1 /*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1982, 1986, 1988, 1993
5 * The Regents of the University of California.
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 *
32 * @(#)raw_ip.c 8.7 (Berkeley) 5/15/95
33 */
34
35 #include <sys/cdefs.h>
36 #include "opt_inet.h"
37 #include "opt_inet6.h"
38 #include "opt_ipsec.h"
39 #include "opt_route.h"
40
41 #include <sys/param.h>
42 #include <sys/jail.h>
43 #include <sys/kernel.h>
44 #include <sys/eventhandler.h>
45 #include <sys/lock.h>
46 #include <sys/malloc.h>
47 #include <sys/mbuf.h>
48 #include <sys/priv.h>
49 #include <sys/proc.h>
50 #include <sys/protosw.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/route/route_ctl.h>
65 #include <net/vnet.h>
66
67 #include <netinet/in.h>
68 #include <netinet/in_systm.h>
69 #include <netinet/in_fib.h>
70 #include <netinet/in_pcb.h>
71 #include <netinet/in_var.h>
72 #include <netinet/if_ether.h>
73 #include <netinet/ip.h>
74 #include <netinet/ip_var.h>
75 #include <netinet/ip_mroute.h>
76 #include <netinet/ip_icmp.h>
77
78 #include <netipsec/ipsec_support.h>
79
80 #include <machine/stdarg.h>
81 #include <security/mac/mac_framework.h>
82
83 extern ipproto_input_t *ip_protox[];
84
85 VNET_DEFINE(int, ip_defttl) = IPDEFTTL;
86 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_VNET | CTLFLAG_RW,
87 &VNET_NAME(ip_defttl), 0,
88 "Maximum TTL on IP packets");
89
90 VNET_DEFINE(struct inpcbinfo, ripcbinfo);
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_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 **, struct ip_fw_args *);
102 void (*ip_divert_ptr)(struct mbuf *, bool);
103 int (*ng_ipfw_input_p)(struct mbuf **, struct ip_fw_args *, bool);
104
105 #ifdef INET
106 /*
107 * Hooks for multicast routing. They all default to NULL, so leave them not
108 * initialized and rely on BSS being set to 0.
109 */
110
111 /*
112 * The socket used to communicate with the multicast routing daemon.
113 */
114 VNET_DEFINE(struct socket *, ip_mrouter);
115
116 /*
117 * The various mrouter and rsvp functions.
118 */
119 int (*ip_mrouter_set)(struct socket *, struct sockopt *);
120 int (*ip_mrouter_get)(struct socket *, struct sockopt *);
121 int (*ip_mrouter_done)(void);
122 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
123 struct ip_moptions *);
124 int (*mrt_ioctl)(u_long, caddr_t, int);
125 int (*legal_vif_num)(int);
126 u_long (*ip_mcast_src)(int);
127
128 int (*rsvp_input_p)(struct mbuf **, int *, int);
129 int (*ip_rsvp_vif)(struct socket *, struct sockopt *);
130 void (*ip_rsvp_force_done)(struct socket *);
131 #endif /* INET */
132
133 u_long rip_sendspace = 9216;
134 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
135 &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
136
137 u_long rip_recvspace = 9216;
138 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW,
139 &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams");
140
141 /*
142 * Hash functions
143 */
144
145 #define INP_PCBHASH_RAW_SIZE 256
146 #define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \
147 (((proto) + (laddr) + (faddr)) % (mask) + 1)
148
149 #ifdef INET
150 static void
rip_inshash(struct inpcb * inp)151 rip_inshash(struct inpcb *inp)
152 {
153 struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
154 struct inpcbhead *pcbhash;
155 int hash;
156
157 INP_HASH_WLOCK_ASSERT(pcbinfo);
158 INP_WLOCK_ASSERT(inp);
159
160 if (inp->inp_ip_p != 0 &&
161 inp->inp_laddr.s_addr != INADDR_ANY &&
162 inp->inp_faddr.s_addr != INADDR_ANY) {
163 hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr,
164 inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask);
