1 /* 2 * Copyright (c) 2004 Jeffrey Hsu. All rights reserved. 3 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 4 * The Regents of the University of California. 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 * 3. All advertising materials mentioning features or use of this software 15 * must display the following acknowledgement: 16 * This product includes software developed by the University of 17 * California, Berkeley and its contributors. 18 * 4. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95 35 * $FreeBSD: src/sys/netinet/udp_usrreq.c,v 1.64.2.18 2003/01/24 05:11:34 sam Exp $ 36 * $DragonFly: src/sys/netinet/udp_usrreq.c,v 1.16 2004/03/17 02:27:59 dillon Exp $ 37 */ 38 39 #include "opt_ipsec.h" 40 #include "opt_inet6.h" 41 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/kernel.h> 45 #include <sys/malloc.h> 46 #include <sys/mbuf.h> 47 #include <sys/domain.h> 48 #include <sys/proc.h> 49 #include <sys/protosw.h> 50 #include <sys/socket.h> 51 #include <sys/socketvar.h> 52 #include <sys/sysctl.h> 53 #include <sys/syslog.h> 54 #include <sys/in_cksum.h> 55 56 #include <vm/vm_zone.h> 57 58 #include <net/if.h> 59 #include <net/route.h> 60 61 #include <netinet/in.h> 62 #include <netinet/in_systm.h> 63 #include <netinet/ip.h> 64 #ifdef INET6 65 #include <netinet/ip6.h> 66 #endif 67 #include <netinet/in_pcb.h> 68 #include <netinet/in_var.h> 69 #include <netinet/ip_var.h> 70 #ifdef INET6 71 #include <netinet6/ip6_var.h> 72 #endif 73 #include <netinet/ip_icmp.h> 74 #include <netinet/icmp_var.h> 75 #include <netinet/udp.h> 76 #include <netinet/udp_var.h> 77 78 #ifdef FAST_IPSEC 79 #include <netipsec/ipsec.h> 80 #endif /*FAST_IPSEC*/ 81 82 #ifdef IPSEC 83 #include <netinet6/ipsec.h> 84 #endif /*IPSEC*/ 85 86 /* 87 * UDP protocol implementation. 88 * Per RFC 768, August, 1980. 89 */ 90 #ifndef COMPAT_42 91 static int udpcksum = 1; 92 #else 93 static int udpcksum = 0; /* XXX */ 94 #endif 95 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW, 96 &udpcksum, 0, ""); 97 98 int log_in_vain = 0; 99 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW, 100 &log_in_vain, 0, "Log all incoming UDP packets"); 101 102 static int blackhole = 0; 103 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW, 104 &blackhole, 0, "Do not send port unreachables for refused connects"); 105 106 static int strict_mcast_mship = 1; 107 SYSCTL_INT(_net_inet_udp, OID_AUTO, strict_mcast_mship, CTLFLAG_RW, 108 &strict_mcast_mship, 0, "Only send multicast to member sockets"); 109 110 struct inpcbinfo udbinfo; 111 112 #ifndef UDBHASHSIZE 113 #define UDBHASHSIZE 16 114 #endif 115 116 struct udpstat udpstat; /* from udp_var.h */ 117 SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RW, 118 &udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)"); 119 120 static struct sockaddr_in udp_in = { sizeof(udp_in), AF_INET }; 121 #ifdef INET6 122 struct udp_in6 { 123 struct sockaddr_in6 uin6_sin; 124 u_char uin6_init_done : 1; 125 } udp_in6 = { 126 { sizeof(udp_in6.uin6_sin), AF_INET6 }, 127 0 128 }; 129 struct udp_ip6 { 130 struct ip6_hdr uip6_ip6; 131 u_char uip6_init_done : 1; 132 } udp_ip6; 133 #endif /* INET6 */ 134 135 static void udp_append (struct inpcb *last, struct ip *ip, 136 struct mbuf *n, int off); 137 #ifdef INET6 138 static void ip_2_ip6_hdr (struct ip6_hdr *ip6, struct ip *ip); 139 #endif 140 