1 /* Kernel communication using routing socket.
2 * Copyright (C) 1999 Kunihiro Ishiguro
4 * This file is part of GNU Zebra.
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
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.
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
26 #include "sockunion.h"
27 #include "connected.h"
36 #include "zebra/interface.h"
37 #include "zebra/zserv.h"
38 #include "zebra/debug.h"
39 #include "zebra/kernel_socket.h"
41 extern struct zebra_privs_t zserv_privs
;
42 extern struct zebra_t zebrad
;
45 * Given a sockaddr length, round it up to include pad bytes following
46 * it. Assumes the kernel pads to sizeof(long).
48 * XXX: why is ROUNDUP(0) sizeof(long)? 0 is an illegal sockaddr
49 * length anyway (< sizeof (struct sockaddr)), so this shouldn't
51 * On OS X, both 32, 64bit syatems align on 4 byte boundary
55 ((a) > 0 ? (1 + (((a) - 1) | (sizeof(int) - 1))) : sizeof(int))
58 ((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
62 * Given a pointer (sockaddr or void *), return the number of bytes
63 * taken up by the sockaddr and any padding needed for alignment.
65 #if defined(HAVE_STRUCT_SOCKADDR_SA_LEN)
66 #define SAROUNDUP(X) ROUNDUP(((struct sockaddr *)(X))->sa_len)
67 #elif defined(HAVE_IPV6)
69 * One would hope all fixed-size structure definitions are aligned,
70 * but round them up nonetheless.
72 #define SAROUNDUP(X) \
73 (((struct sockaddr *)(X))->sa_family == AF_INET ? \
74 ROUNDUP(sizeof(struct sockaddr_in)):\
75 (((struct sockaddr *)(X))->sa_family == AF_INET6 ? \
76 ROUNDUP(sizeof(struct sockaddr_in6)) : \
77 (((struct sockaddr *)(X))->sa_family == AF_LINK ? \
78 ROUNDUP(sizeof(struct sockaddr_dl)) : sizeof(struct sockaddr))))
80 #define SAROUNDUP(X) \
81 (((struct sockaddr *)(X))->sa_family == AF_INET ? \
82 ROUNDUP(sizeof(struct sockaddr_in)):\
83 (((struct sockaddr *)(X))->sa_family == AF_LINK ? \
84 ROUNDUP(sizeof(struct sockaddr_dl)) : sizeof(struct sockaddr)))
85 #endif /* HAVE_STRUCT_SOCKADDR_SA_LEN */
88 * We use a call to an inline function to copy (PNT) to (DEST)
89 * 1. Calculating the length of the copy requires an #ifdef to determine
90 * if sa_len is a field and can't be used directly inside a #define
91 * 2. So the compiler doesn't complain when DEST is NULL, which is only true
92 * when we are skipping the copy and incrementing to the next SA
95 rta_copy (union sockunion
*dest
, caddr_t src
) {
97 #ifdef HAVE_STRUCT_SOCKADDR_SA_LEN
98 len
= (((struct sockaddr
*)src
)->sa_len
> sizeof (*dest
)) ?
99 sizeof (*dest
) : ((struct sockaddr
*)src
)->sa_len
;
101 len
= (SAROUNDUP (src
) > sizeof (*dest
)) ?
102 sizeof (*dest
) : SAROUNDUP (src
) ;
104 memcpy (dest
, src
, len
);
107 #define RTA_ADDR_GET(DEST, RTA, RTMADDRS, PNT) \
108 if ((RTMADDRS) & (RTA)) \
110 int len = SAROUNDUP ((PNT)); \
111 if ( ((DEST) != NULL) && \
112 af_check (((struct sockaddr *)(PNT))->sa_family)) \
113 rta_copy((DEST), (PNT)); \
116 #define RTA_ATTR_GET(DEST, RTA, RTMADDRS, PNT) \
117 if ((RTMADDRS) & (RTA)) \
119 int len = SAROUNDUP ((PNT)); \
120 if ((DEST) != NULL) \
121 rta_copy((DEST), (PNT)); \
125 #define RTA_NAME_GET(DEST, RTA, RTMADDRS, PNT, LEN) \
126 if ((RTMADDRS) & (RTA)) \
128 u_char *pdest = (u_char *) (DEST); \
129 int len = SAROUNDUP ((PNT)); \
130 struct sockaddr_dl *sdl = (struct sockaddr_dl *)(PNT); \
131 if (IS_ZEBRA_DEBUG_KERNEL) \
132 zlog_debug ("%s: RTA_SDL_GET nlen %d, alen %d", \
133 __func__, sdl->sdl_nlen, sdl->sdl_alen); \
134 if ( ((DEST) != NULL) && (sdl->sdl_family == AF_LINK) \
135 && (sdl->sdl_nlen < IFNAMSIZ) && (sdl->sdl_nlen <= len) ) \
137 memcpy (pdest, sdl->sdl_data, sdl->sdl_nlen); \
138 pdest[sdl->sdl_nlen] = '\0'; \
139 (LEN) = sdl->sdl_nlen; \
147 /* Routing socket message types. */
148 const struct message rtm_type_str
[] =
150 {RTM_ADD
, "RTM_ADD"},
151 {RTM_DELETE
, "RTM_DELETE"},
152 {RTM_CHANGE
, "RTM_CHANGE"},
153 {RTM_GET
, "RTM_GET"},
154 {RTM_LOSING
, "RTM_LOSING"},
155 {RTM_REDIRECT
, "RTM_REDIRECT"},
156 {RTM_MISS
, "RTM_MISS"},
157 {RTM_LOCK
, "RTM_LOCK"},
159 {RTM_OLDADD
, "RTM_OLDADD"},
160 #endif /* RTM_OLDADD */
162 {RTM_OLDDEL
, "RTM_OLDDEL"},
