bsd/netinet/udp_usrreq.c

   1 /*
   2  * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
   3  *
   4  * @APPLE_LICENSE_HEADER_START@
   5  *
   6  * The contents of this file constitute Original Code as defined in and
   7  * are subject to the Apple Public Source License Version 1.1 (the
   8  * "License").  You may not use this file except in compliance with the
   9  * License.  Please obtain a copy of the License at
  10  * http://www.apple.com/publicsource and read it before using this file.
  11  *
  12  * This Original Code and all software distributed under the License are
  13  * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  14  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  15  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  16  * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT.  Please see the
  17  * License for the specific language governing rights and limitations
  18  * under the License.
  19  *
  20  * @APPLE_LICENSE_HEADER_END@
  21  */
  22 /*
  23  * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995
  24  *      The Regents of the University of California.  All rights reserved.
  25  *
  26  * Redistribution and use in source and binary forms, with or without
  27  * modification, are permitted provided that the following conditions
  28  * are met:
  29  * 1. Redistributions of source code must retain the above copyright
  30  *    notice, this list of conditions and the following disclaimer.
  31  * 2. Redistributions in binary form must reproduce the above copyright
  32  *    notice, this list of conditions and the following disclaimer in the
  33  *    documentation and/or other materials provided with the distribution.
  34  * 3. All advertising materials mentioning features or use of this software
  35  *    must display the following acknowledgement:
  36  *      This product includes software developed by the University of
  37  *      California, Berkeley and its contributors.
  38  * 4. Neither the name of the University nor the names of its contributors
  39  *    may be used to endorse or promote products derived from this software
  40  *    without specific prior written permission.
  41  *
  42  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  43  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  44  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  45  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  46  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  47  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  48  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  49  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  50  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  51  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  52  * SUCH DAMAGE.
  53  *
  54  *      @(#)udp_usrreq.c        8.6 (Berkeley) 5/23/95
  55  * $FreeBSD: src/sys/netinet/udp_usrreq.c,v 1.64.2.13 2001/08/08 18:59:54 ghelmer Exp $
  56  */
  57
  58 #include <sys/param.h>
  59 #include <sys/systm.h>
  60 #include <sys/kernel.h>
  61 #include <sys/malloc.h>
  62 #include <sys/mbuf.h>
  63 #include <sys/domain.h>
  64 #include <sys/protosw.h>
  65 #include <sys/socket.h>
  66 #include <sys/socketvar.h>
  67 #include <sys/sysctl.h>
  68 #include <sys/syslog.h>
  69
  70 #include <net/if.h>
  71 #include <net/route.h>
  72
  73 #include <netinet/in.h>
  74 #include <netinet/in_systm.h>
  75 #include <netinet/ip.h>
  76 #if INET6
  77 #include <netinet/ip6.h>
  78 #endif
  79 #include <netinet/in_pcb.h>
  80 #include <netinet/in_var.h>
  81 #include <netinet/ip_var.h>
  82 #if INET6
  83 #include <netinet6/ip6_var.h>
  84 #endif
  85 #include <netinet/ip_icmp.h>
  86 #include <netinet/icmp_var.h>
  87 #include <netinet/udp.h>
  88 #include <netinet/udp_var.h>
  89 #include <sys/kdebug.h>
  90
  91 #if IPSEC
  92 #include <netinet6/ipsec.h>
  93 extern int ipsec_bypass;
  94 #endif /*IPSEC*/
  95
  96
  97 #define DBG_LAYER_IN_BEG        NETDBG_CODE(DBG_NETUDP, 0)
  98 #define DBG_LAYER_IN_END        NETDBG_CODE(DBG_NETUDP, 2)
  99 #define DBG_LAYER_OUT_BEG       NETDBG_CODE(DBG_NETUDP, 1)
 100 #define DBG_LAYER_OUT_END       NETDBG_CODE(DBG_NETUDP, 3)
 101 #define DBG_FNC_UDP_INPUT       NETDBG_CODE(DBG_NETUDP, (5 << 8))
 102 #define DBG_FNC_UDP_OUTPUT      NETDBG_CODE(DBG_NETUDP, (6 << 8) | 1)
 103
 104 /*
 105  * UDP protocol implementation.
 106  * Per RFC 768, August, 1980.
