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/bridge.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 (c) 1998-2002 Luigi Rizzo
    3  *
    4  * Work partly supported by: Cisco Systems, Inc. - NSITE lab, RTP, NC
    5  *
    6  * Redistribution and use in source and binary forms, with or without
    7  * modification, are permitted provided that the following conditions
    8  * are met:
    9  * 1. Redistributions of source code must retain the above copyright
   10  *    notice, this list of conditions and the following disclaimer.
   11  * 2. Redistributions in binary form must reproduce the above copyright
   12  *    notice, this list of conditions and the following disclaimer in the
   13  *    documentation and/or other materials provided with the distribution.
   14  *
   15  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND
   16  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   17  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   18  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
   19  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   20  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   21  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   22  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   23  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   24  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   25  * SUCH DAMAGE.
   26  *
   27  * $FreeBSD: releng/6.0/sys/net/bridge.c 149443 2005-08-25 05:01:24Z rwatson $
   28  */
   29 
   30 /*
   31  * This code implements bridging in FreeBSD. It only acts on ethernet
   32  * interfaces, including VLANs (others are still usable for routing).
   33  * A FreeBSD host can implement multiple logical bridges, called
   34  * "clusters". Each cluster is made of a set of interfaces, and
   35  * identified by a "cluster-id" which is a number in the range 1..2^16-1.
   36  *
   37  * Bridging is enabled by the sysctl variable
   38  *      net.link.ether.bridge.enable
   39  * the grouping of interfaces into clusters is done with
   40  *      net.link.ether.bridge.config
   41  * containing a list of interfaces each optionally followed by
   42  * a colon and the cluster it belongs to (1 is the default).
   43  * Separators can be spaces, commas or tabs, e.g.
   44  *      net.link.ether.bridge.config="fxp0:2 fxp1:2 dc0 dc1:1"
   45  * Optionally bridged packets can be passed through the firewall,
   46  * this is controlled by the variable
   47  *      net.link.ether.bridge.ipfw
   48  *
   49  * For each cluster there is a descriptor (cluster_softc) storing
   50  * the following data structures:
   51  * - a hash table with the MAC address and destination interface for each
   52  *   known node. The table is indexed using a hash of the source address.
   53  * - an array with the MAC addresses of the interfaces used in the cluster.
   54  *
   55  * Input packets are tapped near the beginning of ether_input(), and
   56  * analysed by bridge_in(). Depending on the result, the packet
   57  * can be forwarded to one or more output interfaces using bdg_forward(),
   58  * and/or sent to the upper layer (e.g. in case of multicast).
   59  *
   60  * Output packets are intercepted near the end of ether_output().
   61  * The correct destination is selected by bridge_dst_lookup(),
   62  * and then forwarding is done by bdg_forward().
   63  *
   64  * The arp code is also modified to let a machine answer to requests
   65  * irrespective of the port the request came from.
   66  *
   67  * In case of loops in the bridging topology, the bridge detects this
   68  * event and temporarily mutes output bridging on one of the ports.
   69  * Periodically, interfaces are unmuted by bdg_timeout().
   70  * Muting is only implemented as a safety measure, and also as
   71  * a mechanism to support a user-space implementation of the spanning
   72  * tree algorithm.
   73  *
   74  * To build a bridging kernel, use the following option
   75  *    option BRIDGE
   76  * and then at runtime set the sysctl variable to enable bridging.
   77  *
   78  * Only one interface per cluster is supposed to have addresses set (but
   79  * there are no substantial problems if you set addresses for none or
   80  * for more than one interface).
   81  * Bridging will act before routing, but nothing prevents a machine
   82  * from doing both (modulo bugs in the implementation...).
   83  *
   84  * THINGS TO REMEMBER
   85  *  - bridging is incompatible with multicast routing on the same
   86  *    machine. There is not an easy fix to this.
   87  *  - be very careful when bridging VLANs
   88  *  - loop detection is still not very robust.
   89  */
   90 
   91 #include <sys/param.h>
   92 #include <sys/mbuf.h>
   93 #include <sys/malloc.h>
   94 #include <sys/protosw.h>
   95 #include <sys/systm.h>
   96 #include <sys/socket.h> /* for net/if.h */
   97 #include <sys/ctype.h>  /* string functions */
   98 #include <sys/kernel.h>
   99 #include <sys/module.h>
  100 #include <sys/sysctl.h>
  101 
  102 #include <net/ethernet.h>
  103 #include <net/if.h>
  104 #include <net/if_arp.h>         /* for struct arpcom */
  105 #include <net/if_types.h>
  106 #include <net/if_var.h>
  107 #include <net/pfil.h>
  108 
  109 #include <netinet/in.h>
  110 #include <netinet/in_systm.h>
  111 #include <netinet/in_var.h>
  112 #include <netinet/ip.h>
  113 #include <netinet/ip_var.h>
  114 
  115 #include <net/route.h>
  116 #include <netinet/ip_fw.h>
  117 #include <netinet/ip_dummynet.h>
  118 #include <net/bridge.h>
  119 
  120 /*--------------------*/
  121 
  122 #define ETHER_ADDR_COPY(_dst,_src)      bcopy(_src, _dst, ETHER_ADDR_LEN)
  123 #define ETHER_ADDR_EQ(_a1,_a2)          (bcmp(_a1, _a2, ETHER_ADDR_LEN) == 0)
  124 
  125 /*
  126  * For each cluster, source MAC addresses are stored into a hash
  127  * table which locates the port they reside on.
  128  */
  129 #define HASH_SIZE 8192  /* Table size, must be a power of 2 */
  130 
  131 typedef struct hash_table {             /* each entry.          */
  132     struct ifnet *      name;
  133     u_char              etheraddr[ETHER_ADDR_LEN];
  134     u_int16_t           used;           /* also, padding        */
  135 } bdg_hash_table ;
  136 
  137 /*
  138  * The hash function applied to MAC addresses. Out of the 6 bytes,
  139  * the last ones tend to vary more. Since we are on a little endian machine,
  140  * we have to do some gimmick...
  141  */
  142 #define HASH_FN(addr)   (       \
  143     ntohs( ((u_int16_t *)addr)[1] ^ ((u_int16_t *)addr)[2] ) & (HASH_SIZE -1))
  144 
  145 /*
  146  * This is the data structure where local addresses are stored.
  147  */
  148 struct bdg_addr {
  149     u_char      etheraddr[ETHER_ADDR_LEN];
  150     u_int16_t   _padding;
  151 };
  152 
  153 /*
  154  * The configuration of each cluster includes the cluster id, a pointer to
  155  * the hash table, and an array of local MAC addresses (of size "ports").