165 } else
166 hash = 0;
167 pcbhash = &pcbinfo->ipi_hash_exact[hash];
168 CK_LIST_INSERT_HEAD(pcbhash, inp, inp_hash_exact);
169 }
170
171 static void
rip_delhash(struct inpcb * inp)172 rip_delhash(struct inpcb *inp)
173 {
174
175 INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo);
176 INP_WLOCK_ASSERT(inp);
177
178 CK_LIST_REMOVE(inp, inp_hash_exact);
179 }
180 #endif /* INET */
181
182 INPCBSTORAGE_DEFINE(ripcbstor, inpcb, "rawinp", "ripcb", "rip", "riphash");
183
184 static void
rip_init(void * arg __unused)185 rip_init(void *arg __unused)
186 {
187
188 in_pcbinfo_init(&V_ripcbinfo, &ripcbstor, INP_PCBHASH_RAW_SIZE, 1);
189 }
190 VNET_SYSINIT(rip_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, rip_init, NULL);
191
192 #ifdef VIMAGE
193 static void
rip_destroy(void * unused __unused)194 rip_destroy(void *unused __unused)
195 {
196
197 in_pcbinfo_destroy(&V_ripcbinfo);
198 }
199 VNET_SYSUNINIT(raw_ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, rip_destroy, NULL);
200 #endif
201
202 #ifdef INET
203 static int
rip_append(struct inpcb * inp,struct ip * ip,struct mbuf * m,struct sockaddr_in * ripsrc)204 rip_append(struct inpcb *inp, struct ip *ip, struct mbuf *m,
205 struct sockaddr_in *ripsrc)
206 {
207 struct socket *so = inp->inp_socket;
208 struct mbuf *n, *opts = NULL;
209
210 INP_LOCK_ASSERT(inp);
211
212 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
213 /* check AH/ESP integrity. */
214 if (IPSEC_ENABLED(ipv4) && IPSEC_CHECK_POLICY(ipv4, m, inp) != 0)
215 return (0);
216 #endif /* IPSEC */
217 #ifdef MAC
218 if (mac_inpcb_check_deliver(inp, m) != 0)
219 return (0);
220 #endif
221 /* Check the minimum TTL for socket. */
222 if (inp->inp_ip_minttl && inp->inp_ip_minttl > ip->ip_ttl)
223 return (0);
224
225 if ((n = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL)
226 return (0);
227
228 if ((inp->inp_flags & INP_CONTROLOPTS) ||
229 (so->so_options & (SO_TIMESTAMP | SO_BINTIME)))
230 ip_savecontrol(inp, &opts, ip, n);
231 SOCKBUF_LOCK(&so->so_rcv);
232 if (sbappendaddr_locked(&so->so_rcv,
233 (struct sockaddr *)ripsrc, n, opts) == 0) {
234 soroverflow_locked(so);
235 m_freem(n);
236 if (opts)
237 m_freem(opts);
238 return (0);
239 }
240 sorwakeup_locked(so);
241
242 return (1);
243 }
244
245 struct rip_inp_match_ctx {
246 struct ip *ip;
247 int proto;
248 };
249
250 static bool
rip_inp_match1(const struct inpcb * inp,void * v)251 rip_inp_match1(const struct inpcb *inp, void *v)
252 {
253 struct rip_inp_match_ctx *ctx = v;
254
255 if (inp->inp_ip_p != ctx->proto)
256 return (false);
257 #ifdef INET6
258 /* XXX inp locking */
259 if ((inp->inp_vflag & INP_IPV4) == 0)
260 return (false);
261 #endif
262 if (inp->inp_laddr.s_addr != ctx->ip->ip_dst.s_addr)
263 return (false);
264 if (inp->inp_faddr.s_addr != ctx->ip->ip_src.s_addr)
265 return (false);
266 return (true);
267 }
268
269 static bool
rip_inp_match2(const struct inpcb * inp,void * v)270 rip_inp_match2(const struct inpcb *inp, void *v)
271 {
272 struct rip_inp_match_ctx *ctx = v;
273
274 if (inp->inp_ip_p && inp->inp_ip_p != ctx->proto)
275 return (false);
276 #ifdef INET6
277 /* XXX inp locking */
278 if ((inp->inp_vflag & INP_IPV4) == 0)
279 return (false);
280 #endif
281 if (!in_nullhost(inp->inp_laddr) &&
282 !in_hosteq(inp->inp_laddr, ctx->ip->ip_dst))
283 return (false);
284 if (!in_nullhost(inp->inp_faddr) &&
285 !in_hosteq(inp->inp_faddr, ctx->ip->ip_src))
286 return (false);
287 return (true);
288 }
289
290 /*
291 * Setup generic address and protocol structures for raw_input routine, then
292 * pass them along with mbuf chain.
