| 1 | /* |
| 2 | * Copyright (c) 1980, 1986, 1991, 1993 |
| 3 | * The Regents of the University of California. All rights reserved. |
| 4 | * |
| 5 | * %sccs.include.redist.c% |
| 6 | * |
| 7 | * @(#)route.c 8.1 (Berkeley) %G% |
| 8 | */ |
| 9 | |
| 10 | #include <sys/param.h> |
| 11 | #include <sys/systm.h> |
| 12 | #include <sys/proc.h> |
| 13 | #include <sys/mbuf.h> |
| 14 | #include <sys/socket.h> |
| 15 | #include <sys/socketvar.h> |
| 16 | #include <sys/domain.h> |
| 17 | #include <sys/protosw.h> |
| 18 | #include <sys/ioctl.h> |
| 19 | |
| 20 | #include <net/if.h> |
| 21 | #include <net/af.h> |
| 22 | #include <net/route.h> |
| 23 | #include <net/raw_cb.h> |
| 24 | |
| 25 | #include <netinet/in.h> |
| 26 | #include <netinet/in_var.h> |
| 27 | |
| 28 | #ifdef NS |
| 29 | #include <netns/ns.h> |
| 30 | #endif |
| 31 | |
| 32 | #define SA(p) ((struct sockaddr *)(p)) |
| 33 | |
| 34 | int rttrash; /* routes not in table but not freed */ |
| 35 | struct sockaddr wildcard; /* zero valued cookie for wildcard searches */ |
| 36 | |
| 37 | void |
| 38 | rtable_init(table) |
| 39 | void **table; |
| 40 | { |
| 41 | struct domain *dom; |
| 42 | for (dom = domains; dom; dom = dom->dom_next) |
| 43 | if (dom->dom_rtattach) |
| 44 | dom->dom_rtattach(&table[dom->dom_family], |
| 45 | dom->dom_rtoffset); |
| 46 | } |
| 47 | |
| 48 | void |
| 49 | route_init() |
| 50 | { |
| 51 | rn_init(); /* initialize all zeroes, all ones, mask table */ |
| 52 | rtable_init((void **)rt_tables); |
| 53 | } |
| 54 | |
| 55 | /* |
| 56 | * Packet routing routines. |
| 57 | */ |
| 58 | void |
| 59 | rtalloc(ro) |
| 60 | register struct route *ro; |
| 61 | { |
| 62 | if (ro->ro_rt && ro->ro_rt->rt_ifp && (ro->ro_rt->rt_flags & RTF_UP)) |
| 63 | return; /* XXX */ |
| 64 | ro->ro_rt = rtalloc1(&ro->ro_dst, 1); |
| 65 | } |
| 66 | |
| 67 | struct rtentry * |
| 68 | rtalloc1(dst, report) |
| 69 | register struct sockaddr *dst; |
| 70 | int report; |
| 71 | { |
| 72 | register struct radix_node_head *rnh = rt_tables[dst->sa_family]; |
| 73 | register struct rtentry *rt; |
| 74 | register struct radix_node *rn; |
| 75 | struct rtentry *newrt = 0; |
| 76 | struct rt_addrinfo info; |
| 77 | int s = splnet(), err = 0, msgtype = RTM_MISS; |
| 78 | |
| 79 | if (rnh && (rn = rnh->rnh_matchaddr((caddr_t)dst, rnh)) && |
| 80 | ((rn->rn_flags & RNF_ROOT) == 0)) { |
| 81 | newrt = rt = (struct rtentry *)rn; |
| 82 | if (report && (rt->rt_flags & RTF_CLONING)) { |
| 83 | err = rtrequest(RTM_RESOLVE, dst, SA(0), |
| 84 | SA(0), 0, &newrt); |
| 85 | if (err) { |
| 86 | newrt = rt; |
| 87 | rt->rt_refcnt++; |
| 88 | goto miss; |
| 89 | } |
| 90 | if ((rt = newrt) && (rt->rt_flags & RTF_XRESOLVE)) { |
