2 * DECnet An implementation of the DECnet protocol suite for the LINUX
3 * operating system. DECnet is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
6 * DECnet Routing Forwarding Information Base (Glue/Info List)
8 * Author: Steve Whitehouse <SteveW@ACM.org>
12 * Alexey Kuznetsov : SMP locking changes
13 * Steve Whitehouse : Rewrote it... Well to be more correct, I
14 * copied most of it from the ipv4 fib code.
15 * Steve Whitehouse : Updated it in style and fixed a few bugs
16 * which were fixed in the ipv4 code since
17 * this code was copied from it.
20 #include <linux/string.h>
21 #include <linux/net.h>
22 #include <linux/socket.h>
23 #include <linux/slab.h>
24 #include <linux/sockios.h>
25 #include <linux/init.h>
26 #include <linux/skbuff.h>
27 #include <linux/netlink.h>
28 #include <linux/rtnetlink.h>
29 #include <linux/proc_fs.h>
30 #include <linux/netdevice.h>
31 #include <linux/timer.h>
32 #include <linux/spinlock.h>
33 #include <linux/atomic.h>
34 #include <asm/uaccess.h>
35 #include <net/neighbour.h>
38 #include <net/fib_rules.h>
40 #include <net/dn_route.h>
41 #include <net/dn_fib.h>
42 #include <net/dn_neigh.h>
43 #include <net/dn_dev.h>
45 #define RT_MIN_TABLE 1
47 #define for_fib_info() { struct dn_fib_info *fi;\
48 for(fi = dn_fib_info_list; fi; fi = fi->fib_next)
49 #define endfor_fib_info() }
51 #define for_nexthops(fi) { int nhsel; const struct dn_fib_nh *nh;\
52 for(nhsel = 0, nh = (fi)->fib_nh; nhsel < (fi)->fib_nhs; nh++, nhsel++)
54 #define change_nexthops(fi) { int nhsel; struct dn_fib_nh *nh;\
55 for(nhsel = 0, nh = (struct dn_fib_nh *)((fi)->fib_nh); nhsel < (fi)->fib_nhs; nh++, nhsel++)
57 #define endfor_nexthops(fi) }
59 static DEFINE_SPINLOCK(dn_fib_multipath_lock
);
60 static struct dn_fib_info
*dn_fib_info_list
;
61 static DEFINE_SPINLOCK(dn_fib_info_lock
);
67 } dn_fib_props
[RTN_MAX
+1] = {
68 [RTN_UNSPEC
] = { .error
= 0, .scope
= RT_SCOPE_NOWHERE
},
69 [RTN_UNICAST
] = { .error
= 0, .scope
= RT_SCOPE_UNIVERSE
},
70 [RTN_LOCAL
] = { .error
= 0, .scope
= RT_SCOPE_HOST
},
71 [RTN_BROADCAST
] = { .error
= -EINVAL
, .scope
= RT_SCOPE_NOWHERE
},
72 [RTN_ANYCAST
] = { .error
= -EINVAL
, .scope
= RT_SCOPE_NOWHERE
},
73 [RTN_MULTICAST
] = { .error
= -EINVAL
, .scope
= RT_SCOPE_NOWHERE
},
74 [RTN_BLACKHOLE
] = { .error
= -EINVAL
, .scope
= RT_SCOPE_UNIVERSE
},
75 [RTN_UNREACHABLE
] = { .error
= -EHOSTUNREACH
, .scope
= RT_SCOPE_UNIVERSE
},
76 [RTN_PROHIBIT
] = { .error
= -EACCES
, .scope
= RT_SCOPE_UNIVERSE
},
77 [RTN_THROW
] = { .error
= -EAGAIN
, .scope
= RT_SCOPE_UNIVERSE
},
78 [RTN_NAT
] = { .error
= 0, .scope
= RT_SCOPE_NOWHERE
},
79 [RTN_XRESOLVE
] = { .error
= -EINVAL