  156  */
  157 struct cluster_softc {
  158     u_int16_t   cluster_id;
  159     u_int16_t   ports;
  160     bdg_hash_table *ht;
  161     struct bdg_addr     *my_macs;       /* local MAC addresses */
  162 };
  163 
  164 
  165 extern struct protosw inetsw[];                 /* from netinet/ip_input.c */
  166 
  167 static int n_clusters;                          /* number of clusters */
  168 static struct cluster_softc *clusters;
  169 
  170 static struct mtx bdg_mtx;
  171 #define BDG_LOCK_INIT()         mtx_init(&bdg_mtx, "bridge", NULL, MTX_DEF)
  172 #define BDG_LOCK_DESTROY()      mtx_destroy(&bdg_mtx)
  173 #define BDG_LOCK()              mtx_lock(&bdg_mtx)
  174 #define BDG_UNLOCK()            mtx_unlock(&bdg_mtx)
  175 #define BDG_LOCK_ASSERT()       mtx_assert(&bdg_mtx, MA_OWNED)
  176 
  177 #define BDG_MUTED(ifp) (ifp2sc[ifp->if_index].flags & IFF_MUTE)
  178 #define BDG_MUTE(ifp) ifp2sc[ifp->if_index].flags |= IFF_MUTE
  179 #define BDG_CLUSTER(ifp) (ifp2sc[ifp->if_index].cluster)
  180 
  181 #define BDG_SAMECLUSTER(ifp,src) \
  182         (src == NULL || BDG_CLUSTER(ifp) == BDG_CLUSTER(src) )
  183 
  184 #ifdef __i386__
  185 #define BDG_MATCH(a,b) ( \
  186     ((u_int16_t *)(a))[2] == ((u_int16_t *)(b))[2] && \
  187     *((u_int32_t *)(a)) == *((u_int32_t *)(b)) )
  188 #define IS_ETHER_BROADCAST(a) ( \
  189         *((u_int32_t *)(a)) == 0xffffffff && \
  190         ((u_int16_t *)(a))[2] == 0xffff )
  191 #else
  192 /* for machines that do not support unaligned access */
  193 #define BDG_MATCH(a,b)          ETHER_ADDR_EQ(a,b)
  194 #define IS_ETHER_BROADCAST(a)   ETHER_ADDR_EQ(a,"\377\377\377\377\377\377")
  195 #endif
  196 
  197 SYSCTL_DECL(_net_link_ether);
  198 SYSCTL_NODE(_net_link_ether, OID_AUTO, bridge, CTLFLAG_RD, 0,
  199         "Bridge parameters");
  200 static char bridge_version[] = "031224";
  201 SYSCTL_STRING(_net_link_ether_bridge, OID_AUTO, version, CTLFLAG_RD,
  202         bridge_version, 0, "software version");
  203 
  204 #define BRIDGE_DEBUG
  205 #ifdef BRIDGE_DEBUG
  206 int     bridge_debug = 0;
  207 SYSCTL_INT(_net_link_ether_bridge, OID_AUTO, debug, CTLFLAG_RW, &bridge_debug,
  208             0, "control debugging printfs");
  209 #define DPRINTF(X)      if (bridge_debug) printf X
  210 #else
  211 #define DPRINTF(X)
  212 #endif
  213 
  214 #ifdef BRIDGE_TIMING
  215 /*
  216  * For timing-related debugging, you can use the following macros.
  217  * remember, rdtsc() only works on Pentium-class machines
  218 
  219     quad_t ticks;
  220     DDB(ticks = rdtsc();)
  221     ... interesting code ...
  222     DDB(bdg_fw_ticks += (u_long)(rdtsc() - ticks) ; bdg_fw_count++ ;)
  223 
  224  *
  225  */
  226 #define DDB(x)  x
  227 
  228 static int bdg_fw_avg;
  229 SYSCTL_INT(_net_link_ether_bridge, OID_AUTO, fw_avg, CTLFLAG_RW,
  230             &bdg_fw_avg, 0,"Cycle counter avg");
  231 static int bdg_fw_ticks;
  232 SYSCTL_INT(_net_link_ether_bridge, OID_AUTO, fw_ticks, CTLFLAG_RW,
  233             &bdg_fw_ticks, 0,"Cycle counter item");
  234 static int bdg_fw_count;
  235 SYSCTL_INT(_net_link_ether_bridge, OID_AUTO, fw_count, CTLFLAG_RW,
  236             &bdg_fw_count, 0,"Cycle counter count");
  237 #else
  238 #define DDB(x)
  239 #endif
  240 
  241 static int bdginit(void);
  242 static void parse_bdg_cfg(void);
  243 static struct mbuf *bdg_forward(struct mbuf *, struct ifnet *);
  244 
  245 static int bdg_ipf;             /* IPFilter enabled in bridge */
  246 SYSCTL_INT(_net_link_ether_bridge, OID_AUTO, ipf, CTLFLAG_RW,
  247             &bdg_ipf, 0,"Pass bridged pkts through IPFilter");
  248 static int bdg_ipfw;
  249 SYSCTL_INT(_net_link_ether_bridge, OID_AUTO, ipfw, CTLFLAG_RW,
  250             &bdg_ipfw,0,"Pass bridged pkts through firewall");
  251 
  252 static int bdg_copy;
  253 SYSCTL_INT(_net_link_ether_bridge, OID_AUTO, copy, CTLFLAG_RW,
  254         &bdg_copy, 0, "Force packet copy in bdg_forward");
  255 
  256 int bdg_ipfw_drops;
  257 SYSCTL_INT(_net_link_ether_bridge, OID_AUTO, ipfw_drop,
  258         CTLFLAG_RW, &bdg_ipfw_drops,0,"");
  259 int bdg_ipfw_colls;
  260 SYSCTL_INT(_net_link_ether_bridge, OID_AUTO, ipfw_collisions,
  261         CTLFLAG_RW, &bdg_ipfw_colls,0,"");
  262 
  263 static int bdg_thru;
  264 SYSCTL_INT(_net_link_ether_bridge, OID_AUTO, packets, CTLFLAG_RW,
  265         &bdg_thru, 0, "Packets through bridge");
  266 static int bdg_dropped;
  267 SYSCTL_INT(_net_link_ether_bridge, OID_AUTO, dropped, CTLFLAG_RW,
  268         &bdg_dropped, 0, "Packets dropped in bdg_forward");
  269 static int bdg_predict;
  270 SYSCTL_INT(_net_link_ether_bridge, OID_AUTO, predict, CTLFLAG_RW,
  271         &bdg_predict, 0, "Correctly predicted header location");
  272 
  273 #ifdef BRIDGE_DEBUG
  274 static char *bdg_dst_names[] = {
  275         "BDG_NULL    ",
  276         "BDG_BCAST   ",
  277         "BDG_MCAST   ",
  278         "BDG_LOCAL   ",
  279         "BDG_DROP    ",
  280         "BDG_UNKNOWN ",
  281         "BDG_IN      ",
  282         "BDG_OUT     ",
  283         "BDG_FORWARD " };
  284 #endif /* BRIDGE_DEBUG */
  285 
  286 /*
  287  * System initialization
  288  */
  289 static struct bdg_stats bdg_stats ;
  290 SYSCTL_STRUCT(_net_link_ether_bridge, OID_AUTO, stats, CTLFLAG_RD,
  291         &bdg_stats, bdg_stats, "bridge statistics");
  292 
  293 static struct callout bdg_callout;
  294 
  295 /*
  296  * Add an interface to a cluster, possibly creating a new entry in
  297  * the cluster table. This requires reallocation of the table and
  298  * updating pointers in ifp2sc.
