1 /* $NetBSD: rtsock_shared.c,v 1.19 2020/06/26 15:53:59 roy 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 1.19 2020/06/26 15:53:59 roy 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 /* sa_family is after sa_len, rest is data */ 100 #define _SA_MINSIZE (offsetof(struct sockaddr, sa_family) + \ 101 sizeof(((struct sockaddr *)0)->sa_family)) 102 103 #ifdef COMPAT_RTSOCK 104 /* 105 * These are used when #include-d from compat/common/rtsock_50.c 106 */ 107 #define RTM_XVERSION RTM_OVERSION 108 #define RTM_XNEWADDR RTM_ONEWADDR 109 #define RTM_XDELADDR RTM_ODELADDR 110 #define RTM_XCHGADDR RTM_OCHGADDR 111 #define RT_XADVANCE(a,b) RT_OADVANCE(a,b) 112 #define RT_XROUNDUP(n) RT_OROUNDUP(n) 113 #define PF_XROUTE PF_OROUTE 114 #define rt_xmsghdr rt_msghdr50 115 #define if_xmsghdr if_msghdr /* if_msghdr50 is for RTM_OIFINFO */ 116 #define ifa_xmsghdr ifa_msghdr50 117 #define if_xannouncemsghdr if_announcemsghdr50 118 #define COMPATNAME(x) compat_50_ ## x 119 #define DOMAINNAME "oroute" 120 #define COMPATCALL(name, args) \ 121 MODULE_HOOK_CALL_VOID(rtsock_ ## name ## _50_hook, args, __nothing); 122 #define RTS_CTASSERT(x) __nothing 123 CTASSERT(sizeof(struct ifa_xmsghdr) == 20); 124 DOMAIN_DEFINE(compat_50_routedomain); /* forward declare and add to link set */ 125 #else /* COMPAT_RTSOCK */ 126 /* 127 * These are used when #include-d from compat/common/rtsock_50.c 128 */ 129 #define RTM_XVERSION RTM_VERSION 130 #define RTM_XNEWADDR RTM_NEWADDR 131 #define RTM_XDELADDR RTM_DELADDR 132 #define RTM_XCHGADDR RTM_CHGADDR 133 #define RT_XADVANCE(a,b) RT_ADVANCE(a,b) 134 #define RT_XROUNDUP(n) RT_ROUNDUP(n) 135 #define PF_XROUTE PF_ROUTE 136 #define rt_xmsghdr rt_msghdr 137 #define if_xmsghdr if_msghdr 138 #define ifa_xmsghdr ifa_msghdr 139 #define if_xannouncemsghdr if_announcemsghdr 140 #define COMPATNAME(x) x 141 #define DOMAINNAME "route" 142 #define COMPATCALL(name, args) __nothing; 143 #define RTS_CTASSERT(x) CTASSERT(x) 144 CTASSERT(sizeof(struct ifa_xmsghdr) == 32); 145 DOMAIN_DEFINE(routedomain); /* forward declare and add to link set */ 146 #endif /* COMPAT_RTSOCK */ 147 148 #ifdef RTSOCK_DEBUG 149 #define RT_IN_PRINT(info, b, a) (in_print((b), sizeof(b), \ 150 &((const struct sockaddr_in *)(info)->rti_info[(a)])->sin_addr), (b)) 151 #endif /* RTSOCK_DEBUG */ 152 153 struct route_info COMPATNAME(route_info) = { 154 .ri_dst = { .sa_len = 2, .sa_family = PF_XROUTE, }, 155 .ri_src = { .sa_len = 2, .sa_family = PF_XROUTE, }, 156 .ri_maxqlen = IFQ_MAXLEN, 157 }; 158 159 static void COMPATNAME(route_init)(void); 160 static int COMPATNAME(route_output)(struct mbuf *, struct socket *); 161 162 static int rt_xaddrs(u_char, const char *, const char *, struct rt_addrinfo *); 163 static struct mbuf *rt_makeifannouncemsg(struct ifnet *, int, int, 164 struct rt_addrinfo *); 165 static int rt_msg2(int, struct rt_addrinfo *, void *, struct rt_walkarg *, int *); 166 static void _rt_setmetrics(int, const struct rt_xmsghdr *, struct rtentry *); 167 static void rtm_setmetrics(const struct rtentry *, struct rt_xmsghdr *); 168 static void rt_adjustcount(int, int); 169 170 static const struct protosw COMPATNAME(route_protosw)[]; 171 172 struct routecb { 173 struct rawcb rocb_rcb; 174 unsigned int rocb_msgfilter; 175 #define RTMSGFILTER(m) (1U << (m)) 176 char *rocb_missfilter; 177 size_t rocb_missfilterlen; 178 }; 179 #define sotoroutecb(so) ((struct routecb *)(so)->so_pcb) 180 181 static struct rawcbhead rt_rawcb; 182 #ifdef NET_MPSAFE 183 static kmutex_t *rt_so_mtx; 184 185 static bool rt_updating = false; 186 static kcondvar_t rt_update_cv; 187 #endif 188 189 static void 190 rt_adjustcount(int af, int cnt) 191 { 192 struct route_cb * const cb = &COMPATNAME(route_info).ri_cb; 193 194 cb->any_count += cnt; 195 196 switch (af) { 197 case AF_INET: 198 cb->ip_count += cnt; 199 return; 200 #ifdef INET6 201 case AF_INET6: 202 cb->ip6_count += cnt; 203 return; 204 #endif 205 case AF_MPLS: 206 cb->mpls_count += cnt; 207 return; 208 } 209 } 210 211 static int 212 COMPATNAME(route_filter)(struct mbuf *m, struct sockproto *proto, 213 struct rawcb *rp) 214 { 215 struct routecb *rop = (struct routecb *)rp; 216 struct rt_xmsghdr *rtm; 217 218 KASSERT(m != NULL); 219 KASSERT(proto != NULL); 220 KASSERT(rp != NULL); 221 222 /* Wrong family for this socket. */ 223 if (proto->sp_family != PF_ROUTE) 224 return ENOPROTOOPT; 225 226 /* If no filter set, just return. */ 227 if (rop->rocb_msgfilter == 0 && rop->rocb_missfilterlen == 0) 228 return 0; 229 230 /* Ensure we can access rtm_type */ 231 if (m->m_len < 232 offsetof(struct rt_xmsghdr, rtm_type) + sizeof(rtm->rtm_type)) 233 return EINVAL; 234 235 rtm = mtod(m, struct rt_xmsghdr *); 236 if (rtm->rtm_type >= sizeof(rop->rocb_msgfilter) * CHAR_BIT) 237 return EINVAL; 238 /* If the rtm type is filtered out, return a positive. */ 239 if (rop->rocb_msgfilter != 0 && 240 !(rop->rocb_msgfilter & RTMSGFILTER(rtm->rtm_type))) 241 return EEXIST; 242 243 if (rop->rocb_missfilterlen != 0 && rtm->rtm_type == RTM_MISS) { 244 __CTASSERT(RTAX_DST == 0); 245 struct sockaddr_storage ss; 246 struct sockaddr *dst = (struct sockaddr *)&ss, *sa; 247 char *cp = rop->rocb_missfilter; 248 char *ep = cp + rop->rocb_missfilterlen; 249 250 /* Ensure we can access sa_len */ 251 if (m->m_pkthdr.len < sizeof(*rtm) + _SA_MINSIZE) 252 return EINVAL; 253 m_copydata(m, sizeof(*rtm) + offsetof(struct sockaddr, sa_len), 254 sizeof(ss.ss_len), &ss.ss_len); 255 if (ss.ss_len < _SA_MINSIZE || 256 ss.ss_len > sizeof(ss) || 257 m->m_pkthdr.len < sizeof(*rtm) + ss.ss_len) 258 return EINVAL; 259 /* Copy out the destination sockaddr */ 260 m_copydata(m, sizeof(*rtm), ss.ss_len, &ss); 261 262 /* Find a matching sockaddr in the filter */ 263 while (cp < ep) { 264 sa = (struct sockaddr *)cp; 265 if (sa->sa_len == dst->sa_len && 266 memcmp(sa, dst, sa->sa_len) == 0) 267 break; 268 cp += RT_XROUNDUP(sa->sa_len); 269 } 270 if (cp == ep) 271 return EEXIST; 