141 static int udp_detach (struct socket *so); 142 static int udp_output (struct inpcb *, struct mbuf *, struct sockaddr *, 143 struct mbuf *, struct thread *); 144 145 void 146 udp_init() 147 { 148 LIST_INIT(&udbinfo.listhead); 149 udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask); 150 udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB, 151 &udbinfo.porthashmask); 152 udbinfo.bindhashbase = hashinit(UDBHASHSIZE, M_PCB, 153 &udbinfo.bindhashmask); 154 udbinfo.ipi_zone = zinit("udpcb", sizeof(struct inpcb), maxsockets, 155 ZONE_INTERRUPT, 0); 156 udp_thread_init(); 157 } 158 159 /* 160 * Check multicast packets to make sure they are only sent to sockets with 161 * multicast memberships for the packet's destination address and arrival 162 * interface. Multicast packets to multicast-unaware sockets are also 163 * disallowed. 164 * 165 * Returns 0 if the packet is acceptable, -1 if it is not. 166 */ 167 static __inline 168 int 169 check_multicast_membership(struct ip *ip, struct inpcb *inp, struct mbuf *m) 170 { 171 int mshipno; 172 struct ip_moptions *mopt; 173 174 if (strict_mcast_mship == 0 || 175 !IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 176 return(0); 177 } 178 mopt = inp->inp_moptions; 179 if (mopt == NULL) 180 return(-1); 181 for (mshipno = 0; mshipno <= mopt->imo_num_memberships; ++mshipno) { 182 if (ip->ip_dst.s_addr == mopt->imo_membership[mshipno]->inm_addr.s_addr && 183 m->m_pkthdr.rcvif == mopt->imo_membership[mshipno]->inm_ifp) { 184 return(0); 185 } 186 } 187 return(-1); 188 } 189 190 void 191 udp_input(m, off, proto) 192 struct mbuf *m; 193 int off, proto; 194 { 195 int iphlen = off; 196 struct ip *ip; 197 struct udphdr *uh; 198 struct inpcb *inp; 199 struct mbuf *opts = 0; 200 int len; 201 struct ip save_ip; 202 struct sockaddr *append_sa; 203 204 udpstat.udps_ipackets++; 205 206 /* 207 * Strip IP options, if any; should skip this, 208 * make available to user, and use on returned packets, 209 * but we don't yet have a way to check the checksum 210 * with options still present. 211 */ 212 if (iphlen > sizeof (struct ip)) { 213 ip_stripoptions(m); 214 iphlen = sizeof(struct ip); 215 } 216 217 /* 218 * Get IP and UDP header together in first mbuf. 219 */ 220 ip = mtod(m, struct ip *); 221 if (m->m_len < iphlen + sizeof(struct udphdr)) { 222 if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == 0) { 223 udpstat.udps_hdrops++; 224 return; 225 } 226 ip = mtod(m, struct ip *); 227 } 228 uh = (struct udphdr *)((caddr_t)ip + iphlen); 229 230 /* destination port of 0 is illegal, based on RFC768. */ 231 if (uh->uh_dport == 0) 232 goto bad; 233 234 /* 235 * Make mbuf data length reflect UDP length. 236 * If not enough data to reflect UDP length, drop. 237 */ 238 len = ntohs((u_short)uh->uh_ulen); 239 if (ip->ip_len != len) { 240 if (len > ip->ip_len || len < sizeof(struct udphdr)) { 241 udpstat.udps_badlen++; 242 goto bad; 243 } 244 m_adj(m, len - ip->ip_len); 245 /* ip->ip_len = len; */ 246 } 247 /* 248 * Save a copy of the IP header in case we want restore it 249 * for sending an ICMP error message in response. 