  299  */
  300 static struct cluster_softc *
  301 add_cluster(u_int16_t cluster_id, struct ifnet *ifp)
  302 {
  303     struct cluster_softc *c = NULL;
  304     int i;
  305 
  306     BDG_LOCK_ASSERT();
  307 
  308     for (i = 0; i < n_clusters ; i++)
  309         if (clusters[i].cluster_id == cluster_id)
  310             goto found;
  311 
  312     /* Not found, need to reallocate */
  313     c = malloc((1+n_clusters) * sizeof (*c), M_IFADDR, M_NOWAIT | M_ZERO);
  314     if (c == NULL) {/* malloc failure */
  315         printf("-- bridge: cannot add new cluster\n");
  316         goto bad;
  317     }
  318     c[n_clusters].ht = (struct hash_table *)
  319             malloc(HASH_SIZE * sizeof(struct hash_table),
  320                 M_IFADDR, M_NOWAIT | M_ZERO);
  321     if (c[n_clusters].ht == NULL) {
  322         printf("-- bridge: cannot allocate hash table for new cluster\n");
  323         goto bad;
  324     }
  325     c[n_clusters].my_macs = (struct bdg_addr *)
  326             malloc(BDG_MAX_PORTS * sizeof(struct bdg_addr),
  327                 M_IFADDR, M_NOWAIT | M_ZERO);
  328     if (c[n_clusters].my_macs == NULL) {
  329         printf("-- bridge: cannot allocate mac addr table for new cluster\n");
  330         free(c[n_clusters].ht, M_IFADDR);
  331         goto bad;
  332     }
  333 
  334     c[n_clusters].cluster_id = cluster_id;
  335     c[n_clusters].ports = 0;
  336     /*
  337      * now copy old descriptors here
  338      */
  339     if (n_clusters > 0) {
  340         for (i=0; i < n_clusters; i++)
  341             c[i] = clusters[i];
  342         /*
  343          * and finally update pointers in ifp2sc
  344          */
  345         for (i = 0 ; i < if_index && i < BDG_MAX_PORTS; i++)
  346             if (ifp2sc[i].cluster != NULL)
  347                 ifp2sc[i].cluster = c + (ifp2sc[i].cluster - clusters);
  348         free(clusters, M_IFADDR);
  349     }
  350     clusters = c;
  351     i = n_clusters;             /* index of cluster entry */
  352     n_clusters++;
  353 found:
  354     c = clusters + i;           /* the right cluster ... */
  355     ETHER_ADDR_COPY(c->my_macs[c->ports].etheraddr, IFP2ENADDR(ifp));
  356     c->ports++;
  357     return c;
  358 bad:
  359     if (c)
  360         free(c, M_IFADDR);
  361     return NULL;
  362 }
  363 
  364 
  365 /*
  366  * Turn off bridging, by clearing promisc mode on the interface,
  367  * marking the interface as unused, and clearing the name in the
  368  * stats entry.
  369  * Also dispose the hash tables associated with the clusters.
  370  */
  371 static void
  372 bridge_off(void)
  373 {
  374     struct ifnet *ifp ;
  375     int i;
  376 
  377     BDG_LOCK_ASSERT();
  378 
  379     DPRINTF(("%s: n_clusters %d\n", __func__, n_clusters));
  380 
  381     IFNET_RLOCK();
  382     TAILQ_FOREACH(ifp, &ifnet, if_link) {
  383         struct bdg_softc *b;
  384 
  385         if (ifp->if_index >= BDG_MAX_PORTS)
  386             continue;   /* make sure we do not go beyond the end */
  387         b = &ifp2sc[ifp->if_index];
  388 
  389         if ( b->flags & IFF_BDG_PROMISC ) {
  390             ifpromisc(ifp, 0);
  391             b->flags &= ~(IFF_BDG_PROMISC|IFF_MUTE) ;
  392             DPRINTF(("%s: %s promisc OFF if_flags 0x%x "
  393                 "bdg_flags 0x%x\n", __func__, ifp->if_xname,
  394                 ifp->if_flags, b->flags));
  395         }
  396         b->flags &= ~(IFF_USED) ;
  397         b->cluster = NULL;
  398         bdg_stats.s[ifp->if_index].name[0] = '\0';
  399     }
  400     IFNET_RUNLOCK();
  401     /* flush_tables */
  402 
  403     for (i=0; i < n_clusters; i++) {
  404         free(clusters[i].ht, M_IFADDR);
  405         free(clusters[i].my_macs, M_IFADDR);
  406     }
  407     if (clusters != NULL)
  408         free(clusters, M_IFADDR);
  409     clusters = NULL;
  410     n_clusters =0;
  411 }
  412 
  413 /*
  414  * set promisc mode on the interfaces we use.
  415  */
  416 static void
  417 bridge_on(void)
  418 {
  419     struct ifnet *ifp ;
  420 
  421     BDG_LOCK_ASSERT();
  422 
  423     IFNET_RLOCK();
  424     TAILQ_FOREACH(ifp, &ifnet, if_link) {
  425         struct bdg_softc *b = &ifp2sc[ifp->if_index];
  426 
  427         if ( !(b->flags & IFF_USED) )
  428             continue ;
  429         if ( !( ifp->if_flags & IFF_UP) ) {
  430             if_up(ifp);
  431         }
  432         if ( !(b->flags & IFF_BDG_PROMISC) ) {
  433             (void) ifpromisc(ifp, 1);
  434             b->flags |= IFF_BDG_PROMISC ;
  435             DPRINTF(("%s: %s promisc ON if_flags 0x%x bdg_flags 0x%x\n",
  436                 __func__, ifp->if_xname, ifp->if_flags, b->flags));
  437         }
  438         if (b->flags & IFF_MUTE) {
  439             DPRINTF(("%s: unmuting %s\n", __func__, ifp->if_xname));
  440             b->flags &= ~IFF_MUTE;
  441         }
  442     }
  443     IFNET_RUNLOCK();
  444 }
  445 
  446 static char bridge_cfg[1024];           /* NB: in BSS so initialized to zero */
  447 
  448 /**
  449  * reconfigure bridge.
  450  * This is also done every time we attach or detach an interface.
  451  * Main use is to make sure that we do not bridge on some old
  452  * (ejected) device. So, it would be really useful to have a
  453  * pointer to the modified device as an argument. Without it, we
  454  * have to scan all interfaces.