272 } 273 274 /* Passed the filter. */ 275 return 0; 276 } 277 278 static void 279 rt_pr_init(void) 280 { 281 282 LIST_INIT(&rt_rawcb); 283 } 284 285 static int 286 COMPATNAME(route_attach)(struct socket *so, int proto) 287 { 288 struct rawcb *rp; 289 struct routecb *rop; 290 int s, error; 291 292 KASSERT(sotorawcb(so) == NULL); 293 rop = kmem_zalloc(sizeof(*rop), KM_SLEEP); 294 rp = &rop->rocb_rcb; 295 rp->rcb_len = sizeof(*rop); 296 so->so_pcb = rp; 297 298 s = splsoftnet(); 299 300 #ifdef NET_MPSAFE 301 KASSERT(so->so_lock == NULL); 302 mutex_obj_hold(rt_so_mtx); 303 so->so_lock = rt_so_mtx; 304 solock(so); 305 #endif 306 307 if ((error = raw_attach(so, proto, &rt_rawcb)) == 0) { 308 rt_adjustcount(rp->rcb_proto.sp_protocol, 1); 309 rp->rcb_laddr = &COMPATNAME(route_info).ri_src; 310 rp->rcb_faddr = &COMPATNAME(route_info).ri_dst; 311 rp->rcb_filter = COMPATNAME(route_filter); 312 } 313 splx(s); 314 315 if (error) { 316 kmem_free(rop, sizeof(*rop)); 317 so->so_pcb = NULL; 318 return error; 319 } 320 321 soisconnected(so); 322 so->so_options |= SO_USELOOPBACK; 323 KASSERT(solocked(so)); 324 325 return error; 326 } 327 328 static void 329 COMPATNAME(route_detach)(struct socket *so) 330 { 331 struct rawcb *rp = sotorawcb(so); 332 struct routecb *rop = (struct routecb *)rp; 333 int s; 334 335 KASSERT(rp != NULL); 336 KASSERT(solocked(so)); 337 338 s = splsoftnet(); 339 if (rop->rocb_missfilterlen != 0) 340 kmem_free(rop->rocb_missfilter, rop->rocb_missfilterlen); 341 rt_adjustcount(rp->rcb_proto.sp_protocol, -1); 342 raw_detach(so); 343 splx(s); 344 } 345 346 static int 347 COMPATNAME(route_accept)(struct socket *so, struct sockaddr *nam) 348 { 349 KASSERT(solocked(so)); 350 351 panic("route_accept"); 352 353 return EOPNOTSUPP; 354 } 355 356 static int 357 COMPATNAME(route_bind)(struct socket *so, struct sockaddr *nam, struct lwp *l) 358 { 359 KASSERT(solocked(so)); 360 361 return EOPNOTSUPP; 362 } 363 364 static int 365 COMPATNAME(route_listen)(struct socket *so, struct lwp *l) 366 { 367 KASSERT(solocked(so)); 368 369 return EOPNOTSUPP; 370 } 371 372 static int 373 COMPATNAME(route_connect)(struct socket *so, struct sockaddr *nam, struct lwp *l) 374 { 375 KASSERT(solocked(so)); 376 377 return EOPNOTSUPP; 378 } 379 380 static int 381 COMPATNAME(route_connect2)(struct socket *so, struct socket *so2) 382 { 383 KASSERT(solocked(so)); 384 385 return EOPNOTSUPP; 386 } 387 388 static int 389 COMPATNAME(route_disconnect)(struct socket *so) 390 { 391 struct rawcb *rp = sotorawcb(so); 392 int s; 393 394 KASSERT(solocked(so)); 395 KASSERT(rp != NULL); 396 397 s = splsoftnet(); 398 soisdisconnected(so); 399 raw_disconnect(rp); 400 splx(s); 401 402 return 0; 403 } 404 405 static int 406 COMPATNAME(route_shutdown)(struct socket *so) 407 { 408 int s; 409 410 KASSERT(solocked(so)); 411 412 /* 413 * Mark the connection as being incapable of further input. 414 */ 415 s = splsoftnet(); 416 socantsendmore(so); 417 splx(s); 418 return 0; 419 } 420 421 static int 422 COMPATNAME(route_abort)(struct socket *so) 423 { 424 KASSERT(solocked(so)); 425 426 panic("route_abort"); 427 428 return EOPNOTSUPP; 429 } 430 431 static int 432 COMPATNAME(route_ioctl)(struct socket *so, u_long cmd, void *nam, 433 struct ifnet * ifp) 434 { 435 return EOPNOTSUPP; 436 } 437 438 static int 439 COMPATNAME(route_stat)(struct socket *so, struct stat *ub) 440 { 441 KASSERT(solocked(so)); 442 443 return 0; 444 } 445 446 static int 447 COMPATNAME(route_peeraddr)(struct socket *so, struct sockaddr *nam) 448 { 449 struct rawcb *rp = sotorawcb(so); 450 451 KASSERT(solocked(so)); 452 KASSERT(rp != NULL); 453 KASSERT(nam != NULL); 454 455 if (rp->rcb_faddr == NULL) 456 return ENOTCONN; 457 458 raw_setpeeraddr(rp, nam); 459 return 0; 460 } 461 462 static int 463 COMPATNAME(route_sockaddr)(struct socket *so, struct sockaddr *nam) 464 { 465 struct rawcb *rp = sotorawcb(so); 466 467 KASSERT(solocked(so)); 468 KASSERT(rp != NULL); 469 KASSERT(nam != NULL); 470 471 if (rp->rcb_faddr == NULL) 472 return ENOTCONN; 473 474 raw_setsockaddr(rp, nam); 475 return 0; 476 } 477 478 static int 479 COMPATNAME(route_rcvd)(struct socket *so, int flags, struct lwp *l) 480 { 481 KASSERT(solocked(so)); 482 483 return EOPNOTSUPP; 484 } 485 486 static int 487 COMPATNAME(route_recvoob)(struct socket *so, struct mbuf *m, int flags) 488 { 489 KASSERT(solocked(so)); 490 491 return EOPNOTSUPP; 492 } 493 494 static int 495 COMPATNAME(route_send)(struct socket *so, struct mbuf *m, 496 struct sockaddr *nam, struct mbuf *control, struct lwp *l) 497 { 498 int error = 0; 499 int s; 500 501 KASSERT(solocked(so)); 502 KASSERT(so->so_proto == &COMPATNAME(route_protosw)[0]); 503 504 s = splsoftnet(); 505 error = raw_send(so, m, nam, control, l, &COMPATNAME(route_output)); 506 splx(s); 507 508 return error; 509 } 510 511 static int 512 COMPATNAME(route_sendoob)(struct socket *so, struct mbuf *m, 513 struct mbuf *control) 514 { 515 KASSERT(solocked(so)); 516 517 m_freem(m); 518 m_freem(control); 519 520 return EOPNOTSUPP; 521 } 522 static int 523 COMPATNAME(route_purgeif)(struct socket *so, struct ifnet *ifp) 524 { 525 526 panic("route_purgeif"); 527 528 return EOPNOTSUPP; 529 } 530 531 #if defined(INET) || defined(INET6) 532 static int 533 route_get_sdl_index(struct rt_addrinfo *info, int *sdl_index) 534 { 535 struct rtentry *nrt; 536 int error; 537 538 error = rtrequest1(RTM_GET, info, &nrt); 539 if (error != 0) 540 return error; 541 /* 542 * nrt->rt_ifp->if_index may not be correct 543 * due to changing to ifplo0. 