1 /*-
2 * Copyright (c) 1982, 1986, 1988, 1993
3 * The Regents of the University of California.
4 * All rights reserved.
5 *
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
8 * are met:
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
14 * 4. Neither the name of the University nor the names of its contributors
15 * may be used to endorse or promote products derived from this software
16 * without specific prior written permission.
17 *
18 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
19 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
20 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
21 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
22 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
23 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
24 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
25 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
26 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
27 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
28 * SUCH DAMAGE.
29 *
30 * @(#)raw_ip.c 8.7 (Berkeley) 5/15/95
31 */
32
33 #include <sys/cdefs.h>
34 __FBSDID("$FreeBSD$");
35
36 #include "opt_inet.h"
37 #include "opt_inet6.h"
38 #include "opt_ipsec.h"
39
40 #include <sys/param.h>
41 #include <sys/jail.h>
42 #include <sys/kernel.h>
43 #include <sys/eventhandler.h>
44 #include <sys/lock.h>
45 #include <sys/malloc.h>
46 #include <sys/mbuf.h>
47 #include <sys/priv.h>
48 #include <sys/proc.h>
49 #include <sys/protosw.h>
50 #include <sys/rmlock.h>
51 #include <sys/rwlock.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/if_var.h>
63 #include <net/route.h>
64 #include <net/vnet.h>
65
66 #include <netinet/in.h>
67 #include <netinet/in_systm.h>
68 #include <netinet/in_pcb.h>
69 #include <netinet/in_var.h>
70 #include <netinet/if_ether.h>
71 #include <netinet/ip.h>
72 #include <netinet/ip_var.h>
73 #include <netinet/ip_mroute.h>
74
75 #ifdef IPSEC
76 #include <netipsec/ipsec.h>
77 #endif /*IPSEC*/
78
79 #include <machine/stdarg.h>
80 #include <security/mac/mac_framework.h>
81
82 VNET_DEFINE(int, ip_defttl) = IPDEFTTL;
83 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_VNET | CTLFLAG_RW,
84 &VNET_NAME(ip_defttl), 0,
85 "Maximum TTL on IP packets");
86
87 VNET_DEFINE(struct inpcbhead, ripcb);
88 VNET_DEFINE(struct inpcbinfo, ripcbinfo);
89
90 #define V_ripcb VNET(ripcb)
91 #define V_ripcbinfo VNET(ripcbinfo)
92
93 /*
94 * Control and data hooks for ipfw, dummynet, divert and so on.
95 * The data hooks are not used here but it is convenient
96 * to keep them all in one place.
97 */
98 VNET_DEFINE(ip_fw_chk_ptr_t, ip_fw_chk_ptr) = NULL;
99 VNET_DEFINE(ip_fw_ctl_ptr_t, ip_fw_ctl_ptr) = NULL;
100
101 int (*ip_dn_ctl_ptr)(struct sockopt *);
102 int (*ip_dn_io_ptr)(struct mbuf **, int, struct ip_fw_args *);
103 void (*ip_divert_ptr)(struct mbuf *, int);
104 int (*ng_ipfw_input_p)(struct mbuf **, int,
105 struct ip_fw_args *, int);
106
107 #ifdef INET
108 /*
109 * Hooks for multicast routing. They all default to NULL, so leave them not
110 * initialized and rely on BSS being set to 0.
111 */
112
113 /*
114 * The socket used to communicate with the multicast routing daemon.
115 */
116 VNET_DEFINE(struct socket *, ip_mrouter);
117
118 /*
119 * The various mrouter and rsvp functions.
120 */
121 int (*ip_mrouter_set)(struct socket *, struct sockopt *);
122 int (*ip_mrouter_get)(struct socket *, struct sockopt *);
123 int (*ip_mrouter_done)(void);
124 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
125 struct ip_moptions *);
126 int (*mrt_ioctl)(u_long, caddr_t, int);
127 int (*legal_vif_num)(int);
128 u_long (*ip_mcast_src)(int);
129
130 int (*rsvp_input_p)(struct mbuf **, int *, int);
131 int (*ip_rsvp_vif)(struct socket *, struct sockopt *);
132 void (*ip_rsvp_force_done)(struct socket *);
133 #endif /* INET */
134
135 u_long rip_sendspace = 9216;
136 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
137 &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
138
139 u_long rip_recvspace = 9216;
140 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW,
141 &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams");
142
143 /*
144 * Hash functions
145 */
146
147 #define INP_PCBHASH_RAW_SIZE 256
148 #define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \
149 (((proto) + (laddr) + (faddr)) % (mask) + 1)
150
151 #ifdef INET
152 static void
rip_inshash(struct inpcb * inp)153 rip_inshash(struct inpcb *inp)
154 {
155 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
156 struct inpcbhead *pcbhash;
157 int hash;
158
159 INP_INFO_WLOCK_ASSERT(pcbinfo);
160 INP_WLOCK_ASSERT(inp);
161
162 if (inp->inp_ip_p != 0 &&
163 inp->inp_laddr.s_addr != INADDR_ANY &&
164 inp->inp_faddr.s_addr != INADDR_ANY) {
165 hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr,
166 inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask);
167 } else
168 hash = 0;
169 pcbhash = &pcbinfo->ipi_hashbase[hash];
170 LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
171 }
172
173 static void
rip_delhash(struct inpcb * inp)174 rip_delhash(struct inpcb *inp)
175 {
176
177 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
178 INP_WLOCK_ASSERT(inp);
179
180 LIST_REMOVE(inp, inp_hash);
181 }
182 #endif /* INET */
183
184 /*
185 * Raw interface to IP protocol.
