1 /* $NetBSD: rtsock_shared.c,v 1.17 2020/03/13 16:37:12 christos Exp $ */ 2 3 /* 4 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 5 * 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 project 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 PROJECT 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 PROJECT 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 /* 33 * Copyright (c) 1988, 1991, 1993 34 * The Regents of the University of California. All rights reserved. 35 * 36 * Redistribution and use in source and binary forms, with or without 37 * modification, are permitted provided that the following conditions 38 * are met: 39 * 1. Redistributions of source code must retain the above copyright 40 * notice, this list of conditions and the following disclaimer. 41 * 2. Redistributions in binary form must reproduce the above copyright 42 * notice, this list of conditions and the following disclaimer in the 43 * documentation and/or other materials provided with the distribution. 44 * 3. Neither the name of the University nor the names of its contributors 45 * may be used to endorse or promote products derived from this software 46 * without specific prior written permission. 47 * 48 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 49 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 50 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 51 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 52 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 53 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 54 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 55 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 56 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 57 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 * 60 * @(#)rtsock.c 8.7 (Berkeley) 10/12/95 61 */ 62 63 #include <sys/cdefs.h> 64 __KERNEL_RCSID(0, "$NetBSD: rtsock_shared.c,v 1.17 2020/03/13 16:37:12 christos Exp $"); 65 66 #ifdef _KERNEL_OPT 67 #include "opt_inet.h" 68 #include "opt_net_mpsafe.h" 69 #endif 70 71 #include <sys/param.h> 72 #include <sys/systm.h> 73 #include <sys/proc.h> 74 #include <sys/socket.h> 75 #include <sys/socketvar.h> 76 #include <sys/domain.h> 77 #include <sys/protosw.h> 78 #include <sys/sysctl.h> 79 #include <sys/kauth.h> 80 #include <sys/kmem.h> 81 #include <sys/intr.h> 82 #include <sys/condvar.h> 83 #include <sys/compat_stub.h> 84 85 #include <net/if.h> 86 #include <net/if_llatbl.h> 87 #include <net/if_types.h> 88 #include <net/route.h> 89 #include <net/raw_cb.h> 90 91 #include <netinet/in_var.h> 92 #include <netinet/if_inarp.h> 93 94 #include <netmpls/mpls.h> 95 96 #include <compat/net/if.h> 97 #include <compat/net/route.h> 98 99 #ifdef COMPAT_RTSOCK 100 /* 101 * These are used when #include-d from compat/common/rtsock_50.c 102 */ 103 #define RTM_XVERSION RTM_OVERSION 104 #define RTM_XNEWADDR RTM_ONEWADDR 105 #define RTM_XDELADDR RTM_ODELADDR 106 #define RTM_XCHGADDR RTM_OCHGADDR 107 #define RT_XADVANCE(a,b) RT_OADVANCE(a,b) 108 #define RT_XROUNDUP(n) RT_OROUNDUP(n) 109 #define PF_XROUTE PF_OROUTE 110 #define rt_xmsghdr rt_msghdr50 111 #define if_xmsghdr if_msghdr /* if_msghdr50 is for RTM_OIFINFO */ 112 #define ifa_xmsghdr ifa_msghdr50 113 #define if_xannouncemsghdr if_announcemsghdr50 114 #define COMPATNAME(x) compat_50_ ## x 115 #define DOMAINNAME "oroute" 116 #define COMPATCALL(name, args) \ 117 MODULE_HOOK_CALL_VOID(rtsock_ ## name ## _50_hook, args, __nothing); 118 #define RTS_CTASSERT(x) __nothing 119 CTASSERT(sizeof(struct ifa_xmsghdr) == 20); 120 DOMAIN_DEFINE(compat_50_routedomain); /* forward declare and add to link set */ 121 #else /* COMPAT_RTSOCK */ 122 /* 123 * These are used when #include-d from compat/common/rtsock_50.c 124 */ 125 #define RTM_XVERSION RTM_VERSION 126 #define RTM_XNEWADDR RTM_NEWADDR 127 #define RTM_XDELADDR RTM_DELADDR 128 #define RTM_XCHGADDR RTM_CHGADDR 129 #define RT_XADVANCE(a,b) RT_ADVANCE(a,b) 130 #define RT_XROUNDUP(n) RT_ROUNDUP(n) 131 #define PF_XROUTE PF_ROUTE 132 #define rt_xmsghdr rt_msghdr 133 #define if_xmsghdr if_msghdr 134 #define ifa_xmsghdr ifa_msghdr 135 #define if_xannouncemsghdr if_announcemsghdr 136 #define COMPATNAME(x) x 137 #define DOMAINNAME "route" 138 #define COMPATCALL(name, args) __nothing; 139 #define RTS_CTASSERT(x) CTASSERT(x) 140 CTASSERT(sizeof(struct ifa_xmsghdr) == 32); 141 DOMAIN_DEFINE(routedomain); /* forward declare and add to link set */ 142 #endif /* COMPAT_RTSOCK */ 143 144 #ifdef RTSOCK_DEBUG 145 #define RT_IN_PRINT(info, b, a) (in_print((b), sizeof(b), \ 146 &((const struct sockaddr_in *)(info)->rti_info[(a)])->sin_addr), (b)) 147 #endif /* RTSOCK_DEBUG */ 148 149 struct route_info COMPATNAME(route_info) = { 150 .ri_dst = { .sa_len = 2, .sa_family = PF_XROUTE, }, 151 .ri_src = { .sa_len = 2, .sa_family = PF_XROUTE, }, 152 .ri_maxqlen = IFQ_MAXLEN, 153 }; 154 155 static void COMPATNAME(route_init)(void); 156 static int COMPATNAME(route_output)(struct mbuf *, struct socket *); 157 158 static int rt_xaddrs(u_char, const char *, const char *, struct rt_addrinfo *); 159 static struct mbuf *rt_makeifannouncemsg(struct ifnet *, int, int, 160 struct rt_addrinfo *); 161 static int rt_msg2(int, struct rt_addrinfo *, void *, struct rt_walkarg *, int *); 162 static void _rt_setmetrics(int, const struct rt_xmsghdr *, struct rtentry *); 163 static void rtm_setmetrics(const struct rtentry *, struct rt_xmsghdr *); 164 static void rt_adjustcount(int, int); 165 166 static const struct protosw COMPATNAME(route_protosw)[]; 167 168 struct routecb { 169 struct rawcb rocb_rcb; 170 unsigned int rocb_msgfilter; 171 #define RTMSGFILTER(m) (1U << (m)) 172 char *rocb_missfilter; 173 size_t rocb_missfilterlen; 174 }; 175 #define sotoroutecb(so) ((struct routecb *)(so)->so_pcb) 176 177 static struct rawcbhead rt_rawcb; 178 #ifdef NET_MPSAFE 179 static kmutex_t *rt_so_mtx; 180 181 static bool rt_updating = false; 182 static kcondvar_t rt_update_cv; 183 #endif 184 185 static void 186 rt_adjustcount(int af, int cnt) 187 { 188 struct route_cb * const cb = &COMPATNAME(route_info).ri_cb; 189 190 cb->any_count += cnt; 191 192 switch (af) { 193 case AF_INET: 194 cb->ip_count += cnt; 195 return; 196 #ifdef INET6 197 case AF_INET6: 