, .scope
= RT_SCOPE_NOWHERE
},
82 static int dn_fib_sync_down(__le16 local
, struct net_device
*dev
, int force
);
83 static int dn_fib_sync_up(struct net_device
*dev
);
85 void dn_fib_free_info(struct dn_fib_info
*fi
)
87 if (fi
->fib_dead
== 0) {
88 printk(KERN_DEBUG
"DECnet: BUG! Attempt to free alive dn_fib_info\n");
96 } endfor_nexthops(fi
);
100 void dn_fib_release_info(struct dn_fib_info
*fi
)
102 spin_lock(&dn_fib_info_lock
);
103 if (fi
&& --fi
->fib_treeref
== 0) {
105 fi
->fib_next
->fib_prev
= fi
->fib_prev
;
107 fi
->fib_prev
->fib_next
= fi
->fib_next
;
108 if (fi
== dn_fib_info_list
)
109 dn_fib_info_list
= fi
->fib_next
;
113 spin_unlock(&dn_fib_info_lock
);
116 static inline int dn_fib_nh_comp(const struct dn_fib_info
*fi
, const struct dn_fib_info
*ofi
)
118 const struct dn_fib_nh
*onh
= ofi
->fib_nh
;
121 if (nh
->nh_oif
!= onh
->nh_oif
||
122 nh
->nh_gw
!= onh
->nh_gw
||
123 nh
->nh_scope
!= onh
->nh_scope
||
124 nh
->nh_weight
!= onh
->nh_weight
||
125 ((nh
->nh_flags
^onh
->nh_flags
)&~RTNH_F_DEAD
))
128 } endfor_nexthops(fi
);
132 static inline struct dn_fib_info
*dn_fib_find_info(const struct dn_fib_info
*nfi
)
135 if (fi
->fib_nhs
!= nfi
->fib_nhs
)
137 if (nfi
->fib_protocol
== fi
->fib_protocol
&&
138 nfi
->fib_prefsrc
== fi
->fib_prefsrc
&&
139 nfi
->fib_priority
== fi
->fib_priority
&&
140 memcmp(nfi
->fib_metrics
, fi
->fib_metrics
, sizeof(fi
->fib_metrics
)) == 0 &&
141 ((nfi
->fib_flags
^fi
->fib_flags
)&~RTNH_F_DEAD
) == 0 &&
142 (nfi
->fib_nhs
== 0 || dn_fib_nh_comp(fi
, nfi
) == 0))
148 static int dn_fib_count_nhs(const struct nlattr
*attr
)
150 struct rtnexthop
*nhp
= nla_data(attr
);
151 int nhs
= 0, nhlen
= nla_len(attr
);
153 while(nhlen
>= (int)sizeof(struct rtnexthop
)) {
154 if ((nhlen
-= nhp
->rtnh_len
) < 0)
157 nhp
= RTNH_NEXT(nhp
);
163 static int dn_fib_get_nhs(struct dn_fib_info
*fi
, const struct nlattr
*attr
,
164 const struct rtmsg
*r
)
166 struct rtnexthop
*nhp
= nla_data(attr
);
167 int nhlen
= nla_len(attr
);
169 change_nexthops(fi
) {
170 int attrlen
= nhlen
- sizeof(struct rtnexthop
);
171 if (attrlen
< 0 || (nhlen
-= nhp
->rtnh_len
) < 0)
174 nh
->nh_flags
= (r
->rtm_flags
&~0xFF) | nhp
->rtnh_flags
;
175 nh
->nh_oif
= nhp
->rtnh_ifindex
;
176 nh
->nh_weight
= nhp
->rtnh_hops
+ 1;
179 struct nlattr
*gw_attr
;
181 gw_attr
= nla_find((struct nlattr
*) (nhp
+ 1), attrlen
, RTA_GATEWAY
);
182 nh
->nh_gw
= gw_attr
? nla_get_le16(gw_attr
) : 0;
184 nhp
= RTNH_NEXT(nhp
);
185 } endfor_nexthops(fi
);
191 static int dn_fib_check_nh(const struct rtmsg
*r
, struct dn_fib_info
*fi
, struct dn_fib_nh
*nh
)
197 struct dn_fib_res res
;
199 if (nh
->nh_flags
&RTNH_F_ONLINK
