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 #include <netinet/ip_icmp.h>
75
76 #include <netipsec/ipsec_support.h>
77
78 #include <machine/stdarg.h>
79 #include <security/mac/mac_framework.h>
80
81 VNET_DEFINE(int, ip_defttl) = IPDEFTTL;
82 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_VNET | CTLFLAG_RW,
83 &VNET_NAME(ip_defttl), 0,
84 "Maximum TTL on IP packets");
85
86 VNET_DEFINE(struct inpcbhead, ripcb);
87 VNET_DEFINE(struct inpcbinfo, ripcbinfo);
88
89 #define V_ripcb VNET(ripcb)
90 #define V_ripcbinfo VNET(ripcbinfo)
91
92 /*
93 * Control and data hooks for ipfw, dummynet, divert and so on.
94 * The data hooks are not used here but it is convenient
95 * to keep them all in one place.
96 */
97 VNET_DEFINE(ip_fw_chk_ptr_t, ip_fw_chk_ptr) = NULL;
98 VNET_DEFINE(ip_fw_ctl_ptr_t, ip_fw_ctl_ptr) = NULL;
99
100 int (*ip_dn_ctl_ptr)(struct sockopt *);
101 int (*ip_dn_io_ptr)(struct mbuf **, int, struct ip_fw_args *);
102 void (*ip_divert_ptr)(struct mbuf *, int);
103 int (*ng_ipfw_input_p)(struct mbuf **, int,
104 struct ip_fw_args *, int);
105
106 #ifdef INET
107 /*
108 * Hooks for multicast routing. They all default to NULL, so leave them not
109 * initialized and rely on BSS being set to 0.
110 */
111
112 /*
113 * The socket used to communicate with the multicast routing daemon.
114 */
115 VNET_DEFINE(struct socket *, ip_mrouter);
116
117 /*
118 * The various mrouter and rsvp functions.
119 */
120 int (*ip_mrouter_set)(struct socket *, struct sockopt *);
121 int (*ip_mrouter_get)(struct socket *, struct sockopt *);
122 int (*ip_mrouter_done)(void);
123 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
124 struct ip_moptions *);
125 int (*mrt_ioctl)(u_long, caddr_t, int);
126 int (*legal_vif_num)(int);
127 u_long (*ip_mcast_src)(int);
128
129 int (*rsvp_input_p)(struct mbuf **, int *, int);
130 int (*ip_rsvp_vif)(struct socket *, struct sockopt *);
131 void (*ip_rsvp_force_done)(struct socket *);
132 #endif /* INET */
133
134 extern struct protosw inetsw[];
135
136 u_long rip_sendspace = 9216;
137 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
138 &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
139
140 u_long rip_recvspace = 9216;
141 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW,
142 &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams");
143
144 /*
145 * Hash functions
146 */
147
148 #define INP_PCBHASH_RAW_SIZE 256
149 #define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \
150 (((proto) + (laddr) + (faddr)) % (mask) + 1)
151
152 #ifdef INET
153 static void
rip_inshash(struct inpcb * inp)154 rip_inshash(struct inpcb *inp)
155 {
156 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
157 struct inpcbhead *pcbhash;
158 int hash;
159
160 INP_INFO_WLOCK_ASSERT(pcbinfo);
161 INP_WLOCK_ASSERT(inp);
162
163 if (inp->inp_ip_p != 0 &&
164 inp->inp_laddr.s_addr != INADDR_ANY &&
165 inp->inp_faddr.s_addr != INADDR_ANY) {
166 hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr,
167 inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask);
168 } else
169 hash = 0;
170 pcbhash = &pcbinfo->ipi_hashbase[hash];
171 LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
172 }
173
174 static void
rip_delhash(struct inpcb * inp)175 rip_delhash(struct inpcb *inp)
176 {
177
178 INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
179 INP_WLOCK_ASSERT(inp);
180
181 LIST_REMOVE(inp, inp_hash);
182 }
183 #endif /* INET */
184
185 /*
186 * Raw interface to IP protocol.
187 */
188
189 /*
190 * Initialize raw connection block q.
