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: releng/11.0/sys/netinet/raw_ip.c 301114 2016-06-01 10:14:04Z bz $");
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#ifdef IPSEC
77#include <netipsec/ipsec.h>
78#endif /*IPSEC*/
79
80#include <machine/stdarg.h>
81#include <security/mac/mac_framework.h>
82
83VNET_DEFINE(int, ip_defttl) = IPDEFTTL;
84SYSCTL_INT(_net_inet_ip, IPCTL_DEFTTL, ttl, CTLFLAG_VNET | CTLFLAG_RW,
85    &VNET_NAME(ip_defttl), 0,
86    "Maximum TTL on IP packets");
87
88VNET_DEFINE(struct inpcbhead, ripcb);
89VNET_DEFINE(struct inpcbinfo, ripcbinfo);
90
91#define	V_ripcb			VNET(ripcb)
92#define	V_ripcbinfo		VNET(ripcbinfo)
93
94/*
95 * Control and data hooks for ipfw, dummynet, divert and so on.
96 * The data hooks are not used here but it is convenient
97 * to keep them all in one place.
98 */
99VNET_DEFINE(ip_fw_chk_ptr_t, ip_fw_chk_ptr) = NULL;
100VNET_DEFINE(ip_fw_ctl_ptr_t, ip_fw_ctl_ptr) = NULL;
101
102int	(*ip_dn_ctl_ptr)(struct sockopt *);
103int	(*ip_dn_io_ptr)(struct mbuf **, int, struct ip_fw_args *);
104void	(*ip_divert_ptr)(struct mbuf *, int);
105int	(*ng_ipfw_input_p)(struct mbuf **, int,
106			struct ip_fw_args *, int);
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 */
117VNET_DEFINE(struct socket *, ip_mrouter);
118
119/*
120 * The various mrouter and rsvp functions.
121 */
122int (*ip_mrouter_set)(struct socket *, struct sockopt *);
123int (*ip_mrouter_get)(struct socket *, struct sockopt *);
124int (*ip_mrouter_done)(void);
125int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *,
126		   struct ip_moptions *);
127int (*mrt_ioctl)(u_long, caddr_t, int);
128int (*legal_vif_num)(int);
129u_long (*ip_mcast_src)(int);
130
131int (*rsvp_input_p)(struct mbuf **, int *, int);
132int (*ip_rsvp_vif)(struct socket *, struct sockopt *);
133void (*ip_rsvp_force_done)(struct socket *);
134#endif /* INET */
135
136extern	struct protosw inetsw[];
137
138u_long	rip_sendspace = 9216;
139SYSCTL_ULONG(_net_inet_raw, OID_AUTO, maxdgram, CTLFLAG_RW,
140    &rip_sendspace, 0, "Maximum outgoing raw IP datagram size");
141
142u_long	rip_recvspace = 9216;
143SYSCTL_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
155static void
156rip_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	LIST_INSERT_HEAD(pcbhash, inp, inp_hash);
174}
175
176static void
177rip_delhash(struct inpcb *inp)
178{
179
180	INP_INFO_WLOCK_ASSERT(inp->inp_pcbinfo);
181	INP_WLOCK_ASSERT(inp);
182
183	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 */
194static void
195rip_zone_change(void *tag)
196{
197
198	uma_zone_set_max(V_ripcbinfo.ipi_zone, maxsockets);
199}
200
201static int
202rip_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
210void
211rip_init(void)
212{
213
214	in_pcbinfo_init(&V_ripcbinfo, "rip", &V_ripcb, INP_PCBHASH_RAW_SIZE,
215	    1, "ripcb", rip_inpcb_init, NULL, 0, IPI_HASHFIELDS_NONE);
216	EVENTHANDLER_REGISTER(maxsockets_change, rip_zone_change, NULL,
217	    EVENTHANDLER_PRI_ANY);
218}
219
220#ifdef VIMAGE
221static void
222rip_destroy(void *unused __unused)
223{
224
225	in_pcbinfo_destroy(&V_ripcbinfo);
226}
227VNET_SYSUNINIT(raw_ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, rip_destroy, NULL);
228#endif
229
230#ifdef INET
231static int
232rip_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#ifdef IPSEC
240	/* check AH/ESP integrity. */
241	if (ipsec4_in_reject(n, last)) {
242		policyfail = 1;
243	}
244#endif /* IPSEC */
245#ifdef MAC
246	if (!policyfail && mac_inpcb_check_deliver(last, n) != 0)
247		policyfail = 1;
248#endif
249	/* Check the minimum TTL for socket. */