) {
200 struct net_device
*dev
;
202 if (r
->rtm_scope
>= RT_SCOPE_LINK
)
204 if (dnet_addr_type(nh
->nh_gw
) != RTN_UNICAST
)
206 if ((dev
= __dev_get_by_index(&init_net
, nh
->nh_oif
)) == NULL
)
208 if (!(dev
->flags
&IFF_UP
))
212 nh
->nh_scope
= RT_SCOPE_LINK
;
216 memset(&fld
, 0, sizeof(fld
));
217 fld
.daddr
= nh
->nh_gw
;
218 fld
.flowidn_oif
= nh
->nh_oif
;
219 fld
.flowidn_scope
= r
->rtm_scope
+ 1;
221 if (fld
.flowidn_scope
< RT_SCOPE_LINK
)
222 fld
.flowidn_scope
= RT_SCOPE_LINK
;
224 if ((err
= dn_fib_lookup(&fld
, &res
)) != 0)
228 if (res
.type
!= RTN_UNICAST
&& res
.type
!= RTN_LOCAL
)
230 nh
->nh_scope
= res
.scope
;
231 nh
->nh_oif
= DN_FIB_RES_OIF(res
);
232 nh
->nh_dev
= DN_FIB_RES_DEV(res
);
233 if (nh
->nh_dev
== NULL
)
235 dev_hold(nh
->nh_dev
);
237 if (!(nh
->nh_dev
->flags
& IFF_UP
))
241 dn_fib_res_put(&res
);
244 struct net_device
*dev
;
246 if (nh
->nh_flags
&(RTNH_F_PERVASIVE
|RTNH_F_ONLINK
))
249 dev
= __dev_get_by_index(&init_net
, nh
->nh_oif
);
250 if (dev
== NULL
|| dev
->dn_ptr
== NULL
)
252 if (!(dev
->flags
&IFF_UP
))
255 dev_hold(nh
->nh_dev
);
256 nh
->nh_scope
= RT_SCOPE_HOST
;
263 struct dn_fib_info
*dn_fib_create_info(const struct rtmsg
*r
, struct nlattr
*attrs
[],
264 const struct nlmsghdr
*nlh
, int *errp
)
267 struct dn_fib_info
*fi
= NULL
;
268 struct dn_fib_info
*ofi
;
271 if (r
->rtm_type
> RTN_MAX
)
274 if (dn_fib_props
[r
->rtm_type
].scope
> r
->rtm_scope
)
277 if (attrs
[RTA_MULTIPATH
] &&
278 (nhs
= dn_fib_count_nhs(attrs
[RTA_MULTIPATH
])) == 0)
281 fi
= kzalloc(sizeof(*fi
)+nhs
*sizeof(struct dn_fib_nh
), GFP_KERNEL
);
286 fi
->fib_protocol
= r
->rtm_protocol
;
288 fi
->fib_flags
= r
->rtm_flags
;
290 if (attrs
[RTA_PRIORITY
])
291 fi
->fib_priority
= nla_get_u32(attrs
[RTA_PRIORITY
]);
293 if (attrs
[RTA_METRICS
]) {
297 nla_for_each_nested(attr
, attrs
[RTA_METRICS
], rem
) {
298 int type
= nla_type(attr
);
301 if (type
> RTAX_MAX
|| type
== RTAX_CC_ALGO
||
305 fi
->fib_metrics
[type
-1] = nla_get_u32(attr
);
310 if (attrs
[RTA_PREFSRC
])
311 fi
->fib_prefsrc
= nla_get_le16(attrs
[RTA_PREFSRC
]);
313 if (attrs
[RTA_MULTIPATH
]) {
314 if ((err
= dn_fib_get_nhs(fi
, attrs
[RTA_MULTIPATH
], r
)) != 0)
317 if (attrs
[RTA_OIF
] &&
318 fi
->fib_nh
->nh_oif
!= nla_get_u32(attrs
[RTA_OIF
]))
321 if (attrs
[RTA_GATEWAY
] &&
322 fi
->fib_nh
->nh_gw
!= nla_get_le16(attrs
[RTA_GATEWAY
]))
325 struct dn_fib_nh
*nh
= fi
->fib_nh
;
328 nh
->nh_oif
= nla_get_u32(attrs
[RTA_OIF
]);
330 if (attrs
[RTA_GATEWAY
])
331 nh
->nh_gw
= nla_get_le16(attrs
[RTA_GATEWAY
]);
333 nh
->nh_flags
= r
->rtm_flags
;
337 if (r
->rtm_type
== RTN_NAT
) {
338 if (!attrs
[RTA_GATEWAY
] || nhs
