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: src/sys/netinet/ip_divert.c,v 1.98.2.6 2005/05/27 12:32:29 glebius Exp $
30 */
31
32 #include "opt_inet.h"
33 #include "opt_ipfw.h"
34 #include "opt_ipdivert.h"
35 #include "opt_ipsec.h"
36 #include "opt_mac.h"
37
38 #ifndef INET
39 #error "IPDIVERT requires INET."
40 #endif
41
42 #include <sys/param.h>
43 #include <sys/kernel.h>
44 #include <sys/lock.h>
45 #include <sys/malloc.h>
46 #include <sys/mac.h>
47 #include <sys/mbuf.h>
48 #include <sys/proc.h>
49 #include <sys/protosw.h>
50 #include <sys/signalvar.h>
51 #include <sys/socket.h>
52 #include <sys/socketvar.h>
53 #include <sys/sx.h>
54 #include <sys/sysctl.h>
55 #include <sys/systm.h>
56
57 #include <vm/uma.h>
58
59 #include <net/if.h>
60 #include <net/route.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_divert.h>
68 #include <netinet/ip_var.h>
69 #include <netinet/ip_fw.h>
70
71 /*
72 * Divert sockets
73 */
74
75 /*
76 * Allocate enough space to hold a full IP packet
77 */
78 #define DIVSNDQ (65536 + 100)
79 #define DIVRCVQ (65536 + 100)
80
81 /*
82 * Divert sockets work in conjunction with ipfw, see the divert(4)
83 * manpage for features.
84 * Internally, packets selected by ipfw in ip_input() or ip_output(),
85 * and never diverted before, are passed to the input queue of the
86 * divert socket with a given 'divert_port' number (as specified in
87 * the matching ipfw rule), and they are tagged with a 16 bit cookie
88 * (representing the rule number of the matching ipfw rule), which
89 * is passed to process reading from the socket.
90 *
91 * Packets written to the divert socket are again tagged with a cookie
92 * (usually the same as above) and a destination address.
93 * If the destination address is INADDR_ANY then the packet is
94 * treated as outgoing and sent to ip_output(), otherwise it is
95 * treated as incoming and sent to ip_input().
96 * In both cases, the packet is tagged with the cookie.
97 *
98 * On reinjection, processing in ip_input() and ip_output()
99 * will be exactly the same as for the original packet, except that
100 * ipfw processing will start at the rule number after the one
101 * written in the cookie (so, tagging a packet with a cookie of 0
102 * will cause it to be effectively considered as a standard packet).
103 */
104
105 /* Internal variables */
106 static struct inpcbhead divcb;
107 static struct inpcbinfo divcbinfo;
108
109 static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */
110 static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */
111
112 /*
113 * Initialize divert connection block queue.
114 */
115 void
116 div_init(void)
117 {
118 INP_INFO_LOCK_INIT(&divcbinfo, "div");
119 LIST_INIT(&divcb);
120 divcbinfo.listhead = &divcb;
121 /*
122 * XXX We don't use the hash list for divert IP, but it's easier
123 * to allocate a one entry hash list than it is to check all
124 * over the place for hashbase == NULL.
125 */
126 divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask);
127 divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask);
128 divcbinfo.ipi_zone = uma_zcreate("divcb", sizeof(struct inpcb),
129 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
130 uma_zone_set_max(divcbinfo.ipi_zone, maxsockets);
131 }
132
133 /*
134 * IPPROTO_DIVERT is not in the real IP protocol number space; this
135 * function should never be called. Just in case, drop any packets.
136 */
137 void
138 div_input(struct mbuf *m, int off)
139 {
140 ipstat.ips_noproto++;
141 m_freem(m);
142 }
143
144 /*
145 * Divert a packet by passing it up to the divert socket at port 'port'.
146 *
147 * Setup generic address and protocol structures for div_input routine,
148 * then pass them along with mbuf chain.
