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 7.14 (Berkeley) 09/03/90 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) (1 + (((a) - 1) | (sizeof(long) - 1))) 81 #define ADVANCE(x, n) (x += ((n) > 0 ? ROUNDUP(n) : sizeof(long))) 82 83 /*ARGSUSED*/ 84 route_output(m, so) 85 register struct mbuf *m; 86 struct socket *so; 87 { 88 register struct rt_msghdr *rtm = 0; 89 register struct rtentry *rt = 0; 90 struct rtentry *saved_nrt = 0; 91 struct sockaddr *dst = 0, *gate = 0, *netmask = 0, *genmask = 0; 92 struct sockaddr *ifpaddr = 0; 93 caddr_t cp, lim; 94 int len, error = 0; 95 struct ifnet *ifp = 0; 96 struct ifaddr *ifa; 97 extern struct ifaddr *ifaof_ifpforaddr(), *ifa_ifwithroute(); 98 99 #define senderr(e) { error = e; goto flush;} 100 if (m == 0 || m->m_len < sizeof(long)) 101 return (ENOBUFS); 102 if ((m = m_pullup(m, sizeof(long))) == 0) 103 return (ENOBUFS); 104 if ((m->m_flags & M_PKTHDR) == 0) 105 panic("route_output"); 106 len = m->m_pkthdr.len; 107 if (len < sizeof(*rtm) || 108 len != mtod(m, struct rt_msghdr *)->rtm_msglen) 109 senderr(EINVAL); 110 R_Malloc(rtm, struct rt_msghdr *, len); 111 if (rtm == 0) 112 senderr(ENOBUFS); 113 m_copydata(m, 0, len, (caddr_t)rtm); 114 if (rtm->rtm_version != RTM_VERSION) 115 senderr(EPROTONOSUPPORT); 116 rtm->rtm_pid = u.u_procp->p_pid; 117 lim = len + (caddr_t) rtm; 118 cp = (caddr_t) (rtm + 1); 119 if (rtm->rtm_addrs & RTA_DST) { 120 dst = (struct sockaddr *)cp; 121 ADVANCE(cp, dst->sa_len); 122 } else 123 senderr(EINVAL); 124 if ((rtm->rtm_addrs & RTA_GATEWAY) && cp < lim) { 125 gate = (struct sockaddr *)cp; 126 cp += ROUNDUP(gate->sa_len); 127 } 128 if ((rtm->rtm_addrs & RTA_NETMASK) && cp < lim) { 129 netmask = (struct sockaddr *)cp; 130 ADVANCE(cp, netmask->sa_len); 131 } 132 if ((rtm->rtm_addrs & RTA_GENMASK) && cp < lim) { 133 struct radix_node *t, *rn_addmask(); 134 genmask = (struct sockaddr *)cp; 135 ADVANCE(cp, genmask->sa_len); 136 t = rn_addmask(genmask, 1, 2); 137 if (t && Bcmp(genmask, t->rn_key, *(u_char *)genmask) == 0) 138 genmask = (struct sockaddr *)(t->rn_key); 139 else 140 senderr(ENOBUFS); 141 } 142 if ((rtm->rtm_addrs & RTA_IFP) && cp < lim) { 143 ifpaddr = (struct sockaddr *)cp; 144 } 145 switch (rtm->rtm_type) { 146 case RTM_ADD: 147 if (gate == 0) 148 senderr(EINVAL); 149 error = rtrequest(RTM_ADD, dst, gate, netmask, 150 rtm->rtm_flags, &saved_nrt); 151 if (error == 0 && saved_nrt) { 152 rt_setmetrics(rtm->rtm_inits, 153 &rtm->rtm_rmx, &saved_nrt->rt_rmx); 154 saved_nrt->rt_refcnt--; 155 saved_nrt->rt_genmask = genmask; 156 } 157 break; 158 159 case RTM_DELETE: 160 error = rtrequest(RTM_DELETE, dst, gate, netmask, 161 rtm->rtm_flags, (struct rtentry **)0); 162 break; 163 164 case RTM_GET: 165 case RTM_CHANGE: 166 case RTM_LOCK: 167 rt = rtalloc1(dst, 0); 168 if (rt == 0) 169 senderr(ESRCH); 170 if (((struct radix_node *)rt)->rn_dupedkey) { 171 if (netmask == 0 && rtm->rtm_type != RTM_GET) 172 senderr(ETOOMANYREFS); 173 } 174 switch(rtm->rtm_type) { 175 176 case RTM_GET: 177 dst = rt_key(rt); 178 len = sizeof(*rtm) + (dst->sa_len > 0 ? 