1 /* 2 * Copyright (c) 1988 Regents of the University of California. 3 * All rights reserved. 4 * 5 * %sccs.include.redist.c% 6 * 7 * @(#)rtsock.c 7.16 (Berkeley) 03/16/91 8 */ 9 10 #include "param.h" 11 #include "mbuf.h" 12 #include "user.h" 13 #include "proc.h" 14 #include "socket.h" 15 #include "socketvar.h" 16 #include "domain.h" 17 #include "protosw.h" 18 #include "errno.h" 19 20 #include "af.h" 21 #include "if.h" 22 #include "route.h" 23 #include "raw_cb.h" 24 25 #include "machine/mtpr.h" 26 27 struct sockaddr route_dst = { 2, PF_ROUTE, }; 28 struct sockaddr route_src = { 2, PF_ROUTE, }; 29 struct sockproto route_proto = { PF_ROUTE, }; 30 31 /*ARGSUSED*/ 32 route_usrreq(so, req, m, nam, control) 33 register struct socket *so; 34 int req; 35 struct mbuf *m, *nam, *control; 36 { 37 register int error = 0; 38 register struct rawcb *rp = sotorawcb(so); 39 int s; 40 if (req == PRU_ATTACH) { 41 MALLOC(rp, struct rawcb *, sizeof(*rp), M_PCB, M_WAITOK); 42 if (so->so_pcb = (caddr_t)rp) 43 bzero(so->so_pcb, sizeof(*rp)); 44 45 } 46 if (req == PRU_DETACH && rp) { 47 int af = rp->rcb_proto.sp_protocol; 48 if (af == AF_INET) 49 route_cb.ip_count--; 50 else if (af == AF_NS) 51 route_cb.ns_count--; 52 else if (af == AF_ISO) 53 route_cb.iso_count--; 54 route_cb.any_count--; 55 } 56 s = splnet(); 57 error = raw_usrreq(so, req, m, nam, control); 58 rp = sotorawcb(so); 59 if (req == PRU_ATTACH && rp) { 60 int af = rp->rcb_proto.sp_protocol; 61 if (error) { 62 free((caddr_t)rp, M_PCB); 63 splx(s); 64 return (error); 65 } 66 if (af == AF_INET) 67 route_cb.ip_count++; 68 else if (af == AF_NS) 69 route_cb.ns_count++; 70 else if (af == AF_ISO) 71 route_cb.iso_count++; 72 rp->rcb_faddr = &route_src; 73 route_cb.any_count++; 74 soisconnected(so); 75 so->so_options |= SO_USELOOPBACK; 76 } 77 splx(s); 78 return (error); 79 } 80 #define ROUNDUP(a) \ 81 ((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long)) 82 #define ADVANCE(x, n) (x += ROUNDUP((n)->sa_len)) 83 84 /*ARGSUSED*/ 85 route_output(m, so) 86 register struct mbuf *m; 87 struct socket *so; 88 { 89 register struct rt_msghdr *rtm = 0; 90 register struct rtentry *rt = 0; 91 struct rtentry *saved_nrt = 0; 92 struct sockaddr *dst = 0, *gate = 0, *netmask = 0, *genmask = 0; 93 struct sockaddr *ifpaddr = 0, *ifaaddr = 0; 94 caddr_t cp, lim; 95 int len, error = 0; 96 struct ifnet *ifp = 0; 97 struct ifaddr *ifa = 0; 98 struct ifaddr *ifaof_ifpforaddr(), *ifa_ifwithroute(); 99 100 #define senderr(e) { error = e; goto flush;} 101 if (m == 0 || m->m_len < sizeof(long)) 102 return (ENOBUFS); 103 if ((m = m_pullup(m, sizeof(long))) == 0) 104 return (ENOBUFS); 105 if ((m->m_flags & M_PKTHDR) == 0) 106 panic("route_output"); 107 len = m->m_pkthdr.len; 108 if (len < sizeof(*rtm) || 109 len != mtod(m, struct rt_msghdr *)->rtm_msglen) 