 107  */
 108 #ifndef COMPAT_42
 109 static int      udpcksum = 1;
 110 #else
 111 static int      udpcksum = 0;           /* XXX */
 112 #endif
 113 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW,
 114                 &udpcksum, 0, "");
 115
 116 int     log_in_vain = 0;
 117 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW,
 118     &log_in_vain, 0, "Log all incoming UDP packets");
 119
 120 static int      blackhole = 0;
 121 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW,
 122         &blackhole, 0, "Do not send port unreachables for refused connects");
 123
 124 struct  inpcbhead udb;          /* from udp_var.h */
 125 #define udb6    udb  /* for KAME src sync over BSD*'s */
 126 struct  inpcbinfo udbinfo;
 127
 128 #ifndef UDBHASHSIZE
 129 #define UDBHASHSIZE 16
 130 #endif
 131
 132 extern  int apple_hwcksum_rx;
 133 extern  int     esp_udp_encap_port;
 134 extern  u_long  route_generation;
 135
 136 struct  udpstat udpstat;        /* from udp_var.h */
 137 SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RD,
 138     &udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)");
 139
 140 static struct   sockaddr_in udp_in = { sizeof(udp_in), AF_INET };
 141 #if INET6
 142 struct udp_in6 {
 143         struct sockaddr_in6     uin6_sin;
 144         u_char                  uin6_init_done : 1;
 145 } udp_in6 = {
 146         { sizeof(udp_in6.uin6_sin), AF_INET6 },
 147         0
 148 };
 149 struct udp_ip6 {
 150         struct ip6_hdr          uip6_ip6;
 151         u_char                  uip6_init_done : 1;
 152 } udp_ip6;
 153 #endif /* INET6 */
 154
 155 static void udp_append __P((struct inpcb *last, struct ip *ip,
 156                             struct mbuf *n, int off));
 157 #if INET6
 158 static void ip_2_ip6_hdr __P((struct ip6_hdr *ip6, struct ip *ip));
 159 #endif
 160
 161 static int udp_detach __P((struct socket *so));
 162 static  int udp_output __P((struct inpcb *, struct mbuf *, struct sockaddr *,
 163                             struct mbuf *, struct proc *));
 164
 165 void
 166 udp_init()
 167 {
 168         vm_size_t       str_size;
 169         int             stat;
 170         u_char          fake_owner;
 171         struct in_addr  laddr;
 172         struct in_addr  faddr;
 173         u_short         lport;
 174
 175         LIST_INIT(&udb);
 176         udbinfo.listhead = &udb;
 177         udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask);
 178         udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB,
 179                                         &udbinfo.porthashmask);
 180 #ifdef __APPLE__
 181         str_size = (vm_size_t) sizeof(struct inpcb);
 182         udbinfo.ipi_zone = (void *) zinit(str_size, 80000*str_size, 8192, "udpcb");
 183
 184         udbinfo.last_pcb = 0;
 185         in_pcb_nat_init(&udbinfo, AF_INET, IPPROTO_UDP, SOCK_DGRAM);
 186 #else
 187         udbinfo.ipi_zone = zinit("udpcb", sizeof(struct inpcb), maxsockets,
 188                                  ZONE_INTERRUPT, 0);
 189 #endif
 190
 191 #if 0
 192         /* for pcb sharing testing only */
 193         stat = in_pcb_new_share_client(&udbinfo, &fake_owner);
 194         kprintf("udp_init in_pcb_new_share_client - stat = %d\n", stat);
 195
 196         laddr.s_addr = 0x11646464;
 197         faddr.s_addr = 0x11646465;
 198
 199         lport = 1500;
 200         in_pcb_grab_port(&udbinfo, 0, laddr, &lport, faddr, 1600, 0, fake_owner);
 201         kprintf("udp_init in_pcb_grab_port - stat = %d\n", stat);
 202
 203         stat = in_pcb_rem_share_client(&udbinfo, fake_owner);
 204         kprintf("udp_init in_pcb_rem_share_client - stat = %d\n", stat);
 205
 206         stat = in_pcb_new_share_client(&udbinfo, &fake_owner);
 207         kprintf("udp_init in_pcb_new_share_client(2) - stat = %d\n", stat);
 208
 209         laddr.s_addr = 0x11646464;
 210         faddr.s_addr = 0x11646465;
 211
 212         lport = 1500;
 213         stat = in_pcb_grab_port(&udbinfo, 0, laddr, &lport, faddr, 1600, 0, fake_owner);
 214         kprintf("udp_init in_pcb_grab_port(2) - stat = %d\n", stat);
 215 #endif
 216 }
 217
 218 void
 219 udp_input(m, iphlen)
 220         register struct mbuf *m;
 221         int iphlen;
 222 {
 223         register struct ip *ip;
 224         register struct udphdr *uh;
 225         register struct inpcb *inp;
 226         struct mbuf *opts = 0;
 227         int len;
 228         struct ip save_ip;
 229         struct sockaddr *append_sa;
 230
 231         udpstat.udps_ipackets++;
 232
 233         KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_START, 0,0,0,0,0);
 234         if (m->m_pkthdr.csum_flags & CSUM_TCP_SUM16)
 235                 m->m_pkthdr.csum_flags = 0; /* invalidate hwcksum for UDP */
 236
 237         /*
 238          * Strip IP options, if any; should skip this,
 239          * make available to user, and use on returned packets,
 240          * but we don't yet have a way to check the checksum
 241          * with options still present.
 242          */
 243         if (iphlen > sizeof (struct ip)) {
 244                 ip_stripoptions(m, (struct mbuf *)0);
 245                 iphlen = sizeof(struct ip);
 246         }
 247
 248         /*
 249          * Get IP and UDP header together in first mbuf.