250 */ 251 save_ip = *ip; 252 253 /* 254 * Checksum extended UDP header and data. 255 */ 256 if (uh->uh_sum) { 257 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { 258 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) 259 uh->uh_sum = m->m_pkthdr.csum_data; 260 else 261 uh->uh_sum = in_pseudo(ip->ip_src.s_addr, 262 ip->ip_dst.s_addr, htonl((u_short)len + 263 m->m_pkthdr.csum_data + IPPROTO_UDP)); 264 uh->uh_sum ^= 0xffff; 265 } else { 266 char b[9]; 267 bcopy(((struct ipovly *)ip)->ih_x1, b, 9); 268 bzero(((struct ipovly *)ip)->ih_x1, 9); 269 ((struct ipovly *)ip)->ih_len = uh->uh_ulen; 270 uh->uh_sum = in_cksum(m, len + sizeof (struct ip)); 271 bcopy(b, ((struct ipovly *)ip)->ih_x1, 9); 272 } 273 if (uh->uh_sum) { 274 udpstat.udps_badsum++; 275 m_freem(m); 276 return; 277 } 278 } else 279 udpstat.udps_nosum++; 280 281 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || 282 in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) { 283 struct inpcb *last; 284 /* 285 * Deliver a multicast or broadcast datagram to *all* sockets 286 * for which the local and remote addresses and ports match 287 * those of the incoming datagram. This allows more than 288 * one process to receive multi/broadcasts on the same port. 289 * (This really ought to be done for unicast datagrams as 290 * well, but that would cause problems with existing 291 * applications that open both address-specific sockets and 292 * a wildcard socket listening to the same port -- they would 293 * end up receiving duplicates of every unicast datagram. 294 * Those applications open the multiple sockets to overcome an 295 * inadequacy of the UDP socket interface, but for backwards 296 * compatibility we avoid the problem here rather than 297 * fixing the interface. Maybe 4.5BSD will remedy this?) 298 */ 299 300 /* 301 * Construct sockaddr format source address. 302 */ 303 udp_in.sin_port = uh->uh_sport; 304 udp_in.sin_addr = ip->ip_src; 305 /* 306 * Locate pcb(s) for datagram. 307 * (Algorithm copied from raw_intr().) 308 */ 309 last = NULL; 310 #ifdef INET6 311 udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0; 312 #endif 313 LIST_FOREACH(inp, &udbinfo.listhead, inp_list) { 314 #ifdef INET6 315 if ((inp->inp_vflag & INP_IPV4) == 0) 316 continue; 317 #endif 318 if (inp->inp_lport != uh->uh_dport) 319 continue; 320 if (inp->inp_laddr.s_addr != INADDR_ANY) { 321 if (inp->inp_laddr.s_addr != 322 ip->ip_dst.s_addr) 323 continue; 324 } 325 if (inp->inp_faddr.s_addr != INADDR_ANY) { 326 if (inp->inp_faddr.s_addr != 327 ip->ip_src.s_addr || 328 inp->inp_fport != uh->uh_sport) 329 continue; 330 } 331 332 if (check_multicast_membership(ip, inp, m) < 0) 333 continue; 334 335 if (last != NULL) { 336 struct mbuf *n; 337 338 #ifdef IPSEC 339 /* check AH/ESP integrity. */ 340 if (ipsec4_in_reject_so(m, last->inp_socket)) 341 ipsecstat.in_polvio++; 342 /* do not inject data to pcb */ 343 else 344 #endif /*IPSEC*/ 345 #ifdef FAST_IPSEC 346 /* check AH/ESP integrity. */ 347 if (ipsec4_in_reject(m, last)) 348 ; 349 else 350 #endif /*FAST_IPSEC*/ 351 if ((n = m_copy(m, 0, M_COPYALL)) != NULL) 352 udp_append(last, ip, n, 353 iphlen + 354 sizeof(struct udphdr)); 355 } 356 last = inp; 357 /* 358 * Don't look for additional matches if this one does 359 * not have either the SO_REUSEPORT or SO_REUSEADDR 360 * socket options set. This heuristic avoids searching 361 * through all pcbs in the common case of a non-shared 362 * port. It * assumes that an application will never 363 * clear these options after setting them. 364 */ 365 if ((last->inp_socket->so_options&(SO_REUSEPORT|SO_REUSEADDR)) == 0) 366 break; 367 } 368 369 if (last == NULL) { 370 /* 371 * No matching pcb found; discard datagram. 