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: head/sys/netinet/ip_divert.c 95759 2002-04-30 01:54:54Z tanimura $
34 */
35
36#include "opt_inet.h"
37#include "opt_ipfw.h"
38#include "opt_ipdivert.h"
39#include "opt_ipsec.h"
40
41#ifndef INET
42#error "IPDIVERT requires INET."
43#endif
44
45#include <sys/param.h>
46#include <sys/kernel.h>
47#include <sys/lock.h>
48#include <sys/malloc.h>
49#include <sys/mbuf.h>
50#include <sys/proc.h>
51#include <sys/protosw.h>
52#include <sys/signalvar.h>
53#include <sys/socket.h>
54#include <sys/socketvar.h>
55#include <sys/sx.h>
56#include <sys/sysctl.h>
57#include <sys/systm.h>
58
59#include <vm/uma.h>
60
61#include <net/if.h>
62#include <net/route.h>
63
64#include <netinet/in.h>
65#include <netinet/in_pcb.h>
66#include <netinet/in_systm.h>
67#include <netinet/in_var.h>
68#include <netinet/ip.h>
69#include <netinet/ip_var.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 * A 16 bit cookie is passed to and from the user process.
83 * The user process can send it back to help the caller know
84 * something about where the packet originally came from.
85 *
86 * In the case of ipfw, then the cookie is the rule that sent
87 * us here. On reinjection is is the rule after which processing
88 * should continue. Leaving it the same will make processing start
89 * at the rule number after that which sent it here. Setting it to
90 * 0 will restart processing at the beginning.
91 *
92 * For divert_packet(), ip_divert_cookie is an input value only.
93 * For div_output(), ip_divert_cookie is an output value only.
94 */
95u_int16_t ip_divert_cookie;
96
97/* Internal variables */
98static struct inpcbhead divcb;
99static struct inpcbinfo divcbinfo;
100
101static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */
102static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */
103
104/* Optimization: have this preinitialized */
105static struct sockaddr_in divsrc = { sizeof(divsrc), AF_INET };
106
107/* Internal functions */
108static int div_output(struct socket *so,
109 struct mbuf *m, struct sockaddr *addr, struct mbuf *control);
110
111/*
112 * Initialize divert connection block queue.
113 */
114void
115div_init(void)
116{
117 LIST_INIT(&divcb);
118 divcbinfo.listhead = &divcb;
119 /*
120 * XXX We don't use the hash list for divert IP, but it's easier
121 * to allocate a one entry hash list than it is to check all
122 * over the place for hashbase == NULL.
123 */
124 divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask);
125 divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask);
126 divcbinfo.ipi_zone = uma_zcreate("divcb", sizeof(struct inpcb),
127 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
128 uma_zone_set_max(divcbinfo.ipi_zone, maxsockets);
129}
130
131/*
132 * IPPROTO_DIVERT is not a real IP protocol; don't allow any packets
133 * with that protocol number to enter the system from the outside.
134 */
135void
136div_input(struct mbuf *m, int off)
137{
138 ipstat.ips_noproto++;
139 m_freem(m);
140}
141
142/*
143 * Divert a packet by passing it up to the divert socket at port 'port'.
144 *
145 * Setup generic address and protocol structures for div_input routine,
146 * then pass them along with mbuf chain.
147 */
148void
149divert_packet(struct mbuf *m, int incoming, int port)
150{
151 struct ip *ip;
152 struct inpcb *inp;
153 struct socket *sa;
154 u_int16_t nport;
155
156 /* Sanity check */
157 KASSERT(port != 0, ("%s: port=0", __func__));
158
159 /* Record and reset divert cookie */
160 divsrc.sin_port = ip_divert_cookie;
161 ip_divert_cookie = 0;
162
163 /* Assure header */
164 if (m->m_len < sizeof(struct ip) &&
165 (m = m_pullup(m, sizeof(struct ip))) == 0) {
166 return;
167 }
168 ip = mtod(m, struct ip *);
169
170 /*
171 * Record receive interface address, if any.
172 * But only for incoming packets.
