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 (Routing Tables)
8 * Author: Steve Whitehouse <SteveW@ACM.org>
9 * Mostly copied from the IPv4 routing code
15 #include <linux/string.h>
16 #include <linux/net.h>
17 #include <linux/socket.h>
18 #include <linux/sockios.h>
19 #include <linux/init.h>
20 #include <linux/skbuff.h>
21 #include <linux/netlink.h>
22 #include <linux/rtnetlink.h>
23 #include <linux/proc_fs.h>
24 #include <linux/netdevice.h>
25 #include <linux/timer.h>
26 #include <linux/spinlock.h>
27 #include <asm/atomic.h>
28 #include <asm/uaccess.h>
29 #include <linux/route.h> /* RTF_xxx */
30 #include <net/neighbour.h>
34 #include <net/dn_route.h>
35 #include <net/dn_fib.h>
36 #include <net/dn_neigh.h>
37 #include <net/dn_dev.h>
41 struct dn_zone
*dz_next
;
42 struct dn_fib_node
**dz_hash
;
46 #define DZ_HASHMASK(dz) ((dz)->dz_hashmask)
49 #define DZ_MASK(dz) ((dz)->dz_mask)
54 struct dn_zone
*dh_zones
[17];
55 struct dn_zone
*dh_zone_list
;
58 #define dz_key_0(key) ((key).datum = 0)
59 #define dz_prefix(key,dz) ((key).datum)
61 #define for_nexthops(fi) { int nhsel; const struct dn_fib_nh *nh;\
62 for(nhsel = 0, nh = (fi)->fib_nh; nhsel < (fi)->fib_nhs; nh++, nhsel++)
64 #define endfor_nexthops(fi) }
66 #define DN_MAX_DIVISOR 1024
68 #define DN_S_ACCESSED 2
70 #define DN_FIB_SCAN(f, fp) \
71 for( ; ((f) = *(fp)) != NULL; (fp) = &(f)->fn_next)
73 #define DN_FIB_SCAN_KEY(f, fp, key) \
74 for( ; ((f) = *(fp)) != NULL && dn_key_eq((f)->fn_key, (key)); (fp) = &(f)->fn_next)
76 #define RT_TABLE_MIN 1
78 static DEFINE_RWLOCK(dn_fib_tables_lock
);
79 struct dn_fib_table
*dn_fib_tables
[RT_TABLE_MAX
+ 1];
81 static kmem_cache_t
*dn_hash_kmem __read_mostly
;
82 static int dn_fib_hash_zombies
;
84 static inline dn_fib_idx_t
dn_hash(dn_fib_key_t key
, struct dn_zone
*dz
)
86 u16 h
= dn_ntohs(key
.datum
)>>(16 - dz
->dz_order
);
90 return *(dn_fib_idx_t
*)&h
;
93 static inline dn_fib_key_t
dz_key(__le16 dst
, struct dn_zone
*dz
)
96 k
.datum
= dst
& DZ_MASK(dz
);
100 static inline struct dn_fib_node
**dn_chain_p(dn_fib_key_t key
, struct dn_zone
*dz
)
102 return &dz
->dz_hash
[dn_hash(key
, dz
).datum
];
105 static inline struct dn_fib_node
*dz_chain(dn_fib_key_t key
, struct dn_zone
*dz
)
107 return dz
->dz_hash
[dn_hash(key
, dz
).datum
];
110 static inline int dn_key_eq(dn_fib_key_t a
, dn_fib_key_t b
)
112 return a
.datum
== b
.datum
;
115 static inline int dn_key_leq(dn_fib_key_t a
, dn_fib_key_t b
)
117 return a
.datum
<= b
.datum
;
120 static inline void dn_rebuild_zone(struct dn_zone
*dz
,
121 struct dn_fib_node
