bsd/netinet6/in6_rmx.c

   1 /*
   2  * Copyright (c) 2003-2007 Apple Inc. All rights reserved.
   3  *
   4  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
   5  *
   6  * This file contains Original Code and/or Modifications of Original Code
   7  * as defined in and that are subject to the Apple Public Source License
   8  * Version 2.0 (the 'License'). You may not use this file except in
   9  * compliance with the License. The rights granted to you under the License
  10  * may not be used to create, or enable the creation or redistribution of,
  11  * unlawful or unlicensed copies of an Apple operating system, or to
  12  * circumvent, violate, or enable the circumvention or violation of, any
  13  * terms of an Apple operating system software license agreement.
  14  *
  15  * Please obtain a copy of the License at
  16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
  17  *
  18  * The Original Code and all software distributed under the License are
  19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  23  * Please see the License for the specific language governing rights and
  24  * limitations under the License.
  25  *
  26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  27  */
  28
  29 /*      $FreeBSD: src/sys/netinet6/in6_rmx.c,v 1.1.2.2 2001/07/03 11:01:52 ume Exp $    */
  30 /*      $KAME: in6_rmx.c,v 1.10 2001/05/24 05:44:58 itojun Exp $        */
  31
  32 /*
  33  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
  34  * All rights reserved.
  35  *
  36  * Redistribution and use in source and binary forms, with or without
  37  * modification, are permitted provided that the following conditions
  38  * are met:
  39  * 1. Redistributions of source code must retain the above copyright
  40  *    notice, this list of conditions and the following disclaimer.
  41  * 2. Redistributions in binary form must reproduce the above copyright
  42  *    notice, this list of conditions and the following disclaimer in the
  43  *    documentation and/or other materials provided with the distribution.
  44  * 3. Neither the name of the project nor the names of its contributors
  45  *    may be used to endorse or promote products derived from this software
  46  *    without specific prior written permission.
  47  *
  48  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
  49  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  50  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  51  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
  52  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  53  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  54  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  55  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  56  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  57  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  58  * SUCH DAMAGE.
  59  */
  60
  61 /*
  62  * Copyright 1994, 1995 Massachusetts Institute of Technology
  63  *
  64  * Permission to use, copy, modify, and distribute this software and
  65  * its documentation for any purpose and without fee is hereby
  66  * granted, provided that both the above copyright notice and this
  67  * permission notice appear in all copies, that both the above
  68  * copyright notice and this permission notice appear in all
  69  * supporting documentation, and that the name of M.I.T. not be used
  70  * in advertising or publicity pertaining to distribution of the
  71  * software without specific, written prior permission.  M.I.T. makes
  72  * no representations about the suitability of this software for any
  73  * purpose.  It is provided "as is" without express or implied
  74  * warranty.
  75  *
  76  * THIS SOFTWARE IS PROVIDED BY M.I.T. ``AS IS''.  M.I.T. DISCLAIMS
  77  * ALL EXPRESS OR IMPLIED WARRANTIES WITH REGARD TO THIS SOFTWARE,
  78  * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  79  * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT
  80  * SHALL M.I.T. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
  81  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
  82  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
  83  * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
  84  * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
  85  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
  86  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  87  * SUCH DAMAGE.
  88  *
  89  */
  90
  91 /*
  92  * This code does two things necessary for the enhanced TCP metrics to
  93  * function in a useful manner:
  94  *  1) It marks all non-host routes as `cloning', thus ensuring that
  95  *     every actual reference to such a route actually gets turned
  96  *     into a reference to a host route to the specific destination
  97  *     requested.
  98  *  2) When such routes lose all their references, it arranges for them
  99  *     to be deleted in some random collection of circumstances, so that
 100  *     a large quantity of stale routing data is not kept in kernel memory
 101  *     indefinitely.  See in6_rtqtimo() below for the exact mechanism.