191 */
192 static void
rip_zone_change(void * tag)193 rip_zone_change(void *tag)
194 {
195
196 uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets);
197 }
198
199 static int
rip_inpcb_init(void * mem,int size,int flags)200 rip_inpcb_init(void *mem, int size, int flags)
201 {
202 struct inpcb *inp = mem;
203
204 INP_LOCK_INIT(inp, "inp", "rawinp");
205 return (0);
206 }
207
208 void
rip_init(void)209 rip_init(void)
210 {
211
212 in_pcbinfo_init(&V_ripcbinfo, "rip", &V_ripcb, INP_PCBHASH_RAW_SIZE,
213 1, "ripcb", rip_inpcb_init, NULL, 0, IPI_HASHFIELDS_NONE);
214 EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL,
215 EVENTHANDLER_PRI_ANY);
216 }
217
218 #ifdef VIMAGE
219 static void
rip_destroy(void * unused __unused)220 rip_destroy(void *unused __unused)
221 {
222
223 in_pcbinfo_destroy(&V_ripcbinfo);
224 }
225 VNET_SYSUNINIT(raw_ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, rip_destroy, NULL);
226 #endif
227
228 #ifdef INET
229 static int
rip_append(struct inpcb * last,struct ip * ip,struct mbuf * n,struct sockaddr_in * ripsrc)230 rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n,
231 struct sockaddr_in *ripsrc)
232 {
233 int policyfail = 0;
234
235 INP_LOCK_ASSERT(last);
236
237 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
238 /* check AH/ESP integrity. */
239 if (IPSEC_ENABLED(ipv4)) {
240 if (IPSEC_CHECK_POLICY(ipv4, n, last) != 0)
241 policyfail = 1;
242 }
243 #endif /* IPSEC */
244 #ifdef MAC
245 if (!policyfail && mac_inpcb_check_deliver(last, n) != 0)
246 policyfail = 1;
247 #endif
248 /* Check the minimum TTL for socket. */
249 if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl)
250 policyfail = 1;
251 if (!policyfail) {
252 struct mbuf *opts = NULL;
253 struct socket *so;
254
255 so = last->inp_socket;
256 if ((last->inp_flags & INP_CONTROLOPTS) ||
257 (so->so_options & (SO_TIMESTAMP | SO_BINTIME)))
258 ip_savecontrol(last, &opts, ip, n);
259 SOCKBUF_LOCK(&so->so_rcv);
260 if (sbappendaddr_locked(&so->so_rcv,
261 (struct sockaddr *)ripsrc, n, opts) == 0) {
262 /* should notify about lost packet */
263 m_freem(n);
264 if (opts)
265 m_freem(opts);
266 SOCKBUF_UNLOCK(&so->so_rcv);
267 } else
268 sorwakeup_locked(so);
269 } else
270 m_freem(n);
271 return (policyfail);
272 }
273
274 /*
275 * Setup generic address and protocol structures for raw_input routine, then
276 * pass them along with mbuf chain.
277 */
278 int
rip_input(struct mbuf ** mp,int * offp,int proto)279 rip_input(struct mbuf **mp, int *offp, int proto)
280 {
281 struct ifnet *ifp;
282 struct mbuf *m = *mp;
283 struct ip *ip = mtod(m, struct ip *);
284 struct inpcb *inp, *last;
285 struct sockaddr_in ripsrc;
286 int hash;
287
288 *mp = NULL;
289
290 bzero(&ripsrc, sizeof(ripsrc));
291 ripsrc.sin_len = sizeof(ripsrc);
292 ripsrc.sin_family = AF_INET;
293 ripsrc.sin_addr = ip->ip_src;
294 last = NULL;
295
296 ifp = m->m_pkthdr.rcvif;
297
298 hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr,
299 ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask);
300 INP_INFO_RLOCK(&V_ripcbinfo);
301 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) {
302 if (inp->inp_ip_p != proto)
303 continue;
304 #ifdef INET6
305 /* XXX inp locking */
306 if ((inp->inp_vflag & INP_IPV4) == 0)
307 continue;
308 #endif
309 if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
310 continue;
311 if (inp->inp_faddr.s_addr != ip->ip_src.s_addr)
312 continue;
313 if (jailed_without_vnet(inp->inp_cred)) {
314 /*
315 * XXX: If faddr was bound to multicast group,
316 * jailed raw socket will drop datagram.
317 */
318 if (prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0)
319 continue;
320 }
321 if (last != NULL) {
322 struct mbuf *n;
323
324 n = m_copy(m, 0, (int)M_COPYALL);
325 if (n != NULL)
326 (void) rip_append(last, ip, n, &ripsrc);
327 /* XXX count dropped packet */
328 INP_RUNLOCK(last);
329 }
330 INP_RLOCK(inp);
331 last = inp;
332 }
333 LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) {
334 if (inp->inp_ip_p && inp->inp_ip_p != proto)
335 continue;
336 #ifdef INET6
337 /* XXX inp locking */
338 if ((inp->inp_vflag & INP_IPV4) == 0)
339 continue;
340 #endif
341 if (!in_nullhost(inp->inp_laddr) &&
342 !in_hosteq(inp->inp_laddr, ip->ip_dst))
343 continue;
344 if (!in_nullhost(inp->inp_faddr) &&
345 !in_hosteq(inp->inp_faddr, ip->ip_src))
346 continue;
347 if (jailed_without_vnet(inp->inp_cred)) {
348 /*
349 * Allow raw socket in jail to receive multicast;
350 * assume process had PRIV_NETINET_RAW at attach,
351 * and fall through into normal filter path if so.
