1 /* 2 * Copyright (c) 1988, 1991 Regents of the University of California. 3 * All rights reserved. 4 * 5 * %sccs.include.redist.c% 6 * 7 * @(#)rtsock.c 7.18 (Berkeley) 06/27/91 8 */ 9 10 #include "param.h" 11 #include "mbuf.h" 12 #include "proc.h" 13 #include "socket.h" 14 #include "socketvar.h" 15 #include "domain.h" 16 #include "protosw.h" 17 18 #include "af.h" 19 #include "if.h" 20 #include "route.h" 21 #include "raw_cb.h" 22 23 #include "machine/mtpr.h" 24 25 struct sockaddr route_dst = { 2, PF_ROUTE, }; 26 struct sockaddr route_src = { 2, PF_ROUTE, }; 27 struct sockproto route_proto = { PF_ROUTE, }; 28 29 /*ARGSUSED*/ 30 route_usrreq(so, req, m, nam, control) 31 register struct socket *so; 32 int req; 33 struct mbuf *m, *nam, *control; 34 { 35 register int error = 0; 36 register struct rawcb *rp = sotorawcb(so); 37 int s; 38 if (req == PRU_ATTACH) { 39 MALLOC(rp, struct rawcb *, sizeof(*rp), M_PCB, M_WAITOK); 40 if (so->so_pcb = (caddr_t)rp) 41 bzero(so->so_pcb, sizeof(*rp)); 42 43 } 44 if (req == PRU_DETACH && rp) { 45 int af = rp->rcb_proto.sp_protocol; 46 if (af == AF_INET) 47 route_cb.ip_count--; 48 else if (af == AF_NS) 49 route_cb.ns_count--; 50 else if (af == AF_ISO) 51 route_cb.iso_count--; 52 route_cb.any_count--; 53 } 54 s = splnet(); 55 error = raw_usrreq(so, req, m, nam, control); 56 rp = sotorawcb(so); 57 if (req == PRU_ATTACH && rp) { 58 int af = rp->rcb_proto.sp_protocol; 59 if (error) { 60 free((caddr_t)rp, M_PCB); 61 splx(s); 62 return (error); 63 } 64 if (af == AF_INET) 65 route_cb.ip_count++; 66 else if (af == AF_NS) 67 route_cb.ns_count++; 68 else if (af == AF_ISO) 69 route_cb.iso_count++; 70 rp->rcb_faddr = &route_src; 71 route_cb.any_count++; 72 soisconnected(so); 73 so->so_options |= SO_USELOOPBACK; 74 } 75 splx(s); 76 return (error); 77 } 78 #define ROUNDUP(a) \ 79 ((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long)) 80 #define ADVANCE(x, n) (x += ROUNDUP((n)->sa_len)) 81 82 /*ARGSUSED*/ 83 route_output(m, so) 84 register struct mbuf *m; 85 struct socket *so; 86 { 87 register struct rt_msghdr *rtm = 0; 88 register struct rtentry *rt = 0; 89 struct rtentry *saved_nrt = 0; 90 struct sockaddr *dst = 0, *gate = 0, *netmask = 0, *genmask = 0; 91 struct sockaddr *ifpaddr = 0, *ifaaddr = 0; 92 caddr_t cp, lim; 93 int len, error = 0; 94 struct ifnet *ifp = 0; 95 struct ifaddr *ifa = 0; 96 struct ifaddr *ifaof_ifpforaddr(), *ifa_ifwithroute(); 97 98 #define senderr(e) { error = e; goto flush;} 99 if (m == 0 || m->m_len < sizeof(long)) 100 return (ENOBUFS); 101 if ((m = m_pullup(m, sizeof(long))) == 0) 102 return (ENOBUFS); 103 if ((m->m_flags & M_PKTHDR) == 0) 104 panic("route_output"); 105 len = m->m_pkthdr.len; 106 if (len < sizeof(*rtm) || 107 len != mtod(m, struct rt_msghdr *)->rtm_msglen) 108 senderr(EINVAL); 109 R_Malloc(rtm, struct