE, 0, NULL);
203 if (mtag == NULL) {
204 m_freem(m);
205 return;
206 }
207 /* Assure header */
208 if (m->m_len < sizeof(struct ip) &&
209 (m = m_pullup(m, sizeof(struct ip))) == NULL)
210 return;
211 ip = mtod(m, struct ip *);
212
213 /* Delayed checksums are currently not compatible with divert. */
214 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
215 in_delayed_cksum(m);
216 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
217 }
218 #if defined(SCTP) || defined(SCTP_SUPPORT)
219 if (m->m_pkthdr.csum_flags & CSUM_SCTP) {
220 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2));
221 m->m_pkthdr.csum_flags &= ~CSUM_SCTP;
222 }
223 #endif
224 bzero(&divsrc, sizeof(divsrc));
225 divsrc.sin_len = sizeof(divsrc);
226 divsrc.sin_family = AF_INET;
227 /* record matching rule, in host format */
228 divsrc.sin_port = ((struct ipfw_rule_ref *)(mtag+1))->rulenum;
229 /*
230 * Record receive interface address, if any.
231 * But only for incoming packets.
232 */
233 if (incoming) {
234 struct ifaddr *ifa;
235 struct ifnet *ifp;
236
237 /* Sanity check */
238 M_ASSERTPKTHDR(m);
239
240 /* Find IP address for receive interface */
241 ifp = m->m_pkthdr.rcvif;
242 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) {
243 if (ifa->ifa_addr->sa_family != AF_INET)
244 continue;
245 divsrc.sin_addr =
246 ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr;
247 break;
248 }
249 }
250 /*
251 * Record the incoming interface name whenever we have one.
252 */
253 if (m->m_pkthdr.rcvif) {
254 /*
255 * Hide the actual interface name in there in the
256 * sin_zero array. XXX This needs to be moved to a
257 * different sockaddr type for divert, e.g.
258 * sockaddr_div with multiple fields like
259 * sockaddr_dl. Presently we have only 7 bytes
260 * but that will do for now as most interfaces
261 * are 4 or less + 2 or less bytes for unit.
262 * There is probably a faster way of doing this,
263 * possibly taking it from the sockaddr_dl on the iface.
264 * This solves the problem of a P2P link and a LAN interface
265 * having the same address, which can result in the wrong
266 * interface being assigned to the packet when fed back
267 * into the divert socket. Theoretically if the daemon saves
268 * and re-uses the sockaddr_in as suggested in the man pages,
269 * this iface name will come along for the ride.
270 * (see div_output for the other half of this.)
271 */
272 strlcpy(divsrc.sin_zero, m->m_pkthdr.rcvif->if_xname,
273 sizeof(divsrc.sin_zero));
274 }
275
276 /* Put packet on socket queue, if any */
277 sa = NULL;
278 nport = htons((u_int16_t)(((struct ipfw_rule_ref *)(mtag+1))->info));
279 CK_LIST_FOREACH(inp, &V_divcb, inp_list) {
280 /* XXX why does only one socket match? */
281 if (inp->inp_lport == nport) {
282 INP_RLOCK(inp);
283 sa = inp->inp_socket;
284 SOCKBUF_LOCK(&sa->so_rcv);
285 if (sbappendaddr_locked(&sa->so_rcv,
286 (struct sockaddr *)&divsrc, m,
287 (struct mbuf *)0) == 0) {
288 SOCKBUF_UNLOCK(&sa->so_rcv);
289 sa = NULL; /* force mbuf reclaim below */
290 } else
291 sorwakeup_locked(sa);
292 INP_RUNLOCK(inp);
293 break;
294 }
295 }
296 if (sa == NULL) {
297 m_freem(m);
298 KMOD_IPSTAT_INC(ips_noproto);
299 KMOD_IPSTAT_DEC(ips_delivered);
300 }
301 }
302
303 /*
304 * Deliver packet back into the IP processing machinery.
305 *
306 * If no address specified, or address is 0.0.0.0, send to ip_output();
307 * otherwise, send to ip_input() and mark as having been received on
308 * the interface with that address.
309 */
310 static int
div_output(struct socket * so,struct mbuf * m,struct sockaddr_in * sin,struct mbuf * control)311 div_output(struct socket *so, struct mbuf *m, struct sockaddr_in *sin,
312 struct mbuf *control)
313 {
314 struct epoch_tracker et;
315 const struct ip *ip;
316 struct m_tag *mtag;
317 struct ipfw_rule_ref *dt;
318 int error, family;
319
320 /*
321 * An mbuf may hasn't come from userland, but we pretend
322 * that it has.
