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 7.22 (Berkeley) 08/08/91 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 && rt_setgate(rt, rt_key(rt), gate)) 255 senderr(EDQUOT); 256 /* new gateway could require new ifaddr, ifp; 257 flags may also be different; ifp may be specified 258 by ll sockaddr when protocol address is ambiguous */ 259 if (ifpaddr && (ifa = ifa_ifwithnet(ifpaddr)) && 260 (ifp = ifa->ifa_ifp)) 261 ifa = ifaof_ifpforaddr(ifaaddr ? ifaaddr : gate, 262 ifp); 263 else if ((ifaaddr && (ifa = ifa_ifwithaddr(ifaaddr))) || 264 (ifa = ifa_ifwithroute(rt->rt_flags, 265 rt_key(rt), gate))) 266 ifp = ifa->ifa_ifp; 267 if (ifa) { 268 register struct ifaddr *oifa = rt->rt_ifa; 269 if (oifa != ifa) { 270 if (oifa && oifa->ifa_rtrequest) 271 oifa->ifa_rtrequest(RTM_DELETE, 272 rt, gate); 273 rt->rt_ifa = ifa; 274 rt->rt_ifp = ifp; 275 } 276 } 277 rt_setmetrics(rtm->rtm_inits, &rtm->rtm_rmx, 278 &rt->rt_rmx); 279 if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest) 280 rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, gate); 281 if (genmask) 282 rt->rt_genmask = genmask; 283 /* 284 * Fall into 285 */ 286 case RTM_LOCK: 287 rt->rt_rmx.rmx_locks &= ~(rtm->rtm_inits); 288 rt->rt_rmx.rmx_locks |= 289 (rtm->rtm_inits & rtm->rtm_rmx.rmx_locks); 290 break; 291 } 292 goto cleanup; 293 294 default: 295 senderr(EOPNOTSUPP); 296 } 297 298 flush: 299 if (rtm) { 300 if (error) 301 rtm->rtm_errno = error; 302 else 303 rtm->rtm_flags |= RTF_DONE; 304 } 305 cleanup: 306 if (rt) 307 rtfree(rt); 308 { 309 register struct rawcb *rp = 0; 310 /* 311 * Check to see if we don't want our own messages. 312 */ 313 if ((so->so_options & SO_USELOOPBACK) == 0) { 314 if (route_cb.any_count <= 1) { 315 if (rtm) 316 Free(rtm); 317 m_freem(m); 318 return (error); 319 } 320 /* There is another listener, so construct message */ 321 rp = sotorawcb(so); 322 } 323 if (rtm) { 324 m_copyback(m, 0, rtm->rtm_msglen, (caddr_t)rtm); 325 Free(rtm); 326 } 327 if (rp) 328 rp->rcb_proto.sp_family = 0; /* Avoid us */ 329 if (dst) 330 route_proto.sp_protocol = dst->sa_family; 331 raw_input(m, &route_proto, &route_src, &route_dst); 332 if (rp) 333 rp->rcb_proto.sp_family = PF_ROUTE; 334 } 335 return (error); 336 } 337 338 rt_setmetrics(which, in, out) 339 u_long which; 340 register struct rt_metrics *in, *out; 341 { 342 #define metric(f, e) if (which & (f)) out->e = in->e; 343 metric(RTV_RPIPE, rmx_recvpipe); 344 metric(RTV_SPIPE, rmx_sendpipe); 345 metric(RTV_SSTHRESH, rmx_ssthresh); 346 metric(RTV_RTT, rmx_rtt); 347 metric(RTV_RTTVAR, rmx_rttvar); 348 metric(RTV_HOPCOUNT, rmx_hopcount); 349 metric(RTV_MTU, rmx_mtu); 350 metric(RTV_EXPIRE, rmx_expire); 351 #undef metric 352 } 353 354 /* 355 * Copy data from a buffer back into the indicated mbuf chain, 356 * starting "off" bytes from the beginning, extending the mbuf 357 * chain if necessary. 