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/slab.h>
19 #include <linux/sockios.h>
20 #include <linux/init.h>
21 #include <linux/skbuff.h>
22 #include <linux/netlink.h>
23 #include <linux/rtnetlink.h>
24 #include <linux/proc_fs.h>
25 #include <linux/netdevice.h>
26 #include <linux/timer.h>
27 #include <linux/spinlock.h>
28 #include <linux/atomic.h>
29 #include <asm/uaccess.h>
30 #include <linux/route.h> /* RTF_xxx */
31 #include <net/neighbour.h>
32 #include <net/netlink.h>
35 #include <net/fib_rules.h>
37 #include <net/dn_route.h>
38 #include <net/dn_fib.h>
39 #include <net/dn_neigh.h>
40 #include <net/dn_dev.h>
44 struct dn_zone *dz_next;
45 struct dn_fib_node **dz_hash;
49 #define DZ_HASHMASK(dz) ((dz)->dz_hashmask)
52 #define DZ_MASK(dz) ((dz)->dz_mask)
57 struct dn_zone *dh_zones[17];
58 struct dn_zone *dh_zone_list;
61 #define dz_key_0(key) ((key).datum = 0)
63 #define for_nexthops(fi) { int nhsel; const struct dn_fib_nh *nh;\
64 for(nhsel = 0, nh = (fi)->fib_nh; nhsel < (fi)->fib_nhs; nh++, nhsel++)
66 #define endfor_nexthops(fi) }
68 #define DN_MAX_DIVISOR 1024
70 #define DN_S_ACCESSED 2
72 #define DN_FIB_SCAN(f, fp) \
73 for( ; ((f) = *(fp)) != NULL; (fp) = &(f)->fn_next)
75 #define DN_FIB_SCAN_KEY(f, fp, key) \
76 for( ; ((f) = *(fp)) != NULL && dn_key_eq((f)->fn_key, (key)); (fp) = &(f)->fn_next)
78 #define RT_TABLE_MIN 1
79 #define DN_FIB_TABLE_HASHSZ 256
80 static struct hlist_head dn_fib_table_hash[DN_FIB_TABLE_HASHSZ];
81 static DEFINE_RWLOCK(dn_fib_tables_lock);
83 static struct kmem_cache *dn_hash_kmem __read_mostly;
84 static int dn_fib_hash_zombies;
86 static inline dn_fib_idx_t dn_hash(dn_fib_key_t key, struct dn_zone *dz)
88 u16 h = le16_to_cpu(key.datum)>>(16 - dz->dz_order);
92 return *(dn_fib_idx_t *)&h;
95 static inline dn_fib_key_t dz_key(__le16 dst, struct dn_zone *dz)
98 k.datum = dst & DZ_MASK(dz);
102 static inline struct dn_fib_node **dn_chain_p(dn_fib_key_t key, struct dn_zone *dz)
104 return &dz->dz_hash[dn_hash(key, dz).datum];
107 static inline struct dn_fib_node *dz_chain(dn_fib_key_t key, struct dn_zone *dz)
109 return dz->dz_hash[dn_hash(key, dz).datum];
112 static inline int dn_key_eq(dn_fib_key_t a, dn_fib_key_t b)
114 return a.datum == b.datum;
117 static inline int dn_key_leq(dn_fib_key_t a, dn_fib_key_t b)
119 return a.datum <= b.datum;
122 static inline void dn_rebuild_zone(struct dn_zone *dz,
123 struct dn_fib_node **old_ht,
126 struct dn_fib_node *f, **fp, *next;
129 for(i = 0; i < old_divisor; i++) {
130 for(f = old_ht[i]; f; f = next) {
132 for(fp = dn_chain_p(f->fn_key, dz);
133 *fp && dn_key_leq((*fp)->fn_key, f->fn_key);