323 */
324 m->m_pkthdr.rcvif = NULL;
325 m->m_nextpkt = NULL;
326 M_SETFIB(m, so->so_fibnum);
327
328 if (control)
329 m_freem(control); /* XXX */
330
331 mtag = m_tag_locate(m, MTAG_IPFW_RULE, 0, NULL);
332 if (mtag == NULL) {
333 /* this should be normal */
334 mtag = m_tag_alloc(MTAG_IPFW_RULE, 0,
335 sizeof(struct ipfw_rule_ref), M_NOWAIT | M_ZERO);
336 if (mtag == NULL) {
337 m_freem(m);
338 return (ENOBUFS);
339 }
340 m_tag_prepend(m, mtag);
341 }
342 dt = (struct ipfw_rule_ref *)(mtag+1);
343
344 /* Loopback avoidance and state recovery */
345 if (sin) {
346 int i;
347
348 /* set the starting point. We provide a non-zero slot,
349 * but a non_matching chain_id to skip that info and use
350 * the rulenum/rule_id.
351 */
352 dt->slot = 1; /* dummy, chain_id is invalid */
353 dt->chain_id = 0;
354 dt->rulenum = sin->sin_port+1; /* host format ? */
355 dt->rule_id = 0;
356 /* XXX: broken for IPv6 */
357 /*
358 * Find receive interface with the given name, stuffed
359 * (if it exists) in the sin_zero[] field.
360 * The name is user supplied data so don't trust its size
361 * or that it is zero terminated.
362 */
363 for (i = 0; i < sizeof(sin->sin_zero) && sin->sin_zero[i]; i++)
364 ;
365 if ( i > 0 && i < sizeof(sin->sin_zero))
366 m->m_pkthdr.rcvif = ifunit(sin->sin_zero);
367 }
368
369 ip = mtod(m, struct ip *);
370 switch (ip->ip_v) {
371 case IPVERSION:
372 family = AF_INET;
373 break;
374 #ifdef INET6
375 case IPV6_VERSION >> 4:
376 family = AF_INET6;
377 break;
378 #endif
379 default:
380 m_freem(m);
381 return (EAFNOSUPPORT);
382 }
383
384 /* Reinject packet into the system as incoming or outgoing */
385 NET_EPOCH_ENTER(et);
386 if (!sin || sin->sin_addr.s_addr == 0) {
387 dt->info |= IPFW_IS_DIVERT | IPFW_INFO_OUT;
388 error = div_output_outbound(family, so, m);
389 } else {
390 dt->info |= IPFW_IS_DIVERT | IPFW_INFO_IN;
391 error = div_output_inbound(family, so, m, sin);
392 }
393 NET_EPOCH_EXIT(et);
394
395 return (error);
396 }
397
398 /*
399 * Sends mbuf @m to the wire via ip[6]_output().
400 *
401 * Returns 0 on success or an errno value on failure. @m is always consumed.
402 */
403 static int
div_output_outbound(int family,struct socket * so,struct mbuf * m)404 div_output_outbound(int family, struct socket *so, struct mbuf *m)
405 {
406 struct ip *const ip = mtod(m, struct ip *);
407 struct mbuf *options;
408 struct inpcb *inp;
409 int error;
410
411 inp = sotoinpcb(so);
412 INP_RLOCK(inp);
413 switch (family) {
414 case AF_INET:
415 /*
416 * Don't allow both user specified and setsockopt
417 * options, and don't allow packet length sizes that
418 * will crash.
419 */
420 if ((((ip->ip_hl << 2) != sizeof(struct ip)) &&
421 inp->inp_options != NULL) ||
422 ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) {
423 INP_RUNLOCK(inp);
424 m_freem(m);
425 return (EINVAL);
426 }
427 break;
428 #ifdef INET6
429 case AF_INET6:
430 {
431 struct ip6_hdr *const ip6 = mtod(m, struct ip6_hdr *);
432
433 /* Don't allow packet length sizes that will crash */
434 if (((u_short)ntohs(ip6->ip6_plen) > m->m_pkthdr.len)) {
435 INP_RUNLOCK(inp);
436 m_freem(m);
437 return (EINVAL);
438 }
439 break;
440 }
441 #endif
442 }
443
444 /* Send packet to output processing */
445 KMOD_IPSTAT_INC(ips_rawout); /* XXX */
446
447 #ifdef MAC
448 mac_inpcb_create_mbuf(inp, m);
449 #endif
450 /*
451 * Get ready to inject the packet into ip_output().
