xref: /freebsd-13-stable/sys/netinet/raw_ip.c (revision 3bc80996974a61a4223eae4c1ccd47b6ee32a48a)
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/rmlock.h>
52 #include <sys/rwlock.h>
53 #include <sys/signalvar.h>
54 #include <sys/socket.h>
55 #include <sys/socketvar.h>
56 #include <sys/sx.h>
57 #include <sys/sysctl.h>
58 #include <sys/systm.h>
59 
60 #include <vm/uma.h>
61 
62 #include <net/if.h>
63 #include <net/if_var.h>
64 #include <net/route.h>
65 #include <net/route/route_ctl.h>
66 #include <net/vnet.h>
67 
68 #include <netinet/in.h>
69 #include <netinet/in_systm.h>
70 #include <netinet/in_fib.h>
71 #include <netinet/in_pcb.h>
72 #include <netinet/in_var.h>
73 #include <netinet/if_ether.h>
74 #include <netinet/ip.h>
75 #include <netinet/ip_var.h>
76 #include <netinet/ip_mroute.h>
77 #include <netinet/ip_icmp.h>
78 
79 #include <netipsec/ipsec_support.h>
80 
81 #include <machine/stdarg.h>
82 #include <security/mac/mac_framework.h>
83 
84 VNET_DEFINE(int, ip_defttl) = IPDEFTTL;
85 SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_VNET | CTLFLAG_RW,
86     &VNET_NAME(ip_defttl), 0,
87     "Maximum TTL on IP packets");
88 
89 VNET_DEFINE(struct inpcbhead, ripcb);
90 VNET_DEFINE(struct inpcbinfo, ripcbinfo);
91 
92 #define	V_ripcb			VNET(ripcb)
93 #define	V_ripcbinfo		VNET(ripcbinfo)
94 
95 /*
96  * Control and data hooks for ipfw, dummynet, divert and so on.
97  * The data hooks are not used here but it is convenient
98  * to keep them all in one place.
99  */
100 VNET_DEFINE(ip_fw_chk_ptr_t, ip_fw_chk_ptr) = NULL;
101 VNET_DEFINE(ip_fw_ctl_ptr_t, ip_fw_ctl_ptr) = NULL;
102 
103 int	(*ip_dn_ctl_ptr)(struct sockopt *);
104 int	(*ip_dn_io_ptr)(struct mbuf **, struct ip_fw_args *);
105 void	(*ip_divert_ptr)(struct mbuf *, bool);
106 int	(*ng_ipfw_input_p)(struct mbuf **, struct ip_fw_args *, bool);
107 
108 #ifdef INET
109 /*
110  * Hooks for multicast routing. They all default to NULL, so leave them not
111  * initialized and rely on BSS being set to 0.
112  */
113 
114 /*
115  * The socket used to communicate with the multicast routing daemon.
116  */
117 VNET_DEFINE(struct socket *, ip_mrouter);
118 
119 /*
120  * The various mrouter and rsvp functions.
121  */
122 int (*ip_mrouter_set)(struct socket *, struct sockopt *);
123 int (*ip_mrouter_get)(struct socket *, struct sockopt *);
124 int (*ip_mrouter_done)(void);
125 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
126 		   struct ip_moptions *);
127 int (*mrt_ioctl)(u_long, caddr_t, int);
128 int (*legal_vif_num)(int);
129 u_long (*ip_mcast_src)(int);
130 
131 int (*rsvp_input_p)(struct mbuf **, int *, int);
132 int (*ip_rsvp_vif)(struct socket *, struct sockopt *);
133 void (*ip_rsvp_force_done)(struct socket *);
134 #endif /* INET */
135 
136 extern	struct protosw inetsw[];
137 
138 u_long	rip_sendspace = 9216;
139 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
140     &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
141 
142 u_long	rip_recvspace = 9216;
143 SYSCTL_ULONG(_net_inet_raw, OID_AUTO, recvspace, CTLFLAG_RW,
144     &rip_recvspace, 0, "Maximum space for incoming raw IP datagrams");
145 
146 /*
147  * Hash functions
148  */
149 
150 #define INP_PCBHASH_RAW_SIZE	256
151 #define INP_PCBHASH_RAW(proto, laddr, faddr, mask) \
152         (((proto) + (laddr) + (faddr)) % (mask) + 1)
153 
154 #ifdef INET
155 static void
rip_inshash(struct inpcb * inp)156 rip_inshash(struct inpcb *inp)
157 {
158 	struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
159 	struct inpcbhead *pcbhash;
160 	int hash;
161 
162 	INP_INFO_WLOCK_ASSERT(pcbinfo);
163 	INP_WLOCK_ASSERT(inp);
164 
165 	if (inp->inp_ip_p != 0 &&
166 	    inp->inp_laddr.s_addr != INADDR_ANY &&
167 	    inp->inp_faddr.s_addr != INADDR_ANY) {
168 		hash = INP_PCBHASH_RAW(inp->inp_ip_p, inp->inp_laddr.s_addr,
169 		    inp->inp_faddr.s_addr, pcbinfo->ipi_hashmask);
170 	} else
171 		hash = 0;
172 	pcbhash = &pcbinfo->ipi_hashbase[hash];
173 	CK_LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
174 }
175 
176 static void
rip_delhash(struct inpcb * inp)177 rip_delhash(struct inpcb *inp)
178 {
179 
180 	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
181 	INP_WLOCK_ASSERT(inp);
182 
183 	CK_LIST_REMOVE(inp, inp_hash);
184 }
185 #endif /* INET */
186 
187 /*
188  * Raw interface to IP protocol.
