1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1988, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 #include "opt_inet.h" 36 #include "opt_inet6.h" 37 #include "opt_sctp.h" 38 39 #include <sys/param.h> 40 #include <sys/eventhandler.h> 41 #include <sys/kernel.h> 42 #include <sys/lock.h> 43 #include <sys/malloc.h> 44 #include <sys/mbuf.h> 45 #include <sys/module.h> 46 #include <sys/kernel.h> 47 #include <sys/priv.h> 48 #include <sys/proc.h> 49 #include <sys/domain.h> 50 #include <sys/protosw.h> 51 #include <sys/socket.h> 52 #include <sys/socketvar.h> 53 #include <sys/sysctl.h> 54 #include <net/vnet.h> 55 56 #include <net/if.h> 57 #include <net/if_var.h> 58 #include <net/netisr.h> 59 60 #include <netinet/in.h> 61 #include <netinet/in_pcb.h> 62 #include <netinet/in_systm.h> 63 #include <netinet/in_var.h> 64 #include <netinet/ip.h> 65 #include <netinet/ip_var.h> 66 #include <netinet/ip_divert.h> 67 #ifdef INET6 68 #include <netinet/ip6.h> 69 #include <netinet6/ip6_var.h> 70 #endif 71 #if defined(SCTP) || defined(SCTP_SUPPORT) 72 #include <netinet/sctp_crc32.h> 73 #endif 74 75 #include <security/mac/mac_framework.h> 76 /* 77 * Divert sockets 78 */ 79 80 /* 81 * Allocate enough space to hold a full IP packet 82 */ 83 #define DIVSNDQ (65536 + 100) 84 #define DIVRCVQ (65536 + 100) 85 86 /* 87 * Usually a system has very few divert ports. Previous implementation 88 * used a linked list. 89 */ 90 #define DIVHASHSIZE (1 << 3) /* 8 entries, one cache line. */ 91 #define DIVHASH(port) (port % DIVHASHSIZE) 92 #define DCBHASH(dcb) ((dcb)->dcb_port % DIVHASHSIZE) 93 94 /* 95 * Divert sockets work in conjunction with ipfw or other packet filters, 96 * see the divert(4) manpage for features. 97 * Packets are selected by the packet filter and tagged with an 98 * MTAG_IPFW_RULE tag carrying the 'divert port' number (as set by 99 * the packet filter) and information on the matching filter rule for 100 * subsequent reinjection. The divert_port is used to put the packet 101 * on the corresponding divert socket, while the rule number is passed 102 * up (at least partially) as the sin_port in the struct sockaddr. 103 * 104 * Packets written to the divert socket carry in sin_addr a 105 * destination address, and in sin_port the number of the filter rule 106 * after which to continue processing. 107 * If the destination address is INADDR_ANY, the packet is treated as 108 * as outgoing and sent to ip_output(); otherwise it is treated as 109 * incoming and sent to ip_input(). 110 * Further, sin_zero carries some information on the interface, 111 * which can be used in the reinject -- see comments in the code. 112 * 113 * On reinjection, processing in ip_input() and ip_output() 114 * will be exactly the same as for the original packet, except that 115 * packet filter processing will start at the rule number after the one 116 * written in the sin_port (ipfw does not allow a rule #0, so sin_port=0 117 * will apply the entire ruleset to the packet). 