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 "opt_inet.h" 33 #include "opt_inet6.h" 34 #include "opt_sctp.h" 35 36 #include <sys/param.h> 37 #include <sys/ck.h> 38 #include <sys/eventhandler.h> 39 #include <sys/kernel.h> 40 #include <sys/lock.h> 41 #include <sys/malloc.h> 42 #include <sys/mbuf.h> 43 #include <sys/module.h> 44 #include <sys/kernel.h> 45 #include <sys/priv.h> 46 #include <sys/proc.h> 47 #include <sys/domain.h> 48 #include <sys/protosw.h> 49 #include <sys/socket.h> 50 #include <sys/socketvar.h> 51 #include <sys/sysctl.h> 52 #include <net/vnet.h> 53 54 #include <net/if.h> 55 #include <net/if_var.h> 56 #include <net/if_private.h> 57 #include <net/netisr.h> 58 59 #include <netinet/in.h> 60 #include <netinet/in_pcb.h> 61 #include <netinet/in_systm.h> 62 #include <netinet/in_var.h> 63 #include <netinet/ip.h> 64 #include <netinet/ip_var.h> 65 #include <netinet/ip_divert.h> 66 #ifdef INET6 67 #include <netinet/ip6.h> 68 #include <netinet6/ip6_var.h> 69 #endif 70 #if defined(SCTP) || defined(SCTP_SUPPORT) 71 #include <netinet/sctp_crc32.h> 72 #endif 73 74 #include <security/mac/mac_framework.h> 75 /* 76 * Divert sockets 77 */ 78 79 /* 80 * Allocate enough space to hold a full IP packet 81 */ 82 #define DIVSNDQ (65536 + 100) 83 #define DIVRCVQ (65536 + 100) 84 85 /* 86 * Usually a system has very few divert ports. Previous implementation 87 * used a linked list. 88 */ 89 #define DIVHASHSIZE (1 << 3) /* 8 entries, one cache line. */ 90 #define DIVHASH(port) (port % DIVHASHSIZE) 91 #define DCBHASH(dcb) ((dcb)->dcb_port % DIVHASHSIZE) 92 93 /* 94 * Divert sockets work in conjunction with ipfw or other packet filters, 95 * see the divert(4) manpage for features. 96 * Packets are selected by the packet filter and tagged with an 97 * MTAG_IPFW_RULE tag carrying the 'divert port' number (as set by 98 * the packet filter) and information on the matching filter rule for 99 * subsequent reinjection. The divert_port is used to put the packet 100 * on the corresponding divert socket, while the rule number is passed 101 * up (at least partially) as the sin_port in the struct sockaddr. 102 * 103 * Packets written to the divert socket carry in sin_addr a 104 * destination address, and in sin_port the number of the filter rule 105 * after which to continue processing. 106 * If the destination address is INADDR_ANY, the packet is treated as 107 * as outgoing and sent to ip_output(); otherwise it is treated as 108 * incoming and sent to ip_input(). 109 * Further, sin_zero carries some information on the interface, 110 * which can be used in the reinject -- see comments in the code. 111 * 112 * On reinjection, processing in ip_input() and ip_output() 113 * will be exactly the same as for the original packet, except that 114 * packet filter processing will start at the rule number after the one 115 * written in the sin_port (ipfw does not allow a rule #0, so sin_port=0 116 * will apply the entire ruleset to the packet). 