rt_msghdr *, len); 110 if (rtm == 0) 111 senderr(ENOBUFS); 112 m_copydata(m, 0, len, (caddr_t)rtm); 113 if (rtm->rtm_version != RTM_VERSION) 114 senderr(EPROTONOSUPPORT); 115 rtm->rtm_pid = curproc->p_pid; 116 lim = len + (caddr_t) rtm; 117 cp = (caddr_t) (rtm + 1); 118 if (rtm->rtm_addrs & RTA_DST) { 119 dst = (struct sockaddr *)cp; 120 ADVANCE(cp, dst); 121 } else 122 senderr(EINVAL); 123 if ((rtm->rtm_addrs & RTA_GATEWAY) && cp < lim) { 124 gate = (struct sockaddr *)cp; 125 ADVANCE(cp, gate); 126 } 127 if ((rtm->rtm_addrs & RTA_NETMASK) && cp < lim) { 128 netmask = (struct sockaddr *)cp; 129 ADVANCE(cp, netmask); 130 } 131 if ((rtm->rtm_addrs & RTA_GENMASK) && cp < lim) { 132 struct radix_node *t, *rn_addmask(); 133 genmask = (struct sockaddr *)cp; 134 ADVANCE(cp, genmask); 135 t = rn_addmask(genmask, 1, 2); 136 if (t && Bcmp(genmask, t->rn_key, *(u_char *)genmask) == 0) 137 genmask = (struct sockaddr *)(t->rn_key); 138 else 139 senderr(ENOBUFS); 140 } 141 if ((rtm->rtm_addrs & RTA_IFP) && cp < lim) { 142 ifpaddr = (struct sockaddr *)cp; 143 ADVANCE(cp, ifpaddr); 144 } 145 if ((rtm->rtm_addrs & RTA_IFA) && cp < lim) { 146 ifaaddr = (struct sockaddr *)cp; 147 } 148 switch (rtm->rtm_type) { 149 case RTM_ADD: 150 if (gate == 0) 151 senderr(EINVAL); 152 error = rtrequest(RTM_ADD, dst, gate, netmask, 153 rtm->rtm_flags, &saved_nrt); 154 if (error == 0 && saved_nrt) { 155 rt_setmetrics(rtm->rtm_inits, 156 &rtm->rtm_rmx, &saved_nrt->rt_rmx); 157 saved_nrt->rt_refcnt--; 158 saved_nrt->rt_genmask = genmask; 159 } 160 break; 161 162 case RTM_DELETE: 163 error = rtrequest(RTM_DELETE, dst, gate, netmask, 164 rtm->rtm_flags, (struct rtentry **)0); 165 break; 166 167 case RTM_GET: 168 case RTM_CHANGE: 169 case RTM_LOCK: 170 rt = rtalloc1(dst, 0); 171 if (rt == 0) 172 senderr(ESRCH); 173 if (rtm->rtm_type != RTM_GET) { 174 if (Bcmp(dst, rt_key(rt), dst->sa_len) != 0) 175 senderr(ESRCH); 176 if (rt->rt_nodes->rn_dupedkey && 177 (netmask == 0 || 178 Bcmp(netmask, rt_mask(rt), netmask->sa_len))) 179 senderr(ETOOMANYREFS); 180 } 181 switch(rtm->rtm_type) { 182 183 case RTM_GET: 184 dst = rt_key(rt); len = sizeof(*rtm); 185 ADVANCE(len, dst); 186 rtm->rtm_addrs |= RTA_DST; 187 if (gate = rt->rt_gateway) { 188 ADVANCE(len, gate); 189 rtm->rtm_addrs |= RTA_GATEWAY; 190 } else 191 rtm->rtm_addrs &= ~RTA_GATEWAY; 192 if (netmask = rt_mask(rt)) { 193 ADVANCE(len, netmask); 194 rtm->rtm_addrs |= RTA_NETMASK; 195 } else 196 rtm->rtm_addrs &= ~RTA_NETMASK; 197 if (genmask = rt->rt_genmask) { 198 ADVANCE(len, genmask); 199 rtm->rtm_addrs |= RTA_GENMASK; 200 } else 201 rtm->rtm_addrs &= ~RTA_GENMASK; 202 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) { 203 if (rt->rt_ifp == 0) 204 goto badif; 205 for (ifa = rt->rt_ifp->if_addrlist; 206 ifa && ifa->ifa_addr->sa_family != AF_LINK; 207 ifa = ifa->ifa_next){} 208 if (ifa && rt->rt_ifa) { 209 ifpaddr = ifa->ifa_addr; 210 ADVANCE(len, ifpaddr); 211 ifaaddr = rt->rt_ifa->ifa_addr; 212 ADVANCE(len, ifaaddr); 213 rtm->rtm_addrs |= RTA_IFP | RTA_IFA; 214 } else { 215 badif: ifpaddr = 0; 216 rtm->rtm_addrs &= ~(RTA_IFP | RTA_IFA); 217 } 218 } 219 if (len > rtm->rtm_msglen) { 220 struct rt_msghdr *new_rtm; 221 R_Malloc(new_rtm, struct rt_msghdr *, len); 222 if (new_rtm == 0) 223 senderr(ENOBUFS); 224 Bcopy(rtm, new_rtm, rtm->rtm_msglen); 225 Free(rtm); rtm = new_rtm; 226 } 227 rtm->rtm_msglen = len; 228 rtm->rtm_flags = rt->rt_flags; 229 rtm->rtm_rmx = rt->rt_rmx; 230 cp = (caddr_t) (1 + rtm); 231 len = ROUNDUP(dst->sa_len); 232 Bcopy(dst, cp, len); cp += len; 233 if (gate) { 234 len = ROUNDUP(gate->sa_len); 235 Bcopy(gate, cp, len); cp += len; 236 } 237 if (netmask) { 238 len = ROUNDUP(netmask->sa_len); 239 Bcopy(netmask, cp, len); cp += len; 240 } 241 if (genmask) { 242 len = ROUNDUP(genmask->sa_len); 243 Bcopy(genmask, cp, len); cp += len; 244 } 245 if (ifpaddr) { 246 len = ROUNDUP(ifpaddr->sa_len); 247 Bcopy(ifpaddr, cp, len); cp += len; 248 len = ROUNDUP(ifaaddr->sa_len); 249 Bcopy(ifaaddr, cp, len); cp += len; 250 } 251 break; 252 253 case RTM_CHANGE: 254 if (gate && 255 (gate->sa_len > (len = rt->rt_gateway->sa_len))) 256 senderr(EDQUOT); 257 /* new gateway could require new ifaddr, ifp; 258 flags may also be different; ifp may be specified 259 by ll sockaddr when protocol address is ambiguous */ 260 if (ifpaddr && (ifa = ifa_ifwithnet(ifpaddr)) && 261 (ifp = ifa->ifa_ifp)) 262 ifa = ifaof_ifpforaddr(ifaaddr ? ifaaddr : gate, 263 ifp); 264 else if ((ifaaddr && (ifa = ifa_ifwithaddr(ifaaddr))) || 265 (ifa = ifa_ifwithroute(rt->rt_flags, 266 rt_key(rt), gate))) 267 ifp = ifa->ifa_ifp; 268 if (ifa) { 269 register struct ifaddr *oifa = rt->rt_ifa; 270 if (oifa != ifa) { 271 if (oifa && oifa->ifa_rtrequest) 272 oifa->ifa_rtrequest(RTM_DELETE, 273 rt, gate); 274 rt->rt_ifa = ifa; 275 rt->rt_ifp = ifp; 276 } 277 } 278 if (gate) 279 Bcopy(gate, rt->rt_gateway, len); 280 rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx, 281 &rt->rt_rmx); 282 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest) 283 rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, gate); 284 if (genmask) 285 rt->rt_genmask = genmask; 286 /* 287 * Fall into 288 */ 289 case RTM_LOCK: 290 rt->rt_rmx.rmx_locks |= 291 (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks); 292 rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits); 293 break; 294 } 295 goto cleanup; 296 297 default: 298 senderr(EOPNOTSUPP); 299 } 300 301 flush: 302 if (rtm) { 303 if (error) 304 rtm->rtm_errno = error; 305 else 306 rtm->rtm_flags |= RTF_DONE; 307 } 308 cleanup: 309 if (rt) 310 rtfree(rt); 311 { 312 register struct rawcb *rp = 0; 313 /* 314 * Check to see if we don't want our own messages. 