The Design and Implementation of the FreeBSD Operating System, Second Edition
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/net/route.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 

    1 /*-
    2  * SPDX-License-Identifier: BSD-3-Clause
    3  *
    4  * Copyright (c) 1980, 1986, 1991, 1993
    5  *      The Regents of the University of California.  All rights reserved.
    6  *
    7  * Redistribution and use in source and binary forms, with or without
    8  * modification, are permitted provided that the following conditions
    9  * are met:
   10  * 1. Redistributions of source code must retain the above copyright
   11  *    notice, this list of conditions and the following disclaimer.
   12  * 2. Redistributions in binary form must reproduce the above copyright
   13  *    notice, this list of conditions and the following disclaimer in the
   14  *    documentation and/or other materials provided with the distribution.
   15  * 3. Neither the name of the University nor the names of its contributors
   16  *    may be used to endorse or promote products derived from this software
   17  *    without specific prior written permission.
   18  *
   19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
   20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   29  * SUCH DAMAGE.
   30  *
   31  *      @(#)route.c     8.3.1.1 (Berkeley) 2/23/95
   32  * $FreeBSD$
   33  */
   34 /************************************************************************
   35  * Note: In this file a 'fib' is a "forwarding information base"        *
   36  * Which is the new name for an in kernel routing (next hop) table.     *
   37  ***********************************************************************/
   38 
   39 #include "opt_inet.h"
   40 #include "opt_inet6.h"
   41 #include "opt_mrouting.h"
   42 #include "opt_mpath.h"
   43 #include "opt_route.h"
   44 
   45 #include <sys/param.h>
   46 #include <sys/systm.h>
   47 #include <sys/malloc.h>
   48 #include <sys/mbuf.h>
   49 #include <sys/socket.h>
   50 #include <sys/sysctl.h>
   51 #include <sys/syslog.h>
   52 #include <sys/sysproto.h>
   53 #include <sys/proc.h>
   54 #include <sys/domain.h>
   55 #include <sys/eventhandler.h>
   56 #include <sys/kernel.h>
   57 #include <sys/lock.h>
   58 #include <sys/rmlock.h>
   59 
   60 #include <net/if.h>
   61 #include <net/if_var.h>
   62 #include <net/if_dl.h>
   63 #include <net/route.h>
   64 #include <net/route_var.h>
   65 #include <net/vnet.h>
   66 
   67 #ifdef RADIX_MPATH
   68 #include <net/radix_mpath.h>
   69 #endif
   70 
   71 #include <netinet/in.h>
   72 #include <netinet/ip_mroute.h>
   73 
   74 #include <vm/uma.h>
   75 
   76 #define RT_MAXFIBS      UINT16_MAX
   77 
   78 /* Kernel config default option. */
   79 #ifdef ROUTETABLES
   80 #if ROUTETABLES <= 0
   81 #error "ROUTETABLES defined too low"
   82 #endif
   83 #if ROUTETABLES > RT_MAXFIBS
   84 #error "ROUTETABLES defined too big"
   85 #endif
   86 #define RT_NUMFIBS      ROUTETABLES
   87 #endif /* ROUTETABLES */
   88 /* Initialize to default if not otherwise set. */
   89 #ifndef RT_NUMFIBS
   90 #define RT_NUMFIBS      1
   91 #endif
   92 
   93 /* This is read-only.. */
   94 u_int rt_numfibs = RT_NUMFIBS;
   95 SYSCTL_UINT(_net, OID_AUTO, fibs, CTLFLAG_RDTUN, &rt_numfibs, 0, "");
   96 
   97 /*
   98  * By default add routes to all fibs for new interfaces.
   99  * Once this is set to 0 then only allocate routes on interface
  100  * changes for the FIB of the caller when adding a new set of addresses
  101  * to an interface.  XXX this is a shotgun aproach to a problem that needs
  102  * a more fine grained solution.. that will come.
  103  * XXX also has the problems getting the FIB from curthread which will not
  104  * always work given the fib can be overridden and prefixes can be added
  105  * from the network stack context.
  106  */
  107 VNET_DEFINE(u_int, rt_add_addr_allfibs) = 1;
  108 SYSCTL_UINT(_net, OID_AUTO, add_addr_allfibs, CTLFLAG_RWTUN | CTLFLAG_VNET,
  109     &VNET_NAME(rt_add_addr_allfibs), 0, "");
  110 
  111 VNET_DEFINE(struct rtstat, rtstat);
  112 #define V_rtstat        VNET(rtstat)
  113 
  114 VNET_DEFINE(struct rib_head *, rt_tables);
  115 #define V_rt_tables     VNET(rt_tables)
  116 
  117 VNET_DEFINE(int, rttrash);              /* routes not in table but not freed */
  118 #define V_rttrash       VNET(rttrash)
  119 
  120 
  121 /*
  122  * Convert a 'struct radix_node *' to a 'struct rtentry *'.
  123  * The operation can be done safely (in this code) because a
  124  * 'struct rtentry' starts with two 'struct radix_node''s, the first
  125  * one representing leaf nodes in the routing tree, which is
  126  * what the code in radix.c passes us as a 'struct radix_node'.
  127  *
  128  * But because there are a lot of assumptions in this conversion,
  129  * do not cast explicitly, but always use the macro below.
  130  */
  131 #define RNTORT(p)       ((struct rtentry *)(p))
  132 
  133 VNET_DEFINE_STATIC(uma_zone_t, rtzone);         /* Routing table UMA zone. */
  134 #define V_rtzone        VNET(rtzone)
  135 
  136 EVENTHANDLER_LIST_DEFINE(rt_addrmsg);
  137 
  138 static int rtrequest1_fib_change(struct rib_head *, struct rt_addrinfo *,
  139     struct rtentry **, u_int);
  140 static void rt_setmetrics(const struct rt_addrinfo *, struct rtentry *);
  141 static int rt_ifdelroute(const struct rtentry *rt, void *arg);
  142 static struct rtentry *rt_unlinkrte(struct rib_head *rnh,
  143     struct rt_addrinfo *info, int *perror);
  144 static void rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info);
  145 #ifdef RADIX_MPATH
  146 static struct radix_node *rt_mpath_unlink(struct rib_head *rnh,
  147     struct rt_addrinfo *info, struct rtentry *rto, int *perror);
  148 #endif
  149 static int rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info,
  150     int flags);
  151 
  152 struct if_mtuinfo
  153 {
  154         struct ifnet    *ifp;
  155         int             mtu;
  156 };
  157 
  158 static int      if_updatemtu_cb(struct radix_node *, void *);
  159 
  160 /*
  161  * handler for net.my_fibnum
  162  */
  163 static int
  164 sysctl_my_fibnum(SYSCTL_HANDLER_ARGS)
  165 {
  166         int fibnum;
  167         int error;
  168  
  169         fibnum = curthread->td_proc->p_fibnum;
  170         error = sysctl_handle_int(oidp, &fibnum, 0, req);
  171         return (error);
  172 }
  173 
  174 SYSCTL_PROC(_net, OID_AUTO, my_fibnum, CTLTYPE_INT|CTLFLAG_RD,
  175             NULL, 0, &sysctl_my_fibnum, "I", "default FIB of caller");
  176 
  177 static __inline struct rib_head **
  178 rt_tables_get_rnh_ptr(int table, int fam)
  179 {
  180         struct rib_head **rnh;
  181 
  182         KASSERT(table >= 0 && table < rt_numfibs, ("%s: table out of bounds.",
  183             __func__));
  184         KASSERT(fam >= 0 && fam < (AF_MAX+1), ("%s: fam out of bounds.",
  185             __func__));
  186 
  187         /* rnh is [fib=0][af=0]. */
  188         rnh = (struct rib_head **)V_rt_tables;
  189         /* Get the offset to the requested table and fam. */
  190         rnh += table * (AF_MAX+1) + fam;
  191 
  192         return (rnh);
  193 }
  194 
  195 struct rib_head *
  196 rt_tables_get_rnh(int table, int fam)
  197 {
  198 
  199         return (*rt_tables_get_rnh_ptr(table, fam));
  200 }
  201 
  202 u_int
  203 rt_tables_get_gen(int table, int fam)
  204 {
  205         struct rib_head *rnh;
  206 
  207         rnh = *rt_tables_get_rnh_ptr(table, fam);
  208         KASSERT(rnh != NULL, ("%s: NULL rib_head pointer table %d fam %d",
  209             __func__, table, fam));
  210         return (rnh->rnh_gen);
  211 }
  212 
  213 
  214 /*
  215  * route initialization must occur before ip6_init2(), which happenas at
  216  * SI_ORDER_MIDDLE.