117 */ 118 static SYSCTL_NODE(_net_inet, OID_AUTO, divert, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 119 "divert(4)"); 120 121 VNET_PCPUSTAT_DEFINE_STATIC(struct divstat, divstat); 122 VNET_PCPUSTAT_SYSINIT(divstat); 123 #ifdef VIMAGE 124 VNET_PCPUSTAT_SYSUNINIT(divstat); 125 #endif 126 SYSCTL_VNET_PCPUSTAT(_net_inet_divert, OID_AUTO, stats, struct divstat, 127 divstat, "divert(4) socket statistics"); 128 #define DIVSTAT_INC(name) \ 129 VNET_PCPUSTAT_ADD(struct divstat, divstat, div_ ## name, 1) 130 131 static u_long div_sendspace = DIVSNDQ; /* XXX sysctl ? */ 132 static u_long div_recvspace = DIVRCVQ; /* XXX sysctl ? */ 133 134 static int div_output_inbound(int fmaily, struct socket *so, struct mbuf *m, 135 struct sockaddr_in *sin); 136 static int div_output_outbound(int family, struct socket *so, struct mbuf *m); 137 138 struct divcb { 139 union { 140 CK_SLIST_ENTRY(divcb) dcb_next; 141 intptr_t dcb_bound; 142 #define DCB_UNBOUND ((intptr_t)-1) 143 }; 144 struct socket *dcb_socket; 145 uint16_t dcb_port; 146 uint64_t dcb_gencnt; 147 struct epoch_context dcb_epochctx; 148 }; 149 150 CK_SLIST_HEAD(divhashhead, divcb); 151 152 VNET_DEFINE_STATIC(struct divhashhead, divhash[DIVHASHSIZE]) = {}; 153 #define V_divhash VNET(divhash) 154 VNET_DEFINE_STATIC(uint64_t, dcb_count) = 0; 155 #define V_dcb_count VNET(dcb_count) 156 VNET_DEFINE_STATIC(uint64_t, dcb_gencnt) = 0; 157 #define V_dcb_gencnt VNET(dcb_gencnt) 158 159 static struct mtx divert_mtx; 160 MTX_SYSINIT(divert, &divert_mtx, "divert(4) socket pcb lists", MTX_DEF); 161 #define DIVERT_LOCK() mtx_lock(&divert_mtx) 162 #define DIVERT_UNLOCK() mtx_unlock(&divert_mtx) 163 164 /* 165 * Divert a packet by passing it up to the divert socket at port 'port'. 166 */ 167 static void 168 divert_packet(struct mbuf *m, bool incoming) 169 { 170 struct divcb *dcb; 171 u_int16_t nport; 172 struct sockaddr_in divsrc; 173 struct m_tag *mtag; 174 uint16_t cookie; 175 176 NET_EPOCH_ASSERT(); 177 178 mtag = m_tag_locate(m, MTAG_IPFW_RULE, 0, NULL); 179 if (mtag != NULL) { 180 cookie = ((struct ipfw_rule_ref *)(mtag+1))->rulenum; 181 nport = htons((uint16_t) 182 (((struct ipfw_rule_ref *)(mtag+1))->info)); 183 } else if ((mtag = m_tag_locate(m, MTAG_PF_DIVERT, 0, NULL)) != NULL) { 184 cookie = ((struct pf_divert_mtag *)(mtag+1))->idir; 185 nport = htons(((struct pf_divert_mtag *)(mtag+1))->port); 186 } else { 187 m_freem(m); 188 return; 189 } 190 /* Assure header */ 191 if (m->m_len < sizeof(struct ip) && 192 (m = m_pullup(m, sizeof(struct ip))) == NULL) 193 return; 194 #ifdef INET 195 /* Delayed checksums are currently not compatible with divert. */ 196 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 197 in_delayed_cksum(m); 198 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 199 } 200 #if defined(SCTP) || defined(SCTP_SUPPORT) 201 if (m->m_pkthdr.csum_flags & CSUM_SCTP) { 202 struct ip *ip; 203 204 ip = mtod(m, struct ip *); 205 sctp_delayed_cksum(m, (uint32_t)(ip->ip_hl << 2)); 206 m->m_pkthdr.csum_flags &= ~CSUM_SCTP; 207 } 208 #endif 209 #endif 