149 */
150 void
151 divert_packet(struct mbuf *m, int incoming)
152 {
153 struct ip *ip;
154 struct inpcb *inp;
155 struct socket *sa;
156 u_int16_t nport;
157 struct sockaddr_in divsrc;
158 struct m_tag *mtag;
159
160 mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL);
161 if (mtag == NULL) {
162 printf("%s: no divert tag\n", __func__);
163 m_freem(m);
164 return;
165 }
166 /* Assure header */
167 if (m->m_len < sizeof(struct ip) &&
168 (m = m_pullup(m, sizeof(struct ip))) == 0)
169 return;
170 ip = mtod(m, struct ip *);
171
172 /* Delayed checksums are currently not compatible with divert. */
173 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
174 ip->ip_len = ntohs(ip->ip_len);
175 in_delayed_cksum(m);
176 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
177 ip->ip_len = htons(ip->ip_len);
178 }
179
180 /*
181 * Record receive interface address, if any.
182 * But only for incoming packets.
183 */
184 bzero(&divsrc, sizeof(divsrc));
185 divsrc.sin_len = sizeof(divsrc);
186 divsrc.sin_family = AF_INET;
187 divsrc.sin_port = divert_cookie(mtag); /* record matching rule */
188 if (incoming) {
189 struct ifaddr *ifa;
190
191 /* Sanity check */
192 M_ASSERTPKTHDR(m);
193
194 /* Find IP address for receive interface */
195 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
196 if (ifa->ifa_addr == NULL)
197 continue;
198 if (ifa->ifa_addr->sa_family != AF_INET)
199 continue;
200 divsrc.sin_addr =
201 ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr;
202 break;
203 }
204 }
205 /*
206 * Record the incoming interface name whenever we have one.
207 */
208 if (m->m_pkthdr.rcvif) {
209 /*
210 * Hide the actual interface name in there in the
211 * sin_zero array. XXX This needs to be moved to a
212 * different sockaddr type for divert, e.g.
213 * sockaddr_div with multiple fields like
214 * sockaddr_dl. Presently we have only 7 bytes
215 * but that will do for now as most interfaces
216 * are 4 or less + 2 or less bytes for unit.
217 * There is probably a faster way of doing this,
218 * possibly taking it from the sockaddr_dl on the iface.
219 * This solves the problem of a P2P link and a LAN interface
220 * having the same address, which can result in the wrong
221 * interface being assigned to the packet when fed back
222 * into the divert socket. Theoretically if the daemon saves
223 * and re-uses the sockaddr_in as suggested in the man pages,
224 * this iface name will come along for the ride.
225 * (see div_output for the other half of this.)
226 */
227 strlcpy(divsrc.sin_zero, m->m_pkthdr.rcvif->if_xname,
228 sizeof(divsrc.sin_zero));
229 }
230
231 /* Put packet on socket queue, if any */
232 sa = NULL;
233 nport = htons((u_int16_t)divert_info(mtag));
234 INP_INFO_RLOCK(&divcbinfo);
235 LIST_FOREACH(inp, &divcb, inp_list) {
236 INP_LOCK(inp);
237 /* XXX why does only one socket match? */
238 if (inp->inp_lport == nport) {
239 sa = inp->inp_socket;
240 SOCKBUF_LOCK(&sa->so_rcv);
241 if (sbappendaddr_locked(&sa->so_rcv,
242 (struct sockaddr *)&divsrc, m,
243 (struct mbuf *)0) == 0) {
244 SOCKBUF_UNLOCK(&sa->so_rcv);
245 sa = NULL; /* force mbuf reclaim below */
246 } else
247 sorwakeup_locked(sa);
248 INP_UNLOCK(inp);
249 break;
250 }
251 INP_UNLOCK(inp);
252 }
253 INP_INFO_RUNLOCK(&divcbinfo);
254 if (sa == NULL) {
255 m_freem(m);
256 ipstat.ips_noproto++;
257 ipstat.ips_delivered--;
258 }
259 }
260
261 /*
262 * Deliver packet back into the IP processing machinery.
263 *
264 * If no address specified, or address is 0.0.0.0, send to ip_output();
265 * otherwise, send to ip_input() and mark as having been received on
266 * the interface with that address.
267 */
268 static int
269 div_output(struct socket *so, struct mbuf *m,
270 struct sockaddr_in *sin, struct mbuf *control)
271 {
272 struct m_tag *mtag;
273 struct divert_tag *dt;
274 int error = 0;
275
276 /*
277 * An mbuf may hasn't come from userland, but we pretend
278 * that it has.