118 */ 119 static SYSCTL_NODE(_net_inet, OID_AUTO, divert, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 120 "divert(4)"); 121 122 VNET_PCPUSTAT_DEFINE_STATIC(struct divstat, divstat); 123 VNET_PCPUSTAT_SYSINIT(divstat); 124 #ifdef VIMAGE 125 VNET_PCPUSTAT_SYSUNINIT(divstat); 126 #endif 127 SYSCTL_VNET_PCPUSTAT(_net_inet_divert, OID_AUTO, stats, struct divstat, 128 divstat, "divert(4) socket statistics"); 129 #define DIVSTAT_INC(name) \ 130 VNET_PCPUSTAT_ADD(struct divstat, divstat, div_ ## name, 1) 131 132 static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */ 133 static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */ 134 135 static int div_output_inbound(int fmaily, struct socket *so, struct mbuf *m, 136 struct sockaddr_in *sin); 137 static int div_output_outbound(int family, struct socket *so, struct mbuf *m); 138 139 struct divcb { 140 union { 141 SLIST_ENTRY(divcb) dcb_next; 142 intptr_t dcb_bound; 143 #define DCB_UNBOUND ((intptr_t)-1) 144 }; 145 struct socket *dcb_socket; 146 uint16_t dcb_port; 147 uint64_t dcb_gencnt; 148 struct epoch_context dcb_epochctx; 149 }; 150 151 SLIST_HEAD(divhashhead, divcb); 152 153 VNET_DEFINE_STATIC(struct divhashhead, divhash[DIVHASHSIZE]) = {}; 154 #define V_divhash VNET(divhash) 155 VNET_DEFINE_STATIC(uint64_t, dcb_count) = 0; 156 #define V_dcb_count VNET(dcb_count) 157 VNET_DEFINE_STATIC(uint64_t, dcb_gencnt) = 0; 158 #define V_dcb_gencnt VNET(dcb_gencnt) 159 160 static struct mtx divert_mtx; 161 MTX_SYSINIT(divert, &divert_mtx, "divert(4) socket pcb lists", MTX_DEF); 162 #define DIVERT_LOCK() mtx_lock(&divert_mtx) 163 #define DIVERT_UNLOCK() mtx_unlock(&divert_mtx) 164 165 /* 166 * Divert a packet by passing it up to the divert socket at port 'port'. 167 */ 168 static void 169 divert_packet(struct mbuf *m, bool incoming) 170 { 171 struct divcb *dcb; 172 u_int16_t nport; 173 struct sockaddr_in divsrc; 174 struct m_tag *mtag; 175 176 NET_EPOCH_ASSERT(); 177 178 mtag = m_tag_locate(m, MTAG_IPFW_RULE, 0, NULL); 179 if (mtag == NULL) { 180 m_freem(m); 181 return; 182 } 183 /* Assure header */ 184 if (m->m_len < sizeof(struct ip) && 185 (m = m_pullup(m, sizeof(struct ip))) == NULL) 186 return; 187 #ifdef INET 188 /* Delayed checksums are currently not compatible with divert. */ 189 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 190 in_delayed_cksum(m); 191 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 192 } 193 #if defined(SCTP) || defined(SCTP_SUPPORT) 194 if (m->m_pkthdr.csum_flags & CSUM_SCTP) { 195 struct ip *ip; 196 197 ip = mtod(m, struct ip *); 198 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2)); 199 m->m_pkthdr.csum_flags &= ~CSUM_SCTP; 200 } 201 #endif 202 #endif 203 #ifdef INET6 204 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 205 in6_delayed_cksum(m, m->m_pkthdr.len - 206 sizeof(struct ip6_hdr), sizeof(struct ip6_hdr)); 207 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6; 208 } 209 #if defined(SCTP) || defined(SCTP_SUPPORT) 210 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) { 211 sctp_delayed_cksum(m, sizeof(struct ip6_hdr)); 212 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6; 213 } 214 #endif 215 #endif /* INET6 */ 216 bzero(&divsrc, sizeof(divsrc)); 217 divsrc.sin_len = sizeof(divsrc); 218 divsrc.sin_family = AF_INET; 219 /* record matching rule, in host format */ 220 divsrc.sin_port = ((struct ipfw_rule_ref *)(mtag+1))->rulenum; 221 /* 222 * Record receive interface address, if any. 223 * But only for incoming packets. 