1 /* 2 * Copyright (c) 2004 Jeffrey M. Hsu. All rights reserved. 3 * Copyright (c) 2004 The DragonFly Project. All rights reserved. 4 * 5 * This code is derived from software contributed to The DragonFly Project 6 * by Jeffrey M. Hsu. 7 * 8 * Redistribution and use in source and binary forms, with or without 9 * modification, are permitted provided that the following conditions 10 * are met: 11 * 1. Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * 2. Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in the 15 * documentation and/or other materials provided with the distribution. 16 * 3. Neither the name of The DragonFly Project nor the names of its 17 * contributors may be used to endorse or promote products derived 18 * from this software without specific, prior written permission. 19 * 20 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 21 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 22 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS 23 * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE 24 * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, 25 * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING, 26 * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 27 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED 28 * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 29 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 30 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 */ 33 34 /* 35 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1995 36 * The Regents of the University of California. All rights reserved. 37 * 38 * Redistribution and use in source and binary forms, with or without 39 * modification, are permitted provided that the following conditions 40 * are met: 41 * 1. Redistributions of source code must retain the above copyright 42 * notice, this list of conditions and the following disclaimer. 43 * 2. Redistributions in binary form must reproduce the above copyright 44 * notice, this list of conditions and the following disclaimer in the 45 * documentation and/or other materials provided with the distribution. 46 * 3. All advertising materials mentioning features or use of this software 47 * must display the following acknowledgement: 48 * This product includes software developed by the University of 49 * California, Berkeley and its contributors. 50 * 4. Neither the name of the University nor the names of its contributors 51 * may be used to endorse or promote products derived from this software 52 * without specific prior written permission. 53 * 54 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 55 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 56 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 57 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 58 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 59 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 60 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 61 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 62 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 63 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 64 * SUCH DAMAGE. 65 * 66 * @(#)udp_usrreq.c 8.6 (Berkeley) 5/23/95 67 * $FreeBSD: src/sys/netinet/udp_usrreq.c,v 1.64.2.18 2003/01/24 05:11:34 sam Exp $ 68 * $DragonFly: src/sys/netinet/udp_usrreq.c,v 1.47 2008/11/11 10:46:58 sephe Exp $ 69 */ 70 71 #include "opt_ipsec.h" 72 #include "opt_inet6.h" 73 74 #include <sys/param.h> 75 #include <sys/systm.h> 76 #include <sys/kernel.h> 77 #include <sys/malloc.h> 78 #include <sys/mbuf.h> 79 #include <sys/domain.h> 80 #include <sys/proc.h> 81 #include <sys/priv.h> 82 #include <sys/protosw.h> 83 #include <sys/socket.h> 84 #include <sys/socketvar.h> 85 #include <sys/sysctl.h> 86 #include <sys/syslog.h> 87 #include <sys/thread2.h> 88 #include <sys/in_cksum.h> 89 90 #include <machine/stdarg.h> 91 92 #include <net/if.h> 93 #include <net/route.h> 94 #include <net/netmsg2.h> 95 96 #include <netinet/in.h> 97 #include <netinet/in_systm.h> 98 #include <netinet/ip.h> 99 #ifdef INET6 100 #include <netinet/ip6.h> 101 #endif 102 #include <netinet/in_pcb.h> 103 #include <netinet/in_var.h> 104 #include <netinet/ip_var.h> 105 #ifdef INET6 