134 fp = &(*fp)->fn_next)
142 static void dn_rehash_zone(struct dn_zone *dz)
144 struct dn_fib_node **ht, **old_ht;
145 int old_divisor, new_divisor;
148 old_divisor = dz->dz_divisor;
150 switch (old_divisor) {
156 printk(KERN_DEBUG "DECnet: dn_rehash_zone: BUG! %d\n",
160 new_hashmask = 0x3FF;
164 ht = kcalloc(new_divisor, sizeof(struct dn_fib_node*), GFP_KERNEL);
168 write_lock_bh(&dn_fib_tables_lock);
169 old_ht = dz->dz_hash;
171 dz->dz_hashmask = new_hashmask;
172 dz->dz_divisor = new_divisor;
173 dn_rebuild_zone(dz, old_ht, old_divisor);
174 write_unlock_bh(&dn_fib_tables_lock);
178 static void dn_free_node(struct dn_fib_node *f)
180 dn_fib_release_info(DN_FIB_INFO(f));
181 kmem_cache_free(dn_hash_kmem, f);
185 static struct dn_zone *dn_new_zone(struct dn_hash *table, int z)
188 struct dn_zone *dz = kzalloc(sizeof(struct dn_zone), GFP_KERNEL);
194 dz->dz_hashmask = 0x0F;
200 dz->dz_hash = kcalloc(dz->dz_divisor, sizeof(struct dn_fib_node *), GFP_KERNEL);
207 dz->dz_mask = dnet_make_mask(z);
209 for(i = z + 1; i <= 16; i++)
210 if (table->dh_zones[i])
213 write_lock_bh(&dn_fib_tables_lock);
215 dz->dz_next = table->dh_zone_list;
216 table->dh_zone_list = dz;
218 dz->dz_next = table->dh_zones[i]->dz_next;
219 table->dh_zones[i]->dz_next = dz;
221 table->dh_zones[z] = dz;
222 write_unlock_bh(&dn_fib_tables_lock);
227 static int dn_fib_nh_match(struct rtmsg *r, struct nlmsghdr *nlh, struct dn_kern_rta *rta, struct dn_fib_info *fi)
229 struct rtnexthop *nhp;
232 if (rta->rta_priority && *rta->rta_priority != fi->fib_priority)
235 if (rta->rta_oif || rta->rta_gw) {
236 if ((!rta->rta_oif || *rta->rta_oif == fi->fib_nh->nh_oif) &&
237 (!rta->rta_gw || memcmp(rta->rta_gw, &fi->fib_nh->nh_gw, 2) == 0))
242 if (rta->rta_mp == NULL)
245 nhp = RTA_DATA(rta->rta_mp);
246 nhlen = RTA_PAYLOAD(rta->rta_mp);
249 int attrlen = nhlen - sizeof(struct rtnexthop);
252 if (attrlen < 0 || (nhlen -= nhp->rtnh_len) < 0)
254 if (nhp->rtnh_ifindex && nhp->rtnh_ifindex != nh->nh_oif)
257 gw = dn_fib_get_attr16(RTNH_DATA(nhp), attrlen, RTA_GATEWAY);
259 if (gw && gw != nh->nh_gw)
262 nhp = RTNH_NEXT(nhp);
263 } endfor_nexthops(fi);
268 static inline size_t dn_fib_nlmsg_size(struct dn_fib_info *fi)
270 size_t payload = NLMSG_ALIGN(sizeof(struct rtmsg))
271 + nla_total_size(4) /* RTA_TABLE */
272 + nla_total_size(2) /* RTA_DST */
273 + nla_total_size(4); /* RTA_PRIORITY */
275 /* space for nested metrics */
276 payload += nla_total_size((RTAX_MAX * nla_total_size(4)));
279 /* Also handles the special case fib_nhs == 1 */
281 /* each nexthop is packed in an attribute */
