Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]
FreeBSD/Linux Kernel Cross Reference
sys/netinet6/in6_rmx.c
Version:
- FREEBSD - FREEBSD-13-STABLE - FREEBSD-13-0 - FREEBSD-12-STABLE - FREEBSD-12-0 - FREEBSD-11-STABLE - FREEBSD-11-0 - FREEBSD-10-STABLE - FREEBSD-10-0 - FREEBSD-9-STABLE - FREEBSD-9-0 - FREEBSD-8-STABLE - FREEBSD-8-0 - FREEBSD-7-STABLE - FREEBSD-7-0 - FREEBSD-6-STABLE - FREEBSD-6-0 - FREEBSD-5-STABLE - FREEBSD-5-0 - FREEBSD-4-STABLE - FREEBSD-3-STABLE - FREEBSD22 - l41 - OPENBSD - linux-2.6 - MK84 - PLAN9 - xnu-8792
SearchContext: - none - 3 - 10
SearchContext: - none - 3 - 10
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: releng/7.4/sys/netinet6/in6_rmx.c 213687 2010-10-11 11:25:37Z delphij $"); 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 89 #include <net/if.h> 90 #include <net/route.h> 91 #include <netinet/in.h> 92 #include <netinet/ip_var.h> 93 #include <netinet/in_var.h> 94 95 #include <netinet/ip6.h> 96 #include <netinet6/ip6_var.h> 97 98 #include <netinet/icmp6.h> 99 #include <netinet6/nd6.h> 100 101 #include <netinet/tcp.h> 102 #include <netinet/tcp_seq.h> 103 #include <netinet/tcp_timer.h> 104 #include <netinet/tcp_var.h> 105 106 extern int in6_inithead(void **head, int off); 107 108 #define RTPRF_OURS RTF_PROTO3 /* set on routes we manage */ 109 110 /* 111 * Do what we need to do when inserting a route. 112 */ 113 static struct radix_node * 114 in6_addroute(void *v_arg, void *n_arg, struct radix_node_head *head, 115 struct radix_node *treenodes) 116 { 117 struct rtentry *rt = (struct rtentry *)treenodes; 118 struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)rt_key(rt); 119 struct radix_node *ret; 120 121 if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) 122 rt->rt_flags |= RTF_MULTICAST; 123 124 /* 125 * A little bit of help for both IPv6 output and input: 126 * For local addresses, we make sure that RTF_LOCAL is set, 127 * with the thought that this might one day be used to speed up 128 * ip_input(). 129 * 130 * We also mark routes to multicast addresses as such, because 131 * it's easy to do and might be useful (but this is much more 132 * dubious since it's so easy to inspect the address). (This 133 * is done above.) 134 * 135 * XXX 136 * should elaborate the code. 137 */ 138 if (rt->rt_flags & RTF_HOST) { 139 if (IN6_ARE_ADDR_EQUAL(&satosin6(rt->rt_ifa->ifa_addr) 140 ->sin6_addr, 141 &sin6->sin6_addr)) { 142 rt->rt_flags |= RTF_LOCAL; 143 } 144 } 145 146 if (!rt->rt_rmx.rmx_mtu && rt->rt_ifp) 147 rt->rt_rmx.rmx_mtu = IN6_LINKMTU(rt->rt_ifp); 148 149 ret = rn_addroute(v_arg, n_arg, head, treenodes); 150 if (ret == NULL && rt->rt_flags & RTF_HOST) { 151 struct rtentry *rt2; 152 /* 153 * We are trying to add a host route, but can't. 154 * Find out if it is because of an 155 * ARP entry and delete it if so. 156 */ 157 rt2 = rtalloc1((struct sockaddr *)sin6, 0, RTF_RNH_LOCKED|RTF_CLONING); 158 if (rt2) { 159 if (rt2->rt_flags & RTF_LLINFO && 160 rt2->rt_flags & RTF_HOST && 161 rt2->rt_gateway && 162 rt2->rt_gateway->sa_family == AF_LINK) { 163 rtexpunge(rt2); 164 RTFREE_LOCKED(rt2); 165 ret = rn_addroute(v_arg, n_arg, head, 166 treenodes); 167 } else 168 RTFREE_LOCKED(rt2); 169 } 170 } else if (ret == NULL && rt->rt_flags & RTF_CLONING) { 171 struct rtentry *rt2; 172 /* 173 * We are trying to add a net route, but can't. 174 * The following case should be allowed, so we'll make a 175 * special check for this: 176 * Two IPv6 addresses with the same prefix is assigned 177 * to a single interrface. 178 * # ifconfig if0 inet6 3ffe:0501::1 prefix 64 alias (*1) 179 * # ifconfig if0 inet6 3ffe:0501::2 prefix 64 alias (*2) 180 * In this case, (*1) and (*2) want to add the same 181 * net route entry, 3ffe:0501:: -> if0. 182 * This case should not raise an error. 