106 #include <netinet6/ip6_var.h> 107 #endif 108 #include <netinet/ip_icmp.h> 109 #include <netinet/icmp_var.h> 110 #include <netinet/udp.h> 111 #include <netinet/udp_var.h> 112 113 #ifdef FAST_IPSEC 114 #include <netproto/ipsec/ipsec.h> 115 #endif 116 117 #ifdef IPSEC 118 #include <netinet6/ipsec.h> 119 #endif 120 121 int udp_mpsafe_proto = 0; 122 TUNABLE_INT("net.inet.udp.mpsafe_proto", &udp_mpsafe_proto); 123 124 int udp_mpsafe_thread = NETMSG_SERVICE_ADAPTIVE; 125 TUNABLE_INT("net.inet.udp.mpsafe_thread", &udp_mpsafe_thread); 126 SYSCTL_INT(_net_inet_udp, OID_AUTO, mpsafe_thread, CTLFLAG_RW, 127 &udp_mpsafe_thread, 0, 128 "0:BGL, 1:Adaptive BGL, 2:No BGL(experimental)"); 129 130 /* 131 * UDP protocol implementation. 132 * Per RFC 768, August, 1980. 133 */ 134 #ifndef COMPAT_42 135 static int udpcksum = 1; 136 #else 137 static int udpcksum = 0; /* XXX */ 138 #endif 139 SYSCTL_INT(_net_inet_udp, UDPCTL_CHECKSUM, checksum, CTLFLAG_RW, 140 &udpcksum, 0, ""); 141 142 int log_in_vain = 0; 143 SYSCTL_INT(_net_inet_udp, OID_AUTO, log_in_vain, CTLFLAG_RW, 144 &log_in_vain, 0, "Log all incoming UDP packets"); 145 146 static int blackhole = 0; 147 SYSCTL_INT(_net_inet_udp, OID_AUTO, blackhole, CTLFLAG_RW, 148 &blackhole, 0, "Do not send port unreachables for refused connects"); 149 150 static int strict_mcast_mship = 1; 151 SYSCTL_INT(_net_inet_udp, OID_AUTO, strict_mcast_mship, CTLFLAG_RW, 152 &strict_mcast_mship, 0, "Only send multicast to member sockets"); 153 154 struct inpcbinfo udbinfo; 155 156 #ifndef UDBHASHSIZE 157 #define UDBHASHSIZE 16 158 #endif 159 160 struct udpstat udpstat; /* from udp_var.h */ 161 SYSCTL_STRUCT(_net_inet_udp, UDPCTL_STATS, stats, CTLFLAG_RW, 162 &udpstat, udpstat, "UDP statistics (struct udpstat, netinet/udp_var.h)"); 163 164 static struct sockaddr_in udp_in = { sizeof udp_in, AF_INET }; 165 #ifdef INET6 166 struct udp_in6 { 167 struct sockaddr_in6 uin6_sin; 168 u_char uin6_init_done : 1; 169 } udp_in6 = { 170 { sizeof udp_in6.uin6_sin, AF_INET6 }, 171 0 172 }; 173 struct udp_ip6 { 174 struct ip6_hdr uip6_ip6; 175 u_char uip6_init_done : 1; 176 } udp_ip6; 177 #endif /* INET6 */ 178 179 static void udp_append (struct inpcb *last, struct ip *ip, 180 struct mbuf *n, int off); 181 #ifdef INET6 182 static void ip_2_ip6_hdr (struct ip6_hdr *ip6, struct ip *ip); 183 #endif 184 185 static int udp_connect_oncpu(struct socket *so, struct thread *td, 186 struct sockaddr_in *sin, struct sockaddr_in *if_sin); 187 static int udp_detach (struct socket *so); 188 static int udp_output (struct inpcb *, struct mbuf *, struct sockaddr *, 189 struct mbuf *, struct thread *); 190 191 void 192 udp_init(void) 193 { 194 in_pcbinfo_init(&udbinfo); 195 udbinfo.hashbase = hashinit(UDBHASHSIZE, M_PCB, &udbinfo.hashmask); 196 udbinfo.porthashbase = hashinit(UDBHASHSIZE, M_PCB, 197 &udbinfo.porthashmask); 198 udbinfo.wildcardhashbase = hashinit(UDBHASHSIZE, M_PCB, 199 &udbinfo.wildcardhashmask); 200 udbinfo.ipi_size = sizeof(struct inpcb); 201 udp_thread_init(); 202 } 203 204 /* 205 * Check multicast packets to make sure they are only sent to sockets with 206 * multicast memberships for the packet's destination address and arrival 207 * interface. Multicast packets to multicast-unaware sockets are also 208 * disallowed. 209 * 210 * Returns 0 if the packet is acceptable, -1 if it is not. 211 */ 212 static __inline int 213 check_multicast_membership(struct ip *ip, struct inpcb *inp, struct mbuf *m) 214 { 215 int mshipno; 216 struct ip_moptions *mopt; 217 218 if (strict_mcast_mship == 0 || 219 !IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 220 return (0); 221 } 222 mopt = inp->inp_moptions; 223 if (mopt == NULL) 224 return (-1); 225 for (mshipno = 0; mshipno <= mopt->imo_num_memberships; ++mshipno) { 226 struct in_multi *maddr = mopt->imo_membership[mshipno]; 227 228 if (ip->ip_dst.s_addr == maddr->inm_addr.s_addr && 229 m->m_pkthdr.rcvif == maddr->inm_ifp) { 230 return (0); 231 } 232 } 233 return (-1); 234 } 235 236 void 237 udp_input(struct mbuf *m, ...) 