544 */ 545 *sdl_index = satosdl(nrt->rt_gateway)->sdl_index; 546 rt_unref(nrt); 547 548 return 0; 549 } 550 #endif 551 552 static void 553 route_get_sdl(const struct ifnet *ifp, const struct sockaddr *dst, 554 struct sockaddr_dl *sdl, int *flags) 555 { 556 struct llentry *la; 557 558 KASSERT(ifp != NULL); 559 560 IF_AFDATA_RLOCK(ifp); 561 switch (dst->sa_family) { 562 case AF_INET: 563 la = lla_lookup(LLTABLE(ifp), 0, dst); 564 break; 565 case AF_INET6: 566 la = lla_lookup(LLTABLE6(ifp), 0, dst); 567 break; 568 default: 569 la = NULL; 570 KASSERTMSG(0, "Invalid AF=%d\n", dst->sa_family); 571 break; 572 } 573 IF_AFDATA_RUNLOCK(ifp); 574 575 void *a = (LLE_IS_VALID(la) && (la->la_flags & LLE_VALID) == LLE_VALID) 576 ? &la->ll_addr : NULL; 577 578 a = sockaddr_dl_init(sdl, sizeof(*sdl), ifp->if_index, ifp->if_type, 579 NULL, 0, a, ifp->if_addrlen); 580 KASSERT(a != NULL); 581 582 if (la != NULL) { 583 *flags = la->la_flags; 584 LLE_RUNLOCK(la); 585 } 586 } 587 588 static int 589 route_output_report(struct rtentry *rt, struct rt_addrinfo *info, 590 struct rt_xmsghdr *rtm, struct rt_xmsghdr **new_rtm) 591 { 592 int len, error; 593 594 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) { 595 const struct ifaddr *rtifa; 596 const struct ifnet *ifp = rt->rt_ifp; 597 598 info->rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr; 599 /* rtifa used to be simply rt->rt_ifa. 600 * If rt->rt_ifa != NULL, then 601 * rt_get_ifa() != NULL. So this 602 * ought to still be safe. --dyoung 603 */ 604 rtifa = rt_get_ifa(rt); 605 info->rti_info[RTAX_IFA] = rtifa->ifa_addr; 606 #ifdef RTSOCK_DEBUG 607 if (info->rti_info[RTAX_IFA]->sa_family == AF_INET) { 608 char ibuf[INET_ADDRSTRLEN]; 609 char abuf[INET_ADDRSTRLEN]; 610 printf("%s: copying out RTAX_IFA %s " 611 "for info->rti_info[RTAX_DST] %s " 612 "ifa_getifa %p ifa_seqno %p\n", 613 __func__, 614 RT_IN_PRINT(info, ibuf, RTAX_IFA), 615 RT_IN_PRINT(info, abuf, RTAX_DST), 616 (void *)rtifa->ifa_getifa, 617 rtifa->ifa_seqno); 618 } 619 #endif /* RTSOCK_DEBUG */ 620 if (ifp->if_flags & IFF_POINTOPOINT) 621 info->rti_info[RTAX_BRD] = rtifa->ifa_dstaddr; 622 else 623 info->rti_info[RTAX_BRD] = NULL; 624 rtm->rtm_index = ifp->if_index; 625 } 626 error = rt_msg2(rtm->rtm_type, info, NULL, NULL, &len); 627 if (error) 628 return error; 629 if (len > rtm->rtm_msglen) { 630 struct rt_xmsghdr *old_rtm = rtm; 631 R_Malloc(*new_rtm, struct rt_xmsghdr *, len); 632 if (*new_rtm == NULL) 633 return ENOBUFS; 634 (void)memcpy(*new_rtm, old_rtm, old_rtm->rtm_msglen); 635 rtm = *new_rtm; 636 } 637 (void)rt_msg2(rtm->rtm_type, info, rtm, NULL, 0); 638 rtm->rtm_flags = rt->rt_flags; 639 rtm_setmetrics(rt, rtm); 640 rtm->rtm_addrs = info->rti_addrs; 641 642 return 0; 643 } 644 645 /*ARGSUSED*/ 646 int 647 COMPATNAME(route_output)(struct mbuf *m, struct socket *so) 648 { 649 struct sockproto proto = { .sp_family = PF_XROUTE, }; 650 struct rt_xmsghdr *rtm = NULL; 651 struct rt_xmsghdr *old_rtm = NULL, *new_rtm = NULL; 652 struct rtentry *rt = NULL; 653 struct rtentry *saved_nrt = NULL; 654 struct rt_addrinfo info; 655 int len, error = 0; 656 sa_family_t family; 657 struct sockaddr_dl sdl; 658 int bound = curlwp_bind(); 659 bool do_rt_free = false; 660 struct sockaddr_storage netmask; 661 662 #define senderr(e) do { error = e; goto flush;} while (/*CONSTCOND*/ 0) 663 if (m == NULL || ((m->m_len < sizeof(int32_t)) && 664 (m = m_pullup(m, sizeof(int32_t))) == NULL)) { 665 error = ENOBUFS; 666 goto out; 667 } 668 if ((m->m_flags & M_PKTHDR) == 0) 669 panic("%s", __func__); 670 len = m->m_pkthdr.len; 671 if (len < sizeof(*rtm) || 672 len != mtod(m, struct rt_xmsghdr *)->rtm_msglen) { 673 info.rti_info[RTAX_DST] = NULL; 674 senderr(EINVAL); 675 } 676 R_Malloc(rtm, struct rt_xmsghdr *, len); 677 if (rtm == NULL) { 678 info.rti_info[RTAX_DST] = NULL; 679 senderr(ENOBUFS); 680 } 681 m_copydata(m, 0, len, rtm); 682 if (rtm->rtm_version != RTM_XVERSION) { 683 info.rti_info[RTAX_DST] = NULL; 684 senderr(EPROTONOSUPPORT); 685 } 686 rtm->rtm_pid = curproc->p_pid; 687 memset(&info, 0, sizeof(info)); 688 info.rti_addrs = rtm->rtm_addrs; 689 if (rt_xaddrs(rtm->rtm_type, (const char *)(rtm + 1), len + (char *)rtm, 690 &info)) { 691 senderr(EINVAL); 692 } 693 info.rti_flags = rtm->rtm_flags; 694 if (info.rti_info[RTAX_DST] == NULL || 695 (info.rti_info[RTAX_DST]->sa_family >= AF_MAX)) { 696 senderr(EINVAL); 697 } 698 #ifdef RTSOCK_DEBUG 699 if (info.rti_info[RTAX_DST]->sa_family == AF_INET) { 700 char abuf[INET_ADDRSTRLEN]; 701 printf("%s: extracted info.rti_info[RTAX_DST] %s\n", __func__, 702 RT_IN_PRINT(&info, abuf, RTAX_DST)); 703 } 704 #endif /* RTSOCK_DEBUG */ 705 if (info.rti_info[RTAX_GATEWAY] != NULL && 706 (info.rti_info[RTAX_GATEWAY]->sa_family >= AF_MAX)) { 707 senderr(EINVAL); 708 } 709 710 /* 711 * Verify that the socket has the appropriate privilege; RTM_GET 712 * is the only operation the non-superuser is allowed. 713 */ 714 if (kauth_authorize_network(so->so_cred, KAUTH_NETWORK_ROUTE, 715 0, rtm, NULL, NULL) != 0) 716 senderr(EACCES); 717 718 /* 719 * route(8) passes a sockaddr truncated with prefixlen. 720 * The kernel doesn't expect such sockaddr and need to 721 * use a buffer that is big enough for the sockaddr expected 722 * (padded with 0's). We keep the original length of the sockaddr. 723 */ 724 if (info.rti_info[RTAX_NETMASK]) { 725 /* 726 * Use the family of RTAX_DST, because RTAX_NETMASK 727 * can have a zero family if it comes from the radix 728 * tree via rt_mask(). 