352 */
353 if (!IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
354 prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0)
355 continue;
356 }
357 /*
358 * If this raw socket has multicast state, and we
359 * have received a multicast, check if this socket
360 * should receive it, as multicast filtering is now
361 * the responsibility of the transport layer.
362 */
363 if (inp->inp_moptions != NULL &&
364 IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
365 /*
366 * If the incoming datagram is for IGMP, allow it
367 * through unconditionally to the raw socket.
368 *
369 * In the case of IGMPv2, we may not have explicitly
370 * joined the group, and may have set IFF_ALLMULTI
371 * on the interface. imo_multi_filter() may discard
372 * control traffic we actually need to see.
373 *
374 * Userland multicast routing daemons should continue
375 * filter the control traffic appropriately.
376 */
377 int blocked;
378
379 blocked = MCAST_PASS;
380 if (proto != IPPROTO_IGMP) {
381 struct sockaddr_in group;
382
383 bzero(&group, sizeof(struct sockaddr_in));
384 group.sin_len = sizeof(struct sockaddr_in);
385 group.sin_family = AF_INET;
386 group.sin_addr = ip->ip_dst;
387
388 blocked = imo_multi_filter(inp->inp_moptions,
389 ifp,
390 (struct sockaddr *)&group,
391 (struct sockaddr *)&ripsrc);
392 }
393
394 if (blocked != MCAST_PASS) {
395 IPSTAT_INC(ips_notmember);
396 continue;
397 }
398 }
399 if (last != NULL) {
400 struct mbuf *n;
401
402 n = m_copy(m, 0, (int)M_COPYALL);
403 if (n != NULL)
404 (void) rip_append(last, ip, n, &ripsrc);
405 /* XXX count dropped packet */
406 INP_RUNLOCK(last);
407 }
408 INP_RLOCK(inp);
409 last = inp;
410 }
411 INP_INFO_RUNLOCK(&V_ripcbinfo);
412 if (last != NULL) {
413 if (rip_append(last, ip, m, &ripsrc) != 0)
414 IPSTAT_INC(ips_delivered);
415 INP_RUNLOCK(last);
416 } else {
417 if (inetsw[ip_protox[ip->ip_p]].pr_input == rip_input) {
418 IPSTAT_INC(ips_noproto);
419 IPSTAT_DEC(ips_delivered);
420 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PROTOCOL, 0, 0);
421 } else {
422 m_freem(m);
423 }
424 }
425 return (IPPROTO_DONE);
426 }
427
428 /*
429 * Generate IP header and pass packet to ip_output. Tack on options user may
430 * have setup with control call.
431 */
432 int
rip_output(struct mbuf * m,struct socket * so,...)433 rip_output(struct mbuf *m, struct socket *so, ...)
434 {
435 struct ip *ip;
436 int error;
437 struct inpcb *inp = sotoinpcb(so);
438 va_list ap;
439 u_long dst;
440 int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) |
441 IP_ALLOWBROADCAST;
442 int cnt, hlen;
443 u_char opttype, optlen, *cp;
444
445 va_start(ap, so);
446 dst = va_arg(ap, u_long);
447 va_end(ap);
448
449 /*
450 * If the user handed us a complete IP packet, use it. Otherwise,
451 * allocate an mbuf for a header and fill it in.
452 */
453 if ((inp->inp_flags & INP_HDRINCL) == 0) {
454 if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
455 m_freem(m);
456 return(EMSGSIZE);
457 }
458 M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
459 if (m == NULL)
460 return(ENOBUFS);
461
462 INP_RLOCK(inp);
463 ip = mtod(m, struct ip *);
464 ip->ip_tos = inp->inp_ip_tos;
465 if (inp->inp_flags & INP_DONTFRAG)
466 ip->ip_off = htons(IP_DF);
467 else
468 ip->ip_off = htons(0);
469 ip->ip_p = inp->inp_ip_p;
470 ip->ip_len = htons(m->m_pkthdr.len);
471 ip->ip_src = inp->inp_laddr;
472 ip->ip_dst.s_addr = dst;
473 if (jailed(inp->inp_cred)) {
474 /*
475 * prison_local_ip4() would be good enough but would
476 * let a source of INADDR_ANY pass, which we do not
477 * want to see from jails.
478 */
479 if (ip->ip_src.s_addr == INADDR_ANY) {
480 error = in_pcbladdr(inp, &ip->ip_dst, &ip->ip_src,
481 inp->inp_cred);
482 } else {
483 error = prison_local_ip4(inp->inp_cred,
484 &ip->ip_src);
485 }
486 if (error != 0) {
487 INP_RUNLOCK(inp);
488 m_freem(m);
489 return (error);
490 }
491 }
492 ip->ip_ttl = inp->inp_ip_ttl;
493 } else {
494 if (m->m_pkthdr.len > IP_MAXPACKET) {
495 m_freem(m);
496 return(EMSGSIZE);
497 }
498 ip = mtod(m, struct ip *);
499 hlen = ip->ip_hl << 2;
500 if (m->m_len < hlen) {
501 m = m_pullup(m, hlen);
502 if (m == NULL)
503 return (EINVAL);
504 ip = mtod(m, struct ip *);
505 }
506
507 INP_RLOCK(inp);
508 /*
509 * Don't allow both user specified and setsockopt options,
510 * and don't allow packet length sizes that will crash.
