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/netinet/in_rmx.c

Version: -  FREEBSD  -  FREEBSD-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * Copyright 1994, 1995 Massachusetts Institute of Technology
    3  *
    4  * Permission to use, copy, modify, and distribute this software and
    5  * its documentation for any purpose and without fee is hereby
    6  * granted, provided that both the above copyright notice and this
    7  * permission notice appear in all copies, that both the above
    8  * copyright notice and this permission notice appear in all
    9  * supporting documentation, and that the name of M.I.T. not be used
   10  * in advertising or publicity pertaining to distribution of the
   11  * software without specific, written prior permission.  M.I.T. makes
   12  * no representations about the suitability of this software for any
   13  * purpose.  It is provided "as is" without express or implied
   14  * warranty.
   15  *
   16  * THIS SOFTWARE IS PROVIDED BY M.I.T. ``AS IS''.  M.I.T. DISCLAIMS
   17  * ALL EXPRESS OR IMPLIED WARRANTIES WITH REGARD TO THIS SOFTWARE,
   18  * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
   19  * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. IN NO EVENT
   20  * SHALL M.I.T. BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
   21  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
   22  * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF
   23  * USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND
   24  * ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
   25  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
   26  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   27  * SUCH DAMAGE.
   28  */
   29 
   30 #include <sys/cdefs.h>
   31 __FBSDID("$FreeBSD$");
   32 
   33 #include <sys/param.h>
   34 #include <sys/systm.h>
   35 #include <sys/kernel.h>
   36 #include <sys/sysctl.h>
   37 #include <sys/socket.h>
   38 #include <sys/mbuf.h>
   39 #include <sys/syslog.h>
   40 #include <sys/callout.h>
   41 
   42 #include <net/if.h>
   43 #include <net/route.h>
   44 #include <net/vnet.h>
   45 
   46 #include <netinet/in.h>
   47 #include <netinet/in_var.h>
   48 #include <netinet/ip.h>
   49 #include <netinet/ip_icmp.h>
   50 #include <netinet/ip_var.h>
   51 
   52 extern int      in_inithead(void **head, int off);
   53 #ifdef VIMAGE
   54 extern int      in_detachhead(void **head, int off);
   55 #endif
   56 
   57 #define RTPRF_OURS              RTF_PROTO3      /* set on routes we manage */
   58 
   59 /*
   60  * Do what we need to do when inserting a route.
   61  */
   62 static struct radix_node *
   63 in_addroute(void *v_arg, void *n_arg, struct radix_node_head *head,
   64     struct radix_node *treenodes)
   65 {
   66         struct rtentry *rt = (struct rtentry *)treenodes;
   67         struct sockaddr_in *sin = (struct sockaddr_in *)rt_key(rt);
   68 
   69         RADIX_NODE_HEAD_WLOCK_ASSERT(head);
   70         /*
   71          * A little bit of help for both IP output and input:
   72          *   For host routes, we make sure that RTF_BROADCAST
   73          *   is set for anything that looks like a broadcast address.
   74          *   This way, we can avoid an expensive call to in_broadcast()
   75          *   in ip_output() most of the time (because the route passed
   76          *   to ip_output() is almost always a host route).
   77          *
   78          *   We also do the same for local addresses, with the thought
   79          *   that this might one day be used to speed up ip_input().
   80          *
   81          * We also mark routes to multicast addresses as such, because
   82          * it's easy to do and might be useful (but this is much more
   83          * dubious since it's so easy to inspect the address).
   84          */
   85         if (rt->rt_flags & RTF_HOST) {
   86                 if (in_broadcast(sin->sin_addr, rt->rt_ifp)) {
   87                         rt->rt_flags |= RTF_BROADCAST;
   88                 } else if (satosin(rt->rt_ifa->ifa_addr)->sin_addr.s_addr ==
   89                     sin->sin_addr.s_addr) {
   90                         rt->rt_flags |= RTF_LOCAL;
   91                 }
   92         }
   93         if (IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
   94                 rt->rt_flags |= RTF_MULTICAST;
   95 
   96         if (!rt->rt_rmx.rmx_mtu && rt->rt_ifp)
   97                 rt->rt_rmx.rmx_mtu = rt->rt_ifp->if_mtu;
   98 
   99         return (rn_addroute(v_arg, n_arg, head, treenodes));
  100 }
  101 
  102 /*
  103  * This code is the inverse of in_clsroute: on first reference, if we
  104  * were managing the route, stop doing so and set the expiration timer
  105  * back off again.
