Annotation of embedaddon/quagga/zebra/kernel_socket.c, revision 1.1.1.2

1.1       misho       1: /* Kernel communication using routing socket.
                      2:  * Copyright (C) 1999 Kunihiro Ishiguro
                      3:  *
                      4:  * This file is part of GNU Zebra.
                      5:  *
                      6:  * GNU Zebra is free software; you can redistribute it and/or modify it
                      7:  * under the terms of the GNU General Public License as published by the
                      8:  * Free Software Foundation; either version 2, or (at your option) any
                      9:  * later version.
                     10:  *
                     11:  * GNU Zebra is distributed in the hope that it will be useful, but
                     12:  * WITHOUT ANY WARRANTY; without even the implied warranty of
                     13:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
                     14:  * General Public License for more details.
                     15:  *
                     16:  * You should have received a copy of the GNU General Public License
                     17:  * along with GNU Zebra; see the file COPYING.  If not, write to the Free
                     18:  * Software Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
                     19:  * 02111-1307, USA.  
                     20:  */
                     21: 
                     22: #include <zebra.h>
                     23: 
                     24: #include "if.h"
                     25: #include "prefix.h"
                     26: #include "sockunion.h"
                     27: #include "connected.h"
                     28: #include "memory.h"
                     29: #include "ioctl.h"
                     30: #include "log.h"
                     31: #include "str.h"
                     32: #include "table.h"
                     33: #include "rib.h"
                     34: #include "privs.h"
                     35: 
                     36: #include "zebra/interface.h"
                     37: #include "zebra/zserv.h"
                     38: #include "zebra/debug.h"
                     39: #include "zebra/kernel_socket.h"
                     40: 
                     41: extern struct zebra_privs_t zserv_privs;
                     42: extern struct zebra_t zebrad;
                     43: 
                     44: /*
                     45:  * Given a sockaddr length, round it up to include pad bytes following
                     46:  * it.  Assumes the kernel pads to sizeof(long).
                     47:  *
                     48:  * XXX: why is ROUNDUP(0) sizeof(long)?  0 is an illegal sockaddr
                     49:  * length anyway (< sizeof (struct sockaddr)), so this shouldn't
                     50:  * matter.
                     51:  */
                     52: #define ROUNDUP(a) \
                     53:   ((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
                     54: 
                     55: /*
                     56:  * Given a pointer (sockaddr or void *), return the number of bytes
                     57:  * taken up by the sockaddr and any padding needed for alignment.
                     58:  */
                     59: #if defined(HAVE_STRUCT_SOCKADDR_SA_LEN)
                     60: #define SAROUNDUP(X)   ROUNDUP(((struct sockaddr *)(X))->sa_len)
                     61: #elif defined(HAVE_IPV6)
                     62: /*
                     63:  * One would hope all fixed-size structure definitions are aligned,
                     64:  * but round them up nonetheless.
                     65:  */
                     66: #define SAROUNDUP(X) \
                     67:     (((struct sockaddr *)(X))->sa_family == AF_INET ?   \
                     68:       ROUNDUP(sizeof(struct sockaddr_in)):\
                     69:       (((struct sockaddr *)(X))->sa_family == AF_INET6 ? \
                     70:        ROUNDUP(sizeof(struct sockaddr_in6)) :  \
                     71:        (((struct sockaddr *)(X))->sa_family == AF_LINK ? \
                     72:          ROUNDUP(sizeof(struct sockaddr_dl)) : sizeof(struct sockaddr))))
                     73: #else /* HAVE_IPV6 */ 
                     74: #define SAROUNDUP(X) \
                     75:       (((struct sockaddr *)(X))->sa_family == AF_INET ?   \
                     76:         ROUNDUP(sizeof(struct sockaddr_in)):\
                     77:          (((struct sockaddr *)(X))->sa_family == AF_LINK ? \
                     78:            ROUNDUP(sizeof(struct sockaddr_dl)) : sizeof(struct sockaddr)))
                     79: #endif /* HAVE_STRUCT_SOCKADDR_SA_LEN */
                     80: 
                     81: /* We use an additional pointer in following, pdest, rather than (DEST)
                     82:  * directly, because gcc will warn if the macro is expanded and DEST is NULL,
                     83:  * complaining that memcpy is being passed a NULL value, despite the fact
                     84:  * the if (NULL) makes it impossible.
                     85:  */
                     86: #define RTA_ADDR_GET(DEST, RTA, RTMADDRS, PNT) \
                     87:   if ((RTMADDRS) & (RTA)) \
                     88:     { \
                     89:       void *pdest = (DEST); \
                     90:       int len = SAROUNDUP ((PNT)); \
                     91:       if ( ((DEST) != NULL) && \
                     92:            af_check (((struct sockaddr *)(PNT))->sa_family)) \
                     93:         memcpy (pdest, (PNT), len); \
                     94:       (PNT) += len; \
                     95:     }
                     96: #define RTA_ATTR_GET(DEST, RTA, RTMADDRS, PNT) \
                     97:   if ((RTMADDRS) & (RTA)) \
                     98:     { \
                     99:       void *pdest = (DEST); \
                    100:       int len = SAROUNDUP ((PNT)); \
                    101:       if ((DEST) != NULL) \
                    102:         memcpy (pdest, (PNT), len); \
                    103:       (PNT) += len; \
                    104:     }
                    105: 
                    106: #define RTA_NAME_GET(DEST, RTA, RTMADDRS, PNT, LEN) \
                    107:   if ((RTMADDRS) & (RTA)) \
                    108:     { \
                    109:       u_char *pdest = (u_char *) (DEST); \
                    110:       int len = SAROUNDUP ((PNT)); \
                    111:       struct sockaddr_dl *sdl = (struct sockaddr_dl *)(PNT); \
                    112:       if (IS_ZEBRA_DEBUG_KERNEL) \
                    113:         zlog_debug ("%s: RTA_SDL_GET nlen %d, alen %d", \
                    114:                     __func__, sdl->sdl_nlen, sdl->sdl_alen); \
                    115:       if ( ((DEST) != NULL) && (sdl->sdl_family == AF_LINK) \
                    116:            && (sdl->sdl_nlen < IFNAMSIZ) && (sdl->sdl_nlen <= len) ) \
                    117:         { \
                    118:           memcpy (pdest, sdl->sdl_data, sdl->sdl_nlen); \
                    119:           pdest[sdl->sdl_nlen] = '\0'; \
                    120:           (LEN) = sdl->sdl_nlen; \
                    121:         } \
                    122:       (PNT) += len; \
                    123:     } \
                    124:   else \
                    125:     { \
                    126:       (LEN) = 0; \
                    127:     }
                    128: /* Routing socket message types. */
                    129: const struct message rtm_type_str[] =
                    130: {
                    131:   {RTM_ADD,      "RTM_ADD"},
                    132:   {RTM_DELETE,   "RTM_DELETE"},
                    133:   {RTM_CHANGE,   "RTM_CHANGE"},
                    134:   {RTM_GET,      "RTM_GET"},
                    135:   {RTM_LOSING,   "RTM_LOSING"},
                    136:   {RTM_REDIRECT, "RTM_REDIRECT"},
                    137:   {RTM_MISS,     "RTM_MISS"},
                    138:   {RTM_LOCK,     "RTM_LOCK"},
                    139: #ifdef OLDADD
                    140:   {RTM_OLDADD,   "RTM_OLDADD"},
                    141: #endif /* RTM_OLDADD */
                    142: #ifdef RTM_OLDDEL
                    143:   {RTM_OLDDEL,   "RTM_OLDDEL"},
                    144: #endif /* RTM_OLDDEL */
                    145:   {RTM_RESOLVE,  "RTM_RESOLVE"},
                    146:   {RTM_NEWADDR,  "RTM_NEWADDR"},