173 */
174 divsrc.sin_addr.s_addr = 0;
175 if (incoming) {
176 struct ifaddr *ifa;
177
178 /* Sanity check */
179 KASSERT((m->m_flags & M_PKTHDR), ("%s: !PKTHDR", __func__));
180
181 /* Find IP address for receive interface */
182 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
183 if (ifa->ifa_addr == NULL)
184 continue;
185 if (ifa->ifa_addr->sa_family != AF_INET)
186 continue;
187 divsrc.sin_addr =
188 ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr;
189 break;
190 }
191 }
192 /*
193 * Record the incoming interface name whenever we have one.
194 */
195 bzero(&divsrc.sin_zero, sizeof(divsrc.sin_zero));
196 if (m->m_pkthdr.rcvif) {
197 /*
198 * Hide the actual interface name in there in the
199 * sin_zero array. XXX This needs to be moved to a
200 * different sockaddr type for divert, e.g.
201 * sockaddr_div with multiple fields like
202 * sockaddr_dl. Presently we have only 7 bytes
203 * but that will do for now as most interfaces
204 * are 4 or less + 2 or less bytes for unit.
205 * There is probably a faster way of doing this,
206 * possibly taking it from the sockaddr_dl on the iface.
207 * This solves the problem of a P2P link and a LAN interface
208 * having the same address, which can result in the wrong
209 * interface being assigned to the packet when fed back
210 * into the divert socket. Theoretically if the daemon saves
211 * and re-uses the sockaddr_in as suggested in the man pages,
212 * this iface name will come along for the ride.
213 * (see div_output for the other half of this.)
214 */
215 snprintf(divsrc.sin_zero, sizeof(divsrc.sin_zero),
216 "%s%d", m->m_pkthdr.rcvif->if_name,
217 m->m_pkthdr.rcvif->if_unit);
218 }
219
220 /* Put packet on socket queue, if any */
221 sa = NULL;
222 nport = htons((u_int16_t)port);
223 LIST_FOREACH(inp, &divcb, inp_list) {
224 if (inp->inp_lport == nport)
225 sa = inp->inp_socket;
226 }
227 if (sa) {
228 if (sbappendaddr(&sa->so_rcv, (struct sockaddr *)&divsrc,
229 m, (struct mbuf *)0) == 0)
230 m_freem(m);
231 else
232 sorwakeup(sa);
233 } else {
234 m_freem(m);
235 ipstat.ips_noproto++;
236 ipstat.ips_delivered--;
237 }
238}
239
240/*
241 * Deliver packet back into the IP processing machinery.
242 *
243 * If no address specified, or address is 0.0.0.0, send to ip_output();
244 * otherwise, send to ip_input() and mark as having been received on
245 * the interface with that address.
246 */
247static int
248div_output(so, m, addr, control)
249 struct socket *so;
250 register struct mbuf *m;
251 struct sockaddr *addr;
252 struct mbuf *control;
253{
254 register struct inpcb *const inp = sotoinpcb(so);
255 register struct ip *const ip = mtod(m, struct ip *);
256 struct sockaddr_in *sin = (struct sockaddr_in *)addr;
257 int error = 0;
258
259 if (control)
260 m_freem(control); /* XXX */
261
262 /* Loopback avoidance and state recovery */
263 if (sin) {
264 int len = 0;
265 char *c = sin->sin_zero;
266
267 ip_divert_cookie = sin->sin_port;
268
269 /*
270 * Find receive interface with the given name or IP address.
271 * The name is user supplied data so don't trust it's size or
272 * that it is zero terminated. The name has priority.
273 * We are presently assuming that the sockaddr_in
274 * has not been replaced by a sockaddr_div, so we limit it
275 * to 16 bytes in total. the name is stuffed (if it exists)
276 * in the sin_zero[] field.