198 cb->ip6_count += cnt; 199 return; 200 #endif 201 case AF_MPLS: 202 cb->mpls_count += cnt; 203 return; 204 } 205 } 206 207 static int 208 COMPATNAME(route_filter)(struct mbuf *m, struct sockproto *proto, 209 struct rawcb *rp) 210 { 211 struct routecb *rop = (struct routecb *)rp; 212 struct rt_xmsghdr *rtm; 213 214 KASSERT(m != NULL); 215 KASSERT(proto != NULL); 216 KASSERT(rp != NULL); 217 218 /* Wrong family for this socket. */ 219 if (proto->sp_family != PF_ROUTE) 220 return ENOPROTOOPT; 221 222 /* If no filter set, just return. */ 223 if (rop->rocb_msgfilter == 0 && rop->rocb_missfilterlen == 0) 224 return 0; 225 226 /* Ensure we can access rtm_type */ 227 if (m->m_len < 228 offsetof(struct rt_xmsghdr, rtm_type) + sizeof(rtm->rtm_type)) 229 return EINVAL; 230 231 rtm = mtod(m, struct rt_xmsghdr *); 232 if (rtm->rtm_type >= sizeof(rop->rocb_msgfilter) * CHAR_BIT) 233 return EINVAL; 234 /* If the rtm type is filtered out, return a positive. */ 235 if (rop->rocb_msgfilter != 0 && 236 !(rop->rocb_msgfilter & RTMSGFILTER(rtm->rtm_type))) 237 return EEXIST; 238 239 if (rop->rocb_missfilterlen != 0 && rtm->rtm_type == RTM_MISS) { 240 __CTASSERT(RTAX_DST == 0); 241 struct sockaddr_storage ss; 242 struct sockaddr *dst = (struct sockaddr *)&ss, *sa; 243 char *cp = rop->rocb_missfilter; 244 char *ep = cp + rop->rocb_missfilterlen; 245 246 /* Ensure we can access sa_len */ 247 if (m->m_pkthdr.len < sizeof(*rtm) + 248 offsetof(struct sockaddr, sa_len) + sizeof(ss.ss_len)) 249 return EINVAL; 250 m_copydata(m, sizeof(*rtm) + offsetof(struct sockaddr, sa_len), 251 sizeof(ss.ss_len), &ss.ss_len); 252 if (m->m_pkthdr.len < sizeof(*rtm) + ss.ss_len) 253 return EINVAL; 254 /* Copy out the destination sockaddr */ 255 m_copydata(m, sizeof(*rtm), ss.ss_len, &ss); 256 257 /* Find a matching sockaddr in the filter */ 258 while (cp < ep) { 259 sa = (struct sockaddr *)cp; 260 if (sa->sa_len == dst->sa_len && 261 memcmp(sa, dst, sa->sa_len) == 0) 262 break; 263 cp += RT_XROUNDUP(sa->sa_len); 264 } 265 if (cp == ep) 266 return EEXIST; 267 } 268 269 /* Passed the filter. */ 270 return 0; 271 } 272 273 static void 274 rt_pr_init(void) 275 { 276 277 LIST_INIT(&rt_rawcb); 278 } 279 280 static int 281 COMPATNAME(route_attach)(struct socket *so, int proto) 282 { 283 struct rawcb *rp; 284 struct routecb *rop; 285 int s, error; 286 287 KASSERT(sotorawcb(so) == NULL); 288 rop = kmem_zalloc(sizeof(*rop), KM_SLEEP); 289 rp = &rop->rocb_rcb; 290 rp->rcb_len = sizeof(*rop); 291 so->so_pcb = rp; 292 293 s = splsoftnet(); 294 295 #ifdef NET_MPSAFE 296 KASSERT(so->so_lock == NULL); 297 mutex_obj_hold(rt_so_mtx); 298 so->so_lock = rt_so_mtx; 299 solock(so); 300 #endif 301 302 if ((error = raw_attach(so, proto, &rt_rawcb)) == 0) { 303 rt_adjustcount(rp->rcb_proto.sp_protocol, 1); 304 rp->rcb_laddr = &COMPATNAME(route_info).ri_src; 305 rp->rcb_faddr = &COMPATNAME(route_info).ri_dst; 306 rp->rcb_filter = COMPATNAME(route_filter); 307 } 308 splx(s); 309 310 if (error) { 311 kmem_free(rop, sizeof(*rop)); 312 so->so_pcb = NULL; 313 return error; 314 } 315 316 soisconnected(so); 317 so->so_options |= SO_USELOOPBACK; 318 KASSERT(solocked(so)); 319 320 return error; 321 } 322 323 static void 324 COMPATNAME(route_detach)(struct socket *so) 325 { 326 struct rawcb *rp = sotorawcb(so); 327 struct routecb *rop = (struct routecb *)rp; 328 int s; 329 330 KASSERT(rp != NULL); 331 KASSERT(solocked(so)); 332 333 s = splsoftnet(); 334 if (rop->rocb_missfilterlen != 0) 335 kmem_free(rop->rocb_missfilter, rop->rocb_missfilterlen); 336 rt_adjustcount(rp->rcb_proto.sp_protocol, -1); 337 raw_detach(so); 338 splx(s); 339 } 340 341 static int 342 COMPATNAME(route_accept)(struct socket *so, struct sockaddr *nam) 343 { 344 KASSERT(solocked(so)); 345 346 panic("route_accept"); 347 348 return EOPNOTSUPP; 349 } 350 351 static int 352 COMPATNAME(route_bind)(struct socket *so, struct sockaddr *nam, struct lwp *l) 353 { 354 KASSERT(solocked(so)); 355 356 return EOPNOTSUPP; 357 } 358 359 static int 360 COMPATNAME(route_listen)(struct socket *so, struct lwp *l) 361 { 362 KASSERT(solocked(so)); 363 364 return EOPNOTSUPP; 365 } 366 367 static int 368 COMPATNAME(route_connect)(struct socket *so, struct sockaddr *nam, struct lwp *l) 369 { 370 KASSERT(solocked(so)); 371 372 return EOPNOTSUPP; 373 } 374 375 static int 376 COMPATNAME(route_connect2)(struct socket *so, struct socket *so2) 377 { 378 KASSERT(solocked(so)); 379 380 return EOPNOTSUPP; 381 } 382 383 static int 384 COMPATNAME(route_disconnect)(struct socket *so) 385 { 386 struct rawcb *rp = sotorawcb(so); 387 int s; 388 389 KASSERT(solocked(so)); 390 KASSERT(rp != NULL); 391 392 s = splsoftnet(); 393 soisdisconnected(so); 394 raw_disconnect(rp); 395 splx(s); 396 397 return 0; 398 } 399 400 static int 401 COMPATNAME(route_shutdown)(struct socket *so) 402 { 403 int s; 404 405 KASSERT(solocked(so)); 406 407 /* 408 * Mark the connection as being incapable of further input. 409 */ 410 s = splsoftnet(); 411 socantsendmore(so); 412 splx(s); 413 return 0; 414 } 415 416 static int 417 COMPATNAME(route_abort)(struct socket *so) 418 { 419 KASSERT(solocked(so)); 420 421 panic("route_abort"); 422 423 return EOPNOTSUPP; 424 } 425 426 static int 427 COMPATNAME(route_ioctl)(struct socket *so, u_long cmd, void *nam, 428 struct ifnet * ifp) 429 { 430 return EOPNOTSUPP; 431 } 432 433 static int 434 COMPATNAME(route_stat)(struct socket *so, struct stat *ub) 435 { 436 KASSERT(solocked(so)); 437 438 return 0; 439 } 440 441 static int 442 COMPATNAME(route_peeraddr)(struct socket *so, struct sockaddr *nam) 443 { 444 struct rawcb *rp = sotorawcb(so); 445 446 KASSERT(solocked(so)); 447 KASSERT(rp != NULL); 448 KASSERT(nam != NULL); 449 450 if (rp->rcb_faddr == NULL) 451 return ENOTCONN; 452 453 raw_setpeeraddr(rp, nam); 454 return 0; 455 } 456 457 static int 458 COMPATNAME(route_sockaddr)(struct socket *so, struct sockaddr *nam) 459 { 460 struct rawcb *rp = sotorawcb(so); 461 462 KASSERT(solocked(so)); 463 KASSERT(rp != NULL); 464 KASSERT(nam != NULL); 465 466 if (rp->rcb_faddr == NULL) 467 return ENOTCONN; 468 469 raw_setsockaddr(rp, nam); 470 return 0; 471 } 472 473 static int 474 COMPATNAME(route_rcvd)(struct socket *so, int flags, struct lwp *l) 