315 */ 316 if ((so->so_options & SO_USELOOPBACK) == 0) { 317 if (route_cb.any_count <= 1) { 318 if (rtm) 319 Free(rtm); 320 m_freem(m); 321 return (error); 322 } 323 /* There is another listener, so construct message */ 324 rp = sotorawcb(so); 325 } 326 if (rtm) { 327 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm); 328 Free(rtm); 329 } 330 if (rp) 331 rp->rcb_proto.sp_family = 0; /* Avoid us */ 332 if (dst) 333 route_proto.sp_protocol = dst->sa_family; 334 raw_input(m, &route_proto, &route_src, &route_dst); 335 if (rp) 336 rp->rcb_proto.sp_family = PF_ROUTE; 337 } 338 return (error); 339 } 340 341 rt_setmetrics(which, in, out) 342 u_long which; 343 register struct rt_metrics *in, *out; 344 { 345 #define metric(f, e) if (which & (f)) out->e = in->e; 346 metric(RTV_RPIPE, rmx_recvpipe); 347 metric(RTV_SPIPE, rmx_sendpipe); 348 metric(RTV_SSTHRESH, rmx_ssthresh); 349 metric(RTV_RTT, rmx_rtt); 350 metric(RTV_RTTVAR, rmx_rttvar); 351 metric(RTV_HOPCOUNT, rmx_hopcount); 352 metric(RTV_MTU, rmx_mtu); 353 metric(RTV_EXPIRE, rmx_expire); 354 #undef metric 355 } 356 357 /* 358 * Copy data from a buffer back into the indicated mbuf chain, 359 * starting "off" bytes from the beginning, extending the mbuf 360 * chain if necessary. 361 */ 362 m_copyback(m0, off, len, cp) 363 struct mbuf *m0; 364 register int off; 365 register int len; 366 caddr_t cp; 367 368 { 369 register int mlen; 370 register struct mbuf *m = m0, *n; 371 int totlen = 0; 372 373 if (m0 == 0) 374 return; 375 while (off > (mlen = m->m_len)) { 376 off -= mlen; 377 totlen += mlen; 378 if (m->m_next == 0) { 379 n = m_getclr(M_DONTWAIT, m->m_type); 380 if (n == 0) 381 goto out; 382 n->m_len = min(MLEN, len + off); 383 m->m_next = n; 384 } 385 m = m->m_next; 386 } 387 while (len > 0) { 388 mlen = min (m->m_len - off, len); 389 bcopy(cp, off + mtod(m, caddr_t), (unsigned)mlen); 390 cp += mlen; 391 len -= mlen; 392 mlen += off; 393 off = 0; 394 totlen += mlen; 395 if (len == 0) 396 break; 397 if (m->m_next == 0) { 398 n = m_get(M_DONTWAIT, m->m_type); 399 if (n == 0) 400 break; 401 n->m_len = min(MLEN, len); 402 m->m_next = n; 403 } 404 m = m->m_next; 405 } 406 out: if (((m = m0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen)) 407 m->m_pkthdr.len = totlen; 408 } 409 410 /* 411 * The miss message and losing message are very similar. 