293 */
294 int
rip_input(struct mbuf ** mp,int * offp,int proto)295 rip_input(struct mbuf **mp, int *offp, int proto)
296 {
297 struct rip_inp_match_ctx ctx = {
298 .ip = mtod(*mp, struct ip *),
299 .proto = proto,
300 };
301 struct inpcb_iterator inpi = INP_ITERATOR(&V_ripcbinfo,
302 INPLOOKUP_RLOCKPCB, rip_inp_match1, &ctx);
303 struct ifnet *ifp;
304 struct mbuf *m = *mp;
305 struct inpcb *inp;
306 struct sockaddr_in ripsrc;
307 int appended;
308
309 *mp = NULL;
310 appended = 0;
311
312 bzero(&ripsrc, sizeof(ripsrc));
313 ripsrc.sin_len = sizeof(ripsrc);
314 ripsrc.sin_family = AF_INET;
315 ripsrc.sin_addr = ctx.ip->ip_src;
316
317 ifp = m->m_pkthdr.rcvif;
318
319 inpi.hash = INP_PCBHASH_RAW(proto, ctx.ip->ip_src.s_addr,
320 ctx.ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask);
321 while ((inp = inp_next(&inpi)) != NULL) {
322 INP_RLOCK_ASSERT(inp);
323 if (jailed_without_vnet(inp->inp_cred) &&
324 prison_check_ip4(inp->inp_cred, &ctx.ip->ip_dst) != 0) {
325 /*
326 * XXX: If faddr was bound to multicast group,
327 * jailed raw socket will drop datagram.
328 */
329 continue;
330 }
331 appended += rip_append(inp, ctx.ip, m, &ripsrc);
332 }
333
334 inpi.hash = 0;
335 inpi.match = rip_inp_match2;
336 MPASS(inpi.inp == NULL);
337 while ((inp = inp_next(&inpi)) != NULL) {
338 INP_RLOCK_ASSERT(inp);
339 if (jailed_without_vnet(inp->inp_cred) &&
340 !IN_MULTICAST(ntohl(ctx.ip->ip_dst.s_addr)) &&
341 prison_check_ip4(inp->inp_cred, &ctx.ip->ip_dst) != 0)
342 /*
343 * Allow raw socket in jail to receive multicast;
344 * assume process had PRIV_NETINET_RAW at attach,
345 * and fall through into normal filter path if so.
346 */
347 continue;
348 /*
349 * If this raw socket has multicast state, and we
350 * have received a multicast, check if this socket
351 * should receive it, as multicast filtering is now
352 * the responsibility of the transport layer.
353 */
354 if (inp->inp_moptions != NULL &&
355 IN_MULTICAST(ntohl(ctx.ip->ip_dst.s_addr))) {
356 /*
357 * If the incoming datagram is for IGMP, allow it
358 * through unconditionally to the raw socket.
359 *
360 * In the case of IGMPv2, we may not have explicitly
361 * joined the group, and may have set IFF_ALLMULTI
362 * on the interface. imo_multi_filter() may discard
363 * control traffic we actually need to see.
364 *
365 * Userland multicast routing daemons should continue
366 * filter the control traffic appropriately.
367 */
368 int blocked;
369
370 blocked = MCAST_PASS;
371 if (proto != IPPROTO_IGMP) {
372 struct sockaddr_in group;
373
374 bzero(&group, sizeof(struct sockaddr_in));
375 group.sin_len = sizeof(struct sockaddr_in);
376 group.sin_family = AF_INET;
377 group.sin_addr = ctx.ip->ip_dst;
378
379 blocked = imo_multi_filter(inp->inp_moptions,
380 ifp,
381 (struct sockaddr *)&group,
382 (struct sockaddr *)&ripsrc);
383 }
384
385 if (blocked != MCAST_PASS) {
386 IPSTAT_INC(ips_notmember);
387 continue;
388 }
389 }
390 appended += rip_append(inp, ctx.ip, m, &ripsrc);
391 }
392 if (appended == 0 && ip_protox[ctx.ip->ip_p] == rip_input) {
393 IPSTAT_INC(ips_noproto);
394 IPSTAT_DEC(ips_delivered);
395 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PROTOCOL, 0, 0);
396 } else
397 m_freem(m);
398 return (IPPROTO_DONE);
399 }
400
401 /*
402 * Generate IP header and pass packet to ip_output. Tack on options user may
403 * have setup with control call.
404 */
405 static int
rip_send(struct socket * so,int pruflags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct thread * td)406 rip_send(struct socket *so, int pruflags, struct mbuf *m, struct sockaddr *nam,
407 struct mbuf *control, struct thread *td)
408 {
409 struct epoch_tracker et;
410 struct ip *ip;
411 struct inpcb *inp;
412 in_addr_t *dst;
413 int error, flags, cnt, hlen;
414 u_char opttype, optlen, *cp;
415
416 inp = sotoinpcb(so);
417 KASSERT(inp != NULL, ("rip_send: inp == NULL"));
418
419 if (control != NULL) {
420 m_freem(control);
421 control = NULL;
422 }
423
424 if (so->so_state & SS_ISCONNECTED) {
425 if (nam) {
426 error = EISCONN;
427 m_freem(m);
428 return (error);
429 }
430 dst = &inp->inp_faddr.s_addr;
431 } else {
432 if (nam == NULL)
433 error = ENOTCONN;
434 else if (nam->sa_family != AF_INET)
435 error = EAFNOSUPPORT;
436 else if (nam->sa_len != sizeof(struct sockaddr_in))
437 error = EINVAL;
438 else
439 error = 0;
440 if (error != 0) {
441 m_freem(m);
442 return (error);
443 }
444 dst = &((struct sockaddr_in *)nam)->sin_addr.s_addr;
445 }
446
447 flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) |
448 IP_ALLOWBROADCAST;
449
450 /*
451 * If the user handed us a complete IP packet, use it. Otherwise,
452 * allocate an mbuf for a header and fill it in.