  217  */
  218 static void
  219 route_init(void)
  220 {
  221 
  222         /* whack the tunable ints into  line. */
  223         if (rt_numfibs > RT_MAXFIBS)
  224                 rt_numfibs = RT_MAXFIBS;
  225         if (rt_numfibs == 0)
  226                 rt_numfibs = 1;
  227 }
  228 SYSINIT(route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, route_init, NULL);
  229 
  230 static int
  231 rtentry_zinit(void *mem, int size, int how)
  232 {
  233         struct rtentry *rt = mem;
  234 
  235         rt->rt_pksent = counter_u64_alloc(how);
  236         if (rt->rt_pksent == NULL)
  237                 return (ENOMEM);
  238 
  239         RT_LOCK_INIT(rt);
  240 
  241         return (0);
  242 }
  243 
  244 static void
  245 rtentry_zfini(void *mem, int size)
  246 {
  247         struct rtentry *rt = mem;
  248 
  249         RT_LOCK_DESTROY(rt);
  250         counter_u64_free(rt->rt_pksent);
  251 }
  252 
  253 static int
  254 rtentry_ctor(void *mem, int size, void *arg, int how)
  255 {
  256         struct rtentry *rt = mem;
  257 
  258         bzero(rt, offsetof(struct rtentry, rt_endzero));
  259         counter_u64_zero(rt->rt_pksent);
  260         rt->rt_chain = NULL;
  261 
  262         return (0);
  263 }
  264 
  265 static void
  266 rtentry_dtor(void *mem, int size, void *arg)
  267 {
  268         struct rtentry *rt = mem;
  269 
  270         RT_UNLOCK_COND(rt);
  271 }
  272 
  273 static void
  274 vnet_route_init(const void *unused __unused)
  275 {
  276         struct domain *dom;
  277         struct rib_head **rnh;
  278         int table;
  279         int fam;
  280 
  281         V_rt_tables = malloc(rt_numfibs * (AF_MAX+1) *
  282             sizeof(struct rib_head *), M_RTABLE, M_WAITOK|M_ZERO);
  283 
  284         V_rtzone = uma_zcreate("rtentry", sizeof(struct rtentry),
  285             rtentry_ctor, rtentry_dtor,
  286             rtentry_zinit, rtentry_zfini, UMA_ALIGN_PTR, 0);
  287         for (dom = domains; dom; dom = dom->dom_next) {
  288                 if (dom->dom_rtattach == NULL)
  289                         continue;
  290 
  291                 for  (table = 0; table < rt_numfibs; table++) {
  292                         fam = dom->dom_family;
  293                         if (table != 0 && fam != AF_INET6 && fam != AF_INET)
  294                                 break;
  295 
  296                         rnh = rt_tables_get_rnh_ptr(table, fam);
  297                         if (rnh == NULL)
  298                                 panic("%s: rnh NULL", __func__);
  299                         dom->dom_rtattach((void **)rnh, 0);
  300                 }
  301         }
  302 }
  303 VNET_SYSINIT(vnet_route_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH,
  304     vnet_route_init, 0);
  305 
  306 #ifdef VIMAGE
  307 static void
  308 vnet_route_uninit(const void *unused __unused)
  309 {
  310         int table;
  311         int fam;
  312         struct domain *dom;
  313         struct rib_head **rnh;
  314 
  315         for (dom = domains; dom; dom = dom->dom_next) {
  316                 if (dom->dom_rtdetach == NULL)
  317                         continue;
  318 
  319                 for (table = 0; table < rt_numfibs; table++) {
  320                         fam = dom->dom_family;
  321 
  322                         if (table != 0 && fam != AF_INET6 && fam != AF_INET)
  323                                 break;
  324 
  325                         rnh = rt_tables_get_rnh_ptr(table, fam);
  326                         if (rnh == NULL)
  327                                 panic("%s: rnh NULL", __func__);
  328                         dom->dom_rtdetach((void **)rnh, 0);
  329                 }
  330         }
  331 
  332         free(V_rt_tables, M_RTABLE);
  333         uma_zdestroy(V_rtzone);
  334 }
  335 VNET_SYSUNINIT(vnet_route_uninit, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST,
  336     vnet_route_uninit, 0);
  337 #endif
  338 
  339 struct rib_head *
  340 rt_table_init(int offset)
  341 {
  342         struct rib_head *rh;
  343 
  344         rh = malloc(sizeof(struct rib_head), M_RTABLE, M_WAITOK | M_ZERO);
  345 
  346         /* TODO: These details should be hidded inside radix.c */
  347         /* Init masks tree */
  348         rn_inithead_internal(&rh->head, rh->rnh_nodes, offset);
  349         rn_inithead_internal(&rh->rmhead.head, rh->rmhead.mask_nodes, 0);
  350         rh->head.rnh_masks = &rh->rmhead;
  351 
  352         /* Init locks */
  353         RIB_LOCK_INIT(rh);
  354 
  355         /* Finally, set base callbacks */
  356         rh->rnh_addaddr = rn_addroute;
  357         rh->rnh_deladdr = rn_delete;
  358         rh->rnh_matchaddr = rn_match;
  359         rh->rnh_lookup = rn_lookup;
  360         rh->rnh_walktree = rn_walktree;
  361         rh->rnh_walktree_from = rn_walktree_from;
  362 
  363         return (rh);
  364 }
  365 
  366 static int
  367 rt_freeentry(struct radix_node *rn, void *arg)
  368 {
  369         struct radix_head * const rnh = arg;
  370         struct radix_node *x;
  371 
  372         x = (struct radix_node *)rn_delete(rn + 2, NULL, rnh);
  373         if (x != NULL)
  374                 R_Free(x);
  375         return (0);
  376 }
  377 
  378 void
  379 rt_table_destroy(struct rib_head *rh)
  380 {
  381 
  382         rn_walktree(&rh->rmhead.head, rt_freeentry, &rh->rmhead.head);
  383 
  384         /* Assume table is already empty */
  385         RIB_LOCK_DESTROY(rh);
  386         free(rh, M_RTABLE);
  387 }
  388 
  389 
  390 #ifndef _SYS_SYSPROTO_H_
  391 struct setfib_args {
  392         int     fibnum;
  393 };
  394 #endif
  395 int
  396 sys_setfib(struct thread *td, struct setfib_args *uap)
  397 {
  398         if (uap->fibnum < 0 || uap->fibnum >= rt_numfibs)
  399                 return EINVAL;
  400         td->td_proc->p_fibnum = uap->fibnum;
  401         return (0);
  402 }
  403 
  404 /*
  405  * Packet routing routines.
  406  */
  407 void
  408 rtalloc_ign_fib(struct route *ro, u_long ignore, u_int fibnum)
  409 {
  410         struct rtentry *rt;
  411 
  412         if ((rt = ro->ro_rt) != NULL) {
  413                 if (rt->rt_ifp != NULL && rt->rt_flags & RTF_UP)
  414                         return;
  415                 RTFREE(rt);
  416                 ro->ro_rt = NULL;
  417         }
  418         ro->ro_rt = rtalloc1_fib(&ro->ro_dst, 1, ignore, fibnum);
  419         if (ro->ro_rt)
  420                 RT_UNLOCK(ro->ro_rt);
  421 }
  422 
  423 /*
  424  * Look up the route that matches the address given
  425  * Or, at least try.. Create a cloned route if needed.
  426  *
  427  * The returned route, if any, is locked.
  428  */
  429 struct rtentry *
  430 rtalloc1(struct sockaddr *dst, int report, u_long ignflags)
  431 {
  432 
  433         return (rtalloc1_fib(dst, report, ignflags, RT_DEFAULT_FIB));
  434 }
  435 
  436 struct rtentry *
  437 rtalloc1_fib(struct sockaddr *dst, int report, u_long ignflags,
  438                     u_int fibnum)
  439 {
  440         RIB_RLOCK_TRACKER;
  441         struct rib_head *rh;
  442         struct radix_node *rn;
  443         struct rtentry *newrt;
  444         struct rt_addrinfo info;
  445         int err = 0, msgtype = RTM_MISS;
  446 
  447         KASSERT((fibnum < rt_numfibs), ("rtalloc1_fib: bad fibnum"));
  448         rh = rt_tables_get_rnh(fibnum, dst->sa_family);
  449         newrt = NULL;
  450         if (rh == NULL)
  451                 goto miss;
  452 
  453         /*
  454          * Look up the address in the table for that Address Family
  455          */
  456         if ((ignflags & RTF_RNH_LOCKED) == 0)
  457                 RIB_RLOCK(rh);
  458 #ifdef INVARIANTS
  459         else
  460                 RIB_LOCK_ASSERT(rh);
  461 #endif
  462         rn = rh->rnh_matchaddr(dst, &rh->head);
  463         if (rn && ((rn->rn_flags & RNF_ROOT) == 0)) {
  464                 newrt = RNTORT(rn);
  465                 RT_LOCK(newrt);
  466                 RT_ADDREF(newrt);
  467                 if ((ignflags & RTF_RNH_LOCKED) == 0)
  468                         RIB_RUNLOCK(rh);
  469                 return (newrt);
  470 
  471         } else if ((ignflags & RTF_RNH_LOCKED) == 0)
  472                 RIB_RUNLOCK(rh);
  473         /*
  474          * Either we hit the root or could not find any match,
  475          * which basically means: "cannot get there from here".
  476          */
  477 miss:
  478         V_rtstat.rts_unreach++;
  479 
  480         if (report) {
  481                 /*
  482                  * If required, report the failure to the supervising
  483                  * Authorities.
  484                  * For a delete, this is not an error. (report == 0)
  485                  */
  486                 bzero(&info, sizeof(info));
  487                 info.rti_info[RTAX_DST] = dst;
  488                 rt_missmsg_fib(msgtype, &info, 0, err, fibnum);
  489         }
  490         return (newrt);
  491 }
  492 
  493 /*
  494  * Remove a reference count from an rtentry.
  495  * If the count gets low enough, take it out of the routing table
  496  */
  497 void
  498 rtfree(struct rtentry *rt)
  499 {
  500         struct rib_head *rnh;
  501 
  502         KASSERT(rt != NULL,("%s: NULL rt", __func__));
  503         rnh = rt_tables_get_rnh(rt->rt_fibnum, rt_key(rt)->sa_family);
  504         KASSERT(rnh != NULL,("%s: NULL rnh", __func__));
  505 
  506         RT_LOCK_ASSERT(rt);
  507 
  508         /*
  509          * The callers should use RTFREE_LOCKED() or RTFREE(), so
  510          * we should come here exactly with the last reference.
  511          */
  512         RT_REMREF(rt);
  513         if (rt->rt_refcnt > 0) {
  514                 log(LOG_DEBUG, "%s: %p has %d refs\n", __func__, rt, rt->rt_refcnt);
  515                 goto done;
  516         }
  517 
  518         /*
  519          * On last reference give the "close method" a chance
  520          * to cleanup private state.  This also permits (for
  521          * IPv4 and IPv6) a chance to decide if the routing table
  522          * entry should be purged immediately or at a later time.
  523          * When an immediate purge is to happen the close routine
  524          * typically calls rtexpunge which clears the RTF_UP flag
  525          * on the entry so that the code below reclaims the storage.
  526          */
  527         if (rt->rt_refcnt == 0 && rnh->rnh_close)
  528                 rnh->rnh_close((struct radix_node *)rt, &rnh->head);
  529 
  530         /*
  531          * If we are no longer "up" (and ref == 0)
  532          * then we can free the resources associated
  533          * with the route.
  534          */
  535         if ((rt->rt_flags & RTF_UP) == 0) {
  536                 if (rt->rt_nodes->rn_flags & (RNF_ACTIVE | RNF_ROOT))
  537                         panic("rtfree 2");
  538                 /*
  539                  * the rtentry must have been removed from the routing table
  540                  * so it is represented in rttrash.. remove that now.
  541                  */
  542                 V_rttrash--;
  543 #ifdef  DIAGNOSTIC
  544                 if (rt->rt_refcnt < 0) {
  545                         printf("rtfree: %p not freed (neg refs)\n", rt);
  546                         goto done;
  547                 }
  548 #endif
  549                 /*
  550                  * release references on items we hold them on..
  551                  * e.g other routes and ifaddrs.
  552                  */
  553                 if (rt->rt_ifa)
  554                         ifa_free(rt->rt_ifa);
  555                 /*
  556                  * The key is separatly alloc'd so free it (see rt_setgate()).
  557                  * This also frees the gateway, as they are always malloc'd
  558                  * together.
  559                  */
  560                 R_Free(rt_key(rt));
  561 
  562                 /*
  563                  * and the rtentry itself of course
  564                  */
  565                 uma_zfree(V_rtzone, rt);
  566                 return;
  567         }
  568 done:
  569         RT_UNLOCK(rt);
  570 }
  571 
  572 
  573 /*
  574  * Force a routing table entry to the specified
  575  * destination to go through the given gateway.
  576  * Normally called as a result of a routing redirect
  577  * message from the network layer.
