1 /*- 2 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95 30 * $FreeBSD$ 31 */ 32 33 #include "opt_ipsec.h" 34 #include "opt_inet6.h" 35 #include "opt_mac.h" 36 37 #include <sys/param.h> 38 #include <sys/systm.h> 39 #include <sys/domain.h> 40 #include <sys/jail.h> 41 #include <sys/kernel.h> 42 #include <sys/lock.h> 43 #include <sys/mac.h> 44 #include <sys/malloc.h> 45 #include <sys/mbuf.h> 46 #include <sys/proc.h> 47 #include <sys/protosw.h> 48 #include <sys/signalvar.h> 49 #include <sys/socket.h> 50 #include <sys/socketvar.h> 51 #include <sys/sx.h> 52 #include <sys/sysctl.h> 53 #include <sys/syslog.h> 54 55 #include <vm/uma.h> 56 57 #include <net/if.h> 58 #include <net/route.h> 59 60 #include <netinet/in.h> 61 #include <netinet/in_systm.h> 62 #include <netinet/in_pcb.h> 63 #include <netinet/in_var.h> 64 #include <netinet/ip.h> 65 #ifdef INET6 66 #include <netinet/ip6.h> 67 #endif 68 #include <netinet/ip_icmp.h> 69 #include <netinet/icmp_var.h> 70 #include <netinet/ip_var.h> 71 #ifdef INET6 72 #include <netinet6/ip6_var.h> 73 #endif 74 #include <netinet/udp.h> 75 #include <netinet/udp_var.h> 76 77 #ifdef FAST_IPSEC 78 #include <netipsec/ipsec.h> 79 #endif /*FAST_IPSEC*/ 80 81 #ifdef IPSEC 82 #include <netinet6/ipsec.h> 83 #endif /*IPSEC*/ 84 85 #include <machine/in_cksum.h> 86 87 /* 88 * UDP protocol implementation. 89 * Per RFC 768, August, 1980. 90 */ 91 #ifndef COMPAT_42 92 static int udpcksum = 1; 93 #else 94 static int udpcksum = 0; /* XXX */ 95 #endif 96 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW, 97 &udpcksum, 0, ""); 98 99 int log_in_vain = 0; 100 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW, 101 &log_in_vain, 0, "Log all incoming UDP packets"); 102 103 static int blackhole = 0; 104 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW, 105 &blackhole, 0, "Do not send port unreachables for refused connects"); 106 107 static int strict_mcast_mship = 0; 108 SYSCTL_INT(_net_inet_udp, OID_AUTO, strict_mcast_mship, CTLFLAG_RW, 109 &strict_mcast_mship, 0, "Only send multicast to member sockets"); 110 111 struct inpcbhead udb; /* from udp_var.h */ 112 #define udb6 udb /* for KAME src sync over BSD*'s */ 113 struct inpcbinfo udbinfo; 114 115 #ifndef UDBHASHSIZE 116 #define UDBHASHSIZE 16 117 #endif 118 119 struct udpstat udpstat; /* from udp_var.h */ 120 SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RW, 121 &udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)"); 122 123 static void udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, 124 int off, struct sockaddr_in *udp_in); 125 126 static int udp_detach(struct socket *so); 127 static int udp_output(struct inpcb *, struct mbuf *, struct sockaddr *, 128 struct mbuf *, struct thread *); 129 130 void 131 udp_init() 132 { 133 INP_INFO_LOCK_INIT(&udbinfo, "udp"); 134 LIST_INIT(&udb); 135 udbinfo.listhead = &udb; 136 udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask); 137 udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB, 138 &udbinfo.porthashmask); 139 udbinfo.ipi_zone = uma_zcreate("udpcb", sizeof(struct inpcb), NULL, 140 NULL, NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE); 141 uma_zone_set_max(udbinfo.ipi_zone, maxsockets); 142 } 143 144 void 145 udp_input(m, off) 146 register struct mbuf *m; 147 int off; 148 { 149 int iphlen = off; 150 register struct ip *ip; 151 register struct udphdr *uh; 152 register struct inpcb *inp; 153 struct mbuf *opts = 0; 154 int len; 155 struct ip save_ip; 156 struct sockaddr_in udp_in; 157 158 udpstat.udps_ipackets++; 159 160 /* 161 * Strip IP options, if any; should skip this, 162 * make available to user, and use on returned packets, 163 * but we don't yet have a way to check the checksum 164 * with options still present. 