186 */
187
188 /*
189 * Initialize raw connection block q.
190 */
191 static void
rip_zone_change(void * tag)192 rip_zone_change(void *tag)
193 {
194
195 uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets);
196 }
197
198 static int
rip_inpcb_init(void * mem,int size,int flags)199 rip_inpcb_init(void *mem, int size, int flags)
200 {
201 struct inpcb *inp = mem;
202
203 INP_LOCK_INIT(inp, "inp", "rawinp");
204 return (0);
205 }
206
207 void
rip_init(void)208 rip_init(void)
209 {
210
211 in_pcbinfo_init(&V_ripcbinfo, "rip", &V_ripcb, INP_PCBHASH_RAW_SIZE,
212 1, "ripcb", rip_inpcb_init, NULL, UMA_ZONE_NOFREE,
213 IPI_HASHFIELDS_NONE);
214 EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL,
215 EVENTHANDLER_PRI_ANY);
216 }
217
218 #ifdef VIMAGE
219 void
rip_destroy(void)220 rip_destroy(void)
221 {
222
223 in_pcbinfo_destroy(&V_ripcbinfo);
224 }
225 #endif
226
227 #ifdef INET
228 static int
rip_append(struct inpcb * last,struct ip * ip,struct mbuf * n,struct sockaddr_in * ripsrc)229 rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n,
230 struct sockaddr_in *ripsrc)
231 {
232 int policyfail = 0;
233
234 INP_LOCK_ASSERT(last);
235
236 #ifdef IPSEC
237 /* check AH/ESP integrity. */
238 if (ipsec4_in_reject(n, last)) {
239 policyfail = 1;
240 }
241 #endif /* IPSEC */
242 #ifdef MAC
243 if (!policyfail && mac_inpcb_check_deliver(last, n) != 0)
244 policyfail = 1;
245 #endif
246 /* Check the minimum TTL for socket. */
247 if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl)
248 policyfail = 1;
249 if (!policyfail) {
250 struct mbuf *opts = NULL;
251 struct socket *so;
252
253 so = last->inp_socket;
254 if ((last->inp_flags & INP_CONTROLOPTS) ||
255 (so->so_options & (SO_TIMESTAMP | SO_BINTIME)))
256 ip_savecontrol(last, &opts, ip, n);
257 SOCKBUF_LOCK(&so->so_rcv);
258 if (sbappendaddr_locked(&so->so_rcv,
259 (struct sockaddr *)ripsrc, n, opts) == 0) {
260 /* should notify about lost packet */
261 m_freem(n);
262 if (opts)
263 m_freem(opts);
264 SOCKBUF_UNLOCK(&so->so_rcv);
265 } else
266 sorwakeup_locked(so);
267 } else
268 m_freem(n);
269 return (policyfail);
270 }
271
272 /*
273 * Setup generic address and protocol structures for raw_input routine, then
274 * pass them along with mbuf chain.
275 */
276 int
rip_input(struct mbuf ** mp,int * offp,int proto)277 rip_input(struct mbuf **mp, int *offp, int proto)
278 {
279 struct ifnet *ifp;
280 struct mbuf *m = *mp;
281 struct ip *ip = mtod(m, struct ip *);
282 struct inpcb *inp, *last;
283 struct sockaddr_in ripsrc;
284 int hash;
285
286 *mp = NULL;
287
288 bzero(&ripsrc, sizeof(ripsrc));
289 ripsrc.sin_len = sizeof(ripsrc);
290 ripsrc.sin_family = AF_INET;
291 ripsrc.sin_addr = ip->ip_src;
292 last = NULL;
293
294 ifp = m->m_pkthdr.rcvif;
295
296 hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr,
297 ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask);
298 INP_INFO_RLOCK(&V_ripcbinfo);
299 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) {
300 if (inp->inp_ip_p != proto)
301 continue;
302 #ifdef INET6
303 /* XXX inp locking */
304 if ((inp->inp_vflag & INP_IPV4) == 0)
305 continue;
306 #endif
307 if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
308 continue;
309 if (inp->inp_faddr.s_addr != ip->ip_src.s_addr)
310 continue;
311 if (jailed_without_vnet(inp->inp_cred)) {
312 /*
313 * XXX: If faddr was bound to multicast group,
314 * jailed raw socket will drop datagram.