179 ROUNDUP(dst->sa_len) : sizeof(long)); 180 rtm->rtm_addrs |= RTA_DST; 181 if (gate = rt->rt_gateway) { 182 len += ROUNDUP(rt->rt_gateway->sa_len); 183 rtm->rtm_addrs |= RTA_GATEWAY; 184 } else 185 rtm->rtm_addrs &= ~RTA_GATEWAY; 186 if (netmask = rt_mask(rt)) { 187 len += netmask->sa_len; 188 rtm->rtm_addrs |= RTA_NETMASK; 189 } else 190 rtm->rtm_addrs &= ~RTA_NETMASK; 191 if (rt->rt_genmask) { 192 len += ROUNDUP(rt->rt_genmask->sa_len); 193 rtm->rtm_addrs |= RTA_GENMASK; 194 } else 195 rtm->rtm_addrs &= ~RTA_GENMASK; 196 if (len > rtm->rtm_msglen) { 197 struct rt_msghdr *new_rtm; 198 R_Malloc(new_rtm, struct rt_msghdr *, len); 199 if (new_rtm == 0) 200 senderr(ENOBUFS); 201 Bcopy(rtm, new_rtm, rtm->rtm_msglen); 202 Free(rtm); rtm = new_rtm; 203 } 204 rtm->rtm_msglen = len; 205 rtm->rtm_flags = rt->rt_flags; 206 rtm->rtm_rmx = rt->rt_rmx; 207 cp = (caddr_t) (1 + rtm); 208 Bcopy(dst, cp, dst->sa_len); 209 ADVANCE(cp, dst->sa_len); 210 if (gate) { 211 Bcopy(gate, cp, gate->sa_len); 212 ADVANCE(cp, gate->sa_len); 213 } 214 if (netmask) { 215 Bcopy(netmask, cp, netmask->sa_len); 216 ADVANCE(cp, netmask->sa_len); 217 } 218 if (rt->rt_genmask) { 219 Bcopy(rt->rt_genmask, cp, rt->rt_genmask->sa_len); 220 ADVANCE(cp, rt->rt_genmask->sa_len); 221 } 222 break; 223 224 case RTM_CHANGE: 225 if (gate && 226 (gate->sa_len > (len = rt->rt_gateway->sa_len))) 227 senderr(EDQUOT); 228 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest) 229 rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, gate); 230 /* new gateway could require new ifaddr, ifp; 231 flags may also be different; ifp may be specified 232 by ll sockaddr when protocol address is ambiguous */ 233 if (ifpaddr && 234 (ifa = ifa_ifwithnet(ifpaddr)) && 235 (ifp = ifa->ifa_ifp) && 236 (ifa = ifaof_ifpforaddr(gate, ifp))) { 237 /* We got it */ 238 } else { 239 ifa = 0; ifp = 0; 240 } 241 if (gate) 242 Bcopy(gate, rt->rt_gateway, len); 243 rt_setmetrics(rtm->rtm_inits, 244 &rtm->rtm_rmx, &rt->rt_rmx); 245 if (ifa == 0) 246 ifa = ifa_ifwithroute(rt->rt_flags, rt_key(rt), 247 gate); 248 if (ifa) { 249 if (rt->rt_ifa != ifa) { 250 rt->rt_ifa = ifa; 251 rt->rt_ifp = ifa->ifa_ifp; 252 } 253 } 254 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest) 255 rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, gate); 256 if (genmask) 257 rt->rt_genmask = genmask; 258 /* 259 * Fall into 260 */ 261 case RTM_LOCK: 262 rt->rt_rmx.rmx_locks |= 263 (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks); 264 rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits); 265 break; 266 } 267 goto cleanup; 268 269 default: 270 senderr(EOPNOTSUPP); 271 } 272 273 flush: 274 if (rtm) { 275 if (error) 276 rtm->rtm_errno = error; 277 else 278 rtm->rtm_flags |= RTF_DONE; 279 } 280 cleanup: 281 if (rt) 282 rtfree(rt); 283 { 284 register struct rawcb *rp = 0; 285 /* 286 * Check to see if we don't want our own messages. 