163 #endif /* RTM_OLDDEL */
164 {RTM_RESOLVE
, "RTM_RESOLVE"},
165 {RTM_NEWADDR
, "RTM_NEWADDR"},
166 {RTM_DELADDR
, "RTM_DELADDR"},
167 {RTM_IFINFO
, "RTM_IFINFO"},
169 {RTM_OIFINFO
, "RTM_OIFINFO"},
170 #endif /* RTM_OIFINFO */
172 {RTM_NEWMADDR
, "RTM_NEWMADDR"},
173 #endif /* RTM_NEWMADDR */
175 {RTM_DELMADDR
, "RTM_DELMADDR"},
176 #endif /* RTM_DELMADDR */
177 #ifdef RTM_IFANNOUNCE
178 {RTM_IFANNOUNCE
, "RTM_IFANNOUNCE"},
179 #endif /* RTM_IFANNOUNCE */
183 static const struct message rtm_flag_str
[] =
186 {RTF_GATEWAY
, "GATEWAY"},
188 {RTF_REJECT
, "REJECT"},
189 {RTF_DYNAMIC
, "DYNAMIC"},
190 {RTF_MODIFIED
, "MODIFIED"},
194 #endif /* RTF_MASK */
196 {RTF_CLONING
, "CLONING"},
197 #endif /* RTF_CLONING */
198 {RTF_XRESOLVE
, "XRESOLVE"},
199 {RTF_LLINFO
, "LLINFO"},
200 {RTF_STATIC
, "STATIC"},
201 {RTF_BLACKHOLE
, "BLACKHOLE"},
203 {RTF_PRIVATE
, "PRIVATE"},
204 #endif /* RTF_PRIVATE */
205 {RTF_PROTO1
, "PROTO1"},
206 {RTF_PROTO2
, "PROTO2"},
208 {RTF_PRCLONING
, "PRCLONING"},
209 #endif /* RTF_PRCLONING */
211 {RTF_WASCLONED
, "WASCLONED"},
212 #endif /* RTF_WASCLONED */
214 {RTF_PROTO3
, "PROTO3"},
215 #endif /* RTF_PROTO3 */
217 {RTF_PINNED
, "PINNED"},
218 #endif /* RTF_PINNED */
220 {RTF_LOCAL
, "LOCAL"},
221 #endif /* RTF_LOCAL */
223 {RTF_BROADCAST
, "BROADCAST"},
224 #endif /* RTF_BROADCAST */
226 {RTF_MULTICAST
, "MULTICAST"},
227 #endif /* RTF_MULTICAST */
229 {RTF_MULTIRT
, "MULTIRT"},
230 #endif /* RTF_MULTIRT */
232 {RTF_SETSRC
, "SETSRC"},
233 #endif /* RTF_SETSRC */
237 /* Kernel routing update socket. */
238 int routing_sock
= -1;
240 /* Yes I'm checking ugly routing socket behavior. */
243 /* Supported address family check. */
245 af_check (int family
)
247 if (family
== AF_INET
)
250 if (family
== AF_INET6
)
252 #endif /* HAVE_IPV6 */
256 /* Dump routing table flag for debug purpose. */
258 rtm_flag_dump (int flag
)
260 const struct message
*mes
;
261 static char buf
[BUFSIZ
];
264 for (mes
= rtm_flag_str
; mes
->key
!= 0; mes
++)
268 strlcat (buf
, mes
->str
, BUFSIZ
);
269 strlcat (buf
, " ", BUFSIZ
);
272 zlog_debug ("Kernel: %s", buf
);
275 #ifdef RTM_IFANNOUNCE
276 /* Interface adding function */
278 ifan_read (struct if_announcemsghdr
*ifan
)
280 struct interface
*ifp
;
282 ifp
= if_lookup_by_index (ifan
->ifan_index
);
285 assert ( (ifp
->ifindex
== ifan
->ifan_index
)
286 || (ifp
->ifindex
== IFINDEX_INTERNAL
) );
289 || ((ifp
->ifindex
== IFINDEX_INTERNAL
)
290 && (ifan
->ifan_what
== IFAN_ARRIVAL
)) )
292 if (IS_ZEBRA_DEBUG_KERNEL
)
293 zlog_debug ("%s: creating interface for ifindex %d, name %s",
294 __func__
, ifan
->ifan_index
, ifan
->ifan_name
);
296 /* Create Interface */
297 ifp
= if_get_by_name_len(ifan
->ifan_name
,
298 strnlen(ifan
->ifan_name
,
299 sizeof(ifan
->ifan_name
)));
300 ifp
->ifindex
= ifan
->ifan_index
;
305 else if (ifp
!= NULL
&& ifan
->ifan_what
== IFAN_DEPARTURE
)
306 if_delete_update (ifp
);
312 if (IS_ZEBRA_DEBUG_KERNEL
)
313 zlog_debug ("%s: interface %s index %d",
314 __func__
, ifan
->ifan_name
, ifan
->ifan_index
);
318 #endif /* RTM_IFANNOUNCE */
320 #ifdef HAVE_BSD_IFI_LINK_STATE
321 /* BSD link detect translation */
323 bsd_linkdetect_translate (struct if_msghdr
*ifm
)
325 if ((ifm
->ifm_data
.ifi_link_state
>= LINK_STATE_UP
) ||
326 (ifm
->ifm_data
.ifi_link_state
== LINK_STATE_UNKNOWN
))
327 SET_FLAG(ifm
->ifm_flags
, IFF_RUNNING
);
329 UNSET_FLAG(ifm
->ifm_flags
, IFF_RUNNING
);
331 #endif /* HAVE_BSD_IFI_LINK_STATE */
334 * Handle struct if_msghdr obtained from reading routing socket or
335 * sysctl (from interface_list). There may or may not be sockaddrs
336 * present after the header.
339 ifm_read (struct if_msghdr
*ifm
)
341 struct interface
*ifp
= NULL
;
342 struct sockaddr_dl
*sdl
;
343 char ifname
[IFNAMSIZ
];
347 /* terminate ifname at head (for strnlen) and tail (for safety) */
348 ifname
[IFNAMSIZ
- 1] = '\0';
350 /* paranoia: sanity check structure */
351 if (ifm
->ifm_msglen
< sizeof(struct if_msghdr
))
353 zlog_err ("ifm_read: ifm->ifm_msglen %d too short\n",
359 * Check for a sockaddr_dl following the message. First, point to
360 * where a socakddr might be if one follows the message.