315 */
316 if (prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0)
317 continue;
318 }
319 if (last != NULL) {
320 struct mbuf *n;
321
322 n = m_copy(m, 0, (int)M_COPYALL);
323 if (n != NULL)
324 (void) rip_append(last, ip, n, &ripsrc);
325 /* XXX count dropped packet */
326 INP_RUNLOCK(last);
327 }
328 INP_RLOCK(inp);
329 last = inp;
330 }
331 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) {
332 if (inp->inp_ip_p && inp->inp_ip_p != proto)
333 continue;
334 #ifdef INET6
335 /* XXX inp locking */
336 if ((inp->inp_vflag & INP_IPV4) == 0)
337 continue;
338 #endif
339 if (!in_nullhost(inp->inp_laddr) &&
340 !in_hosteq(inp->inp_laddr, ip->ip_dst))
341 continue;
342 if (!in_nullhost(inp->inp_faddr) &&
343 !in_hosteq(inp->inp_faddr, ip->ip_src))
344 continue;
345 if (jailed_without_vnet(inp->inp_cred)) {
346 /*
347 * Allow raw socket in jail to receive multicast;
348 * assume process had PRIV_NETINET_RAW at attach,
349 * and fall through into normal filter path if so.
350 */
351 if (!IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
352 prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0)
353 continue;
354 }
355 /*
356 * If this raw socket has multicast state, and we
357 * have received a multicast, check if this socket
358 * should receive it, as multicast filtering is now
359 * the responsibility of the transport layer.
360 */
361 if (inp->inp_moptions != NULL &&
362 IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
363 /*
364 * If the incoming datagram is for IGMP, allow it
365 * through unconditionally to the raw socket.
366 *
367 * In the case of IGMPv2, we may not have explicitly
368 * joined the group, and may have set IFF_ALLMULTI
369 * on the interface. imo_multi_filter() may discard
370 * control traffic we actually need to see.
371 *
372 * Userland multicast routing daemons should continue
373 * filter the control traffic appropriately.
374 */
375 int blocked;
376
377 blocked = MCAST_PASS;
378 if (proto != IPPROTO_IGMP) {
379 struct sockaddr_in group;
380
381 bzero(&group, sizeof(struct sockaddr_in));
382 group.sin_len = sizeof(struct sockaddr_in);
383 group.sin_family = AF_INET;
384 group.sin_addr = ip->ip_dst;
385
386 blocked = imo_multi_filter(inp->inp_moptions,
387 ifp,
388 (struct sockaddr *)&group,
389 (struct sockaddr *)&ripsrc);
390 }
391
392 if (blocked != MCAST_PASS) {
393 IPSTAT_INC(ips_notmember);
394 continue;
395 }
396 }
397 if (last != NULL) {
398 struct mbuf *n;
399
400 n = m_copy(m, 0, (int)M_COPYALL);
401 if (n != NULL)
402 (void) rip_append(last, ip, n, &ripsrc);
403 /* XXX count dropped packet */
404 INP_RUNLOCK(last);
405 }
406 INP_RLOCK(inp);
407 last = inp;
408 }
409 INP_INFO_RUNLOCK(&V_ripcbinfo);
410 if (last != NULL) {
411 if (rip_append(last, ip, m, &ripsrc) != 0)
412 IPSTAT_INC(ips_delivered);
413 INP_RUNLOCK(last);
414 } else {
415 m_freem(m);
416 IPSTAT_INC(ips_noproto);
417 IPSTAT_DEC(ips_delivered);
418 }
419 return (IPPROTO_DONE);
420 }
421
422 /*
423 * Generate IP header and pass packet to ip_output. Tack on options user may
424 * have setup with control call.
425 */
426 int
rip_output(struct mbuf * m,struct socket * so,...)427 rip_output(struct mbuf *m, struct socket *so, ...)
428 {
429 struct ip *ip;
430 int error;
431 struct inpcb *inp = sotoinpcb(so);
432 va_list ap;
433 u_long dst;
434 int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) |
435 IP_ALLOWBROADCAST;
436
437 va_start(ap, so);
438 dst = va_arg(ap, u_long);
439 va_end(ap);
440
441 /*
442 * If the user handed us a complete IP packet, use it. Otherwise,
443 * allocate an mbuf for a header and fill it in.
