Merge from vendor branch DHCP:
[dragonfly.git] / sys / netinet / in_pcb.c
1 /*
2  * Copyright (c) 2004 Jeffrey M. Hsu.  All rights reserved.
3  * Copyright (c) 2004 The DragonFly Project.  All rights reserved.
4  *
5  * This code is derived from software contributed to The DragonFly Project
6  * by Jeffrey M. Hsu.
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 DragonFly Project nor the names of its
17  *    contributors may be used to endorse or promote products derived
18  *    from this software without specific, prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33
34 /*
35  * Copyright (c) 2004 Jeffrey M. Hsu.  All rights reserved.
36  *
37  * License terms: all terms for the DragonFly license above plus the following:
38  *
39  * 4. All advertising materials mentioning features or use of this software
40  *    must display the following acknowledgement:
41  *
42  *      This product includes software developed by Jeffrey M. Hsu
43  *      for the DragonFly Project.
44  *
45  *    This requirement may be waived with permission from Jeffrey Hsu.
46  *    This requirement will sunset and may be removed on July 8 2005,
47  *    after which the standard DragonFly license (as shown above) will
48  *    apply.
49  */
50
51 /*
52  * Copyright (c) 1982, 1986, 1991, 1993, 1995
53  *      The Regents of the University of California.  All rights reserved.
54  *
55  * Redistribution and use in source and binary forms, with or without
56  * modification, are permitted provided that the following conditions
57  * are met:
58  * 1. Redistributions of source code must retain the above copyright
59  *    notice, this list of conditions and the following disclaimer.
60  * 2. Redistributions in binary form must reproduce the above copyright
61  *    notice, this list of conditions and the following disclaimer in the
62  *    documentation and/or other materials provided with the distribution.
63  * 3. All advertising materials mentioning features or use of this software
64  *    must display the following acknowledgement:
65  *      This product includes software developed by the University of
66  *      California, Berkeley and its contributors.
67  * 4. Neither the name of the University nor the names of its contributors
68  *    may be used to endorse or promote products derived from this software
69  *    without specific prior written permission.
70  *
71  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
72  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
73  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
74  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
75  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
76  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
77  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
78  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
79  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
80  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
81  * SUCH DAMAGE.
82  *
83  *      @(#)in_pcb.c    8.4 (Berkeley) 5/24/95
84  * $FreeBSD: src/sys/netinet/in_pcb.c,v 1.59.2.27 2004/01/02 04:06:42 ambrisko Exp $
85  * $DragonFly: src/sys/netinet/in_pcb.c,v 1.31 2004/12/30 05:14:30 hsu Exp $
86  */
87
88 #include "opt_ipsec.h"
89 #include "opt_inet6.h"
90
91 #include <sys/param.h>
92 #include <sys/systm.h>
93 #include <sys/malloc.h>
94 #include <sys/mbuf.h>
95 #include <sys/domain.h>
96 #include <sys/protosw.h>
97 #include <sys/socket.h>
98 #include <sys/socketvar.h>
99 #include <sys/proc.h>
100 #include <sys/jail.h>
101 #include <sys/kernel.h>
102 #include <sys/sysctl.h>
103
104 #include <machine/limits.h>
105
106 #include <vm/vm_zone.h>
107
108 #include <net/if.h>
109 #include <net/if_types.h>
110 #include <net/route.h>
111
112 #include <netinet/in.h>
113 #include <netinet/in_pcb.h>
114 #include <netinet/in_var.h>
115 #include <netinet/ip_var.h>
116 #ifdef INET6
117 #include <netinet/ip6.h>
118 #include <netinet6/ip6_var.h>
119 #endif /* INET6 */
120
121 #ifdef IPSEC
122 #include <netinet6/ipsec.h>
123 #include <netproto/key/key.h>
124 #endif
125
126 #ifdef FAST_IPSEC
127 #if defined(IPSEC) || defined(IPSEC_ESP)
128 #error "Bad idea: don't compile with both IPSEC and FAST_IPSEC!"
129 #endif
130
131 #include <netproto/ipsec/ipsec.h>
132 #include <netproto/ipsec/key.h>
133 #define IPSEC
134 #endif /* FAST_IPSEC */
135
136 struct in_addr zeroin_addr;
137
138 /*
139  * These configure the range of local port addresses assigned to
140  * "unspecified" outgoing connections/packets/whatever.
141  */
142 int ipport_lowfirstauto = IPPORT_RESERVED - 1;  /* 1023 */
143 int ipport_lowlastauto = IPPORT_RESERVEDSTART;  /* 600 */
144
145 int ipport_firstauto = IPPORT_RESERVED;         /* 1024 */
146 int ipport_lastauto = IPPORT_USERRESERVED;      /* 5000 */
147
148 int ipport_hifirstauto = IPPORT_HIFIRSTAUTO;    /* 49152 */
149 int ipport_hilastauto = IPPORT_HILASTAUTO;      /* 65535 */
150
151 static __inline void
152 RANGECHK(int var, int min, int max)
153 {
154         if (var < min)
155                 var = min;
156         else if (var > max)
157                 var = max;
158 }
159
160 static int
161 sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS)
162 {
163         int error;
164
165         error = sysctl_handle_int(oidp, oidp->oid_arg1, oidp->oid_arg2, req);
166         if (!error) {
167                 RANGECHK(ipport_lowfirstauto, 1, IPPORT_RESERVED - 1);
168                 RANGECHK(ipport_lowlastauto, 1, IPPORT_RESERVED - 1);
169
170                 RANGECHK(ipport_firstauto, IPPORT_RESERVED, USHRT_MAX);
171                 RANGECHK(ipport_lastauto, IPPORT_RESERVED, USHRT_MAX);
172
173                 RANGECHK(ipport_hifirstauto, IPPORT_RESERVED, USHRT_MAX);
174                 RANGECHK(ipport_hilastauto, IPPORT_RESERVED, USHRT_MAX);
175         }
176         return (error);
177 }
178
179 SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW, 0, "IP Ports");
180
181 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLTYPE_INT|CTLFLAG_RW,
182            &ipport_lowfirstauto, 0, &sysctl_net_ipport_check, "I", "");
183 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLTYPE_INT|CTLFLAG_RW,
184            &ipport_lowlastauto, 0, &sysctl_net_ipport_check, "I", "");
185 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLTYPE_INT|CTLFLAG_RW,
186            &ipport_firstauto, 0, &sysctl_net_ipport_check, "I", "");
187 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLTYPE_INT|CTLFLAG_RW,
188            &ipport_lastauto, 0, &sysctl_net_ipport_check, "I", "");
189 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLTYPE_INT|CTLFLAG_RW,
190            &ipport_hifirstauto, 0, &sysctl_net_ipport_check, "I", "");
191 SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLTYPE_INT|CTLFLAG_RW,
192            &ipport_hilastauto, 0, &sysctl_net_ipport_check, "I", "");
193
194 /*
195  * in_pcb.c: manage the Protocol Control Blocks.