511 */
512 if ((hlen < sizeof (*ip))
513 || ((hlen > sizeof (*ip)) && inp->inp_options)
514 || (ntohs(ip->ip_len) != m->m_pkthdr.len)) {
515 INP_RUNLOCK(inp);
516 m_freem(m);
517 return (EINVAL);
518 }
519 error = prison_check_ip4(inp->inp_cred, &ip->ip_src);
520 if (error != 0) {
521 INP_RUNLOCK(inp);
522 m_freem(m);
523 return (error);
524 }
525 /*
526 * Don't allow IP options which do not have the required
527 * structure as specified in section 3.1 of RFC 791 on
528 * pages 15-23.
529 */
530 cp = (u_char *)(ip + 1);
531 cnt = hlen - sizeof (struct ip);
532 for (; cnt > 0; cnt -= optlen, cp += optlen) {
533 opttype = cp[IPOPT_OPTVAL];
534 if (opttype == IPOPT_EOL)
535 break;
536 if (opttype == IPOPT_NOP) {
537 optlen = 1;
538 continue;
539 }
540 if (cnt < IPOPT_OLEN + sizeof(u_char)) {
541 INP_RUNLOCK(inp);
542 m_freem(m);
543 return (EINVAL);
544 }
545 optlen = cp[IPOPT_OLEN];
546 if (optlen < IPOPT_OLEN + sizeof(u_char) ||
547 optlen > cnt) {
548 INP_RUNLOCK(inp);
549 m_freem(m);
550 return (EINVAL);
551 }
552 }
553 /*
554 * This doesn't allow application to specify ID of zero,
555 * but we got this limitation from the beginning of history.
556 */
557 if (ip->ip_id == 0)
558 ip_fillid(ip);
559
560 /*
561 * XXX prevent ip_output from overwriting header fields.
562 */
563 flags |= IP_RAWOUTPUT;
564 IPSTAT_INC(ips_rawout);
565 }
566
567 if (inp->inp_flags & INP_ONESBCAST)
568 flags |= IP_SENDONES;
569
570 #ifdef MAC
571 mac_inpcb_create_mbuf(inp, m);
572 #endif
573
574 error = ip_output(m, inp->inp_options, NULL, flags,
575 inp->inp_moptions, inp);
576 INP_RUNLOCK(inp);
577 return (error);
578 }
579
580 /*
581 * Raw IP socket option processing.
582 *
583 * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could
584 * only be created by a privileged process, and as such, socket option
585 * operations to manage system properties on any raw socket were allowed to
586 * take place without explicit additional access control checks. However,
587 * raw sockets can now also be created in jail(), and therefore explicit
588 * checks are now required. Likewise, raw sockets can be used by a process
589 * after it gives up privilege, so some caution is required. For options
590 * passed down to the IP layer via ip_ctloutput(), checks are assumed to be
591 * performed in ip_ctloutput() and therefore no check occurs here.
592 * Unilaterally checking priv_check() here breaks normal IP socket option
593 * operations on raw sockets.
594 *
595 * When adding new socket options here, make sure to add access control
596 * checks here as necessary.
597 *
598 * XXX-BZ inp locking?