  455  */
  456 static void
  457 reconfigure_bridge_locked(void)
  458 {
  459     BDG_LOCK_ASSERT();
  460 
  461     bridge_off();
  462     if (do_bridge) {
  463         if (if_index >= BDG_MAX_PORTS) {
  464             printf("-- sorry too many interfaces (%d, max is %d),"
  465                 " disabling bridging\n", if_index, BDG_MAX_PORTS);
  466             do_bridge = 0;
  467             return;
  468         }
  469         parse_bdg_cfg();
  470         bridge_on();
  471     }
  472 }
  473 
  474 static void
  475 reconfigure_bridge(void)
  476 {
  477     BDG_LOCK();
  478     reconfigure_bridge_locked();
  479     BDG_UNLOCK();
  480 }
  481 
  482 /*
  483  * parse the config string, set IFF_USED, name and cluster_id
  484  * for all interfaces found.
  485  * The config string is a list of "if[:cluster]" with
  486  * a number of possible separators (see "sep"). In particular the
  487  * use of the space lets you set bridge_cfg with the output from
  488  * "ifconfig -l"
  489  */
  490 static void
  491 parse_bdg_cfg(void)
  492 {
  493     char *p, *beg;
  494     int l, cluster;
  495     static const char *sep = ", \t";
  496 
  497     BDG_LOCK_ASSERT();
  498 
  499     for (p = bridge_cfg; *p ; p++) {
  500         struct ifnet *ifp;
  501         int found = 0;
  502         char c;
  503 
  504         if (index(sep, *p))     /* skip separators */
  505             continue ;
  506         /* names are lowercase and digits */
  507         for ( beg = p ; islower(*p) || isdigit(*p) ; p++ )
  508             ;
  509         l = p - beg ;           /* length of name string */
  510         if (l == 0)             /* invalid name */
  511             break ;
  512         if ( *p != ':' )        /* no ':', assume default cluster 1 */
  513             cluster = 1 ;
  514         else                    /* fetch cluster */
  515             cluster = strtoul( p+1, &p, 10);
  516         c = *p;
  517         *p = '\0';
  518         /*
  519          * now search in interface list for a matching name
  520          */
  521         IFNET_RLOCK();          /* could sleep XXX */
  522         TAILQ_FOREACH(ifp, &ifnet, if_link) {
  523 
  524             if (!strncmp(beg, ifp->if_xname, max(l, strlen(ifp->if_xname)))) {
  525                 struct bdg_softc *b = &ifp2sc[ifp->if_index];
  526                 if (ifp->if_type != IFT_ETHER && ifp->if_type != IFT_L2VLAN) {
  527                     printf("%s is not an ethernet, continue\n", ifp->if_xname);
  528                     continue;
  529                 }
  530                 if (b->flags & IFF_USED) {
  531                     printf("%s already used, skipping\n", ifp->if_xname);
  532                     break;
  533                 }
  534                 b->cluster = add_cluster(htons(cluster), ifp);
  535                 b->flags |= IFF_USED ;
  536                 snprintf(bdg_stats.s[ifp->if_index].name,
  537                     sizeof(bdg_stats.s[ifp->if_index].name),
  538                     "%s:%d", ifp->if_xname, cluster);
  539 
  540                 DPRINTF(("%s: found %s next c %d\n", __func__,
  541                     bdg_stats.s[ifp->if_index].name, c));
  542                 found = 1;
  543                 break ;
  544             }
  545         }
  546         IFNET_RUNLOCK();
  547         if (!found)
  548             printf("interface %s Not found in bridge\n", beg);
  549         *p = c;
  550         if (c == '\0')
  551             break; /* no more */
  552     }
  553 }
  554 
  555 /*
  556  * handler for net.link.ether.bridge
  557  */
  558 static int
  559 sysctl_bdg(SYSCTL_HANDLER_ARGS)
  560 {
  561     int enable = do_bridge;
  562     int error;
  563 
  564     error = sysctl_handle_int(oidp, &enable, 0, req);
  565     enable = (enable) ? 1 : 0;
  566     BDG_LOCK();
  567     if (enable != do_bridge) {
  568         do_bridge = enable;
  569         reconfigure_bridge_locked();
  570     }
  571     BDG_UNLOCK();
  572     return error ;
  573 }
  574 SYSCTL_PROC(_net_link_ether_bridge, OID_AUTO, enable, CTLTYPE_INT|CTLFLAG_RW,
  575             &do_bridge, 0, &sysctl_bdg, "I", "Bridging");
  576 
  577 /*
  578  * handler for net.link.ether.bridge_cfg
  579  */
  580 static int
  581 sysctl_bdg_cfg(SYSCTL_HANDLER_ARGS)
  582 {
  583     int error;
  584     char *new_cfg;
  585 
  586     new_cfg = malloc(sizeof(bridge_cfg), M_TEMP, M_WAITOK);
  587     bcopy(bridge_cfg, new_cfg, sizeof(bridge_cfg));
  588 
  589     error = sysctl_handle_string(oidp, new_cfg, oidp->oid_arg2, req);
  590     if (error == 0) {
  591         BDG_LOCK();
  592         if (strcmp(new_cfg, bridge_cfg)) {
  593             bcopy(new_cfg, bridge_cfg, sizeof(bridge_cfg));
  594             reconfigure_bridge_locked();
  595         }
  596         BDG_UNLOCK();
  597     }
  598 
  599     free(new_cfg, M_TEMP);
  600 
  601     return error;
  602 }
  603 SYSCTL_PROC(_net_link_ether_bridge, OID_AUTO, config, CTLTYPE_STRING|CTLFLAG_RW,
  604             &bridge_cfg, sizeof(bridge_cfg), &sysctl_bdg_cfg, "A",
  605             "Bridge configuration");
  606 
  607 static int
  608 sysctl_refresh(SYSCTL_HANDLER_ARGS)
  609 {
  610     if (req->newptr)
  611         reconfigure_bridge();
  612 
  613     return 0;
  614 }
  615 SYSCTL_PROC(_net_link_ether_bridge, OID_AUTO, refresh, CTLTYPE_INT|CTLFLAG_WR,
  616             NULL, 0, &sysctl_refresh, "I", "iface refresh");
  617 
  618 #ifndef BURN_BRIDGES
  619 #define SYSCTL_OID_COMPAT(parent, nbr, name, kind, a1, a2, handler, fmt, descr)\
  620         static struct sysctl_oid sysctl__##parent##_##name##_compat = {  \
  621                 &sysctl_##parent##_children, { 0 },                      \
  622                 nbr, kind, a1, a2, #name, handler, fmt, 0, descr };      \
  623         DATA_SET(sysctl_set, sysctl__##parent##_##name##_compat)
  624 #define SYSCTL_INT_COMPAT(parent, nbr, name, access, ptr, val, descr)    \
  625         SYSCTL_OID_COMPAT(parent, nbr, name, CTLTYPE_INT|(access),       \
  626                 ptr, val, sysctl_handle_int, "I", descr)
  627 #define SYSCTL_STRUCT_COMPAT(parent, nbr, name, access, ptr, type, descr)\
  628         SYSCTL_OID_COMPAT(parent, nbr, name, CTLTYPE_OPAQUE|(access),    \
  629                 ptr, sizeof(struct type), sysctl_handle_opaque,          \
  630                 "S," #type, descr)
  631 #define SYSCTL_PROC_COMPAT(parent, nbr, name, access, ptr, arg, handler, fmt, descr) \
  632         SYSCTL_OID_COMPAT(parent, nbr, name, (access),                   \
  633                 ptr, arg, handler, fmt, descr)
  634 
  635 SYSCTL_INT_COMPAT(_net_link_ether, OID_AUTO, bridge_ipf, CTLFLAG_RW,
  636             &bdg_ipf, 0,"Pass bridged pkts through IPFilter");
  637 SYSCTL_INT_COMPAT(_net_link_ether, OID_AUTO, bridge_ipfw, CTLFLAG_RW,
  638             &bdg_ipfw,0,"Pass bridged pkts through firewall");
  639 SYSCTL_STRUCT_COMPAT(_net_link_ether, PF_BDG, bdgstats, CTLFLAG_RD,
  640         &bdg_stats, bdg_stats, "bridge statistics");
  641 SYSCTL_PROC_COMPAT(_net_link_ether, OID_AUTO, bridge_cfg, 
  642             CTLTYPE_STRING|CTLFLAG_RW,
  643             &bridge_cfg, sizeof(bridge_cfg), &sysctl_bdg_cfg, "A",
  644             "Bridge configuration");
  645 SYSCTL_PROC_COMPAT(_net_link_ether, OID_AUTO, bridge_refresh,
  646             CTLTYPE_INT|CTLFLAG_WR,
  647             NULL, 0, &sysctl_refresh, "I", "iface refresh");
  648 #endif
  649 
  650 static int bdg_loops;
  651 static int bdg_slowtimer = 0;
  652 static int bdg_age_index = 0;   /* index of table position to age */
  653 
  654 /*
  655  * called periodically to flush entries etc.