**old_ht
,
125 struct dn_fib_node
*f
, **fp
, *next
;
127 for(i
= 0; i
< old_divisor
; i
++) {
128 for(f
= old_ht
[i
]; f
; f
= f
->fn_next
) {
130 for(fp
= dn_chain_p(f
->fn_key
, dz
);
131 *fp
&& dn_key_leq((*fp
)->fn_key
, f
->fn_key
);
132 fp
= &(*fp
)->fn_next
)
140 static void dn_rehash_zone(struct dn_zone
*dz
)
142 struct dn_fib_node
**ht
, **old_ht
;
143 int old_divisor
, new_divisor
;
146 old_divisor
= dz
->dz_divisor
;
148 switch(old_divisor
) {
154 printk(KERN_DEBUG
"DECnet: dn_rehash_zone: BUG! %d\n", old_divisor
);
157 new_hashmask
= 0x3FF;
161 ht
= kcalloc(new_divisor
, sizeof(struct dn_fib_node
*), GFP_KERNEL
);
165 write_lock_bh(&dn_fib_tables_lock
);
166 old_ht
= dz
->dz_hash
;
168 dz
->dz_hashmask
= new_hashmask
;
169 dz
->dz_divisor
= new_divisor
;
170 dn_rebuild_zone(dz
, old_ht
, old_divisor
);
171 write_unlock_bh(&dn_fib_tables_lock
);
175 static void dn_free_node(struct dn_fib_node
*f
)
177 dn_fib_release_info(DN_FIB_INFO(f
));
178 kmem_cache_free(dn_hash_kmem
, f
);
182 static struct dn_zone
*dn_new_zone(struct dn_hash
*table
, int z
)
185 struct dn_zone
*dz
= kzalloc(sizeof(struct dn_zone
), GFP_KERNEL
);
191 dz
->dz_hashmask
= 0x0F;
197 dz
->dz_hash
= kcalloc(dz
->dz_divisor
, sizeof(struct dn_fib_node
*), GFP_KERNEL
);
204 dz
->dz_mask
= dnet_make_mask(z
);
206 for(i
= z
+ 1; i
<= 16; i
++)
207 if (table
->dh_zones
[i
])
210 write_lock_bh(&dn_fib_tables_lock
);
212 dz
->dz_next
= table
->dh_zone_list
;
213 table
->dh_zone_list
= dz
;
215 dz
->dz_next
= table
->dh_zones
[i
]->dz_next
;
216 table
->dh_zones
[i
]->dz_next
= dz
;
218 table
->dh_zones
[z
] = dz
;
219 write_unlock_bh(&dn_fib_tables_lock
);
224 static int dn_fib_nh_match(struct rtmsg
*r
, struct nlmsghdr
*nlh
, struct dn_kern_rta
*rta
, struct dn_fib_info
*fi
)
226 struct rtnexthop
*nhp
;
229 if (rta
->rta_priority
&& *rta
->rta_priority
!= fi
->fib_priority
)
232 if (rta
->rta_oif
|| rta
->rta_gw
) {
233 if ((!rta
->rta_oif
|| *rta
->rta_oif
== fi
->fib_nh
->nh_oif
) &&
234 (!rta
->rta_gw
|| memcmp(rta
->rta_gw
, &fi
->fib_nh
->nh_gw
, 2) == 0))
239 if (rta
->rta_mp
== NULL
)
242 nhp
= RTA_DATA(rta
->rta_mp
);
243 nhlen
= RTA_PAYLOAD(rta
->rta_mp
);
246 int attrlen
= nhlen
- sizeof(struct rtnexthop
);
249 if (attrlen
< 0 || (nhlen
-= nhp
->rtnh_len
) < 0)
251 if (nhp
->rtnh_ifindex
&& nhp
->rtnh_ifindex
!= nh
->nh_oif
)
254 gw
= dn_fib_get_attr16(RTNH_DATA(nhp
), attrlen
, RTA_GATEWAY
);
256 if (gw
&& gw
!= nh
->nh_gw
)
259 nhp
= RTNH_NEXT(nhp
);
260 } endfor_nexthops(fi
);
265 static int dn_fib_dump_info(struct sk_buff