  106  */
  107 static struct radix_node *
  108 in_matroute(void *v_arg, struct radix_node_head *head)
  109 {
  110         struct radix_node *rn = rn_match(v_arg, head);
  111         struct rtentry *rt = (struct rtentry *)rn;
  112 
  113         if (rt) {
  114                 RT_LOCK(rt);
  115                 if (rt->rt_flags & RTPRF_OURS) {
  116                         rt->rt_flags &= ~RTPRF_OURS;
  117                         rt->rt_rmx.rmx_expire = 0;
  118                 }
  119                 RT_UNLOCK(rt);
  120         }
  121         return rn;
  122 }
  123 
  124 static VNET_DEFINE(int, rtq_reallyold) = 60*60; /* one hour is "really old" */
  125 #define V_rtq_reallyold         VNET(rtq_reallyold)
  126 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_RTEXPIRE, rtexpire, CTLFLAG_RW,
  127     &VNET_NAME(rtq_reallyold), 0,
  128     "Default expiration time on dynamically learned routes");
  129 
  130 /* never automatically crank down to less */
  131 static VNET_DEFINE(int, rtq_minreallyold) = 10;
  132 #define V_rtq_minreallyold      VNET(rtq_minreallyold)
  133 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_RTMINEXPIRE, rtminexpire, CTLFLAG_RW,
  134     &VNET_NAME(rtq_minreallyold), 0,
  135     "Minimum time to attempt to hold onto dynamically learned routes");
  136 
  137 /* 128 cached routes is "too many" */
  138 static VNET_DEFINE(int, rtq_toomany) = 128;
  139 #define V_rtq_toomany           VNET(rtq_toomany)
  140 SYSCTL_VNET_INT(_net_inet_ip, IPCTL_RTMAXCACHE, rtmaxcache, CTLFLAG_RW,
  141     &VNET_NAME(rtq_toomany), 0,
  142     "Upper limit on dynamically learned routes");
  143 
  144 /*
  145  * On last reference drop, mark the route as belong to us so that it can be
  146  * timed out.
  147  */
  148 static void
  149 in_clsroute(struct radix_node *rn, struct radix_node_head *head)
  150 {
  151         struct rtentry *rt = (struct rtentry *)rn;
  152 
  153         RT_LOCK_ASSERT(rt);
  154 
  155         if (!(rt->rt_flags & RTF_UP))
  156                 return;                 /* prophylactic measures */
  157 
  158         if (rt->rt_flags & RTPRF_OURS)
  159                 return;
  160 
  161         if (!(rt->rt_flags & RTF_DYNAMIC))
  162                 return;
  163 
  164         /*
  165          * If rtq_reallyold is 0, just delete the route without
  166          * waiting for a timeout cycle to kill it.
  167          */
  168         if (V_rtq_reallyold != 0) {
  169                 rt->rt_flags |= RTPRF_OURS;
  170                 rt->rt_rmx.rmx_expire = time_uptime + V_rtq_reallyold;
  171         } else {
  172                 rtexpunge(rt);
  173         }
  174 }
  175 
  176 struct rtqk_arg {
  177         struct radix_node_head *rnh;
  178         int draining;
  179         int killed;
  180         int found;
  181         int updating;
  182         time_t nextstop;
  183 };
  184 
  185 /*
  186  * Get rid of old routes.  When draining, this deletes everything, even when
  187  * the timeout is not expired yet.  When updating, this makes sure that
  188  * nothing has a timeout longer than the current value of rtq_reallyold.