282 size_t nhsize = nla_total_size(sizeof(struct rtnexthop));
284 /* may contain a gateway attribute */
285 nhsize += nla_total_size(4);
287 /* all nexthops are packed in a nested attribute */
288 payload += nla_total_size(fi->fib_nhs * nhsize);
294 static int dn_fib_dump_info(struct sk_buff *skb, u32 pid, u32 seq, int event,
295 u32 tb_id, u8 type, u8 scope, void *dst, int dst_len,
296 struct dn_fib_info *fi, unsigned int flags)
299 struct nlmsghdr *nlh;
301 nlh = nlmsg_put(skb, pid, seq, event, sizeof(*rtm), flags);
305 rtm = nlmsg_data(nlh);
306 rtm->rtm_family = AF_DECnet;
307 rtm->rtm_dst_len = dst_len;
308 rtm->rtm_src_len = 0;
310 rtm->rtm_table = tb_id;
311 rtm->rtm_flags = fi->fib_flags;
312 rtm->rtm_scope = scope;
313 rtm->rtm_type = type;
314 rtm->rtm_protocol = fi->fib_protocol;
316 if (nla_put_u32(skb, RTA_TABLE, tb_id) < 0)
319 if (rtm->rtm_dst_len &&
320 nla_put(skb, RTA_DST, 2, dst) < 0)
323 if (fi->fib_priority &&
324 nla_put_u32(skb, RTA_PRIORITY, fi->fib_priority) < 0)
327 if (rtnetlink_put_metrics(skb, fi->fib_metrics) < 0)
330 if (fi->fib_nhs == 1) {
331 if (fi->fib_nh->nh_gw &&
332 nla_put_le16(skb, RTA_GATEWAY, fi->fib_nh->nh_gw) < 0)
335 if (fi->fib_nh->nh_oif &&
336 nla_put_u32(skb, RTA_OIF, fi->fib_nh->nh_oif) < 0)
340 if (fi->fib_nhs > 1) {
341 struct rtnexthop *nhp;
342 struct nlattr *mp_head;
344 if (!(mp_head = nla_nest_start(skb, RTA_MULTIPATH)))
348 if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp))))
351 nhp->rtnh_flags = nh->nh_flags & 0xFF;
352 nhp->rtnh_hops = nh->nh_weight - 1;
353 nhp->rtnh_ifindex = nh->nh_oif;
356 nla_put_le16(skb, RTA_GATEWAY, nh->nh_gw) < 0)
359 nhp->rtnh_len = skb_tail_pointer(skb) - (unsigned char *)nhp;
360 } endfor_nexthops(fi);
362 nla_nest_end(skb, mp_head);
365 return nlmsg_end(skb, nlh);
368 nlmsg_cancel(skb, nlh);
373 static void dn_rtmsg_fib(int event, struct dn_fib_node *f, int z, u32 tb_id,
374 struct nlmsghdr *nlh, struct netlink_skb_parms *req)
377 u32 pid = req ? req->pid : 0;
380 skb = nlmsg_new(dn_fib_nlmsg_size(DN_FIB_INFO(f)), GFP_KERNEL);
384 err = dn_fib_dump_info(skb, pid, nlh->nlmsg_seq, event, tb_id,
385 f->fn_type, f->fn_scope, &f->fn_key, z,
388 /* -EMSGSIZE implies BUG in dn_fib_nlmsg_size() */
389 WARN_ON(err == -EMSGSIZE);
393 rtnl_notify(skb, &init_net, pid, RTNLGRP_DECnet_ROUTE, nlh, GFP_KERNEL);
397 rtnl_set_sk_err(&init_net, RTNLGRP_DECnet_ROUTE, err);
400 static __inline__ int dn_hash_dump_bucket(struct sk_buff *skb,
401 struct netlink_callback *cb,
402 struct dn_fib_table *tb,
404 struct dn_fib_node *f)
409 for(i = 0; f; i++, f = f->fn_next) {
