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 <asm/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)
62 #define dz_prefix(key,dz) ((key).datum)
64 #define for_nexthops(fi) { int nhsel; const struct dn_fib_nh *nh;\
65 for(nhsel = 0, nh = (fi)->fib_nh; nhsel < (fi)->fib_nhs; nh++, nhsel++)
67 #define endfor_nexthops(fi) }
69 #define DN_MAX_DIVISOR 1024
71 #define DN_S_ACCESSED 2
73 #define DN_FIB_SCAN(f, fp) \
74 for( ; ((f) = *(fp)) != NULL; (fp) = &(f)->fn_next)
76 #define DN_FIB_SCAN_KEY(f, fp, key) \
77 for( ; ((f) = *(fp)) != NULL && dn_key_eq((f)->fn_key, (key)); (fp) = &(f)->fn_next)
79 #define RT_TABLE_MIN 1
80 #define DN_FIB_TABLE_HASHSZ 256
81 static struct hlist_head dn_fib_table_hash[DN_FIB_TABLE_HASHSZ];
82 static DEFINE_RWLOCK(dn_fib_tables_lock);
84 static struct kmem_cache *dn_hash_kmem __read_mostly;
85 static int dn_fib_hash_zombies;
87 static inline dn_fib_idx_t dn_hash(dn_fib_key_t key, struct dn_zone *dz)
89 u16 h = le16_to_cpu(key.datum)>>(16 - dz->dz_order);
93 return *(dn_fib_idx_t *)&h;
96 static inline dn_fib_key_t dz_key(__le16 dst, struct dn_zone *dz)
99 k.datum = dst & DZ_MASK(dz);
103 static inline struct dn_fib_node **dn_chain_p(dn_fib_key_t key, struct dn_zone *dz)
105 return &dz->dz_hash[dn_hash(key, dz).datum];
108 static inline struct dn_fib_node *dz_chain(dn_fib_key_t key, struct dn_zone *dz)
110 return dz->dz_hash[dn_hash(key, dz).datum];
113 static inline int dn_key_eq(dn_fib_key_t a, dn_fib_key_t b)
115 return a.datum == b.datum;
118 static inline int dn_key_leq(dn_fib_key_t a, dn_fib_key_t b)
120 return a.datum <= b.datum;
123 static inline void dn_rebuild_zone(struct dn_zone *dz,
124 struct dn_fib_node **old_ht,
128 struct dn_fib_node *f, **fp, *next;
130 for(i = 0; i < old_divisor; i++) {
131 for(f = old_ht[i]; f; f = f->fn_next) {
133 for(fp = dn_chain_p(f->fn_key, dz);
134 *fp && dn_key_leq((*fp)->fn_key, f->fn_key);
135 fp = &(*fp)->fn_next)
143 static void dn_rehash_zone(struct dn_zone *dz)
145 struct dn_fib_node **ht, **old_ht;
146 int old_divisor, new_divisor;
149 old_divisor = dz->dz_divisor;
151 switch(old_divisor) {
157 printk(KERN_DEBUG "DECnet: dn_rehash_zone: BUG! %d\n", old_divisor);
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;
300 unsigned char *b = skb_tail_pointer(skb);
302 nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*rtm), flags);
303 rtm = NLMSG_DATA(nlh);
304 rtm->rtm_family = AF_DECnet;
305 rtm->rtm_dst_len = dst_len;
306 rtm->rtm_src_len = 0;
308 rtm->rtm_table = tb_id;
309 RTA_PUT_U32(skb, RTA_TABLE, tb_id);
310 rtm->rtm_flags = fi->fib_flags;
311 rtm->rtm_scope = scope;
312 rtm->rtm_type = type;
313 if (rtm->rtm_dst_len)
314 RTA_PUT(skb, RTA_DST, 2, dst);
315 rtm->rtm_protocol = fi->fib_protocol;
316 if (fi->fib_priority)
317 RTA_PUT(skb, RTA_PRIORITY, 4, &fi->fib_priority);
318 if (rtnetlink_put_metrics(skb, fi->fib_metrics) < 0)
320 if (fi->fib_nhs == 1) {
321 if (fi->fib_nh->nh_gw)
322 RTA_PUT(skb, RTA_GATEWAY, 2, &fi->fib_nh->nh_gw);
323 if (fi->fib_nh->nh_oif)