277 */
278 while (*c++ && (len++ < sizeof(sin->sin_zero)));
279 if ((len > 0) && (len < sizeof(sin->sin_zero)))
280 m->m_pkthdr.rcvif = ifunit(sin->sin_zero);
281 } else {
282 ip_divert_cookie = 0;
283 }
284
285 /* Reinject packet into the system as incoming or outgoing */
286 if (!sin || sin->sin_addr.s_addr == 0) {
287 /*
288 * Don't allow both user specified and setsockopt options,
289 * and don't allow packet length sizes that will crash
290 */
291 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) ||
292 ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) {
293 error = EINVAL;
294 goto cantsend;
295 }
296
297 /* Convert fields to host order for ip_output() */
298 ip->ip_len = ntohs(ip->ip_len);
299 ip->ip_off = ntohs(ip->ip_off);
300
301 /* Send packet to output processing */
302 ipstat.ips_rawout++; /* XXX */
303 error = ip_output(m, inp->inp_options, &inp->inp_route,
304 (so->so_options & SO_DONTROUTE) |
305 IP_ALLOWBROADCAST | IP_RAWOUTPUT,
306 inp->inp_moptions);
307 } else {
308 struct ifaddr *ifa;
309
310 /* If no luck with the name above. check by IP address. */
311 if (m->m_pkthdr.rcvif == NULL) {
312 /*
313 * Make sure there are no distractions
314 * for ifa_ifwithaddr. Clear the port and the ifname.
315 * Maybe zap all 8 bytes at once using a 64bit write?
316 */
317 bzero(sin->sin_zero, sizeof(sin->sin_zero));
318 /* *((u_int64_t *)sin->sin_zero) = 0; */ /* XXX ?? */
319 sin->sin_port = 0;
320 if (!(ifa = ifa_ifwithaddr((struct sockaddr *) sin))) {
321 error = EADDRNOTAVAIL;
322 goto cantsend;
323 }
324 m->m_pkthdr.rcvif = ifa->ifa_ifp;
325 }
326
327 /* Send packet to input processing */
328 ip_input(m);
329 }
330
331 /* paranoid: Reset for next time (and other packets) */
332 /* almost definitly already done in the ipfw filter but.. */
333 ip_divert_cookie = 0;
334 return error;
335
336cantsend:
337 m_freem(m);
338 ip_divert_cookie = 0;
339 return error;
340}
341
342static int
343div_attach(struct socket *so, int proto, struct thread *td)
344{
345 struct inpcb *inp;
346 int error, s;
347
348 inp = sotoinpcb(so);
349 if (inp)
350 panic("div_attach");
351 if (td && (error = suser(td)) != 0)
352 return error;
353
354 error = soreserve(so, div_sendspace, div_recvspace);
355 if (error)
356 return error;
357 s = splnet();
358 error = in_pcballoc(so, &divcbinfo, td);
359 splx(s);
360 if (error)
361 return error;
362 inp = (struct inpcb *)so->so_pcb;
363 inp->inp_ip_p = proto;
364 inp->inp_vflag |= INP_IPV4;
365 inp->inp_flags |= INP_HDRINCL;
366 /* The socket is always "connected" because
367 we always know "where" to send the packet */
368 so->so_state |= SS_ISCONNECTED;
369 return 0;
370}
371
372static int
373div_detach(struct socket *so)
374{
375 struct inpcb *inp;
376
377 inp = sotoinpcb(so);
378 if (inp == 0)
379 panic("div_detach");
380 in_pcbdetach(inp);
381 return 0;
382}
383
384static int
385div_abort(struct socket *so)
386{
387 soisdisconnected(so);
388 return div_detach(so);
389}
390
391static int
392div_disconnect(struct socket *so)
393{
394 if ((so->so_state & SS_ISCONNECTED) == 0)
395 return ENOTCONN;
396 return div_abort(so);
397}
398
399static int
400div_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
401{
402 struct inpcb *inp;
403 int s;
404 int error;
405
406 s = splnet();
407 inp = sotoinpcb(so);
408 /* in_pcbbind assumes that the socket is a sockaddr_in
409 * and in_pcbbind requires a valid address. Since divert
410 * sockets don't we need to make sure the address is
411 * filled in properly.
412 * XXX -- divert should not be abusing in_pcbind
413 * and should probably have its own family.