362 cp
= (void *)(ifm
+ 1);
366 * XXX This behavior should be narrowed to only the kernel versions
367 * for which the structures returned do not match the headers.
369 * if_msghdr_t on 64 bit kernels in Solaris 9 and earlier versions
370 * is 12 bytes larger than the 32 bit version.
372 if (((struct sockaddr
*) cp
)->sa_family
== AF_UNSPEC
)
376 RTA_ADDR_GET (NULL
, RTA_DST
, ifm
->ifm_addrs
, cp
);
377 RTA_ADDR_GET (NULL
, RTA_GATEWAY
, ifm
->ifm_addrs
, cp
);
378 RTA_ATTR_GET (NULL
, RTA_NETMASK
, ifm
->ifm_addrs
, cp
);
379 RTA_ADDR_GET (NULL
, RTA_GENMASK
, ifm
->ifm_addrs
, cp
);
380 sdl
= (struct sockaddr_dl
*)cp
;
381 RTA_NAME_GET (ifname
, RTA_IFP
, ifm
->ifm_addrs
, cp
, ifnlen
);
382 RTA_ADDR_GET (NULL
, RTA_IFA
, ifm
->ifm_addrs
, cp
);
383 RTA_ADDR_GET (NULL
, RTA_AUTHOR
, ifm
->ifm_addrs
, cp
);
384 RTA_ADDR_GET (NULL
, RTA_BRD
, ifm
->ifm_addrs
, cp
);
386 if (IS_ZEBRA_DEBUG_KERNEL
)
387 zlog_debug ("%s: sdl ifname %s", __func__
, (ifnlen
? ifname
: "(nil)"));
390 * Look up on ifindex first, because ifindices are the primary handle for
391 * interfaces across the user/kernel boundary, for most systems. (Some
392 * messages, such as up/down status changes on NetBSD, do not include a
395 if ( (ifp
= if_lookup_by_index (ifm
->ifm_index
)) != NULL
)
397 /* we have an ifp, verify that the name matches as some systems,
398 * eg Solaris, have a 1:many association of ifindex:ifname
399 * if they dont match, we dont have the correct ifp and should
400 * set it back to NULL to let next check do lookup by name
402 if (ifnlen
&& (strncmp (ifp
->name
, ifname
, IFNAMSIZ
) != 0) )
404 if (IS_ZEBRA_DEBUG_KERNEL
)
405 zlog_debug ("%s: ifp name %s doesnt match sdl name %s",
406 __func__
, ifp
->name
, ifname
);
412 * If we dont have an ifp, try looking up by name. Particularly as some
413 * systems (Solaris) have a 1:many mapping of ifindex:ifname - the ifname
414 * is therefore our unique handle to that interface.
416 * Interfaces specified in the configuration file for which the ifindex
417 * has not been determined will have ifindex == IFINDEX_INTERNAL, and such
418 * interfaces are found by this search, and then their ifindex values can
421 if ( (ifp
== NULL
) && ifnlen
)
422 ifp
= if_lookup_by_name (ifname
);
425 * If ifp still does not exist or has an invalid index (IFINDEX_INTERNAL),
426 * create or fill in an interface.
428 if ((ifp
== NULL
) || (ifp
->ifindex
== IFINDEX_INTERNAL
))
431 * To create or fill in an interface, a sockaddr_dl (via
432 * RTA_IFP) is required.
436 zlog_warn ("Interface index %d (new) missing ifname\n",
441 #ifndef RTM_IFANNOUNCE
442 /* Down->Down interface should be ignored here.
443 * See further comment below.
445 if (!CHECK_FLAG (ifm
->ifm_flags
, IFF_UP
))
447 #endif /* !RTM_IFANNOUNCE */
451 /* Interface that zebra was not previously aware of, so create. */
452 ifp
= if_create (ifname
, ifnlen
);
453 if (IS_ZEBRA_DEBUG_KERNEL
)
454 zlog_debug ("%s: creating ifp for ifindex %d",
455 __func__
, ifm
->ifm_index
);
458 if (IS_ZEBRA_DEBUG_KERNEL
)
459 zlog_debug ("%s: updated/created ifp, ifname %s, ifindex %d",
460 __func__
, ifp
->name
, ifp
->ifindex
);
462 * Fill in newly created interface structure, or larval
463 * structure with ifindex IFINDEX_INTERNAL.
465 ifp
->ifindex
= ifm
->ifm_index
;
467 #ifdef HAVE_BSD_IFI_LINK_STATE /* translate BSD kernel msg for link-state */
468 bsd_linkdetect_translate(ifm
);
469 #endif /* HAVE_BSD_IFI_LINK_STATE */
471 if_flags_update (ifp
, ifm
->ifm_flags
);
472 #if defined(__bsdi__)
473 if_kvm_get_mtu (ifp
);
476 #endif /* __bsdi__ */
480 * XXX sockaddr_dl contents can be larger than the structure
481 * definition. There are 2 big families here:
482 * - BSD has sdl_len + sdl_data[16] + overruns sdl_data
483 * we MUST use sdl_len here or we'll truncate data.
484 * - Solaris has no sdl_len, but sdl_data[244]
485 * presumably, it's not going to run past that, so sizeof()
487 * a nonzero ifnlen from RTA_NAME_GET() means sdl is valid
491 #ifdef HAVE_STRUCT_SOCKADDR_DL_SDL_LEN
492 memcpy (&ifp
->sdl
, sdl
, sdl
->sdl_len
);
494 memcpy (&ifp
->sdl
, sdl
, sizeof (struct sockaddr_dl
));
495 #endif /* HAVE_STRUCT_SOCKADDR_DL_SDL_LEN */
502 * Interface structure exists. Adjust stored flags from
503 * notification. If interface has up->down or down->up
504 * transition, call state change routines (to adjust routes,
505 * notify routing daemons, etc.). (Other flag changes are stored
506 * but apparently do not trigger action.)