452 * Just in case socket options were specified on the
453 * divert socket, we duplicate them. This is done
454 * to avoid having to hold the PCB locks over the call
455 * to ip_output(), as doing this results in a number of
456 * lock ordering complexities.
457 *
458 * Note that we set the multicast options argument for
459 * ip_output() to NULL since it should be invariant that
460 * they are not present.
461 */
462 KASSERT(inp->inp_moptions == NULL,
463 ("multicast options set on a divert socket"));
464 /*
465 * XXXCSJP: It is unclear to me whether or not it makes
466 * sense for divert sockets to have options. However,
467 * for now we will duplicate them with the INP locks
468 * held so we can use them in ip_output() without
469 * requring a reference to the pcb.
470 */
471 options = NULL;
472 if (inp->inp_options != NULL) {
473 options = m_dup(inp->inp_options, M_NOWAIT);
474 if (options == NULL) {
475 INP_RUNLOCK(inp);
476 m_freem(m);
477 return (ENOBUFS);
478 }
479 }
480 INP_RUNLOCK(inp);
481
482 error = 0;
483 switch (family) {
484 case AF_INET:
485 error = ip_output(m, options, NULL,
486 ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0)
487 | IP_ALLOWBROADCAST | IP_RAWOUTPUT, NULL, NULL);
488 break;
489 #ifdef INET6
490 case AF_INET6:
491 error = ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL);
492 break;
493 #endif
494 }
495 if (options != NULL)
496 m_freem(options);
497
498 return (error);
499 }
500
501 /*
502 * Schedules mbuf @m for local processing via IPv4/IPv6 netisr queue.
503 *
504 * Returns 0 on success or an errno value on failure. @m is always consumed.
505 */
506 static int
div_output_inbound(int family,struct socket * so,struct mbuf * m,struct sockaddr_in * sin)507 div_output_inbound(int family, struct socket *so, struct mbuf *m,
508 struct sockaddr_in *sin)
509 {
510 const struct ip *ip;
511 struct ifaddr *ifa;
512
513 if (m->m_pkthdr.rcvif == NULL) {
514 /*
515 * No luck with the name, check by IP address.
516 * Clear the port and the ifname to make sure
517 * there are no distractions for ifa_ifwithaddr.
518 */
519
520 /* XXX: broken for IPv6 */
521 bzero(sin->sin_zero, sizeof(sin->sin_zero));
522 sin->sin_port = 0;
523 ifa = ifa_ifwithaddr((struct sockaddr *) sin);
524 if (ifa == NULL) {
525 m_freem(m);
526 return (EADDRNOTAVAIL);
527 }
528 m->m_pkthdr.rcvif = ifa->ifa_ifp;
529 }
530 #ifdef MAC
531 mac_socket_create_mbuf(so, m);
532 #endif
533 /* Send packet to input processing via netisr */
534 switch (family) {
535 case AF_INET:
536 ip = mtod(m, struct ip *);
537 /*
538 * Restore M_BCAST flag when destination address is
539 * broadcast. It is expected by ip_tryforward().
540 */
541 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)))
542 m->m_flags |= M_MCAST;
543 else if (in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif))
544 m->m_flags |= M_BCAST;
545 netisr_queue_src(NETISR_IP, (uintptr_t)so, m);
546 break;
547 #ifdef INET6
548 case AF_INET6:
549 netisr_queue_src(NETISR_IPV6, (uintptr_t)so, m);
550 break;
551 #endif
552 default:
553 m_freem(m);
554 return (EINVAL);
555 }
556
557 return (0);
558 }
559
560 static int
div_attach(struct socket * so,int proto,struct thread * td)561 div_attach(struct socket *so, int proto, struct thread *td)
562 {
563 struct inpcb *inp;
564 int error;
565
566 inp = sotoinpcb(so);
567 KASSERT(inp == NULL, ("div_attach: inp != NULL"));
568 if (td != NULL) {
569 error = priv_check(td, PRIV_NETINET_DIVERT);
570 if (error)
571 return (error);
572 }
573 error = soreserve(so, div_sendspace, div_recvspace);
574 if (error)
575 return error;
576 INP_INFO_WLOCK(&V_divcbinfo);
577 error = in_pcballoc(so, &V_divcbinfo);
578 if (error) {
579 INP_INFO_WUNLOCK(&V_divcbinfo);
580 return error;
581 }
582 inp = (struct inpcb *)so->so_pcb;
583 INP_INFO_WUNLOCK(&V_divcbinfo);
584 inp->inp_ip_p = proto;
585 inp->inp_vflag |= INP_IPV4;
586 inp->inp_flags |= INP_HDRINCL;
587 INP_WUNLOCK(inp);
588 return 0;
589 }
590
591 static void
div_detach(struct socket * so)592 div_detach(struct socket *so)
593 {
594 struct inpcb *inp;
595
596 inp = sotoinpcb(so);
597 KASSERT(inp != NULL, ("div_detach: inp == NULL"));
598 INP_INFO_WLOCK(&V_divcbinfo);
599 INP_WLOCK(inp);
600 in_pcbdetach(inp);
601 in_pcbfree(inp);
602 INP_INFO_WUNLOCK(&V_divcbinfo);
603 }
604
605 static int
div_bind(struct socket * so,struct sockaddr * nam,struct thread * td)606 div_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
607 {
608 struct inpcb *inp;
609 int error;
610
611 inp = sotoinpcb(so);
612 KASSERT(inp != NULL, ("div_bind: inp == NULL"));
613 /* in_pcbbind assumes that nam is a sockaddr_in
614 * and in_pcbbind requires a valid address. Since divert
615 * sockets don't we need to make sure the address is
616 * filled in properly.