238 { 239 int iphlen; 240 struct ip *ip; 241 struct udphdr *uh; 242 struct inpcb *inp; 243 struct mbuf *opts = NULL; 244 int len, off, proto; 245 struct ip save_ip; 246 struct sockaddr *append_sa; 247 __va_list ap; 248 249 __va_start(ap, m); 250 off = __va_arg(ap, int); 251 proto = __va_arg(ap, int); 252 __va_end(ap); 253 254 iphlen = off; 255 udpstat.udps_ipackets++; 256 257 /* 258 * Strip IP options, if any; should skip this, 259 * make available to user, and use on returned packets, 260 * but we don't yet have a way to check the checksum 261 * with options still present. 262 */ 263 if (iphlen > sizeof(struct ip)) { 264 ip_stripoptions(m); 265 iphlen = sizeof(struct ip); 266 } 267 268 /* 269 * IP and UDP headers are together in first mbuf. 270 * Already checked and pulled up in ip_demux(). 271 */ 272 KASSERT(m->m_len >= iphlen + sizeof(struct udphdr), 273 ("UDP header not in one mbuf")); 274 275 ip = mtod(m, struct ip *); 276 uh = (struct udphdr *)((caddr_t)ip + iphlen); 277 278 /* destination port of 0 is illegal, based on RFC768. */ 279 if (uh->uh_dport == 0) 280 goto bad; 281 282 /* 283 * Make mbuf data length reflect UDP length. 284 * If not enough data to reflect UDP length, drop. 285 */ 286 len = ntohs((u_short)uh->uh_ulen); 287 if (ip->ip_len != len) { 288 if (len > ip->ip_len || len < sizeof(struct udphdr)) { 289 udpstat.udps_badlen++; 290 goto bad; 291 } 292 m_adj(m, len - ip->ip_len); 293 /* ip->ip_len = len; */ 294 } 295 /* 296 * Save a copy of the IP header in case we want restore it 297 * for sending an ICMP error message in response. 298 */ 299 save_ip = *ip; 300 301 /* 302 * Checksum extended UDP header and data. 303 */ 304 if (uh->uh_sum) { 305 if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { 306 if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) 307 uh->uh_sum = m->m_pkthdr.csum_data; 308 else 309 uh->uh_sum = in_pseudo(ip->ip_src.s_addr, 310 ip->ip_dst.s_addr, htonl((u_short)len + 311 m->m_pkthdr.csum_data + IPPROTO_UDP)); 312 uh->uh_sum ^= 0xffff; 313 } else { 314 char b[9]; 315 316 bcopy(((struct ipovly *)ip)->ih_x1, b, 9); 317 bzero(((struct ipovly *)ip)->ih_x1, 9); 318 ((struct ipovly *)ip)->ih_len = uh->uh_ulen; 319 uh->uh_sum = in_cksum(m, len + sizeof(struct ip)); 320 bcopy(b, ((struct ipovly *)ip)->ih_x1, 9); 321 } 322 if (uh->uh_sum) { 323 udpstat.udps_badsum++; 324 m_freem(m); 325 return; 326 } 327 } else 328 udpstat.udps_nosum++; 329 330 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr)) || 331 in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) { 332 struct inpcb *last; 333 334 /* 335 * Deliver a multicast or broadcast datagram to *all* sockets 336 * for which the local and remote addresses and ports match 337 * those of the incoming datagram. This allows more than 338 * one process to receive multi/broadcasts on the same port. 339 * (This really ought to be done for unicast datagrams as 340 * well, but that would cause problems with existing 341 * applications that open both address-specific sockets and 342 * a wildcard socket listening to the same port -- they would 343 * end up receiving duplicates of every unicast datagram. 344 * Those applications open the multiple sockets to overcome an 345 * inadequacy of the UDP socket interface, but for backwards 346 * compatibility we avoid the problem here rather than 347 * fixing the interface. Maybe 4.5BSD will remedy this?) 348 */ 349 350 /* 351 * Construct sockaddr format source address. 352 */ 353 udp_in.sin_port = uh->uh_sport; 354 udp_in.sin_addr = ip->ip_src; 355 /* 356 * Locate pcb(s) for datagram. 357 * (Algorithm copied from raw_intr().) 358 */ 359 last = NULL; 360 #ifdef INET6 361 udp_in6.uin6_init_done = udp_ip6.uip6_init_done = 0; 362 #endif 363 LIST_FOREACH(inp, &udbinfo.pcblisthead, inp_list) { 364 if (inp->inp_flags & INP_PLACEMARKER) 365 continue; 366 #ifdef INET6 367 if (!