  656  */
  657 static void
  658 bdg_timeout(void *dummy)
  659 {
  660     if (do_bridge) {
  661         int l, i;
  662 
  663         BDG_LOCK();
  664         /*
  665          * age entries in the forwarding table.
  666          */
  667         l = bdg_age_index + HASH_SIZE/4 ;
  668         if (l > HASH_SIZE)
  669             l = HASH_SIZE;
  670 
  671         for (i = 0; i < n_clusters; i++) {
  672             bdg_hash_table *bdg_table = clusters[i].ht;
  673             for (; bdg_age_index < l; bdg_age_index++)
  674                 if (bdg_table[bdg_age_index].used)
  675                     bdg_table[bdg_age_index].used = 0;
  676                 else if (bdg_table[bdg_age_index].name) {
  677                     DPRINTF(("%s: flushing stale entry %d\n",
  678                         __func__, bdg_age_index));
  679                     bdg_table[bdg_age_index].name = NULL;
  680                 }
  681         }
  682         if (bdg_age_index >= HASH_SIZE)
  683             bdg_age_index = 0;
  684 
  685         if (--bdg_slowtimer <= 0 ) {
  686             bdg_slowtimer = 5;
  687 
  688             bridge_on();        /* we just need unmute, really */
  689             bdg_loops = 0;
  690         }
  691         BDG_UNLOCK();
  692     }
  693     callout_reset(&bdg_callout, 2*hz, bdg_timeout, NULL);
  694 }
  695 
  696 /*
  697  * Find the right pkt destination:
  698  *      BDG_BCAST       is a broadcast
  699  *      BDG_MCAST       is a multicast
  700  *      BDG_LOCAL       is for a local address
  701  *      BDG_DROP        must be dropped
  702  *      other           ifp of the dest. interface (incl.self)
  703  *
  704  * We assume this is only called for interfaces for which bridging
  705  * is enabled, i.e. BDG_USED(ifp) is true.
  706  */
  707 static __inline struct ifnet *
  708 bridge_dst_lookup(struct ether_header *eh, struct cluster_softc *c)
  709 {
  710     bdg_hash_table *bt;         /* pointer to entry in hash table */
  711 
  712     BDG_LOCK_ASSERT();
  713 
  714     if (ETHER_IS_MULTICAST(eh->ether_dhost))
  715         return IS_ETHER_BROADCAST(eh->ether_dhost) ? BDG_BCAST : BDG_MCAST;
  716     /*
  717      * Lookup local addresses in case one matches.  We optimize
  718      * for the common case of two interfaces.
  719      */
  720     KASSERT(c->ports != 0, ("lookup with no ports!"));
  721     switch (c->ports) {
  722         int i;
  723     default:
  724         for (i = c->ports-1; i > 1; i--) {
  725             if (ETHER_ADDR_EQ(c->my_macs[i].etheraddr, eh->ether_dhost))
  726                 return BDG_LOCAL;
  727         }
  728         /* fall thru... */
  729     case 2:
  730         if (ETHER_ADDR_EQ(c->my_macs[1].etheraddr, eh->ether_dhost))
  731             return BDG_LOCAL;
  732     case 1:
  733         if (ETHER_ADDR_EQ(c->my_macs[0].etheraddr, eh->ether_dhost))
  734             return BDG_LOCAL;
  735     }
  736     /*
  737      * Look for a possible destination in table
  738      */
  739     bt = &c->ht[HASH_FN(eh->ether_dhost)];
  740     if (bt->name && ETHER_ADDR_EQ(bt->etheraddr, eh->ether_dhost))
  741         return bt->name;
  742     else
  743         return BDG_UNKNOWN;
  744 }
  745 
  746 /**
  747  * bridge_in() is invoked to perform bridging decision on input packets.
  748  *
  749  * On Input:
  750  *   eh         Ethernet header of the incoming packet.
  751  *   ifp        interface the packet is coming from.
  752  *
  753  * On Return: destination of packet, one of
  754  *   BDG_BCAST  broadcast
  755  *   BDG_MCAST  multicast
  756  *   BDG_LOCAL  is only for a local address (do not forward)
  757  *   BDG_DROP   drop the packet
  758  *   ifp        ifp of the destination interface.
  759  *
  760  * Forwarding is not done directly to give a chance to some drivers
  761  * to fetch more of the packet, or simply drop it completely.
  762  */
  763 
  764 static struct mbuf *
  765 bridge_in(struct ifnet *ifp, struct mbuf *m)
  766 {
  767     struct ether_header *eh;
  768     struct ifnet *dst, *old;
  769     bdg_hash_table *bt;                 /* location in hash table */
  770     int dropit = BDG_MUTED(ifp);
  771     int index;
  772 
  773     eh = mtod(m, struct ether_header *);
  774 
  775     /*
  776      * hash the source address
  777      */
  778     BDG_LOCK();
  779     index = HASH_FN(eh->ether_shost);
  780     bt = &BDG_CLUSTER(ifp)->ht[index];
  781     bt->used = 1;
  782     old = bt->name;
  783     if (old) {                          /* the entry is valid */
  784         if (!ETHER_ADDR_EQ(eh->ether_shost, bt->etheraddr)) {
  785             bdg_ipfw_colls++;
  786             bt->name = NULL;            /* NB: will overwrite below */
  787         } else if (old != ifp) {
  788             /*
  789              * Found a loop. Either a machine has moved, or there
  790              * is a misconfiguration/reconfiguration of the network.