372 * (No need to send an ICMP Port Unreachable 373 * for a broadcast or multicast datgram.) 374 */ 375 udpstat.udps_noportbcast++; 376 goto bad; 377 } 378 #ifdef IPSEC 379 /* check AH/ESP integrity. */ 380 if (ipsec4_in_reject_so(m, last->inp_socket)) { 381 ipsecstat.in_polvio++; 382 goto bad; 383 } 384 #endif /*IPSEC*/ 385 #ifdef FAST_IPSEC 386 /* check AH/ESP integrity. */ 387 if (ipsec4_in_reject(m, last)) 388 goto bad; 389 #endif /*FAST_IPSEC*/ 390 udp_append(last, ip, m, iphlen + sizeof(struct udphdr)); 391 return; 392 } 393 /* 394 * Locate pcb for datagram. 395 */ 396 inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport, 397 ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif); 398 if (inp == NULL) { 399 if (log_in_vain) { 400 char buf[4*sizeof "123"]; 401 402 strcpy(buf, inet_ntoa(ip->ip_dst)); 403 log(LOG_INFO, 404 "Connection attempt to UDP %s:%d from %s:%d\n", 405 buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src), 406 ntohs(uh->uh_sport)); 407 } 408 udpstat.udps_noport++; 409 if (m->m_flags & (M_BCAST | M_MCAST)) { 410 udpstat.udps_noportbcast++; 411 goto bad; 412 } 413 if (blackhole) 414 goto bad; 415 #ifdef ICMP_BANDLIM 416 if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0) 417 goto bad; 418 #endif 419 *ip = save_ip; 420 ip->ip_len += iphlen; 421 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0); 422 return; 423 } 424 #ifdef IPSEC 425 if (ipsec4_in_reject_so(m, inp->inp_socket)) { 426 ipsecstat.in_polvio++; 427 goto bad; 428 } 429 #endif /*IPSEC*/ 430 #ifdef FAST_IPSEC 431 if (ipsec4_in_reject(m, inp)) 432 goto bad; 433 #endif /*FAST_IPSEC*/ 434 435 /* 436 * Construct sockaddr format source address. 437 * Stuff source address and datagram in user buffer. 438 */ 439 udp_in.sin_port = uh->uh_sport; 440 udp_in.sin_addr = ip->ip_src; 441 if (inp->inp_flags & INP_CONTROLOPTS 442 || inp->inp_socket->so_options & SO_TIMESTAMP) { 443 #ifdef INET6 444 if (inp->inp_vflag & INP_IPV6) { 445 int savedflags; 446 447 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip); 448 savedflags = inp->inp_flags; 449 inp->inp_flags &= ~INP_UNMAPPABLEOPTS; 450 ip6_savecontrol(inp, &opts, &udp_ip6.uip6_ip6, m); 451 inp->inp_flags = savedflags; 452 } else 453 #endif 454 ip_savecontrol(inp, &opts, ip, m); 455 } 456 m_adj(m, iphlen + sizeof(struct udphdr)); 457 #ifdef INET6 458 if (inp->inp_vflag & INP_IPV6) { 459 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin); 460 append_sa = (struct sockaddr *)&udp_in6; 461 } else 462 #endif 463 append_sa = (struct sockaddr *)&udp_in; 464 if (sbappendaddr(&inp->inp_socket->so_rcv, append_sa, m, opts) == 0) { 465 udpstat.udps_fullsock++; 466 goto bad; 467 } 468 sorwakeup(inp->inp_socket); 469 return; 470 bad: 471 m_freem(m); 472 if (opts) 473 m_freem(opts); 474 return; 475 } 476 477 #ifdef INET6 478 static void 479 ip_2_ip6_hdr(ip6, ip) 480 struct ip6_hdr *ip6; 481 struct ip *ip; 482 { 483 bzero(ip6, sizeof(*ip6)); 484 485 ip6->ip6_vfc = IPV6_VERSION; 486 ip6->ip6_plen = ip->ip_len; 487 ip6->ip6_nxt = ip->ip_p; 488 ip6->ip6_hlim = ip->ip_ttl; 489 ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] = 490 IPV6_ADDR_INT32_SMP; 491 ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr; 492 ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr; 493 } 494 #endif 495 496 /* 497 * subroutine of udp_input(), mainly for source code readability. 