444 */
445 if ((inp->inp_flags & INP_HDRINCL) == 0) {
446 if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
447 m_freem(m);
448 return(EMSGSIZE);
449 }
450 M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
451 if (m == NULL)
452 return(ENOBUFS);
453
454 INP_RLOCK(inp);
455 ip = mtod(m, struct ip *);
456 ip->ip_tos = inp->inp_ip_tos;
457 if (inp->inp_flags & INP_DONTFRAG)
458 ip->ip_off = htons(IP_DF);
459 else
460 ip->ip_off = htons(0);
461 ip->ip_p = inp->inp_ip_p;
462 ip->ip_len = htons(m->m_pkthdr.len);
463 ip->ip_src = inp->inp_laddr;
464 ip->ip_dst.s_addr = dst;
465 if (jailed(inp->inp_cred)) {
466 /*
467 * prison_local_ip4() would be good enough but would
468 * let a source of INADDR_ANY pass, which we do not
469 * want to see from jails.
470 */
471 if (ip->ip_src.s_addr == INADDR_ANY) {
472 error = in_pcbladdr(inp, &ip->ip_dst, &ip->ip_src,
473 inp->inp_cred);
474 } else {
475 error = prison_local_ip4(inp->inp_cred,
476 &ip->ip_src);
477 }
478 if (error != 0) {
479 INP_RUNLOCK(inp);
480 m_freem(m);
481 return (error);
482 }
483 }
484 ip->ip_ttl = inp->inp_ip_ttl;
485 } else {
486 if (m->m_pkthdr.len > IP_MAXPACKET) {
487 m_freem(m);
488 return(EMSGSIZE);
489 }
490 INP_RLOCK(inp);
491 ip = mtod(m, struct ip *);
492 error = prison_check_ip4(inp->inp_cred, &ip->ip_src);
493 if (error != 0) {
494 INP_RUNLOCK(inp);
495 m_freem(m);
496 return (error);
497 }
498
499 /*
500 * Don't allow both user specified and setsockopt options,
501 * and don't allow packet length sizes that will crash.
502 */
503 if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options)
504 || (ntohs(ip->ip_len) > m->m_pkthdr.len)
505 || (ntohs(ip->ip_len) < (ip->ip_hl << 2))) {
506 INP_RUNLOCK(inp);
507 m_freem(m);
508 return (EINVAL);
509 }
510 /*
511 * This doesn't allow application to specify ID of zero,
512 * but we got this limitation from the beginning of history.
513 */
514 if (ip->ip_id == 0)
515 ip_fillid(ip);
516
517 /*
518 * XXX prevent ip_output from overwriting header fields.
519 */
520 flags |= IP_RAWOUTPUT;
521 IPSTAT_INC(ips_rawout);
522 }
523
524 if (inp->inp_flags & INP_ONESBCAST)
525 flags |= IP_SENDONES;
526
527 #ifdef MAC
528 mac_inpcb_create_mbuf(inp, m);
529 #endif
530
531 error = ip_output(m, inp->inp_options, NULL, flags,
532 inp->inp_moptions, inp);
533 INP_RUNLOCK(inp);
534 return (error);
535 }
536
537 /*
538 * Raw IP socket option processing.
539 *
540 * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could
541 * only be created by a privileged process, and as such, socket option
542 * operations to manage system properties on any raw socket were allowed to
543 * take place without explicit additional access control checks. However,
544 * raw sockets can now also be created in jail(), and therefore explicit
545 * checks are now required. Likewise, raw sockets can be used by a process
546 * after it gives up privilege, so some caution is required. For options
547 * passed down to the IP layer via ip_ctloutput(), checks are assumed to be
548 * performed in ip_ctloutput() and therefore no check occurs here.
549 * Unilaterally checking priv_check() here breaks normal IP socket option
550 * operations on raw sockets.
551 *
552 * When adding new socket options here, make sure to add access control
553 * checks here as necessary.
554 *
555 * XXX-BZ inp locking?