  578  */
  579 void
  580 rtredirect_fib(struct sockaddr *dst,
  581         struct sockaddr *gateway,
  582         struct sockaddr *netmask,
  583         int flags,
  584         struct sockaddr *src,
  585         u_int fibnum)
  586 {
  587         struct rtentry *rt;
  588         int error = 0;
  589         short *stat = NULL;
  590         struct rt_addrinfo info;
  591         struct ifaddr *ifa;
  592         struct rib_head *rnh;
  593 
  594         ifa = NULL;
  595         NET_EPOCH_ENTER();
  596         rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
  597         if (rnh == NULL) {
  598                 error = EAFNOSUPPORT;
  599                 goto out;
  600         }
  601         /* verify the gateway is directly reachable */
  602         if ((ifa = ifa_ifwithnet(gateway, 0, fibnum)) == NULL) {
  603                 error = ENETUNREACH;
  604                 goto out;
  605         }
  606         rt = rtalloc1_fib(dst, 0, 0UL, fibnum); /* NB: rt is locked */
  607         /*
  608          * If the redirect isn't from our current router for this dst,
  609          * it's either old or wrong.  If it redirects us to ourselves,
  610          * we have a routing loop, perhaps as a result of an interface
  611          * going down recently.
  612          */
  613         if (!(flags & RTF_DONE) && rt) {
  614                 if (!sa_equal(src, rt->rt_gateway)) {
  615                         error = EINVAL;
  616                         goto done;
  617                 }
  618                 if (rt->rt_ifa != ifa && ifa->ifa_addr->sa_family != AF_LINK) {
  619                         error = EINVAL;
  620                         goto done;
  621                 }
  622         }
  623         if ((flags & RTF_GATEWAY) && ifa_ifwithaddr_check(gateway)) {
  624                 error = EHOSTUNREACH;
  625                 goto done;
  626         }
  627         /*
  628          * Create a new entry if we just got back a wildcard entry
  629          * or the lookup failed.  This is necessary for hosts
  630          * which use routing redirects generated by smart gateways
  631          * to dynamically build the routing tables.
  632          */
  633         if (rt == NULL || (rt_mask(rt) && rt_mask(rt)->sa_len < 2))
  634                 goto create;
  635         /*
  636          * Don't listen to the redirect if it's
  637          * for a route to an interface.
  638          */
  639         if (rt->rt_flags & RTF_GATEWAY) {
  640                 if (((rt->rt_flags & RTF_HOST) == 0) && (flags & RTF_HOST)) {
  641                         /*
  642                          * Changing from route to net => route to host.
  643                          * Create new route, rather than smashing route to net.
  644                          */
  645                 create:
  646                         if (rt != NULL)
  647                                 RTFREE_LOCKED(rt);
  648                 
  649                         flags |= RTF_DYNAMIC;
  650                         bzero((caddr_t)&info, sizeof(info));
  651                         info.rti_info[RTAX_DST] = dst;
  652                         info.rti_info[RTAX_GATEWAY] = gateway;
  653                         info.rti_info[RTAX_NETMASK] = netmask;
  654                         ifa_ref(ifa);
  655                         info.rti_ifa = ifa;
  656                         info.rti_flags = flags;
  657                         error = rtrequest1_fib(RTM_ADD, &info, &rt, fibnum);
  658                         if (rt != NULL) {
  659                                 RT_LOCK(rt);
  660                                 flags = rt->rt_flags;
  661                         }
  662                         
  663                         stat = &V_rtstat.rts_dynamic;
  664                 } else {
  665 
  666                         /*
  667                          * Smash the current notion of the gateway to
  668                          * this destination.  Should check about netmask!!!
  669                          */
  670                         if ((flags & RTF_GATEWAY) == 0)
  671                                 rt->rt_flags &= ~RTF_GATEWAY;
  672                         rt->rt_flags |= RTF_MODIFIED;
  673                         flags |= RTF_MODIFIED;
  674                         stat = &V_rtstat.rts_newgateway;
  675                         /*
  676                          * add the key and gateway (in one malloc'd chunk).
  677                          */
  678                         RT_UNLOCK(rt);
  679                         RIB_WLOCK(rnh);
  680                         RT_LOCK(rt);
  681                         rt_setgate(rt, rt_key(rt), gateway);
  682                         RIB_WUNLOCK(rnh);
  683                 }
  684         } else
  685                 error = EHOSTUNREACH;
  686 done:
  687         if (rt)
  688                 RTFREE_LOCKED(rt);
  689  out:
  690         NET_EPOCH_EXIT();
  691         if (error)
  692                 V_rtstat.rts_badredirect++;
  693         else if (stat != NULL)
  694                 (*stat)++;
  695         bzero((caddr_t)&info, sizeof(info));
  696         info.rti_info[RTAX_DST] = dst;
  697         info.rti_info[RTAX_GATEWAY] = gateway;
  698         info.rti_info[RTAX_NETMASK] = netmask;
  699         info.rti_info[RTAX_AUTHOR] = src;
  700         rt_missmsg_fib(RTM_REDIRECT, &info, flags, error, fibnum);
  701 }
  702 
  703 /*
  704  * Routing table ioctl interface.
  705  */
  706 int
  707 rtioctl_fib(u_long req, caddr_t data, u_int fibnum)
  708 {
  709 
  710         /*
  711          * If more ioctl commands are added here, make sure the proper
  712          * super-user checks are being performed because it is possible for
  713          * prison-root to make it this far if raw sockets have been enabled
  714          * in jails.
  715          */
  716 #ifdef INET
  717         /* Multicast goop, grrr... */
  718         return mrt_ioctl ? mrt_ioctl(req, data, fibnum) : EOPNOTSUPP;
  719 #else /* INET */
  720         return ENXIO;
  721 #endif /* INET */
  722 }
  723 
  724 struct ifaddr *
  725 ifa_ifwithroute(int flags, const struct sockaddr *dst, struct sockaddr *gateway,
  726                                 u_int fibnum)
  727 {
  728         struct ifaddr *ifa;
  729         int not_found = 0;
  730 
  731         MPASS(in_epoch(net_epoch_preempt));
  732         if ((flags & RTF_GATEWAY) == 0) {
  733                 /*
  734                  * If we are adding a route to an interface,
  735                  * and the interface is a pt to pt link
  736                  * we should search for the destination
  737                  * as our clue to the interface.  Otherwise
  738                  * we can use the local address.
  739                  */
  740                 ifa = NULL;
  741                 if (flags & RTF_HOST)
  742                         ifa = ifa_ifwithdstaddr(dst, fibnum);
  743                 if (ifa == NULL)
  744                         ifa = ifa_ifwithaddr(gateway);
  745         } else {
  746                 /*
  747                  * If we are adding a route to a remote net
  748                  * or host, the gateway may still be on the
  749                  * other end of a pt to pt link.
  750                  */
  751                 ifa = ifa_ifwithdstaddr(gateway, fibnum);
  752         }
  753         if (ifa == NULL)
  754                 ifa = ifa_ifwithnet(gateway, 0, fibnum);
  755         if (ifa == NULL) {
  756                 struct rtentry *rt;
  757 
  758                 rt = rtalloc1_fib(gateway, 0, flags, fibnum);
  759                 if (rt == NULL)
  760                         goto out;
  761                 /*
  762                  * dismiss a gateway that is reachable only
  763                  * through the default router
  764                  */
  765                 switch (gateway->sa_family) {
  766                 case AF_INET:
  767                         if (satosin(rt_key(rt))->sin_addr.s_addr == INADDR_ANY)
  768                                 not_found = 1;
  769                         break;
  770                 case AF_INET6:
  771                         if (IN6_IS_ADDR_UNSPECIFIED(&satosin6(rt_key(rt))->sin6_addr))
  772                                 not_found = 1;
  773                         break;
  774                 default:
  775                         break;
  776                 }
  777                 if (!not_found && rt->rt_ifa != NULL) {
  778                         ifa = rt->rt_ifa;
  779                 }
  780                 RT_REMREF(rt);
  781                 RT_UNLOCK(rt);
  782                 if (not_found || ifa == NULL)
  783                         goto out;
  784         }
  785         if (ifa->ifa_addr->sa_family != dst->sa_family) {
  786                 struct ifaddr *oifa = ifa;
  787                 ifa = ifaof_ifpforaddr(dst, ifa->ifa_ifp);
  788                 if (ifa == NULL)
  789                         ifa = oifa;
  790         }
  791  out:
  792         return (ifa);
  793 }
  794 
  795 /*
  796  * Do appropriate manipulations of a routing tree given
  797  * all the bits of info needed
  798  */
  799 int
  800 rtrequest_fib(int req,
  801         struct sockaddr *dst,
  802         struct sockaddr *gateway,
  803         struct sockaddr *netmask,
  804         int flags,
  805         struct rtentry **ret_nrt,
  806         u_int fibnum)
  807 {
  808         struct rt_addrinfo info;
  809 
  810         if (dst->sa_len == 0)
  811                 return(EINVAL);
  812 
  813         bzero((caddr_t)&info, sizeof(info));
  814         info.rti_flags = flags;
  815         info.rti_info[RTAX_DST] = dst;
  816         info.rti_info[RTAX_GATEWAY] = gateway;
  817         info.rti_info[RTAX_NETMASK] = netmask;
  818         return rtrequest1_fib(req, &info, ret_nrt, fibnum);
  819 }
  820 
  821 
  822 /*
  823  * Copy most of @rt data into @info.
  824  *
  825  * If @flags contains NHR_COPY, copies dst,netmask and gw to the
  826  * pointers specified by @info structure. Assume such pointers
  827  * are zeroed sockaddr-like structures with sa_len field initialized
  828  * to reflect size of the provided buffer. if no NHR_COPY is specified,
  829  * point dst,netmask and gw @info fields to appropriate @rt values.
  830  *
  831  * if @flags contains NHR_REF, do refcouting on rt_ifp and rt_ifa.
  832  *
  833  * Returns 0 on success.
  834  */
  835 int
  836 rt_exportinfo(struct rtentry *rt, struct rt_addrinfo *info, int flags)
  837 {
  838         struct rt_metrics *rmx;
  839         struct sockaddr *src, *dst;
  840         int sa_len;
  841 
  842         if (flags & NHR_COPY) {
  843                 /* Copy destination if dst is non-zero */
  844                 src = rt_key(rt);
  845                 dst = info->rti_info[RTAX_DST];
  846                 sa_len = src->sa_len;
  847                 if (dst != NULL) {
  848                         if (src->sa_len > dst->sa_len)
  849                                 return (ENOMEM);
  850                         memcpy(dst, src, src->sa_len);
  851                         info->rti_addrs |= RTA_DST;
  852                 }
  853 
  854                 /* Copy mask if set && dst is non-zero */
  855                 src = rt_mask(rt);
  856                 dst = info->rti_info[RTAX_NETMASK];
  857                 if (src != NULL && dst != NULL) {
  858 
  859                         /*
  860                          * Radix stores different value in sa_len,
  861                          * assume rt_mask() to have the same length
  862                          * as rt_key()
  863                          */
  864                         if (sa_len > dst->sa_len)
  865                                 return (ENOMEM);
  866                         memcpy(dst, src, src->sa_len);
  867                         info->rti_addrs |= RTA_NETMASK;
  868                 }
  869 
  870                 /* Copy gateway is set && dst is non-zero */
  871                 src = rt->rt_gateway;
  872                 dst = info->rti_info[RTAX_GATEWAY];
  873                 if ((rt->rt_flags & RTF_GATEWAY) && src != NULL && dst != NULL){
  874                         if (src->sa_len > dst->sa_len)
  875                                 return (ENOMEM);
  876                         memcpy(dst, src, src->sa_len);
  877                         info->rti_addrs |= RTA_GATEWAY;
  878                 }
  879         } else {
  880                 info->rti_info[RTAX_DST] = rt_key(rt);
  881                 info->rti_addrs |= RTA_DST;
  882                 if (rt_mask(rt) != NULL) {
  883                         info->rti_info[RTAX_NETMASK] = rt_mask(rt);
  884                         info->rti_addrs |= RTA_NETMASK;
  885                 }
  886                 if (rt->rt_flags & RTF_GATEWAY) {
  887                         info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
  888                         info->rti_addrs |= RTA_GATEWAY;
  889                 }
  890         }
  891 
  892         rmx = info->rti_rmx;
  893         if (rmx != NULL) {
  894                 info->rti_mflags |= RTV_MTU;
  895                 rmx->rmx_mtu = rt->rt_mtu;
  896         }
  897 
  898         info->rti_flags = rt->rt_flags;
  899         info->rti_ifp = rt->rt_ifp;
  900         info->rti_ifa = rt->rt_ifa;
  901         if (flags & NHR_REF) {
  902                 if_ref(info->rti_ifp);
  903                 ifa_ref(info->rti_ifa);
  904         }
  905 
  906         return (0);
  907 }
  908 
  909 /*
  910  * Lookups up route entry for @dst in RIB database for fib @fibnum.