599 */
600 int
rip_ctloutput(struct socket * so,struct sockopt * sopt)601 rip_ctloutput(struct socket *so, struct sockopt *sopt)
602 {
603 struct inpcb *inp = sotoinpcb(so);
604 int error, optval;
605
606 if (sopt->sopt_level != IPPROTO_IP) {
607 if ((sopt->sopt_level == SOL_SOCKET) &&
608 (sopt->sopt_name == SO_SETFIB)) {
609 inp->inp_inc.inc_fibnum = so->so_fibnum;
610 return (0);
611 }
612 return (EINVAL);
613 }
614
615 error = 0;
616 switch (sopt->sopt_dir) {
617 case SOPT_GET:
618 switch (sopt->sopt_name) {
619 case IP_HDRINCL:
620 optval = inp->inp_flags & INP_HDRINCL;
621 error = sooptcopyout(sopt, &optval, sizeof optval);
622 break;
623
624 case IP_FW3: /* generic ipfw v.3 functions */
625 case IP_FW_ADD: /* ADD actually returns the body... */
626 case IP_FW_GET:
627 case IP_FW_TABLE_GETSIZE:
628 case IP_FW_TABLE_LIST:
629 case IP_FW_NAT_GET_CONFIG:
630 case IP_FW_NAT_GET_LOG:
631 if (V_ip_fw_ctl_ptr != NULL)
632 error = V_ip_fw_ctl_ptr(sopt);
633 else
634 error = ENOPROTOOPT;
635 break;
636
637 case IP_DUMMYNET3: /* generic dummynet v.3 functions */
638 case IP_DUMMYNET_GET:
639 if (ip_dn_ctl_ptr != NULL)
640 error = ip_dn_ctl_ptr(sopt);
641 else
642 error = ENOPROTOOPT;
643 break ;
644
645 case MRT_INIT:
646 case MRT_DONE:
647 case MRT_ADD_VIF:
648 case MRT_DEL_VIF:
649 case MRT_ADD_MFC:
650 case MRT_DEL_MFC:
651 case MRT_VERSION:
652 case MRT_ASSERT:
653 case MRT_API_SUPPORT:
654 case MRT_API_CONFIG:
655 case MRT_ADD_BW_UPCALL:
656 case MRT_DEL_BW_UPCALL:
657 error = priv_check(curthread, PRIV_NETINET_MROUTE);
658 if (error != 0)
659 return (error);
660 error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
661 EOPNOTSUPP;
662 break;
663
664 default:
665 error = ip_ctloutput(so, sopt);
666 break;
667 }
668 break;
669
670 case SOPT_SET:
671 switch (sopt->sopt_name) {
672 case IP_HDRINCL:
673 error = sooptcopyin(sopt, &optval, sizeof optval,
674 sizeof optval);
675 if (error)
676 break;
677 if (optval)
678 inp->inp_flags |= INP_HDRINCL;
679 else
680 inp->inp_flags &= ~INP_HDRINCL;
681 break;
682
683 case IP_FW3: /* generic ipfw v.3 functions */
684 case IP_FW_ADD:
685 case IP_FW_DEL:
686 case IP_FW_FLUSH:
687 case IP_FW_ZERO:
688 case IP_FW_RESETLOG:
689 case IP_FW_TABLE_ADD:
690 case IP_FW_TABLE_DEL:
691 case IP_FW_TABLE_FLUSH:
692 case IP_FW_NAT_CFG:
693 case IP_FW_NAT_DEL:
694 if (V_ip_fw_ctl_ptr != NULL)
695 error = V_ip_fw_ctl_ptr(sopt);
696 else
697 error = ENOPROTOOPT;
698 break;
699
700 case IP_DUMMYNET3: /* generic dummynet v.3 functions */
701 case IP_DUMMYNET_CONFIGURE:
702 case IP_DUMMYNET_DEL:
703 case IP_DUMMYNET_FLUSH:
704 if (ip_dn_ctl_ptr != NULL)
705 error = ip_dn_ctl_ptr(sopt);
706 else
707 error = ENOPROTOOPT ;
708 break ;
709
710 case IP_RSVP_ON:
711 error = priv_check(curthread, PRIV_NETINET_MROUTE);
712 if (error != 0)
713 return (error);
714 error = ip_rsvp_init(so);
715 break;
716
717 case IP_RSVP_OFF:
718 error = priv_check(curthread, PRIV_NETINET_MROUTE);
719 if (error != 0)
720 return (error);
721 error = ip_rsvp_done();
722 break;
723
724 case IP_RSVP_VIF_ON:
725 case IP_RSVP_VIF_OFF:
726 error = priv_check(curthread, PRIV_NETINET_MROUTE);
727 if (error != 0)
728 return (error);
729 error = ip_rsvp_vif ?
730 ip_rsvp_vif(so, sopt) : EINVAL;
731 break;
732
733 case MRT_INIT:
734 case MRT_DONE:
735 case MRT_ADD_VIF:
736 case MRT_DEL_VIF:
737 case MRT_ADD_MFC:
738 case MRT_DEL_MFC:
739 case MRT_VERSION:
740 case MRT_ASSERT:
741 case MRT_API_SUPPORT:
742 case MRT_API_CONFIG:
743 case MRT_ADD_BW_UPCALL:
744 case MRT_DEL_BW_UPCALL:
745 error = priv_check(curthread, PRIV_NETINET_MROUTE);
746 if (error != 0)
747 return (error);
748 error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
749 EOPNOTSUPP;
750 break;
751
752 default:
753 error = ip_ctloutput(so, sopt);
754 break;
755 }
756 break;
757 }
758
759 return (error);
760 }
761
762 /*
763 * This function exists solely to receive the PRC_IFDOWN messages which are
764 * sent by if_down(). It looks for an ifaddr whose ifa_addr is sa, and calls
765 * in_ifadown() to remove all routes corresponding to that address. It also
766 * receives the PRC_IFUP messages from if_up() and reinstalls the interface
767 * routes.