  189  */
  190 static int
  191 in_rtqkill(struct radix_node *rn, void *rock)
  192 {
  193         struct rtqk_arg *ap = rock;
  194         struct rtentry *rt = (struct rtentry *)rn;
  195         int err;
  196 
  197         RADIX_NODE_HEAD_WLOCK_ASSERT(ap->rnh);
  198 
  199         if (rt->rt_flags & RTPRF_OURS) {
  200                 ap->found++;
  201 
  202                 if (ap->draining || rt->rt_rmx.rmx_expire <= time_uptime) {
  203                         if (rt->rt_refcnt > 0)
  204                                 panic("rtqkill route really not free");
  205 
  206                         err = in_rtrequest(RTM_DELETE,
  207                                         (struct sockaddr *)rt_key(rt),
  208                                         rt->rt_gateway, rt_mask(rt),
  209                                         rt->rt_flags | RTF_RNH_LOCKED, 0,
  210                                         rt->rt_fibnum);
  211                         if (err) {
  212                                 log(LOG_WARNING, "in_rtqkill: error %d\n", err);
  213                         } else {
  214                                 ap->killed++;
  215                         }
  216                 } else {
  217                         if (ap->updating &&
  218                             (rt->rt_rmx.rmx_expire - time_uptime >
  219                              V_rtq_reallyold)) {
  220                                 rt->rt_rmx.rmx_expire =
  221                                     time_uptime + V_rtq_reallyold;
  222                         }
  223                         ap->nextstop = lmin(ap->nextstop,
  224                                             rt->rt_rmx.rmx_expire);
  225                 }
  226         }
  227 
  228         return 0;
  229 }
  230 
  231 #define RTQ_TIMEOUT     60*10   /* run no less than once every ten minutes */
  232 static VNET_DEFINE(int, rtq_timeout) = RTQ_TIMEOUT;
  233 static VNET_DEFINE(struct callout, rtq_timer);
  234 
  235 #define V_rtq_timeout           VNET(rtq_timeout)
  236 #define V_rtq_timer             VNET(rtq_timer)
  237 
  238 static void in_rtqtimo_one(void *rock);
  239 
  240 static void
  241 in_rtqtimo(void *rock)
  242 {
  243         CURVNET_SET((struct vnet *) rock);
  244         int fibnum;
  245         void *newrock;
  246         struct timeval atv;
  247 
  248         for (fibnum = 0; fibnum < rt_numfibs; fibnum++) {
  249                 newrock = rt_tables_get_rnh(fibnum, AF_INET);
  250                 if (newrock != NULL)
  251                         in_rtqtimo_one(newrock);
  252         }
  253         atv.tv_usec = 0;
  254         atv.tv_sec = V_rtq_timeout;
  255         callout_reset(&V_rtq_timer, tvtohz(&atv), in_rtqtimo, rock);
  256         CURVNET_RESTORE();
  257 }
  258 
  259 static void
  260 in_rtqtimo_one(void *rock)
  261 {
  262         struct radix_node_head *rnh = rock;
  263         struct rtqk_arg arg;
  264         static time_t last_adjusted_timeout = 0;
  265 
  266         arg.found = arg.killed = 0;
  267         arg.rnh = rnh;
  268         arg.nextstop = time_uptime + V_rtq_timeout;
  269         arg.draining = arg.updating = 0;
  270         RADIX_NODE_HEAD_LOCK(rnh);
  271         rnh->rnh_walktree(rnh, in_rtqkill, &arg);
  272         RADIX_NODE_HEAD_UNLOCK(rnh);
  273 
  274         /*
  275          * Attempt to be somewhat dynamic about this:
  276          * If there are ``too many'' routes sitting around taking up space,
  277          * then crank down the timeout, and see if we can't make some more
  278          * go away.  However, we make sure that we will never adjust more
  279          * than once in rtq_timeout seconds, to keep from cranking down too
  280          * hard.