475 { 476 KASSERT(solocked(so)); 477 478 return EOPNOTSUPP; 479 } 480 481 static int 482 COMPATNAME(route_recvoob)(struct socket *so, struct mbuf *m, int flags) 483 { 484 KASSERT(solocked(so)); 485 486 return EOPNOTSUPP; 487 } 488 489 static int 490 COMPATNAME(route_send)(struct socket *so, struct mbuf *m, 491 struct sockaddr *nam, struct mbuf *control, struct lwp *l) 492 { 493 int error = 0; 494 int s; 495 496 KASSERT(solocked(so)); 497 KASSERT(so->so_proto == &COMPATNAME(route_protosw)[0]); 498 499 s = splsoftnet(); 500 error = raw_send(so, m, nam, control, l, &COMPATNAME(route_output)); 501 splx(s); 502 503 return error; 504 } 505 506 static int 507 COMPATNAME(route_sendoob)(struct socket *so, struct mbuf *m, 508 struct mbuf *control) 509 { 510 KASSERT(solocked(so)); 511 512 m_freem(m); 513 m_freem(control); 514 515 return EOPNOTSUPP; 516 } 517 static int 518 COMPATNAME(route_purgeif)(struct socket *so, struct ifnet *ifp) 519 { 520 521 panic("route_purgeif"); 522 523 return EOPNOTSUPP; 524 } 525 526 #if defined(INET) || defined(INET6) 527 static int 528 route_get_sdl_index(struct rt_addrinfo *info, int *sdl_index) 529 { 530 struct rtentry *nrt; 531 int error; 532 533 error = rtrequest1(RTM_GET, info, &nrt); 534 if (error != 0) 535 return error; 536 /* 537 * nrt->rt_ifp->if_index may not be correct 538 * due to changing to ifplo0. 539 */ 540 *sdl_index = satosdl(nrt->rt_gateway)->sdl_index; 541 rt_unref(nrt); 542 543 return 0; 544 } 545 #endif 546 547 static void 548 route_get_sdl(const struct ifnet *ifp, const struct sockaddr *dst, 549 struct sockaddr_dl *sdl, int *flags) 550 { 551 struct llentry *la; 552 553 KASSERT(ifp != NULL); 554 555 IF_AFDATA_RLOCK(ifp); 556 switch (dst->sa_family) { 557 case AF_INET: 558 la = lla_lookup(LLTABLE(ifp), 0, dst); 559 break; 560 case AF_INET6: 561 la = lla_lookup(LLTABLE6(ifp), 0, dst); 562 break; 563 default: 564 la = NULL; 565 KASSERTMSG(0, "Invalid AF=%d\n", dst->sa_family); 566 break; 567 } 568 IF_AFDATA_RUNLOCK(ifp); 569 570 void *a = (LLE_IS_VALID(la) && (la->la_flags & LLE_VALID) == LLE_VALID) 571 ? &la->ll_addr : NULL; 572 573 a = sockaddr_dl_init(sdl, sizeof(*sdl), ifp->if_index, ifp->if_type, 574 NULL, 0, a, ifp->if_addrlen); 575 KASSERT(a != NULL); 576 577 if (la != NULL) { 578 *flags = la->la_flags; 579 LLE_RUNLOCK(la); 580 } 581 } 582 583 static int 584 route_output_report(struct rtentry *rt, struct rt_addrinfo *info, 585 struct rt_xmsghdr *rtm, struct rt_xmsghdr **new_rtm) 586 { 587 int len, error; 588 589 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) { 590 const struct ifaddr *rtifa; 591 const struct ifnet *ifp = rt->rt_ifp; 592 593 info->rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr; 594 /* rtifa used to be simply rt->rt_ifa. 595 * If rt->rt_ifa != NULL, then 596 * rt_get_ifa() != NULL. So this 597 * ought to still be safe. --dyoung 598 */ 599 rtifa = rt_get_ifa(rt); 600 info->rti_info[RTAX_IFA] = rtifa->ifa_addr; 601 #ifdef RTSOCK_DEBUG 602 if (info->rti_info[RTAX_IFA]->sa_family == AF_INET) { 603 char ibuf[INET_ADDRSTRLEN]; 604 char abuf[INET_ADDRSTRLEN]; 605 printf("%s: copying out RTAX_IFA %s " 606 "for info->rti_info[RTAX_DST] %s " 607 "ifa_getifa %p ifa_seqno %p\n", 608 __func__, 609 RT_IN_PRINT(info, ibuf, RTAX_IFA), 610 RT_IN_PRINT(info, abuf, RTAX_DST), 611 (void *)rtifa->ifa_getifa, 612 rtifa->ifa_seqno); 613 } 614 #endif /* RTSOCK_DEBUG */ 615 if (ifp->if_flags & IFF_POINTOPOINT) 616 info->rti_info[RTAX_BRD] = rtifa->ifa_dstaddr; 617 else 618 info->rti_info[RTAX_BRD] = NULL; 619 rtm->rtm_index = ifp->if_index; 620 } 621 error = rt_msg2(rtm->rtm_type, info, NULL, NULL, &len); 622 if (error) 623 return error; 624 if (len > rtm->rtm_msglen) { 625 struct rt_xmsghdr *old_rtm = rtm; 626 R_Malloc(*new_rtm, struct rt_xmsghdr *, len); 627 if (*new_rtm == NULL) 628 return ENOBUFS; 629 (void)memcpy(*new_rtm, old_rtm, old_rtm->rtm_msglen); 630 rtm = *new_rtm; 631 } 632 (void)rt_msg2(rtm->rtm_type, info, rtm, NULL, 0); 633 rtm->rtm_flags = rt->rt_flags; 634 rtm_setmetrics(rt, rtm); 635 rtm->rtm_addrs = info->rti_addrs; 636 637 return 0; 638 } 639 640 /*ARGSUSED*/ 641 int 642 COMPATNAME(route_output)(struct mbuf *m, struct socket *so) 643 { 644 struct sockproto proto = { .sp_family = PF_XROUTE, }; 645 struct rt_xmsghdr *rtm = NULL; 646 struct rt_xmsghdr *old_rtm = NULL, *new_rtm = NULL; 647 struct rtentry *rt = NULL; 648 struct rtentry *saved_nrt = NULL; 649 struct rt_addrinfo info; 650 int len, error = 0; 651 sa_family_t family; 652 struct sockaddr_dl sdl; 653 int bound = curlwp_bind(); 654 bool do_rt_free = false; 655 struct sockaddr_storage netmask; 656 657 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0) 658 if (m == NULL || ((m->m_len < sizeof(int32_t)) && 659 (m = m_pullup(m, sizeof(int32_t))) == NULL)) { 660 error = ENOBUFS; 661 goto out; 662 } 663 if ((m->m_flags & M_PKTHDR) == 0) 664 panic("%s", __func__); 665 len = m->m_pkthdr.len; 666 if (len < sizeof(*rtm) || 667 len != mtod(m, struct rt_xmsghdr *)->rtm_msglen) { 668 info.rti_info[RTAX_DST] = NULL; 669 senderr(EINVAL); 670 } 671 R_Malloc(rtm, struct rt_xmsghdr *, len); 672 if (rtm == NULL) { 673 info.rti_info[RTAX_DST] = NULL; 674 senderr(ENOBUFS); 675 } 676 m_copydata(m, 0, len, rtm); 677 if (rtm->rtm_version != RTM_XVERSION) { 678 info.rti_info[RTAX_DST] = NULL; 679 senderr(EPROTONOSUPPORT); 680 } 681 rtm->rtm_pid = curproc->p_pid; 682 memset(&info, 0, sizeof(info)); 683 info.rti_addrs = rtm->rtm_addrs; 684 if (rt_xaddrs(rtm->rtm_type, (const char *)(rtm + 1), len + (char *)rtm, 685 &info)) { 686 senderr(EINVAL); 687 } 688 info.rti_flags = rtm->rtm_flags; 689 if (info.rti_info[RTAX_DST] == NULL || 690 (info.rti_info[RTAX_DST]->sa_family >= AF_MAX)) { 691 senderr(EINVAL); 692 } 693 #ifdef RTSOCK_DEBUG 694 if (info.rti_info[RTAX_DST]->sa_family == AF_INET) { 695 char abuf[INET_ADDRSTRLEN]; 696 printf("%s: extracted info.rti_info[RTAX_DST] %s\n", __func__, 697 RT_IN_PRINT(&info, abuf, RTAX_DST)); 698 } 699 #endif /* RTSOCK_DEBUG */ 700 if (info.rti_info[RTAX_GATEWAY] != NULL && 701 (info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) { 702 senderr(EINVAL); 703 } 704 705 /* 706 * Verify that the socket has the appropriate privilege; RTM_GET 707 * is the only operation the non-superuser is allowed. 