                    147:   {RTM_DELADDR,  "RTM_DELADDR"},
                    148:   {RTM_IFINFO,   "RTM_IFINFO"},
                    149: #ifdef RTM_OIFINFO
                    150:   {RTM_OIFINFO,   "RTM_OIFINFO"},
                    151: #endif /* RTM_OIFINFO */
                    152: #ifdef RTM_NEWMADDR
                    153:   {RTM_NEWMADDR, "RTM_NEWMADDR"},
                    154: #endif /* RTM_NEWMADDR */
                    155: #ifdef RTM_DELMADDR
                    156:   {RTM_DELMADDR, "RTM_DELMADDR"},
                    157: #endif /* RTM_DELMADDR */
                    158: #ifdef RTM_IFANNOUNCE
                    159:   {RTM_IFANNOUNCE, "RTM_IFANNOUNCE"},
                    160: #endif /* RTM_IFANNOUNCE */
                    161:   {0,            NULL}
                    162: };
                    163: 
                    164: static const struct message rtm_flag_str[] =
                    165: {
                    166:   {RTF_UP,        "UP"},
                    167:   {RTF_GATEWAY,   "GATEWAY"},
                    168:   {RTF_HOST,      "HOST"},
                    169:   {RTF_REJECT,    "REJECT"},
                    170:   {RTF_DYNAMIC,   "DYNAMIC"},
                    171:   {RTF_MODIFIED,  "MODIFIED"},
                    172:   {RTF_DONE,      "DONE"},
                    173: #ifdef RTF_MASK
                    174:   {RTF_MASK,      "MASK"},
                    175: #endif /* RTF_MASK */
                    176: #ifdef RTF_CLONING
                    177:   {RTF_CLONING,   "CLONING"},
                    178: #endif /* RTF_CLONING */
                    179:   {RTF_XRESOLVE,  "XRESOLVE"},
                    180:   {RTF_LLINFO,    "LLINFO"},
                    181:   {RTF_STATIC,    "STATIC"},
                    182:   {RTF_BLACKHOLE, "BLACKHOLE"},
                    183: #ifdef RTF_PRIVATE
                    184:   {RTF_PRIVATE,          "PRIVATE"},
                    185: #endif /* RTF_PRIVATE */
                    186:   {RTF_PROTO1,    "PROTO1"},
                    187:   {RTF_PROTO2,    "PROTO2"},
                    188: #ifdef RTF_PRCLONING
                    189:   {RTF_PRCLONING, "PRCLONING"},
                    190: #endif /* RTF_PRCLONING */
                    191: #ifdef RTF_WASCLONED
                    192:   {RTF_WASCLONED, "WASCLONED"},
                    193: #endif /* RTF_WASCLONED */
                    194: #ifdef RTF_PROTO3
                    195:   {RTF_PROTO3,    "PROTO3"},
                    196: #endif /* RTF_PROTO3 */
                    197: #ifdef RTF_PINNED
                    198:   {RTF_PINNED,    "PINNED"},
                    199: #endif /* RTF_PINNED */
                    200: #ifdef RTF_LOCAL
                    201:   {RTF_LOCAL,    "LOCAL"},
                    202: #endif /* RTF_LOCAL */
                    203: #ifdef RTF_BROADCAST
                    204:   {RTF_BROADCAST, "BROADCAST"},
                    205: #endif /* RTF_BROADCAST */
                    206: #ifdef RTF_MULTICAST
                    207:   {RTF_MULTICAST, "MULTICAST"},
                    208: #endif /* RTF_MULTICAST */
                    209: #ifdef RTF_MULTIRT
                    210:   {RTF_MULTIRT,   "MULTIRT"},
                    211: #endif /* RTF_MULTIRT */
                    212: #ifdef RTF_SETSRC
                    213:   {RTF_SETSRC,    "SETSRC"},
                    214: #endif /* RTF_SETSRC */
                    215:   {0,             NULL}
                    216: };
                    217: 
                    218: /* Kernel routing update socket. */
                    219: int routing_sock = -1;
                    220: 
                    221: /* Yes I'm checking ugly routing socket behavior. */
                    222: /* #define DEBUG */
                    223: 
                    224: /* Supported address family check. */
                    225: static int inline
                    226: af_check (int family)
                    227: {
                    228:   if (family == AF_INET)
                    229:     return 1;
                    230: #ifdef HAVE_IPV6
                    231:   if (family == AF_INET6)
                    232:     return 1;
                    233: #endif /* HAVE_IPV6 */
                    234:   return 0;
                    235: }
                    236: 
                    237: /* Dump routing table flag for debug purpose. */
                    238: static void
                    239: rtm_flag_dump (int flag)
                    240: {
                    241:   const struct message *mes;
                    242:   static char buf[BUFSIZ];
                    243: 
                    244:   buf[0] = '\0';
                    245:   for (mes = rtm_flag_str; mes->key != 0; mes++)
                    246:     {
                    247:       if (mes->key & flag)
                    248:        {
                    249:          strlcat (buf, mes->str, BUFSIZ);
                    250:          strlcat (buf, " ", BUFSIZ);
                    251:        }
                    252:     }
                    253:   zlog_debug ("Kernel: %s", buf);
                    254: }
                    255: 
                    256: #ifdef RTM_IFANNOUNCE
                    257: /* Interface adding function */
                    258: static int
                    259: ifan_read (struct if_announcemsghdr *ifan)
                    260: {
                    261:   struct interface *ifp;
                    262:   
                    263:   ifp = if_lookup_by_index (ifan->ifan_index);
                    264:   
                    265:   if (ifp)
                    266:     assert ( (ifp->ifindex == ifan->ifan_index) 
                    267:              || (ifp->ifindex == IFINDEX_INTERNAL) );
                    268: 
                    269:   if ( (ifp == NULL) 
                    270:       || ((ifp->ifindex == IFINDEX_INTERNAL)
                    271:           && (ifan->ifan_what == IFAN_ARRIVAL)) )
                    272:     {
                    273:       if (IS_ZEBRA_DEBUG_KERNEL)
                    274:         zlog_debug ("%s: creating interface for ifindex %d, name %s",
                    275:                     __func__, ifan->ifan_index, ifan->ifan_name);
                    276:       
                    277:       /* Create Interface */
                    278:       ifp = if_get_by_name_len(ifan->ifan_name,
                    279:                               strnlen(ifan->ifan_name,
                    280:                                       sizeof(ifan->ifan_name)));
                    281:       ifp->ifindex = ifan->ifan_index;
                    282: 
                    283:       if_add_update (ifp);
                    284:     }
                    285:   else if (ifp != NULL && ifan->ifan_what == IFAN_DEPARTURE)
                    286:     if_delete_update (ifp);
                    287: 
                    288:   if_get_flags (ifp);
                    289:   if_get_mtu (ifp);
                    290:   if_get_metric (ifp);
                    291: 
                    292:   if (IS_ZEBRA_DEBUG_KERNEL)
                    293:     zlog_debug ("%s: interface %s index %d", 
                    294:                 __func__, ifan->ifan_name, ifan->ifan_index);
                    295: 
                    296:   return 0;
                    297: }
                    298: #endif /* RTM_IFANNOUNCE */
                    299: 
                    300: #ifdef HAVE_BSD_LINK_DETECT
                    301: /* BSD link detect translation */
                    302: static void
                    303: bsd_linkdetect_translate (struct if_msghdr *ifm)
                    304: {
                    305:   if ((ifm->ifm_data.ifi_link_state >= LINK_STATE_UP) ||
                    306:       (ifm->ifm_data.ifi_link_state == LINK_STATE_UNKNOWN))
                    307:     SET_FLAG(ifm->ifm_flags, IFF_RUNNING);
                    308:   else
                    309:     UNSET_FLAG(ifm->ifm_flags, IFF_RUNNING);
                    310: }
                    311: #endif /* HAVE_BSD_LINK_DETECT */
                    312: 
                    313: /*
                    314:  * Handle struct if_msghdr obtained from reading routing socket or
                    315:  * sysctl (from interface_list).  There may or may not be sockaddrs
                    316:  * present after the header.