196  *
197  * NOTE: It is assumed that most of these functions will be called at
198  * splnet(). XXX - There are, unfortunately, a few exceptions to this
199  * rule that should be fixed.
200  *
201  * NOTE: The caller should initialize the cpu field to the cpu running the
202  * protocol stack associated with this inpcbinfo.
203  */
204
205 void
206 in_pcbinfo_init(struct inpcbinfo *pcbinfo)
207 {
208         LIST_INIT(&pcbinfo->pcblisthead);
209         pcbinfo->cpu = -1;
210 }
211
212 /*
213  * Allocate a PCB and associate it with the socket.
214  */
215 int
216 in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo)
217 {
218         struct inpcb *inp;
219 #ifdef IPSEC
220         int error;
221 #endif
222
223         inp = zalloc(pcbinfo->ipi_zone);
224         if (inp == NULL)
225                 return (ENOBUFS);
226         bzero(inp, sizeof *inp);
227         inp->inp_gencnt = ++pcbinfo->ipi_gencnt;
228         inp->inp_pcbinfo = inp->inp_cpcbinfo = pcbinfo;
229         inp->inp_socket = so;
230 #ifdef IPSEC
231         error = ipsec_init_policy(so, &inp->inp_sp);
232         if (error != 0) {
233                 zfree(pcbinfo->ipi_zone, inp);
234                 return (error);
235         }
236 #endif
237 #ifdef INET6
238         if (INP_SOCKAF(so) == AF_INET6 && ip6_v6only)
239                 inp->inp_flags |= IN6P_IPV6_V6ONLY;
240         if (ip6_auto_flowlabel)
241                 inp->inp_flags |= IN6P_AUTOFLOWLABEL;
242 #endif
243         so->so_pcb = inp;
244         LIST_INSERT_HEAD(&pcbinfo->pcblisthead, inp, inp_list);
245         pcbinfo->ipi_count++;
246         return (0);
247 }
248
249 int
250 in_pcbbind(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
251 {
252         struct socket *so = inp->inp_socket;
253         struct proc *p = td->td_proc;
254         unsigned short *lastport;
255         struct sockaddr_in *sin;
256         struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
257         u_short lport = 0;
258         int wild = 0, reuseport = (so->so_options & SO_REUSEPORT);
259         int error, prison = 0;
260
261         KKASSERT(p);
262
263         if (TAILQ_EMPTY(&in_ifaddrhead)) /* XXX broken! */
264                 return (EADDRNOTAVAIL);
265         if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY)
266                 return (EINVAL);        /* already bound */
267         if (!(so->so_options & (SO_REUSEADDR|SO_REUSEPORT)))
268                 wild = 1;    /* neither SO_REUSEADDR nor SO_REUSEPORT is set */
269         if (nam != NULL) {
270                 sin = (struct sockaddr_in *)nam;
271                 if (nam->sa_len != sizeof *sin)
272                         return (EINVAL);
273 #ifdef notdef
274                 /*
275                  * We should check the family, but old programs
276                  * incorrectly fail to initialize it.
277                  */
278                 if (sin->sin_family != AF_INET)
279                         return (EAFNOSUPPORT);
280 #endif
281                 if (sin->sin_addr.s_addr != INADDR_ANY &&
282                     prison_ip(td, 0, &sin->sin_addr.s_addr))
283                                 return (EINVAL);
284                 lport = sin->sin_port;
285                 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
286                         /*
287                          * Treat SO_REUSEADDR as SO_REUSEPORT for multicast;
288                          * allow complete duplication of binding if
289                          * SO_REUSEPORT is set, or if SO_REUSEADDR is set
290                          * and a multicast address is bound on both
291                          * new and duplicated sockets.