708 */ 709 if (kauth_authorize_network(so->so_cred, KAUTH_NETWORK_ROUTE, 710 0, rtm, NULL, NULL) != 0) 711 senderr(EACCES); 712 713 /* 714 * route(8) passes a sockaddr truncated with prefixlen. 715 * The kernel doesn't expect such sockaddr and need to 716 * use a buffer that is big enough for the sockaddr expected 717 * (padded with 0's). We keep the original length of the sockaddr. 718 */ 719 if (info.rti_info[RTAX_NETMASK]) { 720 /* 721 * Use the family of RTAX_DST, because RTAX_NETMASK 722 * can have a zero family if it comes from the radix 723 * tree via rt_mask(). 724 */ 725 socklen_t sa_len = sockaddr_getsize_by_family( 726 info.rti_info[RTAX_DST]->sa_family); 727 socklen_t masklen = sockaddr_getlen( 728 info.rti_info[RTAX_NETMASK]); 729 if (sa_len != 0 && sa_len > masklen) { 730 KASSERT(sa_len <= sizeof(netmask)); 731 memcpy(&netmask, info.rti_info[RTAX_NETMASK], masklen); 732 memset((char *)&netmask + masklen, 0, sa_len - masklen); 733 info.rti_info[RTAX_NETMASK] = sstocsa(&netmask); 734 } 735 } 736 737 switch (rtm->rtm_type) { 738 739 case RTM_ADD: 740 if (info.rti_info[RTAX_GATEWAY] == NULL) { 741 senderr(EINVAL); 742 } 743 #if defined(INET) || defined(INET6) 744 /* support for new ARP/NDP code with keeping backcompat */ 745 if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) { 746 const struct sockaddr_dl *sdlp = 747 satocsdl(info.rti_info[RTAX_GATEWAY]); 748 749 /* Allow routing requests by interface index */ 750 if (sdlp->sdl_nlen == 0 && sdlp->sdl_alen == 0 751 && sdlp->sdl_slen == 0) 752 goto fallback; 753 /* 754 * Old arp binaries don't set the sdl_index 755 * so we have to complement it. 756 */ 757 int sdl_index = sdlp->sdl_index; 758 if (sdl_index == 0) { 759 error = route_get_sdl_index(&info, &sdl_index); 760 if (error != 0) 761 goto fallback; 762 } else if ( 763 info.rti_info[RTAX_DST]->sa_family == AF_INET) { 764 /* 765 * XXX workaround for SIN_PROXY case; proxy arp 766 * entry should be in an interface that has 767 * a network route including the destination, 768 * not a local (link) route that may not be a 769 * desired place, for example a tap. 770 */ 771 const struct sockaddr_inarp *sina = 772 (const struct sockaddr_inarp *) 773 info.rti_info[RTAX_DST]; 774 if (sina->sin_other & SIN_PROXY) { 775 error = route_get_sdl_index(&info, 776 &sdl_index); 777 if (error != 0) 778 goto fallback; 779 } 780 } 781 error = lla_rt_output(rtm->rtm_type, rtm->rtm_flags, 782 rtm->rtm_rmx.rmx_expire, &info, sdl_index); 783 break; 784 } 785 fallback: 786 #endif /* defined(INET) || defined(INET6) */ 787 error = rtrequest1(rtm->rtm_type, &info, &saved_nrt); 788 if (error == 0) { 789 _rt_setmetrics(rtm->rtm_inits, rtm, saved_nrt); 790 rt_unref(saved_nrt); 791 } 792 break; 793 794 case RTM_DELETE: 795 #if defined(INET) || defined(INET6) 796 /* support for new ARP/NDP code */ 797 if (info.rti_info[RTAX_GATEWAY] && 798 (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) && 799 (rtm->rtm_flags & RTF_LLDATA) != 0) { 800 const struct sockaddr_dl *sdlp = 801 satocsdl(info.rti_info[RTAX_GATEWAY]); 802 error = lla_rt_output(rtm->rtm_type, rtm->rtm_flags, 803 rtm->rtm_rmx.rmx_expire, &info, sdlp->sdl_index); 804 rtm->rtm_flags &= ~RTF_UP; 805 break; 806 } 807 #endif 808 error = rtrequest1(rtm->rtm_type, &info, &saved_nrt); 809 if (error != 0) 810 break; 811 812 rt = saved_nrt; 813 do_rt_free = true; 814 info.rti_info[RTAX_DST] = rt_getkey(rt); 815 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 816 info.rti_info[RTAX_NETMASK] = rt_mask(rt); 817 info.rti_info[RTAX_TAG] = rt_gettag(rt); 818 error = route_output_report(rt, &info, rtm, &new_rtm); 819 if (error) 820 senderr(error); 821 if (new_rtm != NULL) { 822 old_rtm = rtm; 823 rtm = new_rtm; 824 } 825 break; 826 827 case RTM_GET: 828 case RTM_CHANGE: 829 case RTM_LOCK: 830 /* XXX This will mask info.rti_info[RTAX_DST] with 831 * info.rti_info[RTAX_NETMASK] before 832 * searching. It did not used to do that. --dyoung 833 */ 834 rt = NULL; 835 error = rtrequest1(RTM_GET, &info, &rt); 836 if (error != 0) 837 senderr(error); 838 if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */ 839 if (memcmp(info.rti_info[RTAX_DST], rt_getkey(rt), 840 info.rti_info[RTAX_DST]->sa_len) != 0) 841 senderr(ESRCH); 842 if (info.rti_info[RTAX_NETMASK] == NULL && 843 rt_mask(rt) != NULL) 844 senderr(ETOOMANYREFS); 845 } 846 847 /* 848 * XXX if arp/ndp requests an L2 entry, we have to obtain 849 * it from lltable while for the route command we have to 850 * return a route as it is. How to distinguish them? 851 * For newer arp/ndp, RTF_LLDATA flag set by arp/ndp 852 * indicates an L2 entry is requested. For old arp/ndp 853 * binaries, we check RTF_UP flag is NOT set; it works 854 * by the fact that arp/ndp don't set it while the route 855 * command sets it. 856 */ 857 if (((rtm->rtm_flags & RTF_LLDATA) != 0 || 858 (rtm->rtm_flags & RTF_UP) == 0) && 859 rtm->rtm_type == RTM_GET && 860 sockaddr_cmp(rt_getkey(rt), info.rti_info[RTAX_DST]) != 0) { 861 int ll_flags = 0; 862 route_get_sdl(rt->rt_ifp, info.rti_info[RTAX_DST], &sdl, 863 &ll_flags); 864 info.rti_info[RTAX_GATEWAY] = sstocsa(&sdl); 865 error = route_output_report(rt, &info, rtm, &new_rtm); 866 if (error) 867 senderr(error); 868 if (new_rtm != NULL) { 869 old_rtm = rtm; 870 rtm = new_rtm; 871 } 872 rtm->rtm_flags |= RTF_LLDATA; 873 rtm->rtm_flags &= ~RTF_CONNECTED; 874 rtm->rtm_flags |= (ll_flags & LLE_STATIC) ? RTF_STATIC : 0; 875 break; 876 } 877 878 switch (rtm->rtm_type) { 879 case RTM_GET: 880 info.rti_info[RTAX_DST] = rt_getkey(rt); 881 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 882 info.rti_info[RTAX_NETMASK] = rt_mask(rt); 883 info.rti_info[RTAX_TAG] = rt_gettag(rt); 884 error = route_output_report(rt, &info, rtm, &new_rtm); 885 if (error) 886 senderr(error); 887 if (new_rtm != NULL) { 888 old_rtm = rtm; 889 rtm = new_rtm; 890 } 891 break; 892 893 case RTM_CHANGE: 894 #ifdef NET_MPSAFE 895 /* 896 * Release rt_so_mtx to avoid a deadlock with route_intr 897 * and also serialize updating routes to avoid another. 