324 RTA_PUT(skb, RTA_OIF, sizeof(int), &fi->fib_nh->nh_oif);
326 if (fi->fib_nhs > 1) {
327 struct rtnexthop *nhp;
328 struct rtattr *mp_head;
329 if (skb_tailroom(skb) <= RTA_SPACE(0))
331 mp_head = (struct rtattr *)skb_put(skb, RTA_SPACE(0));
334 if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4))
336 nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp)));
337 nhp->rtnh_flags = nh->nh_flags & 0xFF;
338 nhp->rtnh_hops = nh->nh_weight - 1;
339 nhp->rtnh_ifindex = nh->nh_oif;
341 RTA_PUT(skb, RTA_GATEWAY, 2, &nh->nh_gw);
342 nhp->rtnh_len = skb_tail_pointer(skb) - (unsigned char *)nhp;
343 } endfor_nexthops(fi);
344 mp_head->rta_type = RTA_MULTIPATH;
345 mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head;
348 nlh->nlmsg_len = skb_tail_pointer(skb) - b;
359 static void dn_rtmsg_fib(int event, struct dn_fib_node *f, int z, u32 tb_id,
360 struct nlmsghdr *nlh, struct netlink_skb_parms *req)
363 u32 pid = req ? req->pid : 0;
366 skb = nlmsg_new(dn_fib_nlmsg_size(DN_FIB_INFO(f)), GFP_KERNEL);
370 err = dn_fib_dump_info(skb, pid, nlh->nlmsg_seq, event, tb_id,
371 f->fn_type, f->fn_scope, &f->fn_key, z,
374 /* -EMSGSIZE implies BUG in dn_fib_nlmsg_size() */
375 WARN_ON(err == -EMSGSIZE);
379 rtnl_notify(skb, &init_net, pid, RTNLGRP_DECnet_ROUTE, nlh, GFP_KERNEL);
383 rtnl_set_sk_err(&init_net, RTNLGRP_DECnet_ROUTE, err);
386 static __inline__ int dn_hash_dump_bucket(struct sk_buff *skb,
387 struct netlink_callback *cb,
388 struct dn_fib_table *tb,
390 struct dn_fib_node *f)
395 for(i = 0; f; i++, f = f->fn_next) {
398 if (f->fn_state & DN_S_ZOMBIE)
400 if (dn_fib_dump_info(skb, NETLINK_CB(cb->skb).pid,
404 (f->fn_state & DN_S_ZOMBIE) ? 0 : f->fn_type,
405 f->fn_scope, &f->fn_key, dz->dz_order,
406 f->fn_info, NLM_F_MULTI) < 0) {
415 static __inline__ int dn_hash_dump_zone(struct sk_buff *skb,
416 struct netlink_callback *cb,
417 struct dn_fib_table *tb,
423 for(h = 0; h < dz->dz_divisor; h++) {
427 memset(&cb->args[4], 0, sizeof(cb->args) - 4*sizeof(cb->args[0]));
428 if (dz->dz_hash == NULL || dz->dz_hash[h] == NULL)
430 if (dn_hash_dump_bucket(skb, cb, tb, dz, dz->dz_hash[h]) < 0) {
439 static int dn_fib_table_dump(struct dn_fib_table *tb, struct sk_buff *skb,
440 struct netlink_callback *cb)
444 struct dn_hash *table = (struct dn_hash *)tb->data;
447 read_lock(&dn_fib_tables_lock);
448 for(dz = table->dh_zone_list, m = 0; dz; dz = dz->dz_next, m++) {
452 memset(&cb->args[3], 0, sizeof(cb->args) - 3*sizeof(cb->args[0]));
454 if (dn_hash_dump_zone(skb, cb, tb, dz) < 0) {
456 read_unlock(&dn_fib_tables_lock);
460 read_unlock(&dn_fib_tables_lock);
466 int dn_fib_dump(struct sk_buff *skb, struct netlink_callback *cb)
468 struct net *net = sock_net(skb->sk);
470 unsigned int e = 0, s_e;
471 struct dn_fib_table *tb;
472 struct hlist_node *node;
475 if (!net_eq(net, &init_net))
478 if (NLMSG_PAYLOAD(cb->nlh, 0) >= sizeof(struct rtmsg) &&
479 ((struct rtmsg *)NLMSG_DATA(cb->nlh))->rtm_flags&RTM_F_CLONED)
480 return dn_cache_dump(skb, cb);