279 */
280 m->m_pkthdr.rcvif = NULL;
281 m->m_nextpkt = NULL;
282
283 if (control)
284 m_freem(control); /* XXX */
285
286 if ((mtag = m_tag_find(m, PACKET_TAG_DIVERT, NULL)) == NULL) {
287 mtag = m_tag_get(PACKET_TAG_DIVERT, sizeof(struct divert_tag),
288 M_NOWAIT | M_ZERO);
289 if (mtag == NULL) {
290 error = ENOBUFS;
291 goto cantsend;
292 }
293 dt = (struct divert_tag *)(mtag+1);
294 m_tag_prepend(m, mtag);
295 } else
296 dt = (struct divert_tag *)(mtag+1);
297
298 /* Loopback avoidance and state recovery */
299 if (sin) {
300 int i;
301
302 dt->cookie = sin->sin_port;
303 /*
304 * Find receive interface with the given name, stuffed
305 * (if it exists) in the sin_zero[] field.
306 * The name is user supplied data so don't trust its size
307 * or that it is zero terminated.
308 */
309 for (i = 0; i < sizeof(sin->sin_zero) && sin->sin_zero[i]; i++)
310 ;
311 if ( i > 0 && i < sizeof(sin->sin_zero))
312 m->m_pkthdr.rcvif = ifunit(sin->sin_zero);
313 }
314
315 /* Reinject packet into the system as incoming or outgoing */
316 if (!sin || sin->sin_addr.s_addr == 0) {
317 struct ip *const ip = mtod(m, struct ip *);
318 struct inpcb *inp;
319
320 dt->info |= IP_FW_DIVERT_OUTPUT_FLAG;
321 INP_INFO_WLOCK(&divcbinfo);
322 inp = sotoinpcb(so);
323 INP_LOCK(inp);
324 /*
325 * Don't allow both user specified and setsockopt options,
326 * and don't allow packet length sizes that will crash
327 */
328 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) ||
329 ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) {
330 error = EINVAL;
331 m_freem(m);
332 } else {
333 /* Convert fields to host order for ip_output() */
334 ip->ip_len = ntohs(ip->ip_len);
335 ip->ip_off = ntohs(ip->ip_off);
336
337 /* Send packet to output processing */
338 ipstat.ips_rawout++; /* XXX */
339
340 #ifdef MAC
341 mac_create_mbuf_from_inpcb(inp, m);
342 #endif
343 error = ip_output(m,
344 inp->inp_options, NULL,
345 (so->so_options & SO_DONTROUTE) |
346 IP_ALLOWBROADCAST | IP_RAWOUTPUT,
347 inp->inp_moptions, NULL);
348 }
349 INP_UNLOCK(inp);
350 INP_INFO_WUNLOCK(&divcbinfo);
351 } else {
352 dt->info |= IP_FW_DIVERT_LOOPBACK_FLAG;
353 if (m->m_pkthdr.rcvif == NULL) {
354 /*
355 * No luck with the name, check by IP address.
356 * Clear the port and the ifname to make sure
357 * there are no distractions for ifa_ifwithaddr.