768 */
769 void
rip_ctlinput(int cmd,struct sockaddr * sa,void * vip)770 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip)
771 {
772 struct rm_priotracker in_ifa_tracker;
773 struct in_ifaddr *ia;
774 struct ifnet *ifp;
775 int err;
776 int flags;
777
778 switch (cmd) {
779 case PRC_IFDOWN:
780 IN_IFADDR_RLOCK(&in_ifa_tracker);
781 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
782 if (ia->ia_ifa.ifa_addr == sa
783 && (ia->ia_flags & IFA_ROUTE)) {
784 ifa_ref(&ia->ia_ifa);
785 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
786 /*
787 * in_scrubprefix() kills the interface route.
788 */
789 in_scrubprefix(ia, 0);
790 /*
791 * in_ifadown gets rid of all the rest of the
792 * routes. This is not quite the right thing
793 * to do, but at least if we are running a
794 * routing process they will come back.
795 */
796 in_ifadown(&ia->ia_ifa, 0);
797 ifa_free(&ia->ia_ifa);
798 break;
799 }
800 }
801 if (ia == NULL) /* If ia matched, already unlocked. */
802 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
803 break;
804
805 case PRC_IFUP:
806 IN_IFADDR_RLOCK(&in_ifa_tracker);
807 TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
808 if (ia->ia_ifa.ifa_addr == sa)
809 break;
810 }
811 if (ia == NULL || (ia->ia_flags & IFA_ROUTE)) {
812 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
813 return;
814 }
815 ifa_ref(&ia->ia_ifa);
816 IN_IFADDR_RUNLOCK(&in_ifa_tracker);
817 flags = RTF_UP;
818 ifp = ia->ia_ifa.ifa_ifp;
819
820 if ((ifp->if_flags & IFF_LOOPBACK)
821 || (ifp->if_flags & IFF_POINTOPOINT))
822 flags |= RTF_HOST;
823
824 err = ifa_del_loopback_route((struct ifaddr *)ia, sa);
825
826 err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
827 if (err == 0)
828 ia->ia_flags |= IFA_ROUTE;
829
830 err = ifa_add_loopback_route((struct ifaddr *)ia, sa);
831
832 ifa_free(&ia->ia_ifa);
833 break;
834 }
835 }
836
837 static int
rip_attach(struct socket * so,int proto,struct thread * td)838 rip_attach(struct socket *so, int proto, struct thread *td)
839 {
840 struct inpcb *inp;
841 int error;
842
843 inp = sotoinpcb(so);
844 KASSERT(inp == NULL, ("rip_attach: inp != NULL"));
845
846 error = priv_check(td, PRIV_NETINET_RAW);
847 if (error)
848 return (error);
849 if (proto >= IPPROTO_MAX || proto < 0)
850 return EPROTONOSUPPORT;
851 error = soreserve(so, rip_sendspace, rip_recvspace);
852 if (error)
853 return (error);
854 INP_INFO_WLOCK(&V_ripcbinfo);
855 error = in_pcballoc(so, &V_ripcbinfo);
856 if (error) {
857 INP_INFO_WUNLOCK(&V_ripcbinfo);
858 return (error);
859 }
860 inp = (struct inpcb *)so->so_pcb;
861 inp->inp_vflag |= INP_IPV4;
862 inp->inp_ip_p = proto;
863 inp->inp_ip_ttl = V_ip_defttl;
864 rip_inshash(inp);
865 INP_INFO_WUNLOCK(&V_ripcbinfo);
866 INP_WUNLOCK(inp);
867 return (0);
868 }
869
870 static void
rip_detach(struct socket * so)871 rip_detach(struct socket *so)
872 {
873 struct inpcb *inp;
874
875 inp = sotoinpcb(so);
876 KASSERT(inp != NULL, ("rip_detach: inp == NULL"));
877 KASSERT(inp->inp_faddr.s_addr == INADDR_ANY,
878 ("rip_detach: not closed"));
879
880 INP_INFO_WLOCK(&V_ripcbinfo);
881 INP_WLOCK(inp);
882 rip_delhash(inp);
883 if (so == V_ip_mrouter && ip_mrouter_done)
884 ip_mrouter_done();
885 if (ip_rsvp_force_done)
886 ip_rsvp_force_done(so);
887 if (so == V_ip_rsvpd)
888 ip_rsvp_done();
889 in_pcbdetach(inp);
890 in_pcbfree(inp);
891 INP_INFO_WUNLOCK(&V_ripcbinfo);
892 }
893
894 static void
rip_dodisconnect(struct socket * so,struct inpcb * inp)895 rip_dodisconnect(struct socket *so, struct inpcb *inp)
896 {
897 struct inpcbinfo *pcbinfo;