 250          */
 251         ip = mtod(m, struct ip *);
 252         if (m->m_len < iphlen + sizeof(struct udphdr)) {
 253                 if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == 0) {
 254                         udpstat.udps_hdrops++;
 255                         KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0,0,0,0,0);
 256                         return;
 257                 }
 258                 ip = mtod(m, struct ip *);
 259         }
 260         uh = (struct udphdr *)((caddr_t)ip + iphlen);
 261
 262         /* destination port of 0 is illegal, based on RFC768. */
 263         if (uh->uh_dport == 0)
 264                 goto bad;
 265
 266         KERNEL_DEBUG(DBG_LAYER_IN_BEG, uh->uh_dport, uh->uh_sport,
 267                      ip->ip_src.s_addr, ip->ip_dst.s_addr, uh->uh_ulen);
 268
 269         /*
 270          * Make mbuf data length reflect UDP length.
 271          * If not enough data to reflect UDP length, drop.
 272          */
 273         len = ntohs((u_short)uh->uh_ulen);
 274         if (ip->ip_len != len) {
 275                 if (len > ip->ip_len || len < sizeof(struct udphdr)) {
 276                         udpstat.udps_badlen++;
 277                         goto bad;
 278                 }
 279                 m_adj(m, len - ip->ip_len);
 280                 /* ip->ip_len = len; */
 281         }
 282         /*
 283          * Save a copy of the IP header in case we want restore it
 284          * for sending an ICMP error message in response.
 285          */
 286         save_ip = *ip;
 287
 288         /*
 289          * Checksum extended UDP header and data.
 290          */
 291         if (uh->uh_sum) {
 292                 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) {
 293                         if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR)
 294                                 uh->uh_sum = m->m_pkthdr.csum_data;
 295                         else
 296                                 goto doudpcksum;
 297                         uh->uh_sum ^= 0xffff;
 298                 } else {
 299                         char b[9];
 300 doudpcksum:
 301                         *(uint32_t*)&b[0] = *(uint32_t*)&((struct ipovly *)ip)->ih_x1[0];
 302                         *(uint32_t*)&b[4] = *(uint32_t*)&((struct ipovly *)ip)->ih_x1[4];
 303                         *(uint8_t*)&b[8] = *(uint8_t*)&((struct ipovly *)ip)->ih_x1[8];
 304
 305                         bzero(((struct ipovly *)ip)->ih_x1, 9);
 306                         ((struct ipovly *)ip)->ih_len = uh->uh_ulen;
 307                         uh->uh_sum = in_cksum(m, len + sizeof (struct ip));
 308
 309                         *(uint32_t*)&((struct ipovly *)ip)->ih_x1[0] = *(uint32_t*)&b[0];
 310                         *(uint32_t*)&((struct ipovly *)ip)->ih_x1[4] = *(uint32_t*)&b[4];
 311                         *(uint8_t*)&((struct ipovly *)ip)->ih_x1[8] = *(uint8_t*)&b[8];
 312                 }
 313                 if (uh->uh_sum) {
 314                         udpstat.udps_badsum++;
 315                         m_freem(m);
 316                         KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0,0,0,0,0);
 317                         return;
 318                 }
 319         }
 320 #ifndef __APPLE__
 321          else
 322                 udpstat.udps_nosum++;
 323 #endif
 324
 325         if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) ||
 326             in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) {
 327                 struct inpcb *last;
 328                 /*
 329                  * Deliver a multicast or broadcast datagram to *all* sockets
 330                  * for which the local and remote addresses and ports match
 331                  * those of the incoming datagram.  This allows more than
 332                  * one process to receive multi/broadcasts on the same port.
 333                  * (This really ought to be done for unicast datagrams as
 334                  * well, but that would cause problems with existing
 335                  * applications that open both address-specific sockets and
 336                  * a wildcard socket listening to the same port -- they would
 337                  * end up receiving duplicates of every unicast datagram.
 338                  * Those applications open the multiple sockets to overcome an
 339                  * inadequacy of the UDP socket interface, but for backwards
 340                  * compatibility we avoid the problem here rather than
 341                  * fixing the interface.  Maybe 4.5BSD will remedy this?)
 342                  */
 343
 344                 /*
 345                  * Construct sockaddr format source address.
 346                  */
 347                 udp_in.sin_port = uh->uh_sport;
 348                 udp_in.sin_addr = ip->ip_src;
 349                 /*
 350                  * Locate pcb(s) for datagram.
 351                  * (Algorithm copied from raw_intr().)