*skb
, u32 pid
, u32 seq
, int event
,
266 u8 tb_id
, u8 type
, u8 scope
, void *dst
, int dst_len
,
267 struct dn_fib_info
*fi
, unsigned int flags
)
270 struct nlmsghdr
*nlh
;
271 unsigned char *b
= skb
->tail
;
273 nlh
= NLMSG_NEW(skb
, pid
, seq
, event
, sizeof(*rtm
), flags
);
274 rtm
= NLMSG_DATA(nlh
);
275 rtm
->rtm_family
= AF_DECnet
;
276 rtm
->rtm_dst_len
= dst_len
;
277 rtm
->rtm_src_len
= 0;
279 rtm
->rtm_table
= tb_id
;
280 rtm
->rtm_flags
= fi
->fib_flags
;
281 rtm
->rtm_scope
= scope
;
282 rtm
->rtm_type
= type
;
283 if (rtm
->rtm_dst_len
)
284 RTA_PUT(skb
, RTA_DST
, 2, dst
);
285 rtm
->rtm_protocol
= fi
->fib_protocol
;
286 if (fi
->fib_priority
)
287 RTA_PUT(skb
, RTA_PRIORITY
, 4, &fi
->fib_priority
);
288 if (rtnetlink_put_metrics(skb
, fi
->fib_metrics
) < 0)
290 if (fi
->fib_nhs
== 1) {
291 if (fi
->fib_nh
->nh_gw
)
292 RTA_PUT(skb
, RTA_GATEWAY
, 2, &fi
->fib_nh
->nh_gw
);
293 if (fi
->fib_nh
->nh_oif
)
294 RTA_PUT(skb
, RTA_OIF
, sizeof(int), &fi
->fib_nh
->nh_oif
);
296 if (fi
->fib_nhs
> 1) {
297 struct rtnexthop
*nhp
;
298 struct rtattr
*mp_head
;
299 if (skb_tailroom(skb
) <= RTA_SPACE(0))
301 mp_head
= (struct rtattr
*)skb_put(skb
, RTA_SPACE(0));
304 if (skb_tailroom(skb
) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp
)) + 4))
306 nhp
= (struct rtnexthop
*)skb_put(skb
, RTA_ALIGN(sizeof(*nhp
)));
307 nhp
->rtnh_flags
= nh
->nh_flags
& 0xFF;
308 nhp
->rtnh_hops
= nh
->nh_weight
- 1;
309 nhp
->rtnh_ifindex
= nh
->nh_oif
;
311 RTA_PUT(skb
, RTA_GATEWAY
, 2, &nh
->nh_gw
);
312 nhp
->rtnh_len
= skb
->tail
- (unsigned char *)nhp
;
313 } endfor_nexthops(fi
);
314 mp_head
->rta_type
= RTA_MULTIPATH
;
315 mp_head
->rta_len
= skb
->tail
- (u8
*)mp_head
;
318 nlh
->nlmsg_len
= skb
->tail
- b
;
324 skb_trim(skb
, b
- skb
->data
);
329 static void dn_rtmsg_fib(int event
, struct dn_fib_node
*f
, int z
, int tb_id
,
330 struct nlmsghdr
*nlh
, struct netlink_skb_parms
*req
)
333 u32 pid
= req
? req
->pid
: 0;
334 int size
= NLMSG_SPACE(sizeof(struct rtmsg
) + 256);
336 skb
= alloc_skb(size
, GFP_KERNEL
);
340 if (dn_fib_dump_info(skb
, pid
, nlh
->nlmsg_seq
, event
, tb_id
,
341 f
->fn_type
, f
->fn_scope
, &f
->fn_key
, z
,
342 DN_FIB_INFO(f
), 0) < 0) {
346 NETLINK_CB(skb
).dst_group
= RTNLGRP_DECnet_ROUTE
;
347 if (nlh
->nlmsg_flags
& NLM_F_ECHO
)
348 atomic_inc(&skb
->users
);
349 netlink_broadcast(rtnl
, skb
, pid
, RTNLGRP_DECnet_ROUTE
, GFP_KERNEL
);
350 if (nlh
->nlmsg_flags
& NLM_F_ECHO
)
351 netlink_unicast(rtnl
, skb
, pid
, MSG_DONTWAIT
);
354 static __inline__
int dn_hash_dump_bucket(struct sk_buff