414 */
415 if (nam->sa_family != AF_INET) {
416 error = EAFNOSUPPORT;
417 } else {
418 ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY;
419 error = in_pcbbind(inp, nam, td);
420 }
421 splx(s);
422 return error;
423}
424
425static int
426div_shutdown(struct socket *so)
427{
428 socantsendmore(so);
429 return 0;
430}
431
432static int
433div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
434 struct mbuf *control, struct thread *td)
435{
436 /* Packet must have a header (but that's about it) */
437 if (m->m_len < sizeof (struct ip) &&
438 (m = m_pullup(m, sizeof (struct ip))) == 0) {
439 ipstat.ips_toosmall++;
440 m_freem(m);
441 return EINVAL;
442 }
443
444 /* Send packet */
445 return div_output(so, m, nam, control);
446}
447
448static int
449div_pcblist(SYSCTL_HANDLER_ARGS)
450{
451 int error, i, n, s;
452 struct inpcb *inp, **inp_list;
453 inp_gen_t gencnt;
454 struct xinpgen xig;
455
456 /*
457 * The process of preparing the TCB list is too time-consuming and
458 * resource-intensive to repeat twice on every request.
459 */
460 if (req->oldptr == 0) {
461 n = divcbinfo.ipi_count;
462 req->oldidx = 2 * (sizeof xig)
463 + (n + n/8) * sizeof(struct xinpcb);
464 return 0;
465 }
466
467 if (req->newptr != 0)
468 return EPERM;
469
470 /*
471 * OK, now we're committed to doing something.
472 */
473 s = splnet();
474 gencnt = divcbinfo.ipi_gencnt;
475 n = divcbinfo.ipi_count;
476 splx(s);
477
478 xig.xig_len = sizeof xig;
479 xig.xig_count = n;
480 xig.xig_gen = gencnt;
481 xig.xig_sogen = so_gencnt;
482 error = SYSCTL_OUT(req, &xig, sizeof xig);
483 if (error)
484 return error;
485
486 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
487 if (inp_list == 0)
488 return ENOMEM;
489
490 s = splnet();
491 for (inp = LIST_FIRST(divcbinfo.listhead), i = 0; inp && i < n;
492 inp = LIST_NEXT(inp, inp_list)) {
493 if (inp->inp_gencnt <= gencnt && !prison_xinpcb(req->td, inp))
494 inp_list[i++] = inp;
495 }
496 splx(s);
497 n = i;
498
499 error = 0;
500 for (i = 0; i < n; i++) {
501 inp = inp_list[i];
502 if (inp->inp_gencnt <= gencnt) {
503 struct xinpcb xi;
504 xi.xi_len = sizeof xi;
505 /* XXX should avoid extra copy */
506 bcopy(inp, &xi.xi_inp, sizeof *inp);
507 if (inp->inp_socket)
508 sotoxsocket(inp->inp_socket, &xi.xi_socket);
509 error = SYSCTL_OUT(req, &xi, sizeof xi);
510 }
511 }
512 if (!error) {
513 /*
514 * Give the user an updated idea of our state.
515 * If the generation differs from what we told
516 * her before, she knows that something happened
517 * while we were processing this request, and it
518 * might be necessary to retry.
519 */
520 s = splnet();
521 xig.xig_gen = divcbinfo.ipi_gencnt;
522 xig.xig_sogen = so_gencnt;
523 xig.xig_count = divcbinfo.ipi_count;
524 splx(s);
525 error = SYSCTL_OUT(req, &xig, sizeof xig);
526 }
527 free(inp_list, M_TEMP);
528 return error;
529}
530
531SYSCTL_DECL(_net_inet_divert);
532SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLFLAG_RD, 0, 0,
533 div_pcblist, "S,xinpcb", "List of active divert sockets");
534
535struct pr_usrreqs div_usrreqs = {
536 div_abort, pru_accept_notsupp, div_attach, div_bind,
537 pru_connect_notsupp, pru_connect2_notsupp, in_control, div_detach,
538 div_disconnect, pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp,
539 pru_rcvoob_notsupp, div_send, pru_sense_null, div_shutdown,
540 in_setsockaddr, sosend, soreceive, sopoll
541};
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: head/sys/netinet/ip_divert.c 95759 2002-04-30 01:54:54Z tanimura $
34 */
35
36#include "opt_inet.h"
37#include "opt_ipfw.h"
38#include "opt_ipdivert.h"
39#include "opt_ipsec.h"
40
41#ifndef INET
42#error "IPDIVERT requires INET."