110 senderr(EINVAL); 111 R_Malloc(rtm, struct rt_msghdr *, len); 112 if (rtm == 0) 113 senderr(ENOBUFS); 114 m_copydata(m, 0, len, (caddr_t)rtm); 115 if (rtm->rtm_version != RTM_VERSION) 116 senderr(EPROTONOSUPPORT); 117 rtm->rtm_pid = u.u_procp->p_pid; 118 lim = len + (caddr_t) rtm; 119 cp = (caddr_t) (rtm + 1); 120 if (rtm->rtm_addrs & RTA_DST) { 121 dst = (struct sockaddr *)cp; 122 ADVANCE(cp, dst); 123 } else 124 senderr(EINVAL); 125 if ((rtm->rtm_addrs & RTA_GATEWAY) && cp < lim) { 126 gate = (struct sockaddr *)cp; 127 ADVANCE(cp, gate); 128 } 129 if ((rtm->rtm_addrs & RTA_NETMASK) && cp < lim) { 130 netmask = (struct sockaddr *)cp; 131 ADVANCE(cp, netmask); 132 } 133 if ((rtm->rtm_addrs & RTA_GENMASK) && cp < lim) { 134 struct radix_node *t, *rn_addmask(); 135 genmask = (struct sockaddr *)cp; 136 ADVANCE(cp, genmask); 137 t = rn_addmask(genmask, 1, 2); 138 if (t && Bcmp(genmask, t->rn_key, *(u_char *)genmask) == 0) 139 genmask = (struct sockaddr *)(t->rn_key); 140 else 141 senderr(ENOBUFS); 142 } 143 if ((rtm->rtm_addrs & RTA_IFP) && cp < lim) { 144 ifpaddr = (struct sockaddr *)cp; 145 ADVANCE(cp, ifpaddr); 146 } 147 if ((rtm->rtm_addrs & RTA_IFA) && cp < lim) { 148 ifaaddr = (struct sockaddr *)cp; 149 } 150 switch (rtm->rtm_type) { 151 case RTM_ADD: 152 if (gate == 0) 153 senderr(EINVAL); 154 error = rtrequest(RTM_ADD, dst, gate, netmask, 155 rtm->rtm_flags, &saved_nrt); 156 if (error == 0 && saved_nrt) { 157 rt_setmetrics(rtm->rtm_inits, 158 &rtm->rtm_rmx, &saved_nrt->rt_rmx); 159 saved_nrt->rt_refcnt--; 160 saved_nrt->rt_genmask = genmask; 161 } 162 break; 163 164 case RTM_DELETE: 165 error = rtrequest(RTM_DELETE, dst, gate, netmask, 166 rtm->rtm_flags, (struct rtentry **)0); 167 break; 168 169 case RTM_GET: 170 case RTM_CHANGE: 171 case RTM_LOCK: 172 rt = rtalloc1(dst, 0); 173 if (rt == 0) 174 senderr(ESRCH); 175 if (rtm->rtm_type != RTM_GET) { 176 if (Bcmp(dst, rt_key(rt), dst->sa_len) != 0) 177 senderr(ESRCH); 178 if (rt->rt_nodes->rn_dupedkey && 179 (netmask == 0 || 180 Bcmp(netmask, rt_mask(rt), netmask->sa_len))) 181 senderr(ETOOMANYREFS); 182 } 183 switch(rtm->rtm_type) { 184 185 case RTM_GET: 186 dst = rt_key(rt); len = sizeof(*rtm); 187 ADVANCE(len, dst); 188 rtm->rtm_addrs |= RTA_DST; 189 if (gate = rt->rt_gateway) { 190 ADVANCE(len, gate); 191 rtm->rtm_addrs |= RTA_GATEWAY; 192 } else 193 rtm->rtm_addrs &= ~RTA_GATEWAY; 194 if (netmask = rt_mask(rt)) { 195 ADVANCE(len, netmask); 196 rtm->rtm_addrs |= RTA_NETMASK; 197 } else 198 rtm->rtm_addrs &= ~RTA_NETMASK; 199 if (genmask = rt->rt_genmask) { 200 ADVANCE(len, genmask); 201 rtm->rtm_addrs |= RTA_GENMASK; 202 } else 203 rtm->rtm_addrs &= ~RTA_GENMASK; 204 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) { 205 if (rt->rt_ifp == 0) 206 goto badif; 207 for (ifa = rt->rt_ifp->if_addrlist; 