 102  */
 103
 104 #include <sys/param.h>
 105 #include <sys/systm.h>
 106 #include <sys/kernel.h>
 107 #include <sys/sysctl.h>
 108 #include <kern/queue.h>
 109 #include <sys/socket.h>
 110 #include <sys/socketvar.h>
 111 #include <sys/mbuf.h>
 112 #include <sys/syslog.h>
 113 #include <kern/lock.h>
 114
 115 #include <net/if.h>
 116 #include <net/route.h>
 117 #include <netinet/in.h>
 118 #include <netinet/ip_var.h>
 119 #include <netinet/in_var.h>
 120
 121 #include <netinet/ip6.h>
 122 #include <netinet6/ip6_var.h>
 123
 124 #include <netinet/icmp6.h>
 125
 126 #include <netinet/tcp.h>
 127 #include <netinet/tcp_seq.h>
 128 #include <netinet/tcp_timer.h>
 129 #include <netinet/tcp_var.h>
 130
 131 extern int      in6_inithead(void **head, int off);
 132 static void     in6_rtqtimo(void *rock);
 133 static void in6_mtutimo(void *rock);
 134 extern int tvtohz(struct timeval *);
 135
 136 #define RTPRF_OURS              RTF_PROTO3      /* set on routes we manage */
 137
 138 /*
 139  * Do what we need to do when inserting a route.
 140  */
 141 static struct radix_node *
 142 in6_addroute(void *v_arg, void *n_arg, struct radix_node_head *head,
 143             struct radix_node *treenodes)
 144 {
 145         struct rtentry *rt = (struct rtentry *)treenodes;
 146         struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)rt_key(rt);
 147         struct radix_node *ret;
 148
 149         /*
 150          * For IPv6, all unicast non-host routes are automatically cloning.
 151          */
 152         if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
 153                 rt->rt_flags |= RTF_MULTICAST;
 154
 155         if (!(rt->rt_flags & (RTF_HOST | RTF_CLONING | RTF_MULTICAST))) {
 156                 rt->rt_flags |= RTF_PRCLONING;
 157         }
 158
 159         /*
 160          * A little bit of help for both IPv6 output and input:
 161          *   For local addresses, we make sure that RTF_LOCAL is set,
 162          *   with the thought that this might one day be used to speed up
 163          *   ip_input().
 164          *
 165          * We also mark routes to multicast addresses as such, because
 166          * it's easy to do and might be useful (but this is much more
 167          * dubious since it's so easy to inspect the address).  (This
 168          * is done above.)
 169          *
 170          * XXX
 171          * should elaborate the code.
 172          */
 173         if (rt->rt_flags & RTF_HOST) {
 174                 if (IN6_ARE_ADDR_EQUAL(&satosin6(rt->rt_ifa->ifa_addr)
 175                                         ->sin6_addr,
 176                                        &sin6->sin6_addr)) {
 177                         rt->rt_flags |= RTF_LOCAL;
 178                 }
 179         }
 180
 181         if (!rt->rt_rmx.rmx_mtu && !(rt->rt_rmx.rmx_locks & RTV_MTU)
 182             && rt->rt_ifp)
 183                 rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu;
 184
 185         ret = rn_addroute(v_arg, n_arg, head, treenodes);
 186         if (ret == NULL && rt->rt_flags & RTF_HOST) {
 187                 struct rtentry *rt2;
 188                 /*
 189                  * We are trying to add a host route, but can't.
 190                  * Find out if it is because of an
 191                  * ARP entry and delete it if so.
 192                  */
 193                 rt2 = rtalloc1_locked((struct sockaddr *)sin6, 0,
 194                                 RTF_CLONING | RTF_PRCLONING);
 195                 if (rt2) {
 196                         if (rt2->rt_flags & RTF_LLINFO &&
 197                                 rt2->rt_flags & RTF_HOST &&
 198                                 rt2->rt_gateway &&
 199                                 rt2->rt_gateway->sa_family == AF_LINK) {
 200                                 rtrequest_locked(RTM_DELETE,
 201                                           (struct sockaddr *)rt_key(rt2),
 202                                           rt2->rt_gateway,
 203                                           rt_mask(rt2), rt2->rt_flags, 0);
 204                                 ret = rn_addroute(v_arg, n_arg, head,
 205                                         treenodes);
 206                         }
 207                         rtfree_locked(rt2);
 208                 }
 209         } else if (ret == NULL && rt->rt_flags & RTF_CLONING) {
 210                 struct rtentry *rt2;
 211                 /*
 212                  * We are trying to add a net route, but can't.
 213                  * The following case should be allowed, so we'll make a
 214                  * special check for this:
 215                  *      Two IPv6 addresses with the same prefix is assigned
 216                  *      to a single interrface.