485 for (h = s_h; h < DN_FIB_TABLE_HASHSZ; h++, s_h = 0) {
487 hlist_for_each_entry(tb, node, &dn_fib_table_hash[h], hlist) {
491 memset(&cb->args[2], 0, sizeof(cb->args) -
492 2 * sizeof(cb->args[0]));
493 if (tb->dump(tb, skb, cb) < 0)
507 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)
509 struct dn_hash *table = (struct dn_hash *)tb->data;
510 struct dn_fib_node *new_f, *f, **fp, **del_fp;
512 struct dn_fib_info *fi;
513 int z = r->rtm_dst_len;
514 int type = r->rtm_type;
521 dz = table->dh_zones[z];
522 if (!dz && !(dz = dn_new_zone(table, z)))
528 memcpy(&dst, rta->rta_dst, 2);
529 if (dst & ~DZ_MASK(dz))
531 key = dz_key(dst, dz);
534 if ((fi = dn_fib_create_info(r, rta, n, &err)) == NULL)
537 if (dz->dz_nent > (dz->dz_divisor << 2) &&
538 dz->dz_divisor > DN_MAX_DIVISOR &&
539 (z==16 || (1<<z) > dz->dz_divisor))
542 fp = dn_chain_p(key, dz);
545 if (dn_key_leq(key, f->fn_key))
551 if (f && (f->fn_state & DN_S_ZOMBIE) &&
552 dn_key_eq(f->fn_key, key)) {
559 DN_FIB_SCAN_KEY(f, fp, key) {
560 if (fi->fib_priority <= DN_FIB_INFO(f)->fib_priority)
564 if (f && dn_key_eq(f->fn_key, key) &&
565 fi->fib_priority == DN_FIB_INFO(f)->fib_priority) {
566 struct dn_fib_node **ins_fp;
569 if (n->nlmsg_flags & NLM_F_EXCL)
572 if (n->nlmsg_flags & NLM_F_REPLACE) {
582 DN_FIB_SCAN_KEY(f, fp, key) {
583 if (fi->fib_priority != DN_FIB_INFO(f)->fib_priority)
585 if (f->fn_type == type &&
586 f->fn_scope == r->rtm_scope &&
587 DN_FIB_INFO(f) == fi)
591 if (!(n->nlmsg_flags & NLM_F_APPEND)) {
599 if (!(n->nlmsg_flags & NLM_F_CREATE))
604 new_f = kmem_cache_zalloc(dn_hash_kmem, GFP_KERNEL);
609 new_f->fn_type = type;
610 new_f->fn_scope = r->rtm_scope;
611 DN_FIB_INFO(new_f) = fi;
614 write_lock_bh(&dn_fib_tables_lock);
616 write_unlock_bh(&dn_fib_tables_lock);
621 write_lock_bh(&dn_fib_tables_lock);
622 *del_fp = f->fn_next;
623 write_unlock_bh(&dn_fib_tables_lock);
625 if (!(f->fn_state & DN_S_ZOMBIE))
626 dn_rtmsg_fib(RTM_DELROUTE, f, z, tb->n, n, req);
627 if (f->fn_state & DN_S_ACCESSED)
628 dn_rt_cache_flush(-1);
632 dn_rt_cache_flush(-1);
635 dn_rtmsg_fib(RTM_NEWROUTE, new_f, z, tb->n, n, req);
639 dn_fib_release_info(fi);
644 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)
646 struct dn_hash *table = (struct dn_hash*)tb->data;
647 struct dn_fib_node **fp, **del_fp, *f;
648 int z = r->rtm_dst_len;
657 if ((dz = table->dh_zones[z]) == NULL)
663 memcpy(&dst, rta->rta_dst, 2);
664 if (dst & ~DZ_MASK(dz))
666 key = dz_key(dst, dz);
669 fp = dn_chain_p(key, dz);
672 if (dn_key_eq(f->fn_key, key))
674 if (dn_key_leq(key, f->fn_key))
680 DN_FIB_SCAN_KEY(f, fp, key) {
681 struct dn_fib_info *fi = DN_FIB_INFO(f);
683 if (f->fn_state & DN_S_ZOMBIE)
688 if (del_fp == NULL &&
689 (!r->rtm_type || f->fn_type == r->rtm_type) &&
690 (r->rtm_scope == RT_SCOPE_NOWHERE || f->fn_scope == r->rtm_scope) &&
692 fi->fib_protocol == r->rtm_protocol) &&
693 dn_fib_nh_match(r, n, rta, fi) == 0)
699 dn_rtmsg_fib(RTM_DELROUTE, f, z, tb->n, n, req);
702 write_lock_bh(&dn_fib_tables_lock);
703 *del_fp = f->fn_next;