556 */
557 int
rip_ctloutput(struct socket * so,struct sockopt * sopt)558 rip_ctloutput(struct socket *so, struct sockopt *sopt)
559 {
560 struct inpcb *inp = sotoinpcb(so);
561 int error, optval;
562
563 if (sopt->sopt_level != IPPROTO_IP) {
564 if ((sopt->sopt_level == SOL_SOCKET) &&
565 (sopt->sopt_name == SO_SETFIB)) {
566 inp->inp_inc.inc_fibnum = so->so_fibnum;
567 return (0);
568 }
569 return (EINVAL);
570 }
571
572 error = 0;
573 switch (sopt->sopt_dir) {
574 case SOPT_GET:
575 switch (sopt->sopt_name) {
576 case IP_HDRINCL:
577 optval = inp->inp_flags & INP_HDRINCL;
578 error = sooptcopyout(sopt, &optval, sizeof optval);
579 break;
580
581 case IP_FW3: /* generic ipfw v.3 functions */
582 case IP_FW_ADD: /* ADD actually returns the body... */
583 case IP_FW_GET:
584 case IP_FW_TABLE_GETSIZE:
585 case IP_FW_TABLE_LIST:
586 case IP_FW_NAT_GET_CONFIG:
587 case IP_FW_NAT_GET_LOG:
588 if (V_ip_fw_ctl_ptr != NULL)
589 error = V_ip_fw_ctl_ptr(sopt);
590 else
591 error = ENOPROTOOPT;
592 break;
593
594 case IP_DUMMYNET3: /* generic dummynet v.3 functions */
595 case IP_DUMMYNET_GET:
596 if (ip_dn_ctl_ptr != NULL)
597 error = ip_dn_ctl_ptr(sopt);
598 else
599 error = ENOPROTOOPT;
600 break ;
601
602 case MRT_INIT:
603 case MRT_DONE:
604 case MRT_ADD_VIF:
605 case MRT_DEL_VIF:
606 case MRT_ADD_MFC:
607 case MRT_DEL_MFC:
608 case MRT_VERSION:
609 case MRT_ASSERT:
610 case MRT_API_SUPPORT:
611 case MRT_API_CONFIG:
612 case MRT_ADD_BW_UPCALL:
613 case MRT_DEL_BW_UPCALL:
614 error = priv_check(curthread, PRIV_NETINET_MROUTE);
615 if (error != 0)
616 return (error);
617 error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
618 EOPNOTSUPP;
619 break;
620
621 default:
622 error = ip_ctloutput(so, sopt);
623 break;
624 }
625 break;
626
627 case SOPT_SET:
628 switch (sopt->sopt_name) {
629 case IP_HDRINCL:
630 error = sooptcopyin(sopt, &optval, sizeof optval,
631 sizeof optval);
632 if (error)
633 break;
634 if (optval)
635 inp->inp_flags |= INP_HDRINCL;
636 else
637 inp->inp_flags &= ~INP_HDRINCL;
638 break;
639
640 case IP_FW3: /* generic ipfw v.3 functions */
641 case IP_FW_ADD:
642 case IP_FW_DEL:
643 case IP_FW_FLUSH:
644 case IP_FW_ZERO:
645 case IP_FW_RESETLOG:
646 case IP_FW_TABLE_ADD:
647 case IP_FW_TABLE_DEL:
648 case IP_FW_TABLE_FLUSH:
649 case IP_FW_NAT_CFG:
650 case IP_FW_NAT_DEL:
651 if (V_ip_fw_ctl_ptr != NULL)
652 error = V_ip_fw_ctl_ptr(sopt);
653 else
654 error = ENOPROTOOPT;
655 break;
656
657 case IP_DUMMYNET3: /* generic dummynet v.3 functions */
658 case IP_DUMMYNET_CONFIGURE:
659 case IP_DUMMYNET_DEL:
660 case IP_DUMMYNET_FLUSH:
661 if (ip_dn_ctl_ptr != NULL)
662 error = ip_dn_ctl_ptr(sopt);
663 else
664 error = ENOPROTOOPT ;
665 break ;
666
667 case IP_RSVP_ON:
668 error = priv_check(curthread, PRIV_NETINET_MROUTE);
669 if (error != 0)
670 return (error);
671 error = ip_rsvp_init(so);
672 break;
673
674 case IP_RSVP_OFF:
675 error = priv_check(curthread, PRIV_NETINET_MROUTE);
676 if (error != 0)
677 return (error);
678 error = ip_rsvp_done();
679 break;
680
681 case IP_RSVP_VIF_ON:
682 case IP_RSVP_VIF_OFF:
683 error = priv_check(curthread, PRIV_NETINET_MROUTE);
684 if (error != 0)
685 return (error);
686 error = ip_rsvp_vif ?
687 ip_rsvp_vif(so, sopt) : EINVAL;
688 break;
689
690 case MRT_INIT:
691 case MRT_DONE:
692 case MRT_ADD_VIF:
693 case MRT_DEL_VIF:
694 case MRT_ADD_MFC:
695 case MRT_DEL_MFC:
696 case MRT_VERSION:
697 case MRT_ASSERT:
698 case MRT_API_SUPPORT:
699 case MRT_API_CONFIG:
700 case MRT_ADD_BW_UPCALL:
701 case MRT_DEL_BW_UPCALL:
702 error = priv_check(curthread, PRIV_NETINET_MROUTE);
703 if (error != 0)
704 return (error);
705 error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
706 EOPNOTSUPP;
707 break;
708
709 default:
710 error = ip_ctloutput(so, sopt);
711 break;
712 }
713 break;
714 }
715
716 return (error);
717 }
718
719 /*
720 * This function exists solely to receive the PRC_IFDOWN messages which are
721 * sent by if_down(). It looks for an ifaddr whose ifa_addr is sa, and calls
722 * in_ifadown() to remove all routes corresponding to that address. It also
723 * receives the PRC_IFUP messages from if_up() and reinstalls the interface
724 * routes.