292                          */
293                         if (so->so_options & SO_REUSEADDR)
294                                 reuseport = SO_REUSEADDR | SO_REUSEPORT;
295                 } else if (sin->sin_addr.s_addr != INADDR_ANY) {
296                         sin->sin_port = 0;              /* yech... */
297                         bzero(&sin->sin_zero, sizeof sin->sin_zero);
298                         if (ifa_ifwithaddr((struct sockaddr *)sin) == NULL)
299                                 return (EADDRNOTAVAIL);
300                 }
301                 if (lport != 0) {
302                         struct inpcb *t;
303
304                         /* GROSS */
305                         if (ntohs(lport) < IPPORT_RESERVED &&
306                             p && suser_cred(p->p_ucred, PRISON_ROOT))
307                                 return (EACCES);
308                         if (p && p->p_ucred->cr_prison)
309                                 prison = 1;
310                         if (so->so_cred->cr_uid != 0 &&
311                             !IN_MULTICAST(ntohl(sin->sin_addr.s_addr))) {
312                                 t = in_pcblookup_local(inp->inp_pcbinfo,
313                                     sin->sin_addr, lport,
314                                     prison ? 0 : INPLOOKUP_WILDCARD);
315                                 if (t &&
316                                     (!in_nullhost(sin->sin_addr) ||
317                                      !in_nullhost(t->inp_laddr) ||
318                                      (t->inp_socket->so_options &
319                                          SO_REUSEPORT) == 0) &&
320                                     (so->so_cred->cr_uid !=
321                                      t->inp_socket->so_cred->cr_uid)) {
322 #ifdef INET6
323                                         if (!in_nullhost(sin->sin_addr) ||
324                                             !in_nullhost(t->inp_laddr) ||
325                                             INP_SOCKAF(so) ==
326                                             INP_SOCKAF(t->inp_socket))
327 #endif
328                                         return (EADDRINUSE);
329                                 }
330                         }
331                         if (prison && prison_ip(td, 0, &sin->sin_addr.s_addr))
332                                 return (EADDRNOTAVAIL);
333                         t = in_pcblookup_local(pcbinfo, sin->sin_addr,
334                             lport, prison ? 0 : wild);
335                         if (t && !(reuseport & t->inp_socket->so_options)) {
336 #ifdef INET6
337                                 if (!in_nullhost(sin->sin_addr) ||
338                                     !in_nullhost(t->inp_laddr) ||
339                                     INP_SOCKAF(so) == INP_SOCKAF(t->inp_socket))
340 #endif
341                                 return (EADDRINUSE);
342                         }
343                 }
344                 inp->inp_laddr = sin->sin_addr;
345         }
346         if (lport == 0) {
347                 ushort first, last;
348                 int count;
349
350                 if (inp->inp_laddr.s_addr != INADDR_ANY &&
351                     prison_ip(td, 0, &inp->inp_laddr.s_addr )) {
352                         inp->inp_laddr.s_addr = INADDR_ANY;
353                         return (EINVAL);
354                 }
355                 inp->inp_flags |= INP_ANONPORT;
356
357                 if (inp->inp_flags & INP_HIGHPORT) {
358                         first = ipport_hifirstauto;     /* sysctl */
359                         last  = ipport_hilastauto;
360                         lastport = &pcbinfo->lasthi;
361                 } else if (inp->inp_flags & INP_LOWPORT) {
362                         if (p &&
363                             (error = suser_cred(p->p_ucred, PRISON_ROOT))) {
364                                 inp->inp_laddr.s_addr = INADDR_ANY;
365                                 return (error);
366                         }
367                         first = ipport_lowfirstauto;    /* 1023 */
368                         last  = ipport_lowlastauto;     /* 600 */
369                         lastport = &pcbinfo->lastlow;
370                 } else {
371                         first = ipport_firstauto;       /* sysctl */
372                         last  = ipport_lastauto;
373                         lastport = &pcbinfo->lastport;
374                 }
375                 /*
376                  * Simple check to ensure all ports are not used up causing
377                  * a deadlock here.
378                  *
379                  * We split the two cases (up and down) so that the direction
380                  * is not being tested on each round of the loop.
381                  */
382                 if (first > last) {
383                         /*
384                          * counting down
385                          */
386                         count = first - last;
387
388                         do {
389                                 if (count-- < 0) {      /* completely used? */
390                                         inp->inp_laddr.s_addr = INADDR_ANY;
391                                         return (EADDRNOTAVAIL);
392                                 }
393                                 --*lastport;
394                                 if (*lastport > first || *lastport < last)
395                                         *lastport = first;
396                                 lport = htons(*lastport);
397                         } while (in_pcblookup_local(pcbinfo,
398                                  inp->inp_laddr, lport, wild));
399                 } else {
400                         /*
401                          * counting up
402                          */
403                         count = last - first;
404
405                         do {
406                                 if (count-- < 0) {      /* completely used? */
407                                         inp->inp_laddr.s_addr = INADDR_ANY;
408                                         return (EADDRNOTAVAIL);
409                                 }
410                                 ++*lastport;
411                                 if (*lastport < first || *lastport > last)
412                                         *lastport = first;
413                                 lport = htons(*lastport);
414                         } while (in_pcblookup_local(pcbinfo,
415                                  inp->inp_laddr, lport, wild));
416                 }
417         }
418         inp->inp_lport = lport;
419         if (prison_ip(td, 0, &inp->inp_laddr.s_addr)) {
420                 inp->inp_laddr.s_addr = INADDR_ANY;
421                 inp->inp_lport = 0;
422                 return (EINVAL);
423         }
424         if (in_pcbinsporthash(inp) != 0) {
425                 inp->inp_laddr.s_addr = INADDR_ANY;
426                 inp->inp_lport = 0;
427                 return (EAGAIN);
428         }
429         return (0);
430 }
431
432 /*
433  *   Transform old in_pcbconnect() into an inner subroutine for new
434  *   in_pcbconnect(): Do some validity-checking on the remote
435  *   address (in mbuf 'nam') and then determine local host address
436  *   (i.e., which interface) to use to access that remote host.
437  *
438  *   This preserves definition of in_pcbconnect(), while supporting a
439  *   slightly different version for T/TCP.  (This is more than
440  *   a bit of a kludge, but cleaning up the internal interfaces would
441  *   have forced minor changes in every protocol).