412 if (f->fn_state & DN_S_ZOMBIE)
414 if (dn_fib_dump_info(skb, NETLINK_CB(cb->skb).pid,
418 (f->fn_state & DN_S_ZOMBIE) ? 0 : f->fn_type,
419 f->fn_scope, &f->fn_key, dz->dz_order,
420 f->fn_info, NLM_F_MULTI) < 0) {
429 static __inline__ int dn_hash_dump_zone(struct sk_buff *skb,
430 struct netlink_callback *cb,
431 struct dn_fib_table *tb,
437 for(h = 0; h < dz->dz_divisor; h++) {
441 memset(&cb->args[4], 0, sizeof(cb->args) - 4*sizeof(cb->args[0]));
442 if (dz->dz_hash == NULL || dz->dz_hash[h] == NULL)
444 if (dn_hash_dump_bucket(skb, cb, tb, dz, dz->dz_hash[h]) < 0) {
453 static int dn_fib_table_dump(struct dn_fib_table *tb, struct sk_buff *skb,
454 struct netlink_callback *cb)
458 struct dn_hash *table = (struct dn_hash *)tb->data;
461 read_lock(&dn_fib_tables_lock);
462 for(dz = table->dh_zone_list, m = 0; dz; dz = dz->dz_next, m++) {
466 memset(&cb->args[3], 0, sizeof(cb->args) - 3*sizeof(cb->args[0]));
468 if (dn_hash_dump_zone(skb, cb, tb, dz) < 0) {
470 read_unlock(&dn_fib_tables_lock);
474 read_unlock(&dn_fib_tables_lock);
480 int dn_fib_dump(struct sk_buff *skb, struct netlink_callback *cb)
482 struct net *net = sock_net(skb->sk);
484 unsigned int e = 0, s_e;
485 struct dn_fib_table *tb;
486 struct hlist_node *node;
489 if (!net_eq(net, &init_net))
492 if (NLMSG_PAYLOAD(cb->nlh, 0) >= sizeof(struct rtmsg) &&
493 ((struct rtmsg *)nlmsg_data(cb->nlh))->rtm_flags&RTM_F_CLONED)
494 return dn_cache_dump(skb, cb);
499 for (h = s_h; h < DN_FIB_TABLE_HASHSZ; h++, s_h = 0) {
501 hlist_for_each_entry(tb, node, &dn_fib_table_hash[h], hlist) {
505 memset(&cb->args[2], 0, sizeof(cb->args) -
506 2 * sizeof(cb->args[0]));
507 if (tb->dump(tb, skb, cb) < 0)
521 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)
523 struct dn_hash *table = (struct dn_hash *)tb->data;
524 struct dn_fib_node *new_f, *f, **fp, **del_fp;
526 struct dn_fib_info *fi;
527 int z = r->rtm_dst_len;
528 int type = r->rtm_type;
535 dz = table->dh_zones[z];
536 if (!dz && !(dz = dn_new_zone(table, z)))
542 memcpy(&dst, rta->rta_dst, 2);
543 if (dst & ~DZ_MASK(dz))
545 key = dz_key(dst, dz);
548 if ((fi = dn_fib_create_info(r, rta, n, &err)) == NULL)
551 if (dz->dz_nent > (dz->dz_divisor << 2) &&
552 dz->dz_divisor > DN_MAX_DIVISOR &&
553 (z==16 || (1<<z) > dz->dz_divisor))
556 fp = dn_chain_p(key, dz);
559 if (dn_key_leq(key, f->fn_key))
565 if (f && (f->fn_state & DN_S_ZOMBIE) &&
566 dn_key_eq(f->fn_key, key)) {
573 DN_FIB_SCAN_KEY(f, fp, key) {
574 if (fi->fib_priority <= DN_FIB_INFO(f)->fib_priority)
578 if (f && dn_key_eq(f->fn_key, key) &&
579 fi->fib_priority == DN_FIB_INFO(f)->fib_priority) {