43#endif
44
45#include <sys/param.h>
46#include <sys/kernel.h>
47#include <sys/lock.h>
48#include <sys/malloc.h>
49#include <sys/mbuf.h>
50#include <sys/proc.h>
51#include <sys/protosw.h>
52#include <sys/signalvar.h>
53#include <sys/socket.h>
54#include <sys/socketvar.h>
55#include <sys/sx.h>
56#include <sys/sysctl.h>
57#include <sys/systm.h>
58
59#include <vm/uma.h>
60
61#include <net/if.h>
62#include <net/route.h>
63
64#include <netinet/in.h>
65#include <netinet/in_pcb.h>
66#include <netinet/in_systm.h>
67#include <netinet/in_var.h>
68#include <netinet/ip.h>
69#include <netinet/ip_var.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 * A 16 bit cookie is passed to and from the user process.
83 * The user process can send it back to help the caller know
84 * something about where the packet originally came from.
85 *
86 * In the case of ipfw, then the cookie is the rule that sent
87 * us here. On reinjection is is the rule after which processing
88 * should continue. Leaving it the same will make processing start
89 * at the rule number after that which sent it here. Setting it to
90 * 0 will restart processing at the beginning.
91 *
92 * For divert_packet(), ip_divert_cookie is an input value only.
93 * For div_output(), ip_divert_cookie is an output value only.
94 */
95u_int16_t ip_divert_cookie;
96
97/* Internal variables */
98static struct inpcbhead divcb;
99static struct inpcbinfo divcbinfo;
100
101static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */
102static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */
103
104/* Optimization: have this preinitialized */
105static struct sockaddr_in divsrc = { sizeof(divsrc), AF_INET };
106
107/* Internal functions */
108static int div_output(struct socket *so,
109 struct mbuf *m, struct sockaddr *addr, struct mbuf *control);
110
111/*
112 * Initialize divert connection block queue.
113 */
114void
115div_init(void)
116{
117 LIST_INIT(&divcb);
118 divcbinfo.listhead = &divcb;
119 /*
120 * XXX We don't use the hash list for divert IP, but it's easier
121 * to allocate a one entry hash list than it is to check all
122 * over the place for hashbase == NULL.
123 */
124 divcbinfo.hashbase = hashinit(1, M_PCB, &divcbinfo.hashmask);
125 divcbinfo.porthashbase = hashinit(1, M_PCB, &divcbinfo.porthashmask);
126 divcbinfo.ipi_zone = uma_zcreate("divcb", sizeof(struct inpcb),
127 NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
128 uma_zone_set_max(divcbinfo.ipi_zone, maxsockets);
129}
130
131/*
132 * IPPROTO_DIVERT is not a real IP protocol; don't allow any packets
133 * with that protocol number to enter the system from the outside.
134 */
135void
136div_input(struct mbuf *m, int off)
137{
138 ipstat.ips_noproto++;
139 m_freem(m);
140}
141
142/*
143 * Divert a packet by passing it up to the divert socket at port 'port'.
144 *
145 * Setup generic address and protocol structures for div_input routine,
146 * then pass them along with mbuf chain.
147 */
148void
149divert_packet(struct mbuf *m, int incoming, int port)
150{
151 struct ip *ip;
152 struct inpcb *inp;
153 struct socket *sa;
154 u_int16_t nport;
155
156 /* Sanity check */
157 KASSERT(port != 0, ("%s: port=0", __func__));
158
159 /* Record and reset divert cookie */
160 divsrc.sin_port = ip_divert_cookie;
161 ip_divert_cookie = 0;
162
163 /* Assure header */
164 if (m->m_len < sizeof(struct ip) &&
165 (m = m_pullup(m, sizeof(struct ip))) == 0) {
166 return;
167 }
168 ip = mtod(m, struct ip *);
169
170 /*
171 * Record receive interface address, if any.
172 * But only for incoming packets.
