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
53 ((a) > 0 ? (1 + (((a) - 1) | (sizeof(long) - 1))) : sizeof(long))
56 * Given a pointer (sockaddr or void *), return the number of bytes
57 * taken up by the sockaddr and any padding needed for alignment.
59 #if defined(HAVE_STRUCT_SOCKADDR_SA_LEN)
60 #define SAROUNDUP(X) ROUNDUP(((struct sockaddr *)(X))->sa_len)
61 #elif defined(HAVE_IPV6)
63 * One would hope all fixed-size structure definitions are aligned,
64 * but round them up nonetheless.
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))))
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 */
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.
86 #define RTA_ADDR_GET(DEST, RTA, RTMADDRS, PNT) \
87 if ((RTMADDRS) & (RTA)) \
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); \
96 #define RTA_ATTR_GET(DEST, RTA, RTMADDRS, PNT) \
97 if ((RTMADDRS) & (RTA)) \
99 void *pdest = (DEST); \
100 int len = SAROUNDUP ((PNT)); \
101 if ((DEST) != NULL) \
102 memcpy (pdest, (PNT), len); \
106 #define RTA_NAME_GET(DEST, RTA, RTMADDRS, PNT, LEN) \
107 if ((RTMADDRS) & (RTA)) \
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) ) \
118 memcpy (pdest, sdl->sdl_data, sdl->sdl_nlen); \
119 pdest[sdl->sdl_nlen] = '\0'; \
120 (LEN) = sdl->sdl_nlen; \
128 /* Routing socket message types. */
129 const struct message rtm_type_str
[] =
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"},
140 {RTM_OLDADD
, "RTM_OLDADD"},
141 #endif /* RTM_OLDADD */
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"},
150 {RTM_OIFINFO
, "RTM_OIFINFO"},
151 #endif /* RTM_OIFINFO */
153 {RTM_NEWMADDR
, "RTM_NEWMADDR"},
154 #endif /* RTM_NEWMADDR */
156 {RTM_DELMADDR
, "RTM_DELMADDR"},
157 #endif /* RTM_DELMADDR */
158 #ifdef RTM_IFANNOUNCE
159 {RTM_IFANNOUNCE
, "RTM_IFANNOUNCE"},
160 #endif /* RTM_IFANNOUNCE */
164 static const struct message rtm_flag_str
[] =
167 {RTF_GATEWAY
, "GATEWAY"},
169 {RTF_REJECT
, "REJECT"},
170 {RTF_DYNAMIC
, "DYNAMIC"},
171 {RTF_MODIFIED
, "MODIFIED"},
175 #endif /* RTF_MASK */
177 {RTF_CLONING
, "CLONING"},
178 #endif /* RTF_CLONING */
179 {RTF_XRESOLVE
, "XRESOLVE"},
180 {RTF_LLINFO
, "LLINFO"},
181 {RTF_STATIC
, "STATIC"},
182 {RTF_BLACKHOLE
, "BLACKHOLE"},
184 {RTF_PRIVATE
, "PRIVATE"},
185 #endif /* RTF_PRIVATE */
186 {RTF_PROTO1
, "PROTO1"},
187 {RTF_PROTO2
, "PROTO2"},
189 {RTF_PRCLONING
, "PRCLONING"},
190 #endif /* RTF_PRCLONING */
192 {RTF_WASCLONED
, "WASCLONED"},
193 #endif /* RTF_WASCLONED */
195 {RTF_PROTO3
, "PROTO3"},
196 #endif /* RTF_PROTO3 */
198 {RTF_PINNED
, "PINNED"},
199 #endif /* RTF_PINNED */
201 {RTF_LOCAL
, "LOCAL"},
202 #endif /* RTF_LOCAL */
204 {RTF_BROADCAST
, "BROADCAST"},
205 #endif /* RTF_BROADCAST */
207 {RTF_MULTICAST
, "MULTICAST"},
208 #endif /* RTF_MULTICAST */
210 {RTF_MULTIRT
, "MULTIRT"},
211 #endif /* RTF_MULTIRT */
213 {RTF_SETSRC
, "SETSRC"},
214 #endif /* RTF_SETSRC */
218 /* Kernel routing update socket. */
219 int routing_sock
= -1;
221 /* Yes I'm checking ugly routing socket behavior. */
224 /* Supported address family check. */