189  */
190 
191 /*
192  * Initialize raw connection block q.
193  */
194 static void
rip_zone_change(void * tag)195 rip_zone_change(void *tag)
196 {
197 
198 	uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets);
199 }
200 
201 static int
rip_inpcb_init(void * mem,int size,int flags)202 rip_inpcb_init(void *mem, int size, int flags)
203 {
204 	struct inpcb *inp = mem;
205 
206 	INP_LOCK_INIT(inp, "inp", "rawinp");
207 	return (0);
208 }
209 
210 void
rip_init(void)211 rip_init(void)
212 {
213 
214 	in_pcbinfo_init(&V_ripcbinfo, "rip", &V_ripcb, INP_PCBHASH_RAW_SIZE,
215 	    1, "ripcb", rip_inpcb_init, IPI_HASHFIELDS_NONE);
216 	EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL,
217 	    EVENTHANDLER_PRI_ANY);
218 }
219 
220 #ifdef VIMAGE
221 static void
rip_destroy(void * unused __unused)222 rip_destroy(void *unused __unused)
223 {
224 
225 	in_pcbinfo_destroy(&V_ripcbinfo);
226 }
227 VNET_SYSUNINIT(raw_ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, rip_destroy, NULL);
228 #endif
229 
230 #ifdef INET
231 static int
rip_append(struct inpcb * last,struct ip * ip,struct mbuf * n,struct sockaddr_in * ripsrc)232 rip_append(struct inpcb *last, struct ip *ip, struct mbuf *n,
233     struct sockaddr_in *ripsrc)
234 {
235 	int policyfail = 0;
236 
237 	INP_LOCK_ASSERT(last);
238 
239 #if defined(IPSEC) || defined(IPSEC_SUPPORT)
240 	/* check AH/ESP integrity. */
241 	if (IPSEC_ENABLED(ipv4)) {
242 		if (IPSEC_CHECK_POLICY(ipv4, n, last) != 0)
243 			policyfail = 1;
244 	}
245 #endif /* IPSEC */
246 #ifdef MAC
247 	if (!policyfail && mac_inpcb_check_deliver(last, n) != 0)
248 		policyfail = 1;
249 #endif
250 	/* Check the minimum TTL for socket. */
251 	if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl)
252 		policyfail = 1;
253 	if (!policyfail) {
254 		struct mbuf *opts = NULL;
255 		struct socket *so;
256 
257 		so = last->inp_socket;
258 		if ((last->inp_flags & INP_CONTROLOPTS) ||
259 		    (so->so_options & (SO_TIMESTAMP | SO_BINTIME)))
260 			ip_savecontrol(last, &opts, ip, n);
261 		SOCKBUF_LOCK(&so->so_rcv);
262 		if (sbappendaddr_locked(&so->so_rcv,
263 		    (struct sockaddr *)ripsrc, n, opts) == 0) {
264 			soroverflow_locked(so);
265 			m_freem(n);
266 			if (opts)
267 				m_freem(opts);
268 		} else
269 			sorwakeup_locked(so);
270 	} else
271 		m_freem(n);
272 	return (policyfail);
273 }
274 
275 /*
276  * Setup generic address and protocol structures for raw_input routine, then
277  * pass them along with mbuf chain.
278  */
279 int
rip_input(struct mbuf ** mp,int * offp,int proto)280 rip_input(struct mbuf **mp, int *offp, int proto)
281 {
282 	struct ifnet *ifp;
283 	struct mbuf *m = *mp;
284 	struct ip *ip = mtod(m, struct ip *);
285 	struct inpcb *inp, *last;
286 	struct sockaddr_in ripsrc;
287 	int hash;
288 
289 	NET_EPOCH_ASSERT();
290 
291 	*mp = NULL;
292 
293 	bzero(&ripsrc, sizeof(ripsrc));
294 	ripsrc.sin_len = sizeof(ripsrc);
295 	ripsrc.sin_family = AF_INET;
296 	ripsrc.sin_addr = ip->ip_src;
297 	last = NULL;
298 
299 	ifp = m->m_pkthdr.rcvif;
300 
301 	hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr,
302 	    ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask);
303 	CK_LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) {
304 		if (inp->inp_ip_p != proto)
305 			continue;
306 #ifdef INET6
307 		/* XXX inp locking */
308 		if ((inp->inp_vflag & INP_IPV4) == 0)
309 			continue;
310 #endif
311 		if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
312 			continue;
313 		if (inp->inp_faddr.s_addr != ip->ip_src.s_addr)
314 			continue;
315 		if (last != NULL) {
316 			struct mbuf *n;
317 
318 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
319 			if (n != NULL)
320 			    (void) rip_append(last, ip, n, &ripsrc);
321 			/* XXX count dropped packet */
322 			INP_RUNLOCK(last);
323 			last = NULL;
324 		}
325 		INP_RLOCK(inp);
326 		if (__predict_false(inp->inp_flags2 & INP_FREED))
327 			goto skip_1;
328 		if (jailed_without_vnet(inp->inp_cred)) {
329 			/*
330 			 * XXX: If faddr was bound to multicast group,
331 			 * jailed raw socket will drop datagram.