  911  * Exports entry data to @info using rt_exportinfo().
  912  *
  913  * If @flags contains NHR_REF, refcouting is performed on rt_ifp and rt_ifa.
  914  * All references can be released later by calling rib_free_info().
  915  *
  916  * Returns 0 on success.
  917  * Returns ENOENT for lookup failure, ENOMEM for export failure.
  918  */
  919 int
  920 rib_lookup_info(uint32_t fibnum, const struct sockaddr *dst, uint32_t flags,
  921     uint32_t flowid, struct rt_addrinfo *info)
  922 {
  923         RIB_RLOCK_TRACKER;
  924         struct rib_head *rh;
  925         struct radix_node *rn;
  926         struct rtentry *rt;
  927         int error;
  928 
  929         KASSERT((fibnum < rt_numfibs), ("rib_lookup_rte: bad fibnum"));
  930         rh = rt_tables_get_rnh(fibnum, dst->sa_family);
  931         if (rh == NULL)
  932                 return (ENOENT);
  933 
  934         RIB_RLOCK(rh);
  935         rn = rh->rnh_matchaddr(__DECONST(void *, dst), &rh->head);
  936         if (rn != NULL && ((rn->rn_flags & RNF_ROOT) == 0)) {
  937                 rt = RNTORT(rn);
  938                 /* Ensure route & ifp is UP */
  939                 if (RT_LINK_IS_UP(rt->rt_ifp)) {
  940                         flags = (flags & NHR_REF) | NHR_COPY;
  941                         error = rt_exportinfo(rt, info, flags);
  942                         RIB_RUNLOCK(rh);
  943 
  944                         return (error);
  945                 }
  946         }
  947         RIB_RUNLOCK(rh);
  948 
  949         return (ENOENT);
  950 }
  951 
  952 /*
  953  * Releases all references acquired by rib_lookup_info() when
  954  * called with NHR_REF flags.
  955  */
  956 void
  957 rib_free_info(struct rt_addrinfo *info)
  958 {
  959 
  960         ifa_free(info->rti_ifa);
  961         if_rele(info->rti_ifp);
  962 }
  963 
  964 /*
  965  * Iterates over all existing fibs in system calling
  966  *  @setwa_f function prior to traversing each fib.
  967  *  Calls @wa_f function for each element in current fib.
  968  * If af is not AF_UNSPEC, iterates over fibs in particular
  969  * address family.
  970  */
  971 void
  972 rt_foreach_fib_walk(int af, rt_setwarg_t *setwa_f, rt_walktree_f_t *wa_f,
  973     void *arg)
  974 {
  975         struct rib_head *rnh;
  976         uint32_t fibnum;
  977         int i;
  978 
  979         for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
  980                 /* Do we want some specific family? */
  981                 if (af != AF_UNSPEC) {
  982                         rnh = rt_tables_get_rnh(fibnum, af);
  983                         if (rnh == NULL)
  984                                 continue;
  985                         if (setwa_f != NULL)
  986                                 setwa_f(rnh, fibnum, af, arg);
  987 
  988                         RIB_WLOCK(rnh);
  989                         rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
  990                         RIB_WUNLOCK(rnh);
  991                         continue;
  992                 }
  993 
  994                 for (i = 1; i <= AF_MAX; i++) {
  995                         rnh = rt_tables_get_rnh(fibnum, i);
  996                         if (rnh == NULL)
  997                                 continue;
  998                         if (setwa_f != NULL)
  999                                 setwa_f(rnh, fibnum, i, arg);
 1000 
 1001                         RIB_WLOCK(rnh);
 1002                         rnh->rnh_walktree(&rnh->head, (walktree_f_t *)wa_f,arg);
 1003                         RIB_WUNLOCK(rnh);
 1004                 }
 1005         }
 1006 }
 1007 
 1008 struct rt_delinfo
 1009 {
 1010         struct rt_addrinfo info;
 1011         struct rib_head *rnh;
 1012         struct rtentry *head;
 1013 };
 1014 
 1015 /*
 1016  * Conditionally unlinks @rn from radix tree based
 1017  * on info data passed in @arg.
 1018  */
 1019 static int
 1020 rt_checkdelroute(struct radix_node *rn, void *arg)
 1021 {
 1022         struct rt_delinfo *di;
 1023         struct rt_addrinfo *info;
 1024         struct rtentry *rt;
 1025         int error;
 1026 
 1027         di = (struct rt_delinfo *)arg;
 1028         rt = (struct rtentry *)rn;
 1029         info = &di->info;
 1030         error = 0;
 1031 
 1032         info->rti_info[RTAX_DST] = rt_key(rt);
 1033         info->rti_info[RTAX_NETMASK] = rt_mask(rt);
 1034         info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
 1035 
 1036         rt = rt_unlinkrte(di->rnh, info, &error);
 1037         if (rt == NULL) {
 1038                 /* Either not allowed or not matched. Skip entry */
 1039                 return (0);
 1040         }
 1041 
 1042         /* Entry was unlinked. Add to the list and return */
 1043         rt->rt_chain = di->head;
 1044         di->head = rt;
 1045 
 1046         return (0);
 1047 }
 1048 
 1049 /*
 1050  * Iterates over all existing fibs in system.
 1051  * Deletes each element for which @filter_f function returned
 1052  * non-zero value.
 1053  * If @af is not AF_UNSPEC, iterates over fibs in particular
 1054  * address family.
 1055  */
 1056 void
 1057 rt_foreach_fib_walk_del(int af, rt_filter_f_t *filter_f, void *arg)
 1058 {
 1059         struct rib_head *rnh;
 1060         struct rt_delinfo di;
 1061         struct rtentry *rt;
 1062         uint32_t fibnum;
 1063         int i, start, end;
 1064 
 1065         bzero(&di, sizeof(di));
 1066         di.info.rti_filter = filter_f;
 1067         di.info.rti_filterdata = arg;
 1068 
 1069         for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
 1070                 /* Do we want some specific family? */
 1071                 if (af != AF_UNSPEC) {
 1072                         start = af;
 1073                         end = af;
 1074                 } else {
 1075                         start = 1;
 1076                         end = AF_MAX;
 1077                 }
 1078 
 1079                 for (i = start; i <= end; i++) {
 1080                         rnh = rt_tables_get_rnh(fibnum, i);
 1081                         if (rnh == NULL)
 1082                                 continue;
 1083                         di.rnh = rnh;
 1084 
 1085                         RIB_WLOCK(rnh);
 1086                         rnh->rnh_walktree(&rnh->head, rt_checkdelroute, &di);
 1087                         RIB_WUNLOCK(rnh);
 1088 
 1089                         if (di.head == NULL)
 1090                                 continue;
 1091 
 1092                         /* We might have something to reclaim */
 1093                         while (di.head != NULL) {
 1094                                 rt = di.head;
 1095                                 di.head = rt->rt_chain;
 1096                                 rt->rt_chain = NULL;
 1097 
 1098                                 /* TODO std rt -> rt_addrinfo export */
 1099                                 di.info.rti_info[RTAX_DST] = rt_key(rt);
 1100                                 di.info.rti_info[RTAX_NETMASK] = rt_mask(rt);
 1101 
 1102                                 rt_notifydelete(rt, &di.info);
 1103                                 RTFREE_LOCKED(rt);
 1104                         }
 1105 
 1106                 }
 1107         }
 1108 }
 1109 
 1110 /*
 1111  * Delete Routes for a Network Interface
 1112  *
 1113  * Called for each routing entry via the rnh->rnh_walktree() call above
 1114  * to delete all route entries referencing a detaching network interface.
 1115  *
 1116  * Arguments:
 1117  *      rt      pointer to rtentry
 1118  *      arg     argument passed to rnh->rnh_walktree() - detaching interface
 1119  *
 1120  * Returns:
 1121  *      0       successful
 1122  *      errno   failed - reason indicated
 1123  */
 1124 static int
 1125 rt_ifdelroute(const struct rtentry *rt, void *arg)
 1126 {
 1127         struct ifnet    *ifp = arg;
 1128 
 1129         if (rt->rt_ifp != ifp)
 1130                 return (0);
 1131 
 1132         /*
 1133          * Protect (sorta) against walktree recursion problems
 1134          * with cloned routes
 1135          */
 1136         if ((rt->rt_flags & RTF_UP) == 0)
 1137                 return (0);
 1138 
 1139         return (1);
 1140 }
 1141 
 1142 /*
 1143  * Delete all remaining routes using this interface
 1144  * Unfortuneatly the only way to do this is to slog through
 1145  * the entire routing table looking for routes which point
 1146  * to this interface...oh well...
 1147  */
 1148 void
 1149 rt_flushifroutes_af(struct ifnet *ifp, int af)
 1150 {
 1151         KASSERT((af >= 1 && af <= AF_MAX), ("%s: af %d not >= 1 and <= %d",
 1152             __func__, af, AF_MAX));
 1153 
 1154         rt_foreach_fib_walk_del(af, rt_ifdelroute, ifp);
 1155 }
 1156 
 1157 void
 1158 rt_flushifroutes(struct ifnet *ifp)
 1159 {
 1160 
 1161         rt_foreach_fib_walk_del(AF_UNSPEC, rt_ifdelroute, ifp);
 1162 }
 1163 
 1164 /*
 1165  * Conditionally unlinks rtentry matching data inside @info from @rnh.
 1166  * Returns unlinked, locked and referenced @rtentry on success,
 1167  * Returns NULL and sets @perror to:
 1168  * ESRCH - if prefix was not found,
 1169  * EADDRINUSE - if trying to delete PINNED route without appropriate flag.