498 * caller must properly init udp_ip6 and udp_in6 beforehand. 499 */ 500 static void 501 udp_append(last, ip, n, off) 502 struct inpcb *last; 503 struct ip *ip; 504 struct mbuf *n; 505 int off; 506 { 507 struct sockaddr *append_sa; 508 struct mbuf *opts = 0; 509 510 if (last->inp_flags & INP_CONTROLOPTS || 511 last->inp_socket->so_options & SO_TIMESTAMP) { 512 #ifdef INET6 513 if (last->inp_vflag & INP_IPV6) { 514 int savedflags; 515 516 if (udp_ip6.uip6_init_done == 0) { 517 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip); 518 udp_ip6.uip6_init_done = 1; 519 } 520 savedflags = last->inp_flags; 521 last->inp_flags &= ~INP_UNMAPPABLEOPTS; 522 ip6_savecontrol(last, &opts, &udp_ip6.uip6_ip6, n); 523 last->inp_flags = savedflags; 524 } else 525 #endif 526 ip_savecontrol(last, &opts, ip, n); 527 } 528 #ifdef INET6 529 if (last->inp_vflag & INP_IPV6) { 530 if (udp_in6.uin6_init_done == 0) { 531 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin); 532 udp_in6.uin6_init_done = 1; 533 } 534 append_sa = (struct sockaddr *)&udp_in6.uin6_sin; 535 } else 536 #endif 537 append_sa = (struct sockaddr *)&udp_in; 538 m_adj(n, off); 539 if (sbappendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) { 540 m_freem(n); 541 if (opts) 542 m_freem(opts); 543 udpstat.udps_fullsock++; 544 } else 545 sorwakeup(last->inp_socket); 546 } 547 548 /* 549 * Notify a udp user of an asynchronous error; 550 * just wake up so that he can collect error status. 551 */ 552 void 553 udp_notify(inp, errno) 554 struct inpcb *inp; 555 int errno; 556 { 557 inp->inp_socket->so_error = errno; 558 sorwakeup(inp->inp_socket); 559 sowwakeup(inp->inp_socket); 560 } 561 562 void 563 udp_ctlinput(cmd, sa, vip) 564 int cmd; 565 struct sockaddr *sa; 566 void *vip; 567 { 568 struct ip *ip = vip; 569 struct udphdr *uh; 570 void (*notify) (struct inpcb *, int) = udp_notify; 571 struct in_addr faddr; 572 struct inpcb *inp; 573 int s; 574 575 faddr = ((struct sockaddr_in *)sa)->sin_addr; 576 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) 577 return; 578 579 if (PRC_IS_REDIRECT(cmd)) { 580 ip = 0; 581 notify = in_rtchange; 582 } else if (cmd == PRC_HOSTDEAD) 583 ip = 0; 584 else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) 585 return; 586 if (ip) { 587 s = splnet(); 588 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 589 inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport, 590 ip->ip_src, uh->uh_sport, 0, NULL); 591 if (inp != NULL && inp->inp_socket != NULL) 592 (*notify)(inp, inetctlerrmap[cmd]); 593 splx(s); 594 } else 595 in_pcbnotifyall(&udbinfo.listhead, faddr, inetctlerrmap[cmd], 596 notify); 597 } 598 599 static int 600 udp_pcblist(SYSCTL_HANDLER_ARGS) 601 { 602 int error, i, n, s; 603 struct inpcb *inp, **inp_list; 604 inp_gen_t gencnt; 605 struct xinpgen xig; 606 607 /* 608 * The process of preparing the TCB list is too time-consuming and 609 * resource-intensive to repeat twice on every request. 610 */ 611 if (req->oldptr == 0) { 612 n = udbinfo.ipi_count; 613 req->oldidx = 2 * (sizeof xig) 614 + (n + n/8) * sizeof(struct xinpcb); 615 return 0; 616 } 617 618 if (req->newptr != 0) 619 return EPERM; 620 621 /* 622 * OK, now we're committed to doing something. 