173 */
174 divsrc.sin_addr.s_addr = 0;
175 if (incoming) {
176 struct ifaddr *ifa;
177
178 /* Sanity check */
179 KASSERT((m->m_flags & M_PKTHDR), ("%s: !PKTHDR", __func__));
180
181 /* Find IP address for receive interface */
182 TAILQ_FOREACH(ifa, &m->m_pkthdr.rcvif->if_addrhead, ifa_link) {
183 if (ifa->ifa_addr == NULL)
184 continue;
185 if (ifa->ifa_addr->sa_family != AF_INET)
186 continue;
187 divsrc.sin_addr =
188 ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr;
189 break;
190 }
191 }
192 /*
193 * Record the incoming interface name whenever we have one.
194 */
195 bzero(&divsrc.sin_zero, sizeof(divsrc.sin_zero));
196 if (m->m_pkthdr.rcvif) {
197 /*
198 * Hide the actual interface name in there in the
199 * sin_zero array. XXX This needs to be moved to a
200 * different sockaddr type for divert, e.g.
201 * sockaddr_div with multiple fields like
202 * sockaddr_dl. Presently we have only 7 bytes
203 * but that will do for now as most interfaces
204 * are 4 or less + 2 or less bytes for unit.
205 * There is probably a faster way of doing this,
206 * possibly taking it from the sockaddr_dl on the iface.
207 * This solves the problem of a P2P link and a LAN interface
208 * having the same address, which can result in the wrong
209 * interface being assigned to the packet when fed back
210 * into the divert socket. Theoretically if the daemon saves
211 * and re-uses the sockaddr_in as suggested in the man pages,
212 * this iface name will come along for the ride.
213 * (see div_output for the other half of this.)
214 */
215 snprintf(divsrc.sin_zero, sizeof(divsrc.sin_zero),
216 "%s%d", m->m_pkthdr.rcvif->if_name,
217 m->m_pkthdr.rcvif->if_unit);
218 }
219
220 /* Put packet on socket queue, if any */
221 sa = NULL;
222 nport = htons((u_int16_t)port);
223 LIST_FOREACH(inp, &divcb, inp_list) {
224 if (inp->inp_lport == nport)
225 sa = inp->inp_socket;
226 }
227 if (sa) {
228 if (sbappendaddr(&sa->so_rcv, (struct sockaddr *)&divsrc,
229 m, (struct mbuf *)0) == 0)
230 m_freem(m);
231 else
232 sorwakeup(sa);
233 } else {
234 m_freem(m);
235 ipstat.ips_noproto++;
236 ipstat.ips_delivered--;
237 }
238}
239
240/*
241 * Deliver packet back into the IP processing machinery.
242 *
243 * If no address specified, or address is 0.0.0.0, send to ip_output();
244 * otherwise, send to ip_input() and mark as having been received on
245 * the interface with that address.
246 */
247static int
248div_output(so, m, addr, control)
249 struct socket *so;
250 register struct mbuf *m;
251 struct sockaddr *addr;
252 struct mbuf *control;
253{
254 register struct inpcb *const inp = sotoinpcb(so);
255 register struct ip *const ip = mtod(m, struct ip *);
256 struct sockaddr_in *sin = (struct sockaddr_in *)addr;
257 int error = 0;
258
259 if (control)
260 m_freem(control); /* XXX */
261
262 /* Loopback avoidance and state recovery */
263 if (sin) {
264 int len = 0;
265 char *c = sin->sin_zero;
266
267 ip_divert_cookie = sin->sin_port;
268
269 /*
270 * Find receive interface with the given name or IP address.
271 * The name is user supplied data so don't trust it's size or
272 * that it is zero terminated. The name has priority.
273 * We are presently assuming that the sockaddr_in
274 * has not been replaced by a sockaddr_div, so we limit it
275 * to 16 bytes in total. the name is stuffed (if it exists)
276 * in the sin_zero[] field.