226 af_check (int family
)
228 if (family
== AF_INET
)
231 if (family
== AF_INET6
)
233 #endif /* HAVE_IPV6 */
237 /* Dump routing table flag for debug purpose. */
239 rtm_flag_dump (int flag
)
241 const struct message
*mes
;
242 static char buf
[BUFSIZ
];
245 for (mes
= rtm_flag_str
; mes
->key
!= 0; mes
++)
249 strlcat (buf
, mes
->str
, BUFSIZ
);
250 strlcat (buf
, " ", BUFSIZ
);
253 zlog_debug ("Kernel: %s", buf
);
256 #ifdef RTM_IFANNOUNCE
257 /* Interface adding function */
259 ifan_read (struct if_announcemsghdr
*ifan
)
261 struct interface
*ifp
;
263 ifp
= if_lookup_by_index (ifan
->ifan_index
);
266 assert ( (ifp
->ifindex
== ifan
->ifan_index
)
267 || (ifp
->ifindex
== IFINDEX_INTERNAL
) );
270 || ((ifp
->ifindex
== IFINDEX_INTERNAL
)
271 && (ifan
->ifan_what
== IFAN_ARRIVAL
)) )
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
);
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
;
285 else if (ifp
!= NULL
&& ifan
->ifan_what
== IFAN_DEPARTURE
)
286 if_delete_update (ifp
);
292 if (IS_ZEBRA_DEBUG_KERNEL
)
293 zlog_debug ("%s: interface %s index %d",
294 __func__
, ifan
->ifan_name
, ifan
->ifan_index
);
298 #endif /* RTM_IFANNOUNCE */
300 #ifdef HAVE_BSD_LINK_DETECT
301 /* BSD link detect translation */
303 bsd_linkdetect_translate (struct if_msghdr
*ifm
)
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
);
309 UNSET_FLAG(ifm
->ifm_flags
, IFF_RUNNING
);
311 #endif /* HAVE_BSD_LINK_DETECT */
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.
319 ifm_read (struct if_msghdr
*ifm
)
321 struct interface
*ifp
= NULL
;
322 struct sockaddr_dl
*sdl
;
323 char ifname
[IFNAMSIZ
];
327 /* terminate ifname at head (for strnlen) and tail (for safety) */
328 ifname
[IFNAMSIZ
- 1] = '\0';
330 /* paranoia: sanity check structure */
331 if (ifm
->ifm_msglen
< sizeof(struct if_msghdr
))
333 zlog_err ("ifm_read: ifm->ifm_msglen %d too short\n",
339 * Check for a sockaddr_dl following the message. First, point to
340 * where a socakddr might be if one follows the message.
342 cp
= (void *)(ifm
+ 1);
346 * XXX This behavior should be narrowed to only the kernel versions
347 * for which the structures returned do not match the headers.
349 * if_msghdr_t on 64 bit kernels in Solaris 9 and earlier versions
350 * is 12 bytes larger than the 32 bit version.
352 if (((struct sockaddr
*) cp
)->sa_family
== AF_UNSPEC
)
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
);
360 sdl
= (struct sockaddr_dl
*)cp
;
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
);
366 if (IS_ZEBRA_DEBUG_KERNEL
)
367 zlog_debug ("%s: sdl ifname %s", __func__
, (ifnlen
? ifname
: "(nil)"));
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
375 if ( (ifp
= if_lookup_by_index (ifm
->ifm_index
)) != NULL
)
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
382 if (ifnlen
&& (strncmp (ifp
->name
, ifname
, IFNAMSIZ
) != 0) )
384 if (IS_ZEBRA_DEBUG_KERNEL
)
385 zlog_debug ("%s: ifp name %s doesnt match sdl name %s",
386 __func__
, ifp
->name
, ifname
);
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.