(inp->inp_vflag & INP_IPV4)) 368 continue; 369 #endif 370 if (inp->inp_lport != uh->uh_dport) 371 continue; 372 if (inp->inp_laddr.s_addr != INADDR_ANY) { 373 if (inp->inp_laddr.s_addr != 374 ip->ip_dst.s_addr) 375 continue; 376 } 377 if (inp->inp_faddr.s_addr != INADDR_ANY) { 378 if (inp->inp_faddr.s_addr != 379 ip->ip_src.s_addr || 380 inp->inp_fport != uh->uh_sport) 381 continue; 382 } 383 384 if (check_multicast_membership(ip, inp, m) < 0) 385 continue; 386 387 if (last != NULL) { 388 struct mbuf *n; 389 390 #ifdef IPSEC 391 /* check AH/ESP integrity. */ 392 if (ipsec4_in_reject_so(m, last->inp_socket)) 393 ipsecstat.in_polvio++; 394 /* do not inject data to pcb */ 395 else 396 #endif /*IPSEC*/ 397 #ifdef FAST_IPSEC 398 /* check AH/ESP integrity. */ 399 if (ipsec4_in_reject(m, last)) 400 ; 401 else 402 #endif /*FAST_IPSEC*/ 403 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) 404 udp_append(last, ip, n, 405 iphlen + 406 sizeof(struct udphdr)); 407 } 408 last = inp; 409 /* 410 * Don't look for additional matches if this one does 411 * not have either the SO_REUSEPORT or SO_REUSEADDR 412 * socket options set. This heuristic avoids searching 413 * through all pcbs in the common case of a non-shared 414 * port. It * assumes that an application will never 415 * clear these options after setting them. 416 */ 417 if (!(last->inp_socket->so_options & 418 (SO_REUSEPORT | SO_REUSEADDR))) 419 break; 420 } 421 422 if (last == NULL) { 423 /* 424 * No matching pcb found; discard datagram. 425 * (No need to send an ICMP Port Unreachable 426 * for a broadcast or multicast datgram.) 427 */ 428 udpstat.udps_noportbcast++; 429 goto bad; 430 } 431 #ifdef IPSEC 432 /* check AH/ESP integrity. */ 433 if (ipsec4_in_reject_so(m, last->inp_socket)) { 434 ipsecstat.in_polvio++; 435 goto bad; 436 } 437 #endif /*IPSEC*/ 438 #ifdef FAST_IPSEC 439 /* check AH/ESP integrity. */ 440 if (ipsec4_in_reject(m, last)) 441 goto bad; 442 #endif /*FAST_IPSEC*/ 443 udp_append(last, ip, m, iphlen + sizeof(struct udphdr)); 444 return; 445 } 446 /* 447 * Locate pcb for datagram. 448 */ 449 inp = in_pcblookup_hash(&udbinfo, ip->ip_src, uh->uh_sport, 450 ip->ip_dst, uh->uh_dport, 1, m->m_pkthdr.rcvif); 451 if (inp == NULL) { 452 if (log_in_vain) { 453 char buf[sizeof "aaa.bbb.ccc.ddd"]; 454 455 strcpy(buf, inet_ntoa(ip->ip_dst)); 456 log(LOG_INFO, 457 "Connection attempt to UDP %s:%d from %s:%d\n", 458 buf, ntohs(uh->uh_dport), inet_ntoa(ip->ip_src), 459 ntohs(uh->uh_sport)); 460 } 461 udpstat.udps_noport++; 462 if (m->m_flags & (M_BCAST | M_MCAST)) { 463 udpstat.udps_noportbcast++; 464 goto bad; 465 } 466 if (blackhole) 467 goto bad; 468 #ifdef ICMP_BANDLIM 469 if (badport_bandlim(BANDLIM_ICMP_UNREACH) < 0) 470 goto bad; 471 #endif 472 *ip = save_ip; 473 ip->ip_len += iphlen; 474 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_PORT, 0, 0); 475 return; 476 } 477 #ifdef IPSEC 478 if (ipsec4_in_reject_so(m, inp->inp_socket)) { 479 ipsecstat.in_polvio++; 480 goto bad; 481 } 482 #endif /*IPSEC*/ 483 #ifdef FAST_IPSEC 484 if (ipsec4_in_reject(m, inp)) 485 goto bad; 486 #endif /*FAST_IPSEC*/ 487 /* 488 * Check the minimum TTL for socket. 489 */ 490 if (ip->ip_ttl < inp->inp_ip_minttl) 491 goto bad; 492 493 /* 494 * Construct sockaddr format source address. 495 * Stuff source address and datagram in user buffer. 496 */ 497 udp_in.sin_port = uh->uh_sport; 498 udp_in.sin_addr = ip->ip_src; 499 if ((inp->inp_flags & INP_CONTROLOPTS) || 500 (inp->inp_socket->so_options & SO_TIMESTAMP)) { 501 #ifdef INET6 502 if (inp->inp_vflag & INP_IPV6) { 503 int savedflags; 504 505 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip); 506 savedflags = inp->inp_flags; 507 inp->inp_flags &= ~INP_UNMAPPABLEOPTS; 508 ip6_savecontrol(inp, &opts, &udp_ip6.uip6_ip6, m); 509 inp->inp_flags = savedflags; 510 } else 511 #endif 512 ip_savecontrol(inp, &opts, ip, m); 513 } 514 m_adj(m, iphlen + sizeof(struct udphdr)); 515 #ifdef INET6 516 if (inp->inp_vflag & INP_IPV6) { 517 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin); 518 append_sa = (struct sockaddr *)&udp_in6; 519 } else 520 #endif 521 append_sa = (struct sockaddr *)&udp_in; 522 if (ssb_appendaddr(&inp->inp_socket->so_rcv, append_sa, m, opts) == 0) { 523 udpstat.udps_fullsock++; 524 goto bad; 525 } 526 sorwakeup(inp->inp_socket); 527 return; 528 bad: 529 m_freem(m); 530 if (opts) 531 m_freem(opts); 532 return; 533 } 534 535 #ifdef INET6 536 static void 537 ip_2_ip6_hdr(struct ip6_hdr *ip6, struct ip *ip) 538 { 539 bzero(ip6, sizeof *ip6); 540 541 ip6->ip6_vfc = IPV6_VERSION; 542 ip6->ip6_plen = ip->ip_len; 543 ip6->ip6_nxt = ip->ip_p; 544 ip6->ip6_hlim = ip->ip_ttl; 545 ip6->ip6_src.s6_addr32[2] = ip6->ip6_dst.s6_addr32[2] = 546 IPV6_ADDR_INT32_SMP; 547 ip6->ip6_src.s6_addr32[3] = ip->ip_src.s_addr; 548 ip6->ip6_dst.s6_addr32[3] = ip->ip_dst.s_addr; 549 } 550 #endif 551 552 /* 553 * subroutine of udp_input(), mainly for source code readability. 