250	if (last->inp_ip_minttl && last->inp_ip_minttl > ip->ip_ttl)
251		policyfail = 1;
252	if (!policyfail) {
253		struct mbuf *opts = NULL;
254		struct socket *so;
255
256		so = last->inp_socket;
257		if ((last->inp_flags & INP_CONTROLOPTS) ||
258		    (so->so_options & (SO_TIMESTAMP | SO_BINTIME)))
259			ip_savecontrol(last, &opts, ip, n);
260		SOCKBUF_LOCK(&so->so_rcv);
261		if (sbappendaddr_locked(&so->so_rcv,
262		    (struct sockaddr *)ripsrc, n, opts) == 0) {
263			/* should notify about lost packet */
264			m_freem(n);
265			if (opts)
266				m_freem(opts);
267			SOCKBUF_UNLOCK(&so->so_rcv);
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 */
279int
280rip_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	*mp = NULL;
290
291	bzero(&ripsrc, sizeof(ripsrc));
292	ripsrc.sin_len = sizeof(ripsrc);
293	ripsrc.sin_family = AF_INET;
294	ripsrc.sin_addr = ip->ip_src;
295	last = NULL;
296
297	ifp = m->m_pkthdr.rcvif;
298
299	hash = INP_PCBHASH_RAW(proto, ip->ip_src.s_addr,
300	    ip->ip_dst.s_addr, V_ripcbinfo.ipi_hashmask);
301	INP_INFO_RLOCK(&V_ripcbinfo);
302	LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[hash], inp_hash) {
303		if (inp->inp_ip_p != proto)
304			continue;
305#ifdef INET6
306		/* XXX inp locking */
307		if ((inp->inp_vflag & INP_IPV4) == 0)
308			continue;
309#endif
310		if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr)
311			continue;
312		if (inp->inp_faddr.s_addr != ip->ip_src.s_addr)
313			continue;
314		if (jailed_without_vnet(inp->inp_cred)) {
315			/*
316			 * XXX: If faddr was bound to multicast group,
317			 * jailed raw socket will drop datagram.
318			 */
319			if (prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0)
320				continue;
321		}
322		if (last != NULL) {
323			struct mbuf *n;
324
325			n = m_copy(m, 0, (int)M_COPYALL);
326			if (n != NULL)
327		    	    (void) rip_append(last, ip, n, &ripsrc);
328			/* XXX count dropped packet */
329			INP_RUNLOCK(last);
330		}
331		INP_RLOCK(inp);
332		last = inp;
333	}
334	LIST_FOREACH(inp, &V_ripcbinfo.ipi_hashbase[0], inp_hash) {
335		if (inp->inp_ip_p && inp->inp_ip_p != proto)
336			continue;
337#ifdef INET6
338		/* XXX inp locking */
339		if ((inp->inp_vflag & INP_IPV4) == 0)
340			continue;
341#endif
342		if (!in_nullhost(inp->inp_laddr) &&
343		    !in_hosteq(inp->inp_laddr, ip->ip_dst))
344			continue;
345		if (!in_nullhost(inp->inp_faddr) &&
346		    !in_hosteq(inp->inp_faddr, ip->ip_src))
347			continue;
348		if (jailed_without_vnet(inp->inp_cred)) {
349			/*
350			 * Allow raw socket in jail to receive multicast;
351			 * assume process had PRIV_NETINET_RAW at attach,
352			 * and fall through into normal filter path if so.
353			 */
354			if (!IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) &&
355			    prison_check_ip4(inp->inp_cred, &ip->ip_dst) != 0)
356				continue;
357		}
358		/*
359		 * If this raw socket has multicast state, and we
360		 * have received a multicast, check if this socket
361		 * should receive it, as multicast filtering is now
362		 * the responsibility of the transport layer.
363		 */
364		if (inp->inp_moptions != NULL &&
365		    IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) {
366			/*
367			 * If the incoming datagram is for IGMP, allow it
368			 * through unconditionally to the raw socket.
369			 *
370			 * In the case of IGMPv2, we may not have explicitly
371			 * joined the group, and may have set IFF_ALLMULTI
372			 * on the interface. imo_multi_filter() may discard
373			 * control traffic we actually need to see.
374			 *
375			 * Userland multicast routing daemons should continue
376			 * filter the control traffic appropriately.