623 */ 624 s = splnet(); 625 gencnt = udbinfo.ipi_gencnt; 626 n = udbinfo.ipi_count; 627 splx(s); 628 629 xig.xig_len = sizeof xig; 630 xig.xig_count = n; 631 xig.xig_gen = gencnt; 632 xig.xig_sogen = so_gencnt; 633 error = SYSCTL_OUT(req, &xig, sizeof xig); 634 if (error) 635 return error; 636 637 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 638 if (inp_list == 0) 639 return ENOMEM; 640 641 s = splnet(); 642 for (inp = LIST_FIRST(&udbinfo.listhead), i = 0; inp && i < n; 643 inp = LIST_NEXT(inp, inp_list)) { 644 if (inp->inp_gencnt <= gencnt && !prison_xinpcb(req->td, inp)) 645 inp_list[i++] = inp; 646 } 647 splx(s); 648 n = i; 649 650 error = 0; 651 for (i = 0; i < n; i++) { 652 inp = inp_list[i]; 653 if (inp->inp_gencnt <= gencnt) { 654 struct xinpcb xi; 655 xi.xi_len = sizeof xi; 656 /* XXX should avoid extra copy */ 657 bcopy(inp, &xi.xi_inp, sizeof *inp); 658 if (inp->inp_socket) 659 sotoxsocket(inp->inp_socket, &xi.xi_socket); 660 error = SYSCTL_OUT(req, &xi, sizeof xi); 661 } 662 } 663 if (!error) { 664 /* 665 * Give the user an updated idea of our state. 666 * If the generation differs from what we told 667 * her before, she knows that something happened 668 * while we were processing this request, and it 669 * might be necessary to retry. 670 */ 671 s = splnet(); 672 xig.xig_gen = udbinfo.ipi_gencnt; 673 xig.xig_sogen = so_gencnt; 674 xig.xig_count = udbinfo.ipi_count; 675 splx(s); 676 error = SYSCTL_OUT(req, &xig, sizeof xig); 677 } 678 free(inp_list, M_TEMP); 679 return error; 680 } 681 682 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0, 683 udp_pcblist, "S,xinpcb", "List of active UDP sockets"); 684 685 static int 686 udp_getcred(SYSCTL_HANDLER_ARGS) 687 { 688 struct sockaddr_in addrs[2]; 689 struct inpcb *inp; 690 int error, s; 691 692 error = suser(req->td); 693 if (error) 694 return (error); 695 error = SYSCTL_IN(req, addrs, sizeof(addrs)); 696 if (error) 697 return (error); 698 s = splnet(); 699 inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port, 700 addrs[0].sin_addr, addrs[0].sin_port, 1, NULL); 701 if (inp == NULL || inp->inp_socket == NULL) { 702 error = ENOENT; 703 goto out; 704 } 705 error = SYSCTL_OUT(req, inp->inp_socket->so_cred, sizeof(struct ucred)); 706 out: 707 splx(s); 708 return (error); 709 } 710 711 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 712 0, 0, udp_getcred, "S,ucred", "Get the ucred of a UDP connection"); 713 714 static int 715 udp_output(inp, m, dstaddr, control, td) 716 struct inpcb *inp; 717 struct mbuf *m; 718 struct sockaddr *dstaddr; 719 struct mbuf *control; 720 struct thread *td; 721 { 722 struct udpiphdr *ui; 723 int len = m->m_pkthdr.len; 724 struct sockaddr_in *sin; /* really is initialized before use */ 725 int error = 0; 726 727 if (control) 728 m_freem(control); /* XXX */ 729 730 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) { 731 error = EMSGSIZE; 732 goto release; 733 } 734 735 if (inp->inp_lport == 0) { /* unbound socket */ 736 error = in_pcbbind(inp, (struct sockaddr *)NULL, td); 737 if (error) 738 goto release; 739 } 740 741 if (dstaddr != NULL) { /* destination address specified */ 742 if (inp->inp_faddr.s_addr != INADDR_ANY) { 743 /* already connected */ 744 error = EISCONN; 745 goto release; 746 } 747 sin = (struct sockaddr_in *)dstaddr; 748 prison_remote_ip(td, 0, &sin->sin_addr.s_addr); 749 } else { 750 if (inp->inp_faddr.s_addr == INADDR_ANY) { 751 /* no destination specified and not already connected */ 752 error = ENOTCONN; 753 goto release; 754 } 755 sin = NULL; 756 } 757 758 /* 759 * Calculate data length and get a mbuf 760 * for UDP and IP headers. 