332 			 */
333 			if (prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0)
334 				goto skip_1;
335 		}
336 		last = inp;
337 		continue;
338 	skip_1:
339 		INP_RUNLOCK(inp);
340 	}
341 	CK_LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) {
342 		if (inp->inp_ip_p && inp->inp_ip_p != proto)
343 			continue;
344 #ifdef INET6
345 		/* XXX inp locking */
346 		if ((inp->inp_vflag & INP_IPV4) == 0)
347 			continue;
348 #endif
349 		if (!in_nullhost(inp->inp_laddr) &&
350 		    !in_hosteq(inp->inp_laddr, ip->ip_dst))
351 			continue;
352 		if (!in_nullhost(inp->inp_faddr) &&
353 		    !in_hosteq(inp->inp_faddr, ip->ip_src))
354 			continue;
355 		if (last != NULL) {
356 			struct mbuf *n;
357 
358 			n = m_copym(m, 0, M_COPYALL, M_NOWAIT);
359 			if (n != NULL)
360 				(void) rip_append(last, ip, n, &ripsrc);
361 			/* XXX count dropped packet */
362 			INP_RUNLOCK(last);
363 			last = NULL;
364 		}
365 		INP_RLOCK(inp);
366 		if (__predict_false(inp->inp_flags2 & INP_FREED))
367 			goto skip_2;
368 		if (jailed_without_vnet(inp->inp_cred)) {
369 			/*
370 			 * Allow raw socket in jail to receive multicast;
371 			 * assume process had PRIV_NETINET_RAW at attach,
372 			 * and fall through into normal filter path if so.
373 			 */
374 			if (!IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
375 			    prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0)
376 				goto skip_2;
377 		}
378 		/*
379 		 * If this raw socket has multicast state, and we
380 		 * have received a multicast, check if this socket
381 		 * should receive it, as multicast filtering is now
382 		 * the responsibility of the transport layer.
383 		 */
384 		if (inp->inp_moptions != NULL &&
385 		    IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
386 			/*
387 			 * If the incoming datagram is for IGMP, allow it
388 			 * through unconditionally to the raw socket.
389 			 *
390 			 * In the case of IGMPv2, we may not have explicitly
391 			 * joined the group, and may have set IFF_ALLMULTI
392 			 * on the interface. imo_multi_filter() may discard
393 			 * control traffic we actually need to see.
394 			 *
395 			 * Userland multicast routing daemons should continue
396 			 * filter the control traffic appropriately.
397 			 */
398 			int blocked;
399 
400 			blocked = MCAST_PASS;
401 			if (proto != IPPROTO_IGMP) {
402 				struct sockaddr_in group;
403 
404 				bzero(&group, sizeof(struct sockaddr_in));
405 				group.sin_len = sizeof(struct sockaddr_in);
406 				group.sin_family = AF_INET;
407 				group.sin_addr = ip->ip_dst;
408 
409 				blocked = imo_multi_filter(inp->inp_moptions,
410 				    ifp,
411 				    (struct sockaddr *)&group,
412 				    (struct sockaddr *)&ripsrc);
413 			}
414 
415 			if (blocked != MCAST_PASS) {
416 				IPSTAT_INC(ips_notmember);
417 				goto skip_2;
418 			}
419 		}
420 		last = inp;
421 		continue;
422 	skip_2:
423 		INP_RUNLOCK(inp);
424 	}
425 	if (last != NULL) {
426 		if (rip_append(last, ip, m, &ripsrc) != 0)
427 			IPSTAT_INC(ips_delivered);
428 		INP_RUNLOCK(last);
429 	} else {
430 		if (inetsw[ip_protox[ip->ip_p]].pr_input == rip_input) {
431 			IPSTAT_INC(ips_noproto);
432 			IPSTAT_DEC(ips_delivered);
433 			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PROTOCOL, 0, 0);
434 		} else {
435 			m_freem(m);
436 		}
437 	}
438 	return (IPPROTO_DONE);
439 }
440 
441 /*
442  * Generate IP header and pass packet to ip_output.  Tack on options user may
443  * have setup with control call.
444  */
445 int
rip_output(struct mbuf * m,struct socket * so,...)446 rip_output(struct mbuf *m, struct socket *so, ...)
447 {
448 	struct epoch_tracker et;
449 	struct ip *ip;
450 	int error;
451 	struct inpcb *inp = sotoinpcb(so);
452 	va_list ap;
453 	u_long dst;
454 	int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) |
455 	    IP_ALLOWBROADCAST;
456 	int cnt, hlen;
457 	u_char opttype, optlen, *cp;
458 
459 	va_start(ap, so);
460 	dst = va_arg(ap, u_long);
461 	va_end(ap);
462 
463 	/*
464 	 * If the user handed us a complete IP packet, use it.  Otherwise,
465 	 * allocate an mbuf for a header and fill it in.
466 	 */
467 	if ((inp->inp_flags & INP_HDRINCL) == 0) {
468 		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
469 			m_freem(m);
470 			return(EMSGSIZE);
471 		}
472 		M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
473 		if (m == NULL)
474 			return(ENOBUFS);
475 
476 		INP_RLOCK(inp);
477 		ip = mtod(m, struct ip *);
478 		ip->ip_tos = inp->inp_ip_tos;
479 		if (inp->inp_flags & INP_DONTFRAG)
480 			ip->ip_off = htons(IP_DF);
481 		else
482 			ip->ip_off = htons(0);
483 		ip->ip_p = inp->inp_ip_p;
484 		ip->ip_len = htons(m->m_pkthdr.len);
485 		ip->ip_src = inp->inp_laddr;
486 		ip->ip_dst.s_addr = dst;
487 #ifdef ROUTE_MPATH
488 		if (CALC_FLOWID_OUTBOUND) {
489 			uint32_t hash_type, hash_val;
490 
491 			hash_val = fib4_calc_software_hash(ip->ip_src,
492 			    ip->ip_dst, 0, 0, ip->ip_p, &hash_type);
493 			m->m_pkthdr.flowid = hash_val;
494 			M_HASHTYPE_SET(m, hash_type);
495 			flags |= IP_NODEFAULTFLOWID;
496 		}
497 #endif
498 		if (jailed(inp->inp_cred)) {
499 			/*
500 			 * prison_local_ip4() would be good enough but would
501 			 * let a source of INADDR_ANY pass, which we do not
502 			 * want to see from jails.