210 #ifdef INET6 211 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6) { 212 in6_delayed_cksum(m, m->m_pkthdr.len - 213 sizeof(struct ip6_hdr), sizeof(struct ip6_hdr)); 214 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6; 215 } 216 #if defined(SCTP) || defined(SCTP_SUPPORT) 217 if (m->m_pkthdr.csum_flags & CSUM_SCTP_IPV6) { 218 sctp_delayed_cksum(m, sizeof(struct ip6_hdr)); 219 m->m_pkthdr.csum_flags &= ~CSUM_SCTP_IPV6; 220 } 221 #endif 222 #endif /* INET6 */ 223 bzero(&divsrc, sizeof(divsrc)); 224 divsrc.sin_len = sizeof(divsrc); 225 divsrc.sin_family = AF_INET; 226 /* record matching rule, in host format */ 227 divsrc.sin_port = cookie; 228 /* 229 * Record receive interface address, if any. 230 * But only for incoming packets. 231 */ 232 if (incoming) { 233 struct ifaddr *ifa; 234 struct ifnet *ifp; 235 236 /* Sanity check */ 237 M_ASSERTPKTHDR(m); 238 239 /* Find IP address for receive interface */ 240 ifp = m->m_pkthdr.rcvif; 241 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 242 if (ifa->ifa_addr->sa_family != AF_INET) 243 continue; 244 divsrc.sin_addr = 245 ((struct sockaddr_in *) ifa->ifa_addr)->sin_addr; 246 break; 247 } 248 } 249 /* 250 * Record the incoming interface name whenever we have one. 251 */ 252 if (m->m_pkthdr.rcvif) { 253 /* 254 * Hide the actual interface name in there in the 255 * sin_zero array. XXX This needs to be moved to a 256 * different sockaddr type for divert, e.g. 257 * sockaddr_div with multiple fields like 258 * sockaddr_dl. Presently we have only 7 bytes 259 * but that will do for now as most interfaces 260 * are 4 or less + 2 or less bytes for unit. 261 * There is probably a faster way of doing this, 262 * possibly taking it from the sockaddr_dl on the iface. 263 * This solves the problem of a P2P link and a LAN interface 264 * having the same address, which can result in the wrong 265 * interface being assigned to the packet when fed back 266 * into the divert socket. Theoretically if the daemon saves 267 * and re-uses the sockaddr_in as suggested in the man pages, 268 * this iface name will come along for the ride. 269 * (see div_output for the other half of this.) 270 */ 271 strlcpy(divsrc.sin_zero, m->m_pkthdr.rcvif->if_xname, 272 sizeof(divsrc.sin_zero)); 273 } 274 275 /* Put packet on socket queue, if any */ 276 CK_SLIST_FOREACH(dcb, &V_divhash[DIVHASH(nport)], dcb_next) 277 if (dcb->dcb_port == nport) 278 break; 279 280 if (dcb != NULL) { 281 struct socket *sa = dcb->dcb_socket; 282 283 SOCKBUF_LOCK(&sa->so_rcv); 284 if (sbappendaddr_locked(&sa->so_rcv, 285 (struct sockaddr *)&divsrc, m, NULL) == 0) { 286 soroverflow_locked(sa); 287 m_freem(m); 288 } else { 289 sorwakeup_locked(sa); 290 DIVSTAT_INC(diverted); 291 } 292 } else { 293 DIVSTAT_INC(noport); 294 m_freem(m); 295 } 296 } 297 298 /* 299 * Deliver packet back into the IP processing machinery. 300 * 301 * If no address specified, or address is 0.0.0.0, send to ip_output(); 302 * otherwise, send to ip_input() and mark as having been received on 303 * the interface with that address. 