725 */
726 void
rip_ctlinput(int cmd,struct sockaddr * sa,void * vip)727 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip)
728 {
729 struct rm_priotracker in_ifa_tracker;
730 struct in_ifaddr *ia;
731 struct ifnet *ifp;
732 int err;
733 int flags;
734
735 switch (cmd) {
736 case PRC_IFDOWN:
737 IN_IFADDR_RLOCK(&in_ifa_tracker);
738 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
739 if (ia->ia_ifa.ifa_addr == sa
740 && (ia->ia_flags & IFA_ROUTE)) {
741 ifa_ref(&ia->ia_ifa);
742 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
743 /*
744 * in_scrubprefix() kills the interface route.
745 */
746 in_scrubprefix(ia, 0);
747 /*
748 * in_ifadown gets rid of all the rest of the
749 * routes. This is not quite the right thing
750 * to do, but at least if we are running a
751 * routing process they will come back.
752 */
753 in_ifadown(&ia->ia_ifa, 0);
754 ifa_free(&ia->ia_ifa);
755 break;
756 }
757 }
758 if (ia == NULL) /* If ia matched, already unlocked. */
759 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
760 break;
761
762 case PRC_IFUP:
763 IN_IFADDR_RLOCK(&in_ifa_tracker);
764 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
765 if (ia->ia_ifa.ifa_addr == sa)
766 break;
767 }
768 if (ia == NULL || (ia->ia_flags & IFA_ROUTE)) {
769 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
770 return;
771 }
772 ifa_ref(&ia->ia_ifa);
773 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
774 flags = RTF_UP;
775 ifp = ia->ia_ifa.ifa_ifp;
776
777 if ((ifp->if_flags & IFF_LOOPBACK)
778 || (ifp->if_flags & IFF_POINTOPOINT))
779 flags |= RTF_HOST;
780
781 err = ifa_del_loopback_route((struct ifaddr *)ia, sa);
782
783 err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
784 if (err == 0)
785 ia->ia_flags |= IFA_ROUTE;
786
787 err = ifa_add_loopback_route((struct ifaddr *)ia, sa);
788
789 ifa_free(&ia->ia_ifa);
790 break;
791 }
792 }
793
794 static int
rip_attach(struct socket * so,int proto,struct thread * td)795 rip_attach(struct socket *so, int proto, struct thread *td)
796 {
797 struct inpcb *inp;
798 int error;
799
800 inp = sotoinpcb(so);
801 KASSERT(inp == NULL, ("rip_attach: inp != NULL"));
802
803 error = priv_check(td, PRIV_NETINET_RAW);
804 if (error)
805 return (error);
806 if (proto >= IPPROTO_MAX || proto < 0)
807 return EPROTONOSUPPORT;
808 error = soreserve(so, rip_sendspace, rip_recvspace);
809 if (error)
810 return (error);
811 INP_INFO_WLOCK(&V_ripcbinfo);
812 error = in_pcballoc(so, &V_ripcbinfo);
813 if (error) {
814 INP_INFO_WUNLOCK(&V_ripcbinfo);
815 return (error);
816 }
817 inp = (struct inpcb *)so->so_pcb;
818 inp->inp_vflag |= INP_IPV4;
819 inp->inp_ip_p = proto;
820 inp->inp_ip_ttl = V_ip_defttl;
821 rip_inshash(inp);
822 INP_INFO_WUNLOCK(&V_ripcbinfo);
823 INP_WUNLOCK(inp);
824 return (0);
825 }
826
827 static void
rip_detach(struct socket * so)828 rip_detach(struct socket *so)
829 {
830 struct inpcb *inp;
831
832 inp = sotoinpcb(so);
833 KASSERT(inp != NULL, ("rip_detach: inp == NULL"));
834 KASSERT(inp->inp_faddr.s_addr == INADDR_ANY,
835 ("rip_detach: not closed"));
836
837 INP_INFO_WLOCK(&V_ripcbinfo);
838 INP_WLOCK(inp);
839 rip_delhash(inp);
840 if (so == V_ip_mrouter && ip_mrouter_done)
841 ip_mrouter_done();
842 if (ip_rsvp_force_done)
843 ip_rsvp_force_done(so);
844 if (so == V_ip_rsvpd)
845 ip_rsvp_done();
846 in_pcbdetach(inp);
847 in_pcbfree(inp);
848 INP_INFO_WUNLOCK(&V_ripcbinfo);
849 }
850
851 static void
rip_dodisconnect(struct socket * so,struct inpcb * inp)852 rip_dodisconnect(struct socket *so, struct inpcb *inp)
853 {