*skb
,
355 struct netlink_callback
*cb
,
356 struct dn_fib_table
*tb
,
358 struct dn_fib_node
*f
)
363 for(i
= 0; f
; i
++, f
= f
->fn_next
) {
366 if (f
->fn_state
& DN_S_ZOMBIE
)
368 if (dn_fib_dump_info(skb
, NETLINK_CB(cb
->skb
).pid
,
372 (f
->fn_state
& DN_S_ZOMBIE
) ? 0 : f
->fn_type
,
373 f
->fn_scope
, &f
->fn_key
, dz
->dz_order
,
374 f
->fn_info
, NLM_F_MULTI
) < 0) {
383 static __inline__
int dn_hash_dump_zone(struct sk_buff
*skb
,
384 struct netlink_callback
*cb
,
385 struct dn_fib_table
*tb
,
391 for(h
= 0; h
< dz
->dz_divisor
; h
++) {
395 memset(&cb
->args
[3], 0, sizeof(cb
->args
) - 3*sizeof(cb
->args
[0]));
396 if (dz
->dz_hash
== NULL
|| dz
->dz_hash
[h
] == NULL
)
398 if (dn_hash_dump_bucket(skb
, cb
, tb
, dz
, dz
->dz_hash
[h
]) < 0) {
407 static int dn_fib_table_dump(struct dn_fib_table
*tb
, struct sk_buff
*skb
,
408 struct netlink_callback
*cb
)
412 struct dn_hash
*table
= (struct dn_hash
*)tb
->data
;
415 read_lock(&dn_fib_tables_lock
);
416 for(dz
= table
->dh_zone_list
, m
= 0; dz
; dz
= dz
->dz_next
, m
++) {
420 memset(&cb
->args
[2], 0, sizeof(cb
->args
) - 2*sizeof(cb
->args
[0]));
422 if (dn_hash_dump_zone(skb
, cb
, tb
, dz
) < 0) {
424 read_unlock(&dn_fib_tables_lock
);
428 read_unlock(&dn_fib_tables_lock
);
434 static int dn_fib_table_insert(struct dn_fib_table
*tb
, struct rtmsg
*r
, struct dn_kern_rta
*rta
, struct nlmsghdr
*n
, struct netlink_skb_parms
*req
)
436 struct dn_hash
*table
= (struct dn_hash
*)tb
->data
;
437 struct dn_fib_node
*new_f
, *f
, **fp
, **del_fp
;
439 struct dn_fib_info
*fi
;
440 int z
= r
->rtm_dst_len
;
441 int type
= r
->rtm_type
;
448 dz
= table
->dh_zones
[z
];
449 if (!dz
&& !(dz
= dn_new_zone(table
, z
)))
455 memcpy(&dst
, rta
->rta_dst
, 2);
456 if (dst
& ~DZ_MASK(dz
))
458 key
= dz_key(dst
, dz
);
461 if ((fi
= dn_fib_create_info(r
, rta
, n
, &err
)) == NULL
)
464 if (dz
->dz_nent
> (dz
->dz_divisor
<< 2) &&
465 dz
->dz_divisor
> DN_MAX_DIVISOR
&&
466 (z
==16 || (1<<z
) > dz
->dz_divisor
))
469 fp
= dn_chain_p(key
, dz
);
472 if (dn_key_leq(key
, f
->fn_key
))
478 if (f
&& (f
->fn_state
& DN_S_ZOMBIE
) &&
479 dn_key_eq(f
->fn_key
, key
)) {
486 DN_FIB_SCAN_KEY(f
, fp
, key
) {
487 if (fi
->fib_priority
<= DN_FIB_INFO(f
)->fib_priority
)
491 if (f
&& dn_key_eq(f
->fn_key
, key
) &&
492 fi
->fib_priority
== DN_FIB_INFO(f
)->fib_priority
) {
493 struct dn_fib_node
**ins_fp
;
496 if (n
->nlmsg_flags
& NLM_F_EXCL
)
499 if (n
->nlmsg_flags
& NLM_F_REPLACE
) {
509 DN_FIB_SCAN_KEY(f
, fp
, key
) {
510 if (fi
->fib_priority
!= DN_FIB_INFO(f