 352                  */
 353                 last = NULL;
 354 #if INET6
 355                 udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0;
 356 #endif
 357                 LIST_FOREACH(inp, &udb, inp_list) {
 358 #ifdef __APPLE__
 359                         /* Ignore nat/SharedIP dummy pcbs */
 360                         if (inp->inp_socket == &udbinfo.nat_dummy_socket)
 361                                 continue;
 362 #endif
 363 #if INET6
 364                         if ((inp->inp_vflag & INP_IPV4) == 0)
 365                                 continue;
 366 #endif
 367                         if (inp->inp_lport != uh->uh_dport)
 368                                 continue;
 369                         if (inp->inp_laddr.s_addr != INADDR_ANY) {
 370                                 if (inp->inp_laddr.s_addr !=
 371                                     ip->ip_dst.s_addr)
 372                                         continue;
 373                         }
 374                         if (inp->inp_faddr.s_addr != INADDR_ANY) {
 375                                 if (inp->inp_faddr.s_addr !=
 376                                     ip->ip_src.s_addr ||
 377                                     inp->inp_fport != uh->uh_sport)
 378                                         continue;
 379                         }
 380
 381                         if (last != NULL) {
 382                                 struct mbuf *n;
 383
 384 #if IPSEC
 385                                 /* check AH/ESP integrity. */
 386                                 if (ipsec_bypass == 0 && ipsec4_in_reject_so(m, last->inp_socket)) {
 387                                         ipsecstat.in_polvio++;
 388                                         /* do not inject data to pcb */
 389                                 } else
 390 #endif /*IPSEC*/
 391                                 if ((n = m_copy(m, 0, M_COPYALL)) != NULL) {
 392                                         udp_append(last, ip, n,
 393                                                    iphlen +
 394                                                    sizeof(struct udphdr));
 395                                 }
 396                         }
 397                         last = inp;
 398                         /*
 399                          * Don't look for additional matches if this one does
 400                          * not have either the SO_REUSEPORT or SO_REUSEADDR
 401                          * socket options set.  This heuristic avoids searching
 402                          * through all pcbs in the common case of a non-shared
 403                          * port.  It * assumes that an application will never
 404                          * clear these options after setting them.
 405                          */
 406                         if ((last->inp_socket->so_options&(SO_REUSEPORT|SO_REUSEADDR)) == 0)
 407                                 break;
 408                 }
 409
 410                 if (last == NULL) {
 411                         /*
 412                          * No matching pcb found; discard datagram.
 413                          * (No need to send an ICMP Port Unreachable
 414                          * for a broadcast or multicast datgram.)
 415                          */
 416                         udpstat.udps_noportbcast++;
 417                         goto bad;
 418                 }
 419 #if IPSEC
 420                 /* check AH/ESP integrity. */
 421                 if (ipsec_bypass == 0 && m && ipsec4_in_reject_so(m, last->inp_socket)) {
 422                         ipsecstat.in_polvio++;
 423                         goto bad;
 424                 }
 425 #endif /*IPSEC*/
 426                 udp_append(last, ip, m, iphlen + sizeof(struct udphdr));
 427                 return;
 428         }
 429
 430         /*
 431          * UDP to port 4500 with a payload where the first four bytes are
 432          * not zero is a UDP encapsulated IPSec packet. Packets where
 433          * the payload is one byte and that byte is 0xFF are NAT keepalive
 434          * packets. Decapsulate the ESP packet and carry on with IPSec input
 435          * or discard the NAT keep-alive.
 436          */
 437         if (ipsec_bypass == 0 && (esp_udp_encap_port & 0xFFFF) != 0 &&
 438                 uh->uh_dport == ntohs((u_short)esp_udp_encap_port)) {
 439                 int     payload_len = len - sizeof(struct udphdr) > 4 ? 4 : len - sizeof(struct udphdr);
 440                 if (m->m_len < iphlen + sizeof(struct udphdr) + payload_len) {
 441                         if ((m = m_pullup(m, iphlen + sizeof(struct udphdr) + payload_len)) == 0) {
 442                                 udpstat.udps_hdrops++;
 443                                 KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0,0,0,0,0);
 444                                 return;
 445                         }
 446                         ip = mtod(m, struct ip *);
 447                         uh = (struct udphdr *)((caddr_t)ip + iphlen);
 448                 }
 449                 /* Check for NAT keepalive packet */
 450                 if (payload_len == 1 && *(u_int8_t*)((caddr_t)uh + sizeof(struct udphdr)) == 0xFF) {
 451                         m_freem(m);
 452                         KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0,0,0,0,0);
 453                         return;
 454                 }
 455                 else if (payload_len == 4 && *(u_int32_t*)((caddr_t)uh + sizeof(struct udphdr)) != 0) {
 456                         /* UDP encapsulated IPSec packet to pass through NAT */
 457                         size_t stripsiz;
 458
 459                         stripsiz = sizeof(struct udphdr);
 460
 461                         ip = mtod(m, struct ip *);
 462                         ovbcopy((caddr_t)ip, (caddr_t)(((u_char *)ip) + stripsiz), iphlen);
 463                         m->m_data += stripsiz;
 464                         m->m_len -= stripsiz;
 465                         m->m_pkthdr.len -= stripsiz;
 466                         ip = mtod(m, struct ip *);
 467                         ip->ip_len = ip->ip_len - stripsiz;
 468                         ip->ip_p = IPPROTO_ESP;
 469
 470                         KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0,0,0,0,0);
 471                         esp4_input(m, iphlen);
 472                         return;
 473                 }
 474         }
 475
 476         /*
 477          * Locate pcb for datagram.