854 struct inpcbinfo *pcbinfo;
855
856 pcbinfo = inp->inp_pcbinfo;
857 INP_INFO_WLOCK(pcbinfo);
858 INP_WLOCK(inp);
859 rip_delhash(inp);
860 inp->inp_faddr.s_addr = INADDR_ANY;
861 rip_inshash(inp);
862 SOCK_LOCK(so);
863 so->so_state &= ~SS_ISCONNECTED;
864 SOCK_UNLOCK(so);
865 INP_WUNLOCK(inp);
866 INP_INFO_WUNLOCK(pcbinfo);
867 }
868
869 static void
rip_abort(struct socket * so)870 rip_abort(struct socket *so)
871 {
872 struct inpcb *inp;
873
874 inp = sotoinpcb(so);
875 KASSERT(inp != NULL, ("rip_abort: inp == NULL"));
876
877 rip_dodisconnect(so, inp);
878 }
879
880 static void
rip_close(struct socket * so)881 rip_close(struct socket *so)
882 {
883 struct inpcb *inp;
884
885 inp = sotoinpcb(so);
886 KASSERT(inp != NULL, ("rip_close: inp == NULL"));
887
888 rip_dodisconnect(so, inp);
889 }
890
891 static int
rip_disconnect(struct socket * so)892 rip_disconnect(struct socket *so)
893 {
894 struct inpcb *inp;
895
896 if ((so->so_state & SS_ISCONNECTED) == 0)
897 return (ENOTCONN);
898
899 inp = sotoinpcb(so);
900 KASSERT(inp != NULL, ("rip_disconnect: inp == NULL"));
901
902 rip_dodisconnect(so, inp);
903 return (0);
904 }
905
906 static int
rip_bind(struct socket * so,struct sockaddr * nam,struct thread * td)907 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
908 {
909 struct sockaddr_in *addr = (struct sockaddr_in *)nam;
910 struct inpcb *inp;
911 int error;
912
913 if (nam->sa_len != sizeof(*addr))
914 return (EINVAL);
915
916 error = prison_check_ip4(td->td_ucred, &addr->sin_addr);
917 if (error != 0)
918 return (error);
919
920 inp = sotoinpcb(so);
921 KASSERT(inp != NULL, ("rip_bind: inp == NULL"));
922
923 if (TAILQ_EMPTY(&V_ifnet) ||
924 (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
925 (addr->sin_addr.s_addr &&
926 (inp->inp_flags & INP_BINDANY) == 0 &&
927 ifa_ifwithaddr_check((struct sockaddr *)addr) == 0))
928 return (EADDRNOTAVAIL);
929
930 INP_INFO_WLOCK(&V_ripcbinfo);
931 INP_WLOCK(inp);
932 rip_delhash(inp);
933 inp->inp_laddr = addr->sin_addr;
934 rip_inshash(inp);
935 INP_WUNLOCK(inp);
936 INP_INFO_WUNLOCK(&V_ripcbinfo);
937 return (0);
938 }
939
940 static int
rip_connect(struct socket * so,struct sockaddr * nam,struct thread * td)941 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
942 {
943 struct sockaddr_in *addr = (struct sockaddr_in *)nam;
944 struct inpcb *inp;
945
946 if (nam->sa_len != sizeof(*addr))
947 return (EINVAL);
948 if (TAILQ_EMPTY(&V_ifnet))
949 return (EADDRNOTAVAIL);
950 if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
951 return (EAFNOSUPPORT);
952
953 inp = sotoinpcb(so);
954 KASSERT(inp != NULL, ("rip_connect: inp == NULL"));
955
956 INP_INFO_WLOCK(&V_ripcbinfo);
957 INP_WLOCK(inp);
958 rip_delhash(inp);
959 inp->inp_faddr = addr->sin_addr;
960 rip_inshash(inp);
961 soisconnected(so);
962 INP_WUNLOCK(inp);
963 INP_INFO_WUNLOCK(&V_ripcbinfo);
964 return (0);
965 }
966
967 static int
rip_shutdown(struct socket * so)968 rip_shutdown(struct socket *so)
969 {
970 struct inpcb *inp;
971
972 inp = sotoinpcb(so);
973 KASSERT(inp != NULL, ("rip_shutdown: inp == NULL"));
974
975 INP_WLOCK(inp);
976 socantsendmore(so);
977 INP_WUNLOCK(inp);
978 return (0);
979 }
980
981 static int
rip_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct thread * td)982 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
983 struct mbuf *control, struct thread *td)
984 {
985 struct inpcb *inp;
986 u_long dst;
987
988 inp = sotoinpcb(so);
989 KASSERT(inp != NULL, ("rip_send: inp == NULL"));
990
991 /*
992 * Note: 'dst' reads below are unlocked.