377			 */
378			int blocked;
379
380			blocked = MCAST_PASS;
381			if (proto != IPPROTO_IGMP) {
382				struct sockaddr_in group;
383
384				bzero(&group, sizeof(struct sockaddr_in));
385				group.sin_len = sizeof(struct sockaddr_in);
386				group.sin_family = AF_INET;
387				group.sin_addr = ip->ip_dst;
388
389				blocked = imo_multi_filter(inp->inp_moptions,
390				    ifp,
391				    (struct sockaddr *)&group,
392				    (struct sockaddr *)&ripsrc);
393			}
394
395			if (blocked != MCAST_PASS) {
396				IPSTAT_INC(ips_notmember);
397				continue;
398			}
399		}
400		if (last != NULL) {
401			struct mbuf *n;
402
403			n = m_copy(m, 0, (int)M_COPYALL);
404			if (n != NULL)
405				(void) rip_append(last, ip, n, &ripsrc);
406			/* XXX count dropped packet */
407			INP_RUNLOCK(last);
408		}
409		INP_RLOCK(inp);
410		last = inp;
411	}
412	INP_INFO_RUNLOCK(&V_ripcbinfo);
413	if (last != NULL) {
414		if (rip_append(last, ip, m, &ripsrc) != 0)
415			IPSTAT_INC(ips_delivered);
416		INP_RUNLOCK(last);
417	} else {
418		if (inetsw[ip_protox[ip->ip_p]].pr_input == rip_input) {
419			IPSTAT_INC(ips_noproto);
420			IPSTAT_DEC(ips_delivered);
421			icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PROTOCOL, 0, 0);
422		} else {
423			m_freem(m);
424		}
425	}
426	return (IPPROTO_DONE);
427}
428
429/*
430 * Generate IP header and pass packet to ip_output.  Tack on options user may
431 * have setup with control call.
432 */
433int
434rip_output(struct mbuf *m, struct socket *so, ...)
435{
436	struct ip *ip;
437	int error;
438	struct inpcb *inp = sotoinpcb(so);
439	va_list ap;
440	u_long dst;
441	int flags = ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) |
442	    IP_ALLOWBROADCAST;
443
444	va_start(ap, so);
445	dst = va_arg(ap, u_long);
446	va_end(ap);
447
448	/*
449	 * If the user handed us a complete IP packet, use it.  Otherwise,
450	 * allocate an mbuf for a header and fill it in.
451	 */
452	if ((inp->inp_flags & INP_HDRINCL) == 0) {
453		if (m->m_pkthdr.len + sizeof(struct ip) > IP_MAXPACKET) {
454			m_freem(m);
455			return(EMSGSIZE);
456		}
457		M_PREPEND(m, sizeof(struct ip), M_NOWAIT);
458		if (m == NULL)
459			return(ENOBUFS);
460
461		INP_RLOCK(inp);
462		ip = mtod(m, struct ip *);
463		ip->ip_tos = inp->inp_ip_tos;
464		if (inp->inp_flags & INP_DONTFRAG)
465			ip->ip_off = htons(IP_DF);
466		else
467			ip->ip_off = htons(0);
468		ip->ip_p = inp->inp_ip_p;
469		ip->ip_len = htons(m->m_pkthdr.len);
470		ip->ip_src = inp->inp_laddr;
471		ip->ip_dst.s_addr = dst;
472		if (jailed(inp->inp_cred)) {
473			/*
474			 * prison_local_ip4() would be good enough but would
475			 * let a source of INADDR_ANY pass, which we do not
476			 * want to see from jails.
477			 */
478			if (ip->ip_src.s_addr == INADDR_ANY) {
479				error = in_pcbladdr(inp, &ip->ip_dst, &ip->ip_src,
480				    inp->inp_cred);
481			} else {
482				error = prison_local_ip4(inp->inp_cred,
483				    &ip->ip_src);
484			}
485			if (error != 0) {
486				INP_RUNLOCK(inp);
487				m_freem(m);
488				return (error);
489			}
490		}
491		ip->ip_ttl = inp->inp_ip_ttl;
492	} else {
493		if (m->m_pkthdr.len > IP_MAXPACKET) {
494			m_freem(m);
495			return(EMSGSIZE);
496		}
497		INP_RLOCK(inp);
498		ip = mtod(m, struct ip *);
499		error = prison_check_ip4(inp->inp_cred, &ip->ip_src);
500		if (error != 0) {
501			INP_RUNLOCK(inp);
502			m_freem(m);
503			return (error);
504		}
505
506		/*
507		 * Don't allow both user specified and setsockopt options,
508		 * and don't allow packet length sizes that will crash.
