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/lock.h> 82 #include <sys/sysctl.h> 83 #include <sys/queue.h> 84 #include <sys/socket.h> 85 #include <sys/socketvar.h> 86 #include <sys/mbuf.h> 87 #include <sys/rwlock.h> 88 #include <sys/syslog.h> 89 #include <sys/callout.h> 90 91 #include <net/if.h> 92 #include <net/route.h> 93 94 #include <netinet/in.h> 95 #include <netinet/ip_var.h> 96 #include <netinet/in_var.h> 97 98 #include <netinet/ip6.h> 99 #include <netinet6/ip6_var.h> 100 101 #include <netinet/icmp6.h> 102 #include <netinet6/nd6.h> 103 104 #include <netinet/tcp.h> 105 #include <netinet/tcp_seq.h> 106 #include <netinet/tcp_timer.h> 107 #include <netinet/tcp_var.h> 108 109 extern int in6_inithead(void **head, int off); 110 #ifdef VIMAGE 111 extern int in6_detachhead(void **head, int off); 112 #endif 113 114 #define RTPRF_OURS RTF_PROTO3 /* set on routes we manage */ 115 116 /* 117 * Do what we need to do when inserting a route. 118 */ 119 static struct radix_node * 120 in6_addroute(void *v_arg, void *n_arg, struct radix_node_head *head, 121 struct radix_node *treenodes) 122 { 123 struct rtentry *rt = (struct rtentry *)treenodes; 124 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)rt_key(rt); 125 struct radix_node *ret; 126 127 RADIX_NODE_HEAD_WLOCK_ASSERT(head); 128 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 129 rt->rt_flags |= RTF_MULTICAST; 130 131 /* 132 * A little bit of help for both IPv6 output and input: 133 * For local addresses, we make sure that RTF_LOCAL is set, 134 * with the thought that this might one day be used to speed up 135 * ip_input(). 136 * 137 * We also mark routes to multicast addresses as such, because 138 * it's easy to do and might be useful (but this is much more 139 * dubious since it's so easy to inspect the address). (This 140 * is done above.) 141 * 142 * XXX 143 * should elaborate the code. 144 */ 145 if (rt->rt_flags & RTF_HOST) { 146 if (IN6_ARE_ADDR_EQUAL(&satosin6(rt->rt_ifa->ifa_addr) 147 ->sin6_addr, 148 &sin6->sin6_addr)) { 149 rt->rt_flags |= RTF_LOCAL; 150 } 151 } 152 153 if (!rt->rt_rmx.rmx_mtu && rt->rt_ifp) 154 rt->rt_rmx.rmx_mtu = IN6_LINKMTU(rt->rt_ifp); 155 156 ret = rn_addroute(v_arg, n_arg, head, treenodes); 157 if (ret == NULL) { 158 struct rtentry *rt2; 159 /* 160 * We are trying to add a net route, but can't. 161 * The following case should be allowed, so we'll make a 162 * special check for this: 163 * Two IPv6 addresses with the same prefix is assigned 164 * to a single interrface. 165 * # ifconfig if0 inet6 3ffe:0501::1 prefix 64 alias (*1) 166 * # ifconfig if0 inet6 3ffe:0501::2 prefix 64 alias (*2) 167 * In this case, (*1) and (*2) want to add the same 168 * net route entry, 3ffe:0501:: -> if0. 169 * This case should not raise an error. 170 */ 171 rt2 = rtalloc1((struct sockaddr *)sin6, 0, RTF_RNH_LOCKED); 172 if (rt2) { 173 if (((rt2->rt_flags & (RTF_HOST|RTF_GATEWAY)) == 0) 174 && rt2->rt_gateway 175 && rt2->rt_gateway->sa_family == AF_LINK 176 && rt2->rt_ifp == rt->rt_ifp) { 177 ret = rt2->rt_nodes; 178 } 179 RTFREE_LOCKED(rt2); 180 } 181 } 182 return (ret); 183 } 184 185 /* 186 * This code is the inverse of in6_clsroute: on first reference, if we 187 * were managing the route, stop doing so and set the expiration timer 188 * back off again. 