  281          */
  282         if ((arg.found - arg.killed > V_rtq_toomany) &&
  283             (time_uptime - last_adjusted_timeout >= V_rtq_timeout) &&
  284             V_rtq_reallyold > V_rtq_minreallyold) {
  285                 V_rtq_reallyold = 2 * V_rtq_reallyold / 3;
  286                 if (V_rtq_reallyold < V_rtq_minreallyold) {
  287                         V_rtq_reallyold = V_rtq_minreallyold;
  288                 }
  289 
  290                 last_adjusted_timeout = time_uptime;
  291 #ifdef DIAGNOSTIC
  292                 log(LOG_DEBUG, "in_rtqtimo: adjusted rtq_reallyold to %d\n",
  293                     V_rtq_reallyold);
  294 #endif
  295                 arg.found = arg.killed = 0;
  296                 arg.updating = 1;
  297                 RADIX_NODE_HEAD_LOCK(rnh);
  298                 rnh->rnh_walktree(rnh, in_rtqkill, &arg);
  299                 RADIX_NODE_HEAD_UNLOCK(rnh);
  300         }
  301 
  302 }
  303 
  304 void
  305 in_rtqdrain(void)
  306 {
  307         VNET_ITERATOR_DECL(vnet_iter);
  308         struct radix_node_head *rnh;
  309         struct rtqk_arg arg;
  310         int     fibnum;
  311 
  312         VNET_LIST_RLOCK_NOSLEEP();
  313         VNET_FOREACH(vnet_iter) {
  314                 CURVNET_SET(vnet_iter);
  315 
  316                 for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
  317                         rnh = rt_tables_get_rnh(fibnum, AF_INET);
  318                         arg.found = arg.killed = 0;
  319                         arg.rnh = rnh;
  320                         arg.nextstop = 0;
  321                         arg.draining = 1;
  322                         arg.updating = 0;
  323                         RADIX_NODE_HEAD_LOCK(rnh);
  324                         rnh->rnh_walktree(rnh, in_rtqkill, &arg);
  325                         RADIX_NODE_HEAD_UNLOCK(rnh);
  326                 }
  327                 CURVNET_RESTORE();
  328         }
  329         VNET_LIST_RUNLOCK_NOSLEEP();
  330 }
  331 
  332 void
  333 in_setmatchfunc(struct radix_node_head *rnh, int val)
  334 {
  335 
  336         rnh->rnh_matchaddr = (val != 0) ? rn_match : in_matroute;
  337 }
  338 
  339 static int _in_rt_was_here;
  340 /*
  341  * Initialize our routing tree.
  342  */
  343 int
  344 in_inithead(void **head, int off)
  345 {
  346         struct radix_node_head *rnh;
  347 
  348         /* XXX MRT
  349          * This can be called from vfs_export.c too in which case 'off'
  350          * will be 0. We know the correct value so just use that and
  351          * return directly if it was 0.
  352          * This is a hack that replaces an even worse hack on a bad hack
  353          * on a bad design. After RELENG_7 this should be fixed but that
  354          * will change the ABI, so for now do it this way.
  355          */
  356         if (!rn_inithead(head, 32))
  357                 return 0;
  358 
  359         if (off == 0)           /* XXX MRT  see above */
  360                 return 1;       /* only do the rest for a real routing table */
  361 
  362         rnh = *head;
  363         rnh->rnh_addaddr = in_addroute;
  364         in_setmatchfunc(rnh, V_drop_redirect);
  365         rnh->rnh_close = in_clsroute;
  366         if (_in_rt_was_here == 0 ) {
  367                 callout_init(&V_rtq_timer, CALLOUT_MPSAFE);
  368                 callout_reset(&V_rtq_timer, 1, in_rtqtimo, curvnet);
  369                 _in_rt_was_here = 1;
  370         }
  371         return 1;
  372 }
  373 
  374 #ifdef VIMAGE
  375 int
  376 in_detachhead(void **head, int off)
  377 {
  378 
  379         callout_drain(&V_rtq_timer);
  380         return (1);
  381 }
  382 #endif
  383 
  384 /*
  385  * This zaps old routes when the interface goes down or interface
  386  * address is deleted.  In the latter case, it deletes static routes
  387  * that point to this address.  If we don't do this, we may end up
  388  * using the old address in the future.  The ones we always want to
  389  * get rid of are things like ARP entries, since the user might down
  390  * the interface, walk over to a completely different network, and
  391  * plug back in.