509 if (ifp
->ifindex
!= ifm
->ifm_index
)
511 zlog_warn ("%s: index mismatch, ifname %s, ifp index %d, "
513 __func__
, ifp
->name
, ifp
->ifindex
, ifm
->ifm_index
);
517 #ifdef HAVE_BSD_IFI_LINK_STATE /* translate BSD kernel msg for link-state */
518 bsd_linkdetect_translate(ifm
);
519 #endif /* HAVE_BSD_IFI_LINK_STATE */
521 /* update flags and handle operative->inoperative transition, if any */
522 if_flags_update (ifp
, ifm
->ifm_flags
);
524 #ifndef RTM_IFANNOUNCE
527 /* No RTM_IFANNOUNCE on this platform, so we can never
528 * distinguish between ~IFF_UP and delete. We must presume
529 * it has been deleted.
530 * Eg, Solaris will not notify us of unplumb.
532 * XXX: Fixme - this should be runtime detected
533 * So that a binary compiled on a system with IFANNOUNCE
534 * will still behave correctly if run on a platform without
536 if_delete_update (ifp
);
538 #endif /* RTM_IFANNOUNCE */
541 #if defined(__bsdi__)
542 if_kvm_get_mtu (ifp
);
545 #endif /* __bsdi__ */
550 #ifdef HAVE_NET_RT_IFLIST
551 ifp
->stats
= ifm
->ifm_data
;
552 #endif /* HAVE_NET_RT_IFLIST */
554 if (IS_ZEBRA_DEBUG_KERNEL
)
555 zlog_debug ("%s: interface %s index %d",
556 __func__
, ifp
->name
, ifp
->ifindex
);
561 /* Address read from struct ifa_msghdr. */
563 ifam_read_mesg (struct ifa_msghdr
*ifm
,
564 union sockunion
*addr
,
565 union sockunion
*mask
,
566 union sockunion
*brd
,
572 union sockunion gateway
;
574 pnt
= (caddr_t
)(ifm
+ 1);
575 end
= ((caddr_t
)ifm
) + ifm
->ifam_msglen
;
577 /* Be sure structure is cleared */
578 memset (mask
, 0, sizeof (union sockunion
));
579 memset (addr
, 0, sizeof (union sockunion
));
580 memset (brd
, 0, sizeof (union sockunion
));
581 memset (&dst
, 0, sizeof (union sockunion
));
582 memset (&gateway
, 0, sizeof (union sockunion
));
584 /* We fetch each socket variable into sockunion. */
585 RTA_ADDR_GET (&dst
, RTA_DST
, ifm
->ifam_addrs
, pnt
);
586 RTA_ADDR_GET (&gateway
, RTA_GATEWAY
, ifm
->ifam_addrs
, pnt
);
587 RTA_ATTR_GET (mask
, RTA_NETMASK
, ifm
->ifam_addrs
, pnt
);
588 RTA_ADDR_GET (NULL
, RTA_GENMASK
, ifm
->ifam_addrs
, pnt
);
589 RTA_NAME_GET (ifname
, RTA_IFP
, ifm
->ifam_addrs
, pnt
, *ifnlen
);
590 RTA_ADDR_GET (addr
, RTA_IFA
, ifm
->ifam_addrs
, pnt
);
591 RTA_ADDR_GET (NULL
, RTA_AUTHOR
, ifm
->ifam_addrs
, pnt
);
592 RTA_ADDR_GET (brd
, RTA_BRD
, ifm
->ifam_addrs
, pnt
);
594 if (IS_ZEBRA_DEBUG_KERNEL
)
596 switch (sockunion_family(addr
))
600 char buf
[4][INET_ADDRSTRLEN
];
601 zlog_debug ("%s: ifindex %d, ifname %s, ifam_addrs 0x%x, "
602 "ifam_flags 0x%x, addr %s/%d broad %s dst %s "
604 __func__
, ifm
->ifam_index
,
605 (ifnlen
? ifname
: "(nil)"), ifm
->ifam_addrs
,
607 inet_ntop(AF_INET
,&addr
->sin
.sin_addr
,
608 buf
[0],sizeof(buf
[0])),
609 ip_masklen(mask
->sin
.sin_addr
),
610 inet_ntop(AF_INET
,&brd
->sin
.sin_addr
,
611 buf
[1],sizeof(buf
[1])),
612 inet_ntop(AF_INET
,&dst
.sin
.sin_addr
,
613 buf
[2],sizeof(buf
[2])),
614 inet_ntop(AF_INET
,&gateway
.sin
.sin_addr
,
615 buf
[3],sizeof(buf
[3])));
621 char buf
[4][INET6_ADDRSTRLEN
];
622 zlog_debug ("%s: ifindex %d, ifname %s, ifam_addrs 0x%x, "
623 "ifam_flags 0x%x, addr %s/%d broad %s dst %s "
625 __func__
, ifm
->ifam_index
,
626 (ifnlen
? ifname
: "(nil)"), ifm
->ifam_addrs
,
628 inet_ntop(AF_INET6
,&addr
->sin6
.sin6_addr
,
629 buf
[0],sizeof(buf
[0])),
630 ip6_masklen(mask
->sin6
.sin6_addr
),
631 inet_ntop(AF_INET6
,&brd
->sin6
.sin6_addr
,
632 buf
[1],sizeof(buf
[1])),
633 inet_ntop(AF_INET6
,&dst
.sin6
.sin6_addr
,
634 buf
[2],sizeof(buf
[2])),
635 inet_ntop(AF_INET6
,&gateway
.sin6
.sin6_addr
,
636 buf
[3],sizeof(buf
[3])));
639 #endif /* HAVE_IPV6 */
641 zlog_debug ("%s: ifindex %d, ifname %s, ifam_addrs 0x%x",
642 __func__
, ifm
->ifam_index
,
643 (ifnlen
? ifname
: "(nil)"), ifm
->ifam_addrs
);
648 /* Assert read up end point matches to end point */
650 zlog_warn ("ifam_read() doesn't read all socket data");
653 /* Interface's address information get. */
655 ifam_read (struct ifa_msghdr
*ifam
)
657 struct interface
*ifp
= NULL
;
658 union sockunion addr
, mask
, brd
;
659 char ifname
[INTERFACE_NAMSIZ
];
664 ifname
[0] = ifname
[INTERFACE_NAMSIZ
- 1] = '\0';
666 /* Allocate and read address information. */
667 ifam_read_mesg (ifam
, &addr
, &mask
, &brd
, ifname
, &ifnlen
);
669 if ((ifp
= if_lookup_by_index(ifam
->ifam_index
)) == NULL
)
671 zlog_warn ("%s: no interface for ifname %s, index %d",
672 __func__
, ifname
, ifam
->ifam_index
);
676 if (ifnlen
&& strncmp (ifp
->name
, ifname
, INTERFACE_NAMSIZ
))
679 /* N.B. The info in ifa_msghdr does not tell us whether the RTA_BRD
680 field contains a broadcast address or a peer address, so we are forced to
681 rely upon the interface type. */
682 if (if_is_pointopoint(ifp
))
683 SET_FLAG(flags
, ZEBRA_IFA_PEER
);
686 /* it might seem cute to grab the interface metric here, however
687 * we're processing an address update message, and so some systems
688 * (e.g. FBSD) dont bother to fill in ifam_metric. Disabled, but left
689 * in deliberately, as comment.