580 struct dn_fib_node **ins_fp;
583 if (n->nlmsg_flags & NLM_F_EXCL)
586 if (n->nlmsg_flags & NLM_F_REPLACE) {
596 DN_FIB_SCAN_KEY(f, fp, key) {
597 if (fi->fib_priority != DN_FIB_INFO(f)->fib_priority)
599 if (f->fn_type == type &&
600 f->fn_scope == r->rtm_scope &&
601 DN_FIB_INFO(f) == fi)
605 if (!(n->nlmsg_flags & NLM_F_APPEND)) {
613 if (!(n->nlmsg_flags & NLM_F_CREATE))
618 new_f = kmem_cache_zalloc(dn_hash_kmem, GFP_KERNEL);
623 new_f->fn_type = type;
624 new_f->fn_scope = r->rtm_scope;
625 DN_FIB_INFO(new_f) = fi;
628 write_lock_bh(&dn_fib_tables_lock);
630 write_unlock_bh(&dn_fib_tables_lock);
635 write_lock_bh(&dn_fib_tables_lock);
636 *del_fp = f->fn_next;
637 write_unlock_bh(&dn_fib_tables_lock);
639 if (!(f->fn_state & DN_S_ZOMBIE))
640 dn_rtmsg_fib(RTM_DELROUTE, f, z, tb->n, n, req);
641 if (f->fn_state & DN_S_ACCESSED)
642 dn_rt_cache_flush(-1);
646 dn_rt_cache_flush(-1);
649 dn_rtmsg_fib(RTM_NEWROUTE, new_f, z, tb->n, n, req);
653 dn_fib_release_info(fi);
658 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)
660 struct dn_hash *table = (struct dn_hash*)tb->data;
661 struct dn_fib_node **fp, **del_fp, *f;
662 int z = r->rtm_dst_len;
671 if ((dz = table->dh_zones[z]) == NULL)
677 memcpy(&dst, rta->rta_dst, 2);
678 if (dst & ~DZ_MASK(dz))
680 key = dz_key(dst, dz);
683 fp = dn_chain_p(key, dz);
686 if (dn_key_eq(f->fn_key, key))
688 if (dn_key_leq(key, f->fn_key))
694 DN_FIB_SCAN_KEY(f, fp, key) {
695 struct dn_fib_info *fi = DN_FIB_INFO(f);
697 if (f->fn_state & DN_S_ZOMBIE)
702 if (del_fp == NULL &&
703 (!r->rtm_type || f->fn_type == r->rtm_type) &&
704 (r->rtm_scope == RT_SCOPE_NOWHERE || f->fn_scope == r->rtm_scope) &&
706 fi->fib_protocol == r->rtm_protocol) &&
707 dn_fib_nh_match(r, n, rta, fi) == 0)
713 dn_rtmsg_fib(RTM_DELROUTE, f, z, tb->n, n, req);
716 write_lock_bh(&dn_fib_tables_lock);
717 *del_fp = f->fn_next;
718 write_unlock_bh(&dn_fib_tables_lock);
720 if (f->fn_state & DN_S_ACCESSED)
721 dn_rt_cache_flush(-1);
725 f->fn_state |= DN_S_ZOMBIE;
726 if (f->fn_state & DN_S_ACCESSED) {
727 f->fn_state &= ~DN_S_ACCESSED;
728 dn_rt_cache_flush(-1);
730 if (++dn_fib_hash_zombies > 128)
740 static inline int dn_flush_list(struct dn_fib_node **fp, int z, struct dn_hash *table)
743 struct dn_fib_node *f;
745 while((f = *fp) != NULL) {
746 struct dn_fib_info *fi = DN_FIB_INFO(f);
748 if (fi && ((f->fn_state & DN_S_ZOMBIE) || (fi->fib_flags & RTNH_F_DEAD))) {
749 write_lock_bh(&dn_fib_tables_lock);
751 write_unlock_bh(&dn_fib_tables_lock);
763 static int dn_fib_table_flush(struct dn_fib_table *tb)