453 */
454 if ((inp->inp_flags & INP_HDRINCL) == 0) {
455 if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
456 m_freem(m);
457 return(EMSGSIZE);
458 }
459 M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
460 if (m == NULL)
461 return(ENOBUFS);
462
463 INP_RLOCK(inp);
464 ip = mtod(m, struct ip *);
465 ip->ip_tos = inp->inp_ip_tos;
466 if (inp->inp_flags & INP_DONTFRAG)
467 ip->ip_off = htons(IP_DF);
468 else
469 ip->ip_off = htons(0);
470 ip->ip_p = inp->inp_ip_p;
471 ip->ip_len = htons(m->m_pkthdr.len);
472 ip->ip_src = inp->inp_laddr;
473 ip->ip_dst.s_addr = *dst;
474 #ifdef ROUTE_MPATH
475 if (CALC_FLOWID_OUTBOUND) {
476 uint32_t hash_type, hash_val;
477
478 hash_val = fib4_calc_software_hash(ip->ip_src,
479 ip->ip_dst, 0, 0, ip->ip_p, &hash_type);
480 m->m_pkthdr.flowid = hash_val;
481 M_HASHTYPE_SET(m, hash_type);
482 flags |= IP_NODEFAULTFLOWID;
483 }
484 #endif
485 if (jailed(inp->inp_cred)) {
486 /*
487 * prison_local_ip4() would be good enough but would
488 * let a source of INADDR_ANY pass, which we do not
489 * want to see from jails.
490 */
491 if (ip->ip_src.s_addr == INADDR_ANY) {
492 NET_EPOCH_ENTER(et);
493 error = in_pcbladdr(inp, &ip->ip_dst,
494 &ip->ip_src, inp->inp_cred);
495 NET_EPOCH_EXIT(et);
496 } else {
497 error = prison_local_ip4(inp->inp_cred,
498 &ip->ip_src);
499 }
500 if (error != 0) {
501 INP_RUNLOCK(inp);
502 m_freem(m);
503 return (error);
504 }
505 }
506 ip->ip_ttl = inp->inp_ip_ttl;
507 } else {
508 if (m->m_pkthdr.len > IP_MAXPACKET) {
509 m_freem(m);
510 return (EMSGSIZE);
511 }
512 if (m->m_pkthdr.len < sizeof(*ip)) {
513 m_freem(m);
514 return (EINVAL);
515 }
516 m = m_pullup(m, sizeof(*ip));
517 if (m == NULL)
518 return (ENOMEM);
519 ip = mtod(m, struct ip *);
520 hlen = ip->ip_hl << 2;
521 if (m->m_len < hlen) {
522 m = m_pullup(m, hlen);
523 if (m == NULL)
524 return (EINVAL);
525 ip = mtod(m, struct ip *);
526 }
527 #ifdef ROUTE_MPATH
528 if (CALC_FLOWID_OUTBOUND) {
529 uint32_t hash_type, hash_val;
530
531 hash_val = fib4_calc_software_hash(ip->ip_dst,
532 ip->ip_src, 0, 0, ip->ip_p, &hash_type);
533 m->m_pkthdr.flowid = hash_val;
534 M_HASHTYPE_SET(m, hash_type);
535 flags |= IP_NODEFAULTFLOWID;
536 }
537 #endif
538 INP_RLOCK(inp);
539 /*
540 * Don't allow both user specified and setsockopt options,
541 * and don't allow packet length sizes that will crash.
542 */
543 if ((hlen < sizeof (*ip))
544 || ((hlen > sizeof (*ip)) && inp->inp_options)
545 || (ntohs(ip->ip_len) != m->m_pkthdr.len)) {
546 INP_RUNLOCK(inp);
547 m_freem(m);
548 return (EINVAL);
549 }
550 error = prison_check_ip4(inp->inp_cred, &ip->ip_src);
551 if (error != 0) {
552 INP_RUNLOCK(inp);
553 m_freem(m);
554 return (error);
555 }
556 /*
557 * Don't allow IP options which do not have the required
558 * structure as specified in section 3.1 of RFC 791 on
559 * pages 15-23.