                    317:  */
                    318: int
                    319: ifm_read (struct if_msghdr *ifm)
                    320: {
                    321:   struct interface *ifp = NULL;
1.1.1.2 ! misho     322:   struct sockaddr_dl *sdl;
1.1       misho     323:   char ifname[IFNAMSIZ];
                    324:   short ifnlen = 0;
                    325:   caddr_t *cp;
                    326:   
                    327:   /* terminate ifname at head (for strnlen) and tail (for safety) */
                    328:   ifname[IFNAMSIZ - 1] = '\0';
                    329:   
                    330:   /* paranoia: sanity check structure */
                    331:   if (ifm->ifm_msglen < sizeof(struct if_msghdr))
                    332:     {
                    333:       zlog_err ("ifm_read: ifm->ifm_msglen %d too short\n",
                    334:                ifm->ifm_msglen);
                    335:       return -1;
                    336:     }
                    337: 
                    338:   /*
                    339:    * Check for a sockaddr_dl following the message.  First, point to
                    340:    * where a socakddr might be if one follows the message.
                    341:    */
                    342:   cp = (void *)(ifm + 1);
                    343: 
                    344: #ifdef SUNOS_5
                    345:   /* 
                    346:    * XXX This behavior should be narrowed to only the kernel versions
                    347:    * for which the structures returned do not match the headers.
                    348:    *
                    349:    * if_msghdr_t on 64 bit kernels in Solaris 9 and earlier versions
                    350:    * is 12 bytes larger than the 32 bit version.
                    351:    */
                    352:   if (((struct sockaddr *) cp)->sa_family == AF_UNSPEC)
                    353:        cp = cp + 12;
                    354: #endif
                    355: 
                    356:   RTA_ADDR_GET (NULL, RTA_DST, ifm->ifm_addrs, cp);
                    357:   RTA_ADDR_GET (NULL, RTA_GATEWAY, ifm->ifm_addrs, cp);
                    358:   RTA_ATTR_GET (NULL, RTA_NETMASK, ifm->ifm_addrs, cp);
                    359:   RTA_ADDR_GET (NULL, RTA_GENMASK, ifm->ifm_addrs, cp);
1.1.1.2 ! misho     360:   sdl = (struct sockaddr_dl *)cp;
1.1       misho     361:   RTA_NAME_GET (ifname, RTA_IFP, ifm->ifm_addrs, cp, ifnlen);
                    362:   RTA_ADDR_GET (NULL, RTA_IFA, ifm->ifm_addrs, cp);
                    363:   RTA_ADDR_GET (NULL, RTA_AUTHOR, ifm->ifm_addrs, cp);
                    364:   RTA_ADDR_GET (NULL, RTA_BRD, ifm->ifm_addrs, cp);
                    365:   
                    366:   if (IS_ZEBRA_DEBUG_KERNEL)
                    367:     zlog_debug ("%s: sdl ifname %s", __func__, (ifnlen ? ifname : "(nil)"));
                    368:   
                    369:   /* 
                    370:    * Look up on ifindex first, because ifindices are the primary handle for
                    371:    * interfaces across the user/kernel boundary, for most systems.  (Some
                    372:    * messages, such as up/down status changes on NetBSD, do not include a
                    373:    * sockaddr_dl).
                    374:    */
                    375:   if ( (ifp = if_lookup_by_index (ifm->ifm_index)) != NULL )
                    376:     {
                    377:       /* we have an ifp, verify that the name matches as some systems,
                    378:        * eg Solaris, have a 1:many association of ifindex:ifname
                    379:        * if they dont match, we dont have the correct ifp and should
                    380:        * set it back to NULL to let next check do lookup by name
                    381:        */
                    382:       if (ifnlen && (strncmp (ifp->name, ifname, IFNAMSIZ) != 0) )
                    383:         {
                    384:           if (IS_ZEBRA_DEBUG_KERNEL)
                    385:             zlog_debug ("%s: ifp name %s doesnt match sdl name %s",
                    386:                         __func__, ifp->name, ifname);
                    387:           ifp = NULL;
                    388:         }
                    389:     }
                    390:   
                    391:   /* 
                    392:    * If we dont have an ifp, try looking up by name.  Particularly as some
                    393:    * systems (Solaris) have a 1:many mapping of ifindex:ifname - the ifname
                    394:    * is therefore our unique handle to that interface.
                    395:    *
                    396:    * Interfaces specified in the configuration file for which the ifindex
                    397:    * has not been determined will have ifindex == IFINDEX_INTERNAL, and such
                    398:    * interfaces are found by this search, and then their ifindex values can
                    399:    * be filled in.
                    400:    */
                    401:   if ( (ifp == NULL) && ifnlen)
                    402:     ifp = if_lookup_by_name (ifname);
                    403: 
                    404:   /*
                    405:    * If ifp still does not exist or has an invalid index (IFINDEX_INTERNAL),
                    406:    * create or fill in an interface.
                    407:    */
                    408:   if ((ifp == NULL) || (ifp->ifindex == IFINDEX_INTERNAL))
                    409:     {
                    410:       /*
                    411:        * To create or fill in an interface, a sockaddr_dl (via
                    412:        * RTA_IFP) is required.
                    413:        */
                    414:       if (!ifnlen)
                    415:        {
                    416:          zlog_warn ("Interface index %d (new) missing ifname\n",
                    417:                     ifm->ifm_index);
                    418:          return -1;
                    419:        }
                    420: 
                    421: #ifndef RTM_IFANNOUNCE
                    422:       /* Down->Down interface should be ignored here.
                    423:        * See further comment below.
                    424:        */
                    425:       if (!CHECK_FLAG (ifm->ifm_flags, IFF_UP))
                    426:         return 0;
                    427: #endif /* !RTM_IFANNOUNCE */
                    428:       
                    429:       if (ifp == NULL)
                    430:         {
                    431:          /* Interface that zebra was not previously aware of, so create. */ 
                    432:          ifp = if_create (ifname, ifnlen);
                    433:          if (IS_ZEBRA_DEBUG_KERNEL)
                    434:            zlog_debug ("%s: creating ifp for ifindex %d", 
                    435:                        __func__, ifm->ifm_index);
                    436:         }
                    437: 
                    438:       if (IS_ZEBRA_DEBUG_KERNEL)
                    439:         zlog_debug ("%s: updated/created ifp, ifname %s, ifindex %d",
                    440:                     __func__, ifp->name, ifp->ifindex);
                    441:       /* 
                    442:        * Fill in newly created interface structure, or larval
                    443:        * structure with ifindex IFINDEX_INTERNAL.
                    444:        */
                    445:       ifp->ifindex = ifm->ifm_index;
                    446:       
                    447: #ifdef HAVE_BSD_LINK_DETECT /* translate BSD kernel msg for link-state */
                    448:       bsd_linkdetect_translate(ifm);
                    449: #endif /* HAVE_BSD_LINK_DETECT */
                    450: 
                    451:       if_flags_update (ifp, ifm->ifm_flags);
                    452: #if defined(__bsdi__)
                    453:       if_kvm_get_mtu (ifp);
                    454: #else
                    455:       if_get_mtu (ifp);
                    456: #endif /* __bsdi__ */
                    457:       if_get_metric (ifp);
                    458: 
1.1.1.2 ! misho     459:       /*
        !           460:        * XXX sockaddr_dl contents can be larger than the structure
        !           461:        * definition, so the user of the stored structure must be
        !           462:        * careful not to read off the end.
        !           463:        *
        !           464:        * a nonzero ifnlen from RTA_NAME_GET() means sdl is valid
        !           465:        */
        !           466:       if (ifnlen)
        !           467:        memcpy (&ifp->sdl, sdl, sizeof (struct sockaddr_dl));
        !           468: 
1.1       misho     469:       if_add_update (ifp);
                    470:     }
                    471:   else
                    472:     /*
                    473:      * Interface structure exists.  Adjust stored flags from
                    474:      * notification.  If interface has up->down or down->up
                    475:      * transition, call state change routines (to adjust routes,
                    476:      * notify routing daemons, etc.).  (Other flag changes are stored
                    477:      * but apparently do not trigger action.)