765 struct dn_hash *table = (struct dn_hash *)tb->data;
769 dn_fib_hash_zombies = 0;
770 for(dz = table->dh_zone_list; dz; dz = dz->dz_next) {
773 for(i = dz->dz_divisor-1; i >= 0; i--)
774 tmp += dn_flush_list(&dz->dz_hash[i], dz->dz_order, table);
782 static int dn_fib_table_lookup(struct dn_fib_table *tb, const struct flowidn *flp, struct dn_fib_res *res)
786 struct dn_hash *t = (struct dn_hash *)tb->data;
788 read_lock(&dn_fib_tables_lock);
789 for(dz = t->dh_zone_list; dz; dz = dz->dz_next) {
790 struct dn_fib_node *f;
791 dn_fib_key_t k = dz_key(flp->daddr, dz);
793 for(f = dz_chain(k, dz); f; f = f->fn_next) {
794 if (!dn_key_eq(k, f->fn_key)) {
795 if (dn_key_leq(k, f->fn_key))
801 f->fn_state |= DN_S_ACCESSED;
803 if (f->fn_state&DN_S_ZOMBIE)
806 if (f->fn_scope < flp->flowidn_scope)
809 err = dn_fib_semantic_match(f->fn_type, DN_FIB_INFO(f), flp, res);
812 res->type = f->fn_type;
813 res->scope = f->fn_scope;
814 res->prefixlen = dz->dz_order;
823 read_unlock(&dn_fib_tables_lock);
828 struct dn_fib_table *dn_fib_get_table(u32 n, int create)
830 struct dn_fib_table *t;
831 struct hlist_node *node;
834 if (n < RT_TABLE_MIN)
837 if (n > RT_TABLE_MAX)
840 h = n & (DN_FIB_TABLE_HASHSZ - 1);
842 hlist_for_each_entry_rcu(t, node, &dn_fib_table_hash[h], hlist) {
853 if (in_interrupt()) {
854 net_dbg_ratelimited("DECnet: BUG! Attempt to create routing table from interrupt\n");
858 t = kzalloc(sizeof(struct dn_fib_table) + sizeof(struct dn_hash),
864 t->insert = dn_fib_table_insert;
865 t->delete = dn_fib_table_delete;
866 t->lookup = dn_fib_table_lookup;
867 t->flush = dn_fib_table_flush;
868 t->dump = dn_fib_table_dump;
869 hlist_add_head_rcu(&t->hlist, &dn_fib_table_hash[h]);
874 struct dn_fib_table *dn_fib_empty_table(void)
878 for(id = RT_TABLE_MIN; id <= RT_TABLE_MAX; id++)
879 if (dn_fib_get_table(id, 0) == NULL)
880 return dn_fib_get_table(id, 1);
884 void dn_fib_flush(void)
887 struct dn_fib_table *tb;
888 struct hlist_node *node;
891 for (h = 0; h < DN_FIB_TABLE_HASHSZ; h++) {
892 hlist_for_each_entry(tb, node, &dn_fib_table_hash[h], hlist)
893 flushed += tb->flush(tb);
897 dn_rt_cache_flush(-1);
900 void __init dn_fib_table_init(void)
902 dn_hash_kmem = kmem_cache_create("dn_fib_info_cache",
903 sizeof(struct dn_fib_info),
904 0, SLAB_HWCACHE_ALIGN,
908 void __exit dn_fib_table_cleanup(void)
910 struct dn_fib_table *t;
911 struct hlist_node *node, *next;
914 write_lock(&dn_fib_tables_lock);
915 for (h = 0; h < DN_FIB_TABLE_HASHSZ; h++) {
916 hlist_for_each_entry_safe(t, node, next, &dn_fib_table_hash[h],
918 hlist_del(&t->hlist);
922 write_unlock(&dn_fib_tables_lock);