898 */ 899 if (rt_updating) { 900 /* Release to allow the updater to proceed */ 901 rt_unref(rt); 902 rt = NULL; 903 } 904 while (rt_updating) { 905 error = cv_wait_sig(&rt_update_cv, rt_so_mtx); 906 if (error != 0) 907 goto flush; 908 } 909 if (rt == NULL) { 910 error = rtrequest1(RTM_GET, &info, &rt); 911 if (error != 0) 912 goto flush; 913 } 914 rt_updating = true; 915 mutex_exit(rt_so_mtx); 916 917 error = rt_update_prepare(rt); 918 if (error == 0) { 919 error = rt_update(rt, &info, rtm); 920 rt_update_finish(rt); 921 } 922 923 mutex_enter(rt_so_mtx); 924 rt_updating = false; 925 cv_broadcast(&rt_update_cv); 926 #else 927 error = rt_update(rt, &info, rtm); 928 #endif 929 if (error != 0) 930 goto flush; 931 /*FALLTHROUGH*/ 932 case RTM_LOCK: 933 rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits); 934 rt->rt_rmx.rmx_locks |= 935 (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks); 936 break; 937 } 938 break; 939 940 default: 941 senderr(EOPNOTSUPP); 942 } 943 944 flush: 945 if (rtm) { 946 if (error) 947 rtm->rtm_errno = error; 948 else 949 rtm->rtm_flags |= RTF_DONE; 950 } 951 family = info.rti_info[RTAX_DST] ? info.rti_info[RTAX_DST]->sa_family : 952 0; 953 /* We cannot free old_rtm until we have stopped using the 954 * pointers in info, some of which may point to sockaddrs 955 * in old_rtm. 956 */ 957 if (old_rtm != NULL) 958 Free(old_rtm); 959 if (rt) { 960 if (do_rt_free) { 961 #ifdef NET_MPSAFE 962 /* 963 * Release rt_so_mtx to avoid a deadlock with 964 * route_intr. 965 */ 966 mutex_exit(rt_so_mtx); 967 rt_free(rt); 968 mutex_enter(rt_so_mtx); 969 #else 970 rt_free(rt); 971 #endif 972 } else 973 rt_unref(rt); 974 } 975 { 976 struct rawcb *rp = NULL; 977 /* 978 * Check to see if we don't want our own messages. 979 */ 980 if ((so->so_options & SO_USELOOPBACK) == 0) { 981 if (COMPATNAME(route_info).ri_cb.any_count <= 1) { 982 if (rtm) 983 Free(rtm); 984 m_freem(m); 985 goto out; 986 } 987 /* There is another listener, so construct message */ 988 rp = sotorawcb(so); 989 } 990 if (rtm) { 991 m_copyback(m, 0, rtm->rtm_msglen, rtm); 992 if (m->m_pkthdr.len < rtm->rtm_msglen) { 993 m_freem(m); 994 m = NULL; 995 } else if (m->m_pkthdr.len > rtm->rtm_msglen) 996 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len); 997 Free(rtm); 998 } 999 if (rp) 1000 rp->rcb_proto.sp_family = 0; /* Avoid us */ 1001 if (family) 1002 proto.sp_protocol = family; 1003 if (m) 1004 raw_input(m, &proto, &COMPATNAME(route_info).ri_src, 1005 &COMPATNAME(route_info).ri_dst, &rt_rawcb); 1006 if (rp) 1007 rp->rcb_proto.sp_family = PF_XROUTE; 1008 } 1009 out: 1010 curlwp_bindx(bound); 1011 return error; 1012 } 1013 1014 static int 1015 route_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1016 { 1017 struct routecb *rop = sotoroutecb(so); 1018 int error = 0; 1019 unsigned char *rtm_type, *cp, *ep; 1020 size_t len; 1021 unsigned int msgfilter; 1022 struct sockaddr *sa; 1023 1024 KASSERT(solocked(so)); 1025 1026 if (sopt->sopt_level != AF_ROUTE) { 1027 error = ENOPROTOOPT; 1028 } else switch (op) { 1029 case PRCO_SETOPT: 1030 switch (sopt->sopt_name) { 1031 case RO_MSGFILTER: 1032 msgfilter = 0; 1033 for (rtm_type = sopt->sopt_data, len = sopt->sopt_size; 1034 len != 0; 1035 rtm_type++, len -= sizeof(*rtm_type)) 1036 { 1037 /* Guard against overflowing our storage. */ 1038 if (*rtm_type >= sizeof(msgfilter) * CHAR_BIT) { 1039 error = EOVERFLOW; 1040 break; 1041 } 1042 msgfilter |= RTMSGFILTER(*rtm_type); 1043 } 1044 if (error == 0) 1045 rop->rocb_msgfilter = msgfilter; 1046 break; 1047 case RO_MISSFILTER: 1048 /* Validate the data */ 1049 len = 0; 1050 cp = sopt->sopt_data; 1051 ep = cp + sopt->sopt_size; 1052 while (cp < ep) { 1053 if (ep - cp < 1054 offsetof(struct sockaddr, sa_len) + 1055 sizeof(sa->sa_len)) 1056 break; 1057 if (++len > RO_FILTSA_MAX) { 1058 error = ENOBUFS; 1059 break; 1060 } 1061 sa = (struct sockaddr *)cp; 1062 cp += RT_XROUNDUP(sa->sa_len); 1063 } 1064 if (cp != ep) { 1065 if (error == 0) 1066 error = EINVAL; 1067 break; 1068 } 1069 if (rop->rocb_missfilterlen != 0) 1070 kmem_free(rop->rocb_missfilter, 1071 rop->rocb_missfilterlen); 1072 if (sopt->sopt_size != 0) { 1073 rop->rocb_missfilter = 1074 kmem_alloc(sopt->sopt_size, KM_SLEEP); 1075 if (rop->rocb_missfilter == NULL) { 1076 rop->rocb_missfilterlen = 0; 1077 error = ENOBUFS; 1078 break; 1079 } 1080 } else 1081 rop->rocb_missfilter = NULL; 1082 rop->rocb_missfilterlen = sopt->sopt_size; 1083 if (rop->rocb_missfilterlen != 0) 1084 memcpy(rop->rocb_missfilter, sopt->sopt_data, 1085 rop->rocb_missfilterlen); 1086 break; 1087 default: 1088 error = ENOPROTOOPT; 1089 break; 1090 } 1091 break; 1092 case PRCO_GETOPT: 1093 switch (sopt->sopt_name) { 1094 case RO_MSGFILTER: 1095 error = ENOTSUP; 1096 break; 1097 default: 1098 error = ENOPROTOOPT; 1099 break; 1100 } 1101 } 1102 return error; 1103 } 1104 1105 static void 1106 _rt_setmetrics(int which, const struct rt_xmsghdr *in, struct rtentry *out) 1107 { 1108 #define metric(f, e) if (which & (f)) out->rt_rmx.e = in->rtm_rmx.e; 1109 metric(RTV_RPIPE, rmx_recvpipe); 1110 metric(RTV_SPIPE, rmx_sendpipe); 1111 metric(RTV_SSTHRESH, rmx_ssthresh); 1112 metric(RTV_RTT, rmx_rtt); 1113 metric(RTV_RTTVAR, rmx_rttvar); 1114 metric(RTV_HOPCOUNT, rmx_hopcount); 1115 metric(RTV_MTU, rmx_mtu); 1116 #undef metric 1117 if (which & RTV_EXPIRE) { 1118 out->rt_rmx.rmx_expire = in->rtm_rmx.rmx_expire ? 1119 time_wall_to_mono(in->rtm_rmx.rmx_expire) : 0; 1120 } 1121 } 1122 1123 static void 1124 rtm_setmetrics(const struct rtentry *in, struct rt_xmsghdr *out) 1125 { 1126 #define metric(e) out->rtm_rmx.e = in->rt_rmx.e; 1127 metric(rmx_recvpipe); 1128 metric(rmx_sendpipe); 1129 metric(rmx_ssthresh); 1130 metric(rmx_rtt); 1131 metric(rmx_rttvar); 1132 metric(rmx_hopcount); 1133 metric(rmx_mtu); 1134 metric(rmx_locks); 1135 #undef metric 1136 out->rtm_rmx.rmx_expire = in->rt_rmx.rmx_expire ? 