729 */ 730 socklen_t sa_len = sockaddr_getsize_by_family( 731 info.rti_info[RTAX_DST]->sa_family); 732 socklen_t masklen = sockaddr_getlen( 733 info.rti_info[RTAX_NETMASK]); 734 if (sa_len != 0 && sa_len > masklen) { 735 KASSERT(sa_len <= sizeof(netmask)); 736 memcpy(&netmask, info.rti_info[RTAX_NETMASK], masklen); 737 memset((char *)&netmask + masklen, 0, sa_len - masklen); 738 info.rti_info[RTAX_NETMASK] = sstocsa(&netmask); 739 } 740 } 741 742 switch (rtm->rtm_type) { 743 744 case RTM_ADD: 745 if (info.rti_info[RTAX_GATEWAY] == NULL) { 746 senderr(EINVAL); 747 } 748 #if defined(INET) || defined(INET6) 749 /* support for new ARP/NDP code with keeping backcompat */ 750 if (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) { 751 const struct sockaddr_dl *sdlp = 752 satocsdl(info.rti_info[RTAX_GATEWAY]); 753 754 /* Allow routing requests by interface index */ 755 if (sdlp->sdl_nlen == 0 && sdlp->sdl_alen == 0 756 && sdlp->sdl_slen == 0) 757 goto fallback; 758 /* 759 * Old arp binaries don't set the sdl_index 760 * so we have to complement it. 761 */ 762 int sdl_index = sdlp->sdl_index; 763 if (sdl_index == 0) { 764 error = route_get_sdl_index(&info, &sdl_index); 765 if (error != 0) 766 goto fallback; 767 } else if ( 768 info.rti_info[RTAX_DST]->sa_family == AF_INET) { 769 /* 770 * XXX workaround for SIN_PROXY case; proxy arp 771 * entry should be in an interface that has 772 * a network route including the destination, 773 * not a local (link) route that may not be a 774 * desired place, for example a tap. 775 */ 776 const struct sockaddr_inarp *sina = 777 (const struct sockaddr_inarp *) 778 info.rti_info[RTAX_DST]; 779 if (sina->sin_other & SIN_PROXY) { 780 error = route_get_sdl_index(&info, 781 &sdl_index); 782 if (error != 0) 783 goto fallback; 784 } 785 } 786 error = lla_rt_output(rtm->rtm_type, rtm->rtm_flags, 787 rtm->rtm_rmx.rmx_expire, &info, sdl_index); 788 break; 789 } 790 fallback: 791 #endif /* defined(INET) || defined(INET6) */ 792 error = rtrequest1(rtm->rtm_type, &info, &saved_nrt); 793 if (error == 0) { 794 _rt_setmetrics(rtm->rtm_inits, rtm, saved_nrt); 795 rt_unref(saved_nrt); 796 } 797 break; 798 799 case RTM_DELETE: 800 #if defined(INET) || defined(INET6) 801 /* support for new ARP/NDP code */ 802 if (info.rti_info[RTAX_GATEWAY] && 803 (info.rti_info[RTAX_GATEWAY]->sa_family == AF_LINK) && 804 (rtm->rtm_flags & RTF_LLDATA) != 0) { 805 const struct sockaddr_dl *sdlp = 806 satocsdl(info.rti_info[RTAX_GATEWAY]); 807 error = lla_rt_output(rtm->rtm_type, rtm->rtm_flags, 808 rtm->rtm_rmx.rmx_expire, &info, sdlp->sdl_index); 809 rtm->rtm_flags &= ~RTF_UP; 810 break; 811 } 812 #endif 813 error = rtrequest1(rtm->rtm_type, &info, &saved_nrt); 814 if (error != 0) 815 break; 816 817 rt = saved_nrt; 818 do_rt_free = true; 819 info.rti_info[RTAX_DST] = rt_getkey(rt); 820 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 821 info.rti_info[RTAX_NETMASK] = rt_mask(rt); 822 info.rti_info[RTAX_TAG] = rt_gettag(rt); 823 error = route_output_report(rt, &info, rtm, &new_rtm); 824 if (error) 825 senderr(error); 826 if (new_rtm != NULL) { 827 old_rtm = rtm; 828 rtm = new_rtm; 829 } 830 break; 831 832 case RTM_GET: 833 case RTM_CHANGE: 834 case RTM_LOCK: 835 /* XXX This will mask info.rti_info[RTAX_DST] with 836 * info.rti_info[RTAX_NETMASK] before 837 * searching. It did not used to do that. --dyoung 838 */ 839 rt = NULL; 840 error = rtrequest1(RTM_GET, &info, &rt); 841 if (error != 0) 842 senderr(error); 843 if (rtm->rtm_type != RTM_GET) {/* XXX: too grotty */ 844 if (memcmp(info.rti_info[RTAX_DST], rt_getkey(rt), 845 info.rti_info[RTAX_DST]->sa_len) != 0) 846 senderr(ESRCH); 847 if (info.rti_info[RTAX_NETMASK] == NULL && 848 rt_mask(rt) != NULL) 849 senderr(ETOOMANYREFS); 850 } 851 852 /* 853 * XXX if arp/ndp requests an L2 entry, we have to obtain 854 * it from lltable while for the route command we have to 855 * return a route as it is. How to distinguish them? 856 * For newer arp/ndp, RTF_LLDATA flag set by arp/ndp 857 * indicates an L2 entry is requested. For old arp/ndp 858 * binaries, we check RTF_UP flag is NOT set; it works 859 * by the fact that arp/ndp don't set it while the route 860 * command sets it. 861 */ 862 if (((rtm->rtm_flags & RTF_LLDATA) != 0 || 863 (rtm->rtm_flags & RTF_UP) == 0) && 864 rtm->rtm_type == RTM_GET && 865 sockaddr_cmp(rt_getkey(rt), info.rti_info[RTAX_DST]) != 0) { 866 int ll_flags = 0; 867 route_get_sdl(rt->rt_ifp, info.rti_info[RTAX_DST], &sdl, 868 &ll_flags); 869 info.rti_info[RTAX_GATEWAY] = sstocsa(&sdl); 870 error = route_output_report(rt, &info, rtm, &new_rtm); 871 if (error) 872 senderr(error); 873 if (new_rtm != NULL) { 874 old_rtm = rtm; 875 rtm = new_rtm; 876 } 877 rtm->rtm_flags |= RTF_LLDATA; 878 rtm->rtm_flags &= ~RTF_CONNECTED; 879 rtm->rtm_flags |= (ll_flags & LLE_STATIC) ? RTF_STATIC : 0; 880 break; 881 } 882 883 switch (rtm->rtm_type) { 884 case RTM_GET: 885 info.rti_info[RTAX_DST] = rt_getkey(rt); 886 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 887 info.rti_info[RTAX_NETMASK] = rt_mask(rt); 888 info.rti_info[RTAX_TAG] = rt_gettag(rt); 889 error = route_output_report(rt, &info, rtm, &new_rtm); 890 if (error) 891 senderr(error); 892 if (new_rtm != NULL) { 893 old_rtm = rtm; 894 rtm = new_rtm; 895 } 896 break; 897 898 case RTM_CHANGE: 899 #ifdef NET_MPSAFE 900 /* 901 * Release rt_so_mtx to avoid a deadlock with route_intr 902 * and also serialize updating routes to avoid another. 903 */ 904 if (rt_updating) { 905 /* Release to allow the updater to proceed */ 906 rt_unref(rt); 907 rt = NULL; 908 } 909 while (rt_updating) { 910 error = cv_wait_sig(&rt_update_cv, rt_so_mtx); 911 if (error != 0) 912 goto flush; 913 } 914 if (rt == NULL) { 915 error = rtrequest1(RTM_GET, &info, &rt); 916 if (error != 0) 917 goto flush; 918 } 919 rt_updating = true; 920 mutex_exit(rt_so_mtx); 921 922 error = rt_update_prepare(rt); 923 if (error == 0) { 924 error = rt_update(rt, &info, rtm); 925 rt_update_finish(rt); 926 } 927 928 mutex_enter(rt_so_mtx); 929 rt_updating = false; 930 cv_broadcast(&rt_update_cv); 931 #else 932 error = rt_update(rt, &info, rtm); 933 #endif 934 if (error != 0) 935 goto flush; 936 /*FALLTHROUGH*/ 937 case RTM_LOCK: 938 rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits); 939 rt->rt_rmx.rmx_locks |= 940 (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks); 941 break; 942 } 943 break; 944 945 default: 946 senderr(EOPNOTSUPP); 947 } 948 949 flush: 950 if (rtm) { 951 if (error) 952 rtm->rtm_errno = error; 953 else 954 rtm->rtm_flags |= RTF_DONE; 955 } 956 family = info.rti_info[RTAX_DST] ? info.rti_info[RTAX_DST]->sa_family : 957 0; 958 /* We cannot free old_rtm until we have stopped using the 959 * pointers in info, some of which may point to sockaddrs 960 * in old_rtm. 961 */ 962 if (old_rtm != NULL) 963 Free(old_rtm); 964 if (rt) { 965 if (do_rt_free) { 966 #ifdef NET_MPSAFE 967 /* 968 * Release rt_so_mtx to avoid a deadlock with 969 * route_intr. 970 */ 971 mutex_exit(rt_so_mtx); 972 rt_free(rt); 973 mutex_enter(rt_so_mtx); 974 #else 975 rt_free(rt); 976 #endif 977 } else 978 rt_unref(rt); 979 } 980 { 981 struct rawcb *rp = NULL; 982 /* 983 * Check to see if we don't want our own messages. 