 1170  * ENOENT - if supplied filter function returned 0 (not matched).
 1171  */
 1172 static struct rtentry *
 1173 rt_unlinkrte(struct rib_head *rnh, struct rt_addrinfo *info, int *perror)
 1174 {
 1175         struct sockaddr *dst, *netmask;
 1176         struct rtentry *rt;
 1177         struct radix_node *rn;
 1178 
 1179         dst = info->rti_info[RTAX_DST];
 1180         netmask = info->rti_info[RTAX_NETMASK];
 1181 
 1182         rt = (struct rtentry *)rnh->rnh_lookup(dst, netmask, &rnh->head);
 1183         if (rt == NULL) {
 1184                 *perror = ESRCH;
 1185                 return (NULL);
 1186         }
 1187 
 1188         if ((info->rti_flags & RTF_PINNED) == 0) {
 1189                 /* Check if target route can be deleted */
 1190                 if (rt->rt_flags & RTF_PINNED) {
 1191                         *perror = EADDRINUSE;
 1192                         return (NULL);
 1193                 }
 1194         }
 1195 
 1196         if (info->rti_filter != NULL) {
 1197                 if (info->rti_filter(rt, info->rti_filterdata) == 0) {
 1198                         /* Not matched */
 1199                         *perror = ENOENT;
 1200                         return (NULL);
 1201                 }
 1202 
 1203                 /*
 1204                  * Filter function requested rte deletion.
 1205                  * Ease the caller work by filling in remaining info
 1206                  * from that particular entry.
 1207                  */
 1208                 info->rti_info[RTAX_GATEWAY] = rt->rt_gateway;
 1209         }
 1210 
 1211         /*
 1212          * Remove the item from the tree and return it.
 1213          * Complain if it is not there and do no more processing.
 1214          */
 1215         *perror = ESRCH;
 1216 #ifdef RADIX_MPATH
 1217         if (rt_mpath_capable(rnh))
 1218                 rn = rt_mpath_unlink(rnh, info, rt, perror);
 1219         else
 1220 #endif
 1221         rn = rnh->rnh_deladdr(dst, netmask, &rnh->head);
 1222         if (rn == NULL)
 1223                 return (NULL);
 1224 
 1225         if (rn->rn_flags & (RNF_ACTIVE | RNF_ROOT))
 1226                 panic ("rtrequest delete");
 1227 
 1228         rt = RNTORT(rn);
 1229         RT_LOCK(rt);
 1230         RT_ADDREF(rt);
 1231         rt->rt_flags &= ~RTF_UP;
 1232 
 1233         *perror = 0;
 1234 
 1235         return (rt);
 1236 }
 1237 
 1238 static void
 1239 rt_notifydelete(struct rtentry *rt, struct rt_addrinfo *info)
 1240 {
 1241         struct ifaddr *ifa;
 1242 
 1243         /*
 1244          * give the protocol a chance to keep things in sync.
 1245          */
 1246         ifa = rt->rt_ifa;
 1247         if (ifa != NULL && ifa->ifa_rtrequest != NULL)
 1248                 ifa->ifa_rtrequest(RTM_DELETE, rt, info);
 1249 
 1250         /*
 1251          * One more rtentry floating around that is not
 1252          * linked to the routing table. rttrash will be decremented
 1253          * when RTFREE(rt) is eventually called.
 1254          */
 1255         V_rttrash++;
 1256 }
 1257 
 1258 
 1259 /*
 1260  * These (questionable) definitions of apparent local variables apply
 1261  * to the next two functions.  XXXXXX!!!
 1262  */
 1263 #define dst     info->rti_info[RTAX_DST]
 1264 #define gateway info->rti_info[RTAX_GATEWAY]
 1265 #define netmask info->rti_info[RTAX_NETMASK]
 1266 #define ifaaddr info->rti_info[RTAX_IFA]
 1267 #define ifpaddr info->rti_info[RTAX_IFP]
 1268 #define flags   info->rti_flags
 1269 
 1270 /*
 1271  * Look up rt_addrinfo for a specific fib.  Note that if rti_ifa is defined,
 1272  * it will be referenced so the caller must free it.
 1273  */
 1274 int
 1275 rt_getifa_fib(struct rt_addrinfo *info, u_int fibnum)
 1276 {
 1277         struct ifaddr *ifa;
 1278         int needref, error;
 1279 
 1280         /*
 1281          * ifp may be specified by sockaddr_dl
 1282          * when protocol address is ambiguous.
 1283          */
 1284         error = 0;
 1285         needref = (info->rti_ifa == NULL);
 1286         NET_EPOCH_ENTER();
 1287         if (info->rti_ifp == NULL && ifpaddr != NULL &&
 1288             ifpaddr->sa_family == AF_LINK &&
 1289             (ifa = ifa_ifwithnet(ifpaddr, 0, fibnum)) != NULL) {
 1290                 info->rti_ifp = ifa->ifa_ifp;
 1291         }
 1292         if (info->rti_ifa == NULL && ifaaddr != NULL)
 1293                 info->rti_ifa = ifa_ifwithaddr(ifaaddr);
 1294         if (info->rti_ifa == NULL) {
 1295                 struct sockaddr *sa;
 1296 
 1297                 sa = ifaaddr != NULL ? ifaaddr :
 1298                     (gateway != NULL ? gateway : dst);
 1299                 if (sa != NULL && info->rti_ifp != NULL)
 1300                         info->rti_ifa = ifaof_ifpforaddr(sa, info->rti_ifp);
 1301                 else if (dst != NULL && gateway != NULL)
 1302                         info->rti_ifa = ifa_ifwithroute(flags, dst, gateway,
 1303                                                         fibnum);
 1304                 else if (sa != NULL)
 1305                         info->rti_ifa = ifa_ifwithroute(flags, sa, sa,
 1306                                                         fibnum);
 1307         }
 1308         if (needref && info->rti_ifa != NULL) {
 1309                 if (info->rti_ifp == NULL)
 1310                         info->rti_ifp = info->rti_ifa->ifa_ifp;
 1311                 ifa_ref(info->rti_ifa);
 1312         } else
 1313                 error = ENETUNREACH;
 1314         NET_EPOCH_EXIT();
 1315         return (error);
 1316 }
 1317 
 1318 static int
 1319 if_updatemtu_cb(struct radix_node *rn, void *arg)
 1320 {
 1321         struct rtentry *rt;
 1322         struct if_mtuinfo *ifmtu;
 1323 
 1324         rt = (struct rtentry *)rn;
 1325         ifmtu = (struct if_mtuinfo *)arg;
 1326 
 1327         if (rt->rt_ifp != ifmtu->ifp)
 1328                 return (0);
 1329 
 1330         if (rt->rt_mtu >= ifmtu->mtu) {
 1331                 /* We have to decrease mtu regardless of flags */
 1332                 rt->rt_mtu = ifmtu->mtu;
 1333                 return (0);
 1334         }
 1335 
 1336         /*
 1337          * New MTU is bigger. Check if are allowed to alter it
 1338          */
 1339         if ((rt->rt_flags & (RTF_FIXEDMTU | RTF_GATEWAY | RTF_HOST)) != 0) {
 1340 
 1341                 /*
 1342                  * Skip routes with user-supplied MTU and
 1343                  * non-interface routes
 1344                  */
 1345                 return (0);
 1346         }
 1347 
 1348         /* We are safe to update route MTU */
 1349         rt->rt_mtu = ifmtu->mtu;
 1350 
 1351         return (0);
 1352 }
 1353 
 1354 void
 1355 rt_updatemtu(struct ifnet *ifp)
 1356 {
 1357         struct if_mtuinfo ifmtu;
 1358         struct rib_head *rnh;
 1359         int i, j;
 1360 
 1361         ifmtu.ifp = ifp;
 1362 
 1363         /*
 1364          * Try to update rt_mtu for all routes using this interface
 1365          * Unfortunately the only way to do this is to traverse all
 1366          * routing tables in all fibs/domains.
 1367          */
 1368         for (i = 1; i <= AF_MAX; i++) {
 1369                 ifmtu.mtu = if_getmtu_family(ifp, i);
 1370                 for (j = 0; j < rt_numfibs; j++) {
 1371                         rnh = rt_tables_get_rnh(j, i);
 1372                         if (rnh == NULL)
 1373                                 continue;
 1374                         RIB_WLOCK(rnh);
 1375                         rnh->rnh_walktree(&rnh->head, if_updatemtu_cb, &ifmtu);
 1376                         RIB_WUNLOCK(rnh);
 1377                 }
 1378         }
 1379 }
 1380 
 1381 
 1382 #if 0
 1383 int p_sockaddr(char *buf, int buflen, struct sockaddr *s);
 1384 int rt_print(char *buf, int buflen, struct rtentry *rt);
 1385 
 1386 int
 1387 p_sockaddr(char *buf, int buflen, struct sockaddr *s)
 1388 {
 1389         void *paddr = NULL;
 1390 
 1391         switch (s->sa_family) {
 1392         case AF_INET:
 1393                 paddr = &((struct sockaddr_in *)s)->sin_addr;
 1394                 break;
 1395         case AF_INET6:
 1396                 paddr = &((struct sockaddr_in6 *)s)->sin6_addr;
 1397                 break;
 1398         }
 1399 
 1400         if (paddr == NULL)
 1401                 return (0);
 1402 
 1403         if (inet_ntop(s->sa_family, paddr, buf, buflen) == NULL)
 1404                 return (0);
 1405         
 1406         return (strlen(buf));
 1407 }
 1408 
 1409 int
 1410 rt_print(char *buf, int buflen, struct rtentry *rt)
 1411 {
 1412         struct sockaddr *addr, *mask;
 1413         int i = 0;
 1414 
 1415         addr = rt_key(rt);
 1416         mask = rt_mask(rt);
 1417 
 1418         i = p_sockaddr(buf, buflen, addr);
 1419         if (!(rt->rt_flags & RTF_HOST)) {
 1420                 buf[i++] = '/';
 1421                 i += p_sockaddr(buf + i, buflen - i, mask);
 1422         }
 1423 
 1424         if (rt->rt_flags & RTF_GATEWAY) {
 1425                 buf[i++] = '>';
 1426                 i += p_sockaddr(buf + i, buflen - i, rt->rt_gateway);
 1427         }
 1428 
 1429         return (i);
 1430 }
 1431 #endif
 1432 
 1433 #ifdef RADIX_MPATH
 1434 /*
 1435  * Deletes key for single-path routes, unlinks rtentry with
 1436  * gateway specified in @info from multi-path routes.
 1437  *
 1438  * Returnes unlinked entry. In case of failure, returns NULL
 1439  * and sets @perror to ESRCH.
 1440  */
 1441 static struct radix_node *
 1442 rt_mpath_unlink(struct rib_head *rnh, struct rt_addrinfo *info,
 1443     struct rtentry *rto, int *perror)
 1444 {
 1445         /*
 1446          * if we got multipath routes, we require users to specify
 1447          * a matching RTAX_GATEWAY.