304 */ 305 static int 306 div_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, 307 struct mbuf *control, struct thread *td) 308 { 309 struct epoch_tracker et; 310 struct sockaddr_in *sin = (struct sockaddr_in *)nam; 311 const struct ip *ip; 312 struct m_tag *mtag; 313 struct ipfw_rule_ref *dt; 314 struct pf_divert_mtag *pfdt; 315 int error, family; 316 317 if (control) 318 m_freem(control); 319 320 /* Packet must have a header (but that's about it) */ 321 if (m->m_len < sizeof (struct ip) && 322 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 323 m_freem(m); 324 return (EINVAL); 325 } 326 327 if (sin != NULL) { 328 if (sin->sin_family != AF_INET) { 329 m_freem(m); 330 return (EAFNOSUPPORT); 331 } 332 if (sin->sin_len != sizeof(*sin)) { 333 m_freem(m); 334 return (EINVAL); 335 } 336 } 337 338 /* 339 * An mbuf may hasn't come from userland, but we pretend 340 * that it has. 341 */ 342 m->m_pkthdr.rcvif = NULL; 343 m->m_nextpkt = NULL; 344 M_SETFIB(m, so->so_fibnum); 345 346 mtag = m_tag_locate(m, MTAG_IPFW_RULE, 0, NULL); 347 if (mtag == NULL) { 348 /* this should be normal */ 349 mtag = m_tag_alloc(MTAG_IPFW_RULE, 0, 350 sizeof(struct ipfw_rule_ref), M_NOWAIT | M_ZERO); 351 if (mtag == NULL) { 352 m_freem(m); 353 return (ENOBUFS); 354 } 355 m_tag_prepend(m, mtag); 356 } 357 dt = (struct ipfw_rule_ref *)(mtag+1); 358 359 /* Loopback avoidance and state recovery */ 360 if (sin) { 361 int i; 362 363 /* set the starting point. We provide a non-zero slot, 364 * but a non_matching chain_id to skip that info and use 365 * the rulenum/rule_id. 366 */ 367 dt->slot = 1; /* dummy, chain_id is invalid */ 368 dt->chain_id = 0; 369 dt->rulenum = sin->sin_port+1; /* host format ? */ 370 dt->rule_id = 0; 371 /* XXX: broken for IPv6 */ 372 /* 373 * Find receive interface with the given name, stuffed 374 * (if it exists) in the sin_zero[] field. 375 * The name is user supplied data so don't trust its size 376 * or that it is zero terminated. 377 */ 378 for (i = 0; i < sizeof(sin->sin_zero) && sin->sin_zero[i]; i++) 379 ; 380 if ( i > 0 && i < sizeof(sin->sin_zero)) 381 m->m_pkthdr.rcvif = ifunit(sin->sin_zero); 382 } 383 384 ip = mtod(m, struct ip *); 385 switch (ip->ip_v) { 386 #ifdef INET 387 case IPVERSION: 388 family = AF_INET; 389 break; 390 #endif 391 #ifdef INET6 392 case IPV6_VERSION >> 4: 393 family = AF_INET6; 394 break; 395 #endif 396 default: 397 m_freem(m); 398 return (EAFNOSUPPORT); 399 } 400 401 mtag = m_tag_locate(m, MTAG_PF_DIVERT, 0, NULL); 402 if (mtag == NULL) { 403 /* this should be normal */ 404 mtag = m_tag_alloc(MTAG_PF_DIVERT, 0, 405 sizeof(struct pf_divert_mtag), M_NOWAIT | M_ZERO); 406 if (mtag == NULL) { 407 m_freem(m); 408 return (ENOBUFS); 409 } 410 m_tag_prepend(m, mtag); 411 } 412 pfdt = (struct pf_divert_mtag *)(mtag+1); 413 if (sin) 414 pfdt->idir = sin->sin_port; 415 416 /* Reinject packet into the system as incoming or outgoing */ 417 NET_EPOCH_ENTER(et); 418 if (!sin || sin->sin_addr.s_addr == 0) { 419 dt->info |= IPFW_IS_DIVERT | IPFW_INFO_OUT; 420 pfdt->ndir = PF_DIVERT_MTAG_DIR_OUT; 421 error = div_output_outbound(family, so, m); 422 } else { 423 dt->info |= IPFW_IS_DIVERT | IPFW_INFO_IN; 424 pfdt->ndir = PF_DIVERT_MTAG_DIR_IN; 425 error = div_output_inbound(family, so, m, sin); 426 } 427 NET_EPOCH_EXIT(et); 428 429 return (error); 430 } 431 432 /* 433 * Sends mbuf @m to the wire via ip[6]_output(). 