898
899 pcbinfo = inp->inp_pcbinfo;
900 INP_INFO_WLOCK(pcbinfo);
901 INP_WLOCK(inp);
902 rip_delhash(inp);
903 inp->inp_faddr.s_addr = INADDR_ANY;
904 rip_inshash(inp);
905 SOCK_LOCK(so);
906 so->so_state &= ~SS_ISCONNECTED;
907 SOCK_UNLOCK(so);
908 INP_WUNLOCK(inp);
909 INP_INFO_WUNLOCK(pcbinfo);
910 }
911
912 static void
rip_abort(struct socket * so)913 rip_abort(struct socket *so)
914 {
915 struct inpcb *inp;
916
917 inp = sotoinpcb(so);
918 KASSERT(inp != NULL, ("rip_abort: inp == NULL"));
919
920 rip_dodisconnect(so, inp);
921 }
922
923 static void
rip_close(struct socket * so)924 rip_close(struct socket *so)
925 {
926 struct inpcb *inp;
927
928 inp = sotoinpcb(so);
929 KASSERT(inp != NULL, ("rip_close: inp == NULL"));
930
931 rip_dodisconnect(so, inp);
932 }
933
934 static int
rip_disconnect(struct socket * so)935 rip_disconnect(struct socket *so)
936 {
937 struct inpcb *inp;
938
939 if ((so->so_state & SS_ISCONNECTED) == 0)
940 return (ENOTCONN);
941
942 inp = sotoinpcb(so);
943 KASSERT(inp != NULL, ("rip_disconnect: inp == NULL"));
944
945 rip_dodisconnect(so, inp);
946 return (0);
947 }
948
949 static int
rip_bind(struct socket * so,struct sockaddr * nam,struct thread * td)950 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
951 {
952 struct sockaddr_in *addr = (struct sockaddr_in *)nam;
953 struct inpcb *inp;
954 int error;
955
956 if (nam->sa_len != sizeof(*addr))
957 return (EINVAL);
958
959 error = prison_check_ip4(td->td_ucred, &addr->sin_addr);
960 if (error != 0)
961 return (error);
962
963 inp = sotoinpcb(so);
964 KASSERT(inp != NULL, ("rip_bind: inp == NULL"));
965
966 if (TAILQ_EMPTY(&V_ifnet) ||
967 (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
968 (addr->sin_addr.s_addr &&
969 (inp->inp_flags & INP_BINDANY) == 0 &&
970 ifa_ifwithaddr_check((struct sockaddr *)addr) == 0))
971 return (EADDRNOTAVAIL);
972
973 INP_INFO_WLOCK(&V_ripcbinfo);
974 INP_WLOCK(inp);
975 rip_delhash(inp);
976 inp->inp_laddr = addr->sin_addr;
977 rip_inshash(inp);
978 INP_WUNLOCK(inp);
979 INP_INFO_WUNLOCK(&V_ripcbinfo);
980 return (0);
981 }
982
983 static int
rip_connect(struct socket * so,struct sockaddr * nam,struct thread * td)984 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
985 {
986 struct sockaddr_in *addr = (struct sockaddr_in *)nam;
987 struct inpcb *inp;
988
989 if (nam->sa_len != sizeof(*addr))
990 return (EINVAL);
991 if (TAILQ_EMPTY(&V_ifnet))
992 return (EADDRNOTAVAIL);
993 if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
994 return (EAFNOSUPPORT);
995
996 inp = sotoinpcb(so);
997 KASSERT(inp != NULL, ("rip_connect: inp == NULL"));
998
999 INP_INFO_WLOCK(&V_ripcbinfo);
1000 INP_WLOCK(inp);
1001 rip_delhash(inp);
1002 inp->inp_faddr = addr->sin_addr;
1003 rip_inshash(inp);
1004 soisconnected(so);
1005 INP_WUNLOCK(inp);
1006 INP_INFO_WUNLOCK(&V_ripcbinfo);
1007 return (0);
1008 }
1009
1010 static int
rip_shutdown(struct socket * so)1011 rip_shutdown(struct socket *so)
1012 {
1013 struct inpcb *inp;
1014
1015 inp = sotoinpcb(so);
1016 KASSERT(inp != NULL, ("rip_shutdown: inp == NULL"));
1017
1018 INP_WLOCK(inp);
1019 socantsendmore(so);
1020 INP_WUNLOCK(inp);
1021 return (0);
1022 }
1023
1024 static int
rip_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct thread * td)1025 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
1026 struct mbuf *control, struct thread *td)
1027 {
1028 struct inpcb *inp;
1029 u_long dst;
1030
1031 inp = sotoinpcb(so);
1032 KASSERT(inp != NULL, ("rip_send: inp == NULL"));
1033
1034 /*
1035 * Note: 'dst' reads below are unlocked.