691 ifp
->metric
= ifam
->ifam_metric
;
694 /* Add connected address. */
695 switch (sockunion_family (&addr
))
698 if (ifam
->ifam_type
== RTM_NEWADDR
)
699 connected_add_ipv4 (ifp
, flags
, &addr
.sin
.sin_addr
,
700 ip_masklen (mask
.sin
.sin_addr
),
702 (isalias
? ifname
: NULL
));
704 connected_delete_ipv4 (ifp
, flags
, &addr
.sin
.sin_addr
,
705 ip_masklen (mask
.sin
.sin_addr
),
710 /* Unset interface index from link-local address when IPv6 stack
712 if (IN6_IS_ADDR_LINKLOCAL (&addr
.sin6
.sin6_addr
))
713 SET_IN6_LINKLOCAL_IFINDEX (addr
.sin6
.sin6_addr
, 0);
715 if (ifam
->ifam_type
== RTM_NEWADDR
)
716 connected_add_ipv6 (ifp
, flags
, &addr
.sin6
.sin6_addr
,
717 ip6_masklen (mask
.sin6
.sin6_addr
),
719 (isalias
? ifname
: NULL
));
721 connected_delete_ipv6 (ifp
,
722 &addr
.sin6
.sin6_addr
,
723 ip6_masklen (mask
.sin6
.sin6_addr
),
724 &brd
.sin6
.sin6_addr
);
726 #endif /* HAVE_IPV6 */
728 /* Unsupported family silently ignore... */
732 /* Check interface flag for implicit up of the interface. */
736 /* In addition to lacking IFANNOUNCE, on SUNOS IFF_UP is strange.
737 * See comments for SUNOS_5 in interface.c::if_flags_mangle.
739 * Here we take care of case where the real IFF_UP was previously
740 * unset (as kept in struct zebra_if.primary_state) and the mangled
741 * IFF_UP (ie IFF_UP set || listcount(connected) has now transitioned
742 * to unset due to the lost non-primary address having DELADDR'd.
744 * we must delete the interface, because in between here and next
745 * event for this interface-name the administrator could unplumb
746 * and replumb the interface.
749 if_delete_update (ifp
);
755 /* Interface function for reading kernel routing table information. */
757 rtm_read_mesg (struct rt_msghdr
*rtm
,
758 union sockunion
*dest
,
759 union sockunion
*mask
,
760 union sockunion
*gate
,
766 /* Pnt points out socket data start point. */
767 pnt
= (caddr_t
)(rtm
+ 1);
768 end
= ((caddr_t
)rtm
) + rtm
->rtm_msglen
;
770 /* rt_msghdr version check. */
771 if (rtm
->rtm_version
!= RTM_VERSION
)
772 zlog (NULL
, LOG_WARNING
,
773 "Routing message version different %d should be %d."
774 "This may cause problem\n", rtm
->rtm_version
, RTM_VERSION
);
776 /* Be sure structure is cleared */
777 memset (dest
, 0, sizeof (union sockunion
));
778 memset (gate
, 0, sizeof (union sockunion
));
779 memset (mask
, 0, sizeof (union sockunion
));
781 /* We fetch each socket variable into sockunion. */
782 RTA_ADDR_GET (dest
, RTA_DST
, rtm
->rtm_addrs
, pnt
);
783 RTA_ADDR_GET (gate
, RTA_GATEWAY
, rtm
->rtm_addrs
, pnt
);
784 RTA_ATTR_GET (mask
, RTA_NETMASK
, rtm
->rtm_addrs
, pnt
);
785 RTA_ADDR_GET (NULL
, RTA_GENMASK
, rtm
->rtm_addrs
, pnt
);
786 RTA_NAME_GET (ifname
, RTA_IFP
, rtm
->rtm_addrs
, pnt
, *ifnlen
);
787 RTA_ADDR_GET (NULL
, RTA_IFA
, rtm
->rtm_addrs
, pnt
);
788 RTA_ADDR_GET (NULL
, RTA_AUTHOR
, rtm
->rtm_addrs
, pnt
);
789 RTA_ADDR_GET (NULL
, RTA_BRD
, rtm
->rtm_addrs
, pnt
);
791 /* If there is netmask information set it's family same as
793 if (rtm
->rtm_addrs
& RTA_NETMASK
)
794 mask
->sa
.sa_family
= dest
->sa
.sa_family
;
796 /* Assert read up to the end of pointer. */
798 zlog (NULL
, LOG_WARNING
, "rtm_read() doesn't read all socket data.");
800 return rtm
->rtm_flags
;
804 rtm_read (struct rt_msghdr
*rtm
)
808 union sockunion dest
, mask
, gate
;
809 char ifname
[INTERFACE_NAMSIZ
+ 1];
814 /* Read destination and netmask and gateway from rtm message
816 flags
= rtm_read_mesg (rtm
, &dest
, &mask
, &gate
, ifname
, &ifnlen
);
817 if (!(flags
& RTF_DONE
))
819 if (IS_ZEBRA_DEBUG_KERNEL
)
820 zlog_debug ("%s: got rtm of type %d (%s)", __func__
, rtm
->rtm_type
,
821 lookup (rtm_type_str
, rtm
->rtm_type
));
823 #ifdef RTF_CLONED /*bsdi, netbsd 1.6*/
824 if (flags
& RTF_CLONED
)
827 #ifdef RTF_WASCLONED /*freebsd*/
828 if (flags
