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