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
401 if ( (ifp
== NULL
) && ifnlen
)
402 ifp
= if_lookup_by_name (ifname
);
405 * If ifp still does not exist or has an invalid index (IFINDEX_INTERNAL),
406 * create or fill in an interface.
408 if ((ifp
== NULL
) || (ifp
->ifindex
== IFINDEX_INTERNAL
))
411 * To create or fill in an interface, a sockaddr_dl (via
412 * RTA_IFP) is required.
416 zlog_warn ("Interface index %d (new) missing ifname\n",
421 #ifndef RTM_IFANNOUNCE
422 /* Down->Down interface should be ignored here.
423 * See further comment below.
425 if (!CHECK_FLAG (ifm
->ifm_flags
, IFF_UP
))
427 #endif /* !RTM_IFANNOUNCE */
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
);
438 if (IS_ZEBRA_DEBUG_KERNEL
)
439 zlog_debug ("%s: updated/created ifp, ifname %s, ifindex %d",
440 __func__
, ifp
->name
, ifp
->ifindex
);
442 * Fill in newly created interface structure, or larval
443 * structure with ifindex IFINDEX_INTERNAL.
445 ifp
->ifindex
= ifm
->ifm_index
;
447 #ifdef HAVE_BSD_LINK_DETECT /* translate BSD kernel msg for link-state */
448 bsd_linkdetect_translate(ifm
);
449 #endif /* HAVE_BSD_LINK_DETECT */
451 if_flags_update (ifp
, ifm
->ifm_flags
);
452 #if defined(__bsdi__)
453 if_kvm_get_mtu (ifp
);
456 #endif /* __bsdi__ */
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.
464 * a nonzero ifnlen from RTA_NAME_GET() means sdl is valid
467 memcpy (&ifp
->sdl
, sdl
, sizeof (struct sockaddr_dl
));
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.)
480 if (ifp
->ifindex
!= ifm
->ifm_index
)
482 zlog_warn ("%s: index mismatch, ifname %s, ifp index %d, "
484 __func__
, ifp
->name
, ifp
->ifindex
, ifm
->ifm_index
);
488 #ifdef HAVE_BSD_LINK_DETECT /* translate BSD kernel msg for link-state */
489 bsd_linkdetect_translate(ifm
);
490 #endif /* HAVE_BSD_LINK_DETECT */
492 /* update flags and handle operative->inoperative transition, if any */
493 if_flags_update (ifp
, ifm
->ifm_flags
);
495 #ifndef RTM_IFANNOUNCE
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.
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
507 if_delete_update (ifp
);
509 #endif /* RTM_IFANNOUNCE */
512 #if defined(__bsdi__)
513 if_kvm_get_mtu (ifp
);
516 #endif /* __bsdi__ */
521 #ifdef HAVE_NET_RT_IFLIST
522 ifp
->stats
= ifm
->ifm_data
;
523 #endif /* HAVE_NET_RT_IFLIST */
525 if (IS_ZEBRA_DEBUG_KERNEL
)
526 zlog_debug ("%s: interface %s index %d",
527 __func__
, ifp
->name
, ifp
->ifindex
);
532 /* Address read from struct ifa_msghdr. */
534 ifam_read_mesg (struct ifa_msghdr
*ifm
,
535 union sockunion
*addr
,
536 union sockunion
*mask
,
537 union sockunion
*brd
,
543 union sockunion gateway
;
545 pnt
= (caddr_t
)(ifm
+ 1);
546 end
= ((caddr_t
)ifm
) + ifm
->ifam_msglen
;
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
));