1137 time_mono_to_wall(in->rt_rmx.rmx_expire) : 0; 1138 } 1139 1140 static int 1141 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim, 1142 struct rt_addrinfo *rtinfo) 1143 { 1144 const struct sockaddr *sa = NULL; /* Quell compiler warning */ 1145 int i; 1146 1147 for (i = 0; i < RTAX_MAX && cp < cplim; i++) { 1148 if ((rtinfo->rti_addrs & (1 << i)) == 0) 1149 continue; 1150 rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp; 1151 RT_XADVANCE(cp, sa); 1152 } 1153 1154 /* 1155 * Check for extra addresses specified, except RTM_GET asking 1156 * for interface info. 1157 */ 1158 if (rtmtype == RTM_GET) { 1159 if (((rtinfo->rti_addrs & 1160 (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0U << i)) != 0) 1161 return 1; 1162 } else if ((rtinfo->rti_addrs & (~0U << i)) != 0) 1163 return 1; 1164 /* Check for bad data length. */ 1165 if (cp != cplim) { 1166 if (i == RTAX_NETMASK + 1 && sa != NULL && 1167 cp - RT_XROUNDUP(sa->sa_len) + sa->sa_len == cplim) 1168 /* 1169 * The last sockaddr was info.rti_info[RTAX_NETMASK]. 1170 * We accept this for now for the sake of old 1171 * binaries or third party softwares. 1172 */ 1173 ; 1174 else 1175 return 1; 1176 } 1177 return 0; 1178 } 1179 1180 static int 1181 rt_getlen(int type) 1182 { 1183 RTS_CTASSERT(__alignof(struct ifa_msghdr) >= sizeof(uint64_t)); 1184 RTS_CTASSERT(__alignof(struct if_msghdr) >= sizeof(uint64_t)); 1185 RTS_CTASSERT(__alignof(struct if_announcemsghdr) >= sizeof(uint64_t)); 1186 RTS_CTASSERT(__alignof(struct rt_msghdr) >= sizeof(uint64_t)); 1187 1188 switch (type) { 1189 case RTM_ODELADDR: 1190 case RTM_ONEWADDR: 1191 case RTM_OCHGADDR: 1192 if (rtsock_iflist_70_hook.hooked) 1193 return sizeof(struct ifa_msghdr70); 1194 else { 1195 #ifdef RTSOCK_DEBUG 1196 printf("%s: unsupported RTM type %d\n", __func__, type); 1197 #endif 1198 return -1; 1199 } 1200 1201 case RTM_DELADDR: 1202 case RTM_NEWADDR: 1203 case RTM_CHGADDR: 1204 return sizeof(struct ifa_xmsghdr); 1205 1206 case RTM_OOIFINFO: 1207 if (rtsock_iflist_14_hook.hooked) 1208 return sizeof(struct if_msghdr14); 1209 else { 1210 #ifdef RTSOCK_DEBUG 1211 printf("%s: unsupported RTM type RTM_OOIFINFO\n", 1212 __func__); 1213 #endif 1214 return -1; 1215 } 1216 1217 case RTM_OIFINFO: 1218 if (rtsock_iflist_50_hook.hooked) 1219 return sizeof(struct if_msghdr50); 1220 else { 1221 #ifdef RTSOCK_DEBUG 1222 printf("%s: unsupported RTM type RTM_OIFINFO\n", 1223 __func__); 1224 #endif 1225 return -1; 1226 } 1227 1228 case RTM_IFINFO: 1229 return sizeof(struct if_xmsghdr); 1230 1231 case RTM_IFANNOUNCE: 1232 case RTM_IEEE80211: 1233 return sizeof(struct if_xannouncemsghdr); 1234 1235 default: 1236 return sizeof(struct rt_xmsghdr); 1237 } 1238 } 1239 1240 1241 struct mbuf * 1242 COMPATNAME(rt_msg1)(int type, struct rt_addrinfo *rtinfo, void *data, int datalen) 1243 { 1244 struct rt_xmsghdr *rtm; 1245 struct mbuf *m; 1246 int i; 1247 const struct sockaddr *sa; 1248 int len, dlen; 1249 1250 m = m_gethdr(M_DONTWAIT, MT_DATA); 1251 if (m == NULL) 1252 return m; 1253 MCLAIM(m, &COMPATNAME(routedomain).dom_mowner); 1254 1255 if ((len = rt_getlen(type)) == -1) 1256 goto out; 1257 if (len > MHLEN + MLEN) 1258 panic("%s: message too long", __func__); 1259 else if (len > MHLEN) { 1260 m->m_next = m_get(M_DONTWAIT, MT_DATA); 1261 if (m->m_next == NULL) 1262 goto out; 1263 MCLAIM(m->m_next, m->m_owner); 1264 m->m_pkthdr.len = len; 1265 m->m_len = MHLEN; 1266 m->m_next->m_len = len - MHLEN; 1267 } else { 1268 m->m_pkthdr.len = m->m_len = len; 1269 } 1270 m_reset_rcvif(m); 1271 m_copyback(m, 0, datalen, data); 1272 if (len > datalen) 1273 (void)memset(mtod(m, char *) + datalen, 0, len - datalen); 1274 rtm = mtod(m, struct rt_xmsghdr *); 1275 for (i = 0; i < RTAX_MAX; i++) { 1276 if ((sa = rtinfo->rti_info[i]) == NULL) 1277 continue; 1278 rtinfo->rti_addrs |= (1 << i); 1279 dlen = RT_XROUNDUP(sa->sa_len); 1280 m_copyback(m, len, sa->sa_len, sa); 1281 if (dlen != sa->sa_len) { 1282 /* 1283 * Up to 7 + 1 nul's since roundup is to 1284 * sizeof(uint64_t) (8 bytes) 1285 */ 1286 m_copyback(m, len + sa->sa_len, 1287 dlen - sa->sa_len, "\0\0\0\0\0\0\0"); 1288 } 1289 len += dlen; 1290 } 1291 if (m->m_pkthdr.len != len) 1292 goto out; 1293 rtm->rtm_msglen = len; 1294 rtm->rtm_version = RTM_XVERSION; 1295 rtm->rtm_type = type; 1296 return m; 1297 out: 1298 m_freem(m); 1299 return NULL; 1300 } 1301 1302 /* 1303 * rt_msg2 1304 * 1305 * fills 'cp' or 'w'.w_tmem with the routing socket message and 1306 * returns the length of the message in 'lenp'. 1307 * 1308 * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold 1309 * the message 1310 * otherwise walkarg's w_needed is updated and if the user buffer is 1311 * specified and w_needed indicates space exists the information is copied 1312 * into the temp space (w_tmem). w_tmem is [re]allocated if necessary, 1313 * if the allocation fails ENOBUFS is returned. 1314 */ 1315 static int 1316 rt_msg2(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w, 1317 int *lenp) 1318 { 1319 int i; 1320 int len, dlen, second_time = 0; 1321 char *cp0, *cp = cpv; 1322 1323 rtinfo->rti_addrs = 0; 1324 again: 1325 if ((len = rt_getlen(type)) == -1) 1326 return EINVAL; 1327 1328 if ((cp0 = cp) != NULL) 1329 cp += len; 1330 for (i = 0; i < RTAX_MAX; i++) { 1331 const struct sockaddr *sa; 1332 1333 if ((sa = rtinfo->rti_info[i]) == NULL) 1334 continue; 1335 rtinfo->rti_addrs |= (1 << i); 1336 dlen = RT_XROUNDUP(sa->sa_len); 1337 if (cp) { 1338 int diff = dlen - sa->sa_len; 1339 (void)memcpy(cp, sa, (size_t)sa->sa_len); 1340 cp += sa->sa_len; 1341 if (diff > 0) { 1342 (void)memset(cp, 0, (size_t)diff); 1343 cp += diff; 1344 } 1345 } 1346 len += dlen; 1347 } 1348 if (cp == NULL && w != NULL && !second_time) { 1349 struct rt_walkarg *rw = w; 1350 1351 rw->w_needed += len; 1352 if (rw->w_needed <= 0 && rw->w_where) { 1353 if (rw->w_tmemsize < len) { 1354 if (rw->w_tmem) 1355 kmem_free(rw->w_tmem, rw->w_tmemsize); 1356 rw->w_tmem = kmem_zalloc(len, KM_SLEEP); 1357 rw->w_tmemsize = len; 1358 } 1359 if (rw->w_tmem) { 1360 cp = rw->w_tmem; 1361 second_time = 1; 1362 goto again; 1363 } else { 1364 rw->w_tmemneeded = len; 1365 return ENOBUFS; 1366 } 1367 } 1368 } 1369 if (cp) { 1370 struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)cp0; 1371 1372 rtm->rtm_version = RTM_XVERSION; 1373 rtm->rtm_type = type; 1374 rtm->rtm_msglen = len; 1375 } 1376 if (lenp) 1377 *lenp = len; 1378 return 0; 1379 } 1380 1381 /* 1382 * This routine is called to generate a message from the routing 1383 * socket indicating that a redirect has occurred, a routing lookup 1384 * has failed, or that a protocol has detected timeouts to a particular 1385 * destination. 