165 */ 166 if (iphlen > sizeof (struct ip)) { 167 ip_stripoptions(m, (struct mbuf *)0); 168 iphlen = sizeof(struct ip); 169 } 170 171 /* 172 * Get IP and UDP header together in first mbuf. 173 */ 174 ip = mtod(m, struct ip *); 175 if (m->m_len < iphlen + sizeof(struct udphdr)) { 176 if ((m = m_pullup(m, iphlen + sizeof(struct udphdr))) == 0) { 177 udpstat.udps_hdrops++; 178 return; 179 } 180 ip = mtod(m, struct ip *); 181 } 182 uh = (struct udphdr *)((caddr_t)ip + iphlen); 183 184 /* destination port of 0 is illegal, based on RFC768. */ 185 if (uh->uh_dport == 0) 186 goto badunlocked; 187 188 /* 189 * Construct sockaddr format source address. 190 * Stuff source address and datagram in user buffer. 191 */ 192 bzero(&udp_in, sizeof(udp_in)); 193 udp_in.sin_len = sizeof(udp_in); 194 udp_in.sin_family = AF_INET; 195 udp_in.sin_port = uh->uh_sport; 196 udp_in.sin_addr = ip->ip_src; 197 198 /* 199 * Make mbuf data length reflect UDP length. 200 * If not enough data to reflect UDP length, drop. 201 */ 202 len = ntohs((u_short)uh->uh_ulen); 203 if (ip->ip_len != len) { 204 if (len > ip->ip_len || len < sizeof(struct udphdr)) { 205 udpstat.udps_badlen++; 206 goto badunlocked; 207 } 208 m_adj(m, len - ip->ip_len); 209 /* ip->ip_len = len; */ 210 } 211 /* 212 * Save a copy of the IP header in case we want restore it 213 * for sending an ICMP error message in response. 214 */ 215 if (!blackhole) 216 save_ip = *ip; 217 218 /* 219 * Checksum extended UDP header and data. 220 */ 221 if (uh->uh_sum) { 222 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { 223 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) 224 uh->uh_sum = m->m_pkthdr.csum_data; 225 else 226 uh->uh_sum = in_pseudo(ip->ip_src.s_addr, 227 ip->ip_dst.s_addr, htonl((u_short)len + 228 m->m_pkthdr.csum_data + IPPROTO_UDP)); 229 uh->uh_sum ^= 0xffff; 230 } else { 231 char b[9]; 232 bcopy(((struct ipovly *)ip)->ih_x1, b, 9); 233 bzero(((struct ipovly *)ip)->ih_x1, 9); 234 ((struct ipovly *)ip)->ih_len = uh->uh_ulen; 235 uh->uh_sum = in_cksum(m, len + sizeof (struct ip)); 236 bcopy(b, ((struct ipovly *)ip)->ih_x1, 9); 237 } 238 if (uh->uh_sum) { 239 udpstat.udps_badsum++; 240 m_freem(m); 241 return; 242 } 243 } else 244 udpstat.udps_nosum++; 245 246 INP_INFO_RLOCK(&udbinfo); 247 248 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || 249 in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) { 250 struct inpcb *last; 251 /* 252 * Deliver a multicast or broadcast datagram to *all* sockets 253 * for which the local and remote addresses and ports match 254 * those of the incoming datagram. This allows more than 255 * one process to receive multi/broadcasts on the same port. 256 * (This really ought to be done for unicast datagrams as 257 * well, but that would cause problems with existing 258 * applications that open both address-specific sockets and 259 * a wildcard socket listening to the same port -- they would 260 * end up receiving duplicates of every unicast datagram. 261 * Those applications open the multiple sockets to overcome an 262 * inadequacy of the UDP socket interface, but for backwards 263 * compatibility we avoid the problem here rather than 264 * fixing the interface. Maybe 4.5BSD will remedy this?) 265 */ 266 267 /* 268 * Locate pcb(s) for datagram. 269 * (Algorithm copied from raw_intr().) 