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 * 4. 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.
16 *
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
27 * SUCH DAMAGE.
28 *
29 * $FreeBSD: releng/6.4/sys/netinet/ip_divert.c 165644 2006-12-29 19:25:49Z jhb $
30 */
31
32 #if !defined(KLD_MODULE)
33 #include "opt_inet.h"
34 #include "opt_ipfw.h"
35 #include "opt_mac.h"
36 #ifndef INET
37 #error "IPDIVERT requires INET."
38 #endif
39 #ifndef IPFIREWALL
40 #error "IPDIVERT requires IPFIREWALL"
41 #endif
42 #endif
43
44 #include <sys/param.h>
45 #include <sys/kernel.h>
46 #include <sys/lock.h>
47 #include <sys/malloc.h>
48 #include <sys/mac.h>
49 #include <sys/mbuf.h>
50 #include <sys/module.h>
51 #include <sys/kernel.h>
52 #include <sys/proc.h>
53 #include <sys/protosw.h>
54 #include <sys/signalvar.h>
55 #include <sys/socket.h>
56 #include <sys/socketvar.h>
57 #include <sys/sx.h>
58 #include <sys/sysctl.h>
59 #include <sys/systm.h>
60
61 #include <vm/uma.h>
62
63 #include <net/if.h>
64 #include <net/route.h>
65
66 #include <netinet/in.h>
67 #include <netinet/in_pcb.h>
68 #include <netinet/in_systm.h>
69 #include <netinet/in_var.h>
70 #include <netinet/ip.h>
71 #include <netinet/ip_divert.h>
72 #include <netinet/ip_var.h>
73 #include <netinet/ip_fw.h>
74
75 /*
76 * Divert sockets
77 */
78
79 /*
80 * Allocate enough space to hold a full IP packet
81 */
82 #define DIVSNDQ (65536 + 100)
83 #define DIVRCVQ (65536 + 100)
84
85 /*
86 * Divert sockets work in conjunction with ipfw, see the divert(4)
87 * manpage for features.
88 * Internally, packets selected by ipfw in ip_input() or ip_output(),
89 * and never diverted before, are passed to the input queue of the
90 * divert socket with a given 'divert_port' number (as specified in
91 * the matching ipfw rule), and they are tagged with a 16 bit cookie
92 * (representing the rule number of the matching ipfw rule), which
93 * is passed to process reading from the socket.
94 *
95 * Packets written to the divert socket are again tagged with a cookie
96 * (usually the same as above) and a destination address.
97 * If the destination address is INADDR_ANY then the packet is
98 * treated as outgoing and sent to ip_output(), otherwise it is
99 * treated as incoming and sent to ip_input().
100 * In both cases, the packet is tagged with the cookie.
101 *
102 * On reinjection, processing in ip_input() and ip_output()
103 * will be exactly the same as for the original packet, except that
104 * ipfw processing will start at the rule number after the one
105 * written in the cookie (so, tagging a packet with a cookie of 0
106 * will cause it to be effectively considered as a standard packet).
107 */
108
109 /* Internal variables. */
110 static struct inpcbhead divcb;
111 static struct inpcbinfo divcbinfo;
112
113 static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */
114 static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */
115
116 /*
117 * Initialize divert connection block queue.
118 */
119 static void
120 div_zone_change(void *tag)
121 {
122
123 uma_zone_set_max(divcbinfo.ipi_zone, maxsockets);
124 }
125
126 static int
127 div_inpcb_init(void *mem, int size, int flags)
128 {
129 struct inpcb *inp = mem;
130
131 INP_LOCK_INIT(inp, "inp", "divinp");
132 return (0);
133 }
134
135 static void
136 div_inpcb_fini(void *mem, int size)
137 {
138 struct inpcb *inp = mem;
139
140 INP_LOCK_DESTROY(inp);
141 }
142
143 void
144 div_init(void)
145 {
146 INP_INFO_LOCK_INIT(&divcbinfo, "div");
147 LIST_INIT(&divcb);
148 divcbinfo.listhead = &divcb;
149 /*
150 * XXX We don't use the hash list for divert IP, but it's easier
151 * to allocate a one entry hash list than it is to check all
152 * over the place for hashbase == NULL.