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
);
565 if (IS_ZEBRA_DEBUG_KERNEL
)
567 switch (sockunion_family(addr
))
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 "
575 __func__
, ifm
->ifam_index
,
576 (ifnlen
? ifname
: "(nil)"), ifm
->ifam_addrs
,
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])));
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 "
596 __func__
, ifm
->ifam_index
,
597 (ifnlen
? ifname
: "(nil)"), ifm
->ifam_addrs
,
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])));
610 #endif /* HAVE_IPV6 */
612 zlog_debug ("%s: ifindex %d, ifname %s, ifam_addrs 0x%x",
613 __func__
, ifm
->ifam_index
,
614 (ifnlen
? ifname
: "(nil)"), ifm
->ifam_addrs
);
619 /* Assert read up end point matches to end point */
621 zlog_warn ("ifam_read() doesn't read all socket data");
624 /* Interface's address information get. */
626 ifam_read (struct ifa_msghdr
*ifam
)
628 struct interface
*ifp
= NULL
;
629 union sockunion addr
, mask
, brd
;
630 char ifname
[INTERFACE_NAMSIZ
];
635 ifname
[0] = ifname
[INTERFACE_NAMSIZ
- 1] = '\0';
637 /* Allocate and read address information. */
638 ifam_read_mesg (ifam
, &addr
, &mask
, &brd
, ifname
, &ifnlen
);
640 if ((ifp
= if_lookup_by_index(ifam
->ifam_index
)) == NULL
)
642 zlog_warn ("%s: no interface for ifname %s, index %d",
643 __func__
, ifname
, ifam
->ifam_index
);
647 if (ifnlen
&& strncmp (ifp
->name
, ifname
, INTERFACE_NAMSIZ
))
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
);
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.
662 ifp
->metric
= ifam
->ifam_metric
;
665 /* Add connected address. */
666 switch (sockunion_family (&addr
))
669 if (ifam
->ifam_type
== RTM_NEWADDR
)
670 connected_add_ipv4 (ifp
, flags
, &addr
.sin
.sin_addr
,
671 ip_masklen (mask
.sin
.sin_addr
),
673 (isalias
? ifname
: NULL
));
675 connected_delete_ipv4 (ifp
, flags
, &addr
.sin
.sin_addr
,
676 ip_masklen (mask
.sin
.sin_addr
),
681 /* Unset interface index from link-local address when IPv6 stack
683 if (IN6_IS_ADDR_LINKLOCAL (&addr
.sin6
.sin6_addr
))
684 SET_IN6_LINKLOCAL_IFINDEX (addr
.sin6
.sin6_addr
, 0);
686 if (ifam
->ifam_type
== RTM_NEWADDR
)
687 connected_add_ipv6 (ifp
, flags
, &addr
.sin6
.sin6_addr
,
688 ip6_masklen (mask
.sin6
.sin6_addr
),
690 (isalias
? ifname
: NULL
));
692 connected_delete_ipv6 (ifp
,
693 &addr
.sin6
.sin6_addr
,
694 ip6_masklen (mask
.sin6
.sin6_addr
),
695 &brd
.sin6
.sin6_addr
);
697 #endif /* HAVE_IPV6 */
699 /* Unsupported family silently ignore... */
703 /* Check interface flag for implicit up of the interface. */
707 /* In addition to lacking IFANNOUNCE, on SUNOS IFF_UP is strange.
708 * See comments for SUNOS_5 in interface.c::if_flags_mangle.
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.
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.