270 */ 271 last = NULL; 272 LIST_FOREACH(inp, &udb, inp_list) { 273 if (inp->inp_lport != uh->uh_dport) 274 continue; 275 #ifdef INET6 276 if ((inp->inp_vflag & INP_IPV4) == 0) 277 continue; 278 #endif 279 if (inp->inp_laddr.s_addr != INADDR_ANY) { 280 if (inp->inp_laddr.s_addr != ip->ip_dst.s_addr) 281 continue; 282 } 283 if (inp->inp_faddr.s_addr != INADDR_ANY) { 284 if (inp->inp_faddr.s_addr != 285 ip->ip_src.s_addr || 286 inp->inp_fport != uh->uh_sport) 287 continue; 288 } 289 INP_LOCK(inp); 290 291 /* 292 * Check multicast packets to make sure they are only 293 * sent to sockets with multicast memberships for the 294 * packet's destination address and arrival interface 295 */ 296 #define MSHIP(_inp, n) ((_inp)->inp_moptions->imo_membership[(n)]) 297 #define NMSHIPS(_inp) ((_inp)->inp_moptions->imo_num_memberships) 298 if (strict_mcast_mship && inp->inp_moptions != NULL) { 299 int mship, foundmship = 0; 300 301 for (mship = 0; mship < NMSHIPS(inp); mship++) { 302 if (MSHIP(inp, mship)->inm_addr.s_addr 303 == ip->ip_dst.s_addr && 304 MSHIP(inp, mship)->inm_ifp 305 == m->m_pkthdr.rcvif) { 306 foundmship = 1; 307 break; 308 } 309 } 310 if (foundmship == 0) { 311 INP_UNLOCK(inp); 312 continue; 313 } 314 } 315 #undef NMSHIPS 316 #undef MSHIP 317 if (last != NULL) { 318 struct mbuf *n; 319 320 n = m_copy(m, 0, M_COPYALL); 321 if (n != NULL) 322 udp_append(last, ip, n, 323 iphlen + 324 sizeof(struct udphdr), 325 &udp_in); 326 INP_UNLOCK(last); 327 } 328 last = inp; 329 /* 330 * Don't look for additional matches if this one does 331 * not have either the SO_REUSEPORT or SO_REUSEADDR 332 * socket options set. This heuristic avoids searching 333 * through all pcbs in the common case of a non-shared 334 * port. It * assumes that an application will never 335 * clear these options after setting them. 336 */ 337 if ((last->inp_socket->so_options&(SO_REUSEPORT|SO_REUSEADDR)) == 0) 338 break; 339 } 340 341 if (last == NULL) { 342 /* 343 * No matching pcb found; discard datagram. 344 * (No need to send an ICMP Port Unreachable 345 * for a broadcast or multicast datgram.) 346 */ 347 udpstat.udps_noportbcast++; 348 goto badheadlocked; 349 } 350 udp_append(last, ip, m, iphlen + sizeof(struct udphdr), 351 &udp_in); 352 INP_UNLOCK(last); 353 INP_INFO_RUNLOCK(&udbinfo); 354 return; 355 } 356 /* 357 * Locate pcb for datagram. 358 */ 359 inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport, 360 ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif); 361 if (inp == NULL) { 362 if (log_in_vain) { 363 char buf[4*sizeof "123"]; 364 365 strcpy(buf, inet_ntoa(ip->ip_dst)); 366 log(LOG_INFO, 367 "Connection attempt to UDP %s:%d from %s:%d\n", 368 buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src), 369 ntohs(uh->uh_sport)); 370 } 371 udpstat.udps_noport++; 372 if (m->m_flags & (M_BCAST | M_MCAST)) { 373 udpstat.udps_noportbcast++; 374 goto badheadlocked; 375 } 376 if (blackhole) 377 goto badheadlocked; 378 if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0) 379 goto badheadlocked; 380 *ip = save_ip; 381 ip->ip_len += iphlen; 382 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0); 383 INP_INFO_RUNLOCK(&udbinfo); 384 return; 385 } 386 INP_LOCK(inp); 387 udp_append(inp, ip, m, iphlen + sizeof(struct udphdr), &udp_in); 388 INP_UNLOCK(inp); 389 INP_INFO_RUNLOCK(&udbinfo); 390 return; 391 392 badheadlocked: 393 if (inp) 394 INP_UNLOCK(inp); 395 INP_INFO_RUNLOCK(&udbinfo); 396 badunlocked: 397 m_freem(m); 398 if (opts) 399 m_freem(opts); 400 return; 401 } 402 403 /* 404 * Subroutine of udp_input(), which appends the provided mbuf chain to the 405 * passed pcb/socket. The caller must provide a sockaddr_in via udp_in that 406 * contains the source address. If the socket ends up being an IPv6 socket, 407 * udp_append() will convert to a sockaddr_in6 before passing the address 408 * into the socket code. 409 */ 410 static void 411 udp_append(last, ip, n, off, udp_in) 