153 */
154 divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask);
155 divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask);
156 divcbinfo.ipi_zone = uma_zcreate("divcb", sizeof(struct inpcb),
157 NULL, NULL, div_inpcb_init, div_inpcb_fini, UMA_ALIGN_PTR,
158 UMA_ZONE_NOFREE);
159 uma_zone_set_max(divcbinfo.ipi_zone, maxsockets);
160 EVENTHANDLER_REGISTER(maxsockets_change, div_zone_change,
161 NULL, EVENTHANDLER_PRI_ANY);
162 }
163
164 /*
165 * IPPROTO_DIVERT is not in the real IP protocol number space; this
166 * function should never be called. Just in case, drop any packets.
167 */
168 void
169 div_input(struct mbuf *m, int off)
170 {
171 ipstat.ips_noproto++;
172 m_freem(m);
173 }
174
175 /*
176 * Divert a packet by passing it up to the divert socket at port 'port'.
177 *
178 * Setup generic address and protocol structures for div_input routine,
179 * then pass them along with mbuf chain.
180 */
181 static void
182 divert_packet(struct mbuf *m, int incoming)
183 {
184 struct ip *ip;
185 struct inpcb *inp;
186 struct socket *sa;
187 u_int16_t nport;
188 struct sockaddr_in divsrc;
189 struct m_tag *mtag;
190
191 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL);
192 if (mtag == NULL) {
193 printf("%s: no divert tag\n", __func__);
194 m_freem(m);
195 return;
196 }
197 /* Assure header */
198 if (m->m_len < sizeof(struct ip) &&
199 (m = m_pullup(m, sizeof(struct ip))) == 0)
200 return;
201 ip = mtod(m, struct ip *);
202
203 /* Delayed checksums are currently not compatible with divert. */
204 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
205 ip->ip_len = ntohs(ip->ip_len);
206 in_delayed_cksum(m);
207 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
208 ip->ip_len = htons(ip->ip_len);
209 }
210
211 /*
212 * Record receive interface address, if any.
213 * But only for incoming packets.
214 */
215 bzero(&divsrc, sizeof(divsrc));
216 divsrc.sin_len = sizeof(divsrc);
217 divsrc.sin_family = AF_INET;
218 divsrc.sin_port = divert_cookie(mtag); /* record matching rule */
219 if (incoming) {
220 struct ifaddr *ifa;
221
222 /* Sanity check */
223 M_ASSERTPKTHDR(m);
224
225 /* Find IP address for receive interface */
226 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
227 if (ifa->ifa_addr == NULL)
228 continue;
229 if (ifa->ifa_addr->sa_family != AF_INET)
230 continue;
231 divsrc.sin_addr =
232 ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr;
233 break;
234 }
235 }
236 /*
237 * Record the incoming interface name whenever we have one.
238 */
239 if (m->m_pkthdr.rcvif) {
240 /*
241 * Hide the actual interface name in there in the
242 * sin_zero array. XXX This needs to be moved to a
243 * different sockaddr type for divert, e.g.
244 * sockaddr_div with multiple fields like
245 * sockaddr_dl. Presently we have only 7 bytes
246 * but that will do for now as most interfaces
247 * are 4 or less + 2 or less bytes for unit.
248 * There is probably a faster way of doing this,
249 * possibly taking it from the sockaddr_dl on the iface.
250 * This solves the problem of a P2P link and a LAN interface
251 * having the same address, which can result in the wrong
252 * interface being assigned to the packet when fed back
253 * into the divert socket. Theoretically if the daemon saves
254 * and re-uses the sockaddr_in as suggested in the man pages,
255 * this iface name will come along for the ride.
256 * (see div_output for the other half of this.)