560 */
561 cp = (u_char *)(ip + 1);
562 cnt = hlen - sizeof (struct ip);
563 for (; cnt > 0; cnt -= optlen, cp += optlen) {
564 opttype = cp[IPOPT_OPTVAL];
565 if (opttype == IPOPT_EOL)
566 break;
567 if (opttype == IPOPT_NOP) {
568 optlen = 1;
569 continue;
570 }
571 if (cnt < IPOPT_OLEN + sizeof(u_char)) {
572 INP_RUNLOCK(inp);
573 m_freem(m);
574 return (EINVAL);
575 }
576 optlen = cp[IPOPT_OLEN];
577 if (optlen < IPOPT_OLEN + sizeof(u_char) ||
578 optlen > cnt) {
579 INP_RUNLOCK(inp);
580 m_freem(m);
581 return (EINVAL);
582 }
583 }
584 /*
585 * This doesn't allow application to specify ID of zero,
586 * but we got this limitation from the beginning of history.
587 */
588 if (ip->ip_id == 0)
589 ip_fillid(ip);
590
591 /*
592 * XXX prevent ip_output from overwriting header fields.
593 */
594 flags |= IP_RAWOUTPUT;
595 IPSTAT_INC(ips_rawout);
596 }
597
598 if (inp->inp_flags & INP_ONESBCAST)
599 flags |= IP_SENDONES;
600
601 #ifdef MAC
602 mac_inpcb_create_mbuf(inp, m);
603 #endif
604
605 NET_EPOCH_ENTER(et);
606 error = ip_output(m, inp->inp_options, NULL, flags,
607 inp->inp_moptions, inp);
608 NET_EPOCH_EXIT(et);
609 INP_RUNLOCK(inp);
610 return (error);
611 }
612
613 /*
614 * Raw IP socket option processing.
615 *
616 * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could
617 * only be created by a privileged process, and as such, socket option
618 * operations to manage system properties on any raw socket were allowed to
619 * take place without explicit additional access control checks. However,
620 * raw sockets can now also be created in jail(), and therefore explicit
621 * checks are now required. Likewise, raw sockets can be used by a process
622 * after it gives up privilege, so some caution is required. For options
623 * passed down to the IP layer via ip_ctloutput(), checks are assumed to be
624 * performed in ip_ctloutput() and therefore no check occurs here.
625 * Unilaterally checking priv_check() here breaks normal IP socket option
626 * operations on raw sockets.
627 *
628 * When adding new socket options here, make sure to add access control
629 * checks here as necessary.
630 *
631 * XXX-BZ inp locking?
632 */
633 int
rip_ctloutput(struct socket * so,struct sockopt * sopt)634 rip_ctloutput(struct socket *so, struct sockopt *sopt)
635 {
636 struct inpcb *inp = sotoinpcb(so);
637 int error, optval;
638
639 if (sopt->sopt_level != IPPROTO_IP) {
640 if ((sopt->sopt_level == SOL_SOCKET) &&
641 (sopt->sopt_name == SO_SETFIB)) {
642 inp->inp_inc.inc_fibnum = so->so_fibnum;
643 return (0);
644 }
645 return (EINVAL);
646 }
647
648 error = 0;
649 switch (sopt->sopt_dir) {
650 case SOPT_GET:
651 switch (sopt->sopt_name) {
652 case IP_HDRINCL:
653 optval = inp->inp_flags & INP_HDRINCL;
654 error = sooptcopyout(sopt, &optval, sizeof optval);
655 break;
656
657 case IP_FW3: /* generic ipfw v.3 functions */
658 case IP_FW_ADD: /* ADD actually returns the body... */
659 case IP_FW_GET:
660 case IP_FW_TABLE_GETSIZE:
661 case IP_FW_TABLE_LIST:
662 case IP_FW_NAT_GET_CONFIG:
663 case IP_FW_NAT_GET_LOG:
664 if (V_ip_fw_ctl_ptr != NULL)
665 error = V_ip_fw_ctl_ptr(sopt);
666 else
667 error = ENOPROTOOPT;
668 break;
669
670 case IP_DUMMYNET3: /* generic dummynet v.3 functions */
671 case IP_DUMMYNET_GET:
672 if (ip_dn_ctl_ptr != NULL)
673 error = ip_dn_ctl_ptr(sopt);
674 else
675 error = ENOPROTOOPT;
676 break ;
677
678 case MRT_INIT:
679 case MRT_DONE:
680 case MRT_ADD_VIF:
681 case MRT_DEL_VIF:
682 case MRT_ADD_MFC:
683 case MRT_DEL_MFC:
684 case MRT_VERSION:
685 case MRT_ASSERT:
686 case MRT_API_SUPPORT:
687 case MRT_API_CONFIG:
688 case MRT_ADD_BW_UPCALL:
689 case MRT_DEL_BW_UPCALL:
690 error = priv_check(curthread, PRIV_NETINET_MROUTE);
691 if (error != 0)
692 return (error);
693 if (inp->inp_ip_p != IPPROTO_IGMP)
694 return (EOPNOTSUPP);
695 error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
696 EOPNOTSUPP;
697 break;
698
699 default:
700 error = ip_ctloutput(so, sopt);
701 break;
702 }
703 break;
704
705 case SOPT_SET:
706 switch (sopt->sopt_name) {
707 case IP_HDRINCL:
708 error = sooptcopyin(sopt, &optval, sizeof optval,
709 sizeof optval);
710 if (error)
711 break;
712 if (optval)
713 inp->inp_flags |= INP_HDRINCL;
714 else
715 inp->inp_flags &= ~INP_HDRINCL;
716 break;
717
718 case IP_FW3: /* generic ipfw v.3 functions */
719 case IP_FW_ADD:
720 case IP_FW_DEL:
721 case IP_FW_FLUSH:
722 case IP_FW_ZERO:
723 case IP_FW_RESETLOG:
724 case IP_FW_TABLE_ADD:
725 case IP_FW_TABLE_DEL:
726 case IP_FW_TABLE_FLUSH:
727 case IP_FW_NAT_CFG:
728 case IP_FW_NAT_DEL:
729 if (V_ip_fw_ctl_ptr != NULL)
730 error = V_ip_fw_ctl_ptr(sopt);
731 else
732 error = ENOPROTOOPT;
733 break;
734
735 case IP_DUMMYNET3: /* generic dummynet v.3 functions */
736 case IP_DUMMYNET_CONFIGURE:
737 case IP_DUMMYNET_DEL:
738 case IP_DUMMYNET_FLUSH:
739 if (ip_dn_ctl_ptr != NULL)
740 error = ip_dn_ctl_ptr(sopt);
741 else
742 error = ENOPROTOOPT ;
743 break ;
744
745 case IP_RSVP_ON:
746 error = priv_check(curthread, PRIV_NETINET_MROUTE);
747 if (error != 0)
748 return (error);
749 if (inp->inp_ip_p != IPPROTO_RSVP)
750 return (EOPNOTSUPP);
751 error = ip_rsvp_init(so);
752 break;
753
754 case IP_RSVP_OFF:
755 error = priv_check(curthread, PRIV_NETINET_MROUTE);
756 if (error != 0)
757 return (error);
758 error = ip_rsvp_done();
759 break;
760
761 case IP_RSVP_VIF_ON:
762 case IP_RSVP_VIF_OFF:
763 error = priv_check(curthread, PRIV_NETINET_MROUTE);
764 if (error != 0)
765 return (error);
766 if (inp->inp_ip_p != IPPROTO_RSVP)
767 return (EOPNOTSUPP);
768 error = ip_rsvp_vif ?
769 ip_rsvp_vif(so, sopt) : EINVAL;
770 break;
771
772 case MRT_INIT:
773 case MRT_DONE:
774 case MRT_ADD_VIF:
775 case MRT_DEL_VIF:
776 case MRT_ADD_MFC:
777 case MRT_DEL_MFC:
778 case MRT_VERSION:
779 case MRT_ASSERT:
780 case MRT_API_SUPPORT:
781 case MRT_API_CONFIG:
782 case MRT_ADD_BW_UPCALL:
783 case MRT_DEL_BW_UPCALL:
784 error = priv_check(curthread, PRIV_NETINET_MROUTE);
785 if (error != 0)
786 return (error);
787 if (inp->inp_ip_p != IPPROTO_IGMP)
788 return (EOPNOTSUPP);
789 error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
790 EOPNOTSUPP;
791 break;
792
793 default:
794 error = ip_ctloutput(so, sopt);
795 break;
796 }
797 break;
798 }
799
800 return (error);
801 }
802
803 void
rip_ctlinput(struct icmp * icmp)804 rip_ctlinput(struct icmp *icmp)
805 {
806 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
807 if (IPSEC_ENABLED(ipv4))
808 IPSEC_CTLINPUT(ipv4, icmp);
809 #endif
810 }
811
812 static int
rip_attach(struct socket * so,int proto,struct thread * td)813 rip_attach(struct socket *so, int proto, struct thread *td)
814 {
815 struct inpcb *inp;
816 int error;
817
818 inp = sotoinpcb(so);
819 KASSERT(inp == NULL, ("rip_attach: inp != NULL"));
820
821 error = priv_check(td, PRIV_NETINET_RAW);
822 if (error)
823 return (error);
824 if (proto >= IPPROTO_MAX || proto < 0)
825 return EPROTONOSUPPORT;