358 */ 359 m_copyback(m0, off, len, cp) 360 struct mbuf *m0; 361 register int off; 362 register int len; 363 caddr_t cp; 364 365 { 366 register int mlen; 367 register struct mbuf *m = m0, *n; 368 int totlen = 0; 369 370 if (m0 == 0) 371 return; 372 while (off > (mlen = m->m_len)) { 373 off -= mlen; 374 totlen += mlen; 375 if (m->m_next == 0) { 376 n = m_getclr(M_DONTWAIT, m->m_type); 377 if (n == 0) 378 goto out; 379 n->m_len = min(MLEN, len + off); 380 m->m_next = n; 381 } 382 m = m->m_next; 383 } 384 while (len > 0) { 385 mlen = min (m->m_len - off, len); 386 bcopy(cp, off + mtod(m, caddr_t), (unsigned)mlen); 387 cp += mlen; 388 len -= mlen; 389 mlen += off; 390 off = 0; 391 totlen += mlen; 392 if (len == 0) 393 break; 394 if (m->m_next == 0) { 395 n = m_get(M_DONTWAIT, m->m_type); 396 if (n == 0) 397 break; 398 n->m_len = min(MLEN, len); 399 m->m_next = n; 400 } 401 m = m->m_next; 402 } 403 out: if (((m = m0)->m_flags & M_PKTHDR) && (m->m_pkthdr.len < totlen)) 404 m->m_pkthdr.len = totlen; 405 } 406 407 /* 408 * The miss message and losing message are very similar. 409 */ 410 411 rt_missmsg(type, dst, gate, mask, src, flags, error) 412 register struct sockaddr *dst; 413 struct sockaddr *gate, *mask, *src; 414 { 415 register struct rt_msghdr *rtm; 416 register struct mbuf *m; 417 int dlen = ROUNDUP(dst->sa_len); 418 int len = dlen + sizeof(*rtm); 419 420 if (route_cb.any_count == 0) 421 return; 422 m = m_gethdr(M_DONTWAIT, MT_DATA); 423 if (m == 0) 424 return; 425 m->m_pkthdr.len = m->m_len = min(len, MHLEN); 426 m->m_pkthdr.rcvif = 0; 427 rtm = mtod(m, struct rt_msghdr *); 428 bzero((caddr_t)rtm, sizeof(*rtm)); /*XXX assumes sizeof(*rtm) < MHLEN*/ 429 rtm->rtm_flags = RTF_DONE | flags; 430 rtm->rtm_msglen = len; 431 rtm->rtm_version = RTM_VERSION; 432 rtm->rtm_type = type; 433 rtm->rtm_addrs = RTA_DST; 434 if (type == RTM_OLDADD || type == RTM_OLDDEL) { 435 rtm->rtm_pid = curproc->p_pid; 436 } 437 m_copyback(m, sizeof (*rtm), dlen, (caddr_t)dst); 438 if (gate) { 439 dlen = ROUNDUP(gate->sa_len); 440 m_copyback(m, len , dlen, (caddr_t)gate); 441 len += dlen; 442 rtm->rtm_addrs |= RTA_GATEWAY; 443 } 444 if (mask) { 445 dlen = ROUNDUP(mask->sa_len); 446 m_copyback(m, len , dlen, (caddr_t)mask); 447 len += dlen; 448 rtm->rtm_addrs |= RTA_NETMASK; 449 } 450 if (src) { 451 dlen = ROUNDUP(src->sa_len); 452 m_copyback(m, len , dlen, (caddr_t)src); 453 len += dlen; 454 rtm->rtm_addrs |= RTA_AUTHOR; 455 } 456 if (m->m_pkthdr.len != len) { 457 m_freem(m); 458 return; 459 } 460 rtm->rtm_errno = error; 461 rtm->rtm_msglen = len; 462 route_proto.sp_protocol = dst->sa_family; 463 raw_input(m, &route_proto, &route_src, &route_dst); 464 } 465 466 #include "kinfo.h" 467 struct walkarg { 468 int w_op, w_arg; 469 int w_given, w_needed; 470 caddr_t w_where; 471 struct { 472 struct rt_msghdr m_rtm; 473 char m_sabuf[128]; 474 } w_m; 475 #define w_rtm w_m.m_rtm 476 }; 477 /* 478 * This is used in dumping the kernel table via getkinfo(). 479 */ 480 rt_dumpentry(rn, w) 481 struct radix_node *rn; 482 register struct walkarg *w; 483 { 484 register struct sockaddr *sa; 485 int n, error, size = sizeof(w->w_rtm); 486 register struct rtentry *rt = (struct rtentry *)rn; 487 488 if (w->w_op == KINFO_RT_FLAGS && !