224 */ 225 if (incoming) { 226 struct ifaddr *ifa; 227 struct ifnet *ifp; 228 229 /* Sanity check */ 230 M_ASSERTPKTHDR(m); 231 232 /* Find IP address for receive interface */ 233 ifp = m->m_pkthdr.rcvif; 234 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 235 if (ifa->ifa_addr->sa_family != AF_INET) 236 continue; 237 divsrc.sin_addr = 238 ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr; 239 break; 240 } 241 } 242 /* 243 * Record the incoming interface name whenever we have one. 244 */ 245 if (m->m_pkthdr.rcvif) { 246 /* 247 * Hide the actual interface name in there in the 248 * sin_zero array. XXX This needs to be moved to a 249 * different sockaddr type for divert, e.g. 250 * sockaddr_div with multiple fields like 251 * sockaddr_dl. Presently we have only 7 bytes 252 * but that will do for now as most interfaces 253 * are 4 or less + 2 or less bytes for unit. 254 * There is probably a faster way of doing this, 255 * possibly taking it from the sockaddr_dl on the iface. 256 * This solves the problem of a P2P link and a LAN interface 257 * having the same address, which can result in the wrong 258 * interface being assigned to the packet when fed back 259 * into the divert socket. Theoretically if the daemon saves 260 * and re-uses the sockaddr_in as suggested in the man pages, 261 * this iface name will come along for the ride. 262 * (see div_output for the other half of this.) 263 */ 264 strlcpy(divsrc.sin_zero, m->m_pkthdr.rcvif->if_xname, 265 sizeof(divsrc.sin_zero)); 266 } 267 268 /* Put packet on socket queue, if any */ 269 nport = htons((uint16_t)(((struct ipfw_rule_ref *)(mtag+1))->info)); 270 SLIST_FOREACH(dcb, &V_divhash[DIVHASH(nport)], dcb_next) 271 if (dcb->dcb_port == nport) 272 break; 273 274 if (dcb != NULL) { 275 struct socket *sa = dcb->dcb_socket; 276 277 SOCKBUF_LOCK(&sa->so_rcv); 278 if (sbappendaddr_locked(&sa->so_rcv, 279 (struct sockaddr *)&divsrc, m, NULL) == 0) { 280 soroverflow_locked(sa); 281 m_freem(m); 282 } else { 283 sorwakeup_locked(sa); 284 DIVSTAT_INC(diverted); 285 } 286 } else { 287 DIVSTAT_INC(noport); 288 m_freem(m); 289 } 290 } 291 292 /* 293 * Deliver packet back into the IP processing machinery. 294 * 295 * If no address specified, or address is 0.0.0.0, send to ip_output(); 296 * otherwise, send to ip_input() and mark as having been received on 297 * the interface with that address. 298 */ 299 static int 300 div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 301 struct mbuf *control, struct thread *td) 302 { 303 struct epoch_tracker et; 304 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 305 const struct ip *ip; 306 struct m_tag *mtag; 307 struct ipfw_rule_ref *dt; 308 int error, family; 309 310 if (control) 311 m_freem(control); 312 313 /* Packet must have a header (but that's about it) */ 314 if (m->m_len < sizeof (struct ip) && 315 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 316 m_freem(m); 317 return (EINVAL); 318 } 319 320 if (sin != NULL) { 321 if (sin->sin_family != AF_INET) { 322 m_freem(m); 323 return (EAFNOSUPPORT); 324 } 325 if (sin->sin_len != sizeof(*sin)) { 326 m_freem(m); 327 return (EINVAL); 328 } 329 } 330 331 /* 332 * An mbuf may hasn't come from userland, but we pretend 333 * that it has. 334 */ 335 m->m_pkthdr.rcvif = NULL; 336 m->m_nextpkt = NULL; 337 M_SETFIB(m, so->so_fibnum); 338 339 mtag = m_tag_locate(m, MTAG_IPFW_RULE, 0, NULL); 340 if (mtag == NULL) { 341 /* this should be normal */ 342 mtag = m_tag_alloc(MTAG_IPFW_RULE, 0, 343 sizeof(struct ipfw_rule_ref), M_NOWAIT | M_ZERO); 344 if (mtag == NULL) { 345 m_freem(m); 346 return (ENOBUFS); 347 } 348 m_tag_prepend(m, mtag); 349 } 350 dt = (struct ipfw_rule_ref *)(mtag+1); 351 352 /* Loopback avoidance and state recovery */ 353 if (sin) { 354 int i; 355 356 /* set the starting point. We provide a non-zero slot, 357 * but a non_matching chain_id to skip that info and use 358 * the rulenum/rule_id. 359 */ 360 dt->slot = 1; /* dummy, chain_id is invalid */ 361 dt->chain_id = 0; 362 dt->rulenum = sin->sin_port+1; /* host format ? */ 363 dt->rule_id = 0; 364 /* XXX: broken for IPv6 */ 365 /* 366 * Find receive interface with the given name, stuffed 367 * (if it exists) in the sin_zero[] field. 368 * The name is user supplied data so don't trust its size 369 * or that it is zero terminated. 370 */ 371 for (i = 0; i < sizeof(sin->sin_zero) && sin->sin_zero[i]; i++) 372 ; 373 if ( i > 0 && i < sizeof(sin->sin_zero)) 374 m->m_pkthdr.rcvif = ifunit(sin->sin_zero); 375 } 376 377 ip = mtod(m, struct ip *); 378 switch (ip->ip_v) { 379 #ifdef INET 380 case IPVERSION: 381 family = AF_INET; 382 break; 383 #endif 384 #ifdef INET6 385 case IPV6_VERSION >> 4: 386 family = AF_INET6; 387 break; 388 #endif 389 default: 390 m_freem(m); 391 return (EAFNOSUPPORT); 392 } 393 394 /* Reinject packet into the system as incoming or outgoing */ 395 NET_EPOCH_ENTER(et); 396 if (!sin || sin->sin_addr.s_addr == 0) { 397 dt->info |= IPFW_IS_DIVERT | IPFW_INFO_OUT; 398 error = div_output_outbound(family, so, m); 399 } else { 400 dt->info |= IPFW_IS_DIVERT | IPFW_INFO_IN; 401 error = div_output_inbound(family, so, m, sin); 402 } 403 NET_EPOCH_EXIT(et); 404 405 return (error); 406 } 407 408 /* 409 * Sends mbuf @m to the wire via ip[6]_output(). 410 * 411 * Returns 0 on success or an errno value on failure. @m is always consumed. 412 */ 413 static int 414 div_output_outbound(int family, struct socket *so, struct mbuf *m) 415 { 416 int error; 417 418 switch (family) { 419 #ifdef INET 420 case AF_INET: 421 { 422 struct ip *const ip = mtod(m, struct ip *); 423 424 /* Don't allow packet length sizes that will crash. */ 425 if (((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) { 426 m_freem(m); 427 return (EINVAL); 428 } 429 break; 430 } 431 #endif 432 #ifdef INET6 433 case AF_INET6: 434 { 435 struct ip6_hdr *const ip6 = mtod(m, struct ip6_hdr *); 436 437 /* Don't allow packet length sizes that will crash */ 438 if (((u_short)ntohs(ip6->ip6_plen) > m->m_pkthdr.len)) { 439 m_freem(m); 440 return (EINVAL); 441 } 442 break; 443 } 444 #endif 445 } 446 447 #ifdef MAC 448 mac_socket_create_mbuf(so, m); 449 #endif 450 451 error = 0; 452 switch (family) { 453 #ifdef INET 454 case AF_INET: 455 error = ip_output(m, NULL, NULL, 456 ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) 457 | IP_ALLOWBROADCAST | IP_RAWOUTPUT, NULL, NULL); 458 break; 459 #endif 460 #ifdef INET6 461 case AF_INET6: 462 error = ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL); 463 break; 464 #endif 465 } 466 if (error == 0) 467 DIVSTAT_INC(outbound); 468 469 return (error); 470 } 471 472 /* 473 * Schedules mbuf @m for local processing via IPv4/IPv6 netisr queue. 474 * 475 * Returns 0 on success or an errno value on failure. @m is always consumed. 476 */ 477 static int 478 div_output_inbound(int family, struct socket *so, struct mbuf *m, 479 struct sockaddr_in *sin) 480 { 481 struct ifaddr *ifa; 482 483 if (m->m_pkthdr.rcvif == NULL) { 484 /* 485 * No luck with the name, check by IP address. 486 * Clear the port and the ifname to make sure 487 * there are no distractions for ifa_ifwithaddr. 