503 			 */
504 			if (ip->ip_src.s_addr == INADDR_ANY) {
505 				NET_EPOCH_ENTER(et);
506 				error = in_pcbladdr(inp, &ip->ip_dst,
507 				    &ip->ip_src, inp->inp_cred);
508 				NET_EPOCH_EXIT(et);
509 			} else {
510 				error = prison_local_ip4(inp->inp_cred,
511 				    &ip->ip_src);
512 			}
513 			if (error != 0) {
514 				INP_RUNLOCK(inp);
515 				m_freem(m);
516 				return (error);
517 			}
518 		}
519 		ip->ip_ttl = inp->inp_ip_ttl;
520 	} else {
521 		if (m->m_pkthdr.len > IP_MAXPACKET) {
522 			m_freem(m);
523 			return (EMSGSIZE);
524 		}
525 		if (m->m_pkthdr.len < sizeof(*ip)) {
526 			m_freem(m);
527 			return (EINVAL);
528 		}
529 		m = m_pullup(m, sizeof(*ip));
530 		if (m == NULL)
531 			return (ENOMEM);
532 		ip = mtod(m, struct ip *);
533 		hlen = ip->ip_hl << 2;
534 		if (m->m_len < hlen) {
535 			m = m_pullup(m, hlen);
536 			if (m == NULL)
537 				return (EINVAL);
538 			ip = mtod(m, struct ip *);
539 		}
540 #ifdef ROUTE_MPATH
541 		if (CALC_FLOWID_OUTBOUND) {
542 			uint32_t hash_type, hash_val;
543 
544 			hash_val = fib4_calc_software_hash(ip->ip_dst,
545 			    ip->ip_src, 0, 0, ip->ip_p, &hash_type);
546 			m->m_pkthdr.flowid = hash_val;
547 			M_HASHTYPE_SET(m, hash_type);
548 			flags |= IP_NODEFAULTFLOWID;
549 		}
550 #endif
551 		INP_RLOCK(inp);
552 		/*
553 		 * Don't allow both user specified and setsockopt options,
554 		 * and don't allow packet length sizes that will crash.
555 		 */
556 		if ((hlen < sizeof (*ip))
557 		    || ((hlen > sizeof (*ip)) && inp->inp_options)
558 		    || (ntohs(ip->ip_len) != m->m_pkthdr.len)) {
559 			INP_RUNLOCK(inp);
560 			m_freem(m);
561 			return (EINVAL);
562 		}
563 		error = prison_check_ip4(inp->inp_cred, &ip->ip_src);
564 		if (error != 0) {
565 			INP_RUNLOCK(inp);
566 			m_freem(m);
567 			return (error);
568 		}
569 		/*
570 		 * Don't allow IP options which do not have the required
571 		 * structure as specified in section 3.1 of RFC 791 on
572 		 * pages 15-23.
573 		 */
574 		cp = (u_char *)(ip + 1);
575 		cnt = hlen - sizeof (struct ip);
576 		for (; cnt > 0; cnt -= optlen, cp += optlen) {
577 			opttype = cp[IPOPT_OPTVAL];
578 			if (opttype == IPOPT_EOL)
579 				break;
580 			if (opttype == IPOPT_NOP) {
581 				optlen = 1;
582 				continue;
583 			}
584 			if (cnt < IPOPT_OLEN + sizeof(u_char)) {
585 				INP_RUNLOCK(inp);
586 				m_freem(m);
587 				return (EINVAL);
588 			}
589 			optlen = cp[IPOPT_OLEN];
590 			if (optlen < IPOPT_OLEN + sizeof(u_char) ||
591 			    optlen > cnt) {
592 				INP_RUNLOCK(inp);
593 				m_freem(m);
594 				return (EINVAL);
595 			}
596 		}
597 		/*
598 		 * This doesn't allow application to specify ID of zero,
599 		 * but we got this limitation from the beginning of history.
600 		 */
601 		if (ip->ip_id == 0)
602 			ip_fillid(ip);
603 
604 		/*
605 		 * XXX prevent ip_output from overwriting header fields.
606 		 */
607 		flags |= IP_RAWOUTPUT;
608 		IPSTAT_INC(ips_rawout);
609 	}
610 
611 	if (inp->inp_flags & INP_ONESBCAST)
612 		flags |= IP_SENDONES;
613 
614 #ifdef MAC
615 	mac_inpcb_create_mbuf(inp, m);
616 #endif
617 
618 	NET_EPOCH_ENTER(et);
619 	error = ip_output(m, inp->inp_options, NULL, flags,
620 	    inp->inp_moptions, inp);
621 	NET_EPOCH_EXIT(et);
622 	INP_RUNLOCK(inp);
623 	return (error);
624 }
625 
626 /*
627  * Raw IP socket option processing.