412 */ 413 414 rt_missmsg(type, dst, gate, mask, src, flags, error) 415 register struct sockaddr *dst; 416 struct sockaddr *gate, *mask, *src; 417 { 418 register struct rt_msghdr *rtm; 419 register struct mbuf *m; 420 int dlen = ROUNDUP(dst->sa_len); 421 int len = dlen + sizeof(*rtm); 422 423 if (route_cb.any_count == 0) 424 return; 425 m = m_gethdr(M_DONTWAIT, MT_DATA); 426 if (m == 0) 427 return; 428 m->m_pkthdr.len = m->m_len = min(len, MHLEN); 429 m->m_pkthdr.rcvif = 0; 430 rtm = mtod(m, struct rt_msghdr *); 431 bzero((caddr_t)rtm, sizeof(*rtm)); /*XXX assumes sizeof(*rtm) < MHLEN*/ 432 rtm->rtm_flags = RTF_DONE | flags; 433 rtm->rtm_msglen = len; 434 rtm->rtm_version = RTM_VERSION; 435 rtm->rtm_type = type; 436 rtm->rtm_addrs = RTA_DST; 437 if (type == RTM_OLDADD || type == RTM_OLDDEL) { 438 rtm->rtm_pid = curproc->p_pid; 439 } 440 m_copyback(m, sizeof (*rtm), dlen, (caddr_t)dst); 441 if (gate) { 442 dlen = ROUNDUP(gate->sa_len); 443 m_copyback(m, len , dlen, (caddr_t)gate); 444 len += dlen; 445 rtm->rtm_addrs |= RTA_GATEWAY; 446 } 447 if (mask) { 448 dlen = ROUNDUP(mask->sa_len); 449 m_copyback(m, len , dlen, (caddr_t)mask); 450 len += dlen; 451 rtm->rtm_addrs |= RTA_NETMASK; 452 } 453 if (src) { 454 dlen = ROUNDUP(src->sa_len); 455 m_copyback(m, len , dlen, (caddr_t)src); 456 len += dlen; 457 rtm->rtm_addrs |= RTA_AUTHOR; 458 } 459 if (m->m_pkthdr.len != len) { 460 m_freem(m); 461 return; 462 } 463 rtm->rtm_errno = error; 464 rtm->rtm_msglen = len; 465 route_proto.sp_protocol = dst->sa_family; 466 raw_input(m, &route_proto, &route_src, &route_dst); 467 } 468 469 #include "kinfo.h" 470 struct walkarg { 471 int w_op, w_arg; 472 int w_given, w_needed; 473 caddr_t w_where; 474 struct { 475 struct rt_msghdr m_rtm; 476 char m_sabuf[128]; 477 } w_m; 478 #define w_rtm w_m.m_rtm 479 }; 480 /* 481 * This is used in dumping the kernel table via getkinfo(). 482 */ 483 rt_dumpentry(rn, w) 484 struct radix_node *rn; 485 register struct walkarg *w; 486 { 487 register struct sockaddr *sa; 488 int n, error; 489 490 for (; rn; rn = rn->rn_dupedkey) { 491 int count = 0, size = sizeof(w->w_rtm); 492 register struct rtentry *rt = (struct rtentry *)rn; 493 494 if (rn->rn_flags & RNF_ROOT) 495 continue; 496 if (w->w_op == KINFO_RT_FLAGS && !(rt->rt_flags & w->w_arg)) 497 continue; 498 #define next(a, l) {size += (l); w->w_rtm.rtm_addrs |= (a); } 499 w->w_rtm.rtm_addrs = 0; 500 if (sa = rt_key(rt)) 501 next(RTA_DST, ROUNDUP(sa->sa_len)); 502 if (sa = rt->rt_gateway) 503 next(RTA_GATEWAY, ROUNDUP(sa->sa_len)); 504 if (sa = rt_mask(rt)) 505 next(RTA_NETMASK, ROUNDUP(sa->sa_len)); 506 if (sa = rt->rt_genmask) 507 next(RTA_GENMASK, ROUNDUP(sa->sa_len)); 508 w->w_needed += size; 509 if (w->w_where == NULL || w->w_needed > 0) 510 continue; 511 w->w_rtm.rtm_msglen = size; 512 w->w_rtm.rtm_flags = rt->rt_flags; 513 w->w_rtm.rtm_use = rt->rt_use; 514 w->w_rtm.rtm_rmx = rt->rt_rmx; 515 w->w_rtm.rtm_index = rt->rt_ifp->if_index; 516 #undef next 517 #define next(l) {n = (l); Bcopy(sa, cp, n); cp += n;} 518 if (size <= sizeof(w->w_m)) { 519 register caddr_t cp = (caddr_t)(w->w_m.m_sabuf); 520 if (sa = rt_key(rt)) 521 next(ROUNDUP(sa->sa_len)); 522 if (sa = rt->rt_gateway) 523 