  392  */
  393 struct in_ifadown_arg {
  394         struct ifaddr *ifa;
  395         int del;
  396 };
  397 
  398 static int
  399 in_ifadownkill(struct radix_node *rn, void *xap)
  400 {
  401         struct in_ifadown_arg *ap = xap;
  402         struct rtentry *rt = (struct rtentry *)rn;
  403 
  404         RT_LOCK(rt);
  405         if (rt->rt_ifa == ap->ifa &&
  406             (ap->del || !(rt->rt_flags & RTF_STATIC))) {
  407                 /*
  408                  * Aquire a reference so that it can later be freed
  409                  * as the refcount would be 0 here in case of at least
  410                  * ap->del.
  411                  */
  412                 RT_ADDREF(rt);
  413                 /*
  414                  * Disconnect it from the tree and permit protocols
  415                  * to cleanup.
  416                  */
  417                 rtexpunge(rt);
  418                 /*
  419                  * At this point it is an rttrash node, and in case
  420                  * the above is the only reference we must free it.
  421                  * If we do not noone will have a pointer and the
  422                  * rtentry will be leaked forever.
  423                  * In case someone else holds a reference, we are
  424                  * fine as we only decrement the refcount. In that
  425                  * case if the other entity calls RT_REMREF, we
  426                  * will still be leaking but at least we tried.
  427                  */
  428                 RTFREE_LOCKED(rt);
  429                 return (0);
  430         }
  431         RT_UNLOCK(rt);
  432         return 0;
  433 }
  434 
  435 int
  436 in_ifadown(struct ifaddr *ifa, int delete)
  437 {
  438         struct in_ifadown_arg arg;
  439         struct radix_node_head *rnh;
  440         int     fibnum;
  441 
  442         if (ifa->ifa_addr->sa_family != AF_INET)
  443                 return 1;
  444 
  445         for ( fibnum = 0; fibnum < rt_numfibs; fibnum++) {
  446                 rnh = rt_tables_get_rnh(fibnum, AF_INET);
  447                 arg.ifa = ifa;
  448                 arg.del = delete;
  449                 RADIX_NODE_HEAD_LOCK(rnh);
  450                 rnh->rnh_walktree(rnh, in_ifadownkill, &arg);
  451                 RADIX_NODE_HEAD_UNLOCK(rnh);
  452                 ifa->ifa_flags &= ~IFA_ROUTE;           /* XXXlocking? */
  453         }
  454         return 0;
  455 }
  456 
  457 /*
  458  * inet versions of rt functions. These have fib extensions and 
  459  * for now will just reference the _fib variants.
  460  * eventually this order will be reversed,
  461  */
  462 void
  463 in_rtalloc_ign(struct route *ro, u_long ignflags, u_int fibnum)
  464 {
  465         rtalloc_ign_fib(ro, ignflags, fibnum);
  466 }
  467 
  468 int
  469 in_rtrequest( int req,
  470         struct sockaddr *dst,
  471         struct sockaddr *gateway,
  472         struct sockaddr *netmask,
  473         int flags,
  474         struct rtentry **ret_nrt,
  475         u_int fibnum)
  476 {
  477         return (rtrequest_fib(req, dst, gateway, netmask, 
  478             flags, ret_nrt, fibnum));
  479 }
  480 
  481 struct rtentry *
  482 in_rtalloc1(struct sockaddr *dst, int report, u_long ignflags, u_int fibnum)
  483 {
  484         return (rtalloc1_fib(dst, report, ignflags, fibnum));
  485 }
  486 
  487 void
  488 in_rtredirect(struct sockaddr *dst,
  489         struct sockaddr *gateway,
  490         struct sockaddr *netmask,
  491         int flags,
  492         struct sockaddr *src,
  493         u_int fibnum)
  494 {
  495         rtredirect_fib(dst, gateway, netmask, flags, src, fibnum);
  496 }
  497  
  498 void
  499 in_rtalloc(struct route *ro, u_int fibnum)
  500 {
  501         rtalloc_ign_fib(ro, 0UL, fibnum);
  502 }
  503 
  504 #if 0
  505 int      in_rt_getifa(struct rt_addrinfo *, u_int fibnum);
  506 int      in_rtioctl(u_long, caddr_t, u_int);
  507 int      in_rtrequest1(int, struct rt_addrinfo *, struct rtentry **, u_int);
  508 #endif
  509 
  510 

Cache object: 057d48fc902ce4a9a3e950712eec8f33


[ 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.