1 /*- 2 * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project. 3 * All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 3. Neither the name of the project nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * $KAME: in6_rmx.c,v 1.11 2001/07/26 06:53:16 jinmei Exp $ 30 */ 31 32 /*- 33 * Copyright 1994, 1995 Massachusetts Institute of Technology 34 * 35 * Permission to use, copy, modify, and distribute this software and 36 * its documentation for any purpose and without fee is hereby 37 * granted, provided that both the above copyright notice and this 38 * permission notice appear in all copies, that both the above 39 * copyright notice and this permission notice appear in all 40 * supporting documentation, and that the name of M.I.T. not be used 41 * in advertising or publicity pertaining to distribution of the 42 * software without specific, written prior permission. M.I.T. makes 43 * no representations about the suitability of this software for any 44 * purpose. It is provided "as is" without express or implied 45 * warranty. 46 * 47 * THIS SOFTWARE IS PROVIDED BY M.I.T. ``AS IS''. M.I.T. DISCLAIMS 48 * ALL EXPRESS OR IMPLIED WARRANTIES WITH REGARD TO THIS SOFTWARE, 49 * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF 50 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT 51 * SHALL M.I.T. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 52 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 53 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF 54 * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND 55 * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, 56 * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT 57 * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 58 * SUCH DAMAGE. 59 * 60 */ 61 62 /* 63 * This code does two things necessary for the enhanced TCP metrics to 64 * function in a useful manner: 65 * 1) It marks all non-host routes as `cloning', thus ensuring that 66 * every actual reference to such a route actually gets turned 67 * into a reference to a host route to the specific destination 68 * requested. 69 * 2) When such routes lose all their references, it arranges for them 70 * to be deleted in some random collection of circumstances, so that 71 * a large quantity of stale routing data is not kept in kernel memory 72 * indefinitely. See in6_rtqtimo() below for the exact mechanism. 73 */ 74 75 #include <sys/cdefs.h> 76 __FBSDID("$FreeBSD$"); 77 78 #include <sys/param.h> 79 #include <sys/systm.h> 80 #include <sys/kernel.h> 81 #include <sys/sysctl.h> 82 #include <sys/queue.h> 83 #include <sys/socket.h> 84 #include <sys/socketvar.h> 85 #include <sys/mbuf.h> 86 #include <sys/syslog.h> 87 #include <sys/callout.h> 88 #include <sys/vimage.h> 89 90 #include <net/if.h> 91 #include <net/route.h> 92 #include <netinet/in.h> 93 #include <netinet/ip_var.h> 94 #include <netinet/in_var.h> 95 96 #include <netinet/ip6.h> 97 #include <netinet6/ip6_var.h> 98 99 #include <netinet/icmp6.h> 100 #include <netinet6/nd6.h> 101 102 #include <netinet/tcp.h> 103 #include <netinet/tcp_seq.h> 104 #include <netinet/tcp_timer.h> 105 #include <netinet/tcp_var.h> 106 107 extern int in6_inithead(void **head, int off); 108 109 #define RTPRF_OURS RTF_PROTO3 /* set on routes we manage */ 110 111 /* 112 * Do what we need to do when inserting a route. 