984 */ 985 if ((so->so_options & SO_USELOOPBACK) == 0) { 986 if (COMPATNAME(route_info).ri_cb.any_count <= 1) { 987 if (rtm) 988 Free(rtm); 989 m_freem(m); 990 goto out; 991 } 992 /* There is another listener, so construct message */ 993 rp = sotorawcb(so); 994 } 995 if (rtm) { 996 m_copyback(m, 0, rtm->rtm_msglen, rtm); 997 if (m->m_pkthdr.len < rtm->rtm_msglen) { 998 m_freem(m); 999 m = NULL; 1000 } else if (m->m_pkthdr.len > rtm->rtm_msglen) 1001 m_adj(m, rtm->rtm_msglen - m->m_pkthdr.len); 1002 Free(rtm); 1003 } 1004 if (rp) 1005 rp->rcb_proto.sp_family = 0; /* Avoid us */ 1006 if (family) 1007 proto.sp_protocol = family; 1008 if (m) 1009 raw_input(m, &proto, &COMPATNAME(route_info).ri_src, 1010 &COMPATNAME(route_info).ri_dst, &rt_rawcb); 1011 if (rp) 1012 rp->rcb_proto.sp_family = PF_XROUTE; 1013 } 1014 out: 1015 curlwp_bindx(bound); 1016 return error; 1017 } 1018 1019 static int 1020 route_ctloutput(int op, struct socket *so, struct sockopt *sopt) 1021 { 1022 struct routecb *rop = sotoroutecb(so); 1023 int error = 0; 1024 unsigned char *rtm_type, *cp, *ep; 1025 size_t len; 1026 unsigned int msgfilter; 1027 struct sockaddr *sa; 1028 1029 KASSERT(solocked(so)); 1030 1031 if (sopt->sopt_level != AF_ROUTE) { 1032 error = ENOPROTOOPT; 1033 } else switch (op) { 1034 case PRCO_SETOPT: 1035 switch (sopt->sopt_name) { 1036 case RO_MSGFILTER: 1037 msgfilter = 0; 1038 for (rtm_type = sopt->sopt_data, len = sopt->sopt_size; 1039 len != 0; 1040 rtm_type++, len -= sizeof(*rtm_type)) 1041 { 1042 /* Guard against overflowing our storage. */ 1043 if (*rtm_type >= sizeof(msgfilter) * CHAR_BIT) { 1044 error = EOVERFLOW; 1045 break; 1046 } 1047 msgfilter |= RTMSGFILTER(*rtm_type); 1048 } 1049 if (error == 0) 1050 rop->rocb_msgfilter = msgfilter; 1051 break; 1052 case RO_MISSFILTER: 1053 /* Validate the data */ 1054 len = 0; 1055 cp = sopt->sopt_data; 1056 ep = cp + sopt->sopt_size; 1057 while (cp < ep) { 1058 if (ep - cp < 1059 offsetof(struct sockaddr, sa_len) + 1060 sizeof(sa->sa_len)) 1061 break; 1062 if (++len > RO_FILTSA_MAX) { 1063 error = ENOBUFS; 1064 break; 1065 } 1066 sa = (struct sockaddr *)cp; 1067 if (sa->sa_len < _SA_MINSIZE || 1068 sa->sa_len >sizeof(struct sockaddr_storage)) 1069 return EINVAL; 1070 cp += RT_XROUNDUP(sa->sa_len); 1071 } 1072 if (cp != ep) { 1073 if (error == 0) 1074 error = EINVAL; 1075 break; 1076 } 1077 if (rop->rocb_missfilterlen != 0) 1078 kmem_free(rop->rocb_missfilter, 1079 rop->rocb_missfilterlen); 1080 if (sopt->sopt_size != 0) { 1081 rop->rocb_missfilter = 1082 kmem_alloc(sopt->sopt_size, KM_SLEEP); 1083 if (rop->rocb_missfilter == NULL) { 1084 rop->rocb_missfilterlen = 0; 1085 error = ENOBUFS; 1086 break; 1087 } 1088 } else 1089 rop->rocb_missfilter = NULL; 1090 rop->rocb_missfilterlen = sopt->sopt_size; 1091 if (rop->rocb_missfilterlen != 0) 1092 memcpy(rop->rocb_missfilter, sopt->sopt_data, 1093 rop->rocb_missfilterlen); 1094 break; 1095 default: 1096 error = ENOPROTOOPT; 1097 break; 1098 } 1099 break; 1100 case PRCO_GETOPT: 1101 switch (sopt->sopt_name) { 1102 case RO_MSGFILTER: 1103 error = ENOTSUP; 1104 break; 1105 default: 1106 error = ENOPROTOOPT; 1107 break; 1108 } 1109 } 1110 return error; 1111 } 1112 1113 static void 1114 _rt_setmetrics(int which, const struct rt_xmsghdr *in, struct rtentry *out) 1115 { 1116 #define metric(f, e) if (which & (f)) out->rt_rmx.e = in->rtm_rmx.e; 1117 metric(RTV_RPIPE, rmx_recvpipe); 1118 metric(RTV_SPIPE, rmx_sendpipe); 1119 metric(RTV_SSTHRESH, rmx_ssthresh); 1120 metric(RTV_RTT, rmx_rtt); 1121 metric(RTV_RTTVAR, rmx_rttvar); 1122 metric(RTV_HOPCOUNT, rmx_hopcount); 1123 metric(RTV_MTU, rmx_mtu); 1124 #undef metric 1125 if (which & RTV_EXPIRE) { 1126 out->rt_rmx.rmx_expire = in->rtm_rmx.rmx_expire ? 1127 time_wall_to_mono(in->rtm_rmx.rmx_expire) : 0; 1128 } 1129 } 1130 1131 static void 1132 rtm_setmetrics(const struct rtentry *in, struct rt_xmsghdr *out) 1133 { 1134 #define metric(e) out->rtm_rmx.e = in->rt_rmx.e; 1135 metric(rmx_recvpipe); 1136 metric(rmx_sendpipe); 1137 metric(rmx_ssthresh); 1138 metric(rmx_rtt); 1139 metric(rmx_rttvar); 1140 metric(rmx_hopcount); 1141 metric(rmx_mtu); 1142 metric(rmx_locks); 1143 #undef metric 1144 out->rtm_rmx.rmx_expire = in->rt_rmx.rmx_expire ? 1145 time_mono_to_wall(in->rt_rmx.rmx_expire) : 0; 1146 } 1147 1148 static int 1149 rt_xaddrs(u_char rtmtype, const char *cp, const char *cplim, 1150 struct rt_addrinfo *rtinfo) 1151 { 1152 const struct sockaddr *sa = NULL; /* Quell compiler warning */ 1153 int i; 1154 1155 for (i = 0; i < RTAX_MAX && cp < cplim; i++) { 1156 if ((rtinfo->rti_addrs & (1 << i)) == 0) 1157 continue; 1158 rtinfo->rti_info[i] = sa = (const struct sockaddr *)cp; 1159 RT_XADVANCE(cp, sa); 1160 } 1161 1162 /* 1163 * Check for extra addresses specified, except RTM_GET asking 1164 * for interface info. 1165 */ 1166 if (rtmtype == RTM_GET) { 1167 if (((rtinfo->rti_addrs & 1168 (~((1 << RTAX_IFP) | (1 << RTAX_IFA)))) & (~0U << i)) != 0) 1169 return 1; 1170 } else if ((rtinfo->rti_addrs & (~0U << i)) != 0) 1171 return 1; 1172 /* Check for bad data length. */ 1173 if (cp != cplim) { 1174 if (i == RTAX_NETMASK + 1 && sa != NULL && 1175 cp - RT_XROUNDUP(sa->sa_len) + sa->sa_len == cplim) 1176 /* 1177 * The last sockaddr was info.rti_info[RTAX_NETMASK]. 1178 * We accept this for now for the sake of old 1179 * binaries or third party softwares. 