412 struct inpcb *last; 413 struct ip *ip; 414 struct mbuf *n; 415 int off; 416 struct sockaddr_in *udp_in; 417 { 418 struct sockaddr *append_sa; 419 struct socket *so; 420 struct mbuf *opts = 0; 421 #ifdef INET6 422 struct sockaddr_in6 udp_in6; 423 #endif 424 425 INP_LOCK_ASSERT(last); 426 427 #if defined(IPSEC) || defined(FAST_IPSEC) 428 /* check AH/ESP integrity. */ 429 if (ipsec4_in_reject(n, last)) { 430 #ifdef IPSEC 431 ipsecstat.in_polvio++; 432 #endif /*IPSEC*/ 433 m_freem(n); 434 return; 435 } 436 #endif /*IPSEC || FAST_IPSEC*/ 437 #ifdef MAC 438 if (mac_check_inpcb_deliver(last, n) != 0) { 439 m_freem(n); 440 return; 441 } 442 #endif 443 if (last->inp_flags & INP_CONTROLOPTS || 444 last->inp_socket->so_options & (SO_TIMESTAMP | SO_BINTIME)) { 445 #ifdef INET6 446 if (last->inp_vflag & INP_IPV6) { 447 int savedflags; 448 449 savedflags = last->inp_flags; 450 last->inp_flags &= ~INP_UNMAPPABLEOPTS; 451 ip6_savecontrol(last, n, &opts); 452 last->inp_flags = savedflags; 453 } else 454 #endif 455 ip_savecontrol(last, &opts, ip, n); 456 } 457 #ifdef INET6 458 if (last->inp_vflag & INP_IPV6) { 459 bzero(&udp_in6, sizeof(udp_in6)); 460 udp_in6.sin6_len = sizeof(udp_in6); 461 udp_in6.sin6_family = AF_INET6; 462 in6_sin_2_v4mapsin6(udp_in, &udp_in6); 463 append_sa = (struct sockaddr *)&udp_in6; 464 } else 465 #endif 466 append_sa = (struct sockaddr *)udp_in; 467 m_adj(n, off); 468 469 so = last->inp_socket; 470 SOCKBUF_LOCK(&so->so_rcv); 471 if (sbappendaddr_locked(&so->so_rcv, append_sa, n, opts) == 0) { 472 m_freem(n); 473 if (opts) 474 m_freem(opts); 475 udpstat.udps_fullsock++; 476 SOCKBUF_UNLOCK(&so->so_rcv); 477 } else 478 sorwakeup_locked(so); 479 } 480 481 /* 482 * Notify a udp user of an asynchronous error; 483 * just wake up so that he can collect error status. 484 */ 485 struct inpcb * 486 udp_notify(inp, errno) 487 register struct inpcb *inp; 488 int errno; 489 { 490 inp->inp_socket->so_error = errno; 491 sorwakeup(inp->inp_socket); 492 sowwakeup(inp->inp_socket); 493 return inp; 494 } 495 496 void 497 udp_ctlinput(cmd, sa, vip) 498 int cmd; 499 struct sockaddr *sa; 500 void *vip; 501 { 502 struct ip *ip = vip; 503 struct udphdr *uh; 504 struct inpcb *(*notify)(struct inpcb *, int) = udp_notify; 505 struct in_addr faddr; 506 struct inpcb *inp; 507 int s; 508 509 faddr = ((struct sockaddr_in *)sa)->sin_addr; 510 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) 511 return; 512 513 /* 514 * Redirects don't need to be handled up here. 515 */ 516 if (PRC_IS_REDIRECT(cmd)) 517 return; 518 /* 519 * Hostdead is ugly because it goes linearly through all PCBs. 520 * XXX: We never get this from ICMP, otherwise it makes an 521 * excellent DoS attack on machines with many connections. 522 */ 523 if (cmd == PRC_HOSTDEAD) 524 ip = 0; 525 else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) 526 return; 527 if (ip) { 528 s = splnet(); 529 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 530 INP_INFO_RLOCK(&udbinfo); 531 inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport, 532 ip->ip_src, uh->uh_sport, 0, NULL); 533 if (inp != NULL) { 534 INP_LOCK(inp); 535 if (inp->inp_socket != NULL) { 536 (*notify)(inp, inetctlerrmap[cmd]); 537 } 538 INP_UNLOCK(inp); 539 } 540 INP_INFO_RUNLOCK(&udbinfo); 541 splx(s); 542 } else 543 in_pcbnotifyall(&udbinfo, faddr, inetctlerrmap[cmd], notify); 544 } 545 546 static int 547 udp_pcblist(SYSCTL_HANDLER_ARGS) 548 { 549 int error, i, n, s; 550 struct inpcb *inp, **inp_list; 551 inp_gen_t gencnt; 552 struct xinpgen xig; 553 554 /* 555 * The process of preparing the TCB list is too time-consuming and 556 * resource-intensive to repeat twice on every request. 