208 ifa && ifa->ifa_addr->sa_family != AF_LINK; 209 ifa = ifa->ifa_next){} 210 if (ifa && rt->rt_ifa) { 211 ifpaddr = ifa->ifa_addr; 212 ADVANCE(len, ifpaddr); 213 ifaaddr = rt->rt_ifa->ifa_addr; 214 ADVANCE(len, ifaaddr); 215 rtm->rtm_addrs |= RTA_IFP | RTA_IFA; 216 } else { 217 badif: ifpaddr = 0; 218 rtm->rtm_addrs &= ~(RTA_IFP | RTA_IFA); 219 } 220 } 221 if (len > rtm->rtm_msglen) { 222 struct rt_msghdr *new_rtm; 223 R_Malloc(new_rtm, struct rt_msghdr *, len); 224 if (new_rtm == 0) 225 senderr(ENOBUFS); 226 Bcopy(rtm, new_rtm, rtm->rtm_msglen); 227 Free(rtm); rtm = new_rtm; 228 } 229 rtm->rtm_msglen = len; 230 rtm->rtm_flags = rt->rt_flags; 231 rtm->rtm_rmx = rt->rt_rmx; 232 cp = (caddr_t) (1 + rtm); 233 len = ROUNDUP(dst->sa_len); 234 Bcopy(dst, cp, len); cp += len; 235 if (gate) { 236 len = ROUNDUP(gate->sa_len); 237 Bcopy(gate, cp, len); cp += len; 238 } 239 if (netmask) { 240 len = ROUNDUP(netmask->sa_len); 241 Bcopy(netmask, cp, len); cp += len; 242 } 243 if (genmask) { 244 len = ROUNDUP(genmask->sa_len); 245 Bcopy(genmask, cp, len); cp += len; 246 } 247 if (ifpaddr) { 248 len = ROUNDUP(ifpaddr->sa_len); 249 Bcopy(ifpaddr, cp, len); cp += len; 250 len = ROUNDUP(ifaaddr->sa_len); 251 Bcopy(ifaaddr, cp, len); cp += len; 252 } 253 break; 254 255 case RTM_CHANGE: 256 if (gate && 257 (gate->sa_len > (len = rt->rt_gateway->sa_len))) 258 senderr(EDQUOT); 259 /* new gateway could require new ifaddr, ifp; 260 flags may also be different; ifp may be specified 261 by ll sockaddr when protocol address is ambiguous */ 262 if (ifpaddr && (ifa = ifa_ifwithnet(ifpaddr)) && 263 (ifp = ifa->ifa_ifp)) 264 ifa = ifaof_ifpforaddr(ifaaddr ? ifaaddr : gate, 265 ifp); 266 else if ((ifaaddr && (ifa = ifa_ifwithaddr(ifaaddr))) || 267 (ifa = ifa_ifwithroute(rt->rt_flags, 268 rt_key(rt), gate))) 269 ifp = ifa->ifa_ifp; 270 if (ifa) { 271 register struct ifaddr *oifa = rt->rt_ifa; 272 if (oifa != ifa) { 273 if (oifa && oifa->ifa_rtrequest) 274 oifa->ifa_rtrequest(RTM_DELETE, 275 rt, gate); 276 rt->rt_ifa = ifa; 277 rt->rt_ifp = ifp; 278 } 279 } 280 if (gate) 281 Bcopy(gate, rt->rt_gateway, len); 282 rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx, 283 &rt->rt_rmx); 284 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest) 285 rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, gate); 286 if (genmask) 287 rt->rt_genmask = genmask; 288 /* 289 * Fall into 290 */ 291 case RTM_LOCK: 292 rt->rt_rmx.rmx_locks |= 293 (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks); 294 rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits); 295 break; 296 } 297 goto cleanup; 298 299 default: 300 senderr(EOPNOTSUPP); 301 } 302 303 flush: 304 if (rtm) { 305 if (error) 306 rtm->rtm_errno = error; 307 else 308 rtm->rtm_flags |= RTF_DONE; 309 } 310 cleanup: 311 if (rt) 312 rtfree(rt); 313 { 314 register struct rawcb *rp = 0; 315 /* 316 * Check to see if we don't want our own messages. 