1180 */ 1181 ; 1182 else 1183 return 1; 1184 } 1185 return 0; 1186 } 1187 1188 static int 1189 rt_getlen(int type) 1190 { 1191 RTS_CTASSERT(__alignof(struct ifa_msghdr) >= sizeof(uint64_t)); 1192 RTS_CTASSERT(__alignof(struct if_msghdr) >= sizeof(uint64_t)); 1193 RTS_CTASSERT(__alignof(struct if_announcemsghdr) >= sizeof(uint64_t)); 1194 RTS_CTASSERT(__alignof(struct rt_msghdr) >= sizeof(uint64_t)); 1195 1196 switch (type) { 1197 case RTM_ODELADDR: 1198 case RTM_ONEWADDR: 1199 case RTM_OCHGADDR: 1200 if (rtsock_iflist_70_hook.hooked) 1201 return sizeof(struct ifa_msghdr70); 1202 else { 1203 #ifdef RTSOCK_DEBUG 1204 printf("%s: unsupported RTM type %d\n", __func__, type); 1205 #endif 1206 return -1; 1207 } 1208 1209 case RTM_DELADDR: 1210 case RTM_NEWADDR: 1211 case RTM_CHGADDR: 1212 return sizeof(struct ifa_xmsghdr); 1213 1214 case RTM_OOIFINFO: 1215 if (rtsock_iflist_14_hook.hooked) 1216 return sizeof(struct if_msghdr14); 1217 else { 1218 #ifdef RTSOCK_DEBUG 1219 printf("%s: unsupported RTM type RTM_OOIFINFO\n", 1220 __func__); 1221 #endif 1222 return -1; 1223 } 1224 1225 case RTM_OIFINFO: 1226 if (rtsock_iflist_50_hook.hooked) 1227 return sizeof(struct if_msghdr50); 1228 else { 1229 #ifdef RTSOCK_DEBUG 1230 printf("%s: unsupported RTM type RTM_OIFINFO\n", 1231 __func__); 1232 #endif 1233 return -1; 1234 } 1235 1236 case RTM_IFINFO: 1237 return sizeof(struct if_xmsghdr); 1238 1239 case RTM_IFANNOUNCE: 1240 case RTM_IEEE80211: 1241 return sizeof(struct if_xannouncemsghdr); 1242 1243 default: 1244 return sizeof(struct rt_xmsghdr); 1245 } 1246 } 1247 1248 1249 struct mbuf * 1250 COMPATNAME(rt_msg1)(int type, struct rt_addrinfo *rtinfo, void *data, int datalen) 1251 { 1252 struct rt_xmsghdr *rtm; 1253 struct mbuf *m; 1254 int i; 1255 const struct sockaddr *sa; 1256 int len, dlen; 1257 1258 m = m_gethdr(M_DONTWAIT, MT_DATA); 1259 if (m == NULL) 1260 return m; 1261 MCLAIM(m, &COMPATNAME(routedomain).dom_mowner); 1262 1263 if ((len = rt_getlen(type)) == -1) 1264 goto out; 1265 if (len > MHLEN + MLEN) 1266 panic("%s: message too long", __func__); 1267 else if (len > MHLEN) { 1268 m->m_next = m_get(M_DONTWAIT, MT_DATA); 1269 if (m->m_next == NULL) 1270 goto out; 1271 MCLAIM(m->m_next, m->m_owner); 1272 m->m_pkthdr.len = len; 1273 m->m_len = MHLEN; 1274 m->m_next->m_len = len - MHLEN; 1275 } else { 1276 m->m_pkthdr.len = m->m_len = len; 1277 } 1278 m_reset_rcvif(m); 1279 m_copyback(m, 0, datalen, data); 1280 if (len > datalen) 1281 (void)memset(mtod(m, char *) + datalen, 0, len - datalen); 1282 rtm = mtod(m, struct rt_xmsghdr *); 1283 for (i = 0; i < RTAX_MAX; i++) { 1284 if ((sa = rtinfo->rti_info[i]) == NULL) 1285 continue; 1286 rtinfo->rti_addrs |= (1 << i); 1287 dlen = RT_XROUNDUP(sa->sa_len); 1288 m_copyback(m, len, sa->sa_len, sa); 1289 if (dlen != sa->sa_len) { 1290 /* 1291 * Up to 7 + 1 nul's since roundup is to 1292 * sizeof(uint64_t) (8 bytes) 1293 */ 1294 m_copyback(m, len + sa->sa_len, 1295 dlen - sa->sa_len, "\0\0\0\0\0\0\0"); 1296 } 1297 len += dlen; 1298 } 1299 if (m->m_pkthdr.len != len) 1300 goto out; 1301 rtm->rtm_msglen = len; 1302 rtm->rtm_version = RTM_XVERSION; 1303 rtm->rtm_type = type; 1304 return m; 1305 out: 1306 m_freem(m); 1307 return NULL; 1308 } 1309 1310 /* 1311 * rt_msg2 1312 * 1313 * fills 'cp' or 'w'.w_tmem with the routing socket message and 1314 * returns the length of the message in 'lenp'. 1315 * 1316 * if walkarg is 0, cp is expected to be 0 or a buffer large enough to hold 1317 * the message 1318 * otherwise walkarg's w_needed is updated and if the user buffer is 1319 * specified and w_needed indicates space exists the information is copied 1320 * into the temp space (w_tmem). w_tmem is [re]allocated if necessary, 1321 * if the allocation fails ENOBUFS is returned. 1322 */ 1323 static int 1324 rt_msg2(int type, struct rt_addrinfo *rtinfo, void *cpv, struct rt_walkarg *w, 1325 int *lenp) 1326 { 1327 int i; 1328 int len, dlen, second_time = 0; 1329 char *cp0, *cp = cpv; 1330 1331 rtinfo->rti_addrs = 0; 1332 again: 1333 if ((len = rt_getlen(type)) == -1) 1334 return EINVAL; 1335 1336 if ((cp0 = cp) != NULL) 1337 cp += len; 1338 for (i = 0; i < RTAX_MAX; i++) { 1339 const struct sockaddr *sa; 1340 1341 if ((sa = rtinfo->rti_info[i]) == NULL) 1342 continue; 1343 rtinfo->rti_addrs |= (1 << i); 1344 dlen = RT_XROUNDUP(sa->sa_len); 1345 if (cp) { 1346 int diff = dlen - sa->sa_len; 1347 (void)memcpy(cp, sa, (size_t)sa->sa_len); 1348 cp += sa->sa_len; 1349 if (diff > 0) { 1350 (void)memset(cp, 0, (size_t)diff); 1351 cp += diff; 1352 } 1353 } 1354 len += dlen; 1355 } 1356 if (cp == NULL && w != NULL && !second_time) { 1357 struct rt_walkarg *rw = w; 1358 1359 rw->w_needed += len; 1360 if (rw->w_needed <= 0 && rw->w_where) { 1361 if (rw->w_tmemsize < len) { 1362 if (rw->w_tmem) 1363 kmem_free(rw->w_tmem, rw->w_tmemsize); 1364 rw->w_tmem = kmem_zalloc(len, KM_SLEEP); 1365 rw->w_tmemsize = len; 1366 } 1367 if (rw->w_tmem) { 1368 cp = rw->w_tmem; 1369 second_time = 1; 1370 goto again; 1371 } else { 1372 rw->w_tmemneeded = len; 1373 return ENOBUFS; 1374 } 1375 } 1376 } 1377 if (cp) { 1378 struct rt_xmsghdr *rtm = (struct rt_xmsghdr *)cp0; 1379 1380 rtm->rtm_version = RTM_XVERSION; 1381 rtm->rtm_type = type; 1382 rtm->rtm_msglen = len; 1383 } 1384 if (lenp) 1385 *lenp = len; 1386 return 0; 1387 } 1388 1389 /* 1390 * This routine is called to generate a message from the routing 1391 * socket indicating that a redirect has occurred, a routing lookup 1392 * has failed, or that a protocol has detected timeouts to a particular 1393 * destination. 1394 */ 1395 void 1396 COMPATNAME(rt_missmsg)(int type, const struct rt_addrinfo *rtinfo, int flags, 1397 int error) 1398 { 1399 struct rt_xmsghdr rtm; 1400 struct mbuf *m; 1401 const struct sockaddr *sa = rtinfo->rti_info[RTAX_DST]; 1402 struct rt_addrinfo info = *rtinfo; 1403 1404 COMPATCALL(rt_missmsg, (type, rtinfo, flags, error)); 1405 if (COMPATNAME(route_info).ri_cb.any_count == 0) 1406 return; 1407 memset(&rtm, 0, sizeof(rtm)); 1408 rtm.rtm_pid = curproc->p_pid; 1409 rtm.rtm_flags = RTF_DONE | flags; 1410 rtm.rtm_errno = error; 1411 m = COMPATNAME(rt_msg1)(type, &info, &rtm, sizeof(rtm)); 1412 if (m == NULL) 1413 return; 1414 mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs; 1415 COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0); 1416 } 1417 1418 /* 1419 * This routine is called to generate a message from the routing 1420 * socket indicating that the status of a network interface has changed. 