557 */ 558 if (req->oldptr == 0) { 559 n = udbinfo.ipi_count; 560 req->oldidx = 2 * (sizeof xig) 561 + (n + n/8) * sizeof(struct xinpcb); 562 return 0; 563 } 564 565 if (req->newptr != 0) 566 return EPERM; 567 568 /* 569 * OK, now we're committed to doing something. 570 */ 571 s = splnet(); 572 INP_INFO_RLOCK(&udbinfo); 573 gencnt = udbinfo.ipi_gencnt; 574 n = udbinfo.ipi_count; 575 INP_INFO_RUNLOCK(&udbinfo); 576 splx(s); 577 578 error = sysctl_wire_old_buffer(req, 2 * (sizeof xig) 579 + n * sizeof(struct xinpcb)); 580 if (error != 0) 581 return (error); 582 583 xig.xig_len = sizeof xig; 584 xig.xig_count = n; 585 xig.xig_gen = gencnt; 586 xig.xig_sogen = so_gencnt; 587 error = SYSCTL_OUT(req, &xig, sizeof xig); 588 if (error) 589 return error; 590 591 inp_list = malloc(n * sizeof *inp_list, M_TEMP, M_WAITOK); 592 if (inp_list == 0) 593 return ENOMEM; 594 595 s = splnet(); 596 INP_INFO_RLOCK(&udbinfo); 597 for (inp = LIST_FIRST(udbinfo.listhead), i = 0; inp && i < n; 598 inp = LIST_NEXT(inp, inp_list)) { 599 INP_LOCK(inp); 600 if (inp->inp_gencnt <= gencnt && 601 cr_canseesocket(req->td->td_ucred, inp->inp_socket) == 0) 602 inp_list[i++] = inp; 603 INP_UNLOCK(inp); 604 } 605 INP_INFO_RUNLOCK(&udbinfo); 606 splx(s); 607 n = i; 608 609 error = 0; 610 for (i = 0; i < n; i++) { 611 inp = inp_list[i]; 612 if (inp->inp_gencnt <= gencnt) { 613 struct xinpcb xi; 614 bzero(&xi, sizeof(xi)); 615 xi.xi_len = sizeof xi; 616 /* XXX should avoid extra copy */ 617 bcopy(inp, &xi.xi_inp, sizeof *inp); 618 if (inp->inp_socket) 619 sotoxsocket(inp->inp_socket, &xi.xi_socket); 620 xi.xi_inp.inp_gencnt = inp->inp_gencnt; 621 error = SYSCTL_OUT(req, &xi, sizeof xi); 622 } 623 } 624 if (!error) { 625 /* 626 * Give the user an updated idea of our state. 627 * If the generation differs from what we told 628 * her before, she knows that something happened 629 * while we were processing this request, and it 630 * might be necessary to retry. 631 */ 632 s = splnet(); 633 INP_INFO_RLOCK(&udbinfo); 634 xig.xig_gen = udbinfo.ipi_gencnt; 635 xig.xig_sogen = so_gencnt; 636 xig.xig_count = udbinfo.ipi_count; 637 INP_INFO_RUNLOCK(&udbinfo); 638 splx(s); 639 error = SYSCTL_OUT(req, &xig, sizeof xig); 640 } 641 free(inp_list, M_TEMP); 642 return error; 643 } 644 645 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, 0, 0, 646 udp_pcblist, "S,xinpcb", "List of active UDP sockets"); 647 648 static int 649 udp_getcred(SYSCTL_HANDLER_ARGS) 650 { 651 struct xucred xuc; 652 struct sockaddr_in addrs[2]; 653 struct inpcb *inp; 654 int error, s; 655 656 error = suser_cred(req->td->td_ucred, SUSER_ALLOWJAIL); 657 if (error) 658 return (error); 659 error = SYSCTL_IN(req, addrs, sizeof(addrs)); 660 if (error) 661 return (error); 662 s = splnet(); 663 INP_INFO_RLOCK(&udbinfo); 664 inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port, 665 addrs[0].sin_addr, addrs[0].sin_port, 1, NULL); 666 if (inp == NULL || inp->inp_socket == NULL) { 667 error = ENOENT; 668 goto out; 669 } 670 error = cr_canseesocket(req->td->td_ucred, inp->inp_socket); 671 if (error) 672 goto out; 673 cru2x(inp->inp_socket->so_cred, &xuc); 674 out: 675 INP_INFO_RUNLOCK(&udbinfo); 676 splx(s); 677 if (error == 0) 678 error = SYSCTL_OUT(req, &xuc, sizeof(struct xucred)); 679 return (error); 680 } 681 682 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred, 683 CTLTYPE_OPAQUE|CTLFLAG_RW|CTLFLAG_PRISON, 0, 0, 684 udp_getcred, "S,xucred", "Get the xucred of a UDP connection"); 685 686 static int 687 udp_output(inp, m, addr, control, td) 688 register struct inpcb *inp; 689 struct mbuf *m; 690 struct sockaddr *addr; 691 struct mbuf *control; 692 struct thread *td; 693 { 694 register struct udpiphdr *ui; 695 register int len = m->m_pkthdr.len; 696 struct in_addr faddr, laddr; 697 struct cmsghdr *cm; 698 struct sockaddr_in *sin, src; 699 int error = 0; 700 int ipflags; 701 u_short fport, lport; 702 int unlock_udbinfo; 703 704 /* 705 * udp_output() may need to temporarily bind or connect the current 706 * inpcb. As such, we don't know up front what inpcb locks we will 707 * need. Do any work to decide what is needed up front before 708 * acquiring locks. 