434 * 435 * Returns 0 on success or an errno value on failure. @m is always consumed. 436 */ 437 static int 438 div_output_outbound(int family, struct socket *so, struct mbuf *m) 439 { 440 int error; 441 442 switch (family) { 443 #ifdef INET 444 case AF_INET: 445 { 446 struct ip *const ip = mtod(m, struct ip *); 447 448 /* Don't allow packet length sizes that will crash. */ 449 if (((u_short)ntohs(ip->ip_len) > m->m_pkthdr.len)) { 450 m_freem(m); 451 return (EINVAL); 452 } 453 break; 454 } 455 #endif 456 #ifdef INET6 457 case AF_INET6: 458 { 459 struct ip6_hdr *const ip6 = mtod(m, struct ip6_hdr *); 460 461 /* Don't allow packet length sizes that will crash */ 462 if (((u_short)ntohs(ip6->ip6_plen) > m->m_pkthdr.len)) { 463 m_freem(m); 464 return (EINVAL); 465 } 466 break; 467 } 468 #endif 469 } 470 471 #ifdef MAC 472 mac_socket_create_mbuf(so, m); 473 #endif 474 475 error = 0; 476 switch (family) { 477 #ifdef INET 478 case AF_INET: 479 error = ip_output(m, NULL, NULL, 480 ((so->so_options & SO_DONTROUTE) ? IP_ROUTETOIF : 0) 481 | IP_ALLOWBROADCAST | IP_RAWOUTPUT, NULL, NULL); 482 break; 483 #endif 484 #ifdef INET6 485 case AF_INET6: 486 error = ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL); 487 break; 488 #endif 489 } 490 if (error == 0) 491 DIVSTAT_INC(outbound); 492 493 return (error); 494 } 495 496 /* 497 * Schedules mbuf @m for local processing via IPv4/IPv6 netisr queue. 498 * 499 * Returns 0 on success or an errno value on failure. @m is always consumed. 500 */ 501 static int 502 div_output_inbound(int family, struct socket *so, struct mbuf *m, 503 struct sockaddr_in *sin) 504 { 505 #if defined(INET) || defined(INET6) 506 struct divcb *dcb = so->so_pcb; 507 #endif 508 struct ifaddr *ifa; 509 510 if (m->m_pkthdr.rcvif == NULL) { 511 /* 512 * No luck with the name, check by IP address. 513 * Clear the port and the ifname to make sure 514 * there are no distractions for ifa_ifwithaddr. 515 */ 516 517 /* XXX: broken for IPv6 */ 518 bzero(sin->sin_zero, sizeof(sin->sin_zero)); 519 sin->sin_port = 0; 520 ifa = ifa_ifwithaddr((struct sockaddr *) sin); 521 if (ifa == NULL) { 522 m_freem(m); 523 return (EADDRNOTAVAIL); 524 } 525 m->m_pkthdr.rcvif = ifa->ifa_ifp; 526 } 527 #ifdef MAC 528 mac_socket_create_mbuf(so, m); 529 #endif 530 /* Send packet to input processing via netisr */ 531 switch (family) { 532 #ifdef INET 533 case AF_INET: 534 { 535 const struct ip *ip; 536 537 ip = mtod(m, struct ip *); 538 /* 539 * Restore M_BCAST flag when destination address is 540 * broadcast. It is expected by ip_tryforward(). 