1036 */
1037 if (so->so_state & SS_ISCONNECTED) {
1038 if (nam) {
1039 m_freem(m);
1040 return (EISCONN);
1041 }
1042 dst = inp->inp_faddr.s_addr; /* Unlocked read. */
1043 } else {
1044 if (nam == NULL) {
1045 m_freem(m);
1046 return (ENOTCONN);
1047 }
1048 dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr;
1049 }
1050 return (rip_output(m, so, dst));
1051 }
1052 #endif /* INET */
1053
1054 static int
rip_pcblist(SYSCTL_HANDLER_ARGS)1055 rip_pcblist(SYSCTL_HANDLER_ARGS)
1056 {
1057 int error, i, n;
1058 struct inpcb *inp, **inp_list;
1059 inp_gen_t gencnt;
1060 struct xinpgen xig;
1061
1062 /*
1063 * The process of preparing the TCB list is too time-consuming and
1064 * resource-intensive to repeat twice on every request.
1065 */
1066 if (req->oldptr == 0) {
1067 n = V_ripcbinfo.ipi_count;
1068 n += imax(n / 8, 10);
1069 req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb);
1070 return (0);
1071 }
1072
1073 if (req->newptr != 0)
1074 return (EPERM);
1075
1076 /*
1077 * OK, now we're committed to doing something.
1078 */
1079 INP_INFO_RLOCK(&V_ripcbinfo);
1080 gencnt = V_ripcbinfo.ipi_gencnt;
1081 n = V_ripcbinfo.ipi_count;
1082 INP_INFO_RUNLOCK(&V_ripcbinfo);
1083
1084 bzero(&xig, sizeof(xig));
1085 xig.xig_len = sizeof xig;
1086 xig.xig_count = n;
1087 xig.xig_gen = gencnt;
1088 xig.xig_sogen = so_gencnt;
1089 error = SYSCTL_OUT(req, &xig, sizeof xig);
1090 if (error)
1091 return (error);
1092
1093 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
1094
1095 INP_INFO_RLOCK(&V_ripcbinfo);
1096 for (inp = LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n;
1097 inp = LIST_NEXT(inp, inp_list)) {
1098 INP_WLOCK(inp);
1099 if (inp->inp_gencnt <= gencnt &&
1100 cr_canseeinpcb(req->td->td_ucred, inp) == 0) {
1101 in_pcbref(inp);
1102 inp_list[i++] = inp;
1103 }
1104 INP_WUNLOCK(inp);
1105 }
1106 INP_INFO_RUNLOCK(&V_ripcbinfo);
1107 n = i;
1108
1109 error = 0;
1110 for (i = 0; i < n; i++) {
1111 inp = inp_list[i];
1112 INP_RLOCK(inp);
1113 if (inp->inp_gencnt <= gencnt) {
1114 struct xinpcb xi;
1115
1116 bzero(&xi, sizeof(xi));
1117 xi.xi_len = sizeof xi;
1118 /* XXX should avoid extra copy */
1119 bcopy(inp, &xi.xi_inp, sizeof *inp);
1120 if (inp->inp_socket)
1121 sotoxsocket(inp->inp_socket, &xi.xi_socket);
1122 INP_RUNLOCK(inp);
1123 error = SYSCTL_OUT(req, &xi, sizeof xi);
1124 } else
1125 INP_RUNLOCK(inp);
1126 }
1127 INP_INFO_WLOCK(&V_ripcbinfo);
1128 for (i = 0; i < n; i++) {
1129 inp = inp_list[i];
1130 INP_RLOCK(inp);
1131 if (!in_pcbrele_rlocked(inp))
1132 INP_RUNLOCK(inp);
1133 }
1134 INP_INFO_WUNLOCK(&V_ripcbinfo);
1135
1136 if (!error) {
1137 /*
1138 * Give the user an updated idea of our state. If the
1139 * generation differs from what we told her before, she knows
1140 * that something happened while we were processing this
1141 * request, and it might be necessary to retry.
1142 */
1143 INP_INFO_RLOCK(&V_ripcbinfo);
1144 xig.xig_gen = V_ripcbinfo.ipi_gencnt;
1145 xig.xig_sogen = so_gencnt;
1146 xig.xig_count = V_ripcbinfo.ipi_count;
1147 INP_INFO_RUNLOCK(&V_ripcbinfo);
1148 error = SYSCTL_OUT(req, &xig, sizeof xig);
1149 }
1150 free(inp_list, M_TEMP);
1151 return (error);
1152 }
1153
1154 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist,
1155 CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0,
1156 rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
1157
1158 #ifdef INET
1159 struct pr_usrreqs rip_usrreqs = {
1160 .pru_abort = rip_abort,
1161 .pru_attach = rip_attach,
1162 .pru_bind = rip_bind,
1163 .pru_connect = rip_connect,
1164 .pru_control = in_control,
1165 .pru_detach = rip_detach,
1166 .pru_disconnect = rip_disconnect,
1167 .pru_peeraddr = in_getpeeraddr,
1168 .pru_send = rip_send,
1169 .pru_shutdown = rip_shutdown,
1170 .pru_sockaddr = in_getsockaddr,
1171 .pru_sosetlabel = in_pcbsosetlabel,
1172 .pru_close = rip_close,
1173 };
1174 #endif /* INET */
1175