554 * caller must properly init udp_ip6 and udp_in6 beforehand. 555 */ 556 static void 557 udp_append(struct inpcb *last, struct ip *ip, struct mbuf *n, int off) 558 { 559 struct sockaddr *append_sa; 560 struct mbuf *opts = NULL; 561 562 if (last->inp_flags & INP_CONTROLOPTS || 563 last->inp_socket->so_options & SO_TIMESTAMP) { 564 #ifdef INET6 565 if (last->inp_vflag & INP_IPV6) { 566 int savedflags; 567 568 if (udp_ip6.uip6_init_done == 0) { 569 ip_2_ip6_hdr(&udp_ip6.uip6_ip6, ip); 570 udp_ip6.uip6_init_done = 1; 571 } 572 savedflags = last->inp_flags; 573 last->inp_flags &= ~INP_UNMAPPABLEOPTS; 574 ip6_savecontrol(last, &opts, &udp_ip6.uip6_ip6, n); 575 last->inp_flags = savedflags; 576 } else 577 #endif 578 ip_savecontrol(last, &opts, ip, n); 579 } 580 #ifdef INET6 581 if (last->inp_vflag & INP_IPV6) { 582 if (udp_in6.uin6_init_done == 0) { 583 in6_sin_2_v4mapsin6(&udp_in, &udp_in6.uin6_sin); 584 udp_in6.uin6_init_done = 1; 585 } 586 append_sa = (struct sockaddr *)&udp_in6.uin6_sin; 587 } else 588 #endif 589 append_sa = (struct sockaddr *)&udp_in; 590 m_adj(n, off); 591 if (ssb_appendaddr(&last->inp_socket->so_rcv, append_sa, n, opts) == 0) { 592 m_freem(n); 593 if (opts) 594 m_freem(opts); 595 udpstat.udps_fullsock++; 596 } else 597 sorwakeup(last->inp_socket); 598 } 599 600 /* 601 * Notify a udp user of an asynchronous error; 602 * just wake up so that he can collect error status. 603 */ 604 void 605 udp_notify(struct inpcb *inp, int error) 606 { 607 inp->inp_socket->so_error = error; 608 sorwakeup(inp->inp_socket); 609 sowwakeup(inp->inp_socket); 610 } 611 612 struct netmsg_udp_notify { 613 struct netmsg nm_nmsg; 614 void (*nm_notify)(struct inpcb *, int); 615 struct in_addr nm_faddr; 616 int nm_arg; 617 }; 618 619 static void 620 udp_notifyall_oncpu(struct netmsg *netmsg) 621 { 622 struct netmsg_udp_notify *nmsg = (struct netmsg_udp_notify *)netmsg; 623 int nextcpu; 624 625 in_pcbnotifyall(&udbinfo.pcblisthead, nmsg->nm_faddr, nmsg->nm_arg, 626 nmsg->nm_notify); 627 628 nextcpu = mycpuid + 1; 629 if (nextcpu < ncpus2) 630 lwkt_forwardmsg(udp_cport(nextcpu), &netmsg->nm_lmsg); 631 else 632 lwkt_replymsg(&netmsg->nm_lmsg, 0); 633 } 634 635 static void 636 udp_rtchange(struct inpcb *inp, int err) 637 { 638 #ifdef SMP 639 /* XXX Nuke this, once UDP inpcbs are CPU localized */ 640 if (inp->inp_route.ro_rt && inp->inp_route.ro_rt->rt_cpuid == mycpuid) { 641 rtfree(inp->inp_route.ro_rt); 642 inp->inp_route.ro_rt = NULL; 643 /* 644 * A new route can be allocated the next time 645 * output is attempted. 646 */ 647 } 648 #else 649 in_rtchange(inp, err); 650 #endif 651 } 652 653 void 654 udp_ctlinput(int cmd, struct sockaddr *sa, void *vip) 655 { 656 struct ip *ip = vip; 657 struct udphdr *uh; 658 void (*notify) (struct inpcb *, int) = udp_notify; 659 struct in_addr faddr; 660 struct inpcb *inp; 661 662 faddr = ((struct sockaddr_in *)sa)->sin_addr; 663 if (sa->sa_family != AF_INET || faddr.s_addr == INADDR_ANY) 664 return; 665 666 if (PRC_IS_REDIRECT(cmd)) { 667 ip = NULL; 668 notify = udp_rtchange; 669 } else if (cmd == PRC_HOSTDEAD) 670 ip = NULL; 671 else if ((unsigned)cmd >= PRC_NCMDS || inetctlerrmap[cmd] == 0) 672 return; 673 if (ip) { 674 crit_enter(); 675 uh = (struct udphdr *)((caddr_t)ip + (ip->ip_hl << 2)); 676 inp = in_pcblookup_hash(&udbinfo, faddr, uh->uh_dport, 677 ip->ip_src, uh->uh_sport, 0, NULL); 678 if (inp != NULL && inp->inp_socket != NULL) 679 (*notify)(inp, inetctlerrmap[cmd]); 680 crit_exit(); 681 } else if (PRC_IS_REDIRECT(cmd)) { 682 struct netmsg_udp_notify nmsg; 683 684 KKASSERT(&curthread->td_msgport == cpu_portfn(0)); 685 netmsg_init(&nmsg.nm_nmsg, NULL, &curthread->td_msgport, 686 0, udp_notifyall_oncpu); 687 nmsg.nm_faddr = faddr; 688 nmsg.nm_arg = inetctlerrmap[cmd]; 689 nmsg.nm_notify = notify; 690 691 lwkt_domsg(udp_cport(0), &nmsg.nm_nmsg.nm_lmsg, 0); 692 } else { 693 /* 694 * XXX We should forward msg upon PRC_HOSTHEAD and ip == NULL, 695 * once UDP inpcbs are CPU localized 696 */ 697 