358 */
359 struct ifaddr *ifa;
360
361 bzero(sin->sin_zero, sizeof(sin->sin_zero));
362 sin->sin_port = 0;
363 ifa = ifa_ifwithaddr((struct sockaddr *) sin);
364 if (ifa == NULL) {
365 error = EADDRNOTAVAIL;
366 goto cantsend;
367 }
368 m->m_pkthdr.rcvif = ifa->ifa_ifp;
369 }
370 #ifdef MAC
371 SOCK_LOCK(so);
372 mac_create_mbuf_from_socket(so, m);
373 SOCK_UNLOCK(so);
374 #endif
375 /* Send packet to input processing */
376 ip_input(m);
377 }
378
379 return error;
380
381 cantsend:
382 m_freem(m);
383 return error;
384 }
385
386 static int
387 div_attach(struct socket *so, int proto, struct thread *td)
388 {
389 struct inpcb *inp;
390 int error;
391
392 INP_INFO_WLOCK(&divcbinfo);
393 inp = sotoinpcb(so);
394 if (inp != 0) {
395 INP_INFO_WUNLOCK(&divcbinfo);
396 return EINVAL;
397 }
398 if (td && (error = suser(td)) != 0) {
399 INP_INFO_WUNLOCK(&divcbinfo);
400 return error;
401 }
402 error = soreserve(so, div_sendspace, div_recvspace);
403 if (error) {
404 INP_INFO_WUNLOCK(&divcbinfo);
405 return error;
406 }
407 error = in_pcballoc(so, &divcbinfo, "divinp");
408 if (error) {
409 INP_INFO_WUNLOCK(&divcbinfo);
410 return error;
411 }
412 inp = (struct inpcb *)so->so_pcb;
413 INP_LOCK(inp);
414 INP_INFO_WUNLOCK(&divcbinfo);
415 inp->inp_ip_p = proto;
416 inp->inp_vflag |= INP_IPV4;
417 inp->inp_flags |= INP_HDRINCL;
418 INP_UNLOCK(inp);
419 return 0;
420 }
421
422 static int
423 div_detach(struct socket *so)
424 {
425 struct inpcb *inp;
426
427 INP_INFO_WLOCK(&divcbinfo);
428 inp = sotoinpcb(so);
429 if (inp == 0) {
430 INP_INFO_WUNLOCK(&divcbinfo);
431 return EINVAL;
432 }
433 INP_LOCK(inp);
434 in_pcbdetach(inp);
435 INP_INFO_WUNLOCK(&divcbinfo);
436 return 0;
437 }
438
439 static int
440 div_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
441 {
442 struct inpcb *inp;
443 int error;
444
445 INP_INFO_WLOCK(&divcbinfo);
446 inp = sotoinpcb(so);
447 if (inp == 0) {
448 INP_INFO_WUNLOCK(&divcbinfo);
449 return EINVAL;
450 }
451 /* in_pcbbind assumes that nam is a sockaddr_in
452 * and in_pcbbind requires a valid address. Since divert
453 * sockets don't we need to make sure the address is
454 * filled in properly.
455 * XXX -- divert should not be abusing in_pcbind
456 * and should probably have its own family.
457 */
458 if (nam->sa_family != AF_INET)
459 error = EAFNOSUPPORT;
460 else {
461 ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY;
462 INP_LOCK(inp);
463 error = in_pcbbind(inp, nam, td->td_ucred);
464 INP_UNLOCK(inp);
465 }
466 INP_INFO_WUNLOCK(&divcbinfo);
467 return error;
468 }
469
470 static int
471 div_shutdown(struct socket *so)
472 {
473 struct inpcb *inp;
474
475 INP_INFO_RLOCK(&divcbinfo);
476 inp = sotoinpcb(so);
477 if (inp == 0) {
478 INP_INFO_RUNLOCK(&divcbinfo);
479 return EINVAL;
480 }
481 INP_LOCK(inp);
482 INP_INFO_RUNLOCK(&divcbinfo);
483 socantsendmore(so);
484 INP_UNLOCK(inp);
485 return 0;
486 }
487
488 static int
489 div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
490 struct mbuf *control, struct thread *td)
491 {
492 /* Packet must have a header (but that's about it) */
493 if (m->m_len < sizeof (struct ip) &&
494 (m = m_pullup(m, sizeof (struct ip))) == 0) {
495 ipstat.ips_toosmall++;
496 m_freem(m);
497 return EINVAL;
498 }
499
500 /* Send packet */
501 return div_output(so, m, (struct sockaddr_in *)nam, control);
502 }
503
504 void
505 div_ctlinput(int cmd, struct sockaddr *sa, void *vip)
506 {
507 struct in_addr faddr;
508
509 faddr = ((struct sockaddr_in *)sa)->sin_addr;
510 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
511 return;
512 if (PRC_IS_REDIRECT(cmd))
513 return;
514 }
515
516 static int
517 div_pcblist(SYSCTL_HANDLER_ARGS)
518 {
519 int error, i, n;
520 struct inpcb *inp, **inp_list;
521 inp_gen_t gencnt;
522 struct xinpgen xig;
523
524 /*
525 * The process of preparing the TCB list is too time-consuming and
526 * resource-intensive to repeat twice on every request.