826 error = soreserve(so, rip_sendspace, rip_recvspace);
827 if (error)
828 return (error);
829 error = in_pcballoc(so, &V_ripcbinfo);
830 if (error)
831 return (error);
832 inp = (struct inpcb *)so->so_pcb;
833 inp->inp_ip_p = proto;
834 inp->inp_ip_ttl = V_ip_defttl;
835 INP_HASH_WLOCK(&V_ripcbinfo);
836 rip_inshash(inp);
837 INP_HASH_WUNLOCK(&V_ripcbinfo);
838 INP_WUNLOCK(inp);
839 return (0);
840 }
841
842 static void
rip_detach(struct socket * so)843 rip_detach(struct socket *so)
844 {
845 struct inpcb *inp;
846
847 inp = sotoinpcb(so);
848 KASSERT(inp != NULL, ("rip_detach: inp == NULL"));
849 KASSERT(inp->inp_faddr.s_addr == INADDR_ANY,
850 ("rip_detach: not closed"));
851
852 /* Disable mrouter first */
853 if (so == V_ip_mrouter && ip_mrouter_done)
854 ip_mrouter_done();
855
856 INP_WLOCK(inp);
857 INP_HASH_WLOCK(&V_ripcbinfo);
858 rip_delhash(inp);
859 INP_HASH_WUNLOCK(&V_ripcbinfo);
860
861 if (ip_rsvp_force_done)
862 ip_rsvp_force_done(so);
863 if (so == V_ip_rsvpd)
864 ip_rsvp_done();
865 in_pcbfree(inp);
866 }
867
868 static void
rip_dodisconnect(struct socket * so,struct inpcb * inp)869 rip_dodisconnect(struct socket *so, struct inpcb *inp)
870 {
871 struct inpcbinfo *pcbinfo;
872
873 pcbinfo = inp->inp_pcbinfo;
874 INP_WLOCK(inp);
875 INP_HASH_WLOCK(pcbinfo);
876 rip_delhash(inp);
877 inp->inp_faddr.s_addr = INADDR_ANY;
878 rip_inshash(inp);
879 INP_HASH_WUNLOCK(pcbinfo);
880 SOCK_LOCK(so);
881 so->so_state &= ~SS_ISCONNECTED;
882 SOCK_UNLOCK(so);
883 INP_WUNLOCK(inp);
884 }
885
886 static void
rip_abort(struct socket * so)887 rip_abort(struct socket *so)
888 {
889 struct inpcb *inp;
890
891 inp = sotoinpcb(so);
892 KASSERT(inp != NULL, ("rip_abort: inp == NULL"));
893
894 rip_dodisconnect(so, inp);
895 }
896
897 static void
rip_close(struct socket * so)898 rip_close(struct socket *so)
899 {
900 struct inpcb *inp;
901
902 inp = sotoinpcb(so);
903 KASSERT(inp != NULL, ("rip_close: inp == NULL"));
904
905 rip_dodisconnect(so, inp);
906 }
907
908 static int
rip_disconnect(struct socket * so)909 rip_disconnect(struct socket *so)
910 {
911 struct inpcb *inp;
912
913 if ((so->so_state & SS_ISCONNECTED) == 0)
914 return (ENOTCONN);
915
916 inp = sotoinpcb(so);
917 KASSERT(inp != NULL, ("rip_disconnect: inp == NULL"));
918
919 rip_dodisconnect(so, inp);
920 return (0);
921 }
922
923 static int
rip_bind(struct socket * so,struct sockaddr * nam,struct thread * td)924 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
925 {
926 struct sockaddr_in *addr = (struct sockaddr_in *)nam;
927 struct inpcb *inp;
928 int error;
929
930 if (nam->sa_family != AF_INET)
931 return (EAFNOSUPPORT);
932 if (nam->sa_len != sizeof(*addr))
933 return (EINVAL);
934
935 error = prison_check_ip4(td->td_ucred, &addr->sin_addr);
936 if (error != 0)
937 return (error);
938
939 inp = sotoinpcb(so);
940 KASSERT(inp != NULL, ("rip_bind: inp == NULL"));
941
942 if (CK_STAILQ_EMPTY(&V_ifnet) ||
943 (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
944 (addr->sin_addr.s_addr &&
945 (inp->inp_flags & INP_BINDANY) == 0 &&
946 ifa_ifwithaddr_check((struct sockaddr *)addr) == 0))
947 return (EADDRNOTAVAIL);
948
949 INP_WLOCK(inp);
950 INP_HASH_WLOCK(&V_ripcbinfo);
951 rip_delhash(inp);
952 inp->inp_laddr = addr->sin_addr;
953 rip_inshash(inp);
954 INP_HASH_WUNLOCK(&V_ripcbinfo);
955 INP_WUNLOCK(inp);
956 return (0);
957 }
958
959 static int
rip_connect(struct socket * so,struct sockaddr * nam,struct thread * td)960 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