541 */ 542 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) 543 m->m_flags |= M_MCAST; 544 else if (in_ifnet_broadcast(ip->ip_dst, m->m_pkthdr.rcvif)) 545 m->m_flags |= M_BCAST; 546 netisr_queue_src(NETISR_IP, (uintptr_t)dcb->dcb_gencnt, m); 547 DIVSTAT_INC(inbound); 548 break; 549 } 550 #endif 551 #ifdef INET6 552 case AF_INET6: 553 netisr_queue_src(NETISR_IPV6, (uintptr_t)dcb->dcb_gencnt, m); 554 DIVSTAT_INC(inbound); 555 break; 556 #endif 557 default: 558 m_freem(m); 559 return (EINVAL); 560 } 561 562 return (0); 563 } 564 565 static int 566 div_attach(struct socket *so, int proto, struct thread *td) 567 { 568 struct divcb *dcb; 569 int error; 570 571 if (td != NULL) { 572 error = priv_check(td, PRIV_NETINET_DIVERT); 573 if (error) 574 return (error); 575 } 576 error = soreserve(so, div_sendspace, div_recvspace); 577 if (error) 578 return error; 579 dcb = malloc(sizeof(*dcb), M_PCB, M_WAITOK); 580 dcb->dcb_bound = DCB_UNBOUND; 581 dcb->dcb_socket = so; 582 DIVERT_LOCK(); 583 V_dcb_count++; 584 dcb->dcb_gencnt = ++V_dcb_gencnt; 585 DIVERT_UNLOCK(); 586 so->so_pcb = dcb; 587 588 return (0); 589 } 590 591 static void 592 div_free(epoch_context_t ctx) 593 { 594 struct divcb *dcb = __containerof(ctx, struct divcb, dcb_epochctx); 595 596 free(dcb, M_PCB); 597 } 598 599 static void 600 div_detach(struct socket *so) 601 { 602 struct divcb *dcb = so->so_pcb; 603 604 so->so_pcb = NULL; 605 DIVERT_LOCK(); 606 if (dcb->dcb_bound != DCB_UNBOUND) 607 CK_SLIST_REMOVE(&V_divhash[DCBHASH(dcb)], dcb, divcb, dcb_next); 608 V_dcb_count--; 609 V_dcb_gencnt++; 610 DIVERT_UNLOCK(); 611 NET_EPOCH_CALL(div_free, &dcb->dcb_epochctx); 612 } 613 614 static int 615 div_bind(struct socket *so, struct sockaddr *nam, struct thread *td) 616 { 617 struct divcb *dcb; 618 uint16_t port; 619 620 if (nam->sa_family != AF_INET) 621 return EAFNOSUPPORT; 622 if (nam->sa_len != sizeof(struct sockaddr_in)) 623 return EINVAL; 624 port = ((struct sockaddr_in *)nam)->sin_port; 625 DIVERT_LOCK(); 626 CK_SLIST_FOREACH(dcb, &V_divhash[DIVHASH(port)], dcb_next) 627 if (dcb->dcb_port == port) { 628 DIVERT_UNLOCK(); 629 return (EADDRINUSE); 630 } 631 dcb = so->so_pcb; 632 if (dcb->dcb_bound != DCB_UNBOUND) 633 CK_SLIST_REMOVE(&V_divhash[DCBHASH(dcb)], dcb, divcb, dcb_next); 634 dcb->dcb_port = port; 635 CK_SLIST_INSERT_HEAD(&V_divhash[DIVHASH(port)], dcb, dcb_next); 636 DIVERT_UNLOCK(); 637 638 return (0); 639 } 640 641 static int 642 div_pcblist(SYSCTL_HANDLER_ARGS) 643 { 644 struct xinpgen xig; 645 struct divcb *dcb; 646 int error; 647 648 if (req->newptr != 0) 649 return EPERM; 650 651 if (req->oldptr == 0) { 652 u_int n; 653 654 n = V_dcb_count; 655 n += imax(n / 8, 10); 656 req->oldidx = 2 * (sizeof xig) + n * sizeof(struct xinpcb); 657 return 0; 658 } 659 660 if ((error = sysctl_wire_old_buffer(req, 0)) != 0) 661 return (error); 662 663 bzero(&xig, sizeof(xig)); 664 xig.xig_len = sizeof xig; 665 xig.xig_count = V_dcb_count; 666 xig.xig_gen = V_dcb_gencnt; 667 xig.xig_sogen = so_gencnt; 668 error = SYSCTL_OUT(req, &xig, sizeof