628  *
629  * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could
630  * only be created by a privileged process, and as such, socket option
631  * operations to manage system properties on any raw socket were allowed to
632  * take place without explicit additional access control checks.  However,
633  * raw sockets can now also be created in jail(), and therefore explicit
634  * checks are now required.  Likewise, raw sockets can be used by a process
635  * after it gives up privilege, so some caution is required.  For options
636  * passed down to the IP layer via ip_ctloutput(), checks are assumed to be
637  * performed in ip_ctloutput() and therefore no check occurs here.
638  * Unilaterally checking priv_check() here breaks normal IP socket option
639  * operations on raw sockets.
640  *
641  * When adding new socket options here, make sure to add access control
642  * checks here as necessary.
643  *
644  * XXX-BZ inp locking?
645  */
646 int
rip_ctloutput(struct socket * so,struct sockopt * sopt)647 rip_ctloutput(struct socket *so, struct sockopt *sopt)
648 {
649 	struct	inpcb *inp = sotoinpcb(so);
650 	int	error, optval;
651 
652 	if (sopt->sopt_level != IPPROTO_IP) {
653 		if ((sopt->sopt_level == SOL_SOCKET) &&
654 		    (sopt->sopt_name == SO_SETFIB)) {
655 			inp->inp_inc.inc_fibnum = so->so_fibnum;
656 			return (0);
657 		}
658 		return (EINVAL);
659 	}
660 
661 	error = 0;
662 	switch (sopt->sopt_dir) {
663 	case SOPT_GET:
664 		switch (sopt->sopt_name) {
665 		case IP_HDRINCL:
666 			optval = inp->inp_flags & INP_HDRINCL;
667 			error = sooptcopyout(sopt, &optval, sizeof optval);
668 			break;
669 
670 		case IP_FW3:	/* generic ipfw v.3 functions */
671 		case IP_FW_ADD:	/* ADD actually returns the body... */
672 		case IP_FW_GET:
673 		case IP_FW_TABLE_GETSIZE:
674 		case IP_FW_TABLE_LIST:
675 		case IP_FW_NAT_GET_CONFIG:
676 		case IP_FW_NAT_GET_LOG:
677 			if (V_ip_fw_ctl_ptr != NULL)
678 				error = V_ip_fw_ctl_ptr(sopt);
679 			else
680 				error = ENOPROTOOPT;
681 			break;
682 
683 		case IP_DUMMYNET3:	/* generic dummynet v.3 functions */
684 		case IP_DUMMYNET_GET:
685 			if (ip_dn_ctl_ptr != NULL)
686 				error = ip_dn_ctl_ptr(sopt);
687 			else
688 				error = ENOPROTOOPT;
689 			break ;
690 
691 		case MRT_INIT:
692 		case MRT_DONE:
693 		case MRT_ADD_VIF:
694 		case MRT_DEL_VIF:
695 		case MRT_ADD_MFC:
696 		case MRT_DEL_MFC:
697 		case MRT_VERSION:
698 		case MRT_ASSERT:
699 		case MRT_API_SUPPORT:
700 		case MRT_API_CONFIG:
701 		case MRT_ADD_BW_UPCALL:
702 		case MRT_DEL_BW_UPCALL:
703 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
704 			if (error != 0)
705 				return (error);
706 			error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
707 				EOPNOTSUPP;
708 			break;
709 
710 		default:
711 			error = ip_ctloutput(so, sopt);
712 			break;
713 		}
714 		break;
715 
716 	case SOPT_SET:
717 		switch (sopt->sopt_name) {
718 		case IP_HDRINCL:
719 			error = sooptcopyin(sopt, &optval, sizeof optval,
720 					    sizeof optval);
721 			if (error)
722 				break;
723 			if (optval)
724 				inp->inp_flags |= INP_HDRINCL;
725 			else
726 				inp->inp_flags &= ~INP_HDRINCL;
727 			break;
728 
729 		case IP_FW3:	/* generic ipfw v.3 functions */
730 		case IP_FW_ADD:
731 		case IP_FW_DEL:
732 		case IP_FW_FLUSH:
733 		case IP_FW_ZERO:
734 		case IP_FW_RESETLOG:
735 		case IP_FW_TABLE_ADD:
736 		case IP_FW_TABLE_DEL:
737 		case IP_FW_TABLE_FLUSH:
738 		case IP_FW_NAT_CFG:
739 		case IP_FW_NAT_DEL:
740 			if (V_ip_fw_ctl_ptr != NULL)
741 				error = V_ip_fw_ctl_ptr(sopt);
742 			else
743 				error = ENOPROTOOPT;
744 			break;
745 
746 		case IP_DUMMYNET3:	/* generic dummynet v.3 functions */
747 		case IP_DUMMYNET_CONFIGURE:
748 		case IP_DUMMYNET_DEL:
749 		case IP_DUMMYNET_FLUSH:
750 			if (ip_dn_ctl_ptr != NULL)
751 				error = ip_dn_ctl_ptr(sopt);
752 			else
753 				error = ENOPROTOOPT ;
754 			break ;
755 
756 		case IP_RSVP_ON:
757 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
758 			if (error != 0)
759 				return (error);
760 			error = ip_rsvp_init(so);
761 			break;
762 
763 		case IP_RSVP_OFF:
764 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
765 			if (error != 0)
766 				return (error);
767 			error = ip_rsvp_done();
768 			break;
769 
770 		case IP_RSVP_VIF_ON:
771 		case IP_RSVP_VIF_OFF:
772 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
773 			if (error != 0)
774 				return (error);
775 			error = ip_rsvp_vif ?