1421 */ 1422 void 1423 COMPATNAME(rt_ifmsg)(struct ifnet *ifp) 1424 { 1425 struct if_xmsghdr ifm; 1426 struct mbuf *m; 1427 struct rt_addrinfo info; 1428 1429 COMPATCALL(rt_ifmsg, (ifp)); 1430 if (COMPATNAME(route_info).ri_cb.any_count == 0) 1431 return; 1432 (void)memset(&info, 0, sizeof(info)); 1433 (void)memset(&ifm, 0, sizeof(ifm)); 1434 ifm.ifm_index = ifp->if_index; 1435 ifm.ifm_flags = ifp->if_flags; 1436 if_export_if_data(ifp, &ifm.ifm_data, false); 1437 ifm.ifm_addrs = 0; 1438 m = COMPATNAME(rt_msg1)(RTM_IFINFO, &info, &ifm, sizeof(ifm)); 1439 if (m == NULL) 1440 return; 1441 COMPATNAME(route_enqueue)(m, 0); 1442 MODULE_HOOK_CALL_VOID(rtsock_oifmsg_14_hook, (ifp), __nothing); 1443 MODULE_HOOK_CALL_VOID(rtsock_oifmsg_50_hook, (ifp), __nothing); 1444 } 1445 1446 /* 1447 * This is called to generate messages from the routing socket 1448 * indicating a network interface has had addresses associated with it. 1449 * if we ever reverse the logic and replace messages TO the routing 1450 * socket indicate a request to configure interfaces, then it will 1451 * be unnecessary as the routing socket will automatically generate 1452 * copies of it. 1453 */ 1454 static void 1455 COMPATNAME(rt_addrmsg0)(int cmd, struct ifaddr *ifa, int error, 1456 struct rtentry *rt, const struct sockaddr *src) 1457 { 1458 #define cmdpass(__cmd, __pass) (((__cmd) << 2) | (__pass)) 1459 struct rt_addrinfo info; 1460 const struct sockaddr *sa; 1461 int pass; 1462 struct mbuf *m; 1463 struct ifnet *ifp; 1464 struct rt_xmsghdr rtm; 1465 struct ifa_xmsghdr ifam; 1466 int ncmd; 1467 1468 KASSERT(ifa != NULL); 1469 KASSERT(ifa->ifa_addr != NULL); 1470 ifp = ifa->ifa_ifp; 1471 if (cmd == RTM_ADD && vec_sctp_add_ip_address != NULL) { 1472 (*vec_sctp_add_ip_address)(ifa); 1473 } else if (cmd == RTM_DELETE && vec_sctp_delete_ip_address != NULL) { 1474 (*vec_sctp_delete_ip_address)(ifa); 1475 } 1476 1477 COMPATCALL(rt_addrmsg_rt, (cmd, ifa, error, rt)); 1478 if (COMPATNAME(route_info).ri_cb.any_count == 0) 1479 return; 1480 for (pass = 1; pass < 3; pass++) { 1481 memset(&info, 0, sizeof(info)); 1482 switch (cmdpass(cmd, pass)) { 1483 case cmdpass(RTM_ADD, 1): 1484 case cmdpass(RTM_CHANGE, 1): 1485 case cmdpass(RTM_DELETE, 2): 1486 case cmdpass(RTM_NEWADDR, 1): 1487 case cmdpass(RTM_DELADDR, 1): 1488 case cmdpass(RTM_CHGADDR, 1): 1489 switch (cmd) { 1490 case RTM_ADD: 1491 ncmd = RTM_XNEWADDR; 1492 break; 1493 case RTM_DELETE: 1494 ncmd = RTM_XDELADDR; 1495 break; 1496 case RTM_CHANGE: 1497 ncmd = RTM_XCHGADDR; 1498 break; 1499 case RTM_NEWADDR: 1500 ncmd = RTM_XNEWADDR; 1501 break; 1502 case RTM_DELADDR: 1503 ncmd = RTM_XDELADDR; 1504 break; 1505 case RTM_CHGADDR: 1506 ncmd = RTM_XCHGADDR; 1507 break; 1508 default: 1509 panic("%s: unknown command %d", __func__, cmd); 1510 } 1511 MODULE_HOOK_CALL_VOID(rtsock_newaddr_70_hook, 1512 (ncmd, ifa), __nothing); 1513 info.rti_info[RTAX_IFA] = sa = ifa->ifa_addr; 1514 KASSERT(ifp->if_dl != NULL); 1515 info.rti_info[RTAX_IFP] = ifp->if_dl->ifa_addr; 1516 info.rti_info[RTAX_NETMASK] = ifa->ifa_netmask; 1517 info.rti_info[RTAX_BRD] = ifa->ifa_dstaddr; 1518 info.rti_info[RTAX_AUTHOR] = src; 1519 memset(&ifam, 0, sizeof(ifam)); 1520 ifam.ifam_index = ifp->if_index; 1521 ifam.ifam_metric = ifa->ifa_metric; 1522 ifam.ifam_flags = ifa->ifa_flags; 1523 #ifndef COMPAT_RTSOCK 1524 ifam.ifam_pid = curproc->p_pid; 1525 ifam.ifam_addrflags = if_addrflags(ifa); 1526 #endif 1527 m = COMPATNAME(rt_msg1)(ncmd, &info, &ifam, sizeof(ifam)); 1528 if (m == NULL) 1529 continue; 1530 mtod(m, struct ifa_xmsghdr *)->ifam_addrs = 1531 info.rti_addrs; 1532 break; 1533 case cmdpass(RTM_ADD, 2): 1534 case cmdpass(RTM_CHANGE, 2): 1535 case cmdpass(RTM_DELETE, 1): 1536 if (rt == NULL) 1537 continue; 1538 info.rti_info[RTAX_NETMASK] = rt_mask(rt); 1539 info.rti_info[RTAX_DST] = sa = rt_getkey(rt); 1540 info.rti_info[RTAX_GATEWAY] = rt->rt_gateway; 1541 memset(&rtm, 0, sizeof(rtm)); 1542 rtm.rtm_pid = curproc->p_pid; 1543 rtm.rtm_index = ifp->if_index; 1544 rtm.rtm_flags |= rt->rt_flags; 1545 rtm.rtm_errno = error; 1546 m = COMPATNAME(rt_msg1)(cmd, &info, &rtm, sizeof(rtm)); 1547 if (m == NULL) 1548 continue; 1549 mtod(m, struct rt_xmsghdr *)->rtm_addrs = info.rti_addrs; 1550 break; 1551 default: 1552 continue; 1553 } 1554 KASSERTMSG(m != NULL, "called with wrong command"); 1555 COMPATNAME(route_enqueue)(m, sa ? sa->sa_family : 0); 1556 } 1557 #undef cmdpass 1558 } 1559 1560 void 1561 COMPATNAME(rt_addrmsg)(int cmd, struct ifaddr *ifa) 1562 { 1563 1564 COMPATNAME(rt_addrmsg0)(cmd, ifa, 0, NULL, NULL); 1565 } 1566 1567 void 1568 COMPATNAME(rt_addrmsg_rt)(int cmd, struct ifaddr *ifa, int error, 1569 struct rtentry *rt) 1570 { 1571 1572 COMPATNAME(rt_addrmsg0)(cmd, ifa, error, rt, NULL); 1573 } 1574 1575 void 1576 COMPATNAME(rt_addrmsg_src)(int cmd, struct ifaddr *ifa, 1577 const struct sockaddr *src) 1578 { 1579 1580 COMPATNAME(rt_addrmsg0)(cmd, ifa, 0, NULL, src); 1581 } 1582 1583 static struct mbuf * 1584 rt_makeifannouncemsg(struct ifnet *ifp, int type, int what, 1585 struct rt_addrinfo *info) 1586 { 1587 struct if_xannouncemsghdr ifan; 1588 1589 memset(info, 0, sizeof(*info)); 1590 memset(&ifan, 0, sizeof(ifan)); 1591 ifan.ifan_index = ifp->if_index; 1592 strlcpy(ifan.ifan_name, ifp->if_xname, sizeof(ifan.ifan_name)); 1593 ifan.ifan_what = what; 1594 return COMPATNAME(rt_msg1)(type, info, &ifan, sizeof(ifan)); 1595 } 1596 1597 /* 1598 * This is called to generate routing socket messages indicating 1599 * network interface arrival and departure. 