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 7.23 (Berkeley) 01/30/92 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 struct sockaddr route_dst = { 2, PF_ROUTE, }; 24 struct sockaddr route_src = { 2, PF_ROUTE, }; 25 struct sockproto route_proto = { PF_ROUTE, }; 26 27 /*ARGSUSED*/ 28 route_usrreq(so, req, m, nam, control) 29 register struct socket *so; 30 int req; 31 struct mbuf *m, *nam, *control; 32 { 33 register int error = 0; 34 register struct rawcb *rp = sotorawcb(so); 35 int s; 36 if (req == PRU_ATTACH) { 37 MALLOC(rp, struct rawcb *, sizeof(*rp), M_PCB, M_WAITOK); 38 if (so->so_pcb = (caddr_t)rp) 39 bzero(so->so_pcb, sizeof(*rp)); 40 41 } 42 if (req == PRU_DETACH && rp) { 43 int af = rp->rcb_proto.sp_protocol; 44 if (af == AF_INET) 45 route_cb.ip_count--; 46 else if (af == AF_NS) 47 route_cb.ns_count--; 48 else if (af == AF_ISO) 49 route_cb.iso_count--; 50 route_cb.any_count--; 51 } 52 s = splnet(); 53 error = raw_usrreq(so, req, m, nam, control); 54 rp = sotorawcb(so); 55 if (req == PRU_ATTACH && rp) { 56 int af = rp->rcb_proto.sp_protocol; 57 if (error) { 58 free((caddr_t)rp, M_PCB); 59 splx(s); 60 return (error); 61 } 62 if (af == AF_INET) 63 route_cb.ip_count++; 64 else if (af == AF_NS) 65 route_cb.ns_count++; 66 else if (af == AF_ISO) 67 route_cb.iso_count++; 68 rp->rcb_faddr = &route_src; 69 route_cb.any_count++; 70 soisconnected(so); 71 so->so_options |= SO_USELOOPBACK; 72 } 73 splx(s); 74 return (error); 75 } 76 #define ROUNDUP(a) \ 77 ((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long)) 78 #define ADVANCE(x, n) (x += ROUNDUP((n)->sa_len)) 79 80 /*ARGSUSED*/ 81 route_output(m, so) 82 register struct mbuf *m; 83 struct socket *so; 84 { 85 register struct rt_msghdr *rtm = 0; 86 register struct rtentry *rt = 0; 87 struct rtentry *saved_nrt = 0; 88 struct sockaddr *dst = 0, *gate = 0, *netmask = 0, *genmask = 0; 89 struct sockaddr *ifpaddr = 0, *ifaaddr = 0; 90 caddr_t cp, lim; 91 int len, error = 0; 92 struct ifnet *ifp = 0; 93 struct ifaddr *ifa = 0; 94 struct ifaddr *ifaof_ifpforaddr(), *ifa_ifwithroute(); 95 96 #define senderr(e) { error = e; goto flush;} 97 if (m == 0 || m->m_len < sizeof(long)) 98 return (ENOBUFS); 99 if ((m = m_pullup(m, sizeof(long))) == 0) 100 return (ENOBUFS); 101 if ((m->m_flags & M_PKTHDR) == 0) 102 panic("route_output"); 103 len = m->m_pkthdr.len; 104 if (len < sizeof(*rtm) || 105 len != mtod(m, struct rt_msghdr *)->rtm_msglen) 106 senderr(EINVAL); 107 R_Malloc(rtm, struct rt_msghdr *, len); 108 if (rtm == 0) 109 senderr(ENOBUFS); 110 m_copydata(m, 0, len, (caddr_t)rtm); 111 if (rtm->rtm_version != RTM_VERSION) 112 senderr(EPROTONOSUPPORT); 113 rtm->rtm_pid = curproc->p_pid; 114 lim = len + (caddr_t) rtm; 115 cp = (caddr_t) (rtm + 1); 116 if (rtm->rtm_addrs & RTA_DST) { 117 dst = (struct sockaddr *)cp; 118 ADVANCE(cp, dst); 119 } else 120 senderr(EINVAL); 121 if ((rtm->rtm_addrs & RTA_GATEWAY) && cp < lim) { 122 gate = (struct sockaddr *)cp; 123 ADVANCE(cp, gate); 124 } 125 if ((rtm->rtm_addrs & RTA_NETMASK) && cp < lim) { 126 netmask = (struct sockaddr *)cp; 127 ADVANCE(cp, netmask); 128 } 129 if ((rtm->rtm_addrs & RTA_GENMASK) && cp < lim) { 130 struct radix_node *t, *rn_addmask(); 131 genmask = (struct sockaddr *)cp; 132 ADVANCE(cp, genmask); 133 t = rn_addmask(genmask, 1, 2); 134 if (t && Bcmp(genmask, t->rn_key, *(u_char *)genmask) == 0) 135 genmask = (struct sockaddr *)(t->rn_key); 136 else 137 senderr(ENOBUFS); 138 } 139 if ((rtm->rtm_addrs & RTA_IFP) && cp < lim) { 140 ifpaddr = (struct sockaddr *)cp; 141 ADVANCE(cp, ifpaddr); 142 } 143 if ((rtm->rtm_addrs & RTA_IFA) && cp < lim) { 144 ifaaddr = (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 (rtm->rtm_type != RTM_GET) { 172 if (Bcmp(dst, rt_key(rt), dst->sa_len) != 0) 173 senderr(ESRCH); 174 if (rt->rt_nodes->rn_dupedkey && 175 (netmask == 0 || 176 Bcmp(netmask, rt_mask(rt), netmask->sa_len))) 177 senderr(ETOOMANYREFS); 178 } 179 switch(rtm->rtm_type) { 180 181 case RTM_GET: 182 dst = rt_key(rt); len = sizeof(*rtm); 183 ADVANCE(len, dst); 184 rtm->rtm_addrs |= RTA_DST; 185 if (gate = rt->rt_gateway) { 186 ADVANCE(len, gate); 187 rtm->rtm_addrs |= RTA_GATEWAY; 188 } else 189 rtm->rtm_addrs &= ~RTA_GATEWAY; 190 if (netmask = rt_mask(rt)) { 191 ADVANCE(len, netmask); 192 rtm->rtm_addrs |= RTA_NETMASK; 193 } else 194 rtm->rtm_addrs &= ~RTA_NETMASK; 195 if (genmask = rt->rt_genmask) { 196 ADVANCE(len, genmask); 197 rtm->rtm_addrs |= RTA_GENMASK; 198 } else 199 rtm->rtm_addrs &= ~RTA_GENMASK; 200 if (rtm->rtm_addrs & (RTA_IFP | RTA_IFA)) { 201 if (rt->rt_ifp == 0) 202 goto badif; 203 for (ifa = rt->rt_ifp->if_addrlist; 204 ifa && ifa->ifa_addr->sa_family != AF_LINK; 205 ifa = ifa->ifa_next){} 206 if (ifa && rt->rt_ifa) { 207 ifpaddr = ifa->ifa_addr; 208 ADVANCE(len, ifpaddr); 209 ifaaddr = rt->rt_ifa->ifa_addr; 210 ADVANCE(len, ifaaddr); 211 rtm->rtm_addrs |= RTA_IFP | RTA_IFA; 212 } else { 213 badif: ifpaddr = 0; 214 rtm->rtm_addrs &= ~(RTA_IFP | RTA_IFA); 215 } 216 } 217 if (len > rtm->rtm_msglen) { 218 struct rt_msghdr *new_rtm; 219 R_Malloc(new_rtm, struct rt_msghdr *, len); 220 if (new_rtm == 0) 221 senderr(ENOBUFS); 222 Bcopy(rtm, new_rtm, rtm->rtm_msglen); 223 Free(rtm); rtm = new_rtm; 224 } 225 rtm->rtm_msglen = len; 226 rtm->rtm_flags = rt->rt_flags; 227 rtm->rtm_rmx = rt->rt_rmx; 228 cp = (caddr_t) (1 + rtm); 229 len = ROUNDUP(dst->sa_len); 230 Bcopy(dst, cp, len); cp += len; 231 if (gate) { 232 len = ROUNDUP(gate->sa_len); 