113 */ 114 static struct radix_node * 115 in6_addroute(void *v_arg, void *n_arg, struct radix_node_head *head, 116 struct radix_node *treenodes) 117 { 118 struct rtentry *rt = (struct rtentry *)treenodes; 119 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)rt_key(rt); 120 struct radix_node *ret; 121 122 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 123 rt->rt_flags |= RTF_MULTICAST; 124 125 /* 126 * A little bit of help for both IPv6 output and input: 127 * For local addresses, we make sure that RTF_LOCAL is set, 128 * with the thought that this might one day be used to speed up 129 * ip_input(). 130 * 131 * We also mark routes to multicast addresses as such, because 132 * it's easy to do and might be useful (but this is much more 133 * dubious since it's so easy to inspect the address). (This 134 * is done above.) 135 * 136 * XXX 137 * should elaborate the code. 138 */ 139 if (rt->rt_flags & RTF_HOST) { 140 if (IN6_ARE_ADDR_EQUAL(&satosin6(rt->rt_ifa->ifa_addr) 141 ->sin6_addr, 142 &sin6->sin6_addr)) { 143 rt->rt_flags |= RTF_LOCAL; 144 } 145 } 146 147 if (!rt->rt_rmx.rmx_mtu && rt->rt_ifp) 148 rt->rt_rmx.rmx_mtu = IN6_LINKMTU(rt->rt_ifp); 149 150 ret = rn_addroute(v_arg, n_arg, head, treenodes); 151 if (ret == NULL && rt->rt_flags & RTF_HOST) { 152 struct rtentry *rt2; 153 /* 154 * We are trying to add a host route, but can't. 155 * Find out if it is because of an 156 * ARP entry and delete it if so. 157 */ 158 rt2 = rtalloc1((struct sockaddr *)sin6, 0, RTF_CLONING); 159 if (rt2) { 160 if (rt2->rt_flags & RTF_LLINFO && 161 rt2->rt_flags & RTF_HOST && 162 rt2->rt_gateway && 163 rt2->rt_gateway->sa_family == AF_LINK) { 164 rtexpunge(rt2); 165 RTFREE_LOCKED(rt2); 166 ret = rn_addroute(v_arg, n_arg, head, 167 treenodes); 168 } else 169 RTFREE_LOCKED(rt2); 170 } 171 } else if (ret == NULL && rt->rt_flags & RTF_CLONING) { 172 struct rtentry *rt2; 173 /* 174 * We are trying to add a net route, but can't. 175 * The following case should be allowed, so we'll make a 176 * special check for this: 177 * Two IPv6 addresses with the same prefix is assigned 178 * to a single interrface. 179 * # ifconfig if0 inet6 3ffe:0501::1 prefix 64 alias (*1) 180 * # ifconfig if0 inet6 3ffe:0501::2 prefix 64 alias (*2) 181 * In this case, (*1) and (*2) want to add the same 182 * net route entry, 3ffe:0501:: -> if0. 183 * This case should not raise an error. 184 */ 185 rt2 = rtalloc1((struct sockaddr *)sin6, 0, RTF_CLONING); 186 if (rt2) { 187 if ((rt2->rt_flags & (RTF_CLONING|RTF_HOST|RTF_GATEWAY)) 188 == RTF_CLONING 189 && rt2->rt_gateway 190 && rt2->rt_gateway->sa_family == AF_LINK 191 && rt2->rt_ifp == rt->rt_ifp) { 192 ret = rt2->rt_nodes; 193 } 194 RTFREE_LOCKED(rt2); 195 } 196 } 197 return ret; 198 } 199 200 /* 201 * This code is the inverse of in6_clsroute: on first reference, if we 202 * were managing the route, stop doing so and set the expiration timer 203 * back off again. 204 */ 205 static struct radix_node * 206 in6_matroute(void *v_arg, struct radix_node_head *head) 207 { 208 struct radix_node *rn = rn_match(v_arg, head); 209 struct rtentry *rt = (struct rtentry *)rn; 210 211 if (rt && rt->rt_refcnt == 0) { /* this is first reference */ 212 if (rt->rt_flags & RTPRF_OURS) { 213 rt->rt_flags &= ~RTPRF_OURS; 214 rt->rt_rmx.rmx_expire = 0; 215 } 216 } 217 return rn; 218 } 219 220 SYSCTL_DECL(_net_inet6_ip6); 221 222 static int rtq_reallyold6 = 60*60; 223 /* one hour is ``really old'' */ 224 SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTEXPIRE, rtexpire, 225 CTLFLAG_RW, &rtq_reallyold6 , 0, ""); 226 227 static int rtq_minreallyold6 = 10; 228 /* never automatically crank down to less */ 229 SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTMINEXPIRE, rtminexpire, 230 CTLFLAG_RW, &rtq_minreallyold6 , 0, ""); 231 232 static int rtq_toomany6 = 128; 233 /* 128 cached routes is ``too many'' */ 234 SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTMAXCACHE, rtmaxcache, 235 CTLFLAG_RW, &rtq_toomany6 , 0, ""); 236 237 238 /* 239 * On last reference drop, mark the route as belong to us so that it can be 240 * timed out. 241 */ 242 static void 243 in6_clsroute(struct radix_node *rn, struct radix_node_head *head) 244 { 245 INIT_VNET_INET6(curvnet); 246 struct rtentry *rt = (struct rtentry *)rn; 247 248 RT_LOCK_ASSERT(rt); 249 250 if (!