257 */
258 strlcpy(divsrc.sin_zero, m->m_pkthdr.rcvif->if_xname,
259 sizeof(divsrc.sin_zero));
260 }
261
262 /* Put packet on socket queue, if any */
263 sa = NULL;
264 nport = htons((u_int16_t)divert_info(mtag));
265 INP_INFO_RLOCK(&divcbinfo);
266 LIST_FOREACH(inp, &divcb, inp_list) {
267 INP_LOCK(inp);
268 /* XXX why does only one socket match? */
269 if (inp->inp_lport == nport) {
270 sa = inp->inp_socket;
271 SOCKBUF_LOCK(&sa->so_rcv);
272 if (sbappendaddr_locked(&sa->so_rcv,
273 (struct sockaddr *)&divsrc, m,
274 (struct mbuf *)0) == 0) {
275 SOCKBUF_UNLOCK(&sa->so_rcv);
276 sa = NULL; /* force mbuf reclaim below */
277 } else
278 sorwakeup_locked(sa);
279 INP_UNLOCK(inp);
280 break;
281 }
282 INP_UNLOCK(inp);
283 }
284 INP_INFO_RUNLOCK(&divcbinfo);
285 if (sa == NULL) {
286 m_freem(m);
287 ipstat.ips_noproto++;
288 ipstat.ips_delivered--;
289 }
290 }
291
292 /*
293 * Deliver packet back into the IP processing machinery.
294 *
295 * If no address specified, or address is 0.0.0.0, send to ip_output();
296 * otherwise, send to ip_input() and mark as having been received on
297 * the interface with that address.
298 */
299 static int
300 div_output(struct socket *so, struct mbuf *m,
301 struct sockaddr_in *sin, struct mbuf *control)
302 {
303 struct m_tag *mtag;
304 struct divert_tag *dt;
305 int error = 0;
306
307 /*
308 * An mbuf may hasn't come from userland, but we pretend
309 * that it has.
310 */
311 m->m_pkthdr.rcvif = NULL;
312 m->m_nextpkt = NULL;
313
314 if (control)
315 m_freem(control); /* XXX */
316
317 if ((mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL)) == NULL) {
318 mtag = m_tag_get(PACKET_TAG_DIVERT, sizeof(struct divert_tag),
319 M_NOWAIT | M_ZERO);
320 if (mtag == NULL) {
321 error = ENOBUFS;
322 goto cantsend;
323 }
324 dt = (struct divert_tag *)(mtag+1);
325 m_tag_prepend(m, mtag);
326 } else
327 dt = (struct divert_tag *)(mtag+1);
328
329 /* Loopback avoidance and state recovery */
330 if (sin) {
331 int i;
332
333 dt->cookie = sin->sin_port;
334 /*
335 * Find receive interface with the given name, stuffed
336 * (if it exists) in the sin_zero[] field.
337 * The name is user supplied data so don't trust its size
338 * or that it is zero terminated.
339 */
340 for (i = 0; i < sizeof(sin->sin_zero) && sin->sin_zero[i]; i++)
341 ;
342 if ( i > 0 && i < sizeof(sin->sin_zero))
343 m->m_pkthdr.rcvif = ifunit(sin->sin_zero);
344 }
345
346 /* Reinject packet into the system as incoming or outgoing */
347 if (!sin || sin->sin_addr.s_addr == 0) {
348 struct ip *const ip = mtod(m, struct ip *);
349 struct inpcb *inp;
350
351 dt->info |= IP_FW_DIVERT_OUTPUT_FLAG;
352 INP_INFO_WLOCK(&divcbinfo);
353 inp = sotoinpcb(so);
354 INP_LOCK(inp);
355 /*
356 * Don't allow both user specified and setsockopt options,
357 * and don't allow packet length sizes that will crash
358 */
359 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) ||
360 ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) {
361 error = EINVAL;
362 m_freem(m);
363 } else {
364 /* Convert fields to host order for ip_output() */
365 ip->ip_len = ntohs(ip->ip_len);
366 ip->ip_off = ntohs(ip->ip_off);
367
368 /* Send packet to output processing */
369 ipstat.ips_rawout++; /* XXX */
370
371 #ifdef MAC
372 mac_create_mbuf_from_inpcb(inp, m);
373 #endif
374 error = ip_output(m,
375 inp->inp_options, NULL,
376 ((so->so_options & SO_DONTROUTE) ?
377 IP_ROUTETOIF : 0) |
378 IP_ALLOWBROADCAST | IP_RAWOUTPUT,
379 inp->inp_moptions, NULL);
380 }
381 INP_UNLOCK(inp);
382 INP_INFO_WUNLOCK(&divcbinfo);
383 } else {
384 dt->info |= IP_FW_DIVERT_LOOPBACK_FLAG;
385 if (m->m_pkthdr.rcvif == NULL) {
386 /*
387 * No luck with the name, check by IP address.