next(ROUNDUP(sa->sa_len)); 524 if (sa = rt_mask(rt)) 525 next(ROUNDUP(sa->sa_len)); 526 if (sa = rt->rt_genmask) 527 next(ROUNDUP(sa->sa_len)); 528 #undef next 529 #define next(s, l) {n = (l); \ 530 if (error = copyout((caddr_t)(s), w->w_where, n)) return (error); \ 531 w->w_where += n;} 532 533 next(&w->w_m, size); /* Copy rtmsg and sockaddrs back */ 534 continue; 535 } 536 next(&w->w_rtm, sizeof(w->w_rtm)); 537 if (sa = rt_key(rt)) 538 next(sa, ROUNDUP(sa->sa_len)); 539 if (sa = rt->rt_gateway) 540 next(sa, ROUNDUP(sa->sa_len)); 541 if (sa = rt_mask(rt)) 542 next(sa, ROUNDUP(sa->sa_len)); 543 if (sa = rt->rt_genmask) 544 next(sa, ROUNDUP(sa->sa_len)); 545 } 546 return (0); 547 #undef next 548 } 549 550 kinfo_rtable(op, where, given, arg, needed) 551 int op, arg; 552 caddr_t where; 553 int *given, *needed; 554 { 555 register struct radix_node_head *rnh; 556 int s, error = 0; 557 u_char af = ki_af(op); 558 struct walkarg w; 559 560 op &= 0xffff; 561 if (op != KINFO_RT_DUMP && op != KINFO_RT_FLAGS) 562 return (EINVAL); 563 564 Bzero(&w, sizeof(w)); 565 if ((w.w_where = where) && given) 566 w.w_given = *given; 567 w.w_needed = 0 - w.w_given; 568 w.w_arg = arg; 569 w.w_op = op; 570 w.w_rtm.rtm_version = RTM_VERSION; 571 w.w_rtm.rtm_type = RTM_GET; 572 573 s = splnet(); 574 for (rnh = radix_node_head; rnh; rnh = rnh->rnh_next) { 575 if (rnh->rnh_af == 0) 576 continue; 577 if (af && af != rnh->rnh_af) 578 continue; 579 error = rt_walk(rnh->rnh_treetop, rt_dumpentry, &w); 580 if (error) 581 break; 582 } 583 w.w_needed += w.w_given; 584 if (where && given) 585 *given = w.w_where - where; 586 else 587 w.w_needed = (11 * w.w_needed) / 10; 588 *needed = w.w_needed; 589 splx(s); 590 return (error); 591 } 592 593 rt_walk(rn, f, w) 594 register struct radix_node *rn; 595 register int (*f)(); 596 struct walkarg *w; 597 { 598 int error; 599 for (;;) { 600 while (rn->rn_b >= 0) 601 rn = rn->rn_l; /* First time through node, go left */ 602 if (error = (*f)(rn, w)) 603 return (error); /* Process Leaf */ 604 while (rn->rn_p->rn_r == rn) { /* if coming back from right */ 605 rn = rn->rn_p; /* go back up */ 606 if (rn->rn_flags & RNF_ROOT) 607 return 0; 608 } 609 rn = rn->rn_p->rn_r; /* otherwise, go right*/ 610 } 611 } 612 613 /* 614 * Definitions of protocols supported in the ROUTE domain. 615 */ 616 617 int raw_init(),raw_usrreq(),raw_input(),raw_ctlinput(); 618 extern struct domain routedomain; /* or at least forward */ 619 620 struct protosw routesw[] = { 621 { SOCK_RAW, &routedomain, 0, PR_ATOMIC|PR_ADDR, 622 raw_input, route_output, raw_ctlinput, 0, 623 route_usrreq, 624 raw_init, 0, 0, 0, 625 } 626 }; 627 628 int unp_externalize(), unp_dispose(); 629 630 struct domain routedomain = 631 { PF_ROUTE, "route", 0, 0, 0, 632 routesw, &routesw[sizeof(routesw)/sizeof(routesw[0])] }; 633