| 91 | msgtype = RTM_RESOLVE; |
| 92 | goto miss; |
| 93 | } |
| 94 | } else |
| 95 | rt->rt_refcnt++; |
| 96 | } else { |
| 97 | rtstat.rts_unreach++; |
| 98 | miss: if (report) { |
| 99 | bzero((caddr_t)&info, sizeof(info)); |
| 100 | info.rti_info[RTAX_DST] = dst; |
| 101 | rt_missmsg(msgtype, &info, 0, err); |
| 102 | } |
| 103 | } |
| 104 | splx(s); |
| 105 | return (newrt); |
| 106 | } |
| 107 | |
| 108 | void |
| 109 | rtfree(rt) |
| 110 | register struct rtentry *rt; |
| 111 | { |
| 112 | register struct ifaddr *ifa; |
| 113 | |
| 114 | if (rt == 0) |
| 115 | panic("rtfree"); |
| 116 | rt->rt_refcnt--; |
| 117 | if (rt->rt_refcnt <= 0 && (rt->rt_flags & RTF_UP) == 0) { |
| 118 | if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT)) |
| 119 | panic ("rtfree 2"); |
| 120 | rttrash--; |
| 121 | if (rt->rt_refcnt < 0) { |
| 122 | printf("rtfree: %x not freed (neg refs)\n", rt); |
| 123 | return; |
| 124 | } |
| 125 | ifa = rt->rt_ifa; |
| 126 | IFAFREE(ifa); |
| 127 | Free(rt_key(rt)); |
| 128 | Free(rt); |
| 129 | } |
| 130 | } |
| 131 | |
| 132 | void |
| 133 | ifafree(ifa) |
| 134 | register struct ifaddr *ifa; |
| 135 | { |
| 136 | if (ifa == NULL) |
| 137 | panic("ifafree"); |
| 138 | if (ifa->ifa_refcnt == 0) |
| 139 | free(ifa, M_IFADDR); |
| 140 | else |
| 141 | ifa->ifa_refcnt--; |
| 142 | } |
| 143 | |
| 144 | /* |
| 145 | * Force a routing table entry to the specified |
| 146 | * destination to go through the given gateway. |
| 147 | * Normally called as a result of a routing redirect |
| 148 | * message from the network layer. |
| 149 | * |
| 150 | * N.B.: must be called at splnet |
| 151 | * |
| 152 | */ |
| 153 | int |
| 154 | rtredirect(dst, gateway, netmask, flags, src, rtp) |
| 155 | struct sockaddr *dst, *gateway, *netmask, *src; |
| 156 | int flags; |
| 157 | struct rtentry **rtp; |
| 158 | { |
| 159 | register struct rtentry *rt; |
| 160 | int error = 0; |
| 161 | short *stat = 0; |
| 162 | struct rt_addrinfo info; |
| 163 | struct ifaddr *ifa; |
| 164 | |
| 165 | /* verify the gateway is directly reachable */ |
| 166 | if ((ifa = ifa_ifwithnet(gateway)) == 0) { |
| 167 | error = ENETUNREACH; |
| 168 | goto out; |
| 169 | } |
| 170 | rt = rtalloc1(dst, 0); |
| 171 | /* |
| 172 | * If the redirect isn't from our current router for this dst, |
| 173 | * it's either old or wrong. If it redirects us to ourselves, |
| 174 | * we have a routing loop, perhaps as a result of an interface |
| 175 | * going down recently. |
| 176 | */ |
| 177 | #define equal(a1, a2) (bcmp((caddr_t)(a1), (caddr_t)(a2), (a1)->sa_len) == 0) |
| 178 | if (!(flags & RTF_DONE) && rt && |
| 179 | (!equal(src, rt->rt_gateway) || rt->rt_ifa != ifa)) |