617 * XXX -- divert should not be abusing in_pcbind
618 * and should probably have its own family.
619 */
620 if (nam->sa_family != AF_INET)
621 return EAFNOSUPPORT;
622 ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY;
623 INP_INFO_WLOCK(&V_divcbinfo);
624 INP_WLOCK(inp);
625 INP_HASH_WLOCK(&V_divcbinfo);
626 error = in_pcbbind(inp, nam, td->td_ucred);
627 INP_HASH_WUNLOCK(&V_divcbinfo);
628 INP_WUNLOCK(inp);
629 INP_INFO_WUNLOCK(&V_divcbinfo);
630 return error;
631 }
632
633 static int
div_shutdown(struct socket * so)634 div_shutdown(struct socket *so)
635 {
636 struct inpcb *inp;
637
638 inp = sotoinpcb(so);
639 KASSERT(inp != NULL, ("div_shutdown: inp == NULL"));
640 INP_WLOCK(inp);
641 socantsendmore(so);
642 INP_WUNLOCK(inp);
643 return 0;
644 }
645
646 static int
div_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct thread * td)647 div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
648 struct mbuf *control, struct thread *td)
649 {
650
651 /* Packet must have a header (but that's about it) */
652 if (m->m_len < sizeof (struct ip) &&
653 (m = m_pullup(m, sizeof (struct ip))) == NULL) {
654 KMOD_IPSTAT_INC(ips_toosmall);
655 m_freem(m);
656 return EINVAL;
657 }
658
659 /* Send packet */
660 return div_output(so, m, (struct sockaddr_in *)nam, control);
661 }
662
663 static void
div_ctlinput(int cmd,struct sockaddr * sa,void * vip)664 div_ctlinput(int cmd, struct sockaddr *sa, void *vip)
665 {
666 struct in_addr faddr;
667
668 faddr = ((struct sockaddr_in *)sa)->sin_addr;
669 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
670 return;
671 if (PRC_IS_REDIRECT(cmd))
672 return;
673 }
674
675 static int
div_pcblist(SYSCTL_HANDLER_ARGS)676 div_pcblist(SYSCTL_HANDLER_ARGS)
677 {
678 struct xinpgen xig;
679 struct epoch_tracker et;
680 struct inpcb *inp;
681 int error;
682
683 if (req->newptr != 0)
684 return EPERM;
685
686 if (req->oldptr == 0) {
687 int n;
688
689 n = V_divcbinfo.ipi_count;
690 n += imax(n / 8, 10);
691 req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb);
692 return 0;
693 }
694
695 if ((error = sysctl_wire_old_buffer(req, 0)) != 0)
696 return (error);
697
698 bzero(&xig, sizeof(xig));
699 xig.xig_len = sizeof xig;
700 xig.xig_count = V_divcbinfo.ipi_count;
701 xig.xig_gen = V_divcbinfo.ipi_gencnt;
702 xig.xig_sogen = so_gencnt;
703 error = SYSCTL_OUT(req, &xig, sizeof xig);
704 if (error)
705 return error;
706
707 NET_EPOCH_ENTER(et);
708 for (inp = CK_LIST_FIRST(V_divcbinfo.ipi_listhead);
709 inp != NULL;
710 inp = CK_LIST_NEXT(inp, inp_list)) {
711 INP_RLOCK(inp);
712 if (inp->inp_gencnt <= xig.xig_gen) {
713 struct xinpcb xi;
714
715 in_pcbtoxinpcb(inp, &xi);
716 INP_RUNLOCK(inp);
717 error = SYSCTL_OUT(req, &xi, sizeof xi);
718 } else
719 INP_RUNLOCK(inp);
720 }
721 NET_EPOCH_EXIT(et);
722
723 if (!error) {
724 /*
725 * Give the user an updated idea of our state.