183 */ 184 rt2 = rtalloc1((struct sockaddr *)sin6, 0, RTF_RNH_LOCKED|RTF_CLONING); 185 if (rt2) { 186 if ((rt2->rt_flags & (RTF_CLONING|RTF_HOST|RTF_GATEWAY)) 187 == RTF_CLONING 188 && rt2->rt_gateway 189 && rt2->rt_gateway->sa_family == AF_LINK 190 && rt2->rt_ifp == rt->rt_ifp) { 191 ret = rt2->rt_nodes; 192 } 193 RTFREE_LOCKED(rt2); 194 } 195 } 196 return ret; 197 } 198 199 /* 200 * This code is the inverse of in6_clsroute: on first reference, if we 201 * were managing the route, stop doing so and set the expiration timer 202 * back off again. 203 */ 204 static struct radix_node * 205 in6_matroute(void *v_arg, struct radix_node_head *head) 206 { 207 struct radix_node *rn = rn_match(v_arg, head); 208 struct rtentry *rt = (struct rtentry *)rn; 209 210 if (rt) { 211 RT_LOCK(rt); 212 if (rt->rt_flags & RTPRF_OURS) { 213 rt->rt_flags &= ~RTPRF_OURS; 214 rt->rt_rmx.rmx_expire = 0; 215 } 216 RT_UNLOCK(rt); 217 } 218 return rn; 219 } 220 221 SYSCTL_DECL(_net_inet6_ip6); 222 223 static int rtq_reallyold = 60*60; 224 /* one hour is ``really old'' */ 225 SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTEXPIRE, rtexpire, 226 CTLFLAG_RW, &rtq_reallyold , 0, ""); 227 228 static int rtq_minreallyold = 10; 229 /* never automatically crank down to less */ 230 SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTMINEXPIRE, rtminexpire, 231 CTLFLAG_RW, &rtq_minreallyold , 0, ""); 232 233 static int rtq_toomany = 128; 234 /* 128 cached routes is ``too many'' */ 235 SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTMAXCACHE, rtmaxcache, 236 CTLFLAG_RW, &rtq_toomany , 0, ""); 237 238 239 /* 240 * On last reference drop, mark the route as belong to us so that it can be 241 * timed out. 242 */ 243 static void 244 in6_clsroute(struct radix_node *rn, struct radix_node_head *head) 245 { 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_reallyold is 0, just delete the route without 262 * waiting for a timeout cycle to kill it. 263 */ 264 if (rtq_reallyold != 0) { 265 rt->rt_flags |= RTPRF_OURS; 266 rt->rt_rmx.rmx_expire = time_uptime + rtq_reallyold; 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_reallyold. 286 */ 287 static int 288 in6_rtqkill(struct radix_node *rn, void *rock) 289 { 290 struct rtqk_arg *ap = rock; 291 struct rtentry *rt = (struct rtentry *)rn; 292 int err; 293 294 RADIX_NODE_HEAD_LOCK_ASSERT(ap->rnh); 295 296 if (rt->rt_flags & RTPRF_OURS) { 297 ap->found++; 298 299 if (ap->draining || rt->rt_rmx.rmx_expire <= time_uptime) { 300 if (rt->rt_refcnt > 0) 301 panic("rtqkill route really not free"); 302 303 err = rtrequest(RTM_DELETE, 304 (struct sockaddr *)rt_key(rt), 305 rt->rt_gateway, rt_mask(rt), 306 rt->rt_flags|RTF_RNH_LOCKED, 0); 307 if (err) { 308 log(LOG_WARNING, "in6_rtqkill: error %d", err); 309 } else { 310 ap->killed++; 311 } 312 } else { 313 if (ap->updating 314 && (rt->rt_rmx.rmx_expire - time_uptime 315 > rtq_reallyold)) { 316 rt->rt_rmx.rmx_expire = time_uptime 317 + rtq_reallyold; 318 } 319 ap->nextstop = lmin(ap->nextstop, 320 rt->rt_rmx.rmx_expire); 321 } 322 } 323 324 return 0; 325 } 326 327 #define RTQ_TIMEOUT 60*10 /* run no less than once every ten minutes */ 328 static int rtq_timeout = RTQ_TIMEOUT; 329 static struct callout rtq_timer; 330 331 static void 332 in6_rtqtimo(void *rock) 333 { 334 struct radix_node_head *rnh = rock; 335 struct rtqk_arg arg; 336 struct timeval atv; 337 static time_t last_adjusted_timeout = 0; 338 339 arg.found = arg.killed = 0; 340 arg.rnh = rnh; 341 arg.nextstop = time_uptime + rtq_timeout; 342 arg.draining = arg.updating = 0; 343 RADIX_NODE_HEAD_LOCK(rnh); 344 rnh->rnh_walktree(rnh, in6_rtqkill, &arg); 345 RADIX_NODE_HEAD_UNLOCK(rnh); 346 347 /* 348 * Attempt to be somewhat dynamic about this: 349 * If there are ``too many'' routes sitting around taking up