 217                  *      # ifconfig if0 inet6 3ffe:0501::1 prefix 64 alias (*1)
 218                  *      # ifconfig if0 inet6 3ffe:0501::2 prefix 64 alias (*2)
 219                  *      In this case, (*1) and (*2) want to add the same
 220                  *      net route entry, 3ffe:0501:: -> if0.
 221                  *      This case should not raise an error.
 222                  */
 223                 rt2 = rtalloc1_locked((struct sockaddr *)sin6, 0,
 224                                 RTF_CLONING | RTF_PRCLONING);
 225                 if (rt2) {
 226                         if ((rt2->rt_flags & (RTF_CLONING|RTF_HOST|RTF_GATEWAY))
 227                                         == RTF_CLONING
 228                          && rt2->rt_gateway
 229                          && rt2->rt_gateway->sa_family == AF_LINK
 230                          && rt2->rt_ifp == rt->rt_ifp) {
 231                                 ret = rt2->rt_nodes;
 232                         }
 233                         rtfree_locked(rt2);
 234                 }
 235         }
 236         return ret;
 237 }
 238
 239 /*
 240  * This code is the inverse of in6_clsroute: on first reference, if we
 241  * were managing the route, stop doing so and set the expiration timer
 242  * back off again.
 243  */
 244 static struct radix_node *
 245 in6_matroute(void *v_arg, struct radix_node_head *head)
 246 {
 247         struct radix_node *rn = rn_match(v_arg, head);
 248         struct rtentry *rt = (struct rtentry *)rn;
 249
 250         if (rt && rt->rt_refcnt == 0) { /* this is first reference */
 251                 if (rt->rt_flags & RTPRF_OURS) {
 252                         rt->rt_flags &= ~RTPRF_OURS;
 253                         rt->rt_rmx.rmx_expire = 0;
 254                 }
 255         }
 256         return rn;
 257 }
 258
 259 SYSCTL_DECL(_net_inet6_ip6);
 260
 261 static int rtq_reallyold = 60*60;
 262         /* one hour is ``really old'' */
 263 SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTEXPIRE, rtexpire,
 264         CTLFLAG_RW, &rtq_reallyold , 0, "");
 265
 266 static int rtq_minreallyold = 10;
 267         /* never automatically crank down to less */
 268 SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTMINEXPIRE, rtminexpire,
 269         CTLFLAG_RW, &rtq_minreallyold , 0, "");
 270
 271 static int rtq_toomany = 128;
 272         /* 128 cached routes is ``too many'' */
 273 SYSCTL_INT(_net_inet6_ip6, IPV6CTL_RTMAXCACHE, rtmaxcache,
 274         CTLFLAG_RW, &rtq_toomany , 0, "");
 275
 276
 277 /*
 278  * On last reference drop, mark the route as belong to us so that it can be
 279  * timed out.
 280  */
 281 static void
 282 in6_clsroute(struct radix_node *rn, __unused struct radix_node_head *head)
 283 {
 284         struct rtentry *rt = (struct rtentry *)rn;
 285
 286         if (!(rt->rt_flags & RTF_UP))
 287                 return;         /* prophylactic measures */
 288
 289         if ((rt->rt_flags & (RTF_LLINFO | RTF_HOST)) != RTF_HOST)
 290                 return;
 291
 292         if ((rt->rt_flags & (RTF_WASCLONED | RTPRF_OURS)) != RTF_WASCLONED)
 293                 return;
 294
 295         /*
 296          * Delete the route immediately if RTF_DELCLONE is set or
 297          * if route caching is disabled (rtq_reallyold set to 0).
 298          * Otherwise, let it expire and be deleted by in6_rtqkill().
 299          */
 300         if ((rt->rt_flags & RTF_DELCLONE) || rtq_reallyold == 0) {
 301                 /*
 302                  * Delete the route from the radix tree but since we are
 303                  * called when the route's reference count is 0, don't
 304                  * deallocate it until we return from this routine by
 305                  * telling rtrequest that we're interested in it.