KKASSERT(&curthread->td_msgport == udp_cport(0)); 698 in_pcbnotifyall(&udbinfo.pcblisthead, faddr, inetctlerrmap[cmd], 699 notify); 700 } 701 } 702 703 SYSCTL_PROC(_net_inet_udp, UDPCTL_PCBLIST, pcblist, CTLFLAG_RD, &udbinfo, 0, 704 in_pcblist_global, "S,xinpcb", "List of active UDP sockets"); 705 706 static int 707 udp_getcred(SYSCTL_HANDLER_ARGS) 708 { 709 struct sockaddr_in addrs[2]; 710 struct inpcb *inp; 711 int error; 712 713 error = priv_check(req->td, PRIV_ROOT); 714 if (error) 715 return (error); 716 error = SYSCTL_IN(req, addrs, sizeof addrs); 717 if (error) 718 return (error); 719 crit_enter(); 720 inp = in_pcblookup_hash(&udbinfo, addrs[1].sin_addr, addrs[1].sin_port, 721 addrs[0].sin_addr, addrs[0].sin_port, 1, NULL); 722 if (inp == NULL || inp->inp_socket == NULL) { 723 error = ENOENT; 724 goto out; 725 } 726 error = SYSCTL_OUT(req, inp->inp_socket->so_cred, sizeof(struct ucred)); 727 out: 728 crit_exit(); 729 return (error); 730 } 731 732 SYSCTL_PROC(_net_inet_udp, OID_AUTO, getcred, CTLTYPE_OPAQUE|CTLFLAG_RW, 733 0, 0, udp_getcred, "S,ucred", "Get the ucred of a UDP connection"); 734 735 static int 736 udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *dstaddr, 737 struct mbuf *control, struct thread *td) 738 { 739 struct udpiphdr *ui; 740 int len = m->m_pkthdr.len; 741 struct sockaddr_in *sin; /* really is initialized before use */ 742 int error = 0, lport_any = 0; 743 744 if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) { 745 error = EMSGSIZE; 746 goto release; 747 } 748 749 if (inp->inp_lport == 0) { /* unbound socket */ 750 error = in_pcbbind(inp, NULL, td); 751 if (error) 752 goto release; 753 in_pcbinswildcardhash(inp); 754 lport_any = 1; 755 } 756 757 if (dstaddr != NULL) { /* destination address specified */ 758 if (inp->inp_faddr.s_addr != INADDR_ANY) { 759 /* already connected */ 760 error = EISCONN; 761 goto release; 762 } 763 sin = (struct sockaddr_in *)dstaddr; 764 if (!prison_remote_ip(td, (struct sockaddr *)&sin)) { 765 error = EAFNOSUPPORT; /* IPv6 only jail */ 766 goto release; 767 } 768 } else { 769 if (inp->inp_faddr.s_addr == INADDR_ANY) { 770 /* no destination specified and not already connected */ 771 error = ENOTCONN; 772 goto release; 773 } 774 sin = NULL; 775 } 776 777 /* 778 * Calculate data length and get a mbuf 779 * for UDP and IP headers. 780 */ 781 M_PREPEND(m, sizeof(struct udpiphdr), MB_DONTWAIT); 782 if (m == NULL) { 783 error = ENOBUFS; 784 goto release; 785 } 786 787 /* 788 * Fill in mbuf with extended UDP header 789 * and addresses and length put into network format. 790 */ 791 ui = mtod(m, struct udpiphdr *); 792 bzero(ui->ui_x1, sizeof ui->ui_x1); /* XXX still needed? */ 793 ui->ui_pr = IPPROTO_UDP; 794 795 /* 796 * Set destination address. 797 */ 798 if (dstaddr != NULL) { /* use specified destination */ 799 ui->ui_dst = sin->sin_addr; 800 ui->ui_dport = sin->sin_port; 801 } else { /* use connected destination */ 802 ui->ui_dst = inp->inp_faddr; 803 ui->ui_dport = inp->inp_fport; 804 } 805 806 /* 807 * Set source address. 808 */ 809 if (inp->inp_laddr.s_addr == INADDR_ANY) { 810 struct sockaddr_in *if_sin; 811 812 if (dstaddr == NULL) { 813 /* 814 * connect() had (or should have) failed because 815 * the interface had no IP address, but the 816 * application proceeded to call send() anyways. 817 */ 818 error = ENOTCONN; 819 goto release; 820 } 821 822 /* Look up outgoing interface. */ 823 if ((error = in_pcbladdr(inp, dstaddr, &if_sin, td))) 824 goto release; 825 ui->ui_src = if_sin->sin_addr; /* use address of interface */ 826 } else { 827 ui->ui_src = inp->inp_laddr; /* use non-null bound address */ 828 } 829 ui->ui_sport = inp->inp_lport; 830 KASSERT(inp->inp_lport != 0, ("inp lport should have been bound")); 831 832 ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr)); 833 834 /* 835 * Set up checksum and output datagram. 