776 				ip_rsvp_vif(so, sopt) : EINVAL;
777 			break;
778 
779 		case MRT_INIT:
780 		case MRT_DONE:
781 		case MRT_ADD_VIF:
782 		case MRT_DEL_VIF:
783 		case MRT_ADD_MFC:
784 		case MRT_DEL_MFC:
785 		case MRT_VERSION:
786 		case MRT_ASSERT:
787 		case MRT_API_SUPPORT:
788 		case MRT_API_CONFIG:
789 		case MRT_ADD_BW_UPCALL:
790 		case MRT_DEL_BW_UPCALL:
791 			error = priv_check(curthread, PRIV_NETINET_MROUTE);
792 			if (error != 0)
793 				return (error);
794 			error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
795 					EOPNOTSUPP;
796 			break;
797 
798 		default:
799 			error = ip_ctloutput(so, sopt);
800 			break;
801 		}
802 		break;
803 	}
804 
805 	return (error);
806 }
807 
808 /*
809  * This function exists solely to receive the PRC_IFDOWN messages which are
810  * sent by if_down().  It looks for an ifaddr whose ifa_addr is sa, and calls
811  * in_ifadown() to remove all routes corresponding to that address.  It also
812  * receives the PRC_IFUP messages from if_up() and reinstalls the interface
813  * routes.
814  */
815 void
rip_ctlinput(int cmd,struct sockaddr * sa,void * vip)816 rip_ctlinput(int cmd, struct sockaddr *sa, void *vip)
817 {
818 	struct rm_priotracker in_ifa_tracker;
819 	struct in_ifaddr *ia;
820 	struct ifnet *ifp;
821 	int err;
822 	int flags;
823 
824 	switch (cmd) {
825 	case PRC_IFDOWN:
826 		IN_IFADDR_RLOCK(&in_ifa_tracker);
827 		CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
828 			if (ia->ia_ifa.ifa_addr == sa
829 			    && (ia->ia_flags & IFA_ROUTE)) {
830 				ifa_ref(&ia->ia_ifa);
831 				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
832 				/*
833 				 * in_scrubprefix() kills the interface route.
834 				 */
835 				in_scrubprefix(ia, 0);
836 				/*
837 				 * in_ifadown gets rid of all the rest of the
838 				 * routes.  This is not quite the right thing
839 				 * to do, but at least if we are running a
840 				 * routing process they will come back.
841 				 */
842 				in_ifadown(&ia->ia_ifa, 0);
843 				ifa_free(&ia->ia_ifa);
844 				break;
845 			}
846 		}
847 		if (ia == NULL)		/* If ia matched, already unlocked. */
848 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
849 		break;
850 
851 	case PRC_IFUP:
852 		IN_IFADDR_RLOCK(&in_ifa_tracker);
853 		CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
854 			if (ia->ia_ifa.ifa_addr == sa)
855 				break;
856 		}
857 		if (ia == NULL || (ia->ia_flags & IFA_ROUTE)) {
858 			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
859 			return;
860 		}
861 		ifa_ref(&ia->ia_ifa);
862 		IN_IFADDR_RUNLOCK(&in_ifa_tracker);
863 		flags = RTF_UP;
864 		ifp = ia->ia_ifa.ifa_ifp;
865 
866 		if ((ifp->if_flags & IFF_LOOPBACK)
867 		    || (ifp->if_flags & IFF_POINTOPOINT))
868 			flags |= RTF_HOST;
869 
870 		err = ifa_del_loopback_route((struct ifaddr *)ia, sa);
871 
872 		rt_addrmsg(RTM_ADD, &ia->ia_ifa, ia->ia_ifp->if_fib);
873 		err = in_handle_ifaddr_route(RTM_ADD, ia);
874 		if (err == 0)
875 			ia->ia_flags |= IFA_ROUTE;
876 
877 		err = ifa_add_loopback_route((struct ifaddr *)ia, sa);
878 
879 		ifa_free(&ia->ia_ifa);
880 		break;
881 	}
882 }
883 
884 static int
rip_attach(struct socket * so,int proto,struct thread * td)885 rip_attach(struct socket *so, int proto, struct thread *td)
886 {
887 	struct inpcb *inp;
888 	int error;
889 
890 	inp = sotoinpcb(so);
891 	KASSERT(inp == NULL, ("rip_attach: inp != NULL"));
892 
893 	error = priv_check(td, PRIV_NETINET_RAW);
894 	if (error)
895 		return (error);
896 	if (proto >= IPPROTO_MAX || proto < 0)
897 		return EPROTONOSUPPORT;
898 	error = soreserve(so, rip_sendspace, rip_recvspace);
899 	if (error)
900 		return (error);
901 	INP_INFO_WLOCK(&V_ripcbinfo);
902 	error = in_pcballoc(so, &V_ripcbinfo);
903 	if (error) {
904 		INP_INFO_WUNLOCK(&V_ripcbinfo);
905 		return (error);
906 	}
907 	inp = (struct inpcb *)so->so_pcb;
908 	inp->inp_vflag |= INP_IPV4;
909 	inp->inp_ip_p = proto;
910 	inp->inp_ip_ttl = V_ip_defttl;
911 	rip_inshash(inp);
912 	INP_INFO_WUNLOCK(&V_ripcbinfo);
913 	INP_WUNLOCK(inp);
914 	return (0);
915 }
916 
917 static void
rip_detach(struct socket * so)918 rip_detach(struct socket *so)
919 {
920 	struct inpcb *inp;
921 
922 	inp = sotoinpcb(so);
923 	KASSERT(inp != NULL, ("rip_detach: inp == NULL"));
924 	KASSERT(inp->inp_faddr.s_addr == INADDR_ANY,
925 	    ("rip_detach: not closed"));
926 
927 	INP_INFO_WLOCK(&V_ripcbinfo);
928 	INP_WLOCK(inp);