 1448          */
 1449         struct rtentry *rt; // *rto = NULL;
 1450         struct radix_node *rn;
 1451         struct sockaddr *gw;
 1452 
 1453         gw = info->rti_info[RTAX_GATEWAY];
 1454         rt = rt_mpath_matchgate(rto, gw);
 1455         if (rt == NULL) {
 1456                 *perror = ESRCH;
 1457                 return (NULL);
 1458         }
 1459 
 1460         /*
 1461          * this is the first entry in the chain
 1462          */
 1463         if (rto == rt) {
 1464                 rn = rn_mpath_next((struct radix_node *)rt);
 1465                 /*
 1466                  * there is another entry, now it's active
 1467                  */
 1468                 if (rn) {
 1469                         rto = RNTORT(rn);
 1470                         RT_LOCK(rto);
 1471                         rto->rt_flags |= RTF_UP;
 1472                         RT_UNLOCK(rto);
 1473                 } else if (rt->rt_flags & RTF_GATEWAY) {
 1474                         /*
 1475                          * For gateway routes, we need to 
 1476                          * make sure that we we are deleting
 1477                          * the correct gateway. 
 1478                          * rt_mpath_matchgate() does not 
 1479                          * check the case when there is only
 1480                          * one route in the chain.  
 1481                          */
 1482                         if (gw &&
 1483                             (rt->rt_gateway->sa_len != gw->sa_len ||
 1484                                 memcmp(rt->rt_gateway, gw, gw->sa_len))) {
 1485                                 *perror = ESRCH;
 1486                                 return (NULL);
 1487                         }
 1488                 }
 1489 
 1490                 /*
 1491                  * use the normal delete code to remove
 1492                  * the first entry
 1493                  */
 1494                 rn = rnh->rnh_deladdr(dst, netmask, &rnh->head);
 1495                 *perror = 0;
 1496                 return (rn);
 1497         }
 1498                 
 1499         /*
 1500          * if the entry is 2nd and on up
 1501          */
 1502         if (rt_mpath_deldup(rto, rt) == 0)
 1503                 panic ("rtrequest1: rt_mpath_deldup");
 1504         *perror = 0;
 1505         rn = (struct radix_node *)rt;
 1506         return (rn);
 1507 }
 1508 #endif
 1509 
 1510 int
 1511 rtrequest1_fib(int req, struct rt_addrinfo *info, struct rtentry **ret_nrt,
 1512                                 u_int fibnum)
 1513 {
 1514         int error = 0;
 1515         struct rtentry *rt, *rt_old;
 1516         struct radix_node *rn;
 1517         struct rib_head *rnh;
 1518         struct ifaddr *ifa;
 1519         struct sockaddr *ndst;
 1520         struct sockaddr_storage mdst;
 1521 
 1522         KASSERT((fibnum < rt_numfibs), ("rtrequest1_fib: bad fibnum"));
 1523         KASSERT((flags & RTF_RNH_LOCKED) == 0, ("rtrequest1_fib: locked"));
 1524         switch (dst->sa_family) {
 1525         case AF_INET6:
 1526         case AF_INET:
 1527                 /* We support multiple FIBs. */
 1528                 break;
 1529         default:
 1530                 fibnum = RT_DEFAULT_FIB;
 1531                 break;
 1532         }
 1533 
 1534         /*
 1535          * Find the correct routing tree to use for this Address Family
 1536          */
 1537         rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
 1538         if (rnh == NULL)
 1539                 return (EAFNOSUPPORT);
 1540 
 1541         /*
 1542          * If we are adding a host route then we don't want to put
 1543          * a netmask in the tree, nor do we want to clone it.
 1544          */
 1545         if (flags & RTF_HOST)
 1546                 netmask = NULL;
 1547 
 1548         switch (req) {
 1549         case RTM_DELETE:
 1550                 if (netmask) {
 1551                         if (dst->sa_len > sizeof(mdst))
 1552                                 return (EINVAL);
 1553                         rt_maskedcopy(dst, (struct sockaddr *)&mdst, netmask);
 1554                         dst = (struct sockaddr *)&mdst;
 1555                 }
 1556 
 1557                 RIB_WLOCK(rnh);
 1558                 rt = rt_unlinkrte(rnh, info, &error);
 1559                 RIB_WUNLOCK(rnh);
 1560                 if (error != 0)
 1561                         return (error);
 1562 
 1563                 rt_notifydelete(rt, info);
 1564 
 1565                 /*
 1566                  * If the caller wants it, then it can have it,
 1567                  * but it's up to it to free the rtentry as we won't be
 1568                  * doing it.
 1569                  */
 1570                 if (ret_nrt) {
 1571                         *ret_nrt = rt;
 1572                         RT_UNLOCK(rt);
 1573                 } else
 1574                         RTFREE_LOCKED(rt);
 1575                 break;
 1576         case RTM_RESOLVE:
 1577                 /*
 1578                  * resolve was only used for route cloning
 1579                  * here for compat
 1580                  */
 1581                 break;
 1582         case RTM_ADD:
 1583                 if ((flags & RTF_GATEWAY) && !gateway)
 1584                         return (EINVAL);
 1585                 if (dst && gateway && (dst->sa_family != gateway->sa_family) && 
 1586                     (gateway->sa_family != AF_UNSPEC) && (gateway->sa_family != AF_LINK))
 1587                         return (EINVAL);
 1588 
 1589                 if (info->rti_ifa == NULL) {
 1590                         error = rt_getifa_fib(info, fibnum);
 1591                         if (error)
 1592                                 return (error);
 1593                 } else {
 1594                         ifa_ref(info->rti_ifa);
 1595                 }
 1596                 rt = uma_zalloc(V_rtzone, M_NOWAIT);
 1597                 if (rt == NULL) {
 1598                         ifa_free(info->rti_ifa);
 1599                         return (ENOBUFS);
 1600                 }
 1601                 rt->rt_flags = RTF_UP | flags;
 1602                 rt->rt_fibnum = fibnum;
 1603                 /*
 1604                  * Add the gateway. Possibly re-malloc-ing the storage for it.
 1605                  */
 1606                 if ((error = rt_setgate(rt, dst, gateway)) != 0) {
 1607                         ifa_free(info->rti_ifa);
 1608                         uma_zfree(V_rtzone, rt);
 1609                         return (error);
 1610                 }
 1611 
 1612                 /*
 1613                  * point to the (possibly newly malloc'd) dest address.
 1614                  */
 1615                 ndst = (struct sockaddr *)rt_key(rt);
 1616 
 1617                 /*
 1618                  * make sure it contains the value we want (masked if needed).
 1619                  */
 1620                 if (netmask) {
 1621                         rt_maskedcopy(dst, ndst, netmask);
 1622                 } else
 1623                         bcopy(dst, ndst, dst->sa_len);
 1624 
 1625                 /*
 1626                  * We use the ifa reference returned by rt_getifa_fib().
 1627                  * This moved from below so that rnh->rnh_addaddr() can
 1628                  * examine the ifa and  ifa->ifa_ifp if it so desires.
 1629                  */
 1630                 ifa = info->rti_ifa;
 1631                 rt->rt_ifa = ifa;
 1632                 rt->rt_ifp = ifa->ifa_ifp;
 1633                 rt->rt_weight = 1;
 1634 
 1635                 rt_setmetrics(info, rt);
 1636 
 1637                 RIB_WLOCK(rnh);
 1638                 RT_LOCK(rt);
 1639 #ifdef RADIX_MPATH
 1640                 /* do not permit exactly the same dst/mask/gw pair */
 1641                 if (rt_mpath_capable(rnh) &&
 1642                         rt_mpath_conflict(rnh, rt, netmask)) {
 1643                         RIB_WUNLOCK(rnh);
 1644 
 1645                         ifa_free(rt->rt_ifa);
 1646                         R_Free(rt_key(rt));
 1647                         uma_zfree(V_rtzone, rt);
 1648                         return (EEXIST);
 1649                 }
 1650 #endif
 1651 
 1652                 /* XXX mtu manipulation will be done in rnh_addaddr -- itojun */
 1653                 rn = rnh->rnh_addaddr(ndst, netmask, &rnh->head, rt->rt_nodes);
 1654 
 1655                 rt_old = NULL;
 1656                 if (rn == NULL && (info->rti_flags & RTF_PINNED) != 0) {
 1657 
 1658                         /*
 1659                          * Force removal and re-try addition
 1660                          * TODO: better multipath&pinned support
 1661                          */
 1662                         struct sockaddr *info_dst = info->rti_info[RTAX_DST];
 1663                         info->rti_info[RTAX_DST] = ndst;
 1664                         /* Do not delete existing PINNED(interface) routes */
 1665                         info->rti_flags &= ~RTF_PINNED;
 1666                         rt_old = rt_unlinkrte(rnh, info, &error);
 1667                         info->rti_flags |= RTF_PINNED;
 1668                         info->rti_info[RTAX_DST] = info_dst;
 1669                         if (rt_old != NULL)
 1670                                 rn = rnh->rnh_addaddr(ndst, netmask, &rnh->head,
 1671                                     rt->rt_nodes);
 1672                 }
 1673                 RIB_WUNLOCK(rnh);
 1674 
 1675                 if (rt_old != NULL)
 1676                         RT_UNLOCK(rt_old);
 1677 
 1678                 /*
 1679                  * If it still failed to go into the tree,
 1680                  * then un-make it (this should be a function)
 1681                  */
 1682                 if (rn == NULL) {
 1683                         ifa_free(rt->rt_ifa);
 1684                         R_Free(rt_key(rt));
 1685                         uma_zfree(V_rtzone, rt);
 1686                         return (EEXIST);
 1687                 } 
 1688 
 1689                 if (rt_old != NULL) {
 1690                         rt_notifydelete(rt_old, info);
 1691                         RTFREE(rt_old);
 1692                 }
 1693 
 1694                 /*
 1695                  * If this protocol has something to add to this then
 1696                  * allow it to do that as well.
 1697                  */
 1698                 if (ifa->ifa_rtrequest)
 1699                         ifa->ifa_rtrequest(req, rt, info);
 1700 
 1701                 /*
 1702                  * actually return a resultant rtentry and
 1703                  * give the caller a single reference.
 1704                  */
 1705                 if (ret_nrt) {
 1706                         *ret_nrt = rt;
 1707                         RT_ADDREF(rt);
 1708                 }
 1709                 rnh->rnh_gen++;         /* Routing table updated */
 1710                 RT_UNLOCK(rt);
 1711                 break;
 1712         case RTM_CHANGE:
 1713                 RIB_WLOCK(rnh);
 1714                 error = rtrequest1_fib_change(rnh, info, ret_nrt, fibnum);
 1715                 RIB_WUNLOCK(rnh);
 1716                 break;
 1717         default:
 1718                 error = EOPNOTSUPP;
 1719         }
 1720 
 1721         return (error);
 1722 }
 1723 
 1724 #undef dst
 1725 #undef gateway
 1726 #undef netmask
 1727 #undef ifaaddr
 1728 #undef ifpaddr
 1729 #undef flags
 1730 
 1731 static int
 1732 rtrequest1_fib_change(struct rib_head *rnh, struct rt_addrinfo *info,
 1733     struct rtentry **ret_nrt, u_int fibnum)
 1734 {
 1735         struct rtentry *rt = NULL;
 1736         int error = 0;
 1737         int free_ifa = 0;
 1738         int family, mtu;
 1739         struct if_mtuinfo ifmtu;
 1740 
 1741         RIB_WLOCK_ASSERT(rnh);
 1742 
 1743         rt = (struct rtentry *)rnh->rnh_lookup(info->rti_info[RTAX_DST],
 1744             info->rti_info[RTAX_NETMASK], &rnh->head);
 1745 
 1746         if (rt == NULL)
 1747                 return (ESRCH);
 1748 
 1749 #ifdef RADIX_MPATH
 1750         /*
 1751          * If we got multipath routes,
 1752          * we require users to specify a matching RTAX_GATEWAY.