442  */
443 int
444 in_pcbladdr(inp, nam, plocal_sin)
445         struct inpcb *inp;
446         struct sockaddr *nam;
447         struct sockaddr_in **plocal_sin;
448 {
449         struct in_ifaddr *ia;
450         struct sockaddr_in *sin = (struct sockaddr_in *)nam;
451
452         if (nam->sa_len != sizeof *sin)
453                 return (EINVAL);
454         if (sin->sin_family != AF_INET)
455                 return (EAFNOSUPPORT);
456         if (sin->sin_port == 0)
457                 return (EADDRNOTAVAIL);
458         if (!TAILQ_EMPTY(&in_ifaddrhead)) {
459                 ia = TAILQ_FIRST(&in_ifaddrhead);
460                 /*
461                  * If the destination address is INADDR_ANY,
462                  * use the primary local address.
463                  * If the supplied address is INADDR_BROADCAST,
464                  * and the primary interface supports broadcast,
465                  * choose the broadcast address for that interface.
466                  */
467                 if (sin->sin_addr.s_addr == INADDR_ANY)
468                         sin->sin_addr = IA_SIN(ia)->sin_addr;
469                 else if (sin->sin_addr.s_addr == (u_long)INADDR_BROADCAST &&
470                     (ia->ia_ifp->if_flags & IFF_BROADCAST))
471                         sin->sin_addr = satosin(&ia->ia_broadaddr)->sin_addr;
472         }
473         if (inp->inp_laddr.s_addr == INADDR_ANY) {
474                 struct route *ro;
475
476                 ia = (struct in_ifaddr *)NULL;
477                 /*
478                  * If route is known or can be allocated now,
479                  * our src addr is taken from the i/f, else punt.
480                  * Note that we should check the address family of the cached
481                  * destination, in case of sharing the cache with IPv6.
482                  */
483                 ro = &inp->inp_route;
484                 if (ro->ro_rt &&
485                     (!(ro->ro_rt->rt_flags & RTF_UP) ||
486                      ro->ro_dst.sa_family != AF_INET ||
487                      satosin(&ro->ro_dst)->sin_addr.s_addr !=
488                                       sin->sin_addr.s_addr ||
489                      inp->inp_socket->so_options & SO_DONTROUTE)) {
490                         RTFREE(ro->ro_rt);
491                         ro->ro_rt = (struct rtentry *)NULL;
492                 }
493                 if (!(inp->inp_socket->so_options & SO_DONTROUTE) && /*XXX*/
494                     (ro->ro_rt == (struct rtentry *)NULL ||
495                     ro->ro_rt->rt_ifp == (struct ifnet *)NULL)) {
496                         /* No route yet, so try to acquire one */
497                         bzero(&ro->ro_dst, sizeof(struct sockaddr_in));
498                         ro->ro_dst.sa_family = AF_INET;
499                         ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
500                         ((struct sockaddr_in *) &ro->ro_dst)->sin_addr =
501                                 sin->sin_addr;
502                         rtalloc(ro);
503                 }
504                 /*
505                  * If we found a route, use the address
506                  * corresponding to the outgoing interface
507                  * unless it is the loopback (in case a route
508                  * to our address on another net goes to loopback).
509                  */
510                 if (ro->ro_rt && !(ro->ro_rt->rt_ifp->if_flags & IFF_LOOPBACK))
511                         ia = ifatoia(ro->ro_rt->rt_ifa);
512                 if (ia == NULL) {
513                         u_short fport = sin->sin_port;
514
515                         sin->sin_port = 0;
516                         ia = ifatoia(ifa_ifwithdstaddr(sintosa(sin)));
517                         if (ia == NULL)
518                                 ia = ifatoia(ifa_ifwithnet(sintosa(sin)));
519                         sin->sin_port = fport;
520                         if (ia == NULL)
521                                 ia = TAILQ_FIRST(&in_ifaddrhead);
522                         if (ia == NULL)
523                                 return (EADDRNOTAVAIL);
524                 }
525                 /*
526                  * If the destination address is multicast and an outgoing
527                  * interface has been set as a multicast option, use the
528                  * address of that interface as our source address.
529                  */
530                 if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr)) &&
531                     inp->inp_moptions != NULL) {
532                         struct ip_moptions *imo;
533                         struct ifnet *ifp;
534
535                         imo = inp->inp_moptions;
536                         if (imo->imo_multicast_ifp != NULL) {
537                                 ifp = imo->imo_multicast_ifp;
538                                 TAILQ_FOREACH(ia, &in_ifaddrhead, ia_link)
539                                         if (ia->ia_ifp == ifp)
540                                                 break;
541                                 if (ia == NULL)
542                                         return (EADDRNOTAVAIL);
543                         }
544                 }
545                 /*
546                  * Don't do pcblookup call here; return interface in plocal_sin
547                  * and exit to caller, that will do the lookup.
548                  */
549                 *plocal_sin = &ia->ia_addr;
550
551         }
552         return (0);
553 }
554
555 /*
556  * Outer subroutine:
557  * Connect from a socket to a specified address.
558  * Both address and port must be specified in argument sin.
559  * If don't have a local address for this socket yet,
560  * then pick one.