509		 */
510		if (((ip->ip_hl != (sizeof (*ip) >> 2)) && inp->inp_options)
511		    || (ntohs(ip->ip_len) > m->m_pkthdr.len)
512		    || (ntohs(ip->ip_len) < (ip->ip_hl << 2))) {
513			INP_RUNLOCK(inp);
514			m_freem(m);
515			return (EINVAL);
516		}
517		/*
518		 * This doesn't allow application to specify ID of zero,
519		 * but we got this limitation from the beginning of history.
520		 */
521		if (ip->ip_id == 0)
522			ip_fillid(ip);
523
524		/*
525		 * XXX prevent ip_output from overwriting header fields.
526		 */
527		flags |= IP_RAWOUTPUT;
528		IPSTAT_INC(ips_rawout);
529	}
530
531	if (inp->inp_flags & INP_ONESBCAST)
532		flags |= IP_SENDONES;
533
534#ifdef MAC
535	mac_inpcb_create_mbuf(inp, m);
536#endif
537
538	error = ip_output(m, inp->inp_options, NULL, flags,
539	    inp->inp_moptions, inp);
540	INP_RUNLOCK(inp);
541	return (error);
542}
543
544/*
545 * Raw IP socket option processing.
546 *
547 * IMPORTANT NOTE regarding access control: Traditionally, raw sockets could
548 * only be created by a privileged process, and as such, socket option
549 * operations to manage system properties on any raw socket were allowed to
550 * take place without explicit additional access control checks.  However,
551 * raw sockets can now also be created in jail(), and therefore explicit
552 * checks are now required.  Likewise, raw sockets can be used by a process
553 * after it gives up privilege, so some caution is required.  For options
554 * passed down to the IP layer via ip_ctloutput(), checks are assumed to be
555 * performed in ip_ctloutput() and therefore no check occurs here.
556 * Unilaterally checking priv_check() here breaks normal IP socket option
557 * operations on raw sockets.
558 *
559 * When adding new socket options here, make sure to add access control
560 * checks here as necessary.
561 *
562 * XXX-BZ inp locking?
563 */
564int
565rip_ctloutput(struct socket *so, struct sockopt *sopt)
566{
567	struct	inpcb *inp = sotoinpcb(so);
568	int	error, optval;
569
570	if (sopt->sopt_level != IPPROTO_IP) {
571		if ((sopt->sopt_level == SOL_SOCKET) &&
572		    (sopt->sopt_name == SO_SETFIB)) {
573			inp->inp_inc.inc_fibnum = so->so_fibnum;
574			return (0);
575		}
576		return (EINVAL);
577	}
578
579	error = 0;
580	switch (sopt->sopt_dir) {
581	case SOPT_GET:
582		switch (sopt->sopt_name) {
583		case IP_HDRINCL:
584			optval = inp->inp_flags & INP_HDRINCL;
585			error = sooptcopyout(sopt, &optval, sizeof optval);
586			break;
587
588		case IP_FW3:	/* generic ipfw v.3 functions */
589		case IP_FW_ADD:	/* ADD actually returns the body... */
590		case IP_FW_GET:
591		case IP_FW_TABLE_GETSIZE:
592		case IP_FW_TABLE_LIST:
593		case IP_FW_NAT_GET_CONFIG:
594		case IP_FW_NAT_GET_LOG:
595			if (V_ip_fw_ctl_ptr != NULL)
596				error = V_ip_fw_ctl_ptr(sopt);
597			else
598				error = ENOPROTOOPT;
599			break;
600
601		case IP_DUMMYNET3:	/* generic dummynet v.3 functions */
602		case IP_DUMMYNET_GET:
603			if (ip_dn_ctl_ptr != NULL)
604				error = ip_dn_ctl_ptr(sopt);
605			else
606				error = ENOPROTOOPT;
607			break ;
608
609		case MRT_INIT:
610		case MRT_DONE:
611		case MRT_ADD_VIF:
612		case MRT_DEL_VIF:
613		case MRT_ADD_MFC:
614		case MRT_DEL_MFC:
615		case MRT_VERSION:
616		case MRT_ASSERT:
617		case MRT_API_SUPPORT:
618		case MRT_API_CONFIG:
619		case MRT_ADD_BW_UPCALL:
620		case MRT_DEL_BW_UPCALL:
621			error = priv_check(curthread, PRIV_NETINET_MROUTE);
622			if (error != 0)
623				return (error);
624			error = ip_mrouter_get ? ip_mrouter_get(so, sopt) :