189 */ 190 static struct radix_node * 191 in6_matroute(void *v_arg, struct radix_node_head *head) 192 { 193 struct radix_node *rn = rn_match(v_arg, head); 194 struct rtentry *rt = (struct rtentry *)rn; 195 196 if (rt && rt->rt_refcnt == 0) { /* this is first reference */ 197 if (rt->rt_flags & RTPRF_OURS) { 198 rt->rt_flags &= ~RTPRF_OURS; 199 rt->rt_rmx.rmx_expire = 0; 200 } 201 } 202 return rn; 203 } 204 205 SYSCTL_DECL(_net_inet6_ip6); 206 207 static VNET_DEFINE(int, rtq_reallyold6); 208 static VNET_DEFINE(int, rtq_minreallyold6); 209 static VNET_DEFINE(int, rtq_toomany6); 210 211 #define V_rtq_reallyold6 VNET(rtq_reallyold6) 212 #define V_rtq_minreallyold6 VNET(rtq_minreallyold6) 213 #define V_rtq_toomany6 VNET(rtq_toomany6) 214 215 SYSCTL_VNET_INT(_net_inet6_ip6, IPV6CTL_RTEXPIRE, rtexpire, CTLFLAG_RW, 216 &VNET_NAME(rtq_reallyold6) , 0, ""); 217 218 SYSCTL_VNET_INT(_net_inet6_ip6, IPV6CTL_RTMINEXPIRE, rtminexpire, CTLFLAG_RW, 219 &VNET_NAME(rtq_minreallyold6) , 0, ""); 220 221 SYSCTL_VNET_INT(_net_inet6_ip6, IPV6CTL_RTMAXCACHE, rtmaxcache, CTLFLAG_RW, 222 &VNET_NAME(rtq_toomany6) , 0, ""); 223 224 struct rtqk_arg { 225 struct radix_node_head *rnh; 226 int mode; 227 int updating; 228 int draining; 229 int killed; 230 int found; 231 time_t nextstop; 232 }; 233 234 /* 235 * Get rid of old routes. When draining, this deletes everything, even when 236 * the timeout is not expired yet. When updating, this makes sure that 237 * nothing has a timeout longer than the current value of rtq_reallyold6. 238 */ 239 static int 240 in6_rtqkill(struct radix_node *rn, void *rock) 241 { 242 struct rtqk_arg *ap = rock; 243 struct rtentry *rt = (struct rtentry *)rn; 244 int err; 245 246 RADIX_NODE_HEAD_WLOCK_ASSERT(ap->rnh); 247 248 if (rt->rt_flags & RTPRF_OURS) { 249 ap->found++; 250 251 if (ap->draining || rt->rt_rmx.rmx_expire <= time_uptime) { 252 if (rt->rt_refcnt > 0) 253 panic("rtqkill route really not free"); 254 255 err = rtrequest(RTM_DELETE, 256 (struct sockaddr *)rt_key(rt), 257 rt->rt_gateway, rt_mask(rt), 258 rt->rt_flags|RTF_RNH_LOCKED, 0); 259 if (err) { 260 log(LOG_WARNING, "in6_rtqkill: error %d", err); 261 } else { 262 ap->killed++; 263 } 264 } else { 265 if (ap->updating 266 && (rt->rt_rmx.rmx_expire - time_uptime 267 > V_rtq_reallyold6)) { 268 rt->rt_rmx.rmx_expire = time_uptime 269 + V_rtq_reallyold6; 270 } 271 ap->nextstop = lmin(ap->nextstop, 272 rt->rt_rmx.rmx_expire); 273 } 274 } 275 276 return 0; 277 } 278 279 #define RTQ_TIMEOUT 60*10 /* run no less than once every ten minutes */ 280 static VNET_DEFINE(int, rtq_timeout6); 281 static VNET_DEFINE(struct callout, rtq_timer6); 282 283 #define V_rtq_timeout6 VNET(rtq_timeout6) 284 #define V_rtq_timer6 VNET(rtq_timer6) 285 286 static void 287 in6_rtqtimo(void *rock) 288 { 289 CURVNET_SET_QUIET((struct vnet *) rock); 290 struct radix_node_head *rnh; 291 struct rtqk_arg arg; 292 struct timeval atv; 293 static time_t last_adjusted_timeout = 0; 294 295 rnh = rt_tables_get_rnh(0, AF_INET6); 296 if (rnh == NULL) { 297 CURVNET_RESTORE(); 298 return; 299 } 300 arg.found = arg.killed = 0; 301 arg.rnh = rnh; 302 arg.nextstop = time_uptime + V_rtq_timeout6; 303 arg.draining = arg.updating = 0; 304 RADIX_NODE_HEAD_LOCK(rnh); 305 rnh->rnh_walktree(rnh, in6_rtqkill, &arg); 306 RADIX_NODE_HEAD_UNLOCK(rnh); 307 308 /* 309 * Attempt to be somewhat dynamic about this: 310 * If there are ``too many'' routes sitting around taking up space, 311 * then crank down the timeout, and see if we can't make some more 312 * go away. However, we make sure that we will never adjust more 313 * than once in rtq_timeout6 seconds, to keep from cranking down too 314 * hard. 315 */ 316 if ((arg.found - arg.killed > V_rtq_toomany6) 317 && (time_uptime - last_adjusted_timeout >= V_rtq_timeout6) 318 && V_rtq_reallyold6 > V_rtq_minreallyold6) { 319 V_rtq_reallyold6 = 2*V_rtq_reallyold6 / 3; 320 if (V_rtq_reallyold6 < V_rtq_minreallyold6) { 321 V_rtq_reallyold6 = V_rtq_minreallyold6; 322 } 323 324 last_adjusted_timeout = time_uptime; 325 #ifdef DIAGNOSTIC 326 log(LOG_DEBUG, "in6_rtqtimo: adjusted rtq_reallyold6 to %d", 327 V_rtq_reallyold6); 328 #endif 329 arg.found = arg.killed = 0; 330 arg.updating = 1; 331 RADIX_NODE_HEAD_LOCK(rnh); 332 rnh->rnh_walktree(rnh, in6_rtqkill, &arg); 333 RADIX_NODE_HEAD_UNLOCK(rnh); 334 } 335 336 atv.tv_usec = 0; 337 atv.tv_sec = arg.nextstop - time_uptime; 338 callout_reset(&V_rtq_timer6, tvtohz(&atv), in6_rtqtimo, rock); 339 CURVNET_RESTORE(); 340 } 341 342 /* 343 * Age old PMTUs. 344 */ 345 struct mtuex_arg { 346 struct radix_node_head *rnh; 347 time_t nextstop; 348 }; 349 350 static VNET_DEFINE(struct callout, rtq_mtutimer); 351 #define V_rtq_mtutimer VNET(rtq_mtutimer) 352 353 static int 354 in6_mtuexpire(struct radix_node *rn, void *rock) 355 { 356 struct rtentry *rt = (struct rtentry *)rn; 357 struct mtuex_arg *ap = rock; 358 359 /* sanity */ 360 if (!rt) 361 panic("rt == NULL in in6_mtuexpire"); 362 363 if (rt->rt_rmx.rmx_expire && !(rt->rt_flags & RTF_PROBEMTU)) { 364 if (rt->rt_rmx.rmx_expire <= time_uptime) { 365 rt->rt_flags |= RTF_PROBEMTU; 366 } else { 367 ap->nextstop = lmin(ap->nextstop, 368 rt->rt_rmx.rmx_expire); 369 } 370 } 371 372 return 0; 373 } 374 375 #define MTUTIMO_DEFAULT (60*1) 376 377 static void 378 in6_mtutimo(void *rock) 379 { 380 CURVNET_SET_QUIET((struct vnet *) rock); 381 struct radix_node_head *rnh; 382 struct mtuex_arg arg; 383 struct timeval atv; 384 385 rnh = rt_tables_get_rnh(0, AF_INET6); 386 if (rnh == NULL) { 387 CURVNET_RESTORE(); 388 return; 389 } 390 arg.rnh = rnh; 391 arg.nextstop = time_uptime + MTUTIMO_DEFAULT; 392 RADIX_NODE_HEAD_LOCK(rnh); 393 rnh->rnh_walktree(rnh, in6_mtuexpire, &arg); 394 RADIX_NODE_HEAD_UNLOCK(rnh); 395 396 atv.tv_usec = 0; 397 atv.tv_sec = arg.nextstop - time_uptime; 398 if (atv.tv_sec < 0) { 399 printf("invalid mtu expiration time on routing table\n"); 400 arg.nextstop = time_uptime + 30; /* last resort */ 401 atv.tv_sec = 30; 402 } 403 callout_reset(&V_rtq_mtutimer, tvtohz(&atv), in6_mtutimo, rock); 404 CURVNET_RESTORE(); 405 } 406 407 /* 408 * Initialize our routing tree. 409 * XXX MRT When off == 0, we are being called from vfs_export.c 410 * so just set up their table and leave. (we know what the correct 411 * value should be so just use that).. FIX AFTER RELENG_7 is MFC'd 412 * see also comments in in_inithead() vfs_export.c and domain.h 413 */ 414 int 415 in6_inithead(void **head, int off) 416 { 417 struct radix_node_head *rnh; 418 419 if (!rn_inithead(head, offsetof(struct sockaddr_in6, sin6_addr) << 3)) 420 return 0; /* See above */ 421 422 if (off == 0) /* See above */ 423 return 1; /* only do the rest for the real thing */ 424 425 V_rtq_reallyold6 = 60*60; /* one hour is ``really old'' */ 426 V_rtq_minreallyold6 = 10; /* never automatically crank down to less */ 427 V_rtq_toomany6 = 128; /* 128 cached routes is ``too many'' */ 428 V_rtq_timeout6 = RTQ_TIMEOUT; 429 430 rnh = *head; 431 KASSERT(rnh == rt_tables_get_rnh(0, AF_INET6), ("rnh?")); 432 rnh->rnh_addaddr = in6_addroute; 433 rnh->rnh_matchaddr = in6_matroute; 434 callout_init(&V_rtq_timer6, CALLOUT_MPSAFE); 435 callout_init(&V_rtq_mtutimer, CALLOUT_MPSAFE); 436 in6_rtqtimo(curvnet); /* kick off timeout first time */ 437 in6_mtutimo(curvnet); /* kick off timeout first time */ 438 return 1; 439 } 440 441 #ifdef VIMAGE 442 int 443 in6_detachhead(void **head, int off) 444 { 445 446 callout_drain(&V_rtq_timer6); 447 callout_drain(&V_rtq_mtutimer); 448 return (1); 449 } 450 #endif 451