  791              * First, do not forward this packet!
  792              * Record the relocation anyways; then, if loops persist,
  793              * suspect a reconfiguration and disable forwarding
  794              * from the old interface.
  795              */
  796             bt->name = ifp;             /* relocate address */
  797             printf("-- loop (%d) %6D to %s from %s (%s)\n",
  798                         bdg_loops, eh->ether_shost, ".",
  799                         ifp->if_xname, old->if_xname,
  800                         BDG_MUTED(old) ? "muted":"active");
  801             dropit = 1;
  802             if (!BDG_MUTED(old)) {
  803                 if (bdg_loops++ > 10)
  804                     BDG_MUTE(old);
  805             }
  806         }
  807     }
  808 
  809     /*
  810      * now write the source address into the table
  811      */
  812     if (bt->name == NULL) {
  813         DPRINTF(("%s: new addr %6D at %d for %s\n",
  814             __func__, eh->ether_shost, ".", index, ifp->if_xname));
  815         ETHER_ADDR_COPY(bt->etheraddr, eh->ether_shost);
  816         bt->name = ifp;
  817     }
  818     dst = bridge_dst_lookup(eh, BDG_CLUSTER(ifp));
  819     BDG_UNLOCK();
  820 
  821     /*
  822      * bridge_dst_lookup can return the following values:
  823      *   BDG_BCAST, BDG_MCAST, BDG_LOCAL, BDG_UNKNOWN, BDG_DROP, ifp.
  824      * For muted interfaces, or when we detect a loop, the first 3 are
  825      * changed in BDG_LOCAL (we still listen to incoming traffic),
  826      * and others to BDG_DROP (no use for the local host).
  827      * Also, for incoming packets, ifp is changed to BDG_DROP if ifp == src.
  828      * These changes are not necessary for outgoing packets from ether_output().
  829      */
  830     BDG_STAT(ifp, BDG_IN);
  831     switch ((uintptr_t)dst) {
  832     case (uintptr_t)BDG_BCAST:
  833     case (uintptr_t)BDG_MCAST:
  834     case (uintptr_t)BDG_LOCAL:
  835     case (uintptr_t)BDG_UNKNOWN:
  836     case (uintptr_t)BDG_DROP:
  837         BDG_STAT(ifp, dst);
  838         break;
  839     default:
  840         if (dst == ifp || dropit)
  841             BDG_STAT(ifp, BDG_DROP);
  842         else
  843             BDG_STAT(ifp, BDG_FORWARD);
  844         break;
  845     }
  846 
  847     if (dropit) {
  848         if (dst == BDG_BCAST || dst == BDG_MCAST || dst == BDG_LOCAL)
  849             dst = BDG_LOCAL;
  850         else
  851             dst = BDG_DROP;
  852     } else {
  853         if (dst == ifp)
  854             dst = BDG_DROP;
  855     }
  856     DPRINTF(("%s: %6D ->%6D ty 0x%04x dst %s\n", __func__,
  857         eh->ether_shost, ".",
  858         eh->ether_dhost, ".",
  859         ntohs(eh->ether_type),
  860         (dst <= BDG_FORWARD) ? bdg_dst_names[(uintptr_t)dst] :
  861                 dst->if_xname));
  862 
  863     switch ((uintptr_t)dst) {
  864     case (uintptr_t)BDG_DROP:
  865         m_freem(m);
  866         return (NULL);
  867 
  868     case (uintptr_t)BDG_LOCAL:
  869         return (m);
  870 
  871     case (uintptr_t)BDG_BCAST:
  872     case (uintptr_t)BDG_MCAST:
  873         m = bdg_forward(m, dst);
  874 #ifdef  DIAGNOSTIC
  875         if (m == NULL)
  876                 if_printf(ifp, "bridge dropped %s packet\n",
  877                      dst == BDG_BCAST ? "broadcast" : "multicast");
  878 #endif
  879         return (m);
  880 
  881     default:
  882         m = bdg_forward(m, dst);
  883         /*
  884          * But in some cases the bridge may return the
  885          * packet for us to free; sigh.
  886          */
  887         if (m != NULL)
  888                 m_freem(m);
  889 
  890     }
  891 
  892     return (NULL);
  893 }
  894 
  895 /*
  896  * Return 1 if it's ok to send a packet out the specified interface.
  897  * The interface must be:
  898  *      used for bridging,
  899  *      not muted,
  900  *      not full,
  901  *      up and running,
  902  *      not the source interface, and
  903  *      belong to the same cluster as the 'real_dst'.
  904  */
  905 static __inline int
  906 bridge_ifok(struct ifnet *ifp, struct ifnet *src, struct ifnet *dst)
  907 {
  908     return (BDG_USED(ifp)
  909         && !BDG_MUTED(ifp)
  910         && !_IF_QFULL(&ifp->if_snd)
  911         && (ifp->if_flags & IFF_UP)
  912         && (ifp->if_drv_flags & IFF_DRV_RUNNING)
  913         && ifp != src
  914         && BDG_SAMECLUSTER(ifp, dst));
  915 }
  916 
  917 /*
  918  * Forward a packet to dst -- which can be a single interface or
  919  * an entire cluster. The src port and muted interfaces are excluded.
  920  *
  921  * If src == NULL, the pkt comes from ether_output, and dst is the real
  922  * interface the packet is originally sent to. In this case, we must forward
  923  * it to the whole cluster.
  924  * We never call bdg_forward from ether_output on interfaces which are
  925  * not part of a cluster.
  926  *
  927  * If possible (i.e. we can determine that the caller does not need
  928  * a copy), the packet is consumed here, and bdg_forward returns NULL.
  929  * Otherwise, a pointer to a copy of the packet is returned.
  930  */
  931 static struct mbuf *
  932 bdg_forward(struct mbuf *m0, struct ifnet *dst)
  933 {
  934 #define EH_RESTORE(_m) do {                                                \
  935     M_PREPEND((_m), ETHER_HDR_LEN, M_DONTWAIT);                            \
  936     if ((_m) == NULL) {                                                    \
  937         bdg_dropped++;                                                     \
  938         return NULL;                                                       \
  939     }                                                                      \
  940     if (eh != mtod((_m), struct ether_header *))                           \
  941         bcopy(&save_eh, mtod((_m), struct ether_header *), ETHER_HDR_LEN); \
  942     else                                                                   \
  943         bdg_predict++;                                                     \
  944 } while (0);
  945     struct ether_header *eh;
  946     struct ifnet *src;
  947     struct ifnet *ifp, *last;
  948     int shared = bdg_copy;              /* someone else is using the mbuf */
  949     int error;
  950     struct ifnet *real_dst = dst;       /* real dst from ether_output */
  951     struct ip_fw_args args;
  952     struct ether_header save_eh;
  953     struct mbuf *m;
  954 
  955     DDB(quad_t ticks; ticks = rdtsc();)
  956 
  957     args.rule = ip_dn_claim_rule(m0);
  958     if (args.rule)
  959         shared = 0;                     /* For sure this is our own mbuf. */
  960     else
  961         bdg_thru++;                     /* count 1st time through bdg_forward */
  962 
  963     /*
  964      * The packet arrives with the Ethernet header at the front.