)->fib_priority
)
512 if (f
->fn_type
== type
&& f
->fn_scope
== r
->rtm_scope
513 && DN_FIB_INFO(f
) == fi
)
517 if (!(n
->nlmsg_flags
& NLM_F_APPEND
)) {
525 if (!(n
->nlmsg_flags
& NLM_F_CREATE
))
530 new_f
= kmem_cache_alloc(dn_hash_kmem
, SLAB_KERNEL
);
534 memset(new_f
, 0, sizeof(struct dn_fib_node
));
537 new_f
->fn_type
= type
;
538 new_f
->fn_scope
= r
->rtm_scope
;
539 DN_FIB_INFO(new_f
) = fi
;
542 write_lock_bh(&dn_fib_tables_lock
);
544 write_unlock_bh(&dn_fib_tables_lock
);
549 write_lock_bh(&dn_fib_tables_lock
);
550 *del_fp
= f
->fn_next
;
551 write_unlock_bh(&dn_fib_tables_lock
);
553 if (!(f
->fn_state
& DN_S_ZOMBIE
))
554 dn_rtmsg_fib(RTM_DELROUTE
, f
, z
, tb
->n
, n
, req
);
555 if (f
->fn_state
& DN_S_ACCESSED
)
556 dn_rt_cache_flush(-1);
560 dn_rt_cache_flush(-1);
563 dn_rtmsg_fib(RTM_NEWROUTE
, new_f
, z
, tb
->n
, n
, req
);
567 dn_fib_release_info(fi
);
572 static int dn_fib_table_delete(struct dn_fib_table
*tb
, struct rtmsg
*r
, struct dn_kern_rta
*rta
, struct nlmsghdr
*n
, struct netlink_skb_parms
*req
)
574 struct dn_hash
*table
= (struct dn_hash
*)tb
->data
;
575 struct dn_fib_node
**fp
, **del_fp
, *f
;
576 int z
= r
->rtm_dst_len
;
585 if ((dz
= table
->dh_zones
[z
]) == NULL
)
591 memcpy(&dst
, rta
->rta_dst
, 2);
592 if (dst
& ~DZ_MASK(dz
))
594 key
= dz_key(dst
, dz
);
597 fp
= dn_chain_p(key
, dz
);
600 if (dn_key_eq(f
->fn_key
, key
))
602 if (dn_key_leq(key
, f
->fn_key
))
608 DN_FIB_SCAN_KEY(f
, fp
, key
) {
609 struct dn_fib_info
*fi
= DN_FIB_INFO(f
);
611 if (f
->fn_state
& DN_S_ZOMBIE
)
616 if (del_fp
== NULL
&&
617 (!r
->rtm_type
|| f
->fn_type
== r
->rtm_type
) &&
618 (r
->rtm_scope
== RT_SCOPE_NOWHERE
|| f
->fn_scope
== r
->rtm_scope
) &&
620 fi
->fib_protocol
== r
->rtm_protocol
) &&
621 dn_fib_nh_match(r
, n
, rta
, fi
) == 0)
627 dn_rtmsg_fib(RTM_DELROUTE
, f
, z
, tb
->n
, n
, req
);
630 write_lock_bh(&dn_fib_tables_lock
);
631 *del_fp
= f
->fn_next
;
632 write_unlock_bh(&dn_fib_tables_lock
);
634 if (f
->fn_state
& DN_S_ACCESSED
)
635 dn_rt_cache_flush(-1);
639 f
->fn_state
|= DN_S_ZOMBIE
;
640 if (f
->fn_state
& DN_S_ACCESSED
) {
641 f
->fn_state
&= ~DN_S_ACCESSED
;
642 dn_rt_cache_flush(-1);
644 if (++dn_fib_hash_zombies
> 128)
654 static inline int dn_flush_list(struct dn_fib_node
**fp
, int z
, struct dn_hash
*table
)
657 struct dn_fib_node
*f
;
659 while((f
= *fp
) != NULL
) {
660 struct dn_fib_info
*fi
= DN_FIB_INFO(f
);
662 if (fi
&& ((f
->fn_state
& DN_S_ZOMBIE
) || (fi
->fib_flags
& RTNH_F_DEAD
))) {
663 write_lock_bh(&dn_fib_tables_lock
);