929 	rip_delhash(inp);
930 	if (so == V_ip_mrouter && ip_mrouter_done)
931 		ip_mrouter_done();
932 	if (ip_rsvp_force_done)
933 		ip_rsvp_force_done(so);
934 	if (so == V_ip_rsvpd)
935 		ip_rsvp_done();
936 	in_pcbdetach(inp);
937 	in_pcbfree(inp);
938 	INP_INFO_WUNLOCK(&V_ripcbinfo);
939 }
940 
941 static void
rip_dodisconnect(struct socket * so,struct inpcb * inp)942 rip_dodisconnect(struct socket *so, struct inpcb *inp)
943 {
944 	struct inpcbinfo *pcbinfo;
945 
946 	pcbinfo = inp->inp_pcbinfo;
947 	INP_INFO_WLOCK(pcbinfo);
948 	INP_WLOCK(inp);
949 	rip_delhash(inp);
950 	inp->inp_faddr.s_addr = INADDR_ANY;
951 	rip_inshash(inp);
952 	SOCK_LOCK(so);
953 	so->so_state &= ~SS_ISCONNECTED;
954 	SOCK_UNLOCK(so);
955 	INP_WUNLOCK(inp);
956 	INP_INFO_WUNLOCK(pcbinfo);
957 }
958 
959 static void
rip_abort(struct socket * so)960 rip_abort(struct socket *so)
961 {
962 	struct inpcb *inp;
963 
964 	inp = sotoinpcb(so);
965 	KASSERT(inp != NULL, ("rip_abort: inp == NULL"));
966 
967 	rip_dodisconnect(so, inp);
968 }
969 
970 static void
rip_close(struct socket * so)971 rip_close(struct socket *so)
972 {
973 	struct inpcb *inp;
974 
975 	inp = sotoinpcb(so);
976 	KASSERT(inp != NULL, ("rip_close: inp == NULL"));
977 
978 	rip_dodisconnect(so, inp);
979 }
980 
981 static int
rip_disconnect(struct socket * so)982 rip_disconnect(struct socket *so)
983 {
984 	struct inpcb *inp;
985 
986 	if ((so->so_state & SS_ISCONNECTED) == 0)
987 		return (ENOTCONN);
988 
989 	inp = sotoinpcb(so);
990 	KASSERT(inp != NULL, ("rip_disconnect: inp == NULL"));
991 
992 	rip_dodisconnect(so, inp);
993 	return (0);
994 }
995 
996 static int
rip_bind(struct socket * so,struct sockaddr * nam,struct thread * td)997 rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
998 {
999 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
1000 	struct inpcb *inp;
1001 	int error;
1002 
1003 	if (nam->sa_family != AF_INET)
1004 		return (EAFNOSUPPORT);
1005 	if (nam->sa_len != sizeof(*addr))
1006 		return (EINVAL);
1007 
1008 	error = prison_check_ip4(td->td_ucred, &addr->sin_addr);
1009 	if (error != 0)
1010 		return (error);
1011 
1012 	inp = sotoinpcb(so);
1013 	KASSERT(inp != NULL, ("rip_bind: inp == NULL"));
1014 
1015 	if (CK_STAILQ_EMPTY(&V_ifnet) ||
1016 	    (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
1017 	    (addr->sin_addr.s_addr &&
1018 	     (inp->inp_flags & INP_BINDANY) == 0 &&
1019 	     ifa_ifwithaddr_check((struct sockaddr *)addr) == 0))
1020 		return (EADDRNOTAVAIL);
1021 
1022 	INP_INFO_WLOCK(&V_ripcbinfo);
1023 	INP_WLOCK(inp);
1024 	rip_delhash(inp);
1025 	inp->inp_laddr = addr->sin_addr;
1026 	rip_inshash(inp);
1027 	INP_WUNLOCK(inp);
1028 	INP_INFO_WUNLOCK(&V_ripcbinfo);
1029 	return (0);
1030 }
1031 
1032 static int
rip_connect(struct socket * so,struct sockaddr * nam,struct thread * td)1033 rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
1034 {
1035 	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
1036 	struct inpcb *inp;
1037 
1038 	if (nam->sa_len != sizeof(*addr))
1039 		return (EINVAL);
1040 	if (CK_STAILQ_EMPTY(&V_ifnet))
1041 		return (EADDRNOTAVAIL);
1042 	if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
1043 		return (EAFNOSUPPORT);
1044 
1045 	inp = sotoinpcb(so);
1046 	KASSERT(inp != NULL, ("rip_connect: inp == NULL"));
1047 
1048 	INP_INFO_WLOCK(&V_ripcbinfo);
1049 	INP_WLOCK(inp);
1050 	rip_delhash(inp);
1051 	inp->inp_faddr = addr->sin_addr;
1052 	rip_inshash(inp);
1053 	soisconnected(so);
1054 	INP_WUNLOCK(inp);
1055 	INP_INFO_WUNLOCK(&V_ripcbinfo);
1056 	return (0);
1057 }
1058 
1059 static int
rip_shutdown(struct socket * so)1060 rip_shutdown(struct socket *so)
1061 {
1062 	struct inpcb *inp;
1063 
1064 	inp = sotoinpcb(so);
1065 	KASSERT(inp != NULL, ("rip_shutdown: inp == NULL"));
1066 
1067 	INP_WLOCK(inp);
1068 	socantsendmore(so);
1069 	INP_WUNLOCK(inp);
1070 	return (0);
1071 }
1072 
1073 static int
rip_send(struct socket * so,int flags,struct mbuf * m,struct sockaddr * nam,struct mbuf * control,struct thread * td)1074 rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
1075     struct mbuf *control, struct thread *td)
1076 {
1077 	struct inpcb *inp;
1078 	u_long dst;
1079 	int error;
1080 
1081 	inp = sotoinpcb(so);
1082 	KASSERT(inp != NULL, ("rip_send: inp == NULL"));
1083 
1084 	if (control != NULL) {
1085 		m_freem(control);
1086 		control = NULL;
1087 	}
1088 
1089 	/*
1090 	 * Note: 'dst' reads below are unlocked.