720 if_delete_update (ifp
);
726 /* Interface function for reading kernel routing table information. */
728 rtm_read_mesg (struct rt_msghdr
*rtm
,
729 union sockunion
*dest
,
730 union sockunion
*mask
,
731 union sockunion
*gate
,
737 /* Pnt points out socket data start point. */
738 pnt
= (caddr_t
)(rtm
+ 1);
739 end
= ((caddr_t
)rtm
) + rtm
->rtm_msglen
;
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
);
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
));
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
);
762 /* If there is netmask information set it's family same as
764 if (rtm
->rtm_addrs
& RTA_NETMASK
)
765 mask
->sa
.sa_family
= dest
->sa
.sa_family
;
767 /* Assert read up to the end of pointer. */
769 zlog (NULL
, LOG_WARNING
, "rtm_read() doesn't read all socket data.");
771 return rtm
->rtm_flags
;
775 rtm_read (struct rt_msghdr
*rtm
)
779 union sockunion dest
, mask
, gate
;
780 char ifname
[INTERFACE_NAMSIZ
+ 1];
785 /* Read destination and netmask and gateway from rtm message
787 flags
= rtm_read_mesg (rtm
, &dest
, &mask
, &gate
, ifname
, &ifnlen
);
788 if (!(flags
& RTF_DONE
))
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
));
794 #ifdef RTF_CLONED /*bsdi, netbsd 1.6*/
795 if (flags
& RTF_CLONED
)
798 #ifdef RTF_WASCLONED /*freebsd*/
799 if (flags
& RTF_WASCLONED
)
803 if ((rtm
->rtm_type
== RTM_ADD
) && ! (flags
& RTF_UP
))
806 /* This is connected route. */
807 if (! (flags
& RTF_GATEWAY
))
810 if (flags
& RTF_PROTO1
)
811 SET_FLAG (zebra_flags
, ZEBRA_FLAG_SELFROUTE
);
813 /* This is persistent route. */
814 if (flags
& RTF_STATIC
)
815 SET_FLAG (zebra_flags
, ZEBRA_FLAG_STATIC
);
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
);
823 if (dest
.sa
.sa_family
== AF_INET
)
825 struct prefix_ipv4 p
;
828 p
.prefix
= dest
.sin
.sin_addr
;
829 if (flags
& RTF_HOST
)
830 p
.prefixlen
= IPV4_MAX_PREFIXLEN
;
832 p
.prefixlen
= ip_masklen (mask
.sin
.sin_addr
);
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().
838 if (rtm
->rtm_type
!= RTM_GET
&& rtm
->rtm_pid
== pid
)
840 char buf
[INET_ADDRSTRLEN
], gate_buf
[INET_ADDRSTRLEN
];
842 if (! IS_ZEBRA_DEBUG_RIB
)
844 ret
= rib_lookup_ipv4_route (&p
, &gate
);
845 inet_ntop (AF_INET
, &p
.prefix
, buf
, INET_ADDRSTRLEN
);
846 switch (rtm
->rtm_type
)
851 /* The kernel notifies us about a new route in FIB created by us.
852 Do we have a correspondent entry in our RIB? */
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
);
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
);
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
);
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? */
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
);
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
);
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
);
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
));
904 /* Change, delete the old prefix, we have no further information
905 * to specify the route really
907 if (rtm
->rtm_type
== RTM_CHANGE
)
908 rib_delete_ipv4 (ZEBRA_ROUTE_KERNEL
, zebra_flags
, &p
,
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
,
915 &p
, &gate
.sin
.sin_addr
, NULL
, 0, 0, 0, 0);
917 rib_delete_ipv4 (ZEBRA_ROUTE_KERNEL
, zebra_flags
,
918 &p
, &gate
.sin
.sin_addr
, 0, 0);
921 if (dest
.sa
.sa_family
== AF_INET6
)
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.
926 if (rtm
->rtm_type
!= RTM_GET
&& rtm
->rtm_pid
== pid
)
928 struct prefix_ipv6 p
;
929 unsigned int ifindex
= 0;
932 p
.prefix
= dest
.sin6
.sin6_addr
;
933 if (flags
& RTF_HOST
)
934 p
.prefixlen
= IPV6_MAX_PREFIXLEN
;
936 p
.prefixlen
= ip6_masklen (mask
.sin6
.sin6_addr
);
939 if (IN6_IS_ADDR_LINKLOCAL (&gate
.sin6
.sin6_addr
))
941 ifindex
= IN6_LINKLOCAL_IFINDEX (gate
.sin6
.sin6_addr
);
942 SET_IN6_LINKLOCAL_IFINDEX (gate
.sin6
.sin6_addr
, 0);
946 /* CHANGE: delete the old prefix, we have no further information
947 * to specify the route really
949 if (rtm
->rtm_type
== RTM_CHANGE
)
950 rib_delete_ipv6 (ZEBRA_ROUTE_KERNEL
, zebra_flags
, &p
,
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
,
957 &p
, &gate
.sin6
.sin6_addr
, ifindex
, 0, 0, 0);
959 rib_delete_ipv6 (ZEBRA_ROUTE_KERNEL
, zebra_flags
,
960 &p
, &gate
.sin6
.sin6_addr
, ifindex
, 0);
962 #endif /* HAVE_IPV6 */
965 /* Interface function for the kernel routing table updates. Support
966 * for RTM_CHANGE will be needed.