709 */ 710 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) { 711 if (control) 712 m_freem(control); 713 m_freem(m); 714 return EMSGSIZE; 715 } 716 717 src.sin_addr.s_addr = INADDR_ANY; 718 if (control != NULL) { 719 /* 720 * XXX: Currently, we assume all the optional information 721 * is stored in a single mbuf. 722 */ 723 if (control->m_next) { 724 m_freem(control); 725 m_freem(m); 726 return EINVAL; 727 } 728 for (; control->m_len > 0; 729 control->m_data += CMSG_ALIGN(cm->cmsg_len), 730 control->m_len -= CMSG_ALIGN(cm->cmsg_len)) { 731 cm = mtod(control, struct cmsghdr *); 732 if (control->m_len < sizeof(*cm) || cm->cmsg_len == 0 || 733 cm->cmsg_len > control->m_len) { 734 error = EINVAL; 735 break; 736 } 737 if (cm->cmsg_level != IPPROTO_IP) 738 continue; 739 740 switch (cm->cmsg_type) { 741 case IP_SENDSRCADDR: 742 if (cm->cmsg_len != 743 CMSG_LEN(sizeof(struct in_addr))) { 744 error = EINVAL; 745 break; 746 } 747 bzero(&src, sizeof(src)); 748 src.sin_family = AF_INET; 749 src.sin_len = sizeof(src); 750 src.sin_port = inp->inp_lport; 751 src.sin_addr = *(struct in_addr *)CMSG_DATA(cm); 752 break; 753 default: 754 error = ENOPROTOOPT; 755 break; 756 } 757 if (error) 758 break; 759 } 760 m_freem(control); 761 } 762 if (error) { 763 m_freem(m); 764 return error; 765 } 766 767 if (src.sin_addr.s_addr != INADDR_ANY || 768 addr != NULL) { 769 INP_INFO_WLOCK(&udbinfo); 770 unlock_udbinfo = 1; 771 } else 772 unlock_udbinfo = 0; 773 INP_LOCK(inp); 774 775 #ifdef MAC 776 mac_create_mbuf_from_inpcb(inp, m); 777 #endif 778 779 laddr = inp->inp_laddr; 780 lport = inp->inp_lport; 781 if (src.sin_addr.s_addr != INADDR_ANY) { 782 if (lport == 0) { 783 error = EINVAL; 784 goto release; 785 } 786 error = in_pcbbind_setup(inp, (struct sockaddr *)&src, 787 &laddr.s_addr, &lport, td->td_ucred); 788 if (error) 789 goto release; 790 } 791 792 if (addr) { 793 sin = (struct sockaddr_in *)addr; 794 if (jailed(td->td_ucred)) 795 prison_remote_ip(td->td_ucred, 0, &sin->sin_addr.s_addr); 796 if (inp->inp_faddr.s_addr != INADDR_ANY) { 797 error = EISCONN; 798 goto release; 799 } 800 error = in_pcbconnect_setup(inp, addr, &laddr.s_addr, &lport, 801 &faddr.s_addr, &fport, NULL, td->td_ucred); 802 if (error) 803 goto release; 804 805 /* Commit the local port if newly assigned. */ 806 if (inp->inp_laddr.s_addr == INADDR_ANY && 807 inp->inp_lport == 0) { 808 /* 809 * Remember addr if jailed, to prevent rebinding. 810 */ 811 if (jailed(td->td_ucred)) 812 inp->inp_laddr = laddr; 813 inp->inp_lport = lport; 814 if (in_pcbinshash(inp) != 0) { 815 inp->inp_lport = 0; 816 error = EAGAIN; 817 goto release; 818 } 819 inp->inp_flags |= INP_ANONPORT; 820 } 821 } else { 822 faddr = inp->inp_faddr; 823 fport = inp->inp_fport; 824 if (faddr.s_addr == INADDR_ANY) { 825 error = ENOTCONN; 826 goto release; 827 } 828 } 829 830 /* 831 * Calculate data length and get a mbuf for UDP, IP, and possible 832 * link-layer headers. Immediate slide the data pointer back forward 833 * since we won't use that space at this layer. 