625				EOPNOTSUPP;
626			break;
627
628		default:
629			error = ip_ctloutput(so, sopt);
630			break;
631		}
632		break;
633
634	case SOPT_SET:
635		switch (sopt->sopt_name) {
636		case IP_HDRINCL:
637			error = sooptcopyin(sopt, &optval, sizeof optval,
638					    sizeof optval);
639			if (error)
640				break;
641			if (optval)
642				inp->inp_flags |= INP_HDRINCL;
643			else
644				inp->inp_flags &= ~INP_HDRINCL;
645			break;
646
647		case IP_FW3:	/* generic ipfw v.3 functions */
648		case IP_FW_ADD:
649		case IP_FW_DEL:
650		case IP_FW_FLUSH:
651		case IP_FW_ZERO:
652		case IP_FW_RESETLOG:
653		case IP_FW_TABLE_ADD:
654		case IP_FW_TABLE_DEL:
655		case IP_FW_TABLE_FLUSH:
656		case IP_FW_NAT_CFG:
657		case IP_FW_NAT_DEL:
658			if (V_ip_fw_ctl_ptr != NULL)
659				error = V_ip_fw_ctl_ptr(sopt);
660			else
661				error = ENOPROTOOPT;
662			break;
663
664		case IP_DUMMYNET3:	/* generic dummynet v.3 functions */
665		case IP_DUMMYNET_CONFIGURE:
666		case IP_DUMMYNET_DEL:
667		case IP_DUMMYNET_FLUSH:
668			if (ip_dn_ctl_ptr != NULL)
669				error = ip_dn_ctl_ptr(sopt);
670			else
671				error = ENOPROTOOPT ;
672			break ;
673
674		case IP_RSVP_ON:
675			error = priv_check(curthread, PRIV_NETINET_MROUTE);
676			if (error != 0)
677				return (error);
678			error = ip_rsvp_init(so);
679			break;
680
681		case IP_RSVP_OFF:
682			error = priv_check(curthread, PRIV_NETINET_MROUTE);
683			if (error != 0)
684				return (error);
685			error = ip_rsvp_done();
686			break;
687
688		case IP_RSVP_VIF_ON:
689		case IP_RSVP_VIF_OFF:
690			error = priv_check(curthread, PRIV_NETINET_MROUTE);
691			if (error != 0)
692				return (error);
693			error = ip_rsvp_vif ?
694				ip_rsvp_vif(so, sopt) : EINVAL;
695			break;
696
697		case MRT_INIT:
698		case MRT_DONE:
699		case MRT_ADD_VIF:
700		case MRT_DEL_VIF:
701		case MRT_ADD_MFC:
702		case MRT_DEL_MFC:
703		case MRT_VERSION:
704		case MRT_ASSERT:
705		case MRT_API_SUPPORT:
706		case MRT_API_CONFIG:
707		case MRT_ADD_BW_UPCALL:
708		case MRT_DEL_BW_UPCALL:
709			error = priv_check(curthread, PRIV_NETINET_MROUTE);
710			if (error != 0)
711				return (error);
712			error = ip_mrouter_set ? ip_mrouter_set(so, sopt) :
713					EOPNOTSUPP;
714			break;
715
716		default:
717			error = ip_ctloutput(so, sopt);
718			break;
719		}
720		break;
721	}
722
723	return (error);
724}
725
726/*
727 * This function exists solely to receive the PRC_IFDOWN messages which are
728 * sent by if_down().  It looks for an ifaddr whose ifa_addr is sa, and calls
729 * in_ifadown() to remove all routes corresponding to that address.  It also
730 * receives the PRC_IFUP messages from if_up() and reinstalls the interface
731 * routes.
732 */
733void
734rip_ctlinput(int cmd, struct sockaddr *sa, void *vip)
735{
736	struct rm_priotracker in_ifa_tracker;
737	struct in_ifaddr *ia;
738	struct ifnet *ifp;
739	int err;
740	int flags;
741
742	switch (cmd) {
743	case PRC_IFDOWN:
744		IN_IFADDR_RLOCK(&in_ifa_tracker);
745		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
746			if (ia->ia_ifa.ifa_addr == sa
747			    && (ia->ia_flags & IFA_ROUTE)) {
748				ifa_ref(&ia->ia_ifa);
749				IN_IFADDR_RUNLOCK(&in_ifa_tracker);
750				/*
751				 * in_scrubprefix() kills the interface route.
752				 */
753				in_scrubprefix(ia, 0);
754				/*
755				 * in_ifadown gets rid of all the rest of the
756				 * routes.  This is not quite the right thing
757				 * to do, but at least if we are running a
758				 * routing process they will come back.