317 */ 318 if ((so->so_options & SO_USELOOPBACK) == 0) { 319 if (route_cb.any_count <= 1) { 320 if (rtm) 321 Free(rtm); 322 m_freem(m); 323 return (error); 324 } 325 /* There is another listener, so construct message */ 326 rp = sotorawcb(so); 327 } 328 if (rtm) { 329 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm); 330 Free(rtm); 331 } 332 if (rp) 333 rp->rcb_proto.sp_family = 0; /* Avoid us */ 334 if (dst) 335 route_proto.sp_protocol = dst->sa_family; 336 raw_input(m, &route_proto, &route_src, &route_dst); 337 if (rp) 338 rp->rcb_proto.sp_family = PF_ROUTE; 339 } 340 return (error); 341 } 342 343 rt_setmetrics(which, in, out) 344 u_long which; 345 register struct rt_metrics *in, *out; 346 { 347 #define metric(f, e) if (which & (f)) out->e = in->e; 348 metric(RTV_RPIPE, rmx_recvpipe); 349 metric(RTV_SPIPE, rmx_sendpipe); 350 metric(RTV_SSTHRESH, rmx_ssthresh); 351 metric(RTV_RTT, rmx_rtt); 352 metric(RTV_RTTVAR, rmx_rttvar); 353 metric(RTV_HOPCOUNT, rmx_hopcount); 354 metric(RTV_MTU, rmx_mtu); 355 #undef metric 356 } 357 358 /* 359 * Copy data from a buffer back into the indicated mbuf chain, 360 * starting "off" bytes from the beginning, extending the mbuf 361 * chain if necessary. 362 */ 363 m_copyback(m0, off, len, cp) 364 struct mbuf *m0; 365 register int off; 366 register int len; 367 caddr_t cp; 368 369 { 370 register int mlen; 371 register struct mbuf *m = m0, *n; 372 int totlen = 0; 373 374 if (m0 == 0) 375 return; 376 while (off >= (mlen = m->m_len)) { 377 off -= mlen; 378 totlen += mlen; 379 if (m->m_next == 0) { 380 n = m_getclr(M_DONTWAIT, m->m_type); 381 if (n == 0) 382 goto out; 383 n->m_len = min(MLEN, len + off); 384 m->m_next = n; 385 } 386 m = m->m_next; 387 } 388 while (len > 0) { 389 mlen = min (m->m_len - off, len); 390 bcopy(cp, off + mtod(m, caddr_t), (unsigned)mlen); 391 cp += mlen; 392 len -= mlen; 393 mlen += off; 394 off = 0; 395 totlen += mlen; 396 if (len == 0) 397 break; 398 if (m->m_next == 0) { 399 n = m_get(M_DONTWAIT, m->m_type); 400 if (n == 0) 401 break; 402 n->m_len = min(MLEN, len); 403 m->m_next = n; 404 } 405 m = m->m_next; 406 } 407 out: if (((m = m0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen)) 408 m->m_pkthdr.len = totlen; 409 } 410 411 /* 412 * The miss message and losing message are very similar. 