(rt->rt_flags & w->w_arg)) 489 return 0; 490 #define next(a, l) {size += (l); w->w_rtm.rtm_addrs |= (a); } 491 w->w_rtm.rtm_addrs = 0; 492 if (sa = rt_key(rt)) 493 next(RTA_DST, ROUNDUP(sa->sa_len)); 494 if (sa = rt->rt_gateway) 495 next(RTA_GATEWAY, ROUNDUP(sa->sa_len)); 496 if (sa = rt_mask(rt)) 497 next(RTA_NETMASK, ROUNDUP(sa->sa_len)); 498 if (sa = rt->rt_genmask) 499 next(RTA_GENMASK, ROUNDUP(sa->sa_len)); 500 w->w_needed += size; 501 if (w->w_where == NULL || w->w_needed > 0) 502 return 0; 503 w->w_rtm.rtm_msglen = size; 504 w->w_rtm.rtm_flags = rt->rt_flags; 505 w->w_rtm.rtm_use = rt->rt_use; 506 w->w_rtm.rtm_rmx = rt->rt_rmx; 507 w->w_rtm.rtm_index = rt->rt_ifp->if_index; 508 #undef next 509 #define next(l) {n = (l); Bcopy(sa, cp, n); cp += n;} 510 if (size <= sizeof(w->w_m)) { 511 register caddr_t cp = (caddr_t)(w->w_m.m_sabuf); 512 if (sa = rt_key(rt)) 513 next(ROUNDUP(sa->sa_len)); 514 if (sa = rt->rt_gateway) 515 next(ROUNDUP(sa->sa_len)); 516 if (sa = rt_mask(rt)) 517 next(ROUNDUP(sa->sa_len)); 518 if (sa = rt->rt_genmask) 519 next(ROUNDUP(sa->sa_len)); 520 #undef next 521 #define next(s, l) {n = (l); \ 522 if (error = copyout((caddr_t)(s), w->w_where, n)) return (error); \ 523 w->w_where += n;} 524 525 next(&w->w_m, size); /* Copy rtmsg and sockaddrs back */ 526 return 0; 527 } 528 next(&w->w_rtm, sizeof(w->w_rtm)); 529 if (sa = rt_key(rt)) 530 next(sa, ROUNDUP(sa->sa_len)); 531 if (sa = rt->rt_gateway) 532 next(sa, ROUNDUP(sa->sa_len)); 533 if (sa = rt_mask(rt)) 534 next(sa, ROUNDUP(sa->sa_len)); 535 if (sa = rt->rt_genmask) 536 next(sa, ROUNDUP(sa->sa_len)); 537 return (0); 538 #undef next 539 } 540 541 kinfo_rtable(op, where, given, arg, needed) 542 int op, arg; 543 caddr_t where; 544 int *given, *needed; 545 { 546 register struct radix_node_head *rnh; 547 int s, error = 0, i; 548 u_char af = ki_af(op); 549 struct walkarg w; 550 551 op &= 0xffff; 552 if (op != KINFO_RT_DUMP && op != KINFO_RT_FLAGS) 553 return (EINVAL); 554 555 Bzero(&w, sizeof(w)); 556 if ((w.w_where = where) && given) 557 w.w_given = *given; 558 w.w_needed = 0 - w.w_given; 559 w.w_arg = arg; 560 w.w_op = op; 561 w.w_rtm.rtm_version = RTM_VERSION; 562 w.w_rtm.rtm_type = RTM_GET; 563 564 s = splnet(); 565 for (i = 0; i <= AF_MAX; i++) { 566 if ((rnh = rt_tables[i]) == 0 || i == 0 || (af && af != i)) 567 continue; 568 error = rnh->rnh_walk(rnh->rnh_treetop, rt_dumpentry, &w); 569 if (error) 570 break; 571 } 572 w.w_needed += w.w_given; 573 if (where && given) 574 *given = w.w_where - where; 575 else 576 w.w_needed = (11 * w.w_needed) / 10; 577 *needed = w.w_needed; 578 splx(s); 579 return (error); 580 } 581 582 /* 583 * Definitions of protocols supported in the ROUTE domain. 584 */ 585 586 int raw_init(),raw_usrreq(),raw_input(),raw_ctlinput(); 587 extern struct domain routedomain; /* or at least forward */ 588 589 struct protosw routesw[] = { 590 { SOCK_RAW, &routedomain, 0, PR_ATOMIC|PR_ADDR, 591 raw_input, route_output, raw_ctlinput, 0, 592 route_usrreq, 593 raw_init, 0, 0, 0, 594 } 595 }; 596 597 int unp_externalize(), unp_dispose(); 598 599 struct domain routedomain = 600 { PF_ROUTE, "route", 0, 0, 0, 601 routesw, &routesw[sizeof(routesw)/sizeof(routesw[0])] }; 602