| 180 | error = EINVAL; |
| 181 | else if (ifa_ifwithaddr(gateway)) |
| 182 | error = EHOSTUNREACH; |
| 183 | if (error) |
| 184 | goto done; |
| 185 | /* |
| 186 | * Create a new entry if we just got back a wildcard entry |
| 187 | * or the the lookup failed. This is necessary for hosts |
| 188 | * which use routing redirects generated by smart gateways |
| 189 | * to dynamically build the routing tables. |
| 190 | */ |
| 191 | if ((rt == 0) || (rt_mask(rt) && rt_mask(rt)->sa_len < 2)) |
| 192 | goto create; |
| 193 | /* |
| 194 | * Don't listen to the redirect if it's |
| 195 | * for a route to an interface. |
| 196 | */ |
| 197 | if (rt->rt_flags & RTF_GATEWAY) { |
| 198 | if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) { |
| 199 | /* |
| 200 | * Changing from route to net => route to host. |
| 201 | * Create new route, rather than smashing route to net. |
| 202 | */ |
| 203 | create: |
| 204 | flags |= RTF_GATEWAY | RTF_DYNAMIC; |
| 205 | error = rtrequest((int)RTM_ADD, dst, gateway, |
| 206 | netmask, flags, |
| 207 | (struct rtentry **)0); |
| 208 | stat = &rtstat.rts_dynamic; |
| 209 | } else { |
| 210 | /* |
| 211 | * Smash the current notion of the gateway to |
| 212 | * this destination. Should check about netmask!!! |
| 213 | */ |
| 214 | rt->rt_flags |= RTF_MODIFIED; |
| 215 | flags |= RTF_MODIFIED; |
| 216 | stat = &rtstat.rts_newgateway; |
| 217 | rt_setgate(rt, rt_key(rt), gateway); |
| 218 | } |
| 219 | } else |
| 220 | error = EHOSTUNREACH; |
| 221 | done: |
| 222 | if (rt) { |
| 223 | if (rtp && !error) |
| 224 | *rtp = rt; |
| 225 | else |
| 226 | rtfree(rt); |
| 227 | } |
| 228 | out: |
| 229 | if (error) |
| 230 | rtstat.rts_badredirect++; |
| 231 | else if (stat != NULL) |
| 232 | (*stat)++; |
| 233 | bzero((caddr_t)&info, sizeof(info)); |
| 234 | info.rti_info[RTAX_DST] = dst; |
| 235 | info.rti_info[RTAX_GATEWAY] = gateway; |
| 236 | info.rti_info[RTAX_NETMASK] = netmask; |
| 237 | info.rti_info[RTAX_AUTHOR] = src; |
| 238 | rt_missmsg(RTM_REDIRECT, &info, flags, error); |
| 239 | } |
| 240 | |
| 241 | /* |
| 242 | * Routing table ioctl interface. |
| 243 | */ |
| 244 | int |
| 245 | rtioctl(req, data, p) |
| 246 | int req; |
| 247 | caddr_t data; |
| 248 | struct proc *p; |
| 249 | { |
| 250 | return (EOPNOTSUPP); |
| 251 | } |
| 252 | |
| 253 | struct ifaddr * |
| 254 | ifa_ifwithroute(flags, dst, gateway) |
| 255 | int flags; |
| 256 | struct sockaddr *dst, *gateway; |
| 257 | { |
| 258 | register struct ifaddr *ifa; |
| 259 | if ((flags & RTF_GATEWAY) == 0) { |
| 260 | /* |
| 261 | * If we are adding a route to an interface, |
| 262 | * and the interface is a pt to pt link |
| 263 | * we should search for the destination |