388 * Clear the port and the ifname to make sure
389 * there are no distractions for ifa_ifwithaddr.
390 */
391 struct ifaddr *ifa;
392
393 bzero(sin->sin_zero, sizeof(sin->sin_zero));
394 sin->sin_port = 0;
395 ifa = ifa_ifwithaddr((struct sockaddr *) sin);
396 if (ifa == NULL) {
397 error = EADDRNOTAVAIL;
398 goto cantsend;
399 }
400 m->m_pkthdr.rcvif = ifa->ifa_ifp;
401 }
402 #ifdef MAC
403 SOCK_LOCK(so);
404 mac_create_mbuf_from_socket(so, m);
405 SOCK_UNLOCK(so);
406 #endif
407 /* Send packet to input processing */
408 ip_input(m);
409 }
410
411 return error;
412
413 cantsend:
414 m_freem(m);
415 return error;
416 }
417
418 static int
419 div_attach(struct socket *so, int proto, struct thread *td)
420 {
421 struct inpcb *inp;
422 int error;
423
424 INP_INFO_WLOCK(&divcbinfo);
425 inp = sotoinpcb(so);
426 if (inp != 0) {
427 INP_INFO_WUNLOCK(&divcbinfo);
428 return EINVAL;
429 }
430 if (td && (error = suser(td)) != 0) {
431 INP_INFO_WUNLOCK(&divcbinfo);
432 return error;
433 }
434 error = soreserve(so, div_sendspace, div_recvspace);
435 if (error) {
436 INP_INFO_WUNLOCK(&divcbinfo);
437 return error;
438 }
439 error = in_pcballoc(so, &divcbinfo);
440 if (error) {
441 INP_INFO_WUNLOCK(&divcbinfo);
442 return error;
443 }
444 inp = (struct inpcb *)so->so_pcb;
445 INP_INFO_WUNLOCK(&divcbinfo);
446 inp->inp_ip_p = proto;
447 inp->inp_vflag |= INP_IPV4;
448 inp->inp_flags |= INP_HDRINCL;
449 INP_UNLOCK(inp);
450 return 0;
451 }
452
453 static int
454 div_detach(struct socket *so)
455 {
456 struct inpcb *inp;
457
458 INP_INFO_WLOCK(&divcbinfo);
459 inp = sotoinpcb(so);
460 if (inp == 0) {
461 INP_INFO_WUNLOCK(&divcbinfo);
462 return EINVAL;
463 }
464 INP_LOCK(inp);
465 in_pcbdetach(inp);
466 INP_INFO_WUNLOCK(&divcbinfo);
467 return 0;
468 }
469
470 static int
471 div_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
472 {
473 struct inpcb *inp;
474 int error;
475
476 INP_INFO_WLOCK(&divcbinfo);
477 inp = sotoinpcb(so);
478 if (inp == 0) {
479 INP_INFO_WUNLOCK(&divcbinfo);
480 return EINVAL;
481 }
482 /* in_pcbbind assumes that nam is a sockaddr_in
483 * and in_pcbbind requires a valid address. Since divert
484 * sockets don't we need to make sure the address is
485 * filled in properly.
486 * XXX -- divert should not be abusing in_pcbind
487 * and should probably have its own family.