413 */ 414 415 rt_missmsg(type, dst, gate, mask, src, flags, error) 416 register struct sockaddr *dst; 417 struct sockaddr *gate, *mask, *src; 418 { 419 register struct rt_msghdr *rtm; 420 register struct mbuf *m; 421 int dlen = ROUNDUP(dst->sa_len); 422 int len = dlen + sizeof(*rtm); 423 424 if (route_cb.any_count == 0) 425 return; 426 m = m_gethdr(M_DONTWAIT, MT_DATA); 427 if (m == 0) 428 return; 429 m->m_pkthdr.len = m->m_len = min(len, MHLEN); 430 m->m_pkthdr.rcvif = 0; 431 rtm = mtod(m, struct rt_msghdr *); 432 bzero((caddr_t)rtm, sizeof(*rtm)); /*XXX assumes sizeof(*rtm) < MHLEN*/ 433 rtm->rtm_flags = RTF_DONE | flags; 434 rtm->rtm_msglen = len; 435 rtm->rtm_version = RTM_VERSION; 436 rtm->rtm_type = type; 437 rtm->rtm_addrs = RTA_DST; 438 if (type == RTM_OLDADD || type == RTM_OLDDEL) { 439 rtm->rtm_pid = u.u_procp->p_pid; 440 } 441 m_copyback(m, sizeof (*rtm), dlen, (caddr_t)dst); 442 if (gate) { 443 dlen = ROUNDUP(gate->sa_len); 444 m_copyback(m, len , dlen, (caddr_t)gate); 445 len += dlen; 446 rtm->rtm_addrs |= RTA_GATEWAY; 447 } 448 if (mask) { 449 dlen = ROUNDUP(mask->sa_len); 450 m_copyback(m, len , dlen, (caddr_t)mask); 451 len += dlen; 452 rtm->rtm_addrs |= RTA_NETMASK; 453 } 454 if (src) { 455 dlen = ROUNDUP(src->sa_len); 456 m_copyback(m, len , dlen, (caddr_t)src); 457 len += dlen; 458 rtm->rtm_addrs |= RTA_AUTHOR; 459 } 460 if (m->m_pkthdr.len != len) { 461 m_freem(m); 462 return; 463 } 464 rtm->rtm_errno = error; 465 rtm->rtm_msglen = len; 466 route_proto.sp_protocol = dst->sa_family; 467 raw_input(m, &route_proto, &route_src, &route_dst); 468 } 469 470 #include "kinfo.h" 471 struct walkarg { 472 int w_op, w_arg; 473 int w_given, w_needed; 474 caddr_t w_where; 475 struct { 476 struct rt_msghdr m_rtm; 477 char m_sabuf[128]; 478 } w_m; 479 #define w_rtm w_m.m_rtm 480 }; 481 /* 482 * This is used in dumping the kernel table via getkinfo(). 483 */ 484 rt_dumpentry(rn, w) 485 struct radix_node *rn; 486 register struct walkarg *w; 487 { 488 register struct sockaddr *sa; 489 int n, error; 490 491 for (; rn; rn = rn->rn_dupedkey) { 492 int count = 0, size = sizeof(w->w_rtm); 493 register struct rtentry *rt = (struct rtentry *)rn; 494 495 if (rn->rn_flags & RNF_ROOT) 496 continue; 497 if (w->w_op == KINFO_RT_FLAGS && !(rt->rt_flags & w->w_arg)) 498 continue; 499 #define next(a, l) {size += (l); w->w_rtm.rtm_addrs |= (a); } 500 w->w_rtm.rtm_addrs = 0; 501 if (sa = rt_key(rt)) 502 next(RTA_DST, ROUNDUP(sa->sa_len)); 503 if (sa = rt->rt_gateway) 504 next(RTA_GATEWAY, ROUNDUP(sa->sa_len)); 505 if (sa = rt_mask(rt)) 506 next(RTA_NETMASK, ROUNDUP(sa->sa_len)); 507 if (sa = rt->rt_genmask) 508 next(RTA_GENMASK, ROUNDUP(sa->sa_len)); 509 w->w_needed += size; 510 if (w->w_where == NULL || w->w_needed > 0) 511 continue; 512 w->w_rtm.rtm_msglen = size; 513 w->w_rtm.rtm_flags = rt->rt_flags; 514 w->w_rtm.rtm_use = rt->rt_use; 515 w->w_rtm.rtm_rmx = rt->rt_rmx; 516 w->w_rtm.rtm_index = rt->rt_ifp->if_index; 517 #undef next 518 #define next(l) {n = (l); Bcopy(sa, cp, n); cp += n;} 519 if (size <= sizeof(w->w_m)) { 520 register caddr_t cp = (caddr_t)(w->w_m.m_sabuf); 521 if (sa = rt_key(rt)) 522 next(ROUNDUP(sa->sa_len)); 523 if (sa = rt->rt_gateway) 524 