 306                  */
 307                 if (rtrequest_locked(RTM_DELETE, (struct sockaddr *)rt_key(rt),
 308                     rt->rt_gateway, rt_mask(rt), rt->rt_flags, &rt) == 0) {
 309                         /* Now let the caller free it */
 310                         rtunref(rt);
 311                 }
 312         } else {
 313                 struct timeval timenow;
 314
 315                 getmicrotime(&timenow);
 316                 rt->rt_flags |= RTPRF_OURS;
 317                 rt->rt_rmx.rmx_expire = timenow.tv_sec + rtq_reallyold;
 318         }
 319 }
 320
 321 struct rtqk_arg {
 322         struct radix_node_head *rnh;
 323         int mode;
 324         int updating;
 325         int draining;
 326         int killed;
 327         int found;
 328         time_t nextstop;
 329 };
 330
 331 /*
 332  * Get rid of old routes.  When draining, this deletes everything, even when
 333  * the timeout is not expired yet.  When updating, this makes sure that
 334  * nothing has a timeout longer than the current value of rtq_reallyold.
 335  */
 336 static int
 337 in6_rtqkill(struct radix_node *rn, void *rock)
 338 {
 339         struct rtqk_arg *ap = rock;
 340         struct rtentry *rt = (struct rtentry *)rn;
 341         int err;
 342         struct timeval timenow;
 343
 344         getmicrotime(&timenow);
 345         lck_mtx_assert(rt_mtx, LCK_MTX_ASSERT_OWNED);
 346
 347         if (rt->rt_flags & RTPRF_OURS) {
 348                 ap->found++;
 349
 350                 if (ap->draining || rt->rt_rmx.rmx_expire <= timenow.tv_sec) {
 351                         if (rt->rt_refcnt > 0)
 352                                 panic("rtqkill route really not free");
 353
 354                         err = rtrequest_locked(RTM_DELETE,
 355                                         (struct sockaddr *)rt_key(rt),
 356                                         rt->rt_gateway, rt_mask(rt),
 357                                         rt->rt_flags, 0);
 358                         if (err) {
 359                                 log(LOG_WARNING, "in6_rtqkill: error %d", err);
 360                         } else {
 361                                 ap->killed++;
 362                         }
 363                 } else {
 364                         if (ap->updating
 365                            && (rt->rt_rmx.rmx_expire - timenow.tv_sec
 366                                > rtq_reallyold)) {
 367                                 rt->rt_rmx.rmx_expire = timenow.tv_sec
 368                                         + rtq_reallyold;
 369                         }
 370                         ap->nextstop = lmin(ap->nextstop,
 371                                             rt->rt_rmx.rmx_expire);
 372                 }
 373         }
 374
 375         return 0;
 376 }
 377
 378 #define RTQ_TIMEOUT     60*10   /* run no less than once every ten minutes */
 379 static int rtq_timeout = RTQ_TIMEOUT;
 380
 381 static void
 382 in6_rtqtimo(void *rock)
 383 {
 384         struct radix_node_head *rnh = rock;
 385         struct rtqk_arg arg;
 386         struct timeval atv;
 387         static time_t last_adjusted_timeout = 0;
 388         struct timeval timenow;
 389
 390         lck_mtx_lock(rt_mtx);
 391         /* Get the timestamp after we acquire the lock for better accuracy */
 392         getmicrotime(&timenow);
 393
 394         arg.found = arg.killed = 0;
 395         arg.rnh = rnh;
 396         arg.nextstop = timenow.tv_sec + rtq_timeout;
 397         arg.draining = arg.updating = 0;
 398         rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
 399
 400         /*
 401          * Attempt to be somewhat dynamic about this:
 402          * If there are ``too many'' routes sitting around taking up space,
 403          * then crank down the timeout, and see if we can't make some more
 404          * go away.  However, we make sure that we will never adjust more
 405          * than once in rtq_timeout seconds, to keep from cranking down too
 406          * hard.