967 * Exported only for rt_socket.c
970 rtm_write (int message
,
971 union sockunion
*dest
,
972 union sockunion
*mask
,
973 union sockunion
*gate
,
980 struct interface
*ifp
;
982 /* Sequencial number of routing message. */
983 static int msg_seq
= 0;
985 /* Struct of rt_msghdr and buffer for storing socket's data. */
988 struct rt_msghdr rtm
;
992 if (routing_sock
< 0)
993 return ZEBRA_ERR_EPERM
;
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
;
1007 msg
.rtm
.rtm_rmx
.rmx_hopcount
= metric
;
1008 msg
.rtm
.rtm_inits
|= RTV_HOPCOUNT
;
1011 ifp
= if_lookup_by_index (index
);
1013 if (gate
&& message
== RTM_ADD
)
1014 msg
.rtm
.rtm_flags
|= RTF_GATEWAY
;
1016 /* When RTF_CLONING is unavailable on BSD, should we set some
1017 * other flag instead?
1020 if (! gate
&& message
== RTM_ADD
&& ifp
&&
1021 (ifp
->flags
& IFF_POINTOPOINT
) == 0)
1022 msg
.rtm
.rtm_flags
|= RTF_CLONING
;
1023 #endif /* RTF_CLONING */
1025 /* If no protocol specific gateway is specified, use link
1026 address for gateway. */
1031 char dest_buf
[INET_ADDRSTRLEN
] = "NULL", mask_buf
[INET_ADDRSTRLEN
] = "255.255.255.255";
1033 inet_ntop (AF_INET
, &dest
->sin
.sin_addr
, dest_buf
, INET_ADDRSTRLEN
);
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
);
1040 gate
= (union sockunion
*) & ifp
->sdl
;
1044 msg
.rtm
.rtm_addrs
|= RTA_NETMASK
;
1045 else if (message
== RTM_ADD
)
1046 msg
.rtm
.rtm_flags
|= RTF_HOST
;
1048 /* Tagging route with flags */
1049 msg
.rtm
.rtm_flags
|= (RTF_PROTO1
);
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
;
1058 #ifdef HAVE_STRUCT_SOCKADDR_IN_SIN_LEN
1059 #define SOCKADDRSET(X,R) \
1060 if (msg.rtm.rtm_addrs & (R)) \
1062 int len = ROUNDUP ((X)->sa.sa_len); \
1063 memcpy (pnt, (caddr_t)(X), len); \
1067 #define SOCKADDRSET(X,R) \
1068 if (msg.rtm.rtm_addrs & (R)) \
1070 int len = SAROUNDUP (X); \
1071 memcpy (pnt, (caddr_t)(X), len); \
1074 #endif /* HAVE_STRUCT_SOCKADDR_IN_SIN_LEN */
1076 pnt
= (caddr_t
) msg
.buf
;
1078 /* Write each socket data into rtm message buffer */
1079 SOCKADDRSET (dest
, RTA_DST
);
1080 SOCKADDRSET (gate
, RTA_GATEWAY
);
1081 SOCKADDRSET (mask
, RTA_NETMASK
);
1083 msg
.rtm
.rtm_msglen
= pnt
- (caddr_t
) &msg
;
1085 ret
= write (routing_sock
, &msg
, msg
.rtm
.rtm_msglen
);
1087 if (ret
!= msg
.rtm
.rtm_msglen
)
1089 if (errno
== EEXIST
)
1090 return ZEBRA_ERR_RTEXIST
;
1091 if (errno
== ENETUNREACH
)
1092 return ZEBRA_ERR_RTUNREACH
;
1094 return ZEBRA_ERR_RTNOEXIST
;
1096 zlog_warn ("%s: write : %s (%d)", __func__
, safe_strerror (errno
), errno
);
1097 return ZEBRA_ERR_KERNEL
;
1099 return ZEBRA_ERR_NOERROR
;
1104 #include "zebra/zserv.h"
1106 /* For debug purpose. */
1108 rtmsg_debug (struct rt_msghdr
*rtm
)
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
);
1116 /* This is pretty gross, better suggestions welcome -- mhandler */
1119 #define RTAX_MAX RTA_NUMBITS
1122 #endif /* RTA_NUMBITS */
1123 #endif /* RTAX_MAX */
1125 /* Kernel routing table and interface updates via routing socket. */
1127 kernel_read (struct thread
*thread
)
1131 struct rt_msghdr
*rtm
;
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.