next(ROUNDUP(sa->sa_len)); 525 if (sa = rt_mask(rt)) 526 next(ROUNDUP(sa->sa_len)); 527 if (sa = rt->rt_genmask) 528 next(ROUNDUP(sa->sa_len)); 529 #undef next 530 #define next(s, l) {n = (l); \ 531 if (error = copyout((caddr_t)(s), w->w_where, n)) return (error); \ 532 w->w_where += n;} 533 534 next(&w->w_m, size); /* Copy rtmsg and sockaddrs back */ 535 continue; 536 } 537 next(&w->w_rtm, sizeof(w->w_rtm)); 538 if (sa = rt_key(rt)) 539 next(sa, ROUNDUP(sa->sa_len)); 540 if (sa = rt->rt_gateway) 541 next(sa, ROUNDUP(sa->sa_len)); 542 if (sa = rt_mask(rt)) 543 next(sa, ROUNDUP(sa->sa_len)); 544 if (sa = rt->rt_genmask) 545 next(sa, ROUNDUP(sa->sa_len)); 546 } 547 return (0); 548 #undef next 549 } 550 551 kinfo_rtable(op, where, given, arg, needed) 552 int op, arg; 553 caddr_t where; 554 int *given, *needed; 555 { 556 register struct radix_node_head *rnh; 557 int s, error = 0; 558 u_char af = ki_af(op); 559 struct walkarg w; 560 561 op &= 0xffff; 562 if (op != KINFO_RT_DUMP && op != KINFO_RT_FLAGS) 563 return (EINVAL); 564 565 Bzero(&w, sizeof(w)); 566 if ((w.w_where = where) && given) 567 w.w_given = *given; 568 w.w_needed = 0 - w.w_given; 569 w.w_arg = arg; 570 w.w_op = op; 571 w.w_rtm.rtm_version = RTM_VERSION; 572 w.w_rtm.rtm_type = RTM_GET; 573 574 s = splnet(); 575 for (rnh = radix_node_head; rnh; rnh = rnh->rnh_next) { 576 if (rnh->rnh_af == 0) 577 continue; 578 if (af && af != rnh->rnh_af) 579 continue; 580 error = rt_walk(rnh->rnh_treetop, rt_dumpentry, &w); 581 if (error) 582 break; 583 } 584 w.w_needed += w.w_given; 585 if (where && given) 586 *given = w.w_where - where; 587 else 588 w.w_needed = (11 * w.w_needed) / 10; 589 *needed = w.w_needed; 590 splx(s); 591 return (error); 592 } 593 594 rt_walk(rn, f, w) 595 register struct radix_node *rn; 596 register int (*f)(); 597 struct walkarg *w; 598 { 599 int error; 600 for (;;) { 601 while (rn->rn_b >= 0) 602 rn = rn->rn_l; /* First time through node, go left */ 603 if (error = (*f)(rn, w)) 604 return (error); /* Process Leaf */ 605 while (rn->rn_p->rn_r == rn) { /* if coming back from right */ 606 rn = rn->rn_p; /* go back up */ 607 if (rn->rn_flags & RNF_ROOT) 608 return 0; 609 } 610 rn = rn->rn_p->rn_r; /* otherwise, go right*/ 611 } 612 } 613 614 /* 615 * Definitions of protocols supported in the ROUTE domain. 616 */ 617 618 int raw_init(),raw_usrreq(),raw_input(),raw_ctlinput(); 619 extern struct domain routedomain; /* or at least forward */ 620 621 struct protosw routesw[] = { 622 { SOCK_RAW, &routedomain, 0, PR_ATOMIC|PR_ADDR, 623 raw_input, route_output, raw_ctlinput, 0, 624 route_usrreq, 625 raw_init, 0, 0, 0, 626 } 627 }; 628 629 int unp_externalize(), unp_dispose(); 630 631 struct domain routedomain = 632 { PF_ROUTE, "route", 0, 0, 0, 633 routesw, &routesw[sizeof(routesw)/sizeof(routesw[0])] }; 634