                    478:      */
                    479:     {
                    480:       if (ifp->ifindex != ifm->ifm_index)
                    481:         {
                    482:           zlog_warn ("%s: index mismatch, ifname %s, ifp index %d, "
                    483:                      "ifm index %d", 
                    484:                      __func__, ifp->name, ifp->ifindex, ifm->ifm_index);
                    485:           return -1;
                    486:         }
                    487:       
                    488: #ifdef HAVE_BSD_LINK_DETECT /* translate BSD kernel msg for link-state */
                    489:       bsd_linkdetect_translate(ifm);
                    490: #endif /* HAVE_BSD_LINK_DETECT */
                    491: 
                    492:       /* update flags and handle operative->inoperative transition, if any */
                    493:       if_flags_update (ifp, ifm->ifm_flags);
                    494:       
                    495: #ifndef RTM_IFANNOUNCE
                    496:       if (!if_is_up (ifp))
                    497:           {
                    498:             /* No RTM_IFANNOUNCE on this platform, so we can never
                    499:              * distinguish between ~IFF_UP and delete. We must presume
                    500:              * it has been deleted.
                    501:              * Eg, Solaris will not notify us of unplumb.
                    502:              *
                    503:              * XXX: Fixme - this should be runtime detected
                    504:              * So that a binary compiled on a system with IFANNOUNCE
                    505:              * will still behave correctly if run on a platform without
                    506:              */
                    507:             if_delete_update (ifp);
                    508:           }
                    509: #endif /* RTM_IFANNOUNCE */
                    510:       if (if_is_up (ifp))
                    511:       {
                    512: #if defined(__bsdi__)
                    513:         if_kvm_get_mtu (ifp);
                    514: #else
                    515:         if_get_mtu (ifp);
                    516: #endif /* __bsdi__ */
                    517:         if_get_metric (ifp);
                    518:       }
                    519:     }
                    520: 
                    521: #ifdef HAVE_NET_RT_IFLIST
                    522:   ifp->stats = ifm->ifm_data;
                    523: #endif /* HAVE_NET_RT_IFLIST */
                    524: 
                    525:   if (IS_ZEBRA_DEBUG_KERNEL)
                    526:     zlog_debug ("%s: interface %s index %d", 
                    527:                 __func__, ifp->name, ifp->ifindex);
                    528: 
                    529:   return 0;
                    530: }
                    531: 
                    532: /* Address read from struct ifa_msghdr. */
                    533: static void
                    534: ifam_read_mesg (struct ifa_msghdr *ifm,
                    535:                union sockunion *addr,
                    536:                union sockunion *mask,
                    537:                union sockunion *brd,
                    538:                char *ifname,
                    539:                short *ifnlen)
                    540: {
                    541:   caddr_t pnt, end;
                    542:   union sockunion dst;
                    543:   union sockunion gateway;
                    544: 
                    545:   pnt = (caddr_t)(ifm + 1);
                    546:   end = ((caddr_t)ifm) + ifm->ifam_msglen;
                    547: 
                    548:   /* Be sure structure is cleared */
                    549:   memset (mask, 0, sizeof (union sockunion));
                    550:   memset (addr, 0, sizeof (union sockunion));
                    551:   memset (brd, 0, sizeof (union sockunion));
                    552:   memset (&dst, 0, sizeof (union sockunion));
                    553:   memset (&gateway, 0, sizeof (union sockunion));
                    554: 
                    555:   /* We fetch each socket variable into sockunion. */
                    556:   RTA_ADDR_GET (&dst, RTA_DST, ifm->ifam_addrs, pnt);
                    557:   RTA_ADDR_GET (&gateway, RTA_GATEWAY, ifm->ifam_addrs, pnt);
                    558:   RTA_ATTR_GET (mask, RTA_NETMASK, ifm->ifam_addrs, pnt);
                    559:   RTA_ADDR_GET (NULL, RTA_GENMASK, ifm->ifam_addrs, pnt);
                    560:   RTA_NAME_GET (ifname, RTA_IFP, ifm->ifam_addrs, pnt, *ifnlen);
                    561:   RTA_ADDR_GET (addr, RTA_IFA, ifm->ifam_addrs, pnt);
                    562:   RTA_ADDR_GET (NULL, RTA_AUTHOR, ifm->ifam_addrs, pnt);
                    563:   RTA_ADDR_GET (brd, RTA_BRD, ifm->ifam_addrs, pnt);
                    564: 
                    565:   if (IS_ZEBRA_DEBUG_KERNEL)
                    566:     {
                    567:       switch (sockunion_family(addr))
                    568:         {
                    569:        case AF_INET:
                    570:          {
                    571:            char buf[4][INET_ADDRSTRLEN];
                    572:            zlog_debug ("%s: ifindex %d, ifname %s, ifam_addrs 0x%x, "
                    573:                        "ifam_flags 0x%x, addr %s/%d broad %s dst %s "
                    574:                        "gateway %s",
                    575:                        __func__, ifm->ifam_index,
                    576:                        (ifnlen ? ifname : "(nil)"), ifm->ifam_addrs,
                    577:                        ifm->ifam_flags,
                    578:                        inet_ntop(AF_INET,&addr->sin.sin_addr,
                    579:                                  buf[0],sizeof(buf[0])),
                    580:                        ip_masklen(mask->sin.sin_addr),
                    581:                        inet_ntop(AF_INET,&brd->sin.sin_addr,
                    582:                                  buf[1],sizeof(buf[1])),
                    583:                        inet_ntop(AF_INET,&dst.sin.sin_addr,
                    584:                                  buf[2],sizeof(buf[2])),
                    585:                        inet_ntop(AF_INET,&gateway.sin.sin_addr,
                    586:                                  buf[3],sizeof(buf[3])));
                    587:          }
                    588:          break;
                    589: #ifdef HAVE_IPV6
                    590:        case AF_INET6:
                    591:          {
                    592:            char buf[4][INET6_ADDRSTRLEN];
                    593:            zlog_debug ("%s: ifindex %d, ifname %s, ifam_addrs 0x%x, "
                    594:                        "ifam_flags 0x%x, addr %s/%d broad %s dst %s "
                    595:                        "gateway %s",
                    596:                        __func__, ifm->ifam_index, 
                    597:                        (ifnlen ? ifname : "(nil)"), ifm->ifam_addrs,
                    598:                        ifm->ifam_flags,
                    599:                        inet_ntop(AF_INET6,&addr->sin6.sin6_addr,
                    600:                                  buf[0],sizeof(buf[0])),
                    601:                        ip6_masklen(mask->sin6.sin6_addr),
                    602:                        inet_ntop(AF_INET6,&brd->sin6.sin6_addr,
                    603:                                  buf[1],sizeof(buf[1])),
                    604:                        inet_ntop(AF_INET6,&dst.sin6.sin6_addr,
                    605:                                  buf[2],sizeof(buf[2])),
                    606:                        inet_ntop(AF_INET6,&gateway.sin6.sin6_addr,
                    607:                                  buf[3],sizeof(buf[3])));
                    608:          }
                    609:          break;
                    610: #endif /* HAVE_IPV6 */
                    611:         default:
                    612:          zlog_debug ("%s: ifindex %d, ifname %s, ifam_addrs 0x%x",
                    613:                      __func__, ifm->ifam_index, 
                    614:                      (ifnlen ? ifname : "(nil)"), ifm->ifam_addrs);
                    615:          break;
                    616:         }
                    617:     }
                    618: 
                    619:   /* Assert read up end point matches to end point */
                    620:   if (pnt != end)
1.1.1.2 ! misho     621:     zlog_warn ("ifam_read() doesn't read all socket data");
1.1       misho     622: }
                    623: 
                    624: /* Interface's address information get. */
                    625: int
                    626: ifam_read (struct ifa_msghdr *ifam)
                    627: {
                    628:   struct interface *ifp = NULL;
                    629:   union sockunion addr, mask, brd;
                    630:   char ifname[INTERFACE_NAMSIZ];
                    631:   short ifnlen = 0;
                    632:   char isalias = 0;
                    633:   int flags = 0;
                    634:   
                    635:   ifname[0] = ifname[INTERFACE_NAMSIZ - 1] = '\0';
                    636:   
                    637:   /* Allocate and read address information. */
                    638:   ifam_read_mesg (ifam, &addr, &mask, &brd, ifname, &ifnlen);
                    639:   
                    640:   if ((ifp = if_lookup_by_index(ifam->ifam_index)) == NULL)
                    641:     {
                    642:       zlog_warn ("%s: no interface for ifname %s, index %d", 
                    643:                  __func__, ifname, ifam->ifam_index);
                    644:       return -1;
                    645:     }
                    646:   
                    647:   if (ifnlen && strncmp (ifp->name, ifname, INTERFACE_NAMSIZ))
                    648:     isalias = 1;
                    649:   
                    650:   /* N.B. The info in ifa_msghdr does not tell us whether the RTA_BRD
                    651:      field contains a broadcast address or a peer address, so we are forced to
                    652:      rely upon the interface type. */
                    653:   if (if_is_pointopoint(ifp))
                    654:     SET_FLAG(flags, ZEBRA_IFA_PEER);
                    655: 
                    656: #if 0
                    657:   /* it might seem cute to grab the interface metric here, however
                    658:    * we're processing an address update message, and so some systems
                    659:    * (e.g. FBSD) dont bother to fill in ifam_metric. Disabled, but left
                    660:    * in deliberately, as comment.