287 */ 288 if ((so->so_options & SO_USELOOPBACK) == 0) { 289 if (route_cb.any_count <= 1) { 290 if (rtm) 291 Free(rtm); 292 m_freem(m); 293 return (error); 294 } 295 /* There is another listener, so construct message */ 296 rp = sotorawcb(so); 297 } 298 if (cp = (caddr_t)rtm) { 299 m_copyback(m, 0, len, cp); 300 Free(rtm); 301 } 302 if (rp) 303 rp->rcb_proto.sp_family = 0; /* Avoid us */ 304 if (dst) 305 route_proto.sp_protocol = dst->sa_family; 306 raw_input(m, &route_proto, &route_src, &route_dst); 307 if (rp) 308 rp->rcb_proto.sp_family = PF_ROUTE; 309 } 310 return (error); 311 } 312 313 rt_setmetrics(which, in, out) 314 u_long which; 315 register struct rt_metrics *in, *out; 316 { 317 #define metric(f, e) if (which & (f)) out->e = in->e; 318 metric(RTV_RPIPE, rmx_recvpipe); 319 metric(RTV_SPIPE, rmx_sendpipe); 320 metric(RTV_SSTHRESH, rmx_ssthresh); 321 metric(RTV_RTT, rmx_rtt); 322 metric(RTV_RTTVAR, rmx_rttvar); 323 metric(RTV_HOPCOUNT, rmx_hopcount); 324 metric(RTV_MTU, rmx_mtu); 325 #undef metric 326 } 327 328 /* 329 * Copy data from a buffer back into the indicated mbuf chain, 330 * starting "off" bytes from the beginning, extending the mbuf 331 * chain if necessary. 332 */ 333 m_copyback(m0, off, len, cp) 334 struct mbuf *m0; 335 register int off; 336 register int len; 337 caddr_t cp; 338 339 { 340 register int mlen; 341 register struct mbuf *m = m0, *n; 342 int totlen = 0; 343 344 if (m0 == 0) 345 return; 346 while (off >= (mlen = m->m_len)) { 347 off -= mlen; 348 totlen += mlen; 349 if (m->m_next == 0) { 350 n = m_getclr(M_DONTWAIT, m->m_type); 351 if (n == 0) 352 goto out; 353 n->m_len = min(MLEN, len + off); 354 m->m_next = n; 355 } 356 m = m->m_next; 357 } 358 while (len > 0) { 359 mlen = min (m->m_len - off, len); 360 bcopy(cp, off + mtod(m, caddr_t), (unsigned)mlen); 361 cp += mlen; 362 len -= mlen; 363 mlen += off; 364 off = 0; 365 totlen += mlen; 366 if (len == 0) 367 break; 368 if (m->m_next == 0) { 369 n = m_get(M_DONTWAIT, m->m_type); 370 if (n == 0) 371 break; 372 n->m_len = min(MLEN, len); 373 m->m_next = n; 374 } 375 m = m->m_next; 376 } 377 out: if (((m = m0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen)) 378 m->m_pkthdr.len = totlen; 379 } 380 381 /* 382 * The miss message and losing message are very similar. 