665 write_unlock_bh(&dn_fib_tables_lock
);
677 static int dn_fib_table_flush(struct dn_fib_table
*tb
)
679 struct dn_hash
*table
= (struct dn_hash
*)tb
->data
;
683 dn_fib_hash_zombies
= 0;
684 for(dz
= table
->dh_zone_list
; dz
; dz
= dz
->dz_next
) {
687 for(i
= dz
->dz_divisor
-1; i
>= 0; i
--)
688 tmp
+= dn_flush_list(&dz
->dz_hash
[i
], dz
->dz_order
, table
);
696 static int dn_fib_table_lookup(struct dn_fib_table
*tb
, const struct flowi
*flp
, struct dn_fib_res
*res
)
700 struct dn_hash
*t
= (struct dn_hash
*)tb
->data
;
702 read_lock(&dn_fib_tables_lock
);
703 for(dz
= t
->dh_zone_list
; dz
; dz
= dz
->dz_next
) {
704 struct dn_fib_node
*f
;
705 dn_fib_key_t k
= dz_key(flp
->fld_dst
, dz
);
707 for(f
= dz_chain(k
, dz
); f
; f
= f
->fn_next
) {
708 if (!dn_key_eq(k
, f
->fn_key
)) {
709 if (dn_key_leq(k
, f
->fn_key
))
715 f
->fn_state
|= DN_S_ACCESSED
;
717 if (f
->fn_state
&DN_S_ZOMBIE
)
720 if (f
->fn_scope
< flp
->fld_scope
)
723 err
= dn_fib_semantic_match(f
->fn_type
, DN_FIB_INFO(f
), flp
, res
);
726 res
->type
= f
->fn_type
;
727 res
->scope
= f
->fn_scope
;
728 res
->prefixlen
= dz
->dz_order
;
737 read_unlock(&dn_fib_tables_lock
);
742 struct dn_fib_table
*dn_fib_get_table(int n
, int create
)
744 struct dn_fib_table
*t
;
746 if (n
< RT_TABLE_MIN
)
749 if (n
> RT_TABLE_MAX
)
752 if (dn_fib_tables
[n
])
753 return dn_fib_tables
[n
];
758 if (in_interrupt() && net_ratelimit()) {
759 printk(KERN_DEBUG
"DECnet: BUG! Attempt to create routing table from interrupt\n");
762 if ((t
= kmalloc(sizeof(struct dn_fib_table
) + sizeof(struct dn_hash
), GFP_KERNEL
)) == NULL
)
765 memset(t
, 0, sizeof(struct dn_fib_table
));
768 t
->insert
= dn_fib_table_insert
;
769 t
->delete = dn_fib_table_delete
;
770 t
->lookup
= dn_fib_table_lookup
;
771 t
->flush
= dn_fib_table_flush
;
772 t
->dump
= dn_fib_table_dump
;
773 memset(t
->data
, 0, sizeof(struct dn_hash
));
774 dn_fib_tables
[n
] = t
;
779 static void dn_fib_del_tree(int n
)
781 struct dn_fib_table
*t
;
783 write_lock(&dn_fib_tables_lock
);
784 t
= dn_fib_tables
[n
];
785 dn_fib_tables
[n
] = NULL
;
786 write_unlock(&dn_fib_tables_lock
);
791 struct dn_fib_table
*dn_fib_empty_table(void)
795 for(id
= RT_TABLE_MIN
; id
<= RT_TABLE_MAX
; id
++)
796 if (dn_fib_tables
[id
] == NULL
)
797 return dn_fib_get_table(id
, 1);
801 void __init
dn_fib_table_init(void)
803 dn_hash_kmem
= kmem_cache_create("dn_fib_info_cache",
804 sizeof(struct dn_fib_info
),
805 0, SLAB_HWCACHE_ALIGN
,
809 void __exit
dn_fib_table_cleanup(void)
813 for (i
= RT_TABLE_MIN
; i
<= RT_TABLE_MAX
; ++i
)