233 Bcopy(gate, cp, len); cp += len; 234 } 235 if (netmask) { 236 len = ROUNDUP(netmask->sa_len); 237 Bcopy(netmask, cp, len); cp += len; 238 } 239 if (genmask) { 240 len = ROUNDUP(genmask->sa_len); 241 Bcopy(genmask, cp, len); cp += len; 242 } 243 if (ifpaddr) { 244 len = ROUNDUP(ifpaddr->sa_len); 245 Bcopy(ifpaddr, cp, len); cp += len; 246 len = ROUNDUP(ifaaddr->sa_len); 247 Bcopy(ifaaddr, cp, len); cp += len; 248 } 249 break; 250 251 case RTM_CHANGE: 252 if (gate && rt_setgate(rt, rt_key(rt), gate)) 253 senderr(EDQUOT); 254 /* new gateway could require new ifaddr, ifp; 255 flags may also be different; ifp may be specified 256 by ll sockaddr when protocol address is ambiguous */ 257 if (ifpaddr && (ifa = ifa_ifwithnet(ifpaddr)) && 258 (ifp = ifa->ifa_ifp)) 259 ifa = ifaof_ifpforaddr(ifaaddr ? ifaaddr : gate, 260 ifp); 261 else if ((ifaaddr && (ifa = ifa_ifwithaddr(ifaaddr))) || 262 (ifa = ifa_ifwithroute(rt->rt_flags, 263 rt_key(rt), gate))) 264 ifp = ifa->ifa_ifp; 265 if (ifa) { 266 register struct ifaddr *oifa = rt->rt_ifa; 267 if (oifa != ifa) { 268 if (oifa && oifa->ifa_rtrequest) 269 oifa->ifa_rtrequest(RTM_DELETE, 270 rt, gate); 271 rt->rt_ifa = ifa; 272 rt->rt_ifp = ifp; 273 } 274 } 275 rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx, 276 &rt->rt_rmx); 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 &= ~(rtm->rtm_inits); 286 rt->rt_rmx.rmx_locks |= 287 (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks); 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 metric(RTV_EXPIRE, rmx_expire); 349 #undef metric 350 } 351 352 /* 353 * Copy data from a buffer back into the indicated mbuf chain, 354 * starting "off" bytes from the beginning, extending the mbuf 355 * chain if necessary. 356 */ 357 m_copyback(m0, off, len, cp) 358 struct mbuf *m0; 359 register int off; 360 register int len; 361 caddr_t cp; 362 363 { 364 register int mlen; 365 register struct mbuf *m = m0, *n; 366 int totlen = 0; 367 368 if (m0 == 0) 369 return; 370 while (off > (mlen = m->m_len)) { 371 off -= mlen; 372 totlen += mlen; 373 if (m->m_next == 0) { 374 n = m_getclr(M_DONTWAIT, m->m_type); 375 if (n == 0) 376 goto out; 377 n->m_len = min(MLEN, len + off); 378 m->m_next = n; 379 } 380 m = m->m_next; 381 } 382 while (len > 0) { 383 mlen = min (m->m_len - off, len); 384 bcopy(cp, off + mtod(m, caddr_t), (unsigned)mlen); 385 cp += mlen; 386 len -= mlen; 387 mlen += off; 388 off = 0; 389 totlen += mlen; 390 if (len == 0) 391 break; 392 if (m->m_next == 0) { 393 n = m_get(M_DONTWAIT, m->m_type); 394 if (n == 0) 395 break; 396 n->m_len = min(MLEN, len); 397 m->m_next = n; 398 } 399 m = m->m_next; 400 } 401 out: if (((m = m0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen)) 402 m->m_pkthdr.len = totlen; 403 } 404 405 /* 406 * The miss message and losing message are very similar. 