space, 350 * then crank down the timeout, and see if we can't make some more 351 * go away. However, we make sure that we will never adjust more 352 * than once in rtq_timeout seconds, to keep from cranking down too 353 * hard. 354 */ 355 if ((arg.found - arg.killed > rtq_toomany) 356 && (time_uptime - last_adjusted_timeout >= rtq_timeout) 357 && rtq_reallyold > rtq_minreallyold) { 358 rtq_reallyold = 2*rtq_reallyold / 3; 359 if (rtq_reallyold < rtq_minreallyold) { 360 rtq_reallyold = rtq_minreallyold; 361 } 362 363 last_adjusted_timeout = time_uptime; 364 #ifdef DIAGNOSTIC 365 log(LOG_DEBUG, "in6_rtqtimo: adjusted rtq_reallyold to %d", 366 rtq_reallyold); 367 #endif 368 arg.found = arg.killed = 0; 369 arg.updating = 1; 370 RADIX_NODE_HEAD_LOCK(rnh); 371 rnh->rnh_walktree(rnh, in6_rtqkill, &arg); 372 RADIX_NODE_HEAD_UNLOCK(rnh); 373 } 374 375 atv.tv_usec = 0; 376 atv.tv_sec = arg.nextstop - time_uptime; 377 callout_reset(&rtq_timer, tvtohz(&atv), in6_rtqtimo, rock); 378 } 379 380 /* 381 * Age old PMTUs. 382 */ 383 struct mtuex_arg { 384 struct radix_node_head *rnh; 385 time_t nextstop; 386 }; 387 static struct callout rtq_mtutimer; 388 389 static int 390 in6_mtuexpire(struct radix_node *rn, void *rock) 391 { 392 struct rtentry *rt = (struct rtentry *)rn; 393 struct mtuex_arg *ap = rock; 394 395 /* sanity */ 396 if (!rt) 397 panic("rt == NULL in in6_mtuexpire"); 398 399 if (rt->rt_rmx.rmx_expire && !(rt->rt_flags & RTF_PROBEMTU)) { 400 if (rt->rt_rmx.rmx_expire <= time_uptime) { 401 rt->rt_flags |= RTF_PROBEMTU; 402 } else { 403 ap->nextstop = lmin(ap->nextstop, 404 rt->rt_rmx.rmx_expire); 405 } 406 } 407 408 return 0; 409 } 410 411 #define MTUTIMO_DEFAULT (60*1) 412 413 static void 414 in6_mtutimo(void *rock) 415 { 416 struct radix_node_head *rnh = rock; 417 struct mtuex_arg arg; 418 struct timeval atv; 419 420 arg.rnh = rnh; 421 arg.nextstop = time_uptime + MTUTIMO_DEFAULT; 422 RADIX_NODE_HEAD_LOCK(rnh); 423 rnh->rnh_walktree(rnh, in6_mtuexpire, &arg); 424 RADIX_NODE_HEAD_UNLOCK(rnh); 425 426 atv.tv_usec = 0; 427 atv.tv_sec = arg.nextstop - time_uptime; 428 if (atv.tv_sec < 0) { 429 printf("invalid mtu expiration time on routing table\n"); 430 arg.nextstop = time_uptime + 30; /* last resort */ 431 atv.tv_sec = 30; 432 } 433 callout_reset(&rtq_mtutimer, tvtohz(&atv), in6_mtutimo, rock); 434 } 435 436 #if 0 437 void 438 in6_rtqdrain(void) 439 { 440 struct radix_node_head *rnh = rt_tables[AF_INET6]; 441 struct rtqk_arg arg; 442 443 arg.found = arg.killed = 0; 444 arg.rnh = rnh; 445 arg.nextstop = 0; 446 arg.draining = 1; 447 arg.updating = 0; 448 RADIX_NODE_HEAD_LOCK(rnh); 449 rnh->rnh_walktree(rnh, in6_rtqkill, &arg); 450 RADIX_NODE_HEAD_UNLOCK(rnh); 451 } 452 #endif 453 454 /* 455 * Initialize our routing tree. 456 * XXX MRT When off == 0, we are being called from vfs_export.c 457 * so just set up their table and leave. (we know what the correct 458 * value should be so just use that).. FIX AFTER RELENG_7 is MFC'd 459 * see also comments in in_inithead() vfs_export.c and domain.h 460 */ 461 int 462 in6_inithead(void **head, int off) 463 { 464 struct radix_node_head *rnh; 465 466 if (!rn_inithead(head, offsetof(struct sockaddr_in6, sin6_addr) << 3)) 467 return 0; /* See above */ 468 469 if (off == 0) /* See above */ 470 return 1; /* only do the rest for the real thing */ 471 472 rnh = *head; 473 rnh->rnh_addaddr = in6_addroute; 474 rnh->rnh_matchaddr = in6_matroute; 475 rnh->rnh_close = in6_clsroute; 476 callout_init(&rtq_timer, CALLOUT_MPSAFE); 477 in6_rtqtimo(rnh); /* kick off timeout first time */ 478 callout_init(&rtq_mtutimer, CALLOUT_MPSAFE); 479 in6_mtutimo(rnh); /* kick off timeout first time */ 480 return 1; 481 }
Cache object: eae3cc8a7c1c0bd669e7690568eeec34
[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]