277 */
278 while (*c++ && (len++ < sizeof(sin->sin_zero)));
279 if ((len > 0) && (len < sizeof(sin->sin_zero)))
280 m->m_pkthdr.rcvif = ifunit(sin->sin_zero);
281 } else {
282 ip_divert_cookie = 0;
283 }
284
285 /* Reinject packet into the system as incoming or outgoing */
286 if (!sin || sin->sin_addr.s_addr == 0) {
287 /*
288 * Don't allow both user specified and setsockopt options,
289 * and don't allow packet length sizes that will crash
290 */
291 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options) ||
292 ((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) {
293 error = EINVAL;
294 goto cantsend;
295 }
296
297 /* Convert fields to host order for ip_output() */
298 ip->ip_len = ntohs(ip->ip_len);
299 ip->ip_off = ntohs(ip->ip_off);
300
301 /* Send packet to output processing */
302 ipstat.ips_rawout++; /* XXX */
303 error = ip_output(m, inp->inp_options, &inp->inp_route,
304 (so->so_options & SO_DONTROUTE) |
305 IP_ALLOWBROADCAST | IP_RAWOUTPUT,
306 inp->inp_moptions);
307 } else {
308 struct ifaddr *ifa;
309
310 /* If no luck with the name above. check by IP address. */
311 if (m->m_pkthdr.rcvif == NULL) {
312 /*
313 * Make sure there are no distractions
314 * for ifa_ifwithaddr. Clear the port and the ifname.
315 * Maybe zap all 8 bytes at once using a 64bit write?
316 */
317 bzero(sin->sin_zero, sizeof(sin->sin_zero));
318 /* *((u_int64_t *)sin->sin_zero) = 0; */ /* XXX ?? */
319 sin->sin_port = 0;
320 if (!(ifa = ifa_ifwithaddr((struct sockaddr *) sin))) {
321 error = EADDRNOTAVAIL;
322 goto cantsend;
323 }
324 m->m_pkthdr.rcvif = ifa->ifa_ifp;
325 }
326
327 /* Send packet to input processing */
328 ip_input(m);
329 }
330
331 /* paranoid: Reset for next time (and other packets) */
332 /* almost definitly already done in the ipfw filter but.. */
333 ip_divert_cookie = 0;
334 return error;
335
336cantsend:
337 m_freem(m);
338 ip_divert_cookie = 0;
339 return error;
340}
341
342static int
343div_attach(struct socket *so, int proto, struct thread *td)
344{
345 struct inpcb *inp;
346 int error, s;
347
348 inp = sotoinpcb(so);
349 if (inp)
350 panic("div_attach");
351 if (td && (error = suser(td)) != 0)
352 return error;
353
354 error = soreserve(so, div_sendspace, div_recvspace);
355 if (error)
356 return error;
357 s = splnet();
358 error = in_pcballoc(so, &divcbinfo, td);
359 splx(s);
360 if (error)
361 return error;
362 inp = (struct inpcb *)so->so_pcb;
363 inp->inp_ip_p = proto;
364 inp->inp_vflag |= INP_IPV4;
365 inp->inp_flags |= INP_HDRINCL;
366 /* The socket is always "connected" because
367 we always know "where" to send the packet */
368 so->so_state |= SS_ISCONNECTED;
369 return 0;
370}
371
372static int
373div_detach(struct socket *so)
374{
375 struct inpcb *inp;
376
377 inp = sotoinpcb(so);
378 if (inp == 0)
379 panic("div_detach");
380 in_pcbdetach(inp);
381 return 0;
382}
383
384static int
385div_abort(struct socket *so)
386{
387 soisdisconnected(so);
388 return div_detach(so);
389}
390
391static int
392div_disconnect(struct socket *so)
393{
394 if ((so->so_state & SS_ISCONNECTED) == 0)
395 return ENOTCONN;
396 return div_abort(so);
397}
398
399static int
400div_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
401{
402 struct inpcb *inp;
403 int s;
404 int error;
405
406 s = splnet();
407 inp = sotoinpcb(so);
408 /* in_pcbbind assumes that the socket is a sockaddr_in
409 * and in_pcbbind requires a valid address. Since divert
410 * sockets don't we need to make sure the address is
411 * filled in properly.
412 * XXX -- divert should not be abusing in_pcbind
413 * and should probably have its own family.