759				 */
760				in_ifadown(&ia->ia_ifa, 0);
761				ifa_free(&ia->ia_ifa);
762				break;
763			}
764		}
765		if (ia == NULL)		/* If ia matched, already unlocked. */
766			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
767		break;
768
769	case PRC_IFUP:
770		IN_IFADDR_RLOCK(&in_ifa_tracker);
771		TAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) {
772			if (ia->ia_ifa.ifa_addr == sa)
773				break;
774		}
775		if (ia == NULL || (ia->ia_flags & IFA_ROUTE)) {
776			IN_IFADDR_RUNLOCK(&in_ifa_tracker);
777			return;
778		}
779		ifa_ref(&ia->ia_ifa);
780		IN_IFADDR_RUNLOCK(&in_ifa_tracker);
781		flags = RTF_UP;
782		ifp = ia->ia_ifa.ifa_ifp;
783
784		if ((ifp->if_flags & IFF_LOOPBACK)
785		    || (ifp->if_flags & IFF_POINTOPOINT))
786			flags |= RTF_HOST;
787
788		err = ifa_del_loopback_route((struct ifaddr *)ia, sa);
789
790		err = rtinit(&ia->ia_ifa, RTM_ADD, flags);
791		if (err == 0)
792			ia->ia_flags |= IFA_ROUTE;
793
794		err = ifa_add_loopback_route((struct ifaddr *)ia, sa);
795
796		ifa_free(&ia->ia_ifa);
797		break;
798	}
799}
800
801static int
802rip_attach(struct socket *so, int proto, struct thread *td)
803{
804	struct inpcb *inp;
805	int error;
806
807	inp = sotoinpcb(so);
808	KASSERT(inp == NULL, ("rip_attach: inp != NULL"));
809
810	error = priv_check(td, PRIV_NETINET_RAW);
811	if (error)
812		return (error);
813	if (proto >= IPPROTO_MAX || proto < 0)
814		return EPROTONOSUPPORT;
815	error = soreserve(so, rip_sendspace, rip_recvspace);
816	if (error)
817		return (error);
818	INP_INFO_WLOCK(&V_ripcbinfo);
819	error = in_pcballoc(so, &V_ripcbinfo);
820	if (error) {
821		INP_INFO_WUNLOCK(&V_ripcbinfo);
822		return (error);
823	}
824	inp = (struct inpcb *)so->so_pcb;
825	inp->inp_vflag |= INP_IPV4;
826	inp->inp_ip_p = proto;
827	inp->inp_ip_ttl = V_ip_defttl;
828	rip_inshash(inp);
829	INP_INFO_WUNLOCK(&V_ripcbinfo);
830	INP_WUNLOCK(inp);
831	return (0);
832}
833
834static void
835rip_detach(struct socket *so)
836{
837	struct inpcb *inp;
838
839	inp = sotoinpcb(so);
840	KASSERT(inp != NULL, ("rip_detach: inp == NULL"));
841	KASSERT(inp->inp_faddr.s_addr == INADDR_ANY,
842	    ("rip_detach: not closed"));
843
844	INP_INFO_WLOCK(&V_ripcbinfo);
845	INP_WLOCK(inp);
846	rip_delhash(inp);
847	if (so == V_ip_mrouter && ip_mrouter_done)
848		ip_mrouter_done();
849	if (ip_rsvp_force_done)
850		ip_rsvp_force_done(so);
851	if (so == V_ip_rsvpd)
852		ip_rsvp_done();
853	in_pcbdetach(inp);
854	in_pcbfree(inp);
855	INP_INFO_WUNLOCK(&V_ripcbinfo);
856}
857
858static void
859rip_dodisconnect(struct socket *so, struct inpcb *inp)
860{
861	struct inpcbinfo *pcbinfo;
862
863	pcbinfo = inp->inp_pcbinfo;
864	INP_INFO_WLOCK(pcbinfo);
865	INP_WLOCK(inp);
866	rip_delhash(inp);
867	inp->inp_faddr.s_addr = INADDR_ANY;
868	rip_inshash(inp);
869	SOCK_LOCK(so);
870	so->so_state &= ~SS_ISCONNECTED;
871	SOCK_UNLOCK(so);
872	INP_WUNLOCK(inp);
873	INP_INFO_WUNLOCK(pcbinfo);
874}
875
876static void
877rip_abort(struct socket *so)
878{
879	struct inpcb *inp;
880
881	inp = sotoinpcb(so);
882	KASSERT(inp != NULL, ("rip_abort: inp == NULL"));
883
884	rip_dodisconnect(so, inp);
885}
886
887static void
888rip_close(struct socket *so)
889{
890	struct inpcb *inp;
891
892	inp = sotoinpcb(so);
893	KASSERT(inp != NULL, ("rip_close: inp == NULL"));
894