726 * If the generation differs from what we told
727 * her before, she knows that something happened
728 * while we were processing this request, and it
729 * might be necessary to retry.
730 */
731 xig.xig_gen = V_divcbinfo.ipi_gencnt;
732 xig.xig_sogen = so_gencnt;
733 xig.xig_count = V_divcbinfo.ipi_count;
734 error = SYSCTL_OUT(req, &xig, sizeof xig);
735 }
736
737 return (error);
738 }
739
740 #ifdef SYSCTL_NODE
741 static SYSCTL_NODE(_net_inet, IPPROTO_DIVERT, divert,
742 CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
743 "IPDIVERT");
744 SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist,
745 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE,
746 NULL, 0, div_pcblist, "S,xinpcb",
747 "List of active divert sockets");
748 #endif
749
750 struct pr_usrreqs div_usrreqs = {
751 .pru_attach = div_attach,
752 .pru_bind = div_bind,
753 .pru_control = in_control,
754 .pru_detach = div_detach,
755 .pru_peeraddr = in_getpeeraddr,
756 .pru_send = div_send,
757 .pru_shutdown = div_shutdown,
758 .pru_sockaddr = in_getsockaddr,
759 .pru_sosetlabel = in_pcbsosetlabel
760 };
761
762 struct protosw div_protosw = {
763 .pr_type = SOCK_RAW,
764 .pr_protocol = IPPROTO_DIVERT,
765 .pr_flags = PR_ATOMIC|PR_ADDR,
766 .pr_input = div_input,
767 .pr_ctlinput = div_ctlinput,
768 .pr_ctloutput = ip_ctloutput,
769 .pr_init = div_init,
770 .pr_usrreqs = &div_usrreqs
771 };
772
773 static int
div_modevent(module_t mod,int type,void * unused)774 div_modevent(module_t mod, int type, void *unused)
775 {
776 int err = 0;
777
778 switch (type) {
779 case MOD_LOAD:
780 /*
781 * Protocol will be initialized by pf_proto_register().
782 * We don't have to register ip_protox because we are not
783 * a true IP protocol that goes over the wire.
784 */
785 err = pf_proto_register(PF_INET, &div_protosw);
786 if (err != 0)
787 return (err);
788 ip_divert_ptr = divert_packet;
789 ip_divert_event_tag = EVENTHANDLER_REGISTER(maxsockets_change,
790 div_zone_change, NULL, EVENTHANDLER_PRI_ANY);
791 break;
792 case MOD_QUIESCE:
793 /*
794 * IPDIVERT may normally not be unloaded because of the
795 * potential race conditions. Tell kldunload we can't be
796 * unloaded unless the unload is forced.
797 */
798 err = EPERM;
799 break;
800 case MOD_UNLOAD:
801 /*
802 * Forced unload.
803 *
804 * Module ipdivert can only be unloaded if no sockets are
805 * connected. Maybe this can be changed later to forcefully
806 * disconnect any open sockets.
807 *
808 * XXXRW: Note that there is a slight race here, as a new
809 * socket open request could be spinning on the lock and then
810 * we destroy the lock.
811 */
812 INP_INFO_WLOCK(&V_divcbinfo);
813 if (V_divcbinfo.ipi_count != 0) {
814 err = EBUSY;
815 INP_INFO_WUNLOCK(&V_divcbinfo);
816 break;
817 }
818 ip_divert_ptr = NULL;
819 err = pf_proto_unregister(PF_INET, IPPROTO_DIVERT, SOCK_RAW);
820 INP_INFO_WUNLOCK(&V_divcbinfo);
821 #ifndef VIMAGE
822 div_destroy(NULL);
823 #endif
824 EVENTHANDLER_DEREGISTER(maxsockets_change, ip_divert_event_tag);
825 break;
826 default:
827 err = EOPNOTSUPP;
828 break;
829 }
830 return err;
831 }
832
833 static moduledata_t ipdivertmod = {
834 "ipdivert",
835 div_modevent,
836 0
837 };
838
839 DECLARE_MODULE(ipdivert, ipdivertmod, SI_SUB_PROTO_FIREWALL, SI_ORDER_ANY);
840 MODULE_DEPEND(ipdivert, ipfw, 3, 3, 3);
841 MODULE_VERSION(ipdivert, 1);
842