1386 */ 1387 void 1388 COMPATNAME(rt_missmsg)(int type, const struct rt_addrinfo *rtinfo, int flags, 1389 int error) 1390 { 1391 struct rt_xmsghdr rtm; 1392 struct mbuf *m; 1393 const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST]; 1394 struct rt_addrinfo info = *rtinfo; 1395 1396 COMPATCALL(rt_missmsg, (type, rtinfo, flags, error)); 1397 if (COMPATNAME(route_info).ri_cb.any_count == 0) 1398 return; 1399 memset(&rtm, 0, sizeof(rtm)); 1400 rtm.rtm_pid = curproc->p_pid; 1401 rtm.rtm_flags = RTF_DONE | flags; 1402 rtm.rtm_errno = error; 1403 m = COMPATNAME(rt_msg1)(type, &info, &rtm, sizeof(rtm)); 1404 if (m == NULL) 1405 return; 1406 mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs; 1407 COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0); 1408 } 1409 1410 /* 1411 * This routine is called to generate a message from the routing 1412 * socket indicating that the status of a network interface has changed. 1413 */ 1414 void 1415 COMPATNAME(rt_ifmsg)(struct ifnet *ifp) 1416 { 1417 struct if_xmsghdr ifm; 1418 struct mbuf *m; 1419 struct rt_addrinfo info; 1420 1421 COMPATCALL(rt_ifmsg, (ifp)); 1422 if (COMPATNAME(route_info).ri_cb.any_count == 0) 1423 return; 1424 (void)memset(&info, 0, sizeof(info)); 1425 (void)memset(&ifm, 0, sizeof(ifm)); 1426 ifm.ifm_index = ifp->if_index; 1427 ifm.ifm_flags = ifp->if_flags; 1428 if_export_if_data(ifp, &ifm.ifm_data, false); 1429 ifm.ifm_addrs = 0; 1430 m = COMPATNAME(rt_msg1)(RTM_IFINFO, &info, &ifm, sizeof(ifm)); 1431 if (m == NULL) 1432 return; 1433 COMPATNAME(route_enqueue)(m, 0); 1434 MODULE_HOOK_CALL_VOID(rtsock_oifmsg_14_hook, (ifp), __nothing); 1435 MODULE_HOOK_CALL_VOID(rtsock_oifmsg_50_hook, (ifp), __nothing); 1436 } 1437 1438 /* 1439 * This is called to generate messages from the routing socket 1440 * indicating a network interface has had addresses associated with it. 1441 * if we ever reverse the logic and replace messages TO the routing 1442 * socket indicate a request to configure interfaces, then it will 1443 * be unnecessary as the routing socket will automatically generate 1444 * copies of it. 1445 */ 1446 static void 1447 COMPATNAME(rt_addrmsg0)(int cmd, struct ifaddr *ifa, int error, 1448 struct rtentry *rt, const struct sockaddr *src) 1449 { 1450 #define cmdpass(__cmd, __pass) (((__cmd) << 2) | (__pass)) 1451 struct rt_addrinfo info; 1452 const struct sockaddr *sa; 1453 int pass; 1454 struct mbuf *m; 1455 struct ifnet *ifp; 1456 struct rt_xmsghdr rtm; 1457 struct ifa_xmsghdr ifam; 1458 int ncmd; 1459 1460 KASSERT(ifa != NULL); 1461 KASSERT(ifa->ifa_addr != NULL); 1462 ifp = ifa->ifa_ifp; 1463 if (cmd == RTM_ADD && vec_sctp_add_ip_address != NULL) { 1464 (*vec_sctp_add_ip_address)(ifa); 1465 } else if (cmd == RTM_DELETE && vec_sctp_delete_ip_address != NULL) { 1466 (*vec_sctp_delete_ip_address)(ifa); 1467 } 1468 1469 COMPATCALL(rt_addrmsg_rt, (cmd, ifa, error, rt)); 1470 if (COMPATNAME(route_info).ri_cb.any_count == 0) 1471 return; 1472 for (pass = 1; pass < 3; pass++) { 1473 memset(&info, 0, sizeof(info)); 1474 switch (cmdpass(cmd, pass)) { 1475 case cmdpass(RTM_ADD, 1): 1476 case cmdpass(RTM_CHANGE, 1): 1477 case cmdpass(RTM_DELETE, 2): 1478 case cmdpass(RTM_NEWADDR, 1): 1479 case cmdpass(RTM_DELADDR, 1): 1480 case cmdpass(RTM_CHGADDR, 1): 1481 switch (cmd) { 1482 case RTM_ADD: 1483 ncmd = RTM_XNEWADDR; 1484 break; 1485 case RTM_DELETE: 1486 ncmd = RTM_XDELADDR; 1487 break; 1488 case RTM_CHANGE: 1489 ncmd = RTM_XCHGADDR; 1490 break; 1491 case RTM_NEWADDR: 1492 ncmd = RTM_XNEWADDR; 1493 break; 1494 case RTM_DELADDR: 1495 ncmd = RTM_XDELADDR; 1496 break; 1497 case RTM_CHGADDR: 1498 ncmd = RTM_XCHGADDR; 1499 break; 1500 default: 1501 panic("%s: unknown command %d", __func__, cmd); 1502 } 1503 MODULE_HOOK_CALL_VOID(rtsock_newaddr_70_hook, 1504 (ncmd, ifa), __nothing); 1505 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr; 1506 KASSERT(ifp->if_dl != NULL); 1507 info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr; 1508 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask; 1509 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 1510 info.rti_info[RTAX_AUTHOR] = src; 1511 memset(&ifam, 0, sizeof(ifam)); 1512 ifam.ifam_index = ifp->if_index; 1513 ifam.ifam_metric = ifa->ifa_metric; 1514 ifam.ifam_flags = ifa->ifa_flags; 1515 #ifndef COMPAT_RTSOCK 1516 ifam.ifam_pid = curproc->p_pid; 1517 ifam.ifam_addrflags = if_addrflags(ifa); 1518 #endif 1519 m = COMPATNAME(rt_msg1)(ncmd, &info, &ifam, sizeof(ifam)); 1520 if (m == NULL) 1521 continue; 1522 mtod(m, struct ifa_xmsghdr *)->ifam_addrs = 1523 info.rti_addrs; 1524 break; 1525 case cmdpass(RTM_ADD, 2): 1526 case cmdpass(RTM_CHANGE, 2): 1527 case cmdpass(RTM_DELETE, 1): 1528 if (rt == NULL) 1529 continue; 1530 info.rti_info[RTAX_NETMASK] = rt_mask(rt); 1531 info.rti_info[RTAX_DST] = sa = rt_getkey(rt); 1532 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 1533 memset(&rtm, 0, sizeof(rtm)); 1534 rtm.rtm_pid = curproc->p_pid; 1535 rtm.rtm_index = ifp->if_index; 1536 rtm.rtm_flags |= rt->rt_flags; 1537 rtm.rtm_errno = error; 1538 m = COMPATNAME(rt_msg1)(cmd, &info, &rtm, sizeof(rtm)); 1539 if (m == NULL) 1540 continue; 1541 mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs; 1542 break; 1543 default: 1544 continue; 1545 } 1546 KASSERTMSG(m != NULL, "called with wrong command"); 1547 COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0); 1548 } 1549 #undef cmdpass 1550 } 1551 1552 void 1553 COMPATNAME(rt_addrmsg)(int cmd, struct ifaddr *ifa) 1554 { 1555 1556 COMPATNAME(rt_addrmsg0)(cmd, ifa, 0, NULL, NULL); 1557 } 1558 1559 void 1560 COMPATNAME(rt_addrmsg_rt)(int cmd, struct ifaddr *ifa, int error, 1561 struct rtentry *rt) 1562 { 1563 1564 COMPATNAME(rt_addrmsg0)(cmd, ifa, error, rt, NULL); 1565 } 1566 1567 void 1568 COMPATNAME(rt_addrmsg_src)(int cmd, struct ifaddr *ifa, 1569 const struct sockaddr *src) 1570 { 1571 1572 COMPATNAME(rt_addrmsg0)(cmd, ifa, 0, NULL, src); 1573 } 1574 1575 static struct mbuf * 1576 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what, 1577 struct rt_addrinfo *info) 1578 { 1579 struct if_xannouncemsghdr ifan; 1580 1581 memset(info, 0, sizeof(*info)); 1582 memset(&ifan, 0, sizeof(ifan)); 1583 ifan.ifan_index = ifp->if_index; 1584 strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name)); 1585 ifan.ifan_what = what; 1586 return COMPATNAME(rt_msg1)(type, info, &ifan, sizeof(ifan)); 1587 } 1588 1589 /* 1590 * This is called to generate routing socket messages indicating 1591 * network interface arrival and departure. 