561  */
562 int
563 in_pcbconnect(struct inpcb *inp, struct sockaddr *nam, struct thread *td)
564 {
565         struct sockaddr_in *if_sin;
566         struct sockaddr_in *sin = (struct sockaddr_in *)nam;
567         struct sockaddr_in sa;
568         struct ucred *cr = td->td_proc ? td->td_proc->p_ucred : NULL;
569         int error;
570
571         if (cr && cr->cr_prison != NULL && in_nullhost(inp->inp_laddr)) {
572                 bzero(&sa, sizeof sa);
573                 sa.sin_addr.s_addr = htonl(cr->cr_prison->pr_ip);
574                 sa.sin_len = sizeof sa;
575                 sa.sin_family = AF_INET;
576                 error = in_pcbbind(inp, (struct sockaddr *)&sa, td);
577                 if (error)
578                         return (error);
579         }
580
581         /* Call inner routine to assign local interface address. */
582         if ((error = in_pcbladdr(inp, nam, &if_sin)) != 0)
583                 return (error);
584
585         if (in_pcblookup_hash(inp->inp_cpcbinfo, sin->sin_addr, sin->sin_port,
586             inp->inp_laddr.s_addr ? inp->inp_laddr : if_sin->sin_addr,
587             inp->inp_lport, FALSE, NULL) != NULL) {
588                 return (EADDRINUSE);
589         }
590         if (inp->inp_laddr.s_addr == INADDR_ANY) {
591                 if (inp->inp_lport == 0) {
592                         error = in_pcbbind(inp, (struct sockaddr *)NULL, td);
593                         if (error)
594                                 return (error);
595                 }
596                 inp->inp_laddr = if_sin->sin_addr;
597         }
598         inp->inp_faddr = sin->sin_addr;
599         inp->inp_fport = sin->sin_port;
600         in_pcbinsconnhash(inp);
601         return (0);
602 }
603
604 void
605 in_pcbdisconnect(inp)
606         struct inpcb *inp;
607 {
608
609         inp->inp_faddr.s_addr = INADDR_ANY;
610         inp->inp_fport = 0;
611         in_pcbremconnhash(inp);
612         if (inp->inp_socket->so_state & SS_NOFDREF)
613                 in_pcbdetach(inp);
614 }
615
616 void
617 in_pcbdetach(inp)
618         struct inpcb *inp;
619 {
620         struct socket *so = inp->inp_socket;
621         struct inpcbinfo *ipi = inp->inp_pcbinfo;
622
623 #ifdef IPSEC
624         ipsec4_delete_pcbpolicy(inp);
625 #endif /*IPSEC*/
626         inp->inp_gencnt = ++ipi->ipi_gencnt;
627         in_pcbremlists(inp);
628         so->so_pcb = 0;
629         sofree(so);
630         if (inp->inp_options)
631                 m_free(inp->inp_options);
632         if (inp->inp_route.ro_rt)
633                 rtfree(inp->inp_route.ro_rt);
634         ip_freemoptions(inp->inp_moptions);
635         inp->inp_vflag = 0;
636         zfree(ipi->ipi_zone, inp);
637 }
638
639 /*
640  * The calling convention of in_setsockaddr() and in_setpeeraddr() was
641  * modified to match the pru_sockaddr() and pru_peeraddr() entry points
642  * in struct pr_usrreqs, so that protocols can just reference then directly
643  * without the need for a wrapper function.  The socket must have a valid
644  * (i.e., non-nil) PCB, but it should be impossible to get an invalid one
645  * except through a kernel programming error, so it is acceptable to panic
646  * (or in this case trap) if the PCB is invalid.  (Actually, we don't trap
647  * because there actually /is/ a programming error somewhere... XXX)
648  */
649 int
650 in_setsockaddr(so, nam)
651         struct socket *so;
652         struct sockaddr **nam;
653 {
654         int s;
655         struct inpcb *inp;
656         struct sockaddr_in *sin;
657
658         /*
659          * Do the malloc first in case it blocks.
660          */
661         MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
662                 M_WAITOK | M_ZERO);
663         sin->sin_family = AF_INET;
664         sin->sin_len = sizeof *sin;
665
666         s = splnet();
667         inp = sotoinpcb(so);
668         if (!inp) {
669                 splx(s);
670                 free(sin, M_SONAME);
671                 return (ECONNRESET);
672         }
673         sin->sin_port = inp->inp_lport;
674         sin->sin_addr = inp->inp_laddr;
675         splx(s);
676
677         *nam = (struct sockaddr *)sin;
678         return (0);
679 }
680
681 int
682 in_setpeeraddr(so, nam)
683         struct socket *so;
684         struct sockaddr **nam;
685 {
686         int s;
687         struct inpcb *inp;
688         struct sockaddr_in *sin;
689
690         /*
691          * Do the malloc first in case it blocks.
692          */
693         MALLOC(sin, struct sockaddr_in *, sizeof *sin, M_SONAME,
694                 M_WAITOK | M_ZERO);
695         sin->sin_family = AF_INET;
696         sin->sin_len = sizeof *sin;
697
698         s = splnet();
699         inp = sotoinpcb(so);
700         if (!inp) {
701                 splx(s);
702                 free(sin, M_SONAME);
703                 return (ECONNRESET);
704         }
705         sin->sin_port = inp->inp_fport;
706         sin->sin_addr = inp->inp_faddr;
707         splx(s);
708
709         *nam = (struct sockaddr *)sin;
710         return (0);
711 }
712
713 void
714 in_pcbnotifyall(head, faddr, errno, notify)
715         struct inpcbhead *head;
716         struct in_addr faddr;
717         void (*notify) (struct inpcb *, int);
718 {
719         struct inpcb *inp, *ninp;
720         int s;
721
722         /*
723          * note: if INP_PLACEMARKER is set we must ignore the rest of
724          * the structure and skip it.
725          */
726         s = splnet();
727         for (inp = LIST_FIRST(head); inp != NULL; inp = ninp) {
728                 ninp = LIST_NEXT(inp, inp_list);
729                 if (inp->inp_flags & INP_PLACEMARKER)
730                         continue;
731 #ifdef INET6
732                 if (!(inp->inp_vflag & INP_IPV4))
733                         continue;
734 #endif
735                 if (inp->inp_faddr.s_addr != faddr.s_addr ||
736                     inp->inp_socket == NULL)
737                         continue;
738                 (*notify)(inp, errno);
739         }
740         splx(s);
741 }
742
743 void
744 in_pcbpurgeif0(head, ifp)
745         struct inpcb *head;
746         struct ifnet *ifp;
747 {
748         struct inpcb *inp;
749         struct ip_moptions *imo;
750         int i, gap;
751
752         for (inp = head; inp != NULL; inp = LIST_NEXT(inp, inp_list)) {
753                 if (inp->inp_flags & INP_PLACEMARKER)
754                         continue;
755                 imo = inp->inp_moptions;
756                 if ((inp->inp_vflag & INP_IPV4) && imo != NULL) {
757                         /*
758                          * Unselect the outgoing interface if it is being
759                          * detached.