| 264 | * as our clue to the interface. Otherwise |
| 265 | * we can use the local address. |
| 266 | */ |
| 267 | ifa = 0; |
| 268 | if (flags & RTF_HOST) |
| 269 | ifa = ifa_ifwithdstaddr(dst); |
| 270 | if (ifa == 0) |
| 271 | ifa = ifa_ifwithaddr(gateway); |
| 272 | } else { |
| 273 | /* |
| 274 | * If we are adding a route to a remote net |
| 275 | * or host, the gateway may still be on the |
| 276 | * other end of a pt to pt link. |
| 277 | */ |
| 278 | ifa = ifa_ifwithdstaddr(gateway); |
| 279 | } |
| 280 | if (ifa == 0) |
| 281 | ifa = ifa_ifwithnet(gateway); |
| 282 | if (ifa == 0) { |
| 283 | struct rtentry *rt = rtalloc1(dst, 0); |
| 284 | if (rt == 0) |
| 285 | return (0); |
| 286 | rt->rt_refcnt--; |
| 287 | if ((ifa = rt->rt_ifa) == 0) |
| 288 | return (0); |
| 289 | } |
| 290 | if (ifa->ifa_addr->sa_family != dst->sa_family) { |
| 291 | struct ifaddr *oifa = ifa; |
| 292 | ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp); |
| 293 | if (ifa == 0) |
| 294 | ifa = oifa; |
| 295 | } |
| 296 | return (ifa); |
| 297 | } |
| 298 | |
| 299 | #define ROUNDUP(a) (a>0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long)) |
| 300 | |
| 301 | int |
| 302 | rtrequest(req, dst, gateway, netmask, flags, ret_nrt) |
| 303 | int req, flags; |
| 304 | struct sockaddr *dst, *gateway, *netmask; |
| 305 | struct rtentry **ret_nrt; |
| 306 | { |
| 307 | int s = splnet(); int error = 0; |
| 308 | register struct rtentry *rt; |
| 309 | register struct radix_node *rn; |
| 310 | register struct radix_node_head *rnh; |
| 311 | struct ifaddr *ifa; |
| 312 | struct sockaddr *ndst; |
| 313 | #define senderr(x) { error = x ; goto bad; } |
| 314 | |
| 315 | if ((rnh = rt_tables[dst->sa_family]) == 0) |
| 316 | senderr(ESRCH); |
| 317 | if (flags & RTF_HOST) |
| 318 | netmask = 0; |
| 319 | switch (req) { |
| 320 | case RTM_DELETE: |
| 321 | if ((rn = rnh->rnh_deladdr(dst, netmask, rnh)) == 0) |
| 322 | senderr(ESRCH); |
| 323 | if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT)) |
| 324 | panic ("rtrequest delete"); |
| 325 | rt = (struct rtentry *)rn; |
| 326 | rt->rt_flags &= ~RTF_UP; |
| 327 | if (rt->rt_gwroute) { |
| 328 | rt = rt->rt_gwroute; RTFREE(rt); |
| 329 | (rt = (struct rtentry *)rn)->rt_gwroute = 0; |
| 330 | } |
| 331 | if ((ifa = rt->rt_ifa) && ifa->ifa_rtrequest) |
| 332 | ifa->ifa_rtrequest(RTM_DELETE, rt, SA(0)); |
| 333 | rttrash++; |
| 334 | if (ret_nrt) |
| 335 | *ret_nrt = rt; |
| 336 | else if (rt->rt_refcnt <= 0) { |
| 337 | rt->rt_refcnt++; |
| 338 | rtfree(rt); |
| 339 | } |
| 340 | break; |
| 341 | |
| 342 | case RTM_RESOLVE: |
| 343 | if (ret_nrt == 0 || (rt = *ret_nrt) == 0) |
| 344 | senderr(EINVAL); |