                    661:    */
                    662:   ifp->metric = ifam->ifam_metric;
                    663: #endif
                    664: 
                    665:   /* Add connected address. */
                    666:   switch (sockunion_family (&addr))
                    667:     {
                    668:     case AF_INET:
                    669:       if (ifam->ifam_type == RTM_NEWADDR)
                    670:        connected_add_ipv4 (ifp, flags, &addr.sin.sin_addr, 
                    671:                            ip_masklen (mask.sin.sin_addr),
                    672:                            &brd.sin.sin_addr,
                    673:                            (isalias ? ifname : NULL));
                    674:       else
                    675:        connected_delete_ipv4 (ifp, flags, &addr.sin.sin_addr, 
                    676:                               ip_masklen (mask.sin.sin_addr),
                    677:                               &brd.sin.sin_addr);
                    678:       break;
                    679: #ifdef HAVE_IPV6
                    680:     case AF_INET6:
                    681:       /* Unset interface index from link-local address when IPv6 stack
                    682:         is KAME. */
                    683:       if (IN6_IS_ADDR_LINKLOCAL (&addr.sin6.sin6_addr))
                    684:        SET_IN6_LINKLOCAL_IFINDEX (addr.sin6.sin6_addr, 0);
                    685: 
                    686:       if (ifam->ifam_type == RTM_NEWADDR)
                    687:        connected_add_ipv6 (ifp, flags, &addr.sin6.sin6_addr, 
                    688:                            ip6_masklen (mask.sin6.sin6_addr),
                    689:                            &brd.sin6.sin6_addr,
                    690:                            (isalias ? ifname : NULL));
                    691:       else
                    692:        connected_delete_ipv6 (ifp,
                    693:                               &addr.sin6.sin6_addr, 
                    694:                               ip6_masklen (mask.sin6.sin6_addr),
                    695:                               &brd.sin6.sin6_addr);
                    696:       break;
                    697: #endif /* HAVE_IPV6 */
                    698:     default:
                    699:       /* Unsupported family silently ignore... */
                    700:       break;
                    701:     }
                    702:   
                    703:   /* Check interface flag for implicit up of the interface. */
                    704:   if_refresh (ifp);
                    705: 
                    706: #ifdef SUNOS_5
                    707:   /* In addition to lacking IFANNOUNCE, on SUNOS IFF_UP is strange. 
                    708:    * See comments for SUNOS_5 in interface.c::if_flags_mangle.
                    709:    * 
                    710:    * Here we take care of case where the real IFF_UP was previously
                    711:    * unset (as kept in struct zebra_if.primary_state) and the mangled
                    712:    * IFF_UP (ie IFF_UP set || listcount(connected) has now transitioned
                    713:    * to unset due to the lost non-primary address having DELADDR'd.
                    714:    *
                    715:    * we must delete the interface, because in between here and next
                    716:    * event for this interface-name the administrator could unplumb
                    717:    * and replumb the interface.
                    718:    */
                    719:   if (!if_is_up (ifp))
                    720:     if_delete_update (ifp);
                    721: #endif /* SUNOS_5 */
                    722:   
                    723:   return 0;
                    724: }
                    725: 
                    726: /* Interface function for reading kernel routing table information. */
                    727: static int
                    728: rtm_read_mesg (struct rt_msghdr *rtm,
                    729:               union sockunion *dest,
                    730:               union sockunion *mask,
                    731:               union sockunion *gate,
                    732:               char *ifname,
                    733:               short *ifnlen)
                    734: {
                    735:   caddr_t pnt, end;
                    736: 
                    737:   /* Pnt points out socket data start point. */
                    738:   pnt = (caddr_t)(rtm + 1);
                    739:   end = ((caddr_t)rtm) + rtm->rtm_msglen;
                    740: 
                    741:   /* rt_msghdr version check. */
                    742:   if (rtm->rtm_version != RTM_VERSION) 
                    743:       zlog (NULL, LOG_WARNING,
                    744:              "Routing message version different %d should be %d."
                    745:              "This may cause problem\n", rtm->rtm_version, RTM_VERSION);
                    746:   
                    747:   /* Be sure structure is cleared */
                    748:   memset (dest, 0, sizeof (union sockunion));
                    749:   memset (gate, 0, sizeof (union sockunion));
                    750:   memset (mask, 0, sizeof (union sockunion));
                    751: 
                    752:   /* We fetch each socket variable into sockunion. */
                    753:   RTA_ADDR_GET (dest, RTA_DST, rtm->rtm_addrs, pnt);
                    754:   RTA_ADDR_GET (gate, RTA_GATEWAY, rtm->rtm_addrs, pnt);
                    755:   RTA_ATTR_GET (mask, RTA_NETMASK, rtm->rtm_addrs, pnt);
                    756:   RTA_ADDR_GET (NULL, RTA_GENMASK, rtm->rtm_addrs, pnt);
                    757:   RTA_NAME_GET (ifname, RTA_IFP, rtm->rtm_addrs, pnt, *ifnlen);
                    758:   RTA_ADDR_GET (NULL, RTA_IFA, rtm->rtm_addrs, pnt);
                    759:   RTA_ADDR_GET (NULL, RTA_AUTHOR, rtm->rtm_addrs, pnt);
                    760:   RTA_ADDR_GET (NULL, RTA_BRD, rtm->rtm_addrs, pnt);
                    761: 
                    762:   /* If there is netmask information set it's family same as
                    763:      destination family*/
                    764:   if (rtm->rtm_addrs & RTA_NETMASK)
                    765:     mask->sa.sa_family = dest->sa.sa_family;
                    766: 
                    767:   /* Assert read up to the end of pointer. */
                    768:   if (pnt != end) 
1.1.1.2 ! misho     769:       zlog (NULL, LOG_WARNING, "rtm_read() doesn't read all socket data.");
1.1       misho     770: 
                    771:   return rtm->rtm_flags;
                    772: }
                    773: 
                    774: void
                    775: rtm_read (struct rt_msghdr *rtm)
                    776: {
                    777:   int flags;
                    778:   u_char zebra_flags;
                    779:   union sockunion dest, mask, gate;
                    780:   char ifname[INTERFACE_NAMSIZ + 1];
                    781:   short ifnlen = 0;
                    782: 
                    783:   zebra_flags = 0;
                    784: 
                    785:   /* Read destination and netmask and gateway from rtm message
                    786:      structure. */
                    787:   flags = rtm_read_mesg (rtm, &dest, &mask, &gate, ifname, &ifnlen);
                    788:   if (!(flags & RTF_DONE))
                    789:     return;
                    790:   if (IS_ZEBRA_DEBUG_KERNEL)
                    791:     zlog_debug ("%s: got rtm of type %d (%s)", __func__, rtm->rtm_type,
                    792:       lookup (rtm_type_str, rtm->rtm_type));
                    793: 
                    794: #ifdef RTF_CLONED      /*bsdi, netbsd 1.6*/
                    795:   if (flags & RTF_CLONED)
                    796:     return;
                    797: #endif
                    798: #ifdef RTF_WASCLONED   /*freebsd*/
                    799:   if (flags & RTF_WASCLONED)
                    800:     return;
                    801: #endif
                    802: 
                    803:   if ((rtm->rtm_type == RTM_ADD) && ! (flags & RTF_UP))
                    804:     return;
                    805: 
                    806:   /* This is connected route. */
                    807:   if (! (flags & RTF_GATEWAY))
                    808:       return;
                    809: 
                    810:   if (flags & RTF_PROTO1)
                    811:     SET_FLAG (zebra_flags, ZEBRA_FLAG_SELFROUTE);
                    812: 
                    813:   /* This is persistent route. */
                    814:   if (flags & RTF_STATIC)
                    815:     SET_FLAG (zebra_flags, ZEBRA_FLAG_STATIC);
                    816: 
                    817:   /* This is a reject or blackhole route */
                    818:   if (flags & RTF_REJECT)
                    819:     SET_FLAG (zebra_flags, ZEBRA_FLAG_REJECT);
                    820:   if (flags & RTF_BLACKHOLE)
                    821:     SET_FLAG (zebra_flags, ZEBRA_FLAG_BLACKHOLE);
                    822: 
                    823:   if (dest.sa.sa_family == AF_INET)
                    824:     {
                    825:       struct prefix_ipv4 p;
                    826: 
                    827:       p.family = AF_INET;
                    828:       p.prefix = dest.sin.sin_addr;
                    829:       if (flags & RTF_HOST)
                    830:        p.prefixlen = IPV4_MAX_PREFIXLEN;
                    831:       else
                    832:        p.prefixlen = ip_masklen (mask.sin.sin_addr);
                    833:       
                    834:       /* Catch self originated messages and match them against our current RIB.