760                          */
761                         if (imo->imo_multicast_ifp == ifp)
762                                 imo->imo_multicast_ifp = NULL;
763
764                         /*
765                          * Drop multicast group membership if we joined
766                          * through the interface being detached.
767                          */
768                         for (i = 0, gap = 0; i < imo->imo_num_memberships;
769                             i++) {
770                                 if (imo->imo_membership[i]->inm_ifp == ifp) {
771                                         in_delmulti(imo->imo_membership[i]);
772                                         gap++;
773                                 } else if (gap != 0)
774                                         imo->imo_membership[i - gap] =
775                                             imo->imo_membership[i];
776                         }
777                         imo->imo_num_memberships -= gap;
778                 }
779         }
780 }
781
782 /*
783  * Check for alternatives when higher level complains
784  * about service problems.  For now, invalidate cached
785  * routing information.  If the route was created dynamically
786  * (by a redirect), time to try a default gateway again.
787  */
788 void
789 in_losing(inp)
790         struct inpcb *inp;
791 {
792         struct rtentry *rt;
793         struct rt_addrinfo info;
794
795         if ((rt = inp->inp_route.ro_rt)) {
796                 bzero(&info, sizeof info);
797                 info.rti_flags = rt->rt_flags;
798                 info.rti_info[RTAX_DST] = rt_key(rt);
799                 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway;
800                 info.rti_info[RTAX_NETMASK] = rt_mask(rt);
801                 rt_missmsg(RTM_LOSING, &info, rt->rt_flags, 0);
802                 if (rt->rt_flags & RTF_DYNAMIC)
803                         rtrequest1(RTM_DELETE, &info, NULL);
804                 inp->inp_route.ro_rt = NULL;
805                 rtfree(rt);
806                 /*
807                  * A new route can be allocated
808                  * the next time output is attempted.
809                  */
810         }
811 }
812
813 /*
814  * After a routing change, flush old routing
815  * and allocate a (hopefully) better one.
816  */
817 void
818 in_rtchange(inp, errno)
819         struct inpcb *inp;
820         int errno;
821 {
822         if (inp->inp_route.ro_rt) {
823                 rtfree(inp->inp_route.ro_rt);
824                 inp->inp_route.ro_rt = 0;
825                 /*
826                  * A new route can be allocated the next time
827                  * output is attempted.
828                  */
829         }
830 }
831
832 /*
833  * Lookup a PCB based on the local address and port.
834  */
835 struct inpcb *
836 in_pcblookup_local(pcbinfo, laddr, lport_arg, wild_okay)
837         struct inpcbinfo *pcbinfo;
838         struct in_addr laddr;
839         u_int lport_arg;
840         int wild_okay;
841 {
842         struct inpcb *inp;
843         int matchwild = 3, wildcard;
844         u_short lport = lport_arg;
845
846         struct inpcbporthead *porthash;
847         struct inpcbport *phd;
848         struct inpcb *match = NULL;
849
850         /*
851          * Best fit PCB lookup.
852          *
853          * First see if this local port is in use by looking on the
854          * port hash list.
855          */
856         porthash = &pcbinfo->porthashbase[INP_PCBPORTHASH(lport,
857             pcbinfo->porthashmask)];
858         LIST_FOREACH(phd, porthash, phd_hash) {
859                 if (phd->phd_port == lport)
860                         break;
861         }
862         if (phd != NULL) {
863                 /*
864                  * Port is in use by one or more PCBs. Look for best
865                  * fit.
866                  */
867                 LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) {
868                         wildcard = 0;
869 #ifdef INET6
870                         if ((inp->inp_vflag & INP_IPV4) == 0)
871                                 continue;
872 #endif
873                         if (inp->inp_faddr.s_addr != INADDR_ANY)
874                                 wildcard++;
875                         if (inp->inp_laddr.s_addr != INADDR_ANY) {
876                                 if (laddr.s_addr == INADDR_ANY)
877                                         wildcard++;
878                                 else if (inp->inp_laddr.s_addr != laddr.s_addr)
879                                         continue;
880                         } else {
881                                 if (laddr.s_addr != INADDR_ANY)
882                                         wildcard++;
883                         }
884                         if (wildcard && !wild_okay)
885                                 continue;
886                         if (wildcard < matchwild) {
887                                 match = inp;
888                                 matchwild = wildcard;
889                                 if (matchwild == 0) {
890                                         break;
891                                 }
892                         }
893                 }
894         }
895         return (match);
896 }
897
898 /*
899  * Lookup PCB in hash list.
900  */
901 struct inpcb *
902 in_pcblookup_hash(pcbinfo, faddr, fport_arg, laddr, lport_arg, wildcard, ifp)
903         struct inpcbinfo *pcbinfo;
904         struct in_addr faddr, laddr;
905         u_int fport_arg, lport_arg;
906         boolean_t wildcard;
907         struct ifnet *ifp;
908 {
909         struct inpcbhead *head;
910         struct inpcb *inp;
911         u_short fport = fport_arg, lport = lport_arg;
912
913         /*
914          * First look for an exact match.