xig); 669 if (error) 670 return error; 671 672 DIVERT_LOCK(); 673 for (int i = 0; i < DIVHASHSIZE; i++) 674 CK_SLIST_FOREACH(dcb, &V_divhash[i], dcb_next) { 675 if (dcb->dcb_gencnt <= xig.xig_gen) { 676 struct xinpcb xi; 677 678 bzero(&xi, sizeof(xi)); 679 xi.xi_len = sizeof(struct xinpcb); 680 sotoxsocket(dcb->dcb_socket, &xi.xi_socket); 681 xi.inp_gencnt = dcb->dcb_gencnt; 682 xi.inp_vflag = INP_IPV4; /* XXX: netstat(1) */ 683 xi.inp_inc.inc_ie.ie_lport = dcb->dcb_port; 684 error = SYSCTL_OUT(req, &xi, sizeof xi); 685 if (error) 686 goto errout; 687 } 688 } 689 690 /* 691 * Give the user an updated idea of our state. 692 * If the generation differs from what we told 693 * her before, she knows that something happened 694 * while we were processing this request, and it 695 * might be necessary to retry. 696 */ 697 xig.xig_gen = V_dcb_gencnt; 698 xig.xig_sogen = so_gencnt; 699 xig.xig_count = V_dcb_count; 700 error = SYSCTL_OUT(req, &xig, sizeof xig); 701 702 errout: 703 DIVERT_UNLOCK(); 704 705 return (error); 706 } 707 SYSCTL_PROC(_net_inet_divert, OID_AUTO, pcblist, 708 CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0, div_pcblist, 709 "S,xinpcb", "List of active divert sockets"); 710 711 static struct protosw div_protosw = { 712 .pr_type = SOCK_RAW, 713 .pr_flags = PR_ATOMIC|PR_ADDR, 714 .pr_attach = div_attach, 715 .pr_bind = div_bind, 716 .pr_detach = div_detach, 717 .pr_send = div_send, 718 }; 719 720 static struct domain divertdomain = { 721 .dom_family = PF_DIVERT, 722 .dom_name = "divert", 723 .dom_nprotosw = 1, 724 .dom_protosw = { &div_protosw }, 725 }; 726 727 static int 728 div_modevent(module_t mod, int type, void *unused) 729 { 730 int err = 0; 731 732 switch (type) { 733 case MOD_LOAD: 734 domain_add(&divertdomain); 735 ip_divert_ptr = divert_packet; 736 break; 737 case MOD_QUIESCE: 738 /* 739 * IPDIVERT may normally not be unloaded because of the 740 * potential race conditions. Tell kldunload we can't be 741 * unloaded unless the unload is forced. 742 */ 743 err = EPERM; 744 break; 745 case MOD_UNLOAD: 746 /* 747 * Forced unload. 748 * 749 * Module ipdivert can only be unloaded if no sockets are 750 * connected. Maybe this can be changed later to forcefully 751 * disconnect any open sockets. 752 * 753 * XXXRW: Note that there is a slight race here, as a new 754 * socket open request could be spinning on the lock and then 755 * we destroy the lock. 756 * 757 * XXXGL: One more reason this code is incorrect is that it 758 * checks only the current vnet. 759 */ 760 DIVERT_LOCK(); 761 if (V_dcb_count != 0) { 762 DIVERT_UNLOCK(); 763 err = EBUSY; 764 break; 765 } 766 DIVERT_UNLOCK(); 767 ip_divert_ptr = NULL; 768 domain_remove(&divertdomain); 769 break; 770 default: 771 err = EOPNOTSUPP; 772 break; 773 } 774 return err; 775 } 776 777 static moduledata_t ipdivertmod = { 778 "ipdivert", 779 div_modevent, 780 0 781 }; 782 783 DECLARE_MODULE(ipdivert, ipdivertmod, SI_SUB_PROTO_FIREWALL, SI_ORDER_ANY); 784 MODULE_VERSION(ipdivert, 1); 785