761 */ 762 M_PREPEND(m, sizeof(struct udpiphdr), M_DONTWAIT); 763 if (m == 0) { 764 error = ENOBUFS; 765 goto release; 766 } 767 768 /* 769 * Fill in mbuf with extended UDP header 770 * and addresses and length put into network format. 771 */ 772 ui = mtod(m, struct udpiphdr *); 773 bzero(ui->ui_x1, sizeof(ui->ui_x1)); /* XXX still needed? */ 774 ui->ui_pr = IPPROTO_UDP; 775 776 /* 777 * Set destination address. 778 */ 779 if (dstaddr != NULL) { /* use specified destination */ 780 ui->ui_dst = sin->sin_addr; 781 ui->ui_dport = sin->sin_port; 782 } else { /* use connected destination */ 783 ui->ui_dst = inp->inp_faddr; 784 ui->ui_dport = inp->inp_fport; 785 } 786 787 /* 788 * Set source address. If the source address is INADDR_ANY we 789 * have to lookup the outgoing interface based on the target address 790 * and assign the source address from that. 791 * 792 * If dstaddr is NULL the socket is connected and we have to use the 793 * connected target (in_faddr) as the target address, otherwise we 794 * just use dstaddr. 795 */ 796 if (inp->inp_laddr.s_addr == INADDR_ANY) { /* need to pick an address */ 797 struct sockaddr_in *if_sin; 798 struct sockaddr_in sin_tmp; 799 800 if (dstaddr == NULL) { 801 sin_tmp.sin_len = sizeof(sin_tmp); 802 sin_tmp.sin_family = AF_INET; 803 sin_tmp.sin_port = inp->inp_fport; 804 sin_tmp.sin_addr = inp->inp_faddr; 805 error = in_pcbladdr(inp, (void *)&sin_tmp, &if_sin); 806 } else { 807 error = in_pcbladdr(inp, dstaddr, &if_sin); 808 } 809 if (error) 810 goto release; 811 ui->ui_src = if_sin->sin_addr; 812 } else { /* use bound non-null address */ 813 ui->ui_src = inp->inp_laddr; 814 } 815 ui->ui_sport = inp->inp_lport; 816 KASSERT(inp->inp_lport != 0, ("inp lport should have been bound")); 817 818 ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr)); 819 820 /* 821 * Set up checksum and output datagram. 822 */ 823 if (udpcksum) { 824 ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr, 825 htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP)); 826 m->m_pkthdr.csum_flags = CSUM_UDP; 827 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); 828 } else { 829 ui->ui_sum = 0; 830 } 831 ((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len; 832 ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */ 833 ((struct ip *)ui)->ip_tos = inp->inp_ip_tos; /* XXX */ 834 udpstat.udps_opackets++; 835 836 error = ip_output(m, inp->inp_options, &inp->inp_route, 837 (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)), 838 inp->inp_moptions, inp); 839 840 return (error); 841 842 release: 843 m_freem(m); 844 return (error); 845 } 846 847 u_long udp_sendspace = 9216; /* really max datagram size */ 848 /* 40 1K datagrams */ 849 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW, 850 &udp_sendspace, 0, "Maximum outgoing UDP datagram size"); 851 852 u_long udp_recvspace = 40 * (1024 + 853 #ifdef INET6 854 sizeof(struct sockaddr_in6) 855 #else 856 sizeof(struct sockaddr_in) 857 #endif 858 ); 859 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW, 860 &udp_recvspace, 0, "Maximum incoming UDP datagram size"); 861 862 