1592 */ 1593 void 1594 COMPATNAME(rt_ifannouncemsg)(struct ifnet *ifp, int what) 1595 { 1596 struct mbuf *m; 1597 struct rt_addrinfo info; 1598 1599 COMPATCALL(rt_ifannouncemsg, (ifp, what)); 1600 if (COMPATNAME(route_info).ri_cb.any_count == 0) 1601 return; 1602 m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info); 1603 if (m == NULL) 1604 return; 1605 COMPATNAME(route_enqueue)(m, 0); 1606 } 1607 1608 /* 1609 * This is called to generate routing socket messages indicating 1610 * IEEE80211 wireless events. 1611 * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way. 1612 */ 1613 void 1614 COMPATNAME(rt_ieee80211msg)(struct ifnet *ifp, int what, void *data, 1615 size_t data_len) 1616 { 1617 struct mbuf *m; 1618 struct rt_addrinfo info; 1619 1620 COMPATCALL(rt_ieee80211msg, (ifp, what, data, data_len)); 1621 if (COMPATNAME(route_info).ri_cb.any_count == 0) 1622 return; 1623 m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info); 1624 if (m == NULL) 1625 return; 1626 /* 1627 * Append the ieee80211 data. Try to stick it in the 1628 * mbuf containing the ifannounce msg; otherwise allocate 1629 * a new mbuf and append. 1630 * 1631 * NB: we assume m is a single mbuf. 1632 */ 1633 if (data_len > M_TRAILINGSPACE(m)) { 1634 struct mbuf *n = m_get(M_NOWAIT, MT_DATA); 1635 if (n == NULL) { 1636 m_freem(m); 1637 return; 1638 } 1639 (void)memcpy(mtod(n, void *), data, data_len); 1640 n->m_len = data_len; 1641 m->m_next = n; 1642 } else if (data_len > 0) { 1643 (void)memcpy(mtod(m, uint8_t *) + m->m_len, data, data_len); 1644 m->m_len += data_len; 1645 } 1646 if (m->m_flags & M_PKTHDR) 1647 m->m_pkthdr.len += data_len; 1648 mtod(m, struct if_xannouncemsghdr *)->ifan_msglen += data_len; 1649 COMPATNAME(route_enqueue)(m, 0); 1650 } 1651 1652 /* 1653 * Routing message software interrupt routine 1654 */ 1655 static void 1656 COMPATNAME(route_intr)(void *cookie) 1657 { 1658 struct sockproto proto = { .sp_family = PF_XROUTE, }; 1659 struct route_info * const ri = &COMPATNAME(route_info); 1660 struct mbuf *m; 1661 1662 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 1663 for (;;) { 1664 IFQ_LOCK(&ri->ri_intrq); 1665 IF_DEQUEUE(&ri->ri_intrq, m); 1666 IFQ_UNLOCK(&ri->ri_intrq); 1667 if (m == NULL) 1668 break; 1669 proto.sp_protocol = M_GETCTX(m, uintptr_t); 1670 #ifdef NET_MPSAFE 1671 mutex_enter(rt_so_mtx); 1672 #endif 1673 raw_input(m, &proto, &ri->ri_src, &ri->ri_dst, &rt_rawcb); 1674 #ifdef NET_MPSAFE 1675 mutex_exit(rt_so_mtx); 1676 #endif 1677 } 1678 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 1679 } 1680 1681 /* 1682 * Enqueue a message to the software interrupt routine. 1683 */ 1684 void 1685 COMPATNAME(route_enqueue)(struct mbuf *m, int family) 1686 { 1687 struct route_info * const ri = &COMPATNAME(route_info); 1688 int wasempty; 1689 1690 IFQ_LOCK(&ri->ri_intrq); 1691 if (IF_QFULL(&ri->ri_intrq)) { 1692 printf("%s: queue full, dropped message\n", __func__); 1693 IF_DROP(&ri->ri_intrq); 1694 IFQ_UNLOCK(&ri->ri_intrq); 1695 m_freem(m); 1696 } else { 1697 wasempty = IF_IS_EMPTY(&ri->ri_intrq); 1698 M_SETCTX(m, (uintptr_t)family); 1699 IF_ENQUEUE(&ri->ri_intrq, m); 1700 IFQ_UNLOCK(&ri->ri_intrq); 1701 if (wasempty) { 1702 kpreempt_disable(); 1703 softint_schedule(ri->ri_sih); 1704 kpreempt_enable(); 1705 } 1706 } 1707 } 1708 1709 static void 1710 COMPATNAME(route_init)(void) 1711 { 1712 struct route_info * const ri = &COMPATNAME(route_info); 1713 1714 #ifndef COMPAT_RTSOCK 1715 rt_init(); 1716 #ifdef NET_MPSAFE 1717 rt_so_mtx = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 1718 1719 cv_init(&rt_update_cv, "rtsock_cv"); 1720 #endif 1721 1722 sysctl_net_route_setup(NULL, PF_ROUTE, "rtable"); 1723 #endif 1724 ri->ri_intrq.ifq_maxlen = ri->ri_maxqlen; 1725 ri->ri_sih = softint_establish(SOFTINT_NET | SOFTINT_MPSAFE, 1726 COMPATNAME(route_intr), NULL); 1727 IFQ_LOCK_INIT(&ri->ri_intrq); 1728 1729 #ifdef MBUFTRACE 1730 MOWNER_ATTACH(&COMPATNAME(routedomain).dom_mowner); 1731 #endif 1732 } 1733 1734 /* 1735 * Definitions of protocols supported in the ROUTE domain. 1736 */ 1737 #ifndef COMPAT_RTSOCK 1738 PR_WRAP_USRREQS(route); 1739 #else 1740 PR_WRAP_USRREQS(compat_50_route); 1741 #endif 1742 1743 static const struct pr_usrreqs route_usrreqs = { 1744 .pr_attach = COMPATNAME(route_attach_wrapper), 1745 .pr_detach = COMPATNAME(route_detach_wrapper), 1746 .pr_accept = COMPATNAME(route_accept_wrapper), 1747 .pr_bind = COMPATNAME(route_bind_wrapper), 1748 .pr_listen = COMPATNAME(route_listen_wrapper), 1749 .pr_connect = COMPATNAME(route_connect_wrapper), 1750 .pr_connect2 = COMPATNAME(route_connect2_wrapper), 1751 .pr_disconnect = COMPATNAME(route_disconnect_wrapper), 1752 .pr_shutdown = COMPATNAME(route_shutdown_wrapper), 1753 .pr_abort = COMPATNAME(route_abort_wrapper), 1754 .pr_ioctl = COMPATNAME(route_ioctl_wrapper), 1755 .pr_stat = COMPATNAME(route_stat_wrapper), 1756 .pr_peeraddr = COMPATNAME(route_peeraddr_wrapper), 1757 .pr_sockaddr = COMPATNAME(route_sockaddr_wrapper), 1758 .pr_rcvd = COMPATNAME(route_rcvd_wrapper), 1759 .pr_recvoob = COMPATNAME(route_recvoob_wrapper), 1760 .pr_send = COMPATNAME(route_send_wrapper), 1761 .pr_sendoob = COMPATNAME(route_sendoob_wrapper), 1762 .pr_purgeif = COMPATNAME(route_purgeif_wrapper), 1763 }; 1764 1765 static const struct protosw COMPATNAME(route_protosw)[] = { 1766 { 1767 .pr_type = SOCK_RAW, 1768 .pr_domain = &COMPATNAME(routedomain), 1769 .pr_flags = PR_ATOMIC|PR_ADDR, 1770 .pr_ctlinput = raw_ctlinput, 1771 .pr_ctloutput = route_ctloutput, 1772 .pr_usrreqs = &route_usrreqs, 1773 .pr_init = rt_pr_init, 1774 }, 1775 }; 1776 1777 struct domain COMPATNAME(routedomain) = { 1778 .dom_family = PF_XROUTE, 1779 .dom_name = DOMAINNAME, 1780 .dom_init = COMPATNAME(route_init), 1781 .dom_protosw = COMPATNAME(route_protosw), 1782 .dom_protoswNPROTOSW = 1783 &COMPATNAME(route_protosw)[__arraycount(COMPATNAME(route_protosw))], 1784 #ifdef MBUFTRACE 1785 .dom_mowner = MOWNER_INIT("route", "rtm"), 1786 #endif 1787 }; 1788