383 */ 384 385 rt_missmsg(type, dst, gate, mask, src, flags, error) 386 register struct sockaddr *dst; 387 struct sockaddr *gate, *mask, *src; 388 { 389 register struct rt_msghdr *rtm; 390 register struct mbuf *m; 391 int dlen = ROUNDUP(dst->sa_len); 392 int len = dlen + sizeof(*rtm); 393 394 if (route_cb.any_count == 0) 395 return; 396 m = m_gethdr(M_DONTWAIT, MT_DATA); 397 if (m == 0) 398 return; 399 m->m_pkthdr.len = m->m_len = min(len, MHLEN); 400 m->m_pkthdr.rcvif = 0; 401 rtm = mtod(m, struct rt_msghdr *); 402 bzero((caddr_t)rtm, sizeof(*rtm)); /*XXX assumes sizeof(*rtm) < MHLEN*/ 403 rtm->rtm_flags = RTF_DONE | flags; 404 rtm->rtm_msglen = len; 405 rtm->rtm_version = RTM_VERSION; 406 rtm->rtm_type = type; 407 rtm->rtm_addrs = RTA_DST; 408 if (type == RTM_OLDADD || type == RTM_OLDDEL) { 409 rtm->rtm_pid = u.u_procp->p_pid; 410 } 411 m_copyback(m, sizeof (*rtm), dlen, (caddr_t)dst); 412 if (gate) { 413 dlen = ROUNDUP(gate->sa_len); 414 m_copyback(m, len , dlen, (caddr_t)gate); 415 len += dlen; 416 rtm->rtm_addrs |= RTA_GATEWAY; 417 } 418 if (mask) { 419 if (mask->sa_len) 420 dlen = ROUNDUP(mask->sa_len); 421 else 422 dlen = sizeof(long); 423 m_copyback(m, len , dlen, (caddr_t)mask); 424 len += dlen; 425 rtm->rtm_addrs |= RTA_NETMASK; 426 } 427 if (src) { 428 dlen = ROUNDUP(src->sa_len); 429 m_copyback(m, len , dlen, (caddr_t)src); 430 len += dlen; 431 rtm->rtm_addrs |= RTA_AUTHOR; 432 } 433 if (m->m_pkthdr.len != len) { 434 m_freem(m); 435 return; 436 } 437 rtm->rtm_errno = error; 438 rtm->rtm_msglen = len; 439 route_proto.sp_protocol = dst->sa_family; 440 raw_input(m, &route_proto, &route_src, &route_dst); 441 } 442 443 #include "kinfo.h" 444 struct walkarg { 445 int w_op, w_arg; 446 int w_given, w_needed; 447 caddr_t w_where; 448 struct { 449 struct rt_msghdr m_rtm; 450 char m_sabuf[128]; 451 } w_m; 452 #define w_rtm w_m.m_rtm 453 }; 454 /* 455 * This is used in dumping the kernel table via getkinfo(). 456 */ 457 rt_dumpentry(rn, w) 458 struct radix_node *rn; 459 register struct walkarg *w; 460 { 461 register struct sockaddr *sa; 462 int n, error; 463 464 for (; rn; rn = rn->rn_dupedkey) { 465 int count = 0, size = sizeof(w->w_rtm); 466 register struct rtentry *rt = (struct rtentry *)rn; 467 468 if (rn->rn_flags & RNF_ROOT) 469 continue; 470 if (w->w_op == KINFO_RT_FLAGS && !(rt->rt_flags & w->w_arg)) 471 continue; 472 #define next(a, l) {size += (l); w->w_rtm.rtm_addrs |= (a); } 473 w->w_rtm.rtm_addrs = 0; 474 if (sa = rt_key(rt)) 475 next(RTA_DST, ROUNDUP(sa->sa_len)); 476 if (sa = rt->rt_gateway) 477 next(RTA_GATEWAY, ROUNDUP(sa->sa_len)); 478 if (sa = rt_mask(rt)) 479 next(RTA_NETMASK, 480 sa->sa_len ? ROUNDUP(sa->sa_len) : sizeof(long)); 481 if (sa = rt->rt_genmask) 482 next(RTA_GENMASK, ROUNDUP(sa->sa_len)); 483 w->w_needed += size; 484 if (w->w_where == NULL || w->w_needed > 0) 485 continue; 486 w->w_rtm.rtm_msglen = size; 487 w->w_rtm.rtm_flags = rt->rt_flags; 488 w->w_rtm.rtm_use = rt->rt_use; 489 w->w_rtm.rtm_rmx = rt->rt_rmx; 490 w->w_rtm.rtm_index = rt->rt_ifp->if_index; 491 #undef next 492 #define next(l) {n = (l); Bcopy(sa, cp, n); cp += n;} 493 if (size <= sizeof(w->w_m)) { 494 register caddr_t cp = (caddr_t)(w->w_m.m_sabuf); 495 if (sa = rt_key(rt)) 496 next(ROUNDUP(sa->sa_len)); 497 if (sa = rt->rt_gateway) 498 next(ROUNDUP(sa->sa_len)); 499 if (sa = rt_mask(rt)) 500 next(sa->sa_len ? ROUNDUP(sa->sa_len) : sizeof(long)); 501 if (sa = rt->rt_genmask) 502 next(ROUNDUP(sa->sa_len)); 503 #undef next 504 #define next(s, l) {n = (l); \ 505 if (error = copyout((caddr_t)(s), w->w_where, n)) return (error); \ 506 w->w_where += n;} 507 508 next(&w->w_m, size); /* Copy rtmsg and sockaddrs back */ 509 continue; 510 } 511 next(&w->w_rtm, sizeof(w->w_rtm)); 512 if (sa = rt_key(rt)) 513 next(sa, ROUNDUP(sa->sa_len)); 514 if (sa = rt->rt_gateway) 515 next(sa, ROUNDUP(sa->sa_len)); 516 if (sa = rt_mask(rt)) 517 next(sa, sa->sa_len ? ROUNDUP(sa->sa_len) : sizeof(long)); 518 if (sa = rt->rt_genmask) 519 next(sa, ROUNDUP(sa->sa_len)); 520 } 521 return (0); 522 #undef next 523 } 524 525 kinfo_rtable(op, where, given, arg, needed) 526 int op, arg; 527 caddr_t where; 528 int *given, *needed; 529 { 530 register struct radix_node_head *rnh; 531 int s, error = 0; 532 u_char af = ki_af(op); 533 struct walkarg w; 534 535 op &= 0xffff; 536 if (op != KINFO_RT_DUMP && op != KINFO_RT_FLAGS) 537 return (EINVAL); 538 539 Bzero(&w, sizeof(w)); 540 if ((w.w_where = where) && given) 541 w.w_given = *given; 542 w.w_needed = 0 - w.w_given; 543 w.w_arg = arg; 544 w.w_op = op; 545 w.w_rtm.rtm_version = RTM_VERSION; 546 w.w_rtm.rtm_type = RTM_GET; 547 548 s = splnet(); 549 for (rnh = radix_node_head; rnh; rnh = rnh->rnh_next) { 550 if (rnh->rnh_af == 0) 551 continue; 552 if (af && af != rnh->rnh_af) 553 continue; 554 error = rt_walk(rnh->rnh_treetop, rt_dumpentry, &w); 555 if (error) 556 break; 557 } 558 w.w_needed += w.w_given; 559 if (where && given) 560 *given = w.w_where - where; 561 else 562 w.w_needed = (11 * w.w_needed) / 10; 563 *needed = w.w_needed; 564 splx(s); 565 return (error); 566 } 567 568 rt_walk(rn, f, w) 569 register struct radix_node *rn; 570 register int (*f)(); 571 struct walkarg *w; 572 { 573 int error; 574 for (;;) { 575 while (rn->rn_b >= 0) 576 rn = rn->rn_l; /* First time through node, go left */ 577 if (error = (*f)(rn, w)) 578 return (error); /* Process Leaf */ 579 while (rn->rn_p->rn_r == rn) { /* if coming back from right */ 580 rn = rn->rn_p; /* go back up */ 581 if (rn->rn_flags & RNF_ROOT) 582 return 0; 583 } 584 rn = rn->rn_p->rn_r; /* otherwise, go right*/ 585 } 586 } 587 588 /* 589 * Definitions of protocols supported in the ROUTE domain. 590 */ 591 592 int raw_init(),raw_usrreq(),raw_input(),raw_ctlinput(); 593 extern struct domain routedomain; /* or at least forward */ 594 595 struct protosw routesw[] = { 596 { SOCK_RAW, &routedomain, 0, PR_ATOMIC|PR_ADDR, 597 raw_input, route_output, raw_ctlinput, 0, 598 route_usrreq, 599 raw_init, 0, 0, 0, 600 } 601 }; 602 603 int unp_externalize(), unp_dispose(); 604 605 struct domain routedomain = 606 { PF_ROUTE, "route", 0, 0, 0, 607 routesw, &routesw[sizeof(routesw)/sizeof(routesw[0])] }; 608