961 {
962 struct sockaddr_in *addr = (struct sockaddr_in *)nam;
963 struct inpcb *inp;
964
965 if (nam->sa_len != sizeof(*addr))
966 return (EINVAL);
967 if (CK_STAILQ_EMPTY(&V_ifnet))
968 return (EADDRNOTAVAIL);
969 if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
970 return (EAFNOSUPPORT);
971
972 inp = sotoinpcb(so);
973 KASSERT(inp != NULL, ("rip_connect: inp == NULL"));
974
975 INP_WLOCK(inp);
976 INP_HASH_WLOCK(&V_ripcbinfo);
977 rip_delhash(inp);
978 inp->inp_faddr = addr->sin_addr;
979 rip_inshash(inp);
980 INP_HASH_WUNLOCK(&V_ripcbinfo);
981 soisconnected(so);
982 INP_WUNLOCK(inp);
983 return (0);
984 }
985
986 static int
rip_shutdown(struct socket * so)987 rip_shutdown(struct socket *so)
988 {
989 struct inpcb *inp;
990
991 inp = sotoinpcb(so);
992 KASSERT(inp != NULL, ("rip_shutdown: inp == NULL"));
993
994 INP_WLOCK(inp);
995 socantsendmore(so);
996 INP_WUNLOCK(inp);
997 return (0);
998 }
999 #endif /* INET */
1000
1001 static int
rip_pcblist(SYSCTL_HANDLER_ARGS)1002 rip_pcblist(SYSCTL_HANDLER_ARGS)
1003 {
1004 struct inpcb_iterator inpi = INP_ALL_ITERATOR(&V_ripcbinfo,
1005 INPLOOKUP_RLOCKPCB);
1006 struct xinpgen xig;
1007 struct inpcb *inp;
1008 int error;
1009
1010 if (req->newptr != 0)
1011 return (EPERM);
1012
1013 if (req->oldptr == 0) {
1014 int n;
1015
1016 n = V_ripcbinfo.ipi_count;
1017 n += imax(n / 8, 10);
1018 req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb);
1019 return (0);
1020 }
1021
1022 if ((error = sysctl_wire_old_buffer(req, 0)) != 0)
1023 return (error);
1024
1025 bzero(&xig, sizeof(xig));
1026 xig.xig_len = sizeof xig;
1027 xig.xig_count = V_ripcbinfo.ipi_count;
1028 xig.xig_gen = V_ripcbinfo.ipi_gencnt;
1029 xig.xig_sogen = so_gencnt;
1030 error = SYSCTL_OUT(req, &xig, sizeof xig);
1031 if (error)
1032 return (error);
1033
1034 while ((inp = inp_next(&inpi)) != NULL) {
1035 if (inp->inp_gencnt <= xig.xig_gen &&
1036 cr_canseeinpcb(req->td->td_ucred, inp) == 0) {
1037 struct xinpcb xi;
1038
1039 in_pcbtoxinpcb(inp, &xi);
1040 error = SYSCTL_OUT(req, &xi, sizeof xi);
1041 if (error) {
1042 INP_RUNLOCK(inp);
1043 break;
1044 }
1045 }
1046 }
1047
1048 if (!error) {
1049 /*
1050 * Give the user an updated idea of our state. If the
1051 * generation differs from what we told her before, she knows
1052 * that something happened while we were processing this
1053 * request, and it might be necessary to retry.
1054 */
1055 xig.xig_gen = V_ripcbinfo.ipi_gencnt;
1056 xig.xig_sogen = so_gencnt;
1057 xig.xig_count = V_ripcbinfo.ipi_count;
1058 error = SYSCTL_OUT(req, &xig, sizeof xig);
1059 }
1060
1061 return (error);
1062 }
1063
1064 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist,
1065 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
1066 rip_pcblist, "S,xinpcb",
1067 "List of active raw IP sockets");
1068
1069 #ifdef INET
1070 struct protosw rip_protosw = {
1071 .pr_type = SOCK_RAW,
1072 .pr_flags = PR_ATOMIC|PR_ADDR,
1073 .pr_ctloutput = rip_ctloutput,
1074 .pr_abort = rip_abort,
1075 .pr_attach = rip_attach,
1076 .pr_bind = rip_bind,
1077 .pr_connect = rip_connect,
1078 .pr_control = in_control,
1079 .pr_detach = rip_detach,
1080 .pr_disconnect = rip_disconnect,
1081 .pr_peeraddr = in_getpeeraddr,
1082 .pr_send = rip_send,
1083 .pr_shutdown = rip_shutdown,
1084 .pr_sockaddr = in_getsockaddr,
1085 .pr_sosetlabel = in_pcbsosetlabel,
1086 .pr_close = rip_close
1087 };
1088 #endif /* INET */
1089