407 */ 408 409 rt_missmsg(type, dst, gate, mask, src, flags, error) 410 register struct sockaddr *dst; 411 struct sockaddr *gate, *mask, *src; 412 { 413 register struct rt_msghdr *rtm; 414 register struct mbuf *m; 415 int dlen = ROUNDUP(dst->sa_len); 416 int len = dlen + sizeof(*rtm); 417 418 if (route_cb.any_count == 0) 419 return; 420 m = m_gethdr(M_DONTWAIT, MT_DATA); 421 if (m == 0) 422 return; 423 m->m_pkthdr.len = m->m_len = min(len, MHLEN); 424 m->m_pkthdr.rcvif = 0; 425 rtm = mtod(m, struct rt_msghdr *); 426 bzero((caddr_t)rtm, sizeof(*rtm)); /*XXX assumes sizeof(*rtm) < MHLEN*/ 427 rtm->rtm_flags = RTF_DONE | flags; 428 rtm->rtm_msglen = len; 429 rtm->rtm_version = RTM_VERSION; 430 rtm->rtm_type = type; 431 rtm->rtm_addrs = RTA_DST; 432 if (type == RTM_OLDADD || type == RTM_OLDDEL) { 433 rtm->rtm_pid = curproc->p_pid; 434 } 435 m_copyback(m, sizeof (*rtm), dlen, (caddr_t)dst); 436 if (gate) { 437 dlen = ROUNDUP(gate->sa_len); 438 m_copyback(m, len , dlen, (caddr_t)gate); 439 len += dlen; 440 rtm->rtm_addrs |= RTA_GATEWAY; 441 } 442 if (mask) { 443 dlen = ROUNDUP(mask->sa_len); 444 m_copyback(m, len , dlen, (caddr_t)mask); 445 len += dlen; 446 rtm->rtm_addrs |= RTA_NETMASK; 447 } 448 if (src) { 449 dlen = ROUNDUP(src->sa_len); 450 m_copyback(m, len , dlen, (caddr_t)src); 451 len += dlen; 452 rtm->rtm_addrs |= RTA_AUTHOR; 453 } 454 if (m->m_pkthdr.len != len) { 455 m_freem(m); 456 return; 457 } 458 rtm->rtm_errno = error; 459 rtm->rtm_msglen = len; 460 route_proto.sp_protocol = dst->sa_family; 461 raw_input(m, &route_proto, &route_src, &route_dst); 462 } 463 464 #include "kinfo.h" 465 struct walkarg { 466 int w_op, w_arg; 467 int w_given, w_needed; 468 caddr_t w_where; 469 struct { 470 struct rt_msghdr m_rtm; 471 char m_sabuf[128]; 472 } w_m; 473 #define w_rtm w_m.m_rtm 474 }; 475 /* 476 * This is used in dumping the kernel table via getkinfo(). 477 */ 478 rt_dumpentry(rn, w) 479 struct radix_node *rn; 480 register struct walkarg *w; 481 { 482 register struct sockaddr *sa; 483 int n, error, size = sizeof(w->w_rtm); 484 register struct rtentry *rt = (struct rtentry *)rn; 485 486 if (w->w_op == KINFO_RT_FLAGS && !(rt->rt_flags & w->w_arg)) 487 return 0; 488 #define next(a, l) {size += (l); w->w_rtm.rtm_addrs |= (a); } 489 w->w_rtm.rtm_addrs = 0; 490 if (sa = rt_key(rt)) 491 next(RTA_DST, ROUNDUP(sa->sa_len)); 492 if (sa = rt->rt_gateway) 493 next(RTA_GATEWAY, ROUNDUP(sa->sa_len)); 494 if (sa = rt_mask(rt)) 495 next(RTA_NETMASK, ROUNDUP(sa->sa_len)); 496 if (sa = rt->rt_genmask) 497 next(RTA_GENMASK, ROUNDUP(sa->sa_len)); 498 w->w_needed += size; 499 if (w->w_where == NULL || w->w_needed > 0) 500 return 0; 501 w->w_rtm.rtm_msglen = size; 502 w->w_rtm.rtm_flags = rt->rt_flags; 503 w->w_rtm.rtm_use = rt->rt_use; 504 w->w_rtm.rtm_rmx = rt->rt_rmx; 505 w->w_rtm.rtm_index = rt->rt_ifp->if_index; 506 #undef next 507 #define next(l) {n = (l); Bcopy(sa, cp, n); cp += n;} 508 if (size <= sizeof(w->w_m)) { 509 register caddr_t cp = (caddr_t)(w->w_m.m_sabuf); 510 if (sa = rt_key(rt)) 511 next(ROUNDUP(sa->sa_len)); 512 if (sa = rt->rt_gateway) 513 next(ROUNDUP(sa->sa_len)); 514 if (sa = rt_mask(rt)) 515 next(ROUNDUP(sa->sa_len)); 516 if (sa = rt->rt_genmask) 517 next(ROUNDUP(sa->sa_len)); 518 #undef next 519 #define next(s, l) {n = (l); \ 520 if (error = copyout((caddr_t)(s), w->w_where, n)) return (error); \ 521 w->w_where += n;} 522 523 next(&w->w_m, size); /* Copy rtmsg and sockaddrs back */ 524 return 0; 525 } 526 next(&w->w_rtm, sizeof(w->w_rtm)); 527 if (sa = rt_key(rt)) 528 next(sa, ROUNDUP(sa->sa_len)); 529 if (sa = rt->rt_gateway) 530 next(sa, ROUNDUP(sa->sa_len)); 531 if (sa = rt_mask(rt)) 532 next(sa, ROUNDUP(sa->sa_len)); 533 if (sa = rt->rt_genmask) 534 next(sa, ROUNDUP(sa->sa_len)); 535 return (0); 536 #undef next 537 } 538 539 kinfo_rtable(op, where, given, arg, needed) 540 int op, arg; 541 caddr_t where; 542 int *given, *needed; 543 { 544 register struct radix_node_head *rnh; 545 int s, error = 0, i; 546 u_char af = ki_af(op); 547 struct walkarg w; 548 549 op &= 0xffff; 550 if (op != KINFO_RT_DUMP && op != KINFO_RT_FLAGS) 551 return (EINVAL); 552 553 Bzero(&w, sizeof(w)); 554 if ((w.w_where = where) && given) 555 w.w_given = *given; 556 w.w_needed = 0 - w.w_given; 557 w.w_arg = arg; 558 w.w_op = op; 559 w.w_rtm.rtm_version = RTM_VERSION; 560 w.w_rtm.rtm_type = RTM_GET; 561 562 s = splnet(); 563 for (i = 0; i <= AF_MAX; i++) { 564 if ((rnh = rt_tables[i]) == 0 || i == 0 || (af && af != i)) 565 continue; 566 error = rnh->rnh_walk(rnh->rnh_treetop, rt_dumpentry, &w); 567 if (error) 568 break; 569 } 570 w.w_needed += w.w_given; 571 if (where && given) 572 *given = w.w_where - where; 573 else 574 w.w_needed = (11 * w.w_needed) / 10; 575 *needed = w.w_needed; 576 splx(s); 577 return (error); 578 } 579 580 /* 581 * Definitions of protocols supported in the ROUTE domain. 582 */ 583 584 int raw_init(),raw_usrreq(),raw_input(),raw_ctlinput(); 585 extern struct domain routedomain; /* or at least forward */ 586 587 struct protosw routesw[] = { 588 { SOCK_RAW, &routedomain, 0, PR_ATOMIC|PR_ADDR, 589 raw_input, route_output, raw_ctlinput, 0, 590 route_usrreq, 591 raw_init, 0, 0, 0, 592 } 593 }; 594 595 int unp_externalize(), unp_dispose(); 596 597 struct domain routedomain = 598 { PF_ROUTE, "route", 0, 0, 0, 599 routesw, &routesw[sizeof(routesw)/sizeof(routesw[0])] }; 600