1600 */ 1601 void 1602 COMPATNAME(rt_ifannouncemsg)(struct ifnet *ifp, int what) 1603 { 1604 struct mbuf *m; 1605 struct rt_addrinfo info; 1606 1607 COMPATCALL(rt_ifannouncemsg, (ifp, what)); 1608 if (COMPATNAME(route_info).ri_cb.any_count == 0) 1609 return; 1610 m = rt_makeifannouncemsg(ifp, RTM_IFANNOUNCE, what, &info); 1611 if (m == NULL) 1612 return; 1613 COMPATNAME(route_enqueue)(m, 0); 1614 } 1615 1616 /* 1617 * This is called to generate routing socket messages indicating 1618 * IEEE80211 wireless events. 1619 * XXX we piggyback on the RTM_IFANNOUNCE msg format in a clumsy way. 1620 */ 1621 void 1622 COMPATNAME(rt_ieee80211msg)(struct ifnet *ifp, int what, void *data, 1623 size_t data_len) 1624 { 1625 struct mbuf *m; 1626 struct rt_addrinfo info; 1627 1628 COMPATCALL(rt_ieee80211msg, (ifp, what, data, data_len)); 1629 if (COMPATNAME(route_info).ri_cb.any_count == 0) 1630 return; 1631 m = rt_makeifannouncemsg(ifp, RTM_IEEE80211, what, &info); 1632 if (m == NULL) 1633 return; 1634 /* 1635 * Append the ieee80211 data. Try to stick it in the 1636 * mbuf containing the ifannounce msg; otherwise allocate 1637 * a new mbuf and append. 1638 * 1639 * NB: we assume m is a single mbuf. 1640 */ 1641 if (data_len > M_TRAILINGSPACE(m)) { 1642 struct mbuf *n = m_get(M_NOWAIT, MT_DATA); 1643 if (n == NULL) { 1644 m_freem(m); 1645 return; 1646 } 1647 (void)memcpy(mtod(n, void *), data, data_len); 1648 n->m_len = data_len; 1649 m->m_next = n; 1650 } else if (data_len > 0) { 1651 (void)memcpy(mtod(m, uint8_t *) + m->m_len, data, data_len); 1652 m->m_len += data_len; 1653 } 1654 if (m->m_flags & M_PKTHDR) 1655 m->m_pkthdr.len += data_len; 1656 mtod(m, struct if_xannouncemsghdr *)->ifan_msglen += data_len; 1657 COMPATNAME(route_enqueue)(m, 0); 1658 } 1659 1660 /* 1661 * Routing message software interrupt routine 1662 */ 1663 static void 1664 COMPATNAME(route_intr)(void *cookie) 1665 { 1666 struct sockproto proto = { .sp_family = PF_XROUTE, }; 1667 struct route_info * const ri = &COMPATNAME(route_info); 1668 struct mbuf *m; 1669 1670 SOFTNET_KERNEL_LOCK_UNLESS_NET_MPSAFE(); 1671 for (;;) { 1672 IFQ_LOCK(&ri->ri_intrq); 1673 IF_DEQUEUE(&ri->ri_intrq, m); 1674 IFQ_UNLOCK(&ri->ri_intrq); 1675 if (m == NULL) 1676 break; 1677 proto.sp_protocol = M_GETCTX(m, uintptr_t); 1678 #ifdef NET_MPSAFE 1679 mutex_enter(rt_so_mtx); 1680 #endif 1681 raw_input(m, &proto, &ri->ri_src, &ri->ri_dst, &rt_rawcb); 1682 #ifdef NET_MPSAFE 1683 mutex_exit(rt_so_mtx); 1684 #endif 1685 } 1686 SOFTNET_KERNEL_UNLOCK_UNLESS_NET_MPSAFE(); 1687 } 1688 1689 /* 1690 * Enqueue a message to the software interrupt routine. 1691 */ 1692 void 1693 COMPATNAME(route_enqueue)(struct mbuf *m, int family) 1694 { 1695 struct route_info * const ri = &COMPATNAME(route_info); 1696 int wasempty; 1697 1698 IFQ_LOCK(&ri->ri_intrq); 1699 if (IF_QFULL(&ri->ri_intrq)) { 1700 printf("%s: queue full, dropped message\n", __func__); 1701 IF_DROP(&ri->ri_intrq); 1702 IFQ_UNLOCK(&ri->ri_intrq); 1703 m_freem(m); 1704 } else { 1705 wasempty = IF_IS_EMPTY(&ri->ri_intrq); 1706 M_SETCTX(m, (uintptr_t)family); 1707 IF_ENQUEUE(&ri->ri_intrq, m); 1708 IFQ_UNLOCK(&ri->ri_intrq); 1709 if (wasempty) { 1710 kpreempt_disable(); 1711 softint_schedule(ri->ri_sih); 1712 kpreempt_enable(); 1713 } 1714 } 1715 } 1716 1717 static void 1718 COMPATNAME(route_init)(void) 1719 { 1720 struct route_info * const ri = &COMPATNAME(route_info); 1721 1722 #ifndef COMPAT_RTSOCK 1723 rt_init(); 1724 #ifdef NET_MPSAFE 1725 rt_so_mtx = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE); 1726 1727 cv_init(&rt_update_cv, "rtsock_cv"); 1728 #endif 1729 1730 sysctl_net_route_setup(NULL, PF_ROUTE, "rtable"); 1731 #endif 1732 ri->ri_intrq.ifq_maxlen = ri->ri_maxqlen; 1733 ri->ri_sih = softint_establish(SOFTINT_NET | SOFTINT_MPSAFE, 1734 COMPATNAME(route_intr), NULL); 1735 IFQ_LOCK_INIT(&ri->ri_intrq); 1736 1737 #ifdef MBUFTRACE 1738 MOWNER_ATTACH(&COMPATNAME(routedomain).dom_mowner); 1739 #endif 1740 } 1741 1742 /* 1743 * Definitions of protocols supported in the ROUTE domain. 1744 */ 1745 #ifndef COMPAT_RTSOCK 1746 PR_WRAP_USRREQS(route); 1747 #else 1748 PR_WRAP_USRREQS(compat_50_route); 1749 #endif 1750 1751 static const struct pr_usrreqs route_usrreqs = { 1752 .pr_attach = COMPATNAME(route_attach_wrapper), 1753 .pr_detach = COMPATNAME(route_detach_wrapper), 1754 .pr_accept = COMPATNAME(route_accept_wrapper), 1755 .pr_bind = COMPATNAME(route_bind_wrapper), 1756 .pr_listen = COMPATNAME(route_listen_wrapper), 1757 .pr_connect = COMPATNAME(route_connect_wrapper), 1758 .pr_connect2 = COMPATNAME(route_connect2_wrapper), 1759 .pr_disconnect = COMPATNAME(route_disconnect_wrapper), 1760 .pr_shutdown = COMPATNAME(route_shutdown_wrapper), 1761 .pr_abort = COMPATNAME(route_abort_wrapper), 1762 .pr_ioctl = COMPATNAME(route_ioctl_wrapper), 1763 .pr_stat = COMPATNAME(route_stat_wrapper), 1764 .pr_peeraddr = COMPATNAME(route_peeraddr_wrapper), 1765 .pr_sockaddr = COMPATNAME(route_sockaddr_wrapper), 1766 .pr_rcvd = COMPATNAME(route_rcvd_wrapper), 1767 .pr_recvoob = COMPATNAME(route_recvoob_wrapper), 1768 .pr_send = COMPATNAME(route_send_wrapper), 1769 .pr_sendoob = COMPATNAME(route_sendoob_wrapper), 1770 .pr_purgeif = COMPATNAME(route_purgeif_wrapper), 1771 }; 1772 1773 static const struct protosw COMPATNAME(route_protosw)[] = { 1774 { 1775 .pr_type = SOCK_RAW, 1776 .pr_domain = &COMPATNAME(routedomain), 1777 .pr_flags = PR_ATOMIC|PR_ADDR, 1778 .pr_ctlinput = raw_ctlinput, 1779 .pr_ctloutput = route_ctloutput, 1780 .pr_usrreqs = &route_usrreqs, 1781 .pr_init = rt_pr_init, 1782 }, 1783 }; 1784 1785 struct domain COMPATNAME(routedomain) = { 1786 .dom_family = PF_XROUTE, 1787 .dom_name = DOMAINNAME, 1788 .dom_init = COMPATNAME(route_init), 1789 .dom_protosw = COMPATNAME(route_protosw), 1790 .dom_protoswNPROTOSW = 1791 &COMPATNAME(route_protosw)[__arraycount(COMPATNAME(route_protosw))], 1792 #ifdef MBUFTRACE 1793 .dom_mowner = MOWNER_INIT("route", "rtm"), 1794 #endif 1795 }; 1796