& RTF_WASCLONED
)
832 if ((rtm
->rtm_type
== RTM_ADD
) && ! (flags
& RTF_UP
))
835 /* This is connected route. */
836 if (! (flags
& RTF_GATEWAY
))
839 if (flags
& RTF_PROTO1
)
840 SET_FLAG (zebra_flags
, ZEBRA_FLAG_SELFROUTE
);
842 /* This is persistent route. */
843 if (flags
& RTF_STATIC
)
844 SET_FLAG (zebra_flags
, ZEBRA_FLAG_STATIC
);
846 /* This is a reject or blackhole route */
847 if (flags
& RTF_REJECT
)
848 SET_FLAG (zebra_flags
, ZEBRA_FLAG_REJECT
);
849 if (flags
& RTF_BLACKHOLE
)
850 SET_FLAG (zebra_flags
, ZEBRA_FLAG_BLACKHOLE
);
852 if (dest
.sa
.sa_family
== AF_INET
)
854 struct prefix_ipv4 p
;
857 p
.prefix
= dest
.sin
.sin_addr
;
858 if (flags
& RTF_HOST
)
859 p
.prefixlen
= IPV4_MAX_PREFIXLEN
;
861 p
.prefixlen
= ip_masklen (mask
.sin
.sin_addr
);
863 /* Catch self originated messages and match them against our current RIB.
864 * At the same time, ignore unconfirmed messages, they should be tracked
865 * by rtm_write() and kernel_rtm_ipv4().
867 if (rtm
->rtm_type
!= RTM_GET
&& rtm
->rtm_pid
== pid
)
869 char buf
[INET_ADDRSTRLEN
], gate_buf
[INET_ADDRSTRLEN
];
871 if (! IS_ZEBRA_DEBUG_RIB
)
873 ret
= rib_lookup_ipv4_route (&p
, &gate
);
874 inet_ntop (AF_INET
, &p
.prefix
, buf
, INET_ADDRSTRLEN
);
875 switch (rtm
->rtm_type
)
880 /* The kernel notifies us about a new route in FIB created by us.
881 Do we have a correspondent entry in our RIB? */
884 case ZEBRA_RIB_NOTFOUND
:
885 zlog_debug ("%s: %s %s/%d: desync: RR isn't yet in RIB, while already in FIB",
886 __func__
, lookup (rtm_type_str
, rtm
->rtm_type
), buf
, p
.prefixlen
);
888 case ZEBRA_RIB_FOUND_CONNECTED
:
889 case ZEBRA_RIB_FOUND_NOGATE
:
890 inet_ntop (AF_INET
, &gate
.sin
.sin_addr
, gate_buf
, INET_ADDRSTRLEN
);
891 zlog_debug ("%s: %s %s/%d: desync: RR is in RIB, but gate differs (ours is %s)",
892 __func__
, lookup (rtm_type_str
, rtm
->rtm_type
), buf
, p
.prefixlen
, gate_buf
);
894 case ZEBRA_RIB_FOUND_EXACT
: /* RIB RR == FIB RR */
895 zlog_debug ("%s: %s %s/%d: done Ok",
896 __func__
, lookup (rtm_type_str
, rtm
->rtm_type
), buf
, p
.prefixlen
);
897 rib_lookup_and_dump (&p
);
903 /* The kernel notifies us about a route deleted by us. Do we still
904 have it in the RIB? Do we have anything instead? */
907 case ZEBRA_RIB_FOUND_EXACT
:
908 zlog_debug ("%s: %s %s/%d: desync: RR is still in RIB, while already not in FIB",
909 __func__
, lookup (rtm_type_str
, rtm
->rtm_type
), buf
, p
.prefixlen
);
910 rib_lookup_and_dump (&p
);
912 case ZEBRA_RIB_FOUND_CONNECTED
:
913 case ZEBRA_RIB_FOUND_NOGATE
:
914 zlog_debug ("%s: %s %s/%d: desync: RR is still in RIB, plus gate differs",
915 __func__
, lookup (rtm_type_str
, rtm
->rtm_type
), buf
, p
.prefixlen
);
916 rib_lookup_and_dump (&p
);
918 case ZEBRA_RIB_NOTFOUND
: /* RIB RR == FIB RR */
919 zlog_debug ("%s: %s %s/%d: done Ok",
920 __func__
, lookup (rtm_type_str
, rtm
->rtm_type
), buf
, p
.prefixlen
);
921 rib_lookup_and_dump (&p
);
927 zlog_debug ("%s: %s/%d: warning: loopback RTM of type %s received",
928 __func__
, buf
, p
.prefixlen
, lookup (rtm_type_str
, rtm
->rtm_type
));
933 /* Change, delete the old prefix, we have no further information
934 * to specify the route really
936 if (rtm
->rtm_type
== RTM_CHANGE
)
937 rib_delete_ipv4 (ZEBRA_ROUTE_KERNEL
, zebra_flags
, &p
,
938 NULL
, 0, 0, SAFI_UNICAST
);
940 if (rtm
->rtm_type
== RTM_GET
941 || rtm
->rtm_type
== RTM_ADD
942 || rtm
->rtm_type
== RTM_CHANGE
)
943 rib_add_ipv4 (ZEBRA_ROUTE_KERNEL
, zebra_flags
,
944 &p
, &gate
.sin
.sin_addr
, NULL
, 0, 0, 0, 0, SAFI_UNICAST
);
946 rib_delete_ipv4 (ZEBRA_ROUTE_KERNEL
, zebra_flags
,
947 &p
, &gate
.sin
.sin_addr
, 0, 0, SAFI_UNICAST
);
950 if (dest
.sa
.sa_family
== AF_INET6
)
952 /* One day we might have a debug section here like one in the
953 * IPv4 case above. Just ignore own messages at the moment.