488 */
489 if (nam->sa_family != AF_INET)
490 error = EAFNOSUPPORT;
491 else {
492 ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY;
493 INP_LOCK(inp);
494 error = in_pcbbind(inp, nam, td->td_ucred);
495 INP_UNLOCK(inp);
496 }
497 INP_INFO_WUNLOCK(&divcbinfo);
498 return error;
499 }
500
501 static int
502 div_shutdown(struct socket *so)
503 {
504 struct inpcb *inp;
505
506 INP_INFO_RLOCK(&divcbinfo);
507 inp = sotoinpcb(so);
508 if (inp == 0) {
509 INP_INFO_RUNLOCK(&divcbinfo);
510 return EINVAL;
511 }
512 INP_LOCK(inp);
513 INP_INFO_RUNLOCK(&divcbinfo);
514 socantsendmore(so);
515 INP_UNLOCK(inp);
516 return 0;
517 }
518
519 static int
520 div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
521 struct mbuf *control, struct thread *td)
522 {
523 /* Packet must have a header (but that's about it) */
524 if (m->m_len < sizeof (struct ip) &&
525 (m = m_pullup(m, sizeof (struct ip))) == 0) {
526 ipstat.ips_toosmall++;
527 m_freem(m);
528 return EINVAL;
529 }
530
531 /* Send packet */
532 return div_output(so, m, (struct sockaddr_in *)nam, control);
533 }
534
535 void
536 div_ctlinput(int cmd, struct sockaddr *sa, void *vip)
537 {
538 struct in_addr faddr;
539
540 faddr = ((struct sockaddr_in *)sa)->sin_addr;
541 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
542 return;
543 if (PRC_IS_REDIRECT(cmd))
544 return;
545 }
546
547 static int
548 div_pcblist(SYSCTL_HANDLER_ARGS)
549 {
550 int error, i, n;
551 struct inpcb *inp, **inp_list;
552 inp_gen_t gencnt;
553 struct xinpgen xig;
554
555 /*
556 * The process of preparing the TCB list is too time-consuming and
557 * resource-intensive to repeat twice on every request.
558 */
559 if (req->oldptr == 0) {
560 n = divcbinfo.ipi_count;
561 req->oldidx = 2 * (sizeof xig)
562 + (n + n/8) * sizeof(struct xinpcb);
563 return 0;
564 }
565
566 if (req->newptr != 0)
567 return EPERM;
568
569 /*
570 * OK, now we're committed to doing something.
571 */
572 INP_INFO_RLOCK(&divcbinfo);
573 gencnt = divcbinfo.ipi_gencnt;
574 n = divcbinfo.ipi_count;
575 INP_INFO_RUNLOCK(&divcbinfo);
576
577 error = sysctl_wire_old_buffer(req,
578 2 * sizeof(xig) + n*sizeof(struct xinpcb));
579 if (error != 0)
580 return (error);
581
582 xig.xig_len = sizeof xig;
583 xig.xig_count = n;
584 xig.xig_gen = gencnt;
585 xig.xig_sogen = so_gencnt;
586 error = SYSCTL_OUT(req, &xig, sizeof xig);
587 if (error)
588 return error;
589
590 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
591 if (inp_list == 0)
592 return ENOMEM;
593
594 INP_INFO_RLOCK(&divcbinfo);
595 for (inp = LIST_FIRST(divcbinfo.listhead), i = 0; inp && i < n;
596 inp = LIST_NEXT(inp, inp_list)) {
597 INP_LOCK(inp);
598 if (inp->inp_gencnt <= gencnt &&
599 cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0)
600 inp_list[i++] = inp;
601 INP_UNLOCK(inp);
602 }
603 INP_INFO_RUNLOCK(&divcbinfo);
604 n = i;
605
606 error = 0;
607 for (i = 0; i < n; i++) {
608 inp = inp_list[i];
609 INP_LOCK(inp);
610 if (inp->inp_gencnt <= gencnt) {
611 struct xinpcb xi;
612 bzero(&xi, sizeof(xi));
613 xi.xi_len = sizeof xi;
614 /* XXX should avoid extra copy */
615 bcopy(inp, &xi.xi_inp, sizeof *inp);
616 if (inp->inp_socket)
617 sotoxsocket(inp->inp_socket, &xi.xi_socket);
618 INP_UNLOCK(inp);
619 error = SYSCTL_OUT(req, &xi, sizeof xi);
620 } else
621 INP_UNLOCK(inp);
622 }
623 if (!error) {
624 /*
625 * Give the user an updated idea of our state.
626 * If the generation differs from what we told
627 * her before, she knows that something happened
628 * while we were processing this request, and it
629 * might be necessary to retry.
630 */
631 INP_INFO_RLOCK(&divcbinfo);
632 xig.xig_gen = divcbinfo.ipi_gencnt;
633 xig.xig_sogen = so_gencnt;
634 xig.xig_count = divcbinfo.ipi_count;
635 INP_INFO_RUNLOCK(&divcbinfo);
636 error = SYSCTL_OUT(req, &xig, sizeof xig);
637 }
638 free(inp_list, M_TEMP);
639 return error;
640 }
641
642 /*
643 * This is the wrapper function for in_setsockaddr. We just pass down
644 * the pcbinfo for in_setpeeraddr to lock.