895	rip_dodisconnect(so, inp);
896}
897
898static int
899rip_disconnect(struct socket *so)
900{
901	struct inpcb *inp;
902
903	if ((so->so_state & SS_ISCONNECTED) == 0)
904		return (ENOTCONN);
905
906	inp = sotoinpcb(so);
907	KASSERT(inp != NULL, ("rip_disconnect: inp == NULL"));
908
909	rip_dodisconnect(so, inp);
910	return (0);
911}
912
913static int
914rip_bind(struct socket *so, struct sockaddr *nam, struct thread *td)
915{
916	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
917	struct inpcb *inp;
918	int error;
919
920	if (nam->sa_len != sizeof(*addr))
921		return (EINVAL);
922
923	error = prison_check_ip4(td->td_ucred, &addr->sin_addr);
924	if (error != 0)
925		return (error);
926
927	inp = sotoinpcb(so);
928	KASSERT(inp != NULL, ("rip_bind: inp == NULL"));
929
930	if (TAILQ_EMPTY(&V_ifnet) ||
931	    (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK) ||
932	    (addr->sin_addr.s_addr &&
933	     (inp->inp_flags & INP_BINDANY) == 0 &&
934	     ifa_ifwithaddr_check((struct sockaddr *)addr) == 0))
935		return (EADDRNOTAVAIL);
936
937	INP_INFO_WLOCK(&V_ripcbinfo);
938	INP_WLOCK(inp);
939	rip_delhash(inp);
940	inp->inp_laddr = addr->sin_addr;
941	rip_inshash(inp);
942	INP_WUNLOCK(inp);
943	INP_INFO_WUNLOCK(&V_ripcbinfo);
944	return (0);
945}
946
947static int
948rip_connect(struct socket *so, struct sockaddr *nam, struct thread *td)
949{
950	struct sockaddr_in *addr = (struct sockaddr_in *)nam;
951	struct inpcb *inp;
952
953	if (nam->sa_len != sizeof(*addr))
954		return (EINVAL);
955	if (TAILQ_EMPTY(&V_ifnet))
956		return (EADDRNOTAVAIL);
957	if (addr->sin_family != AF_INET && addr->sin_family != AF_IMPLINK)
958		return (EAFNOSUPPORT);
959
960	inp = sotoinpcb(so);
961	KASSERT(inp != NULL, ("rip_connect: inp == NULL"));
962
963	INP_INFO_WLOCK(&V_ripcbinfo);
964	INP_WLOCK(inp);
965	rip_delhash(inp);
966	inp->inp_faddr = addr->sin_addr;
967	rip_inshash(inp);
968	soisconnected(so);
969	INP_WUNLOCK(inp);
970	INP_INFO_WUNLOCK(&V_ripcbinfo);
971	return (0);
972}
973
974static int
975rip_shutdown(struct socket *so)
976{
977	struct inpcb *inp;
978
979	inp = sotoinpcb(so);
980	KASSERT(inp != NULL, ("rip_shutdown: inp == NULL"));
981
982	INP_WLOCK(inp);
983	socantsendmore(so);
984	INP_WUNLOCK(inp);
985	return (0);
986}
987
988static int
989rip_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam,
990    struct mbuf *control, struct thread *td)
991{
992	struct inpcb *inp;
993	u_long dst;
994
995	inp = sotoinpcb(so);
996	KASSERT(inp != NULL, ("rip_send: inp == NULL"));
997
998	/*
999	 * Note: 'dst' reads below are unlocked.
1000	 */
1001	if (so->so_state & SS_ISCONNECTED) {
1002		if (nam) {
1003			m_freem(m);
1004			return (EISCONN);
1005		}
1006		dst = inp->inp_faddr.s_addr;	/* Unlocked read. */
1007	} else {
1008		if (nam == NULL) {
1009			m_freem(m);
1010			return (ENOTCONN);
1011		}
1012		dst = ((struct sockaddr_in *)nam)->sin_addr.s_addr;
1013	}
1014	return (rip_output(m, so, dst));
1015}
1016#endif /* INET */
1017
1018static int
1019rip_pcblist(SYSCTL_HANDLER_ARGS)
1020{
1021	int error, i, n;
1022	struct inpcb *inp, **inp_list;
1023	inp_gen_t gencnt;
1024	struct xinpgen xig;
1025
1026	/*
1027	 * The process of preparing the TCB list is too time-consuming and
1028	 * resource-intensive to repeat twice on every request.