!= 1 || attrs
[RTA_OIF
])
341 fi
->fib_nh
->nh_gw
= nla_get_le16(attrs
[RTA_GATEWAY
]);
345 if (dn_fib_props
[r
->rtm_type
].error
) {
346 if (attrs
[RTA_GATEWAY
] || attrs
[RTA_OIF
] || attrs
[RTA_MULTIPATH
])
352 if (r
->rtm_scope
> RT_SCOPE_HOST
)
355 if (r
->rtm_scope
== RT_SCOPE_HOST
) {
356 struct dn_fib_nh
*nh
= fi
->fib_nh
;
358 /* Local address is added */
359 if (nhs
!= 1 || nh
->nh_gw
)
361 nh
->nh_scope
= RT_SCOPE_NOWHERE
;
362 nh
->nh_dev
= dev_get_by_index(&init_net
, fi
->fib_nh
->nh_oif
);
364 if (nh
->nh_dev
== NULL
)
367 change_nexthops(fi
) {
368 if ((err
= dn_fib_check_nh(r
, fi
, nh
)) != 0)
370 } endfor_nexthops(fi
)
373 if (fi
->fib_prefsrc
) {
374 if (r
->rtm_type
!= RTN_LOCAL
|| !attrs
[RTA_DST
] ||
375 fi
->fib_prefsrc
!= nla_get_le16(attrs
[RTA_DST
]))
376 if (dnet_addr_type(fi
->fib_prefsrc
) != RTN_LOCAL
)
381 if ((ofi
= dn_fib_find_info(fi
)) != NULL
) {
383 dn_fib_free_info(fi
);
389 atomic_inc(&fi
->fib_clntref
);
390 spin_lock(&dn_fib_info_lock
);
391 fi
->fib_next
= dn_fib_info_list
;
393 if (dn_fib_info_list
)
394 dn_fib_info_list
->fib_prev
= fi
;
395 dn_fib_info_list
= fi
;
396 spin_unlock(&dn_fib_info_lock
);
406 dn_fib_free_info(fi
);
412 int dn_fib_semantic_match(int type
, struct dn_fib_info
*fi
, const struct flowidn
*fld
, struct dn_fib_res
*res
)
414 int err
= dn_fib_props
[type
].error
;
417 if (fi
->fib_flags
& RTNH_F_DEAD
)
424 DN_FIB_RES_RESET(*res
);
425 atomic_inc(&fi
->fib_clntref
);
430 if (nh
->nh_flags
& RTNH_F_DEAD
)
432 if (!fld
->flowidn_oif
||
433 fld
->flowidn_oif
== nh
->nh_oif
)
436 if (nhsel
< fi
->fib_nhs
) {
438 atomic_inc(&fi
->fib_clntref
);
445 net_err_ratelimited("DECnet: impossible routing event : dn_fib_semantic_match type=%d\n",
454 void dn_fib_select_multipath(const struct flowidn
*fld
, struct dn_fib_res
*res
)
456 struct dn_fib_info
*fi
= res
->fi
;
459 spin_lock_bh(&dn_fib_multipath_lock
);
460 if (fi
->fib_power
<= 0) {
462 change_nexthops(fi
) {
463 if (!(nh
->nh_flags
&RTNH_F_DEAD
)) {
464 power
+= nh
->nh_weight
;
465 nh
->nh_power
= nh
->nh_weight
;
467 } endfor_nexthops(fi
);
468 fi
->fib_power
= power
;
470 spin_unlock_bh(&dn_fib_multipath_lock
);
476 w
= jiffies
% fi
->fib_power
;
478 change_nexthops(fi
) {
479 if (!(nh
->nh_flags
&RTNH_F_DEAD
) && nh
->nh_power
) {
480 if ((w
-= nh
->nh_power
) <= 0) {
484 spin_unlock_bh(&dn_fib_multipath_lock
);
488 } endfor_nexthops(fi
);
490 spin_unlock_bh(&dn_fib_multipath_lock
);
493 static inline u32
rtm_get_table(struct nlattr
*attrs
[], u8 table
)
495 if (attrs
[RTA_TABLE
])
496 table
= nla_get_u32(attrs
[RTA_TABLE
]);
501 static int dn_fib_rtm_delroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