915          */
916         head = &pcbinfo->hashbase[INP_PCBCONNHASH(faddr.s_addr, fport,
917             laddr.s_addr, lport, pcbinfo->hashmask)];
918         LIST_FOREACH(inp, head, inp_hash) {
919 #ifdef INET6
920                 if (!(inp->inp_vflag & INP_IPV4))
921                         continue;
922 #endif
923                 if (in_hosteq(inp->inp_faddr, faddr) &&
924                     in_hosteq(inp->inp_laddr, laddr) &&
925                     inp->inp_fport == fport && inp->inp_lport == lport) {
926                         /* found */
927                         return (inp);
928                 }
929         }
930
931         if (wildcard) {
932                 struct inpcb *local_wild = NULL;
933 #ifdef INET6
934                 struct inpcb *local_wild_mapped = NULL;
935 #endif
936                 struct inpcontainer *ic;
937                 struct inpcontainerhead *chead;
938
939                 chead = &pcbinfo->wildcardhashbase[
940                     INP_PCBWILDCARDHASH(lport, pcbinfo->wildcardhashmask)];
941                 LIST_FOREACH(ic, chead, ic_list) {
942                         inp = ic->ic_inp;
943 #ifdef INET6
944                         if (!(inp->inp_vflag & INP_IPV4))
945                                 continue;
946 #endif
947                         if (inp->inp_lport == lport) {
948                                 if (ifp && ifp->if_type == IFT_FAITH &&
949                                     !(inp->inp_flags & INP_FAITH))
950                                         continue;
951                                 if (inp->inp_laddr.s_addr == laddr.s_addr)
952                                         return (inp);
953                                 if (inp->inp_laddr.s_addr == INADDR_ANY) {
954 #ifdef INET6
955                                         if (INP_CHECK_SOCKAF(inp->inp_socket,
956                                                              AF_INET6))
957                                                 local_wild_mapped = inp;
958                                         else
959 #endif
960                                                 local_wild = inp;
961                                 }
962                         }
963                 }
964 #ifdef INET6
965                 if (local_wild == NULL)
966                         return (local_wild_mapped);
967 #endif
968                 return (local_wild);
969         }
970
971         /*
972          * Not found.
973          */
974         return (NULL);
975 }
976
977 /*
978  * Insert PCB into connection hash table.
979  */
980 void
981 in_pcbinsconnhash(struct inpcb *inp)
982 {
983         struct inpcbinfo *pcbinfo = inp->inp_cpcbinfo;
984         struct inpcbhead *bucket;
985         u_int32_t hashkey_faddr, hashkey_laddr;
986
987 #ifdef INET6
988         if (inp->inp_vflag & INP_IPV6) {
989                 hashkey_faddr = inp->in6p_faddr.s6_addr32[3] /* XXX JH */;
990                 hashkey_laddr = inp->in6p_laddr.s6_addr32[3] /* XXX JH */;
991         } else {
992 #endif
993                 hashkey_faddr = inp->inp_faddr.s_addr;
994                 hashkey_laddr = inp->inp_laddr.s_addr;
995 #ifdef INET6
996         }
997 #endif
998
999         KASSERT(!(inp->inp_flags & INP_CONNECTED), ("already on hash list"));
1000         inp->inp_flags |= INP_CONNECTED;
1001
1002         /*
1003          * Insert into the connection hash table.
1004          */
1005         bucket = &pcbinfo->hashbase[INP_PCBCONNHASH(hashkey_faddr,
1006             inp->inp_fport, hashkey_laddr, inp->inp_lport, pcbinfo->hashmask)];
1007         LIST_INSERT_HEAD(bucket, inp, inp_hash);
1008 }
1009
1010 /*
1011  * Remove PCB from connection hash table.
1012  */
1013 void
1014 in_pcbremconnhash(struct inpcb *inp)
1015 {
1016         KASSERT(inp->inp_flags & INP_CONNECTED, ("inp not connected"));
1017         LIST_REMOVE(inp, inp_hash);
1018         inp->inp_flags &= ~INP_CONNECTED;
1019 }
1020
1021 /*
1022  * Insert PCB into port hash table.
1023  */
1024 int
1025 in_pcbinsporthash(struct inpcb *inp)
1026 {
1027         struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1028         struct inpcbporthead *pcbporthash;
1029         struct inpcbport *phd;
1030
1031         /*
1032          * Insert into the port hash table.
1033          */
1034         pcbporthash = &pcbinfo->porthashbase[
1035             INP_PCBPORTHASH(inp->inp_lport, pcbinfo->porthashmask)];
1036
1037         /* Go through port list and look for a head for this lport. */
1038         LIST_FOREACH(phd, pcbporthash, phd_hash)
1039                 if (phd->phd_port == inp->inp_lport)
1040                         break;
1041
1042         /* If none exists, malloc one and tack it on. */
1043         if (phd == NULL) {
1044                 MALLOC(phd, struct inpcbport *, sizeof(struct inpcbport),
1045                     M_PCB, M_INTWAIT | M_NULLOK);
1046                 if (phd == NULL)
1047                         return (ENOBUFS); /* XXX */
1048                 phd->phd_port = inp->inp_lport;
1049                 LIST_INIT(&phd->phd_pcblist);
1050                 LIST_INSERT_HEAD(pcbporthash, phd, phd_hash);
1051         }
1052
1053         inp->inp_phd = phd;
1054         LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist);
1055
1056         return (0);
1057 }
1058
1059 void
1060 in_pcbinswildcardhash_oncpu(struct inpcb *inp, struct inpcbinfo *pcbinfo)
1061 {
1062         struct inpcontainer *ic;
1063         struct inpcontainerhead *bucket;
1064
1065         bucket = &pcbinfo->wildcardhashbase[
1066             INP_PCBWILDCARDHASH(inp->inp_lport, pcbinfo->wildcardhashmask)];
1067
1068         ic = malloc(sizeof(struct inpcontainer), M_TEMP, M_INTWAIT);
1069         ic->ic_inp = inp;
1070         LIST_INSERT_HEAD(bucket, ic, ic_list);
1071 }
1072
1073 /*
1074  * Insert PCB into wildcard hash table.