 1753          */
 1754         if (rt_mpath_capable(rnh)) {
 1755                 rt = rt_mpath_matchgate(rt, info->rti_info[RTAX_GATEWAY]);
 1756                 if (rt == NULL)
 1757                         return (ESRCH);
 1758         }
 1759 #endif
 1760 
 1761         RT_LOCK(rt);
 1762 
 1763         rt_setmetrics(info, rt);
 1764 
 1765         /*
 1766          * New gateway could require new ifaddr, ifp;
 1767          * flags may also be different; ifp may be specified
 1768          * by ll sockaddr when protocol address is ambiguous
 1769          */
 1770         if (((rt->rt_flags & RTF_GATEWAY) &&
 1771             info->rti_info[RTAX_GATEWAY] != NULL) ||
 1772             info->rti_info[RTAX_IFP] != NULL ||
 1773             (info->rti_info[RTAX_IFA] != NULL &&
 1774              !sa_equal(info->rti_info[RTAX_IFA], rt->rt_ifa->ifa_addr))) {
 1775                 /*
 1776                  * XXX: Temporarily set RTF_RNH_LOCKED flag in the rti_flags
 1777                  *      to avoid rlock in the ifa_ifwithroute().
 1778                  */
 1779                 info->rti_flags |= RTF_RNH_LOCKED;
 1780                 error = rt_getifa_fib(info, fibnum);
 1781                 info->rti_flags &= ~RTF_RNH_LOCKED;
 1782                 if (info->rti_ifa != NULL)
 1783                         free_ifa = 1;
 1784 
 1785                 if (error != 0)
 1786                         goto bad;
 1787         }
 1788 
 1789         /* Check if outgoing interface has changed */
 1790         if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa &&
 1791             rt->rt_ifa != NULL) {
 1792                 if (rt->rt_ifa->ifa_rtrequest != NULL)
 1793                         rt->rt_ifa->ifa_rtrequest(RTM_DELETE, rt, info);
 1794                 ifa_free(rt->rt_ifa);
 1795                 rt->rt_ifa = NULL;
 1796         }
 1797         /* Update gateway address */
 1798         if (info->rti_info[RTAX_GATEWAY] != NULL) {
 1799                 error = rt_setgate(rt, rt_key(rt), info->rti_info[RTAX_GATEWAY]);
 1800                 if (error != 0)
 1801                         goto bad;
 1802 
 1803                 rt->rt_flags &= ~RTF_GATEWAY;
 1804                 rt->rt_flags |= (RTF_GATEWAY & info->rti_flags);
 1805         }
 1806 
 1807         if (info->rti_ifa != NULL && info->rti_ifa != rt->rt_ifa) {
 1808                 ifa_ref(info->rti_ifa);
 1809                 rt->rt_ifa = info->rti_ifa;
 1810                 rt->rt_ifp = info->rti_ifp;
 1811         }
 1812         /* Allow some flags to be toggled on change. */
 1813         rt->rt_flags &= ~RTF_FMASK;
 1814         rt->rt_flags |= info->rti_flags & RTF_FMASK;
 1815 
 1816         if (rt->rt_ifa && rt->rt_ifa->ifa_rtrequest != NULL)
 1817                rt->rt_ifa->ifa_rtrequest(RTM_ADD, rt, info);
 1818 
 1819         /* Alter route MTU if necessary */
 1820         if (rt->rt_ifp != NULL) {
 1821                 family = info->rti_info[RTAX_DST]->sa_family;
 1822                 mtu = if_getmtu_family(rt->rt_ifp, family);
 1823                 /* Set default MTU */
 1824                 if (rt->rt_mtu == 0)
 1825                         rt->rt_mtu = mtu;
 1826                 if (rt->rt_mtu != mtu) {
 1827                         /* Check if we really need to update */
 1828                         ifmtu.ifp = rt->rt_ifp;
 1829                         ifmtu.mtu = mtu;
 1830                         if_updatemtu_cb(rt->rt_nodes, &ifmtu);
 1831                 }
 1832         }
 1833 
 1834         /*
 1835          * This route change may have modified the route's gateway.  In that
 1836          * case, any inpcbs that have cached this route need to invalidate their
 1837          * llentry cache.
 1838          */
 1839         rnh->rnh_gen++;
 1840 
 1841         if (ret_nrt) {
 1842                 *ret_nrt = rt;
 1843                 RT_ADDREF(rt);
 1844         }
 1845 bad:
 1846         RT_UNLOCK(rt);
 1847         if (free_ifa != 0) {
 1848                 ifa_free(info->rti_ifa);
 1849                 info->rti_ifa = NULL;
 1850         }
 1851         return (error);
 1852 }
 1853 
 1854 static void
 1855 rt_setmetrics(const struct rt_addrinfo *info, struct rtentry *rt)
 1856 {
 1857 
 1858         if (info->rti_mflags & RTV_MTU) {
 1859                 if (info->rti_rmx->rmx_mtu != 0) {
 1860 
 1861                         /*
 1862                          * MTU was explicitly provided by user.
 1863                          * Keep it.
 1864                          */
 1865                         rt->rt_flags |= RTF_FIXEDMTU;
 1866                 } else {
 1867 
 1868                         /*
 1869                          * User explicitly sets MTU to 0.
 1870                          * Assume rollback to default.
 1871                          */
 1872                         rt->rt_flags &= ~RTF_FIXEDMTU;
 1873                 }
 1874                 rt->rt_mtu = info->rti_rmx->rmx_mtu;
 1875         }
 1876         if (info->rti_mflags & RTV_WEIGHT)
 1877                 rt->rt_weight = info->rti_rmx->rmx_weight;
 1878         /* Kernel -> userland timebase conversion. */
 1879         if (info->rti_mflags & RTV_EXPIRE)
 1880                 rt->rt_expire = info->rti_rmx->rmx_expire ?
 1881                     info->rti_rmx->rmx_expire - time_second + time_uptime : 0;
 1882 }
 1883 
 1884 int
 1885 rt_setgate(struct rtentry *rt, struct sockaddr *dst, struct sockaddr *gate)
 1886 {
 1887         /* XXX dst may be overwritten, can we move this to below */
 1888         int dlen = SA_SIZE(dst), glen = SA_SIZE(gate);
 1889 
 1890         /*
 1891          * Prepare to store the gateway in rt->rt_gateway.
 1892          * Both dst and gateway are stored one after the other in the same
 1893          * malloc'd chunk. If we have room, we can reuse the old buffer,
 1894          * rt_gateway already points to the right place.
 1895          * Otherwise, malloc a new block and update the 'dst' address.
 1896          */
 1897         if (rt->rt_gateway == NULL || glen > SA_SIZE(rt->rt_gateway)) {
 1898                 caddr_t new;
 1899 
 1900                 R_Malloc(new, caddr_t, dlen + glen);
 1901                 if (new == NULL)
 1902                         return ENOBUFS;
 1903                 /*
 1904                  * XXX note, we copy from *dst and not *rt_key(rt) because
 1905                  * rt_setgate() can be called to initialize a newly
 1906                  * allocated route entry, in which case rt_key(rt) == NULL
 1907                  * (and also rt->rt_gateway == NULL).
 1908                  * Free()/free() handle a NULL argument just fine.
 1909                  */
 1910                 bcopy(dst, new, dlen);
 1911                 R_Free(rt_key(rt));     /* free old block, if any */
 1912                 rt_key(rt) = (struct sockaddr *)new;
 1913                 rt->rt_gateway = (struct sockaddr *)(new + dlen);
 1914         }
 1915 
 1916         /*
 1917          * Copy the new gateway value into the memory chunk.
 1918          */
 1919         bcopy(gate, rt->rt_gateway, glen);
 1920 
 1921         return (0);
 1922 }
 1923 
 1924 void
 1925 rt_maskedcopy(struct sockaddr *src, struct sockaddr *dst, struct sockaddr *netmask)
 1926 {
 1927         u_char *cp1 = (u_char *)src;
 1928         u_char *cp2 = (u_char *)dst;
 1929         u_char *cp3 = (u_char *)netmask;
 1930         u_char *cplim = cp2 + *cp3;
 1931         u_char *cplim2 = cp2 + *cp1;
 1932 
 1933         *cp2++ = *cp1++; *cp2++ = *cp1++; /* copies sa_len & sa_family */
 1934         cp3 += 2;
 1935         if (cplim > cplim2)
 1936                 cplim = cplim2;
 1937         while (cp2 < cplim)
 1938                 *cp2++ = *cp1++ & *cp3++;
 1939         if (cp2 < cplim2)
 1940                 bzero((caddr_t)cp2, (unsigned)(cplim2 - cp2));
 1941 }
 1942 
 1943 /*
 1944  * Set up a routing table entry, normally
 1945  * for an interface.
 1946  */
 1947 #define _SOCKADDR_TMPSIZE 128 /* Not too big.. kernel stack size is limited */
 1948 static inline  int
 1949 rtinit1(struct ifaddr *ifa, int cmd, int flags, int fibnum)
 1950 {
 1951         RIB_RLOCK_TRACKER;
 1952         struct sockaddr *dst;
 1953         struct sockaddr *netmask;
 1954         struct rtentry *rt = NULL;
 1955         struct rt_addrinfo info;
 1956         int error = 0;
 1957         int startfib, endfib;
 1958         char tempbuf[_SOCKADDR_TMPSIZE];
 1959         int didwork = 0;
 1960         int a_failure = 0;
 1961         static struct sockaddr_dl null_sdl = {sizeof(null_sdl), AF_LINK};
 1962         struct rib_head *rnh;
 1963 
 1964         if (flags & RTF_HOST) {
 1965                 dst = ifa->ifa_dstaddr;
 1966                 netmask = NULL;
 1967         } else {
 1968                 dst = ifa->ifa_addr;
 1969                 netmask = ifa->ifa_netmask;
 1970         }
 1971         if (dst->sa_len == 0)
 1972                 return(EINVAL);
 1973         switch (dst->sa_family) {
 1974         case AF_INET6:
 1975         case AF_INET:
 1976                 /* We support multiple FIBs. */
 1977                 break;
 1978         default:
 1979                 fibnum = RT_DEFAULT_FIB;
 1980                 break;
 1981         }
 1982         if (fibnum == RT_ALL_FIBS) {
 1983                 if (V_rt_add_addr_allfibs == 0 && cmd == (int)RTM_ADD)
 1984                         startfib = endfib = ifa->ifa_ifp->if_fib;
 1985                 else {
 1986                         startfib = 0;
 1987                         endfib = rt_numfibs - 1;
 1988                 }
 1989         } else {
 1990                 KASSERT((fibnum < rt_numfibs), ("rtinit1: bad fibnum"));
 1991                 startfib = fibnum;
 1992                 endfib = fibnum;
 1993         }
 1994 
 1995         /*
 1996          * If it's a delete, check that if it exists,
 1997          * it's on the correct interface or we might scrub
 1998          * a route to another ifa which would
 1999          * be confusing at best and possibly worse.