 407          */
 408         if ((arg.found - arg.killed > rtq_toomany)
 409            && (timenow.tv_sec - last_adjusted_timeout >= rtq_timeout)
 410            && rtq_reallyold > rtq_minreallyold) {
 411                 rtq_reallyold = 2*rtq_reallyold / 3;
 412                 if (rtq_reallyold < rtq_minreallyold) {
 413                         rtq_reallyold = rtq_minreallyold;
 414                 }
 415
 416                 last_adjusted_timeout = timenow.tv_sec;
 417 #if DIAGNOSTIC
 418                 log(LOG_DEBUG, "in6_rtqtimo: adjusted rtq_reallyold to %d",
 419                     rtq_reallyold);
 420 #endif
 421                 arg.found = arg.killed = 0;
 422                 arg.updating = 1;
 423                 rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
 424         }
 425
 426         atv.tv_usec = 0;
 427         atv.tv_sec = arg.nextstop - timenow.tv_sec;
 428         lck_mtx_unlock(rt_mtx);
 429         timeout(in6_rtqtimo, rock, tvtohz(&atv));
 430 }
 431
 432 /*
 433  * Age old PMTUs.
 434  */
 435 struct mtuex_arg {
 436         struct radix_node_head *rnh;
 437         time_t nextstop;
 438 };
 439
 440 static int
 441 in6_mtuexpire(struct radix_node *rn, void *rock)
 442 {
 443         struct rtentry *rt = (struct rtentry *)rn;
 444         struct mtuex_arg *ap = rock;
 445         struct timeval timenow;
 446
 447         getmicrotime(&timenow);
 448
 449         /* sanity */
 450         if (!rt)
 451                 panic("rt == NULL in in6_mtuexpire");
 452
 453         if (rt->rt_rmx.rmx_expire && !(rt->rt_flags & RTF_PROBEMTU)) {
 454                 if (rt->rt_rmx.rmx_expire <= timenow.tv_sec) {
 455                         rt->rt_flags |= RTF_PROBEMTU;
 456                 } else {
 457                         ap->nextstop = lmin(ap->nextstop,
 458                                         rt->rt_rmx.rmx_expire);
 459                 }
 460         }
 461
 462         return 0;
 463 }
 464
 465 #define MTUTIMO_DEFAULT (60*1)
 466
 467 static void
 468 in6_mtutimo(void *rock)
 469 {
 470         struct radix_node_head *rnh = rock;
 471         struct mtuex_arg arg;
 472         struct timeval atv;
 473         struct timeval timenow;
 474
 475         getmicrotime(&timenow);
 476
 477         arg.rnh = rnh;
 478         arg.nextstop = timenow.tv_sec + MTUTIMO_DEFAULT;
 479         lck_mtx_lock(rt_mtx);
 480         rnh->rnh_walktree(rnh, in6_mtuexpire, &arg);
 481
 482         atv.tv_usec = 0;
 483         atv.tv_sec = arg.nextstop;
 484         if (atv.tv_sec < timenow.tv_sec) {
 485 #if DIAGNOSTIC
 486                 log(LOG_DEBUG, "IPv6: invalid mtu expiration time on routing table\n");
 487 #endif
 488                 arg.nextstop = timenow.tv_sec + 30;     /*last resort*/
 489         }
 490         atv.tv_sec -= timenow.tv_sec;
 491         lck_mtx_unlock(rt_mtx);
 492         timeout(in6_mtutimo, rock, tvtohz(&atv));
 493 }
 494
 495 #if 0
 496 void
 497 in6_rtqdrain()
 498 {
 499         struct radix_node_head *rnh = rt_tables[AF_INET6];
 500         struct rtqk_arg arg;
 501         int s;
 502         arg.found = arg.killed = 0;
 503         arg.rnh = rnh;
 504         arg.nextstop = 0;
 505         arg.draining = 1;
 506         arg.updating = 0;
 507         s = splnet();
 508         rnh->rnh_walktree(rnh, in6_rtqkill, &arg);
 509         splx(s);
 510 }
 511 #endif
 512
 513 /*
 514  * Initialize our routing tree.
 515  */
 516 int
 517 in6_inithead(void **head, int off)
 518 {
 519         struct radix_node_head *rnh;
 520
 521         if (!rn_inithead(head, off))
 522                 return 0;
 523
 524         if (head != (void **)&rt_tables[AF_INET6]) /* BOGUS! */
 525                 return 1;       /* only do this for the real routing table */
 526
 527         rnh = *head;
 528         rnh->rnh_addaddr = in6_addroute;
 529         rnh->rnh_matchaddr = in6_matroute;
 530         rnh->rnh_close = in6_clsroute;
 531         in6_rtqtimo(rnh);       /* kick off timeout first time */
 532         in6_mtutimo(rnh);       /* kick off timeout first time */
 533         return 1;
 534 }