static int 863 udp_abort(struct socket *so) 864 { 865 struct inpcb *inp; 866 int s; 867 868 inp = sotoinpcb(so); 869 if (inp == 0) 870 return EINVAL; /* ??? possible? panic instead? */ 871 soisdisconnected(so); 872 s = splnet(); 873 in_pcbdetach(inp); 874 splx(s); 875 return 0; 876 } 877 878 static int 879 udp_attach(struct socket *so, int proto, struct pru_attach_info *ai) 880 { 881 struct inpcb *inp; 882 int s, error; 883 884 inp = sotoinpcb(so); 885 if (inp != 0) 886 return EINVAL; 887 888 error = soreserve(so, udp_sendspace, udp_recvspace, ai->sb_rlimit); 889 if (error) 890 return error; 891 s = splnet(); 892 error = in_pcballoc(so, &udbinfo); 893 splx(s); 894 if (error) 895 return error; 896 897 inp = (struct inpcb *)so->so_pcb; 898 inp->inp_vflag |= INP_IPV4; 899 inp->inp_ip_ttl = ip_defttl; 900 return 0; 901 } 902 903 static int 904 udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 905 { 906 struct inpcb *inp; 907 int s, error; 908 909 inp = sotoinpcb(so); 910 if (inp == 0) 911 return EINVAL; 912 s = splnet(); 913 error = in_pcbbind(inp, nam, td); 914 splx(s); 915 return error; 916 } 917 918 static int 919 udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 920 { 921 struct inpcb *inp; 922 int s, error; 923 struct sockaddr_in *sin; 924 925 inp = sotoinpcb(so); 926 if (inp == 0) 927 return EINVAL; 928 if (inp->inp_faddr.s_addr != INADDR_ANY) 929 return EISCONN; 930 error = 0; 931 s = splnet(); 932 if (inp->inp_laddr.s_addr == INADDR_ANY && 933 td->td_proc && 934 td->td_proc->p_ucred->cr_prison != NULL) { 935 error = in_pcbbind(inp, NULL, td); 936 } 937 if (error == 0) { 938 sin = (struct sockaddr_in *)nam; 939 prison_remote_ip(td, 0, &sin->sin_addr.s_addr); 940 error = in_pcbconnect(inp, nam, td); 941 } 942 splx(s); 943 if (error == 0) 944 soisconnected(so); 945 return error; 946 } 947 948 static int 949 udp_detach(struct socket *so) 950 { 951 struct inpcb *inp; 952 int s; 953 954 inp = sotoinpcb(so); 955 if (inp == 0) 956 return EINVAL; 957 s = splnet(); 958 in_pcbdetach(inp); 959 splx(s); 960 return 0; 961 } 962 963 static int 964 udp_disconnect(struct socket *so) 965 { 966 struct inpcb *inp; 967 int s; 968 969 inp = sotoinpcb(so); 970 if (inp == 0) 971 return EINVAL; 972 if (inp->inp_faddr.s_addr == INADDR_ANY) 973 return ENOTCONN; 974 975 s = splnet(); 976 in_pcbdisconnect(inp); 977 inp->inp_laddr.s_addr = INADDR_ANY; 978 in_pcbinsbindhash(inp); 979 splx(s); 980 so->so_state &= ~SS_ISCONNECTED; /* XXX */ 981 return 0; 982 } 983 984 static int 985 udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, 986 struct mbuf *control, struct thread *td) 987 { 988 struct inpcb *inp; 989 990 inp = sotoinpcb(so); 991 if (inp == 0) { 992 m_freem(m); 993 return EINVAL; 994 } 995 return udp_output(inp, m, addr, control, td); 996 } 997 998 int 999 udp_shutdown(struct socket *so) 1000 { 1001 struct inpcb *inp; 1002 1003 inp = sotoinpcb(so); 1004 if (inp == 0) 1005 return EINVAL; 1006 socantsendmore(so); 1007 return 0; 1008 } 1009 1010 struct pr_usrreqs udp_usrreqs = { 1011 udp_abort, pru_accept_notsupp, udp_attach, udp_bind, udp_connect, 1012 pru_connect2_notsupp, in_control, udp_detach, udp_disconnect, 1013 pru_listen_notsupp, in_setpeeraddr, pru_rcvd_notsupp, 1014 pru_rcvoob_notsupp, udp_send, pru_sense_null, udp_shutdown, 1015 in_setsockaddr, sosend, soreceive, sopoll 1016 }; 1017 1018