955 if (rtm
->rtm_type
!= RTM_GET
&& rtm
->rtm_pid
== pid
)
957 struct prefix_ipv6 p
;
958 unsigned int ifindex
= 0;
961 p
.prefix
= dest
.sin6
.sin6_addr
;
962 if (flags
& RTF_HOST
)
963 p
.prefixlen
= IPV6_MAX_PREFIXLEN
;
965 p
.prefixlen
= ip6_masklen (mask
.sin6
.sin6_addr
);
968 if (IN6_IS_ADDR_LINKLOCAL (&gate
.sin6
.sin6_addr
))
970 ifindex
= IN6_LINKLOCAL_IFINDEX (gate
.sin6
.sin6_addr
);
971 SET_IN6_LINKLOCAL_IFINDEX (gate
.sin6
.sin6_addr
, 0);
975 /* CHANGE: delete the old prefix, we have no further information
976 * to specify the route really
978 if (rtm
->rtm_type
== RTM_CHANGE
)
979 rib_delete_ipv6 (ZEBRA_ROUTE_KERNEL
, zebra_flags
, &p
,
980 NULL
, 0, 0, SAFI_UNICAST
);
982 if (rtm
->rtm_type
== RTM_GET
983 || rtm
->rtm_type
== RTM_ADD
984 || rtm
->rtm_type
== RTM_CHANGE
)
985 rib_add_ipv6 (ZEBRA_ROUTE_KERNEL
, zebra_flags
,
986 &p
, &gate
.sin6
.sin6_addr
, ifindex
, 0, 0, 0, SAFI_UNICAST
);
988 rib_delete_ipv6 (ZEBRA_ROUTE_KERNEL
, zebra_flags
,
989 &p
, &gate
.sin6
.sin6_addr
, ifindex
, 0, SAFI_UNICAST
);
991 #endif /* HAVE_IPV6 */
994 /* Interface function for the kernel routing table updates. Support
995 * for RTM_CHANGE will be needed.
996 * Exported only for rt_socket.c
999 rtm_write (int message
,
1000 union sockunion
*dest
,
1001 union sockunion
*mask
,
1002 union sockunion
*gate
,
1009 struct interface
*ifp
;
1011 /* Sequencial number of routing message. */
1012 static int msg_seq
= 0;
1014 /* Struct of rt_msghdr and buffer for storing socket's data. */
1017 struct rt_msghdr rtm
;
1021 if (routing_sock
< 0)
1022 return ZEBRA_ERR_EPERM
;
1024 /* Clear and set rt_msghdr values */
1025 memset (&msg
, 0, sizeof (struct rt_msghdr
));
1026 msg
.rtm
.rtm_version
= RTM_VERSION
;
1027 msg
.rtm
.rtm_type
= message
;
1028 msg
.rtm
.rtm_seq
= msg_seq
++;
1029 msg
.rtm
.rtm_addrs
= RTA_DST
;
1030 msg
.rtm
.rtm_addrs
|= RTA_GATEWAY
;
1031 msg
.rtm
.rtm_flags
= RTF_UP
;
1032 msg
.rtm
.rtm_index
= index
;
1036 msg
.rtm
.rtm_rmx
.rmx_hopcount
= metric
;
1037 msg
.rtm
.rtm_inits
|= RTV_HOPCOUNT
;
1040 ifp
= if_lookup_by_index (index
);
1042 if (gate
&& message
== RTM_ADD
)
1043 msg
.rtm
.rtm_flags
|= RTF_GATEWAY
;
1045 /* When RTF_CLONING is unavailable on BSD, should we set some
1046 * other flag instead?
1049 if (! gate
&& message
== RTM_ADD
&& ifp
&&
1050 (ifp
->flags
& IFF_POINTOPOINT
) == 0)
1051 msg
.rtm
.rtm_flags
|= RTF_CLONING
;
1052 #endif /* RTF_CLONING */
1054 /* If no protocol specific gateway is specified, use link
1055 address for gateway. */
1060 char dest_buf
[INET_ADDRSTRLEN
] = "NULL", mask_buf
[INET_ADDRSTRLEN
] = "255.255.255.255";
1062 inet_ntop (AF_INET
, &dest
->sin
.sin_addr
, dest_buf
, INET_ADDRSTRLEN
);
1064 inet_ntop (AF_INET
, &mask
->sin
.sin_addr
, mask_buf
, INET_ADDRSTRLEN
);
1065 zlog_warn ("%s: %s/%s: gate == NULL and no gateway found for ifindex %d",
1066 __func__
, dest_buf
, mask_buf
, index
);
1069 gate
= (union sockunion
*) & ifp
->sdl
;
1073 msg
.rtm
.rtm_addrs
|= RTA_NETMASK
;
1074 else if (message
== RTM_ADD
)
1075 msg
.rtm
.rtm_flags
|= RTF_HOST
;
1077 /* Tagging route with flags */
1078 msg
.rtm
.rtm_flags
|= (RTF_PROTO1
);
1080 /* Additional flags. */
1081 if (zebra_flags
& ZEBRA_FLAG_BLACKHOLE
)
1082 msg
.rtm
.rtm_flags
|= RTF_BLACKHOLE
;
1083 if (zebra_flags
& ZEBRA_FLAG_REJECT
)
1084 msg
.rtm
.rtm_flags
|= RTF_REJECT
;
1087 #ifdef HAVE_STRUCT_SOCKADDR_IN_SIN_LEN
1088 #define SOCKADDRSET(X,R) \
1089 if (msg.rtm.rtm_addrs & (R)) \
1091 int len = ROUNDUP ((X)->sa.sa_len); \
1092 memcpy (pnt, (caddr_t)(X), len); \
1096 #define SOCKADDRSET(X,R) \
1097 if (msg.rtm.rtm_addrs & (R)) \
1099 int len = SAROUNDUP (X); \
1100 memcpy (pnt, (caddr_t)(X), len); \
1103 #endif /* HAVE_STRUCT_SOCKADDR_IN_SIN_LEN */
1105 pnt
= (caddr_t
) msg
.buf
;
1107 /* Write each socket data into rtm message buffer */
1108 SOCKADDRSET (dest
, RTA_DST
);
1109 SOCKADDRSET (gate
, RTA_GATEWAY
);
1110 SOCKADDRSET (mask
, RTA_NETMASK
);
1112 msg
.rtm
.rtm_msglen
= pnt
- (caddr_t
) &msg
;
1114 ret
= write (routing_sock
, &msg