1091 	 */
1092 	if (so->so_state & SS_ISCONNECTED) {
1093 		if (nam) {
1094 			error = EISCONN;
1095 			goto release;
1096 		}
1097 		dst = inp->inp_faddr.s_addr;	/* Unlocked read. */
1098 	} else {
1099 		error = 0;
1100 		if (nam == NULL)
1101 			error = ENOTCONN;
1102 		else if (nam->sa_family != AF_INET)
1103 			error = EAFNOSUPPORT;
1104 		else if (nam->sa_len != sizeof(struct sockaddr_in))
1105 			error = EINVAL;
1106 		if (error != 0)
1107 			goto release;
1108 		dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr;
1109 	}
1110 	return (rip_output(m, so, dst));
1111 
1112 release:
1113 	m_freem(m);
1114 	return (error);
1115 }
1116 #endif /* INET */
1117 
1118 static int
rip_pcblist(SYSCTL_HANDLER_ARGS)1119 rip_pcblist(SYSCTL_HANDLER_ARGS)
1120 {
1121 	struct xinpgen xig;
1122 	struct epoch_tracker et;
1123 	struct inpcb *inp;
1124 	int error;
1125 
1126 	if (req->newptr != 0)
1127 		return (EPERM);
1128 
1129 	if (req->oldptr == 0) {
1130 		int n;
1131 
1132 		n = V_ripcbinfo.ipi_count;
1133 		n += imax(n / 8, 10);
1134 		req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb);
1135 		return (0);
1136 	}
1137 
1138 	if ((error = sysctl_wire_old_buffer(req, 0)) != 0)
1139 		return (error);
1140 
1141 	bzero(&xig, sizeof(xig));
1142 	xig.xig_len = sizeof xig;
1143 	xig.xig_count = V_ripcbinfo.ipi_count;
1144 	xig.xig_gen = V_ripcbinfo.ipi_gencnt;
1145 	xig.xig_sogen = so_gencnt;
1146 	error = SYSCTL_OUT(req, &xig, sizeof xig);
1147 	if (error)
1148 		return (error);
1149 
1150 	NET_EPOCH_ENTER(et);
1151 	for (inp = CK_LIST_FIRST(V_ripcbinfo.ipi_listhead);
1152 	    inp != NULL;
1153 	    inp = CK_LIST_NEXT(inp, inp_list)) {
1154 		INP_RLOCK(inp);
1155 		if (inp->inp_gencnt <= xig.xig_gen &&
1156 		    cr_canseeinpcb(req->td->td_ucred, inp) == 0) {
1157 			struct xinpcb xi;
1158 
1159 			in_pcbtoxinpcb(inp, &xi);
1160 			INP_RUNLOCK(inp);
1161 			error = SYSCTL_OUT(req, &xi, sizeof xi);
1162 			if (error)
1163 				break;
1164 		} else
1165 			INP_RUNLOCK(inp);
1166 	}
1167 	NET_EPOCH_EXIT(et);
1168 
1169 	if (!error) {
1170 		/*
1171 		 * Give the user an updated idea of our state.  If the
1172 		 * generation differs from what we told her before, she knows
1173 		 * that something happened while we were processing this
1174 		 * request, and it might be necessary to retry.
1175 		 */
1176 		xig.xig_gen = V_ripcbinfo.ipi_gencnt;
1177 		xig.xig_sogen = so_gencnt;
1178 		xig.xig_count = V_ripcbinfo.ipi_count;
1179 		error = SYSCTL_OUT(req, &xig, sizeof xig);
1180 	}
1181 
1182 	return (error);
1183 }
1184 
1185 SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist,
1186     CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
1187     rip_pcblist, "S,xinpcb",
1188     "List of active raw IP sockets");
1189 
1190 #ifdef INET
1191 struct pr_usrreqs rip_usrreqs = {
1192 	.pru_abort =		rip_abort,
1193 	.pru_attach =		rip_attach,
1194 	.pru_bind =		rip_bind,
1195 	.pru_connect =		rip_connect,
1196 	.pru_control =		in_control,
1197 	.pru_detach =		rip_detach,
1198 	.pru_disconnect =	rip_disconnect,
1199 	.pru_peeraddr =		in_getpeeraddr,
1200 	.pru_send =		rip_send,
1201 	.pru_shutdown =		rip_shutdown,
1202 	.pru_sockaddr =		in_getsockaddr,
1203 	.pru_sosetlabel =	in_pcbsosetlabel,
1204 	.pru_close =		rip_close,
1205 };
1206 #endif /* INET */
1207