503 struct net
*net
= sock_net(skb
->sk
);
504 struct dn_fib_table
*tb
;
505 struct rtmsg
*r
= nlmsg_data(nlh
);
506 struct nlattr
*attrs
[RTA_MAX
+1];
509 if (!netlink_capable(skb
, CAP_NET_ADMIN
))
512 if (!net_eq(net
, &init_net
))
515 err
= nlmsg_parse(nlh
, sizeof(*r
), attrs
, RTA_MAX
, rtm_dn_policy
);
519 tb
= dn_fib_get_table(rtm_get_table(attrs
, r
->rtm_table
), 0);
523 return tb
->delete(tb
, r
, attrs
, nlh
, &NETLINK_CB(skb
));
526 static int dn_fib_rtm_newroute(struct sk_buff
*skb
, struct nlmsghdr
*nlh
)
528 struct net
*net
= sock_net(skb
->sk
);
529 struct dn_fib_table
*tb
;
530 struct rtmsg
*r
= nlmsg_data(nlh
);
531 struct nlattr
*attrs
[RTA_MAX
+1];
534 if (!netlink_capable(skb
, CAP_NET_ADMIN
))
537 if (!net_eq(net
, &init_net
))
540 err
= nlmsg_parse(nlh
, sizeof(*r
), attrs
, RTA_MAX
, rtm_dn_policy
);
544 tb
= dn_fib_get_table(rtm_get_table(attrs
, r
->rtm_table
), 1);
548 return tb
->insert(tb
, r
, attrs
, nlh
, &NETLINK_CB(skb
));
551 static void fib_magic(int cmd
, int type
, __le16 dst
, int dst_len
, struct dn_ifaddr
*ifa
)
553 struct dn_fib_table
*tb
;
568 .prefsrc
= ifa
->ifa_local
,
574 .oif
= ifa
->ifa_dev
->dev
->ifindex
,
576 struct nlattr
*attrs
[RTA_MAX
+1] = {
577 [RTA_DST
] = (struct nlattr
*) &dst_attr
,
578 [RTA_PREFSRC
] = (struct nlattr
* ) &prefsrc_attr
,
579 [RTA_OIF
] = (struct nlattr
*) &oif_attr
,
582 memset(&req
.rtm
, 0, sizeof(req
.rtm
));
584 if (type
== RTN_UNICAST
)
585 tb
= dn_fib_get_table(RT_MIN_TABLE
, 1);
587 tb
= dn_fib_get_table(RT_TABLE_LOCAL
, 1);
592 req
.nlh
.nlmsg_len
= sizeof(req
);
593 req
.nlh
.nlmsg_type
= cmd
;
594 req
.nlh
.nlmsg_flags
= NLM_F_REQUEST
|NLM_F_CREATE
|NLM_F_APPEND
;
595 req
.nlh
.nlmsg_pid
= 0;
596 req
.nlh
.nlmsg_seq
= 0;
598 req
.rtm
.rtm_dst_len
= dst_len
;
599 req
.rtm
.rtm_table
= tb
->n
;
600 req
.rtm
.rtm_protocol
= RTPROT_KERNEL
;
601 req
.rtm
.rtm_scope
= (type
!= RTN_LOCAL
? RT_SCOPE_LINK
: RT_SCOPE_HOST
);
602 req
.rtm
.rtm_type
= type
;
604 if (cmd
== RTM_NEWROUTE
)
605 tb
->insert(tb
, &req
.rtm
, attrs
, &req
.nlh
, NULL
);
607 tb
->delete(tb
, &req
.rtm
, attrs
, &req
.nlh
, NULL
);
610 static void dn_fib_add_ifaddr(struct dn_ifaddr
*ifa
)
613 fib_magic(RTM_NEWROUTE
, RTN_LOCAL
, ifa
->ifa_local
, 16, ifa
);
616 if (!(dev
->flags
&IFF_UP
))
618 /* In the future, we will want to add default routes here */
623 static void dn_fib_del_ifaddr(struct dn_ifaddr
*ifa
)
626 struct net_device
*dev
;
627 struct dn_dev
*dn_db
;
628 struct dn_ifaddr
*ifa2
;
632 /* Scan device list */
634 for_each_netdev_rcu(&init_net
, dev
) {
635 dn_db
= rcu_dereference(dev
->dn_ptr
);
638 for (ifa2
= rcu_dereference(dn_db