, msg
.rtm
.rtm_msglen
);
1116 if (ret
!= msg
.rtm
.rtm_msglen
)
1118 if (errno
== EEXIST
)
1119 return ZEBRA_ERR_RTEXIST
;
1120 if (errno
== ENETUNREACH
)
1121 return ZEBRA_ERR_RTUNREACH
;
1123 return ZEBRA_ERR_RTNOEXIST
;
1125 zlog_warn ("%s: write : %s (%d)", __func__
, safe_strerror (errno
), errno
);
1126 return ZEBRA_ERR_KERNEL
;
1128 return ZEBRA_ERR_NOERROR
;
1133 #include "zebra/zserv.h"
1135 /* For debug purpose. */
1137 rtmsg_debug (struct rt_msghdr
*rtm
)
1139 zlog_debug ("Kernel: Len: %d Type: %s", rtm
->rtm_msglen
, lookup (rtm_type_str
, rtm
->rtm_type
));
1140 rtm_flag_dump (rtm
->rtm_flags
);
1141 zlog_debug ("Kernel: message seq %d", rtm
->rtm_seq
);
1142 zlog_debug ("Kernel: pid %d, rtm_addrs 0x%x", rtm
->rtm_pid
, rtm
->rtm_addrs
);
1145 /* This is pretty gross, better suggestions welcome -- mhandler */
1148 #define RTAX_MAX RTA_NUMBITS
1151 #endif /* RTA_NUMBITS */
1152 #endif /* RTAX_MAX */
1154 /* Kernel routing table and interface updates via routing socket. */
1156 kernel_read (struct thread
*thread
)
1160 struct rt_msghdr
*rtm
;
1163 * This must be big enough for any message the kernel might send.
1164 * Rather than determining how many sockaddrs of what size might be
1165 * in each particular message, just use RTAX_MAX of sockaddr_storage
1166 * for each. Note that the sockaddrs must be after each message
1167 * definition, or rather after whichever happens to be the largest,
1168 * since the buffer needs to be big enough for a message and the
1169 * sockaddrs together.
1173 /* Routing information. */
1176 struct rt_msghdr rtm
;
1177 struct sockaddr_storage addr
[RTAX_MAX
];
1180 /* Interface information. */
1183 struct if_msghdr ifm
;
1184 struct sockaddr_storage addr
[RTAX_MAX
];
1187 /* Interface address information. */
1190 struct ifa_msghdr ifa
;
1191 struct sockaddr_storage addr
[RTAX_MAX
];
1194 #ifdef RTM_IFANNOUNCE
1195 /* Interface arrival/departure */
1198 struct if_announcemsghdr ifan
;
1199 struct sockaddr_storage addr
[RTAX_MAX
];
1201 #endif /* RTM_IFANNOUNCE */
1205 /* Fetch routing socket. */
1206 sock
= THREAD_FD (thread
);
1208 nbytes
= read (sock
, &buf
, sizeof buf
);
1212 if (nbytes
< 0 && errno
!= EWOULDBLOCK
&& errno
!= EAGAIN
)
1213 zlog_warn ("routing socket error: %s", safe_strerror (errno
));
1217 thread_add_read (zebrad
.master
, kernel_read
, NULL
, sock
);
1219 if (IS_ZEBRA_DEBUG_KERNEL
)
1220 rtmsg_debug (&buf
.r
.rtm
);
1225 * Ensure that we didn't drop any data, so that processing routines
1226 * can assume they have the whole message.
1228 if (rtm
->rtm_msglen
!= nbytes
)
1230 zlog_warn ("kernel_read: rtm->rtm_msglen %d, nbytes %d, type %d\n",
1231 rtm
->rtm_msglen
, nbytes
, rtm
->rtm_type
);
1235 switch (rtm
->rtm_type
)
1243 ifm_read (&buf
.im
.ifm
);
1247 ifam_read (&buf
.ia
.ifa
);
1249 #ifdef RTM_IFANNOUNCE
1250 case RTM_IFANNOUNCE
:
1251 ifan_read (&buf
.ian
.ifan
);
1253 #endif /* RTM_IFANNOUNCE */
1255 if (IS_ZEBRA_DEBUG_KERNEL
)
1256 zlog_debug("Unprocessed RTM_type: %d", rtm
->rtm_type
);
1262 /* Make routing socket. */
1264 routing_socket (void)
1266 if ( zserv_privs
.change (ZPRIVS_RAISE
) )
1267 zlog_err ("routing_socket: Can't raise privileges");
1269 routing_sock
= socket (AF_ROUTE
, SOCK_RAW
, 0);
1271 if (routing_sock
< 0)
1273 if ( zserv_privs
.change (ZPRIVS_LOWER
) )
1274 zlog_err ("routing_socket: Can't lower privileges");
1275 zlog_warn ("Can't init kernel routing socket");
1279 /* XXX: Socket should be NONBLOCK, however as we currently
1280 * discard failed writes, this will lead to inconsistencies.
1281 * For now, socket must be blocking.
1283 /*if (fcntl (routing_sock, F_SETFL, O_NONBLOCK) < 0)
1284 zlog_warn ("Can't set O_NONBLOCK to routing socket");*/
1286 if ( zserv_privs
.change (ZPRIVS_LOWER
) )
1287 zlog_err ("routing_socket: Can't lower privileges");
1289 /* kernel_read needs rewrite. */
1290 thread_add_read (zebrad
.master
, kernel_read
, NULL
, routing_sock
);
1293 /* Exported interface function. This function simply calls
1294 routing_socket (). */