1144 /* Routing information. */
1147 struct rt_msghdr rtm
;
1148 struct sockaddr_storage addr
[RTAX_MAX
];
1151 /* Interface information. */
1154 struct if_msghdr ifm
;
1155 struct sockaddr_storage addr
[RTAX_MAX
];
1158 /* Interface address information. */
1161 struct ifa_msghdr ifa
;
1162 struct sockaddr_storage addr
[RTAX_MAX
];
1165 #ifdef RTM_IFANNOUNCE
1166 /* Interface arrival/departure */
1169 struct if_announcemsghdr ifan
;
1170 struct sockaddr_storage addr
[RTAX_MAX
];
1172 #endif /* RTM_IFANNOUNCE */
1176 /* Fetch routing socket. */
1177 sock
= THREAD_FD (thread
);
1179 nbytes
= read (sock
, &buf
, sizeof buf
);
1183 if (nbytes
< 0 && errno
!= EWOULDBLOCK
&& errno
!= EAGAIN
)
1184 zlog_warn ("routing socket error: %s", safe_strerror (errno
));
1188 thread_add_read (zebrad
.master
, kernel_read
, NULL
, sock
);
1190 if (IS_ZEBRA_DEBUG_KERNEL
)
1191 rtmsg_debug (&buf
.r
.rtm
);
1196 * Ensure that we didn't drop any data, so that processing routines
1197 * can assume they have the whole message.
1199 if (rtm
->rtm_msglen
!= nbytes
)
1201 zlog_warn ("kernel_read: rtm->rtm_msglen %d, nbytes %d, type %d\n",
1202 rtm
->rtm_msglen
, nbytes
, rtm
->rtm_type
);
1206 switch (rtm
->rtm_type
)
1214 ifm_read (&buf
.im
.ifm
);
1218 ifam_read (&buf
.ia
.ifa
);
1220 #ifdef RTM_IFANNOUNCE
1221 case RTM_IFANNOUNCE
:
1222 ifan_read (&buf
.ian
.ifan
);
1224 #endif /* RTM_IFANNOUNCE */
1226 if (IS_ZEBRA_DEBUG_KERNEL
)
1227 zlog_debug("Unprocessed RTM_type: %d", rtm
->rtm_type
);
1233 /* Make routing socket. */
1235 routing_socket (void)
1237 if ( zserv_privs
.change (ZPRIVS_RAISE
) )
1238 zlog_err ("routing_socket: Can't raise privileges");
1240 routing_sock
= socket (AF_ROUTE
, SOCK_RAW
, 0);
1242 if (routing_sock
< 0)
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");
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.
1254 /*if (fcntl (routing_sock, F_SETFL, O_NONBLOCK) < 0)
1255 zlog_warn ("Can't set O_NONBLOCK to routing socket");*/
1257 if ( zserv_privs
.change (ZPRIVS_LOWER
) )
1258 zlog_err ("routing_socket: Can't lower privileges");
1260 /* kernel_read needs rewrite. */
1261 thread_add_read (zebrad
.master
, kernel_read
, NULL
, routing_sock
);
1264 /* Exported interface function. This function simply calls
1265 routing_socket (). */