488 */ 489 490 /* XXX: broken for IPv6 */ 491 bzero(sin->sin_zero, sizeof(sin->sin_zero)); 492 sin->sin_port = 0; 493 ifa = ifa_ifwithaddr((struct sockaddr *) sin); 494 if (ifa == NULL) { 495 m_freem(m); 496 return (EADDRNOTAVAIL); 497 } 498 m->m_pkthdr.rcvif = ifa->ifa_ifp; 499 } 500 #ifdef MAC 501 mac_socket_create_mbuf(so, m); 502 #endif 503 /* Send packet to input processing via netisr */ 504 switch (family) { 505 #ifdef INET 506 case AF_INET: 507 { 508 const struct ip *ip; 509 510 ip = mtod(m, struct ip *); 511 /* 512 * Restore M_BCAST flag when destination address is 513 * broadcast. It is expected by ip_tryforward(). 514 */ 515 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) 516 m->m_flags |= M_MCAST; 517 else if (in_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) 518 m->m_flags |= M_BCAST; 519 netisr_queue_src(NETISR_IP, (uintptr_t)so, m); 520 DIVSTAT_INC(inbound); 521 break; 522 } 523 #endif 524 #ifdef INET6 525 case AF_INET6: 526 netisr_queue_src(NETISR_IPV6, (uintptr_t)so, m); 527 DIVSTAT_INC(inbound); 528 break; 529 #endif 530 default: 531 m_freem(m); 532 return (EINVAL); 533 } 534 535 return (0); 536 } 537 538 static int 539 div_attach(struct socket *so, int proto, struct thread *td) 540 { 541 struct divcb *dcb; 542 int error; 543 544 if (td != NULL) { 545 error = priv_check(td, PRIV_NETINET_DIVERT); 546 if (error) 547 return (error); 548 } 549 error = soreserve(so, div_sendspace, div_recvspace); 550 if (error) 551 return error; 552 dcb = malloc(sizeof(*dcb), M_PCB, M_WAITOK); 553 dcb->dcb_bound = DCB_UNBOUND; 554 dcb->dcb_socket = so; 555 DIVERT_LOCK(); 556 V_dcb_count++; 557 dcb->dcb_gencnt = ++V_dcb_gencnt; 558 DIVERT_UNLOCK(); 559 so->so_pcb = dcb; 560 561 return (0); 562 } 563 564 static void 565 div_free(epoch_context_t ctx) 566 { 567 struct divcb *dcb = __containerof(ctx, struct divcb, dcb_epochctx); 568 569 free(dcb, M_PCB); 570 } 571 572 static void 573 div_detach(struct socket *so) 574 { 575 struct divcb *dcb = so->so_pcb; 576 577 so->so_pcb = NULL; 578 DIVERT_LOCK(); 579 if (dcb->dcb_bound != DCB_UNBOUND) 580 SLIST_REMOVE(&V_divhash[DCBHASH(dcb)], dcb, divcb, dcb_next); 581 V_dcb_count--; 582 V_dcb_gencnt++; 583 DIVERT_UNLOCK(); 584 NET_EPOCH_CALL(div_free, &dcb->dcb_epochctx); 585 } 586 587 static int 588 div_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 589 { 590 struct divcb *dcb; 591 uint16_t port; 592 593 if (nam->sa_family != AF_INET) 594 return EAFNOSUPPORT; 595 if (nam->sa_len != sizeof(struct sockaddr_in)) 596 return EINVAL; 597 port = ((struct sockaddr_in *)nam)->sin_port; 598 DIVERT_LOCK(); 599 SLIST_FOREACH(dcb, &V_divhash[DIVHASH(port)], dcb_next) 600 if (dcb->dcb_port == port) { 601 DIVERT_UNLOCK(); 602 return (EADDRINUSE); 603 } 604 dcb = so->so_pcb; 605 if (dcb->dcb_bound != DCB_UNBOUND) 606 SLIST_REMOVE(&V_divhash[DCBHASH(dcb)], dcb, divcb, dcb_next); 607 dcb->dcb_port = port; 608 SLIST_INSERT_HEAD(&V_divhash[DIVHASH(port)], dcb, dcb_next); 609 DIVERT_UNLOCK(); 610 611 return (0); 612 } 613 614 static int 615 div_shutdown(struct socket *so) 616 { 617 618 socantsendmore(so); 619 return 0; 620 } 621 622 static int 623 div_pcblist(SYSCTL_HANDLER_ARGS) 624 { 625 struct xinpgen xig; 626 struct divcb *dcb; 627 int error; 628 629 if (req->newptr != 0) 630 return EPERM; 631 632 if (req->oldptr == 0) { 633 u_int n; 634 635 n = V_dcb_count; 636 n += imax(n / 8, 10); 637 