                    835:        * At the same time, ignore unconfirmed messages, they should be tracked
                    836:        * by rtm_write() and kernel_rtm_ipv4().
                    837:        */
                    838:       if (rtm->rtm_type != RTM_GET && rtm->rtm_pid == pid)
                    839:       {
                    840:         char buf[INET_ADDRSTRLEN], gate_buf[INET_ADDRSTRLEN];
                    841:         int ret;
                    842:         if (! IS_ZEBRA_DEBUG_RIB)
                    843:           return;
                    844:         ret = rib_lookup_ipv4_route (&p, &gate); 
                    845:         inet_ntop (AF_INET, &p.prefix, buf, INET_ADDRSTRLEN);
                    846:         switch (rtm->rtm_type)
                    847:         {
                    848:           case RTM_ADD:
                    849:           case RTM_GET:
                    850:           case RTM_CHANGE:
                    851:             /* The kernel notifies us about a new route in FIB created by us.
                    852:                Do we have a correspondent entry in our RIB? */
                    853:             switch (ret)
                    854:             {
                    855:               case ZEBRA_RIB_NOTFOUND:
                    856:                 zlog_debug ("%s: %s %s/%d: desync: RR isn't yet in RIB, while already in FIB",
                    857:                   __func__, lookup (rtm_type_str, rtm->rtm_type), buf, p.prefixlen);
                    858:                 break;
                    859:               case ZEBRA_RIB_FOUND_CONNECTED:
                    860:               case ZEBRA_RIB_FOUND_NOGATE:
                    861:                 inet_ntop (AF_INET, &gate.sin.sin_addr, gate_buf, INET_ADDRSTRLEN);
                    862:                 zlog_debug ("%s: %s %s/%d: desync: RR is in RIB, but gate differs (ours is %s)",
                    863:                   __func__, lookup (rtm_type_str, rtm->rtm_type), buf, p.prefixlen, gate_buf);
                    864:                 break;
                    865:               case ZEBRA_RIB_FOUND_EXACT: /* RIB RR == FIB RR */
                    866:                 zlog_debug ("%s: %s %s/%d: done Ok",
                    867:                   __func__, lookup (rtm_type_str, rtm->rtm_type), buf, p.prefixlen);
                    868:                 rib_lookup_and_dump (&p);
                    869:                 return;
                    870:                 break;
                    871:             }
                    872:             break;
                    873:           case RTM_DELETE:
                    874:             /* The kernel notifies us about a route deleted by us. Do we still
                    875:                have it in the RIB? Do we have anything instead? */
                    876:             switch (ret)
                    877:             {
                    878:               case ZEBRA_RIB_FOUND_EXACT:
                    879:                 zlog_debug ("%s: %s %s/%d: desync: RR is still in RIB, while already not in FIB",
                    880:                   __func__, lookup (rtm_type_str, rtm->rtm_type), buf, p.prefixlen);
                    881:                 rib_lookup_and_dump (&p);
                    882:                 break;
                    883:               case ZEBRA_RIB_FOUND_CONNECTED:
                    884:               case ZEBRA_RIB_FOUND_NOGATE:
                    885:                 zlog_debug ("%s: %s %s/%d: desync: RR is still in RIB, plus gate differs",
                    886:                   __func__, lookup (rtm_type_str, rtm->rtm_type), buf, p.prefixlen);
                    887:                 rib_lookup_and_dump (&p);
                    888:                 break;
                    889:               case ZEBRA_RIB_NOTFOUND: /* RIB RR == FIB RR */
                    890:                 zlog_debug ("%s: %s %s/%d: done Ok",
                    891:                   __func__, lookup (rtm_type_str, rtm->rtm_type), buf, p.prefixlen);
                    892:                 rib_lookup_and_dump (&p);
                    893:                 return;
                    894:                 break;
                    895:             }
                    896:             break;
                    897:           default:
                    898:             zlog_debug ("%s: %s/%d: warning: loopback RTM of type %s received",
                    899:               __func__, buf, p.prefixlen, lookup (rtm_type_str, rtm->rtm_type));
                    900:         }
                    901:         return;
                    902:       }
                    903: 
                    904:       /* Change, delete the old prefix, we have no further information
                    905:        * to specify the route really
                    906:        */
                    907:       if (rtm->rtm_type == RTM_CHANGE)
                    908:         rib_delete_ipv4 (ZEBRA_ROUTE_KERNEL, zebra_flags, &p,
1.1.1.2 ! misho     909:                          NULL, 0, 0, SAFI_UNICAST);
1.1       misho     910:       
                    911:       if (rtm->rtm_type == RTM_GET 
                    912:           || rtm->rtm_type == RTM_ADD
                    913:           || rtm->rtm_type == RTM_CHANGE)
                    914:        rib_add_ipv4 (ZEBRA_ROUTE_KERNEL, zebra_flags, 
1.1.1.2 ! misho     915:                      &p, &gate.sin.sin_addr, NULL, 0, 0, 0, 0, SAFI_UNICAST);
1.1       misho     916:       else
                    917:        rib_delete_ipv4 (ZEBRA_ROUTE_KERNEL, zebra_flags, 
1.1.1.2 ! misho     918:                      &p, &gate.sin.sin_addr, 0, 0, SAFI_UNICAST);
1.1       misho     919:     }
                    920: #ifdef HAVE_IPV6
                    921:   if (dest.sa.sa_family == AF_INET6)
                    922:     {
                    923:       /* One day we might have a debug section here like one in the
                    924:        * IPv4 case above. Just ignore own messages at the moment.