834 */ 835 M_PREPEND(m, sizeof(struct udpiphdr) + max_linkhdr, M_DONTWAIT); 836 if (m == NULL) { 837 error = ENOBUFS; 838 goto release; 839 } 840 m->m_data += max_linkhdr; 841 m->m_len -= max_linkhdr; 842 m->m_pkthdr.len -= max_linkhdr; 843 844 /* 845 * Fill in mbuf with extended UDP header 846 * and addresses and length put into network format. 847 */ 848 ui = mtod(m, struct udpiphdr *); 849 bzero(ui->ui_x1, sizeof(ui->ui_x1)); /* XXX still needed? */ 850 ui->ui_pr = IPPROTO_UDP; 851 ui->ui_src = laddr; 852 ui->ui_dst = faddr; 853 ui->ui_sport = lport; 854 ui->ui_dport = fport; 855 ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr)); 856 857 ipflags = 0; 858 if (inp->inp_socket->so_options & SO_DONTROUTE) 859 ipflags |= IP_ROUTETOIF; 860 if (inp->inp_socket->so_options & SO_BROADCAST) 861 ipflags |= IP_ALLOWBROADCAST; 862 if (inp->inp_flags & INP_ONESBCAST) 863 ipflags |= IP_SENDONES; 864 865 /* 866 * Set up checksum and output datagram. 867 */ 868 if (udpcksum) { 869 if (inp->inp_flags & INP_ONESBCAST) 870 faddr.s_addr = INADDR_BROADCAST; 871 ui->ui_sum = in_pseudo(ui->ui_src.s_addr, faddr.s_addr, 872 htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP)); 873 m->m_pkthdr.csum_flags = CSUM_UDP; 874 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); 875 } else { 876 ui->ui_sum = 0; 877 } 878 ((struct ip *)ui)->ip_len = sizeof (struct udpiphdr) + len; 879 ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */ 880 ((struct ip *)ui)->ip_tos = inp->inp_ip_tos; /* XXX */ 881 udpstat.udps_opackets++; 882 883 if (unlock_udbinfo) 884 INP_INFO_WUNLOCK(&udbinfo); 885 error = ip_output(m, inp->inp_options, NULL, ipflags, 886 inp->inp_moptions, inp); 887 INP_UNLOCK(inp); 888 return (error); 889 890 release: 891 INP_UNLOCK(inp); 892 if (unlock_udbinfo) 893 INP_INFO_WUNLOCK(&udbinfo); 894 m_freem(m); 895 return (error); 896 } 897 898 u_long udp_sendspace = 9216; /* really max datagram size */ 899 /* 40 1K datagrams */ 900 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW, 901 &udp_sendspace, 0, "Maximum outgoing UDP datagram size"); 902 903 u_long udp_recvspace = 40 * (1024 + 904 #ifdef INET6 905 sizeof(struct sockaddr_in6) 906 #else 907 sizeof(struct sockaddr_in) 908 #endif 909 ); 910 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW, 911 &udp_recvspace, 0, "Maximum space for incoming UDP datagrams"); 912 913 static int 914 udp_abort(struct socket *so) 915 { 916 struct inpcb *inp; 917 int s; 918 919 INP_INFO_WLOCK(&udbinfo); 920 inp = sotoinpcb(so); 921 if (inp == 0) { 922 INP_INFO_WUNLOCK(&udbinfo); 923 return EINVAL; /* ??? possible? panic instead? */ 924 } 925 INP_LOCK(inp); 926 soisdisconnected(so); 927 s = splnet(); 928 in_pcbdetach(inp); 929 INP_INFO_WUNLOCK(&udbinfo); 930 splx(s); 931 return 0; 932 } 933 934 static int 935 udp_attach(struct socket *so, int proto, struct thread *td) 936 { 937 struct inpcb *inp; 938 int s, error; 939 940 INP_INFO_WLOCK(&udbinfo); 941 inp = sotoinpcb(so); 942 if (inp != 0) { 943 INP_INFO_WUNLOCK(&udbinfo); 944 return EINVAL; 945 } 946 error = soreserve(so, udp_sendspace, udp_recvspace); 947 if (error) { 948 INP_INFO_WUNLOCK(&udbinfo); 949 return error; 950 } 951 s = splnet(); 952 error = in_pcballoc(so, &udbinfo, "udpinp"); 953 splx(s); 954 if (error) { 955 INP_INFO_WUNLOCK(&udbinfo); 956 return error; 957 } 958 959 inp = (struct inpcb *)so->so_pcb; 960 INP_LOCK(inp); 961 INP_INFO_WUNLOCK(&udbinfo); 962 inp->inp_vflag |= INP_IPV4; 963 inp->inp_ip_ttl = ip_defttl; 964 INP_UNLOCK(inp); 965 return 0; 966 } 967 968 static int 969 udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 970 { 971 struct inpcb *inp; 972 int s, error; 973 974 INP_INFO_WLOCK(&udbinfo); 975 inp = sotoinpcb(so); 976 if (inp == 0) { 977 INP_INFO_WUNLOCK(&udbinfo); 978 return EINVAL; 979 } 980 