->ifa_list
);
640 ifa2
= rcu_dereference(ifa2
->ifa_next
)) {
641 if (ifa2
->ifa_local
== ifa
->ifa_local
) {
650 fib_magic(RTM_DELROUTE
, RTN_LOCAL
, ifa
->ifa_local
, 16, ifa
);
652 if (dnet_addr_type(ifa
->ifa_local
) != RTN_LOCAL
) {
653 if (dn_fib_sync_down(ifa
->ifa_local
, NULL
, 0))
659 static void dn_fib_disable_addr(struct net_device
*dev
, int force
)
661 if (dn_fib_sync_down(0, dev
, force
))
663 dn_rt_cache_flush(0);
664 neigh_ifdown(&dn_neigh_table
, dev
);
667 static int dn_fib_dnaddr_event(struct notifier_block
*this, unsigned long event
, void *ptr
)
669 struct dn_ifaddr
*ifa
= (struct dn_ifaddr
*)ptr
;
673 dn_fib_add_ifaddr(ifa
);
674 dn_fib_sync_up(ifa
->ifa_dev
->dev
);
675 dn_rt_cache_flush(-1);
678 dn_fib_del_ifaddr(ifa
);
679 if (ifa
->ifa_dev
&& ifa
->ifa_dev
->ifa_list
== NULL
) {
680 dn_fib_disable_addr(ifa
->ifa_dev
->dev
, 1);
682 dn_rt_cache_flush(-1);
689 static int dn_fib_sync_down(__le16 local
, struct net_device
*dev
, int force
)
692 int scope
= RT_SCOPE_NOWHERE
;
699 * This makes no sense for DECnet.... we will almost
700 * certainly have more than one local address the same
701 * over all our interfaces. It needs thinking about
704 if (local
&& fi
->fib_prefsrc
== local
) {
705 fi
->fib_flags
|= RTNH_F_DEAD
;
707 } else if (dev
&& fi
->fib_nhs
) {
710 change_nexthops(fi
) {
711 if (nh
->nh_flags
&RTNH_F_DEAD
)
713 else if (nh
->nh_dev
== dev
&&
714 nh
->nh_scope
!= scope
) {
715 spin_lock_bh(&dn_fib_multipath_lock
);
716 nh
->nh_flags
|= RTNH_F_DEAD
;
717 fi
->fib_power
-= nh
->nh_power
;
719 spin_unlock_bh(&dn_fib_multipath_lock
);
722 } endfor_nexthops(fi
)
723 if (dead
== fi
->fib_nhs
) {
724 fi
->fib_flags
|= RTNH_F_DEAD
;
733 static int dn_fib_sync_up(struct net_device
*dev
)
737 if (!(dev
->flags
&IFF_UP
))
743 change_nexthops(fi
) {
744 if (!(nh
->nh_flags
&RTNH_F_DEAD
)) {
748 if (nh
->nh_dev
== NULL
|| !(nh
->nh_dev
->flags
&IFF_UP
))
750 if (nh
->nh_dev
!= dev
|| dev
->dn_ptr
== NULL
)
753 spin_lock_bh(&dn_fib_multipath_lock
);
755 nh
->nh_flags
&= ~RTNH_F_DEAD
;
756 spin_unlock_bh(&dn_fib_multipath_lock
);
757 } endfor_nexthops(fi
);
760 fi
->fib_flags
&= ~RTNH_F_DEAD
;
767 static struct notifier_block dn_fib_dnaddr_notifier
= {
768 .notifier_call
= dn_fib_dnaddr_event
,
771 void __exit
dn_fib_cleanup(void)
773 dn_fib_table_cleanup();
774 dn_fib_rules_cleanup();
776 unregister_dnaddr_notifier(&dn_fib_dnaddr_notifier
);
780 void __init
dn_fib_init(void)
785 register_dnaddr_notifier(&dn_fib_dnaddr_notifier
);
787 rtnl_register(PF_DECnet
, RTM_NEWROUTE
, dn_fib_rtm_newroute
, NULL
, NULL
);
788 rtnl_register(PF_DECnet
, RTM_DELROUTE
, dn_fib_rtm_delroute
, NULL
, NULL
);