527 */
528 if (req->oldptr == 0) {
529 n = divcbinfo.ipi_count;
530 req->oldidx = 2 * (sizeof xig)
531 + (n + n/8) * sizeof(struct xinpcb);
532 return 0;
533 }
534
535 if (req->newptr != 0)
536 return EPERM;
537
538 /*
539 * OK, now we're committed to doing something.
540 */
541 INP_INFO_RLOCK(&divcbinfo);
542 gencnt = divcbinfo.ipi_gencnt;
543 n = divcbinfo.ipi_count;
544 INP_INFO_RUNLOCK(&divcbinfo);
545
546 error = sysctl_wire_old_buffer(req,
547 2 * sizeof(xig) + n*sizeof(struct xinpcb));
548 if (error != 0)
549 return (error);
550
551 xig.xig_len = sizeof xig;
552 xig.xig_count = n;
553 xig.xig_gen = gencnt;
554 xig.xig_sogen = so_gencnt;
555 error = SYSCTL_OUT(req, &xig, sizeof xig);
556 if (error)
557 return error;
558
559 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
560 if (inp_list == 0)
561 return ENOMEM;
562
563 INP_INFO_RLOCK(&divcbinfo);
564 for (inp = LIST_FIRST(divcbinfo.listhead), i = 0; inp && i < n;
565 inp = LIST_NEXT(inp, inp_list)) {
566 INP_LOCK(inp);
567 if (inp->inp_gencnt <= gencnt &&
568 cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0)
569 inp_list[i++] = inp;
570 INP_UNLOCK(inp);
571 }
572 INP_INFO_RUNLOCK(&divcbinfo);
573 n = i;
574
575 error = 0;
576 for (i = 0; i < n; i++) {
577 inp = inp_list[i];
578 if (inp->inp_gencnt <= gencnt) {
579 struct xinpcb xi;
580 bzero(&xi, sizeof(xi));
581 xi.xi_len = sizeof xi;
582 /* XXX should avoid extra copy */
583 bcopy(inp, &xi.xi_inp, sizeof *inp);
584 if (inp->inp_socket)
585 sotoxsocket(inp->inp_socket, &xi.xi_socket);
586 error = SYSCTL_OUT(req, &xi, sizeof xi);
587 }
588 }
589 if (!error) {
590 /*
591 * Give the user an updated idea of our state.
592 * If the generation differs from what we told
593 * her before, she knows that something happened
594 * while we were processing this request, and it
595 * might be necessary to retry.
596 */
597 INP_INFO_RLOCK(&divcbinfo);
598 xig.xig_gen = divcbinfo.ipi_gencnt;
599 xig.xig_sogen = so_gencnt;
600 xig.xig_count = divcbinfo.ipi_count;
601 INP_INFO_RUNLOCK(&divcbinfo);
602 error = SYSCTL_OUT(req, &xig, sizeof xig);
603 }
604 free(inp_list, M_TEMP);
605 return error;
606 }
607
608 /*
609 * This is the wrapper function for in_setsockaddr. We just pass down
610 * the pcbinfo for in_setpeeraddr to lock.
611 */
612 static int
613 div_sockaddr(struct socket *so, struct sockaddr **nam)
614 {
615 return (in_setsockaddr(so, nam, &divcbinfo));
616 }
617
618 /*
619 * This is the wrapper function for in_setpeeraddr. We just pass down
620 * the pcbinfo for in_setpeeraddr to lock.
621 */
622 static int
623 div_peeraddr(struct socket *so, struct sockaddr **nam)
624 {
625 return (in_setpeeraddr(so, nam, &divcbinfo));
626 }
627
628
629 SYSCTL_DECL(_net_inet_divert);
630 SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLFLAG_RD, 0, 0,
631 div_pcblist, "S,xinpcb", "List of active divert sockets");
632
633 struct pr_usrreqs div_usrreqs = {
634 NULL, pru_accept_notsupp, div_attach, div_bind,
635 pru_connect_notsupp, pru_connect2_notsupp, in_control, div_detach,
636 NULL, pru_listen_notsupp, div_peeraddr, pru_rcvd_notsupp,
637 pru_rcvoob_notsupp, div_send, pru_sense_null, div_shutdown,
638 div_sockaddr, sosend, soreceive, sopoll, in_pcbsosetlabel
639 };
Cache object: 742292622a2317f97dec958cff52d258
|