  965      */
  966     eh = mtod(m0, struct ether_header *);
  967 
  968     src = m0->m_pkthdr.rcvif;
  969     if (src == NULL) {                  /* packet from ether_output */
  970         BDG_LOCK();
  971         dst = bridge_dst_lookup(eh, BDG_CLUSTER(real_dst));
  972         BDG_UNLOCK();
  973     }
  974 
  975     if (dst == BDG_DROP) {              /* this should not happen */
  976         printf("xx bdg_forward for BDG_DROP\n");
  977         m_freem(m0);
  978         bdg_dropped++;
  979         return NULL;
  980     }
  981     if (dst == BDG_LOCAL) {             /* this should not happen as well */
  982         printf("xx ouch, bdg_forward for local pkt\n");
  983         return m0;
  984     }
  985     if (dst == BDG_BCAST || dst == BDG_MCAST) {
  986          /* need a copy for the local stack */
  987          shared = 1;
  988     }
  989 
  990     /*
  991      * Do filtering in a very similar way to what is done in ip_output.
  992      * Only if firewall is loaded, enabled, and the packet is not
  993      * from ether_output() (src==NULL, or we would filter it twice).
  994      * Additional restrictions may apply e.g. non-IP, short packets,
  995      * and pkts already gone through a pipe.
  996      */
  997     if (src != NULL && (
  998         (inet_pfil_hook.ph_busy_count >= 0 && bdg_ipf != 0) ||
  999         (IPFW_LOADED && bdg_ipfw != 0))) {
 1000 
 1001         int i;
 1002 
 1003         if (args.rule != NULL && fw_one_pass)
 1004             goto forward; /* packet already partially processed */
 1005         /*
 1006          * i need some amt of data to be contiguous, and in case others need
 1007          * the packet (shared==1) also better be in the first mbuf.
 1008          */
 1009         i = min(m0->m_pkthdr.len, max_protohdr) ;
 1010         if (shared || m0->m_len < i) {
 1011             m0 = m_pullup(m0, i);
 1012             if (m0 == NULL) {
 1013                 printf("%s: m_pullup failed\n", __func__);      /* XXXDPRINTF*/
 1014                 bdg_dropped++;
 1015                 return NULL;
 1016             }
 1017             eh = mtod(m0, struct ether_header *);
 1018         }
 1019 
 1020         /*
 1021          * Processing below expects the Ethernet header is stripped.
 1022          * Furthermore, the mbuf chain might be replaced at various
 1023          * places.  To deal with this we copy the header to a temporary
 1024          * location, strip the header, and restore it as needed.
 1025          */
 1026         bcopy(eh, &save_eh, ETHER_HDR_LEN);     /* local copy for restore */
 1027         m_adj(m0, ETHER_HDR_LEN);               /* temporarily strip header */
 1028 
 1029         /*
 1030          * Check that the IP header is aligned before passing up to the packet
 1031          * filter.
 1032          */
 1033         if (ntohs(save_eh.ether_type) == ETHERTYPE_IP && 
 1034             IP_HDR_ALIGNED_P(mtod(m0, caddr_t)) == 0) {
 1035                 if ((m0 = m_copyup(m0, sizeof(struct ip),
 1036                         (max_linkhdr + 3) & ~3)) == NULL) {
 1037                         bdg_dropped++;
 1038                         return NULL;
 1039                 }
 1040         }
 1041 
 1042         /*
 1043          * NetBSD-style generic packet filter, pfil(9), hooks.
 1044          * Enables ipf(8) in bridging.
 1045          */
 1046         if (!IPFW_LOADED) { /* XXX: Prevent ipfw from being run twice. */
 1047         if (inet_pfil_hook.ph_busy_count >= 0 &&
 1048             m0->m_pkthdr.len >= sizeof(struct ip) &&
 1049             ntohs(save_eh.ether_type) == ETHERTYPE_IP) {
 1050             /*
 1051              * before calling the firewall, swap fields the same as IP does.
 1052              * here we assume the pkt is an IP one and the header is contiguous
 1053              */
 1054             struct ip *ip = mtod(m0, struct ip *);
 1055 
 1056             ip->ip_len = ntohs(ip->ip_len);
 1057             ip->ip_off = ntohs(ip->ip_off);
 1058 
 1059             if (pfil_run_hooks(&inet_pfil_hook, &m0, src, PFIL_IN, NULL) != 0) {
 1060                 /* NB: hook should consume packet */
 1061                 return NULL;
 1062             }
 1063             if (m0 == NULL)                     /* consumed by filter */
 1064                 return m0;
 1065             /*
 1066              * If we get here, the firewall has passed the pkt, but the mbuf
 1067              * pointer might have changed. Restore ip and the fields ntohs()'d.
 1068              */
 1069             ip = mtod(m0, struct ip *);
 1070             ip->ip_len = htons(ip->ip_len);
 1071             ip->ip_off = htons(ip->ip_off);
 1072         }
 1073         } /* XXX: Prevent ipfw from being run twice. */
 1074 
 1075         /*
 1076          * Prepare arguments and call the firewall.
 1077          */
 1078         if (!IPFW_LOADED || bdg_ipfw == 0) {
 1079             EH_RESTORE(m0);     /* restore Ethernet header */
 1080             goto forward;       /* not using ipfw, accept the packet */
 1081         }
 1082 
 1083         /*
 1084          * XXX The following code is very similar to the one in
 1085          * if_ethersubr.c:ether_ipfw_chk()
 1086          */
 1087 
 1088         args.m = m0;            /* the packet we are looking at         */
 1089         args.oif = NULL;        /* this is an input packet              */
 1090         args.next_hop = NULL;   /* we do not support forward yet        */
 1091         args.eh = &save_eh;     /* MAC header for bridged/MAC packets   */
 1092         i = ip_fw_chk_ptr(&args);
 1093         m0 = args.m;            /* in case the firewall used the mbuf   */
 1094 
 1095         if (m0 != NULL)
 1096                 EH_RESTORE(m0); /* restore Ethernet header */
 1097 
 1098         if (i == IP_FW_DENY) /* drop */
 1099             return m0;
 1100 
 1101         KASSERT(m0 != NULL, ("bdg_forward: m0 is NULL"));
 1102 
 1103         if (i == 0) /* a PASS rule.  */
 1104             goto forward;
 1105         if (DUMMYNET_LOADED && (i == IP_FW_DUMMYNET)) {
 1106             /*
 1107              * Pass the pkt to dummynet, which consumes it.
 1108              * If shared, make a copy and keep the original.