| 345 | ifa = rt->rt_ifa; |
| 346 | flags = rt->rt_flags & ~RTF_CLONING; |
| 347 | gateway = rt->rt_gateway; |
| 348 | if ((netmask = rt->rt_genmask) == 0) |
| 349 | flags |= RTF_HOST; |
| 350 | goto makeroute; |
| 351 | |
| 352 | case RTM_ADD: |
| 353 | if ((ifa = ifa_ifwithroute(flags, dst, gateway)) == 0) |
| 354 | senderr(ENETUNREACH); |
| 355 | makeroute: |
| 356 | R_Malloc(rt, struct rtentry *, sizeof(*rt)); |
| 357 | if (rt == 0) |
| 358 | senderr(ENOBUFS); |
| 359 | Bzero(rt, sizeof(*rt)); |
| 360 | rt->rt_flags = RTF_UP | flags; |
| 361 | if (rt_setgate(rt, dst, gateway)) { |
| 362 | Free(rt); |
| 363 | senderr(ENOBUFS); |
| 364 | } |
| 365 | ndst = rt_key(rt); |
| 366 | if (netmask) { |
| 367 | rt_maskedcopy(dst, ndst, netmask); |
| 368 | } else |
| 369 | Bcopy(dst, ndst, dst->sa_len); |
| 370 | rn = rnh->rnh_addaddr((caddr_t)ndst, (caddr_t)netmask, |
| 371 | rnh, rt->rt_nodes); |
| 372 | if (rn == 0) { |
| 373 | if (rt->rt_gwroute) |
| 374 | rtfree(rt->rt_gwroute); |
| 375 | Free(rt_key(rt)); |
| 376 | Free(rt); |
| 377 | senderr(EEXIST); |
| 378 | } |
| 379 | ifa->ifa_refcnt++; |
| 380 | rt->rt_ifa = ifa; |
| 381 | rt->rt_ifp = ifa->ifa_ifp; |
| 382 | if (req == RTM_RESOLVE) |
| 383 | rt->rt_rmx = (*ret_nrt)->rt_rmx; /* copy metrics */ |
| 384 | if (ifa->ifa_rtrequest) |
| 385 | ifa->ifa_rtrequest(req, rt, SA(ret_nrt ? *ret_nrt : 0)); |
| 386 | if (ret_nrt) { |
| 387 | *ret_nrt = rt; |
| 388 | rt->rt_refcnt++; |
| 389 | } |
| 390 | break; |
| 391 | } |
| 392 | bad: |
| 393 | splx(s); |
| 394 | return (error); |
| 395 | } |
| 396 | |
| 397 | int |
| 398 | rt_setgate(rt0, dst, gate) |
| 399 | struct rtentry *rt0; |
| 400 | struct sockaddr *dst, *gate; |
| 401 | { |
| 402 | caddr_t new, old; |
| 403 | int dlen = ROUNDUP(dst->sa_len), glen = ROUNDUP(gate->sa_len); |
| 404 | register struct rtentry *rt = rt0; |
| 405 | |
| 406 | if (rt->rt_gateway == 0 || glen > ROUNDUP(rt->rt_gateway->sa_len)) { |
| 407 | old = (caddr_t)rt_key(rt); |
| 408 | R_Malloc(new, caddr_t, dlen + glen); |
| 409 | if (new == 0) |
| 410 | return 1; |
| 411 | rt->rt_nodes->rn_key = new; |
| 412 | } else { |
| 413 | new = rt->rt_nodes->rn_key; |
| 414 | old = 0; |
| 415 | } |
| 416 | Bcopy(gate, (rt->rt_gateway = (struct sockaddr *)(new + dlen)), glen); |
| 417 | if (old) { |
| 418 | Bcopy(dst, new, dlen); |
| 419 | Free(old); |
| 420 | } |
| 421 | if (rt->rt_gwroute) { |
| 422 | rt = rt->rt_gwroute; RTFREE(rt); |
| 423 | rt = rt0; rt->rt_gwroute = 0; |
| 424 | } |
| 425 | if (rt->rt_flags & RTF_GATEWAY) { |
| 426 | rt->rt_gwroute = rtalloc1(gate, 1); |
| 427 | } |
| 428 | return 0; |
| 429 | } |
| 430 | |
| 431 | void |
| 432 | rt_maskedcopy(src, dst, netmask) |