                    925:        */
                    926:       if (rtm->rtm_type != RTM_GET && rtm->rtm_pid == pid)
                    927:         return;
                    928:       struct prefix_ipv6 p;
                    929:       unsigned int ifindex = 0;
                    930: 
                    931:       p.family = AF_INET6;
                    932:       p.prefix = dest.sin6.sin6_addr;
                    933:       if (flags & RTF_HOST)
                    934:        p.prefixlen = IPV6_MAX_PREFIXLEN;
                    935:       else
                    936:        p.prefixlen = ip6_masklen (mask.sin6.sin6_addr);
                    937: 
                    938: #ifdef KAME
                    939:       if (IN6_IS_ADDR_LINKLOCAL (&gate.sin6.sin6_addr))
                    940:        {
                    941:          ifindex = IN6_LINKLOCAL_IFINDEX (gate.sin6.sin6_addr);
                    942:          SET_IN6_LINKLOCAL_IFINDEX (gate.sin6.sin6_addr, 0);
                    943:        }
                    944: #endif /* KAME */
                    945: 
                    946:       /* CHANGE: delete the old prefix, we have no further information
                    947:        * to specify the route really
                    948:        */
                    949:       if (rtm->rtm_type == RTM_CHANGE)
                    950:         rib_delete_ipv6 (ZEBRA_ROUTE_KERNEL, zebra_flags, &p,
1.1.1.2 ! misho     951:                          NULL, 0, 0, SAFI_UNICAST);
1.1       misho     952:       
                    953:       if (rtm->rtm_type == RTM_GET 
                    954:           || rtm->rtm_type == RTM_ADD
                    955:           || rtm->rtm_type == RTM_CHANGE)
                    956:        rib_add_ipv6 (ZEBRA_ROUTE_KERNEL, zebra_flags,
1.1.1.2 ! misho     957:                      &p, &gate.sin6.sin6_addr, ifindex, 0, 0, 0, SAFI_UNICAST);
1.1       misho     958:       else
                    959:        rib_delete_ipv6 (ZEBRA_ROUTE_KERNEL, zebra_flags,
1.1.1.2 ! misho     960:                         &p, &gate.sin6.sin6_addr, ifindex, 0, SAFI_UNICAST);
1.1       misho     961:     }
                    962: #endif /* HAVE_IPV6 */
                    963: }
                    964: 
                    965: /* Interface function for the kernel routing table updates.  Support
                    966:  * for RTM_CHANGE will be needed.
                    967:  * Exported only for rt_socket.c
                    968:  */
                    969: int
                    970: rtm_write (int message,
                    971:           union sockunion *dest,
                    972:           union sockunion *mask,
                    973:           union sockunion *gate,
                    974:           unsigned int index,
                    975:           int zebra_flags,
                    976:           int metric)
                    977: {
                    978:   int ret;
                    979:   caddr_t pnt;
                    980:   struct interface *ifp;
                    981: 
                    982:   /* Sequencial number of routing message. */
                    983:   static int msg_seq = 0;
                    984: 
                    985:   /* Struct of rt_msghdr and buffer for storing socket's data. */
                    986:   struct 
                    987:   {
                    988:     struct rt_msghdr rtm;
                    989:     char buf[512];
                    990:   } msg;
                    991:   
                    992:   if (routing_sock < 0)
                    993:     return ZEBRA_ERR_EPERM;
                    994: 
                    995:   /* Clear and set rt_msghdr values */
                    996:   memset (&msg, 0, sizeof (struct rt_msghdr));
                    997:   msg.rtm.rtm_version = RTM_VERSION;
                    998:   msg.rtm.rtm_type = message;
                    999:   msg.rtm.rtm_seq = msg_seq++;
                   1000:   msg.rtm.rtm_addrs = RTA_DST;
                   1001:   msg.rtm.rtm_addrs |= RTA_GATEWAY;
                   1002:   msg.rtm.rtm_flags = RTF_UP;
                   1003:   msg.rtm.rtm_index = index;
                   1004: 
                   1005:   if (metric != 0)
                   1006:     {
                   1007:       msg.rtm.rtm_rmx.rmx_hopcount = metric;
                   1008:       msg.rtm.rtm_inits |= RTV_HOPCOUNT;
                   1009:     }
                   1010: 
                   1011:   ifp = if_lookup_by_index (index);
                   1012: 
                   1013:   if (gate && message == RTM_ADD)
                   1014:     msg.rtm.rtm_flags |= RTF_GATEWAY;
                   1015: 
                   1016:   /* When RTF_CLONING is unavailable on BSD, should we set some
                   1017:    * other flag instead?
                   1018:    */
                   1019: #ifdef RTF_CLONING
                   1020:   if (! gate && message == RTM_ADD && ifp &&
                   1021:       (ifp->flags & IFF_POINTOPOINT) == 0)
                   1022:     msg.rtm.rtm_flags |= RTF_CLONING;
                   1023: #endif /* RTF_CLONING */
                   1024: 
                   1025:   /* If no protocol specific gateway is specified, use link
                   1026:      address for gateway. */
                   1027:   if (! gate)
                   1028:     {
                   1029:       if (!ifp)
                   1030:         {
                   1031:           char dest_buf[INET_ADDRSTRLEN] = "NULL", mask_buf[INET_ADDRSTRLEN] = "255.255.255.255";
                   1032:           if (dest)
                   1033:             inet_ntop (AF_INET, &dest->sin.sin_addr, dest_buf, INET_ADDRSTRLEN);
                   1034:           if (mask)
                   1035:             inet_ntop (AF_INET, &mask->sin.sin_addr, mask_buf, INET_ADDRSTRLEN);
                   1036:           zlog_warn ("%s: %s/%s: gate == NULL and no gateway found for ifindex %d",
                   1037:             __func__, dest_buf, mask_buf, index);
                   1038:           return -1;
                   1039:         }
                   1040:       gate = (union sockunion *) & ifp->sdl;
                   1041:     }
                   1042: 
                   1043:   if (mask)
                   1044:     msg.rtm.rtm_addrs |= RTA_NETMASK;
                   1045:   else if (message == RTM_ADD) 
                   1046:     msg.rtm.rtm_flags |= RTF_HOST;
                   1047: 
                   1048:   /* Tagging route with flags */
                   1049:   msg.rtm.rtm_flags |= (RTF_PROTO1);
                   1050: 
                   1051:   /* Additional flags. */
                   1052:   if (zebra_flags & ZEBRA_FLAG_BLACKHOLE)
                   1053:     msg.rtm.rtm_flags |= RTF_BLACKHOLE;
                   1054:   if (zebra_flags & ZEBRA_FLAG_REJECT)
                   1055:     msg.rtm.rtm_flags |= RTF_REJECT;
                   1056: 
                   1057: 
                   1058: #ifdef HAVE_STRUCT_SOCKADDR_IN_SIN_LEN
                   1059: #define SOCKADDRSET(X,R) \
                   1060:   if (msg.rtm.rtm_addrs & (R)) \
                   1061:     { \
                   1062:       int len = ROUNDUP ((X)->sa.sa_len); \
                   1063:       memcpy (pnt, (caddr_t)(X), len); \
                   1064:       pnt += len; \
                   1065:     }
                   1066: #else 
                   1067: #define SOCKADDRSET(X,R) \
                   1068:   if (msg.rtm.rtm_addrs & (R)) \
                   1069:     { \
                   1070:       int len = SAROUNDUP (X); \
                   1071:       memcpy (pnt, (caddr_t)(X), len); \
                   1072:       pnt += len; \
                   1073:     }
                   1074: #endif /* HAVE_STRUCT_SOCKADDR_IN_SIN_LEN */
                   1075: 
                   1076:   pnt = (caddr_t) msg.buf;
                   1077: 
                   1078:   /* Write each socket data into rtm message buffer */
                   1079:   SOCKADDRSET (dest, RTA_DST);
                   1080:   SOCKADDRSET (gate, RTA_GATEWAY);
                   1081:   SOCKADDRSET (mask, RTA_NETMASK);
                   1082: 
                   1083:   msg.rtm.rtm_msglen = pnt - (caddr_t) &msg;
                   1084: 
                   1085:   ret = write (routing_sock, &msg, msg.rtm.rtm_msglen);
                   1086: 
                   1087:   if (ret != msg.rtm.rtm_msglen) 
                   1088:     {
                   1089:       if (errno == EEXIST) 
                   1090:        return ZEBRA_ERR_RTEXIST;
                   1091:       if (errno == ENETUNREACH)
                   1092:        return ZEBRA_ERR_RTUNREACH;
                   1093:       if (errno == ESRCH)
                   1094:        return ZEBRA_ERR_RTNOEXIST;
                   1095:       
                   1096:       zlog_warn ("%s: write : %s (%d)", __func__, safe_strerror (errno), errno);
                   1097:       return ZEBRA_ERR_KERNEL;
                   1098:     }
                   1099:   return ZEBRA_ERR_NOERROR;
                   1100: }
                   1101: 
                   1102: 
                   1103: #include "thread.h"
                   1104: #include "zebra/zserv.h"
                   1105: 
                   1106: /* For debug purpose. */
                   1107: static void
                   1108: rtmsg_debug (struct rt_msghdr *rtm)
                   1109: {
                   1110:   zlog_debug ("Kernel: Len: %d Type: %s", rtm->rtm_msglen, lookup (rtm_type_str, rtm->rtm_type));
                   1111:   rtm_flag_dump (rtm->rtm_flags);
                   1112:   zlog_debug ("Kernel: message seq %d", rtm->rtm_seq);
                   1113:   zlog_debug ("Kernel: pid %d, rtm_addrs 0x%x", rtm->rtm_pid, rtm->rtm_addrs);
                   1114: }
                   1115: 
                   1116: /* This is pretty gross, better suggestions welcome -- mhandler */
                   1117: #ifndef RTAX_MAX
                   1118: #ifdef RTA_NUMBITS
                   1119: #define RTAX_MAX       RTA_NUMBITS
                   1120: #else
                   1121: #define RTAX_MAX       8
                   1122: #endif /* RTA_NUMBITS */
                   1123: #endif /* RTAX_MAX */
                   1124: 
                   1125: /* Kernel routing table and interface updates via routing socket. */
                   1126: static int
                   1127: kernel_read (struct thread *thread)
                   1128: {
                   1129:   int sock;
                   1130:   int nbytes;
                   1131:   struct rt_msghdr *rtm;
                   1132: 
                   1133:   /*
                   1134:    * This must be big enough for any message the kernel might send.