836 */ 837 if (udpcksum) { 838 ui->ui_sum = in_pseudo(ui->ui_src.s_addr, ui->ui_dst.s_addr, 839 htons((u_short)len + sizeof(struct udphdr) + IPPROTO_UDP)); 840 m->m_pkthdr.csum_flags = CSUM_UDP; 841 m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum); 842 } else { 843 ui->ui_sum = 0; 844 } 845 ((struct ip *)ui)->ip_len = sizeof(struct udpiphdr) + len; 846 ((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */ 847 ((struct ip *)ui)->ip_tos = inp->inp_ip_tos; /* XXX */ 848 udpstat.udps_opackets++; 849 850 error = ip_output(m, inp->inp_options, &inp->inp_route, 851 (inp->inp_socket->so_options & (SO_DONTROUTE | SO_BROADCAST)) | 852 IP_DEBUGROUTE, 853 inp->inp_moptions, inp); 854 855 /* 856 * If this is the first data gram sent on an unbound and unconnected 857 * UDP socket, lport will be changed in this function. If target 858 * CPU after this lport changing is no longer the current CPU, then 859 * free the route entry allocated on the current CPU. 860 */ 861 if (lport_any) { 862 if (udp_addrcpu(inp->inp_faddr.s_addr, inp->inp_fport, 863 inp->inp_laddr.s_addr, inp->inp_lport) != mycpuid) { 864 struct route *ro = &inp->inp_route; 865 866 if (ro->ro_rt != NULL) 867 RTFREE(ro->ro_rt); 868 bzero(ro, sizeof(*ro)); 869 } 870 } 871 return (error); 872 873 release: 874 m_freem(m); 875 return (error); 876 } 877 878 u_long udp_sendspace = 9216; /* really max datagram size */ 879 /* 40 1K datagrams */ 880 SYSCTL_INT(_net_inet_udp, UDPCTL_MAXDGRAM, maxdgram, CTLFLAG_RW, 881 &udp_sendspace, 0, "Maximum outgoing UDP datagram size"); 882 883 u_long udp_recvspace = 40 * (1024 + 884 #ifdef INET6 885 sizeof(struct sockaddr_in6) 886 #else 887 sizeof(struct sockaddr_in) 888 #endif 889 ); 890 SYSCTL_INT(_net_inet_udp, UDPCTL_RECVSPACE, recvspace, CTLFLAG_RW, 891 &udp_recvspace, 0, "Maximum incoming UDP datagram size"); 892 893 static int 894 udp_abort(struct socket *so) 895 { 896 struct inpcb *inp; 897 898 inp = so->so_pcb; 899 if (inp == NULL) 900 return EINVAL; /* ??? possible? panic instead? */ 901 soisdisconnected(so); 902 crit_enter(); 903 in_pcbdetach(inp); 904 crit_exit(); 905 return 0; 906 } 907 908 static int 909 udp_attach(struct socket *so, int proto, struct pru_attach_info *ai) 910 { 911 struct inpcb *inp; 912 int error; 913 914 inp = so->so_pcb; 915 if (inp != NULL) 916 return EINVAL; 917 918 error = soreserve(so, udp_sendspace, udp_recvspace, ai->sb_rlimit); 919 if (error) 920 return error; 921 crit_enter(); 922 error = in_pcballoc(so, &udbinfo); 923 crit_exit(); 924 if (error) 925 return error; 926 so->so_port = udp_soport_attach(so); 927 928 inp = (struct inpcb *)so->so_pcb; 929 inp->inp_vflag |= INP_IPV4; 930 inp->inp_ip_ttl = ip_defttl; 931 return 0; 932 } 933 934 static int 935 udp_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 936 { 937 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 938 struct inpcb *inp; 939 int error; 940 941 inp = so->so_pcb; 942 if (inp == NULL) 943 return EINVAL; 944 crit_enter(); 945 error = in_pcbbind(inp, nam, td); 946 crit_exit(); 947 if (error == 0) { 948 if (sin->sin_addr.s_addr != INADDR_ANY) 949 inp->inp_flags |= INP_WASBOUND_NOTANY; 950 in_pcbinswildcardhash(inp); 951 } 952 return error; 953 } 954 955 #ifdef SMP 956 957 struct netmsg_udp_connect { 958 struct netmsg nm_netmsg; 959 struct socket *nm_so; 960 struct sockaddr_in *nm_sin; 961 struct sockaddr_in *nm_ifsin; 962 struct thread *nm_td; 963 }; 964 965 static void 966 udp_connect_handler(netmsg_t netmsg) 967 { 968 struct netmsg_udp_connect *msg = (void *)netmsg; 969 int error; 970 971 error = udp_connect_oncpu(msg->nm_so, msg->nm_td, 972 msg->nm_sin, msg->nm_ifsin); 973 lwkt_replymsg(&msg->nm_netmsg.nm_lmsg, error); 974 } 975 976 #endif 977 978 static int 979 udp_connect(struct socket *so, struct sockaddr *nam, struct thread *td) 980 { 981 struct inpcb *inp; 982 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 983 struct sockaddr_in *if_sin; 984 lwkt_port_t port; 985 int error; 986 987 inp = so->so_pcb; 988 if (inp == NULL) 989 return EINVAL; 990 if (inp->inp_faddr.s_addr != INADDR_ANY) 991 return EISCONN; 992 error = 0; 993 994 /* 995 * Bind if we have to 996 */ 997 if (td->td_proc && td->td_proc->p_ucred->cr_prison != NULL && 998 inp->inp_laddr.s_addr == INADDR_ANY) { 999 error = in_pcbbind(inp, NULL, td); 1000 if (error) 1001 return (error); 1002 } 1003 1004 /* 1005 * Calculate the correct protocol processing thread. The connect 1006 * operation must run there. 