993 */
994 if (so->so_state & SS_ISCONNECTED) {
995 if (nam) {
996 m_freem(m);
997 return (EISCONN);
998 }
999 dst = inp->inp_faddr.s_addr; /* Unlocked read. */
1000 } else {
1001 if (nam == NULL) {
1002 m_freem(m);
1003 return (ENOTCONN);
1004 }
1005 dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr;
1006 }
1007 return (rip_output(m, so, dst));
1008 }
1009 #endif /* INET */
1010
1011 static int
rip_pcblist(SYSCTL_HANDLER_ARGS)1012 rip_pcblist(SYSCTL_HANDLER_ARGS)
1013 {
1014 int error, i, n;
1015 struct inpcb *inp, **inp_list;
1016 inp_gen_t gencnt;
1017 struct xinpgen xig;
1018
1019 /*
1020 * The process of preparing the TCB list is too time-consuming and
1021 * resource-intensive to repeat twice on every request.
1022 */
1023 if (req->oldptr == 0) {
1024 n = V_ripcbinfo.ipi_count;
1025 n += imax(n / 8, 10);
1026 req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb);
1027 return (0);
1028 }
1029
1030 if (req->newptr != 0)
1031 return (EPERM);
1032
1033 /*
1034 * OK, now we're committed to doing something.
1035 */
1036 INP_INFO_RLOCK(&V_ripcbinfo);
1037 gencnt = V_ripcbinfo.ipi_gencnt;
1038 n = V_ripcbinfo.ipi_count;
1039 INP_INFO_RUNLOCK(&V_ripcbinfo);
1040
1041 xig.xig_len = sizeof xig;
1042 xig.xig_count = n;
1043 xig.xig_gen = gencnt;
1044 xig.xig_sogen = so_gencnt;
1045 error = SYSCTL_OUT(req, &xig, sizeof xig);
1046 if (error)
1047 return (error);
1048
1049 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
1050 if (inp_list == 0)
1051 return (ENOMEM);
1052
1053 INP_INFO_RLOCK(&V_ripcbinfo);
1054 for (inp = LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n;
1055 inp = LIST_NEXT(inp, inp_list)) {
1056 INP_WLOCK(inp);
1057 if (inp->inp_gencnt <= gencnt &&
1058 cr_canseeinpcb(req->td->td_ucred, inp) == 0) {
1059 in_pcbref(inp);
1060 inp_list[i++] = inp;
1061 }
1062 INP_WUNLOCK(inp);
1063 }
1064 INP_INFO_RUNLOCK(&V_ripcbinfo);
1065 n = i;
1066
1067 error = 0;
1068 for (i = 0; i < n; i++) {
1069 inp = inp_list[i];
1070 INP_RLOCK(inp);
1071 if (inp->inp_gencnt <= gencnt) {
1072 struct xinpcb xi;
1073
1074 bzero(&xi, sizeof(xi));
1075 xi.xi_len = sizeof xi;
1076 /* XXX should avoid extra copy */
1077 bcopy(inp, &xi.xi_inp, sizeof *inp);
1078 if (inp->inp_socket)
1079 sotoxsocket(inp->inp_socket, &xi.xi_socket);
1080 INP_RUNLOCK(inp);
1081 error = SYSCTL_OUT(req, &xi, sizeof xi);
1082 } else
1083 INP_RUNLOCK(inp);
1084 }
1085 INP_INFO_WLOCK(&V_ripcbinfo);
1086 for (i = 0; i < n; i++) {
1087 inp = inp_list[i];
1088 INP_RLOCK(inp);
1089 if (!in_pcbrele_rlocked(inp))
1090 INP_RUNLOCK(inp);
1091 }
1092 INP_INFO_WUNLOCK(&V_ripcbinfo);
1093
1094 if (!error) {
1095 /*
1096 * Give the user an updated idea of our state. If the
1097 * generation differs from what we told her before, she knows
1098 * that something happened while we were processing this
1099 * request, and it might be necessary to retry.
1100 */
1101 INP_INFO_RLOCK(&V_ripcbinfo);
1102 xig.xig_gen = V_ripcbinfo.ipi_gencnt;
1103 xig.xig_sogen = so_gencnt;
1104 xig.xig_count = V_ripcbinfo.ipi_count;
1105 INP_INFO_RUNLOCK(&V_ripcbinfo);
1106 error = SYSCTL_OUT(req, &xig, sizeof xig);
1107 }
1108 free(inp_list, M_TEMP);
1109 return (error);
1110 }
1111
1112 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist,
1113 CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0,
1114 rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
1115
1116 #ifdef INET
1117 struct pr_usrreqs rip_usrreqs = {
1118 .pru_abort = rip_abort,
1119 .pru_attach = rip_attach,
1120 .pru_bind = rip_bind,
1121 .pru_connect = rip_connect,
1122 .pru_control = in_control,
1123 .pru_detach = rip_detach,
1124 .pru_disconnect = rip_disconnect,
1125 .pru_peeraddr = in_getpeeraddr,
1126 .pru_send = rip_send,
1127 .pru_shutdown = rip_shutdown,
1128 .pru_sockaddr = in_getsockaddr,
1129 .pru_sosetlabel = in_pcbsosetlabel,
1130 .pru_close = rip_close,
1131 };
1132 #endif /* INET */
1133