                   1135:    * Rather than determining how many sockaddrs of what size might be
                   1136:    * in each particular message, just use RTAX_MAX of sockaddr_storage
                   1137:    * for each.  Note that the sockaddrs must be after each message
                   1138:    * definition, or rather after whichever happens to be the largest,
                   1139:    * since the buffer needs to be big enough for a message and the
                   1140:    * sockaddrs together.
                   1141:    */
                   1142:   union 
                   1143:   {
                   1144:     /* Routing information. */
                   1145:     struct 
                   1146:     {
                   1147:       struct rt_msghdr rtm;
                   1148:       struct sockaddr_storage addr[RTAX_MAX];
                   1149:     } r;
                   1150: 
                   1151:     /* Interface information. */
                   1152:     struct
                   1153:     {
                   1154:       struct if_msghdr ifm;
                   1155:       struct sockaddr_storage addr[RTAX_MAX];
                   1156:     } im;
                   1157: 
                   1158:     /* Interface address information. */
                   1159:     struct
                   1160:     {
                   1161:       struct ifa_msghdr ifa;
                   1162:       struct sockaddr_storage addr[RTAX_MAX];
                   1163:     } ia;
                   1164: 
                   1165: #ifdef RTM_IFANNOUNCE
                   1166:     /* Interface arrival/departure */
                   1167:     struct
                   1168:     {
                   1169:       struct if_announcemsghdr ifan;
                   1170:       struct sockaddr_storage addr[RTAX_MAX];
                   1171:     } ian;
                   1172: #endif /* RTM_IFANNOUNCE */
                   1173: 
                   1174:   } buf;
                   1175: 
                   1176:   /* Fetch routing socket. */
                   1177:   sock = THREAD_FD (thread);
                   1178: 
                   1179:   nbytes= read (sock, &buf, sizeof buf);
                   1180: 
                   1181:   if (nbytes <= 0)
                   1182:     {
                   1183:       if (nbytes < 0 && errno != EWOULDBLOCK && errno != EAGAIN)
                   1184:        zlog_warn ("routing socket error: %s", safe_strerror (errno));
                   1185:       return 0;
                   1186:     }
                   1187: 
                   1188:   thread_add_read (zebrad.master, kernel_read, NULL, sock);
                   1189: 
                   1190:   if (IS_ZEBRA_DEBUG_KERNEL)
                   1191:     rtmsg_debug (&buf.r.rtm);
                   1192: 
                   1193:   rtm = &buf.r.rtm;
                   1194: 
                   1195:   /*
                   1196:    * Ensure that we didn't drop any data, so that processing routines
                   1197:    * can assume they have the whole message.
                   1198:    */
                   1199:   if (rtm->rtm_msglen != nbytes)
                   1200:     {
                   1201:       zlog_warn ("kernel_read: rtm->rtm_msglen %d, nbytes %d, type %d\n",
                   1202:                 rtm->rtm_msglen, nbytes, rtm->rtm_type);
                   1203:       return -1;
                   1204:     }
                   1205: 
                   1206:   switch (rtm->rtm_type)
                   1207:     {
                   1208:     case RTM_ADD:
                   1209:     case RTM_DELETE:
                   1210:     case RTM_CHANGE:
                   1211:       rtm_read (rtm);
                   1212:       break;
                   1213:     case RTM_IFINFO:
                   1214:       ifm_read (&buf.im.ifm);
                   1215:       break;
                   1216:     case RTM_NEWADDR:
                   1217:     case RTM_DELADDR:
                   1218:       ifam_read (&buf.ia.ifa);
                   1219:       break;
                   1220: #ifdef RTM_IFANNOUNCE
                   1221:     case RTM_IFANNOUNCE:
                   1222:       ifan_read (&buf.ian.ifan);
                   1223:       break;
                   1224: #endif /* RTM_IFANNOUNCE */
                   1225:     default:
                   1226:       if (IS_ZEBRA_DEBUG_KERNEL)
                   1227:         zlog_debug("Unprocessed RTM_type: %d", rtm->rtm_type);
                   1228:       break;
                   1229:     }
                   1230:   return 0;
                   1231: }
                   1232: 
                   1233: /* Make routing socket. */
                   1234: static void
                   1235: routing_socket (void)
                   1236: {
                   1237:   if ( zserv_privs.change (ZPRIVS_RAISE) )
                   1238:     zlog_err ("routing_socket: Can't raise privileges");
                   1239: 
                   1240:   routing_sock = socket (AF_ROUTE, SOCK_RAW, 0);
                   1241: 
                   1242:   if (routing_sock < 0) 
                   1243:     {
                   1244:       if ( zserv_privs.change (ZPRIVS_LOWER) )
                   1245:         zlog_err ("routing_socket: Can't lower privileges");
                   1246:       zlog_warn ("Can't init kernel routing socket");
                   1247:       return;
                   1248:     }
                   1249: 
                   1250:   /* XXX: Socket should be NONBLOCK, however as we currently 
                   1251:    * discard failed writes, this will lead to inconsistencies.
                   1252:    * For now, socket must be blocking.
                   1253:    */
                   1254:   /*if (fcntl (routing_sock, F_SETFL, O_NONBLOCK) < 0) 
                   1255:     zlog_warn ("Can't set O_NONBLOCK to routing socket");*/
                   1256:     
                   1257:   if ( zserv_privs.change (ZPRIVS_LOWER) )
                   1258:     zlog_err ("routing_socket: Can't lower privileges");
                   1259: 
                   1260:   /* kernel_read needs rewrite. */
                   1261:   thread_add_read (zebrad.master, kernel_read, NULL, routing_sock);
                   1262: }
                   1263: 
                   1264: /* Exported interface function.  This function simply calls
                   1265:    routing_socket (). */
                   1266: void
                   1267: kernel_init (void)
                   1268: {
                   1269:   routing_socket ();
                   1270: }

FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>