1007 */ 1008 error = in_pcbladdr(inp, nam, &if_sin, td); 1009 if (error) 1010 return(error); 1011 if (!prison_remote_ip(td, nam)) 1012 return(EAFNOSUPPORT); /* IPv6 only jail */ 1013 1014 port = udp_addrport(sin->sin_addr.s_addr, sin->sin_port, 1015 inp->inp_laddr.s_addr, inp->inp_lport); 1016 #ifdef SMP 1017 if (port != &curthread->td_msgport) { 1018 struct netmsg_udp_connect msg; 1019 struct route *ro = &inp->inp_route; 1020 1021 /* 1022 * in_pcbladdr() may have allocated a route entry for us 1023 * on the current CPU, but we need a route entry on the 1024 * inpcb's owner CPU, so free it here. 1025 */ 1026 if (ro->ro_rt != NULL) 1027 RTFREE(ro->ro_rt); 1028 bzero(ro, sizeof(*ro)); 1029 1030 /* 1031 * NOTE: We haven't set so->so_port yet do not pass so 1032 * to netmsg_init() or it will be improperly forwarded. 1033 */ 1034 netmsg_init(&msg.nm_netmsg, NULL, &curthread->td_msgport, 1035 0, udp_connect_handler); 1036 msg.nm_so = so; 1037 msg.nm_sin = sin; 1038 msg.nm_ifsin = if_sin; 1039 msg.nm_td = td; 1040 error = lwkt_domsg(port, &msg.nm_netmsg.nm_lmsg, 0); 1041 } else { 1042 error = udp_connect_oncpu(so, td, sin, if_sin); 1043 } 1044 #else 1045 error = udp_connect_oncpu(so, td, sin, if_sin); 1046 #endif 1047 return (error); 1048 } 1049 1050 static int 1051 udp_connect_oncpu(struct socket *so, struct thread *td, 1052 struct sockaddr_in *sin, struct sockaddr_in *if_sin) 1053 { 1054 struct inpcb *inp; 1055 int error; 1056 1057 inp = so->so_pcb; 1058 if (inp->inp_flags & INP_WILDCARD) 1059 in_pcbremwildcardhash(inp); 1060 error = in_pcbconnect(inp, (struct sockaddr *)sin, td); 1061 1062 if (error == 0) { 1063 /* 1064 * No more errors can occur, finish adjusting the socket 1065 * and change the processing port to reflect the connected 1066 * socket. Once set we can no longer safely mess with the 1067 * socket. 1068 */ 1069 soisconnected(so); 1070 sosetport(so, &curthread->td_msgport); 1071 } else if (error == EAFNOSUPPORT) { /* connection dissolved */ 1072 /* 1073 * Follow traditional BSD behavior and retain 1074 * the local port binding. But, fix the old misbehavior 1075 * of overwriting any previously bound local address. 1076 */ 1077 if (!(inp->inp_flags & INP_WASBOUND_NOTANY)) 1078 inp->inp_laddr.s_addr = INADDR_ANY; 1079 in_pcbinswildcardhash(inp); 1080 } 1081 return error; 1082 } 1083 1084 static int 1085 udp_detach(struct socket *so) 1086 { 1087 struct inpcb *inp; 1088 1089 inp = so->so_pcb; 1090 if (inp == NULL) 1091 return EINVAL; 1092 crit_enter(); 1093 in_pcbdetach(inp); 1094 crit_exit(); 1095 return 0; 1096 } 1097 1098 static int 1099 udp_disconnect(struct socket *so) 1100 { 1101 struct route *ro; 1102 struct inpcb *inp; 1103 1104 inp = so->so_pcb; 1105 if (inp == NULL) 1106 return EINVAL; 1107 if (inp->inp_faddr.s_addr == INADDR_ANY) 1108 return ENOTCONN; 1109 1110 crit_enter(); 1111 in_pcbdisconnect(inp); 1112 crit_exit(); 1113 so->so_state &= ~SS_ISCONNECTED; /* XXX */ 1114 1115 ro = &inp->inp_route; 1116 if (ro->ro_rt != NULL) 1117 RTFREE(ro->ro_rt); 1118 bzero(ro, sizeof(*ro)); 1119 1120 return 0; 1121 } 1122 1123 static int 1124 udp_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *addr, 1125 struct mbuf *control, struct thread *td) 1126 { 1127 struct inpcb *inp; 1128 1129 inp = so->so_pcb; 1130 if (inp == NULL) { 1131 m_freem(m); 1132 return EINVAL; 1133 } 1134 return udp_output(inp, m, addr, control, td); 1135 } 1136 1137 int 1138 udp_shutdown(struct socket *so) 1139 { 1140 struct inpcb *inp; 1141 1142 inp = so->so_pcb; 1143 if (inp == NULL) 1144 return EINVAL; 1145 socantsendmore(so); 1146 return 0; 1147 } 1148 1149 struct pr_usrreqs udp_usrreqs = { 1150 .pru_abort = udp_abort, 1151 .pru_accept = pru_accept_notsupp, 1152 .pru_attach = udp_attach, 1153 .pru_bind = udp_bind, 1154 .pru_connect = udp_connect, 1155 .pru_connect2 = pru_connect2_notsupp, 1156 .pru_control = in_control, 1157 .pru_detach = udp_detach, 1158 .pru_disconnect = udp_disconnect, 1159 .pru_listen = pru_listen_notsupp, 1160 .pru_peeraddr = in_setpeeraddr, 1161 .pru_rcvd = pru_rcvd_notsupp, 1162 .pru_rcvoob = pru_rcvoob_notsupp, 1163 .pru_send = udp_send, 1164 .pru_sense = pru_sense_null, 1165 .pru_shutdown = udp_shutdown, 1166 .pru_sockaddr = in_setsockaddr, 1167 .pru_sosend = sosendudp, 1168 .pru_soreceive = soreceive 1169 }; 1170 1171