645 */
646 static int
647 div_sockaddr(struct socket *so, struct sockaddr **nam)
648 {
649 return (in_setsockaddr(so, nam, &divcbinfo));
650 }
651
652 /*
653 * This is the wrapper function for in_setpeeraddr. We just pass down
654 * the pcbinfo for in_setpeeraddr to lock.
655 */
656 static int
657 div_peeraddr(struct socket *so, struct sockaddr **nam)
658 {
659 return (in_setpeeraddr(so, nam, &divcbinfo));
660 }
661
662 #ifdef SYSCTL_NODE
663 SYSCTL_NODE(_net_inet, IPPROTO_DIVERT, divert, CTLFLAG_RW, 0, "IPDIVERT");
664 SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLFLAG_RD, 0, 0,
665 div_pcblist, "S,xinpcb", "List of active divert sockets");
666 #endif
667
668 struct pr_usrreqs div_usrreqs = {
669 .pru_attach = div_attach,
670 .pru_bind = div_bind,
671 .pru_control = in_control,
672 .pru_detach = div_detach,
673 .pru_peeraddr = div_peeraddr,
674 .pru_send = div_send,
675 .pru_shutdown = div_shutdown,
676 .pru_sockaddr = div_sockaddr,
677 .pru_sosetlabel = in_pcbsosetlabel
678 };
679
680 struct protosw div_protosw = {
681 .pr_type = SOCK_RAW,
682 .pr_protocol = IPPROTO_DIVERT,
683 .pr_flags = PR_ATOMIC|PR_ADDR,
684 .pr_input = div_input,
685 .pr_ctlinput = div_ctlinput,
686 .pr_ctloutput = ip_ctloutput,
687 .pr_init = div_init,
688 .pr_usrreqs = &div_usrreqs
689 };
690
691 static int
692 div_modevent(module_t mod, int type, void *unused)
693 {
694 int err = 0;
695 int n;
696
697 switch (type) {
698 case MOD_LOAD:
699 /*
700 * Protocol will be initialized by pf_proto_register().
701 * We don't have to register ip_protox because we are not
702 * a true IP protocol that goes over the wire.
703 */
704 err = pf_proto_register(PF_INET, &div_protosw);
705 ip_divert_ptr = divert_packet;
706 break;
707 case MOD_QUIESCE:
708 /*
709 * IPDIVERT may normally not be unloaded because of the
710 * potential race conditions. Tell kldunload we can't be
711 * unloaded unless the unload is forced.
712 */
713 err = EPERM;
714 break;
715 case MOD_UNLOAD:
716 /*
717 * Forced unload.
718 *
719 * Module ipdivert can only be unloaded if no sockets are
720 * connected. Maybe this can be changed later to forcefully
721 * disconnect any open sockets.
722 *
723 * XXXRW: Note that there is a slight race here, as a new
724 * socket open request could be spinning on the lock and then
725 * we destroy the lock.
726 */
727 INP_INFO_WLOCK(&divcbinfo);
728 n = divcbinfo.ipi_count;
729 if (n != 0) {
730 err = EBUSY;
731 INP_INFO_WUNLOCK(&divcbinfo);
732 break;
733 }
734 ip_divert_ptr = NULL;
735 err = pf_proto_unregister(PF_INET, IPPROTO_DIVERT, SOCK_RAW);
736 INP_INFO_WUNLOCK(&divcbinfo);
737 INP_INFO_LOCK_DESTROY(&divcbinfo);
738 uma_zdestroy(divcbinfo.ipi_zone);
739 break;
740 default:
741 err = EOPNOTSUPP;
742 break;
743 }
744 return err;
745 }
746
747 static moduledata_t ipdivertmod = {
748 "ipdivert",
749 div_modevent,
750 0
751 };
752
753 DECLARE_MODULE(ipdivert, ipdivertmod, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY);
754 MODULE_DEPEND(dummynet, ipfw, 2, 2, 2);
755 MODULE_VERSION(ipdivert, 1);
Cache object: 28b1d04c183bc635c5763c42d889efdd
|