 478          */
 479         inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport,
 480             ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif);
 481         if (inp == NULL) {
 482                 if (log_in_vain) {
 483                         char buf[4*sizeof "123"];
 484
 485                         strcpy(buf, inet_ntoa(ip->ip_dst));
 486                         log(LOG_INFO,
 487                             "Connection attempt to UDP %s:%d from %s:%d\n",
 488                             buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src),
 489                             ntohs(uh->uh_sport));
 490                 }
 491                 udpstat.udps_noport++;
 492                 if (m->m_flags & (M_BCAST | M_MCAST)) {
 493                         udpstat.udps_noportbcast++;
 494                         goto bad;
 495                 }
 496 #if ICMP_BANDLIM
 497                 if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0)
 498                         goto bad;
 499 #endif
 500                 if (blackhole)
 501                         goto bad;
 502                 *ip = save_ip;
 503                 ip->ip_len += iphlen;
 504                 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0);
 505                 KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0,0,0,0,0);
 506                 return;
 507         }
 508 #if IPSEC
 509         if (ipsec_bypass == 0 && inp != NULL && ipsec4_in_reject_so(m, inp->inp_socket)) {
 510                 ipsecstat.in_polvio++;
 511                 goto bad;
 512         }
 513 #endif /*IPSEC*/
 514
 515         /*
 516          * Construct sockaddr format source address.
 517          * Stuff source address and datagram in user buffer.
 518          */
 519         udp_in.sin_port = uh->uh_sport;
 520         udp_in.sin_addr = ip->ip_src;
 521         if (inp->inp_flags & INP_CONTROLOPTS
 522             || inp->inp_socket->so_options & SO_TIMESTAMP) {
 523 #if INET6
 524                 if (inp->inp_vflag & INP_IPV6) {
 525                         int savedflags;
 526
 527                         ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
 528                         savedflags = inp->inp_flags;
 529                         inp->inp_flags &= ~INP_UNMAPPABLEOPTS;
 530                         ip6_savecontrol(inp, &opts, &udp_ip6.uip6_ip6, m);
 531                         inp->inp_flags = savedflags;
 532                 } else
 533 #endif
 534                 ip_savecontrol(inp, &opts, ip, m);
 535         }
 536         m_adj(m, iphlen + sizeof(struct udphdr));
 537
 538         KERNEL_DEBUG(DBG_LAYER_IN_END, uh->uh_dport, uh->uh_sport,
 539                      save_ip.ip_src.s_addr, save_ip.ip_dst.s_addr, uh->uh_ulen);
 540
 541 #if INET6
 542         if (inp->inp_vflag & INP_IPV6) {
 543                 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
 544                 append_sa = (struct sockaddr *)&udp_in6;
 545         } else
 546 #endif
 547         append_sa = (struct sockaddr *)&udp_in;
 548         if (sbappendaddr(&inp->inp_socket->so_rcv, append_sa, m, opts) == 0) {
 549                 udpstat.udps_fullsock++;
 550                 goto bad;
 551         }
 552         sorwakeup(inp->inp_socket);
 553         KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0,0,0,0,0);
 554         return;
 555 bad:
 556         m_freem(m);
 557         if (opts)
 558                 m_freem(opts);
 559         KERNEL_DEBUG(DBG_FNC_UDP_INPUT | DBG_FUNC_END, 0,0,0,0,0);
 560         return;
 561 }
 562
 563 #if INET6
 564 static void
 565 ip_2_ip6_hdr(ip6, ip)
 566         struct ip6_hdr *ip6;
 567         struct ip *ip;
 568 {
 569         bzero(ip6, sizeof(*ip6));
 570
 571         ip6->ip6_vfc = IPV6_VERSION;
 572         ip6->ip6_plen = ip->ip_len;
 573         ip6->ip6_nxt = ip->ip_p;
 574         ip6->ip6_hlim = ip->ip_ttl;
 575         ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] =
 576                 IPV6_ADDR_INT32_SMP;
 577         ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr;
 578         ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr;
 579 }
 580 #endif
 581
 582 /*
 583  * subroutine of udp_input(), mainly for source code readability.
 584  * caller must properly init udp_ip6 and udp_in6 beforehand.
 585  */
 586 static void
 587 udp_append(last, ip, n, off)
 588         struct inpcb *last;
 589         struct ip *ip;
 590         struct mbuf *n;
 591         int off;
 592 {
 593         struct sockaddr *append_sa;
 594         struct mbuf *opts = 0;
 595
 596         if (last->inp_flags & INP_CONTROLOPTS ||
 597             last->inp_socket->so_options & SO_TIMESTAMP) {
 598 #if INET6
 599                 if (last->inp_vflag & INP_IPV6) {
 600                         int savedflags;
 601
 602                         if (udp_ip6.uip6_init_done == 0) {
 603                                 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip);
 604                                 udp_ip6.uip6_init_done = 1;
 605                         }
 606                         savedflags = last->inp_flags;
 607                         last->inp_flags &= ~INP_UNMAPPABLEOPTS;
 608                         ip6_savecontrol(last, &opts, &udp_ip6.uip6_ip6, n);
 609                         last->inp_flags = savedflags;
 610                 } else
 611 #endif
 612                 ip_savecontrol(last, &opts, ip, n);
 613         }
 614 #if INET6
 615         if (last->inp_vflag & INP_IPV6) {
 616                 if (udp_in6.uin6_init_done == 0) {
 617                         in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin);
 618                         udp_in6.uin6_init_done = 1;
 619                 }
 620                 append_sa = (struct sockaddr *)&udp_in6.uin6_sin;
 621         } else
 622 #endif
 623         append_sa = (struct sockaddr *)&udp_in;
 624         m_adj(n, off);
 625         if (sbappendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) {
 626                 m_freem(n);
 627                 if (opts)
 628                         m_freem(opts);
 629                 udpstat.udps_fullsock++;
 630         } else
 631                 sorwakeup(last->inp_socket);
 632 }
 633
 634 /*
 635  * Notify a udp user of an asynchronous error;
 636  * just wake up so that he can collect error status.