1029	 */
1030	if (req->oldptr == 0) {
1031		n = V_ripcbinfo.ipi_count;
1032		n += imax(n / 8, 10);
1033		req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb);
1034		return (0);
1035	}
1036
1037	if (req->newptr != 0)
1038		return (EPERM);
1039
1040	/*
1041	 * OK, now we're committed to doing something.
1042	 */
1043	INP_INFO_RLOCK(&V_ripcbinfo);
1044	gencnt = V_ripcbinfo.ipi_gencnt;
1045	n = V_ripcbinfo.ipi_count;
1046	INP_INFO_RUNLOCK(&V_ripcbinfo);
1047
1048	xig.xig_len = sizeof xig;
1049	xig.xig_count = n;
1050	xig.xig_gen = gencnt;
1051	xig.xig_sogen = so_gencnt;
1052	error = SYSCTL_OUT(req, &xig, sizeof xig);
1053	if (error)
1054		return (error);
1055
1056	inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK);
1057	if (inp_list == NULL)
1058		return (ENOMEM);
1059
1060	INP_INFO_RLOCK(&V_ripcbinfo);
1061	for (inp = LIST_FIRST(V_ripcbinfo.ipi_listhead), i = 0; inp && i < n;
1062	     inp = LIST_NEXT(inp, inp_list)) {
1063		INP_WLOCK(inp);
1064		if (inp->inp_gencnt <= gencnt &&
1065		    cr_canseeinpcb(req->td->td_ucred, inp) == 0) {
1066			in_pcbref(inp);
1067			inp_list[i++] = inp;
1068		}
1069		INP_WUNLOCK(inp);
1070	}
1071	INP_INFO_RUNLOCK(&V_ripcbinfo);
1072	n = i;
1073
1074	error = 0;
1075	for (i = 0; i < n; i++) {
1076		inp = inp_list[i];
1077		INP_RLOCK(inp);
1078		if (inp->inp_gencnt <= gencnt) {
1079			struct xinpcb xi;
1080
1081			bzero(&xi, sizeof(xi));
1082			xi.xi_len = sizeof xi;
1083			/* XXX should avoid extra copy */
1084			bcopy(inp, &xi.xi_inp, sizeof *inp);
1085			if (inp->inp_socket)
1086				sotoxsocket(inp->inp_socket, &xi.xi_socket);
1087			INP_RUNLOCK(inp);
1088			error = SYSCTL_OUT(req, &xi, sizeof xi);
1089		} else
1090			INP_RUNLOCK(inp);
1091	}
1092	INP_INFO_WLOCK(&V_ripcbinfo);
1093	for (i = 0; i < n; i++) {
1094		inp = inp_list[i];
1095		INP_RLOCK(inp);
1096		if (!in_pcbrele_rlocked(inp))
1097			INP_RUNLOCK(inp);
1098	}
1099	INP_INFO_WUNLOCK(&V_ripcbinfo);
1100
1101	if (!error) {
1102		/*
1103		 * Give the user an updated idea of our state.  If the
1104		 * generation differs from what we told her before, she knows
1105		 * that something happened while we were processing this
1106		 * request, and it might be necessary to retry.
1107		 */
1108		INP_INFO_RLOCK(&V_ripcbinfo);
1109		xig.xig_gen = V_ripcbinfo.ipi_gencnt;
1110		xig.xig_sogen = so_gencnt;
1111		xig.xig_count = V_ripcbinfo.ipi_count;
1112		INP_INFO_RUNLOCK(&V_ripcbinfo);
1113		error = SYSCTL_OUT(req, &xig, sizeof xig);
1114	}
1115	free(inp_list, M_TEMP);
1116	return (error);
1117}
1118
1119SYSCTL_PROC(_net_inet_raw, OID_AUTO/*XXX*/, pcblist,
1120    CTLTYPE_OPAQUE | CTLFLAG_RD, NULL, 0,
1121    rip_pcblist, "S,xinpcb", "List of active raw IP sockets");
1122
1123#ifdef INET
1124struct pr_usrreqs rip_usrreqs = {
1125	.pru_abort =		rip_abort,
1126	.pru_attach =		rip_attach,
1127	.pru_bind =		rip_bind,
1128	.pru_connect =		rip_connect,
1129	.pru_control =		in_control,
1130	.pru_detach =		rip_detach,
1131	.pru_disconnect =	rip_disconnect,
1132	.pru_peeraddr =		in_getpeeraddr,
1133	.pru_send =		rip_send,
1134	.pru_shutdown =		rip_shutdown,
1135	.pru_sockaddr =		in_getsockaddr,
1136	.pru_sosetlabel =	in_pcbsosetlabel,
1137	.pru_close =		rip_close,
1138};
1139#endif /* INET */
1140