INP_LOCK(inp); 981 s = splnet(); 982 error = in_pcbbind(inp, nam, td->td_ucred); 983 splx(s); 984 INP_UNLOCK(inp); 985 INP_INFO_WUNLOCK(&udbinfo); 986 return error; 987 } 988 989 static int 990 udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 991 { 992 struct inpcb *inp; 993 int s, error; 994 struct sockaddr_in *sin; 995 996 INP_INFO_WLOCK(&udbinfo); 997 inp = sotoinpcb(so); 998 if (inp == 0) { 999 INP_INFO_WUNLOCK(&udbinfo); 1000 return EINVAL; 1001 } 1002 INP_LOCK(inp); 1003 if (inp->inp_faddr.s_addr != INADDR_ANY) { 1004 INP_UNLOCK(inp); 1005 INP_INFO_WUNLOCK(&udbinfo); 1006 return EISCONN; 1007 } 1008 s = splnet(); 1009 sin = (struct sockaddr_in *)nam; 1010 if (jailed(td->td_ucred)) 1011 prison_remote_ip(td->td_ucred, 0, &sin->sin_addr.s_addr); 1012 error = in_pcbconnect(inp, nam, td->td_ucred); 1013 splx(s); 1014 if (error == 0) 1015 soisconnected(so); 1016 INP_UNLOCK(inp); 1017 INP_INFO_WUNLOCK(&udbinfo); 1018 return error; 1019 } 1020 1021 static int 1022 udp_detach(struct socket *so) 1023 { 1024 struct inpcb *inp; 1025 int s; 1026 1027 INP_INFO_WLOCK(&udbinfo); 1028 inp = sotoinpcb(so); 1029 if (inp == 0) { 1030 INP_INFO_WUNLOCK(&udbinfo); 1031 return EINVAL; 1032 } 1033 INP_LOCK(inp); 1034 s = splnet(); 1035 in_pcbdetach(inp); 1036 INP_INFO_WUNLOCK(&udbinfo); 1037 splx(s); 1038 return 0; 1039 } 1040 1041 static int 1042 udp_disconnect(struct socket *so) 1043 { 1044 struct inpcb *inp; 1045 int s; 1046 1047 INP_INFO_WLOCK(&udbinfo); 1048 inp = sotoinpcb(so); 1049 if (inp == 0) { 1050 INP_INFO_WUNLOCK(&udbinfo); 1051 return EINVAL; 1052 } 1053 INP_LOCK(inp); 1054 if (inp->inp_faddr.s_addr == INADDR_ANY) { 1055 INP_INFO_WUNLOCK(&udbinfo); 1056 INP_UNLOCK(inp); 1057 return ENOTCONN; 1058 } 1059 1060 s = splnet(); 1061 in_pcbdisconnect(inp); 1062 inp->inp_laddr.s_addr = INADDR_ANY; 1063 INP_UNLOCK(inp); 1064 INP_INFO_WUNLOCK(&udbinfo); 1065 splx(s); 1066 so->so_state &= ~SS_ISCONNECTED; /* XXX */ 1067 return 0; 1068 } 1069 1070 static int 1071 udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, 1072 struct mbuf *control, struct thread *td) 1073 { 1074 struct inpcb *inp; 1075 1076 inp = sotoinpcb(so); 1077 return udp_output(inp, m, addr, control, td); 1078 } 1079 1080 int 1081 udp_shutdown(struct socket *so) 1082 { 1083 struct inpcb *inp; 1084 1085 INP_INFO_RLOCK(&udbinfo); 1086 inp = sotoinpcb(so); 1087 if (inp == 0) { 1088 INP_INFO_RUNLOCK(&udbinfo); 1089 return EINVAL; 1090 } 1091 INP_LOCK(inp); 1092 INP_INFO_RUNLOCK(&udbinfo); 1093 socantsendmore(so); 1094 INP_UNLOCK(inp); 1095 return 0; 1096 } 1097 1098 /* 1099 * This is the wrapper function for in_setsockaddr. We just pass down 1100 * the pcbinfo for in_setsockaddr to lock. We don't want to do the locking 1101 * here because in_setsockaddr will call malloc and might block. 1102 */ 1103 static int 1104 udp_sockaddr(struct socket *so, struct sockaddr **nam) 1105 { 1106 return (in_setsockaddr(so, nam, &udbinfo)); 1107 } 1108 1109 /* 1110 * This is the wrapper function for in_setpeeraddr. We just pass down 1111 * the pcbinfo for in_setpeeraddr to lock. 1112 */ 1113 static int 1114 udp_peeraddr(struct socket *so, struct sockaddr **nam) 1115 { 1116 return (in_setpeeraddr(so, nam, &udbinfo)); 1117 } 1118 1119 struct pr_usrreqs udp_usrreqs = { 1120 .pru_abort = udp_abort, 1121 .pru_attach = udp_attach, 1122 .pru_bind = udp_bind, 1123 .pru_connect = udp_connect, 1124 .pru_control = in_control, 1125 .pru_detach = udp_detach, 1126 .pru_disconnect = udp_disconnect, 1127 .pru_peeraddr = udp_peeraddr, 1128 .pru_send = udp_send, 1129 .pru_shutdown = udp_shutdown, 1130 .pru_sockaddr = udp_sockaddr, 1131 .pru_sosetlabel = in_pcbsosetlabel 1132 }; 1133