 637  */
 638 void
 639 udp_notify(inp, errno)
 640         register struct inpcb *inp;
 641         int errno;
 642 {
 643         inp->inp_socket->so_error = errno;
 644         sorwakeup(inp->inp_socket);
 645         sowwakeup(inp->inp_socket);
 646 }
 647
 648 void
 649 udp_ctlinput(cmd, sa, vip)
 650         int cmd;
 651         struct sockaddr *sa;
 652         void *vip;
 653 {
 654         struct ip *ip = vip;
 655         struct udphdr *uh;
 656         void (*notify) __P((struct inpcb *, int)) = udp_notify;
 657         struct in_addr faddr;
 658         struct inpcb *inp;
 659         int s;
 660
 661         faddr = ((struct sockaddr_in *)sa)->sin_addr;
 662         if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY)
 663                 return;
 664
 665         if (PRC_IS_REDIRECT(cmd)) {
 666                 ip = 0;
 667                 notify = in_rtchange;
 668         } else if (cmd == PRC_HOSTDEAD)
 669                 ip = 0;
 670         else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0)
 671                 return;
 672         if (ip) {
 673                 s = splnet();
 674                 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2));
 675                 inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport,
 676                     ip->ip_src, uh->uh_sport, 0, NULL);
 677                 if (inp != NULL && inp->inp_socket != NULL)
 678                         (*notify)(inp, inetctlerrmap[cmd]);
 679                 splx(s);
 680         } else
 681                 in_pcbnotifyall(&udb, faddr, inetctlerrmap[cmd], notify);
 682 }
 683
 684 static int
 685 udp_pcblist SYSCTL_HANDLER_ARGS
 686 {
 687         int error, i, n, s;
 688         struct inpcb *inp, **inp_list;
 689         inp_gen_t gencnt;
 690         struct xinpgen xig;
 691
 692         /*
 693          * The process of preparing the TCB list is too time-consuming and
 694          * resource-intensive to repeat twice on every request.
 695          */
 696         if (req->oldptr == 0) {
 697                 n = udbinfo.ipi_count;
 698                 req->oldidx = 2 * (sizeof xig)
 699                         + (n + n/8) * sizeof(struct xinpcb);
 700                 return 0;
 701         }
 702
 703         if (req->newptr != 0)
 704                 return EPERM;
 705
 706         /*
 707          * OK, now we're committed to doing something.
 708          */
 709         s = splnet();
 710         gencnt = udbinfo.ipi_gencnt;
 711         n = udbinfo.ipi_count;
 712         splx(s);
 713
 714         xig.xig_len = sizeof xig;
 715         xig.xig_count = n;
 716         xig.xig_gen = gencnt;
 717         xig.xig_sogen = so_gencnt;
 718         error = SYSCTL_OUT(req, &xig, sizeof xig);
 719         if (error)
 720                 return error;
 721     /*
 722      * We are done if there is no pcb
 723      */
 724     if (n == 0)
 725         return 0;
 726
 727         inp_list = _MALLOC(n * sizeof *inp_list, M_TEMP, M_WAITOK);
 728         if (inp_list == 0) {
 729                 return ENOMEM;
 730         }
 731
 732         for (inp = LIST_FIRST(udbinfo.listhead), i = 0; inp && i < n;
 733              inp = LIST_NEXT(inp, inp_list)) {
 734                 if (inp->inp_gencnt <= gencnt)
 735                         inp_list[i++] = inp;
 736         }
 737         splx(s);
 738         n = i;
 739
 740         error = 0;
 741         for (i = 0; i < n; i++) {
 742                 inp = inp_list[i];
 743                 if (inp->inp_gencnt <= gencnt) {
 744                         struct xinpcb xi;
 745                         xi.xi_len = sizeof xi;
 746                         /* XXX should avoid extra copy */
 747                         bcopy(inp, &xi.xi_inp, sizeof *inp);
 748                         if (inp->inp_socket)
 749                                 sotoxsocket(inp->inp_socket, &xi.xi_socket);
 750                         error = SYSCTL_OUT(req, &xi, sizeof xi);
 751                 }
 752         }
 753         if (!error) {
 754                 /*
 755                  * Give the user an updated idea of our state.
 756                  * If the generation differs from what we told
 757                  * her before, she knows that something happened
 758                  * while we were processing this request, and it
 759                  * might be necessary to retry.