414 */
415 if (nam->sa_family != AF_INET) {
416 error = EAFNOSUPPORT;
417 } else {
418 ((struct sockaddr_in *)nam)->sin_addr.s_addr = INADDR_ANY;
419 error = in_pcbbind(inp, nam, td);
420 }
421 splx(s);
422 return error;
423}
424
425static int
426div_shutdown(struct socket *so)
427{
428 socantsendmore(so);
429 return 0;
430}
431
432static int
433div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
434 struct mbuf *control, struct thread *td)
435{
436 /* Packet must have a header (but that's about it) */
437 if (m->m_len < sizeof (struct ip) &&
438 (m = m_pullup(m, sizeof (struct ip))) == 0) {
439 ipstat.ips_toosmall++;
440 m_freem(m);
441 return EINVAL;
442 }
443
444 /* Send packet */
445 return div_output(so, m, nam, control);
446}
447
448static int
449div_pcblist(SYSCTL_HANDLER_ARGS)
450{
451 int error, i, n, s;
452 struct inpcb *inp, **inp_list;
453 inp_gen_t gencnt;
454 struct xinpgen xig;
455
456 /*
457 * The process of preparing the TCB list is too time-consuming and
458 * resource-intensive to repeat twice on every request.
459 */
460 if (req->oldptr == 0) {
461 n = divcbinfo.ipi_count;
462 req->oldidx = 2 * (sizeof xig)
463 + (n + n/8) * sizeof(struct xinpcb);
464 return 0;
465 }
466
467 if (req->newptr != 0)
468 return EPERM;
469
470 /*
471 * OK, now we're committed to doing something.
472 */
473 s = splnet();
474 gencnt = divcbinfo.ipi_gencnt;
475 n = divcbinfo.ipi_count;
476 splx(s);
477
478 xig.xig_len = sizeof xig;
479 xig.xig_count = n;
480 xig.xig_gen = gencnt;
481 xig.xig_sogen = so_gencnt;
482 error = SYSCTL_OUT(req, &xig, sizeof xig);
483 if (error)
484 return error;
485
486 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
487 if (inp_list == 0)
488 return ENOMEM;
489
490 s = splnet();
491 for (inp = LIST_FIRST(divcbinfo.listhead), i = 0; inp && i < n;
492 inp = LIST_NEXT(inp, inp_list)) {
493 if (inp->inp_gencnt <= gencnt && !prison_xinpcb(req->td, inp))
494 inp_list[i++] = inp;
495 }
496 splx(s);
497 n = i;
498
499 error = 0;
500 for (i = 0; i < n; i++) {
501 inp = inp_list[i];
502 if (inp->inp_gencnt <= gencnt) {
503 struct xinpcb xi;
504 xi.xi_len = sizeof xi;
505 /* XXX should avoid extra copy */
506 bcopy(inp, &xi.xi_inp, sizeof *inp);
507 if (inp->inp_socket)
508 sotoxsocket(inp->inp_socket, &xi.xi_socket);
509 error = SYSCTL_OUT(req, &xi, sizeof xi);
510 }
511 }
512 if (!error) {
513 /*
514 * Give the user an updated idea of our state.
515 * If the generation differs from what we told
516 * her before, she knows that something happened
517 * while we were processing this request, and it
518 * might be necessary to retry.
519 */
520 s = splnet();
521 xig.xig_gen = divcbinfo.ipi_gencnt;
522 xig.xig_sogen = so_gencnt;
523 xig.xig_count = divcbinfo.ipi_count;
524 splx(s);
525 error = SYSCTL_OUT(req, &xig, sizeof xig);
526 }
527 free(inp_list, M_TEMP);
528 return error;
529}
530
531SYSCTL_DECL(_net_inet_divert);
532SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, CTLFLAG_RD, 0, 0,
533 div_pcblist, "S,xinpcb", "List of active divert sockets");
534
535struct pr_usrreqs div_usrreqs = {
536 div_abort, pru_accept_notsupp, div_attach, div_bind,
537 pru_connect_notsupp, pru_connect2_notsupp, in_control, div_detach,
538 div_disconnect, pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp,
539 pru_rcvoob_notsupp, div_send, pru_sense_null, div_shutdown,
540 in_setsockaddr, sosend, soreceive, sopoll
541};