req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb); 638 return 0; 639 } 640 641 if ((error = sysctl_wire_old_buffer(req, 0)) != 0) 642 return (error); 643 644 bzero(&xig, sizeof(xig)); 645 xig.xig_len = sizeof xig; 646 xig.xig_count = V_dcb_count; 647 xig.xig_gen = V_dcb_gencnt; 648 xig.xig_sogen = so_gencnt; 649 error = SYSCTL_OUT(req, &xig, sizeof xig); 650 if (error) 651 return error; 652 653 DIVERT_LOCK(); 654 for (int i = 0; i < DIVHASHSIZE; i++) 655 SLIST_FOREACH(dcb, &V_divhash[i], dcb_next) { 656 if (dcb->dcb_gencnt <= xig.xig_gen) { 657 struct xinpcb xi; 658 659 bzero(&xi, sizeof(xi)); 660 xi.xi_len = sizeof(struct xinpcb); 661 sotoxsocket(dcb->dcb_socket, &xi.xi_socket); 662 xi.inp_gencnt = dcb->dcb_gencnt; 663 xi.inp_vflag = INP_IPV4; /* XXX: netstat(1) */ 664 xi.inp_inc.inc_ie.ie_lport = dcb->dcb_port; 665 error = SYSCTL_OUT(req, &xi, sizeof xi); 666 if (error) 667 goto errout; 668 } 669 } 670 671 /* 672 * Give the user an updated idea of our state. 673 * If the generation differs from what we told 674 * her before, she knows that something happened 675 * while we were processing this request, and it 676 * might be necessary to retry. 677 */ 678 xig.xig_gen = V_dcb_gencnt; 679 xig.xig_sogen = so_gencnt; 680 xig.xig_count = V_dcb_count; 681 error = SYSCTL_OUT(req, &xig, sizeof xig); 682 683 errout: 684 DIVERT_UNLOCK(); 685 686 return (error); 687 } 688 SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, 689 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, div_pcblist, 690 "S,xinpcb", "List of active divert sockets"); 691 692 static struct protosw div_protosw = { 693 .pr_type = SOCK_RAW, 694 .pr_flags = PR_ATOMIC|PR_ADDR, 695 .pr_attach = div_attach, 696 .pr_bind = div_bind, 697 .pr_detach = div_detach, 698 .pr_send = div_send, 699 .pr_shutdown = div_shutdown, 700 }; 701 702 static struct domain divertdomain = { 703 .dom_family = PF_DIVERT, 704 .dom_name = "divert", 705 .dom_nprotosw = 1, 706 .dom_protosw = { &div_protosw }, 707 }; 708 709 static int 710 div_modevent(module_t mod, int type, void *unused) 711 { 712 int err = 0; 713 714 switch (type) { 715 case MOD_LOAD: 716 domain_add(&divertdomain); 717 ip_divert_ptr = divert_packet; 718 break; 719 case MOD_QUIESCE: 720 /* 721 * IPDIVERT may normally not be unloaded because of the 722 * potential race conditions. Tell kldunload we can't be 723 * unloaded unless the unload is forced. 724 */ 725 err = EPERM; 726 break; 727 case MOD_UNLOAD: 728 /* 729 * Forced unload. 730 * 731 * Module ipdivert can only be unloaded if no sockets are 732 * connected. Maybe this can be changed later to forcefully 733 * disconnect any open sockets. 734 * 735 * XXXRW: Note that there is a slight race here, as a new 736 * socket open request could be spinning on the lock and then 737 * we destroy the lock. 738 * 739 * XXXGL: One more reason this code is incorrect is that it 740 * checks only the current vnet. 741 */ 742 DIVERT_LOCK(); 743 if (V_dcb_count != 0) { 744 DIVERT_UNLOCK(); 745 err = EBUSY; 746 break; 747 } 748 DIVERT_UNLOCK(); 749 ip_divert_ptr = NULL; 750 domain_remove(&divertdomain); 751 break; 752 default: 753 err = EOPNOTSUPP; 754 break; 755 } 756 return err; 757 } 758 759 static moduledata_t ipdivertmod = { 760 "ipdivert", 761 div_modevent, 762 0 763 }; 764 765 DECLARE_MODULE(ipdivert, ipdivertmod, SI_SUB_PROTO_FIREWALL, SI_ORDER_ANY); 766 MODULE_VERSION(ipdivert, 1); 767