 760                  */
 761                 s = splnet();
 762                 xig.xig_gen = udbinfo.ipi_gencnt;
 763                 xig.xig_sogen = so_gencnt;
 764                 xig.xig_count = udbinfo.ipi_count;
 765                 splx(s);
 766                 error = SYSCTL_OUT(req, &xig, sizeof xig);
 767         }
 768         FREE(inp_list, M_TEMP);
 769         return error;
 770 }
 771
 772 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0,
 773             udp_pcblist, "S,xinpcb", "List of active UDP sockets");
 774
 775
 776
 777 static int
 778 udp_output(inp, m, addr, control, p)
 779         register struct inpcb *inp;
 780         struct mbuf *m;
 781         struct sockaddr *addr;
 782         struct mbuf *control;
 783         struct proc *p;
 784 {
 785         register struct udpiphdr *ui;
 786         register int len = m->m_pkthdr.len;
 787         struct in_addr laddr;
 788         int s = 0, error = 0;
 789
 790         KERNEL_DEBUG(DBG_FNC_UDP_OUTPUT | DBG_FUNC_START, 0,0,0,0,0);
 791
 792         if (control)
 793                 m_freem(control);               /* XXX */
 794
 795         KERNEL_DEBUG(DBG_LAYER_OUT_BEG, inp->inp_fport, inp->inp_lport,
 796                      inp->inp_laddr.s_addr, inp->inp_faddr.s_addr,
 797                      (htons((u_short)len + sizeof (struct udphdr))));
 798
 799         if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
 800                 error = EMSGSIZE;
 801                 goto release;
 802         }
 803
 804         /* If there was a routing change, discard cached route and check
 805          * that we have a valid source address.
 806          * Reacquire a new source address if INADDR_ANY was specified
 807          */
 808
 809         if (inp->inp_route.ro_rt && inp->inp_route.ro_rt->generation_id != route_generation) {
 810                 if (ifa_foraddr(inp->inp_laddr.s_addr) == NULL) { /* src address is gone */
 811                         if (inp->inp_flags & INP_INADDR_ANY)
 812                                 inp->inp_faddr.s_addr = INADDR_ANY; /* new src will be set later */
 813                         else {
 814                                 error = EADDRNOTAVAIL;
 815                                 goto release;
 816                         }
 817                 }
 818                 rtfree(inp->inp_route.ro_rt);
 819                 inp->inp_route.ro_rt = (struct rtentry *)0;
 820         }
 821
 822         if (addr) {
 823                 laddr = inp->inp_laddr;
 824                 if (inp->inp_faddr.s_addr != INADDR_ANY) {
 825                         error = EISCONN;
 826                         goto release;
 827                 }
 828                 /*
 829                  * Must block input while temporarily connected.
 830                  */
 831                 s = splnet();
 832                 error = in_pcbconnect(inp, addr, p);
 833                 if (error) {
 834                         splx(s);
 835                         goto release;
 836                 }
 837         } else {
 838                 if (inp->inp_faddr.s_addr == INADDR_ANY) {
 839                         error = ENOTCONN;
 840                         goto release;
 841                 }
 842         }
 843
 844
 845         /*
 846          * Calculate data length and get a mbuf
 847          * for UDP and IP headers.
 848          */
 849         M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT);
 850         if (m == 0) {
 851                 error = ENOBUFS;
 852                 goto abort;
 853         }
 854
 855         /*
 856          * Fill in mbuf with extended UDP header
 857          * and addresses and length put into network format.
 858          */
 859         ui = mtod(m, struct udpiphdr *);
 860         bzero(ui->ui_x1, sizeof(ui->ui_x1));    /* XXX still needed? */
 861         ui->ui_pr = IPPROTO_UDP;
 862         ui->ui_src = inp->inp_laddr;
 863         ui->ui_dst = inp->inp_faddr;
 864         ui->ui_sport = inp->inp_lport;
 865         ui->ui_dport = inp->inp_fport;
 866         ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
 867
 868         /*
 869          * Set up checksum and output datagram.