 1109              */
 1110             if (shared) {
 1111                 m = m_copypacket(m0, M_DONTWAIT);
 1112                 if (m == NULL) {        /* copy failed, give up */
 1113                     bdg_dropped++;
 1114                     return NULL;
 1115                 }
 1116             } else {
 1117                 m = m0 ; /* pass the original to dummynet */
 1118                 m0 = NULL ; /* and nothing back to the caller */
 1119             }
 1120 
 1121             args.oif = real_dst;
 1122             ip_dn_io_ptr(m, DN_TO_BDG_FWD, &args);
 1123             return m0;
 1124         }
 1125         /*
 1126          * XXX at some point, add support for divert/forward actions.
 1127          * If none of the above matches, we have to drop the packet.
 1128          */
 1129         bdg_ipfw_drops++;
 1130         return m0;
 1131     }
 1132 forward:
 1133     /*
 1134      * Again, bring up the headers in case of shared bufs to avoid
 1135      * corruptions in the future.
 1136      */
 1137     if (shared) {
 1138         int i = min(m0->m_pkthdr.len, max_protohdr);
 1139 
 1140         m0 = m_pullup(m0, i);
 1141         if (m0 == NULL) {
 1142             bdg_dropped++;
 1143             return NULL;
 1144         }
 1145         /* NB: eh is not used below; no need to recalculate it */
 1146     }
 1147 
 1148     /*
 1149      * now real_dst is used to determine the cluster where to forward.
 1150      * For packets coming from ether_input, this is the one of the 'src'
 1151      * interface, whereas for locally generated packets (src==NULL) it
 1152      * is the cluster of the original destination interface, which
 1153      * was already saved into real_dst.
 1154      */
 1155     if (src != NULL)
 1156         real_dst = src;
 1157 
 1158     last = NULL;
 1159     if (dst == BDG_BCAST || dst == BDG_MCAST || dst == BDG_UNKNOWN) {
 1160         /*
 1161          * Scan all ports and send copies to all but the last.
 1162          */
 1163         IFNET_RLOCK();          /* XXX replace with generation # */
 1164         TAILQ_FOREACH(ifp, &ifnet, if_link) {
 1165             if (bridge_ifok(ifp, src, real_dst)) {
 1166                 if (last) {
 1167                     /*
 1168                      * At this point we know two interfaces need a copy
 1169                      * of the packet (last + ifp) so we must create a
 1170                      * copy to handoff to last.
 1171                      */
 1172                     m = m_copypacket(m0, M_DONTWAIT);
 1173                     if (m == NULL) {
 1174                         IFNET_RUNLOCK();
 1175                         printf("%s: , m_copypacket failed!\n", __func__);
 1176                         bdg_dropped++;
 1177                         return m0;      /* the original is still there... */
 1178                     }
 1179                     IFQ_HANDOFF(last, m, error);
 1180                     if (!error)
 1181                         BDG_STAT(last, BDG_OUT);
 1182                     else
 1183                         bdg_dropped++;
 1184                 }
 1185                 last = ifp;
 1186             }
 1187         }
 1188         IFNET_RUNLOCK();
 1189     } else {
 1190         if (bridge_ifok(dst, src, real_dst))
 1191             last = dst;
 1192     }
 1193     if (last) {
 1194         if (shared) {                   /* need to copy */
 1195             m = m_copypacket(m0, M_DONTWAIT);
 1196             if (m == NULL) {
 1197                 printf("%s: sorry, m_copypacket failed!\n", __func__);
 1198                 bdg_dropped++ ;
 1199                 return m0;              /* the original is still there... */
 1200             }
 1201         } else {                        /* consume original */
 1202             m = m0, m0 = NULL;
 1203         }
 1204         IFQ_HANDOFF(last, m, error);
 1205         if (!error)
 1206             BDG_STAT(last, BDG_OUT);
 1207         else
 1208             bdg_dropped++;
 1209     }
 1210 
 1211     DDB(bdg_fw_ticks += (u_long)(rdtsc() - ticks) ; bdg_fw_count++ ;
 1212         if (bdg_fw_count != 0) bdg_fw_avg = bdg_fw_ticks/bdg_fw_count; )
 1213     return m0;
 1214 #undef EH_RESTORE
 1215 }
 1216 
 1217 /*
 1218  * initialization of bridge code.
 1219  */
 1220 static int
 1221 bdginit(void)
 1222 {
 1223     if (bootverbose)
 1224             printf("BRIDGE %s loaded\n", bridge_version);
 1225 
 1226     ifp2sc = malloc(BDG_MAX_PORTS * sizeof(struct bdg_softc),
 1227                 M_IFADDR, M_WAITOK | M_ZERO );
 1228     if (ifp2sc == NULL)
 1229         return ENOMEM;
 1230 
 1231     BDG_LOCK_INIT();
 1232 
 1233     n_clusters = 0;
 1234     clusters = NULL;
 1235     do_bridge = 0;
 1236 
 1237     bzero(&bdg_stats, sizeof(bdg_stats));
 1238 
 1239     bridge_in_ptr = bridge_in;
 1240     bdg_forward_ptr = bdg_forward;
 1241     bdgtakeifaces_ptr = reconfigure_bridge;
 1242 
 1243     bdgtakeifaces_ptr();                /* XXX does this do anything? */
 1244 
 1245     callout_init(&bdg_callout, NET_CALLOUT_MPSAFE);
 1246     bdg_timeout(0);
 1247     return 0 ;
 1248 }
 1249 
 1250 static void
 1251 bdgdestroy(void)
 1252 {
 1253     bridge_in_ptr = NULL;
 1254     bdg_forward_ptr = NULL;
 1255     bdgtakeifaces_ptr = NULL;
 1256 
 1257     callout_stop(&bdg_callout);
 1258     BDG_LOCK();
 1259     bridge_off();
 1260 
 1261     if (ifp2sc) {
 1262         free(ifp2sc, M_IFADDR);
 1263         ifp2sc = NULL;
 1264     }
 1265     BDG_LOCK_DESTROY();
 1266 }
 1267 
 1268 /*
 1269  * initialization code, both for static and dynamic loading.
 1270  */
 1271 static int
 1272 bridge_modevent(module_t mod, int type, void *unused)
 1273 {
 1274         int err;
 1275 
 1276         switch (type) {
 1277         case MOD_LOAD:
 1278                 if (BDG_LOADED)
 1279                         err = EEXIST;
 1280                 else
 1281                         err = bdginit();
 1282                 break;
 1283         case MOD_UNLOAD:
 1284                 do_bridge = 0;
 1285                 bdgdestroy();
 1286                 err = 0;
 1287                 break;
 1288         default:
 1289                 err = EINVAL;
 1290                 break;
 1291         }
 1292         return err;
 1293 }
 1294 
 1295 static moduledata_t bridge_mod = {
 1296         "bridge",
 1297         bridge_modevent,
 1298         0
 1299 };
 1300 
 1301 DECLARE_MODULE(bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 1302 MODULE_VERSION(bridge, 1);

Cache object: f539972b0f9e3732df69c7aff307598b


[ 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.