(rt->rt_flags & RTF_UP)) 251 return; /* prophylactic measures */ 252 253 if ((rt->rt_flags & (RTF_LLINFO | RTF_HOST)) != RTF_HOST) 254 return; 255 256 if ((rt->rt_flags & (RTF_WASCLONED | RTPRF_OURS)) != RTF_WASCLONED) 257 return; 258 259 /* 260 * As requested by David Greenman: 261 * If rtq_reallyold6 is 0, just delete the route without 262 * waiting for a timeout cycle to kill it. 263 */ 264 if (V_rtq_reallyold6 != 0) { 265 rt->rt_flags |= RTPRF_OURS; 266 rt->rt_rmx.rmx_expire = time_uptime + V_rtq_reallyold6; 267 } else { 268 rtexpunge(rt); 269 } 270 } 271 272 struct rtqk_arg { 273 struct radix_node_head *rnh; 274 int mode; 275 int updating; 276 int draining; 277 int killed; 278 int found; 279 time_t nextstop; 280 }; 281 282 /* 283 * Get rid of old routes. When draining, this deletes everything, even when 284 * the timeout is not expired yet. When updating, this makes sure that 285 * nothing has a timeout longer than the current value of rtq_reallyold6. 286 */ 287 static int 288 in6_rtqkill(struct radix_node *rn, void *rock) 289 { 290 INIT_VNET_INET6(curvnet); 291 struct rtqk_arg *ap = rock; 292 struct rtentry *rt = (struct rtentry *)rn; 293 int err; 294 295 if (rt->rt_flags & RTPRF_OURS) { 296 ap->found++; 297 298 if (ap->draining || rt->rt_rmx.rmx_expire <= time_uptime) { 299 if (rt->rt_refcnt > 0) 300 panic("rtqkill route really not free"); 301 302 err = rtrequest(RTM_DELETE, 303 (struct sockaddr *)rt_key(rt), 304 rt->rt_gateway, rt_mask(rt), 305 rt->rt_flags, 0); 306 if (err) { 307 log(LOG_WARNING, "in6_rtqkill: error %d", err); 308 } else { 309 ap->killed++; 310 } 311 } else { 312 if (ap->updating 313 && (rt->rt_rmx.rmx_expire - time_uptime 314 > V_rtq_reallyold6)) { 315 rt->rt_rmx.rmx_expire = time_uptime 316 + V_rtq_reallyold6; 317 } 318 ap->nextstop = lmin(ap->nextstop, 319 rt->rt_rmx.rmx_expire); 320 } 321 } 322 323 return 0; 324 } 325 326 #define RTQ_TIMEOUT 60*10 /* run no less than once every ten minutes */ 327 static int rtq_timeout6 = RTQ_TIMEOUT; 328 static struct callout rtq_timer6; 329 330 static void 331 in6_rtqtimo(void *rock) 332 { 333 CURVNET_SET_QUIET((struct vnet *) rock); 334 INIT_VNET_NET((struct vnet *) rock); 335 INIT_VNET_INET6((struct vnet *) rock); 336 struct radix_node_head *rnh = rock; 337 struct rtqk_arg arg; 338 struct timeval atv; 339 static time_t last_adjusted_timeout = 0; 340 341 arg.found = arg.killed = 0; 342 arg.rnh = rnh; 343 arg.nextstop = time_uptime + V_rtq_timeout6; 344 arg.draining = arg.updating = 0; 345 RADIX_NODE_HEAD_LOCK(rnh); 346 rnh->rnh_walktree(rnh, in6_rtqkill, &arg); 347 RADIX_NODE_HEAD_UNLOCK(rnh); 348 349 /* 350 * Attempt to be somewhat dynamic about this: 351 * If there are ``too many'' routes sitting around taking up space, 352 * then crank down the timeout, and see if we can't make some more 353 * go away. However, we make sure that we will never adjust more 354 * than once in rtq_timeout6 seconds, to keep from cranking down too 355 * hard. 356 */ 357 if ((arg.found - arg.killed > V_rtq_toomany6) 358 && (time_uptime - last_adjusted_timeout >= V_rtq_timeout6) 359 && V_rtq_reallyold6 > V_rtq_minreallyold6) { 360 V_rtq_reallyold6 = 2*V_rtq_reallyold6 / 3; 361 if (V_rtq_reallyold6 < V_rtq_minreallyold6) { 362 V_rtq_reallyold6 = V_rtq_minreallyold6; 363 } 364 365 last_adjusted_timeout = time_uptime; 366 #ifdef DIAGNOSTIC 367 log(LOG_DEBUG, "in6_rtqtimo: adjusted rtq_reallyold6 to %d", 368 V_rtq_reallyold6); 369 #endif 370 arg.found = arg.killed = 0; 371 arg.updating = 1; 372 RADIX_NODE_HEAD_LOCK(rnh); 373 rnh->rnh_walktree(rnh, in6_rtqkill, &arg); 374 RADIX_NODE_HEAD_UNLOCK(rnh); 375 } 376 377 atv.tv_usec = 0; 378 atv.tv_sec = arg.nextstop - time_uptime; 379 callout_reset(&V_rtq_timer6, tvtohz(&atv), in6_rtqtimo, rock); 380 CURVNET_RESTORE(); 381 } 382 383 /* 384 * Age old PMTUs. 385 */ 386 struct mtuex_arg { 387 struct radix_node_head *rnh; 388 time_t nextstop; 389 }; 390 static struct callout rtq_mtutimer; 391 392 static int 393 in6_mtuexpire(struct radix_node *rn, void *rock) 394 { 395 struct rtentry *rt = (struct rtentry *)rn; 396 struct mtuex_arg *ap = rock; 397 398 /* sanity */ 399 if (!rt) 400 panic("rt == NULL in in6_mtuexpire"); 401 402 if (rt->rt_rmx.rmx_expire && !(rt->rt_flags & RTF_PROBEMTU)) { 403 if (rt->rt_rmx.rmx_expire <= time_uptime) { 404 rt->rt_flags |= RTF_PROBEMTU; 405 } else { 406 ap->nextstop = lmin(ap->nextstop, 407 rt->rt_rmx.rmx_expire); 408 } 409 } 410 411 return 0; 412 } 413 414 #define MTUTIMO_DEFAULT (60*1) 415 416 static void 417 in6_mtutimo(void *rock) 418 { 419 CURVNET_SET_QUIET((struct vnet *) rock); 420 INIT_VNET_NET((struct vnet *) rock); 421 INIT_VNET_INET6((struct vnet *) rock); 422 struct radix_node_head *rnh = rock; 423 struct mtuex_arg arg; 424 struct timeval atv; 425 426 arg.rnh = rnh; 427 arg.nextstop = time_uptime + MTUTIMO_DEFAULT; 428 RADIX_NODE_HEAD_LOCK(rnh); 429 rnh->rnh_walktree(rnh, in6_mtuexpire, &arg); 430 RADIX_NODE_HEAD_UNLOCK(rnh); 431 432 atv.tv_usec = 0; 433 atv.tv_sec = arg.nextstop - time_uptime; 434 if (atv.tv_sec < 0) { 435 printf("invalid mtu expiration time on routing table\n"); 436 arg.nextstop = time_uptime + 30; /* last resort */ 437 atv.tv_sec = 30; 438 } 439 callout_reset(&V_rtq_mtutimer, tvtohz(&atv), in6_mtutimo, rock); 440 CURVNET_RESTORE(); 441 } 442 443 #if 0 444 void 445 in6_rtqdrain(void) 446 { 447 INIT_VNET_NET(curvnet); 448 struct radix_node_head *rnh = V_rt_tables[AF_INET6]; 449 struct rtqk_arg arg; 450 451 arg.found = arg.killed = 0; 452 arg.rnh = rnh; 453 arg.nextstop = 0; 454 arg.draining = 1; 455 arg.updating = 0; 456 RADIX_NODE_HEAD_LOCK(rnh); 457 rnh->rnh_walktree(rnh, in6_rtqkill, &arg); 458 RADIX_NODE_HEAD_UNLOCK(rnh); 459 } 460 #endif 461 462 /* 463 * Initialize our routing tree. 464 * XXX MRT When off == 0, we are being called from vfs_export.c 465 * so just set up their table and leave. (we know what the correct 466 * value should be so just use that).. FIX AFTER RELENG_7 is MFC'd 467 * see also comments in in_inithead() vfs_export.c and domain.h 468 */ 469 int 470 in6_inithead(void **head, int off) 471 { 472 INIT_VNET_INET6(curvnet); 473 struct radix_node_head *rnh; 474 475 if (!rn_inithead(head, offsetof(struct sockaddr_in6, sin6_addr) << 3)) 476 return 0; /* See above */ 477 478 if (off == 0) /* See above */ 479 return 1; /* only do the rest for the real thing */ 480 481 rnh = *head; 482 rnh->rnh_addaddr = in6_addroute; 483 rnh->rnh_matchaddr = in6_matroute; 484 rnh->rnh_close = in6_clsroute; 485 callout_init(&V_rtq_timer6, CALLOUT_MPSAFE); 486 in6_rtqtimo(rnh); /* kick off timeout first time */ 487 callout_init(&V_rtq_mtutimer, CALLOUT_MPSAFE); 488 in6_mtutimo(rnh); /* kick off timeout first time */ 489 return 1; 490 } 491