 870          */
 871         if (udpcksum) {
 872                 ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr,
 873                     htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP));
 874                 m->m_pkthdr.csum_flags = CSUM_UDP;
 875                 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
 876         } else {
 877                 ui->ui_sum = 0;
 878         }
 879         ((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len;
 880         ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl;    /* XXX */
 881         ((struct ip *)ui)->ip_tos = inp->inp_ip_tos;    /* XXX */
 882         udpstat.udps_opackets++;
 883
 884         KERNEL_DEBUG(DBG_LAYER_OUT_END, ui->ui_dport, ui->ui_sport,
 885                      ui->ui_src.s_addr, ui->ui_dst.s_addr, ui->ui_ulen);
 886
 887 #if IPSEC
 888         if (ipsec_bypass == 0 && ipsec_setsocket(m, inp->inp_socket) != 0) {
 889                 error = ENOBUFS;
 890                 goto abort;
 891         }
 892 #endif /*IPSEC*/
 893         error = ip_output(m, inp->inp_options, &inp->inp_route,
 894             (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)),
 895             inp->inp_moptions);
 896
 897         if (addr) {
 898                 in_pcbdisconnect(inp);
 899                 inp->inp_laddr = laddr; /* XXX rehash? */
 900                 splx(s);
 901         }
 902         KERNEL_DEBUG(DBG_FNC_UDP_OUTPUT | DBG_FUNC_END, error, 0,0,0,0);
 903         return (error);
 904
 905 abort:
 906         if (addr) {
 907                 in_pcbdisconnect(inp);
 908                 inp->inp_laddr = laddr; /* XXX rehash? */
 909                 splx(s);
 910         }
 911
 912 release:
 913         m_freem(m);
 914         KERNEL_DEBUG(DBG_FNC_UDP_OUTPUT | DBG_FUNC_END, error, 0,0,0,0);
 915         return (error);
 916 }
 917
 918 u_long  udp_sendspace = 9216;           /* really max datagram size */
 919                                         /* 40 1K datagrams */
 920 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW,
 921     &udp_sendspace, 0, "Maximum outgoing UDP datagram size");
 922
 923 u_long  udp_recvspace = 40 * (1024 +
 924 #if INET6
 925                                       sizeof(struct sockaddr_in6)
 926 #else
 927                                       sizeof(struct sockaddr_in)
 928 #endif
 929                                       );
 930 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW,
 931     &udp_recvspace, 0, "Maximum incoming UDP datagram size");
 932
 933 static int
 934 udp_abort(struct socket *so)
 935 {
 936         struct inpcb *inp;
 937         int s;
 938
 939         inp = sotoinpcb(so);
 940         if (inp == 0)
 941                 return EINVAL;  /* ??? possible? panic instead? */
 942         soisdisconnected(so);
 943         s = splnet();
 944         in_pcbdetach(inp);
 945         splx(s);
 946         return 0;
 947 }
 948
 949 static int
 950 udp_attach(struct socket *so, int proto, struct proc *p)
 951 {
 952         struct inpcb *inp;
 953         int error; long s;
 954
 955         inp = sotoinpcb(so);
 956         if (inp != 0)
 957                 return EINVAL;
 958
 959         error = soreserve(so, udp_sendspace, udp_recvspace);
 960         if (error)
 961                 return error;
 962         s = splnet();
 963         error = in_pcballoc(so, &udbinfo, p);
 964         splx(s);
 965         if (error)
 966                 return error;
 967         inp = (struct inpcb *)so->so_pcb;
 968         inp->inp_vflag |= INP_IPV4;
 969         inp->inp_ip_ttl = ip_defttl;
 970         return 0;
 971 }
 972
 973 static int
 974 udp_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
 975 {
 976         struct inpcb *inp;
 977         int s, error;
 978
 979         inp = sotoinpcb(so);
 980         if (inp == 0)
 981                 return EINVAL;
 982         s = splnet();
 983         error = in_pcbbind(inp, nam, p);
 984         splx(s);
 985         return error;
 986 }
 987
 988 static int
 989 udp_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
 990 {
 991         struct inpcb *inp;
 992         int s, error;
 993
 994         inp = sotoinpcb(so);
 995         if (inp == 0)
 996                 return EINVAL;
 997         if (inp->inp_faddr.s_addr != INADDR_ANY)
 998                 return EISCONN;
 999         s = splnet();
1000         error = in_pcbconnect(inp, nam, p);
1001         splx(s);
1002         if (error == 0)
1003                 soisconnected(so);
1004         return error;
1005 }
1006
1007 static int
1008 udp_detach(struct socket *so)
1009 {
1010         struct inpcb *inp;
1011         int s;
1012
1013         inp = sotoinpcb(so);
1014         if (inp == 0)
1015                 return EINVAL;
1016         s = splnet();
1017         in_pcbdetach(inp);
1018         splx(s);
1019         return 0;
1020 }
1021
1022 static int
1023 udp_disconnect(struct socket *so)
1024 {
1025         struct inpcb *inp;
1026         int s;
1027
1028         inp = sotoinpcb(so);
1029         if (inp == 0)
1030                 return EINVAL;
1031         if (inp->inp_faddr.s_addr == INADDR_ANY)
1032                 return ENOTCONN;
1033
1034         s = splnet();
1035         in_pcbdisconnect(inp);
1036         inp->inp_laddr.s_addr = INADDR_ANY;
1037         splx(s);
1038         so->so_state &= ~SS_ISCONNECTED;                /* XXX */
1039         return 0;
1040 }
1041
1042 static int
1043 udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr,
1044             struct mbuf *control, struct proc *p)
1045 {
1046         struct inpcb *inp;
1047
1048         inp = sotoinpcb(so);
1049         if (inp == 0) {
1050                 m_freem(m);
1051                 return EINVAL;
1052         }
1053         return udp_output(inp, m, addr, control, p);
1054 }
1055
1056 int
1057 udp_shutdown(struct socket *so)
1058 {
1059         struct inpcb *inp;
1060
1061         inp = sotoinpcb(so);
1062         if (inp == 0)
1063                 return EINVAL;
1064         socantsendmore(so);
1065         return 0;
1066 }
1067
1068 struct pr_usrreqs udp_usrreqs = {
1069         udp_abort, pru_accept_notsupp, udp_attach, udp_bind, udp_connect,
1070         pru_connect2_notsupp, in_control, udp_detach, udp_disconnect,
1071         pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp,
1072         pru_rcvoob_notsupp, udp_send, pru_sense_null, udp_shutdown,
1073         in_setsockaddr, sosend, soreceive, sopoll
1074 };
1075