 2000          */
 2001         if (cmd == RTM_DELETE) {
 2002                 /*
 2003                  * It's a delete, so it should already exist..
 2004                  * If it's a net, mask off the host bits
 2005                  * (Assuming we have a mask)
 2006                  * XXX this is kinda inet specific..
 2007                  */
 2008                 if (netmask != NULL) {
 2009                         rt_maskedcopy(dst, (struct sockaddr *)tempbuf, netmask);
 2010                         dst = (struct sockaddr *)tempbuf;
 2011                 }
 2012         }
 2013         /*
 2014          * Now go through all the requested tables (fibs) and do the
 2015          * requested action. Realistically, this will either be fib 0
 2016          * for protocols that don't do multiple tables or all the
 2017          * tables for those that do.
 2018          */
 2019         for ( fibnum = startfib; fibnum <= endfib; fibnum++) {
 2020                 if (cmd == RTM_DELETE) {
 2021                         struct radix_node *rn;
 2022                         /*
 2023                          * Look up an rtentry that is in the routing tree and
 2024                          * contains the correct info.
 2025                          */
 2026                         rnh = rt_tables_get_rnh(fibnum, dst->sa_family);
 2027                         if (rnh == NULL)
 2028                                 /* this table doesn't exist but others might */
 2029                                 continue;
 2030                         RIB_RLOCK(rnh);
 2031                         rn = rnh->rnh_lookup(dst, netmask, &rnh->head);
 2032 #ifdef RADIX_MPATH
 2033                         if (rt_mpath_capable(rnh)) {
 2034 
 2035                                 if (rn == NULL) 
 2036                                         error = ESRCH;
 2037                                 else {
 2038                                         rt = RNTORT(rn);
 2039                                         /*
 2040                                          * for interface route the
 2041                                          * rt->rt_gateway is sockaddr_intf
 2042                                          * for cloning ARP entries, so
 2043                                          * rt_mpath_matchgate must use the
 2044                                          * interface address
 2045                                          */
 2046                                         rt = rt_mpath_matchgate(rt,
 2047                                             ifa->ifa_addr);
 2048                                         if (rt == NULL) 
 2049                                                 error = ESRCH;
 2050                                 }
 2051                         }
 2052 #endif
 2053                         error = (rn == NULL ||
 2054                             (rn->rn_flags & RNF_ROOT) ||
 2055                             RNTORT(rn)->rt_ifa != ifa);
 2056                         RIB_RUNLOCK(rnh);
 2057                         if (error) {
 2058                                 /* this is only an error if bad on ALL tables */
 2059                                 continue;
 2060                         }
 2061                 }
 2062                 /*
 2063                  * Do the actual request
 2064                  */
 2065                 bzero((caddr_t)&info, sizeof(info));
 2066                 info.rti_ifa = ifa;
 2067                 info.rti_flags = flags |
 2068                     (ifa->ifa_flags & ~IFA_RTSELF) | RTF_PINNED;
 2069                 info.rti_info[RTAX_DST] = dst;
 2070                 /* 
 2071                  * doing this for compatibility reasons
 2072                  */
 2073                 if (cmd == RTM_ADD)
 2074                         info.rti_info[RTAX_GATEWAY] =
 2075                             (struct sockaddr *)&null_sdl;
 2076                 else
 2077                         info.rti_info[RTAX_GATEWAY] = ifa->ifa_addr;
 2078                 info.rti_info[RTAX_NETMASK] = netmask;
 2079                 error = rtrequest1_fib(cmd, &info, &rt, fibnum);
 2080                 if (error == 0 && rt != NULL) {
 2081                         /*
 2082                          * notify any listening routing agents of the change
 2083                          */
 2084                         RT_LOCK(rt);
 2085 #ifdef RADIX_MPATH
 2086                         /*
 2087                          * in case address alias finds the first address
 2088                          * e.g. ifconfig bge0 192.0.2.246/24
 2089                          * e.g. ifconfig bge0 192.0.2.247/24
 2090                          * the address set in the route is 192.0.2.246
 2091                          * so we need to replace it with 192.0.2.247
 2092                          */
 2093                         if (memcmp(rt->rt_ifa->ifa_addr,
 2094                             ifa->ifa_addr, ifa->ifa_addr->sa_len)) {
 2095                                 ifa_free(rt->rt_ifa);
 2096                                 ifa_ref(ifa);
 2097                                 rt->rt_ifp = ifa->ifa_ifp;
 2098                                 rt->rt_ifa = ifa;
 2099                         }
 2100 #endif
 2101                         /* 
 2102                          * doing this for compatibility reasons
 2103                          */
 2104                         if (cmd == RTM_ADD) {
 2105                             ((struct sockaddr_dl *)rt->rt_gateway)->sdl_type  =
 2106                                 rt->rt_ifp->if_type;
 2107                             ((struct sockaddr_dl *)rt->rt_gateway)->sdl_index =
 2108                                 rt->rt_ifp->if_index;
 2109                         }
 2110                         RT_ADDREF(rt);
 2111                         RT_UNLOCK(rt);
 2112                         rt_newaddrmsg_fib(cmd, ifa, error, rt, fibnum);
 2113                         RT_LOCK(rt);
 2114                         RT_REMREF(rt);
 2115                         if (cmd == RTM_DELETE) {
 2116                                 /*
 2117                                  * If we are deleting, and we found an entry,
 2118                                  * then it's been removed from the tree..
 2119                                  * now throw it away.
 2120                                  */
 2121                                 RTFREE_LOCKED(rt);
 2122                         } else {
 2123                                 if (cmd == RTM_ADD) {
 2124                                         /*
 2125                                          * We just wanted to add it..
 2126                                          * we don't actually need a reference.
 2127                                          */
 2128                                         RT_REMREF(rt);
 2129                                 }
 2130                                 RT_UNLOCK(rt);
 2131                         }
 2132                         didwork = 1;
 2133                 }
 2134                 if (error)
 2135                         a_failure = error;
 2136         }
 2137         if (cmd == RTM_DELETE) {
 2138                 if (didwork) {
 2139                         error = 0;
 2140                 } else {
 2141                         /* we only give an error if it wasn't in any table */
 2142                         error = ((flags & RTF_HOST) ?
 2143                             EHOSTUNREACH : ENETUNREACH);
 2144                 }
 2145         } else {
 2146                 if (a_failure) {
 2147                         /* return an error if any of them failed */
 2148                         error = a_failure;
 2149                 }
 2150         }
 2151         return (error);
 2152 }
 2153 
 2154 /*
 2155  * Set up a routing table entry, normally
 2156  * for an interface.
 2157  */
 2158 int
 2159 rtinit(struct ifaddr *ifa, int cmd, int flags)
 2160 {
 2161         struct sockaddr *dst;
 2162         int fib = RT_DEFAULT_FIB;
 2163 
 2164         if (flags & RTF_HOST) {
 2165                 dst = ifa->ifa_dstaddr;
 2166         } else {
 2167                 dst = ifa->ifa_addr;
 2168         }
 2169 
 2170         switch (dst->sa_family) {
 2171         case AF_INET6:
 2172         case AF_INET:
 2173                 /* We do support multiple FIBs. */
 2174                 fib = RT_ALL_FIBS;
 2175                 break;
 2176         }
 2177         return (rtinit1(ifa, cmd, flags, fib));
 2178 }
 2179 
 2180 /*
 2181  * Announce interface address arrival/withdraw
 2182  * Returns 0 on success.
 2183  */
 2184 int
 2185 rt_addrmsg(int cmd, struct ifaddr *ifa, int fibnum)
 2186 {
 2187 
 2188         KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
 2189             ("unexpected cmd %d", cmd));
 2190         KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
 2191             ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
 2192 
 2193         EVENTHANDLER_DIRECT_INVOKE(rt_addrmsg, ifa, cmd);
 2194         return (rtsock_addrmsg(cmd, ifa, fibnum));
 2195 }
 2196 
 2197 /*
 2198  * Announce route addition/removal.
 2199  * Users of this function MUST validate input data BEFORE calling.
 2200  * However we have to be able to handle invalid data:
 2201  * if some userland app sends us "invalid" route message (invalid mask,
 2202  * no dst, wrong address families, etc...) we need to pass it back
 2203  * to app (and any other rtsock consumers) with rtm_errno field set to
 2204  * non-zero value.
 2205  * Returns 0 on success.
 2206  */
 2207 int
 2208 rt_routemsg(int cmd, struct ifnet *ifp, int error, struct rtentry *rt,
 2209     int fibnum)
 2210 {
 2211 
 2212         KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
 2213             ("unexpected cmd %d", cmd));
 2214         
 2215         KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
 2216             ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
 2217 
 2218         KASSERT(rt_key(rt) != NULL, (":%s: rt_key must be supplied", __func__));
 2219 
 2220         return (rtsock_routemsg(cmd, ifp, error, rt, fibnum));
 2221 }
 2222 
 2223 void
 2224 rt_newaddrmsg(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt)
 2225 {
 2226 
 2227         rt_newaddrmsg_fib(cmd, ifa, error, rt, RT_ALL_FIBS);
 2228 }
 2229 
 2230 /*
 2231  * This is called to generate messages from the routing socket
 2232  * indicating a network interface has had addresses associated with it.
 2233  */
 2234 void
 2235 rt_newaddrmsg_fib(int cmd, struct ifaddr *ifa, int error, struct rtentry *rt,
 2236     int fibnum)
 2237 {
 2238 
 2239         KASSERT(cmd == RTM_ADD || cmd == RTM_DELETE,
 2240                 ("unexpected cmd %u", cmd));
 2241         KASSERT(fibnum == RT_ALL_FIBS || (fibnum >= 0 && fibnum < rt_numfibs),
 2242             ("%s: fib out of range 0 <=%d<%d", __func__, fibnum, rt_numfibs));
 2243 
 2244         if (cmd == RTM_ADD) {
 2245                 rt_addrmsg(cmd, ifa, fibnum);
 2246                 if (rt != NULL)
 2247                         rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum);
 2248         } else {
 2249                 if (rt != NULL)
 2250                         rt_routemsg(cmd, ifa->ifa_ifp, error, rt, fibnum);
 2251                 rt_addrmsg(cmd, ifa, fibnum);
 2252         }
 2253 }
 2254 

Cache object: c376987d232cac108b463135f9d16170


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.