1075  */
1076 void
1077 in_pcbinswildcardhash(struct inpcb *inp)
1078 {
1079         struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1080         
1081         KKASSERT(pcbinfo != NULL);
1082
1083         in_pcbinswildcardhash_oncpu(inp, pcbinfo);
1084         inp->inp_flags |= INP_WILDCARD;
1085 }
1086
1087 void
1088 in_pcbremwildcardhash_oncpu(struct inpcb *inp, struct inpcbinfo *pcbinfo)
1089 {
1090         struct inpcontainer *ic;
1091         struct inpcontainerhead *head;
1092
1093         /* find bucket */
1094         head = &pcbinfo->wildcardhashbase[
1095             INP_PCBWILDCARDHASH(inp->inp_lport, pcbinfo->wildcardhashmask)];
1096
1097         LIST_FOREACH(ic, head, ic_list) {
1098                 if (ic->ic_inp == inp)
1099                         goto found;
1100         }
1101         return;                 /* not found! */
1102
1103 found:
1104         LIST_REMOVE(ic, ic_list);       /* remove container from bucket chain */
1105         free(ic, M_TEMP);               /* deallocate container */
1106 }
1107
1108 /*
1109  * Remove PCB from wildcard hash table.
1110  */
1111 void
1112 in_pcbremwildcardhash(struct inpcb *inp)
1113 {
1114         struct inpcbinfo *pcbinfo = inp->inp_pcbinfo;
1115
1116         KASSERT(inp->inp_flags & INP_WILDCARD, ("inp not wildcard"));
1117         in_pcbremwildcardhash_oncpu(inp, pcbinfo);
1118         inp->inp_flags &= ~INP_WILDCARD;
1119 }
1120
1121 /*
1122  * Remove PCB from various lists.
1123  */
1124 void
1125 in_pcbremlists(inp)
1126         struct inpcb *inp;
1127 {
1128         if (inp->inp_lport) {
1129                 struct inpcbport *phd = inp->inp_phd;
1130
1131                 LIST_REMOVE(inp, inp_portlist);
1132                 if (LIST_FIRST(&phd->phd_pcblist) == NULL) {
1133                         LIST_REMOVE(phd, phd_hash);
1134                         free(phd, M_PCB);
1135                 }
1136         }
1137         if (inp->inp_flags & INP_WILDCARD) {
1138                 in_pcbremwildcardhash(inp);
1139         } else if (inp->inp_flags & INP_CONNECTED) {
1140                 in_pcbremconnhash(inp);
1141         }
1142         LIST_REMOVE(inp, inp_list);
1143         inp->inp_pcbinfo->ipi_count--;
1144 }
1145
1146 int
1147 prison_xinpcb(struct thread *td, struct inpcb *inp)
1148 {
1149         struct ucred *cr;
1150
1151         if (td->td_proc == NULL)
1152                 return (0);
1153         cr = td->td_proc->p_ucred;
1154         if (cr->cr_prison == NULL)
1155                 return (0);
1156         if (ntohl(inp->inp_laddr.s_addr) == cr->cr_prison->pr_ip)
1157                 return (0);
1158         return (1);
1159 }
1160
1161 int
1162 in_pcblist_global(SYSCTL_HANDLER_ARGS)
1163 {
1164         struct inpcbinfo *pcbinfo = arg1;
1165         struct inpcb *inp, *marker;
1166         struct xinpcb xi;
1167         int error, i, n;
1168         inp_gen_t gencnt;
1169
1170         /*
1171          * The process of preparing the TCB list is too time-consuming and
1172          * resource-intensive to repeat twice on every request.
1173          */
1174         if (req->oldptr == NULL) {
1175                 n = pcbinfo->ipi_count;
1176                 req->oldidx = (n + n/8 + 10) * sizeof(struct xinpcb);
1177                 return 0;
1178         }
1179
1180         if (req->newptr != NULL)
1181                 return EPERM;
1182
1183         /*
1184          * OK, now we're committed to doing something.  Re-fetch ipi_count
1185          * after obtaining the generation count.
1186          */
1187         gencnt = pcbinfo->ipi_gencnt;
1188         n = pcbinfo->ipi_count;
1189
1190         marker = malloc(sizeof(struct inpcb), M_TEMP, M_WAITOK|M_ZERO);
1191         marker->inp_flags |= INP_PLACEMARKER;
1192         LIST_INSERT_HEAD(&pcbinfo->pcblisthead, marker, inp_list);
1193
1194         i = 0;
1195         error = 0;
1196
1197         while ((inp = LIST_NEXT(marker, inp_list)) != NULL && i < n) {
1198                 LIST_REMOVE(marker, inp_list);
1199                 LIST_INSERT_AFTER(inp, marker, inp_list);
1200
1201                 if (inp->inp_flags & INP_PLACEMARKER)
1202                         continue;
1203                 if (inp->inp_gencnt > gencnt)
1204                         continue;
1205                 if (prison_xinpcb(req->td, inp))
1206                         continue;
1207                 xi.xi_len = sizeof xi;
1208                 bcopy(inp, &xi.xi_inp, sizeof *inp);
1209                 if (inp->inp_socket)
1210                         sotoxsocket(inp->inp_socket, &xi.xi_socket);
1211                 if ((error = SYSCTL_OUT(req, &xi, sizeof xi)) != 0)
1212                         break;
1213                 ++i;
1214         }
1215         LIST_REMOVE(marker, inp_list);
1216         if (error == 0 && i < n) {
1217                 bzero(&xi, sizeof(xi));
1218                 xi.xi_len = sizeof(xi);
1219                 while (i < n) {
1220                         error = SYSCTL_OUT(req, &xi, sizeof xi);
1221                         ++i;
1222                 }
1223         }
1224         free(marker, M_TEMP);
1225         return(error);
1226 }