| 433 | struct sockaddr *src, *dst, *netmask; |
| 434 | { |
| 435 | register u_char *cp1 = (u_char *)src; |
| 436 | register u_char *cp2 = (u_char *)dst; |
| 437 | register u_char *cp3 = (u_char *)netmask; |
| 438 | u_char *cplim = cp2 + *cp3; |
| 439 | u_char *cplim2 = cp2 + *cp1; |
| 440 | |
| 441 | *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */ |
| 442 | cp3 += 2; |
| 443 | if (cplim > cplim2) |
| 444 | cplim = cplim2; |
| 445 | while (cp2 < cplim) |
| 446 | *cp2++ = *cp1++ & *cp3++; |
| 447 | if (cp2 < cplim2) |
| 448 | bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2)); |
| 449 | } |
| 450 | |
| 451 | /* |
| 452 | * Set up a routing table entry, normally |
| 453 | * for an interface. |
| 454 | */ |
| 455 | int |
| 456 | rtinit(ifa, cmd, flags) |
| 457 | register struct ifaddr *ifa; |
| 458 | int cmd, flags; |
| 459 | { |
| 460 | register struct rtentry *rt; |
| 461 | register struct sockaddr *dst; |
| 462 | register struct sockaddr *deldst; |
| 463 | struct mbuf *m = 0; |
| 464 | struct rtentry *nrt = 0; |
| 465 | int error; |
| 466 | |
| 467 | dst = flags & RTF_HOST ? ifa->ifa_dstaddr : ifa->ifa_addr; |
| 468 | if (cmd == RTM_DELETE) { |
| 469 | if ((flags & RTF_HOST) == 0 && ifa->ifa_netmask) { |
| 470 | m = m_get(M_WAIT, MT_SONAME); |
| 471 | deldst = mtod(m, struct sockaddr *); |
| 472 | rt_maskedcopy(dst, deldst, ifa->ifa_netmask); |
| 473 | dst = deldst; |
| 474 | } |
| 475 | if (rt = rtalloc1(dst, 0)) { |
| 476 | rt->rt_refcnt--; |
| 477 | if (rt->rt_ifa != ifa) { |
| 478 | if (m) |
| 479 | (void) m_free(m); |
| 480 | return (flags & RTF_HOST ? EHOSTUNREACH |
| 481 | : ENETUNREACH); |
| 482 | } |
| 483 | } |
| 484 | } |
| 485 | error = rtrequest(cmd, dst, ifa->ifa_addr, ifa->ifa_netmask, |
| 486 | flags | ifa->ifa_flags, &nrt); |
| 487 | if (m) |
| 488 | (void) m_free(m); |
| 489 | if (cmd == RTM_DELETE && error == 0 && (rt = nrt)) { |
| 490 | rt_newaddrmsg(cmd, ifa, error, nrt); |
| 491 | if (rt->rt_refcnt <= 0) { |
| 492 | rt->rt_refcnt++; |
| 493 | rtfree(rt); |
| 494 | } |
| 495 | } |
| 496 | if (cmd == RTM_ADD && error == 0 && (rt = nrt)) { |
| 497 | rt->rt_refcnt--; |
| 498 | if (rt->rt_ifa != ifa) { |
| 499 | printf("rtinit: wrong ifa (%x) was (%x)\n", ifa, |
| 500 | rt->rt_ifa); |
| 501 | if (rt->rt_ifa->ifa_rtrequest) |
| 502 | rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, SA(0)); |
| 503 | IFAFREE(rt->rt_ifa); |
| 504 | rt->rt_ifa = ifa; |
| 505 | rt->rt_ifp = ifa->ifa_ifp; |
| 506 | ifa->ifa_refcnt++; |
| 507 | if (ifa->ifa_rtrequest) |
| 508 | ifa->ifa_rtrequest(RTM_ADD, rt, SA(0)); |
| 509 | } |
| 510 | rt_newaddrmsg(cmd, ifa, error, nrt); |
| 511 | } |
| 512 | return (error); |
| 513 | } |