704 write_unlock_bh(&dn_fib_tables_lock);
706 if (f->fn_state & DN_S_ACCESSED)
707 dn_rt_cache_flush(-1);
711 f->fn_state |= DN_S_ZOMBIE;
712 if (f->fn_state & DN_S_ACCESSED) {
713 f->fn_state &= ~DN_S_ACCESSED;
714 dn_rt_cache_flush(-1);
716 if (++dn_fib_hash_zombies > 128)
726 static inline int dn_flush_list(struct dn_fib_node **fp, int z, struct dn_hash *table)
729 struct dn_fib_node *f;
731 while((f = *fp) != NULL) {
732 struct dn_fib_info *fi = DN_FIB_INFO(f);
734 if (fi && ((f->fn_state & DN_S_ZOMBIE) || (fi->fib_flags & RTNH_F_DEAD))) {
735 write_lock_bh(&dn_fib_tables_lock);
737 write_unlock_bh(&dn_fib_tables_lock);
749 static int dn_fib_table_flush(struct dn_fib_table *tb)
751 struct dn_hash *table = (struct dn_hash *)tb->data;
755 dn_fib_hash_zombies = 0;
756 for(dz = table->dh_zone_list; dz; dz = dz->dz_next) {
759 for(i = dz->dz_divisor-1; i >= 0; i--)
760 tmp += dn_flush_list(&dz->dz_hash[i], dz->dz_order, table);
768 static int dn_fib_table_lookup(struct dn_fib_table *tb, const struct flowi *flp, struct dn_fib_res *res)
772 struct dn_hash *t = (struct dn_hash *)tb->data;
774 read_lock(&dn_fib_tables_lock);
775 for(dz = t->dh_zone_list; dz; dz = dz->dz_next) {
776 struct dn_fib_node *f;
777 dn_fib_key_t k = dz_key(flp->fld_dst, dz);
779 for(f = dz_chain(k, dz); f; f = f->fn_next) {
780 if (!dn_key_eq(k, f->fn_key)) {
781 if (dn_key_leq(k, f->fn_key))
787 f->fn_state |= DN_S_ACCESSED;
789 if (f->fn_state&DN_S_ZOMBIE)
792 if (f->fn_scope < flp->fld_scope)
795 err = dn_fib_semantic_match(f->fn_type, DN_FIB_INFO(f), flp, res);
798 res->type = f->fn_type;
799 res->scope = f->fn_scope;
800 res->prefixlen = dz->dz_order;
809 read_unlock(&dn_fib_tables_lock);
814 struct dn_fib_table *dn_fib_get_table(u32 n, int create)
816 struct dn_fib_table *t;
817 struct hlist_node *node;
820 if (n < RT_TABLE_MIN)
823 if (n > RT_TABLE_MAX)
826 h = n & (DN_FIB_TABLE_HASHSZ - 1);
828 hlist_for_each_entry_rcu(t, node, &dn_fib_table_hash[h], hlist) {
839 if (in_interrupt() && net_ratelimit()) {
840 printk(KERN_DEBUG "DECnet: BUG! Attempt to create routing table from interrupt\n");
844 t = kzalloc(sizeof(struct dn_fib_table) + sizeof(struct dn_hash),
850 t->insert = dn_fib_table_insert;
851 t->delete = dn_fib_table_delete;
852 t->lookup = dn_fib_table_lookup;
853 t->flush = dn_fib_table_flush;
854 t->dump = dn_fib_table_dump;
855 hlist_add_head_rcu(&t->hlist, &dn_fib_table_hash[h]);
860 struct dn_fib_table *dn_fib_empty_table(void)
864 for(id = RT_TABLE_MIN; id <= RT_TABLE_MAX; id++)
865 if (dn_fib_get_table(id, 0) == NULL)
866 return dn_fib_get_table(id, 1);
870 void dn_fib_flush(void)
873 struct dn_fib_table *tb;
874 struct hlist_node *node;
877 for (h = 0; h < DN_FIB_TABLE_HASHSZ; h++) {
878 hlist_for_each_entry(tb, node, &dn_fib_table_hash[h], hlist)
879 flushed += tb->flush(tb);
883 dn_rt_cache_flush(-1);
886 void __init dn_fib_table_init(void)
888 dn_hash_kmem = kmem_cache_create("dn_fib_info_cache",
889 sizeof(struct dn_fib_info),
890 0, SLAB_HWCACHE_ALIGN,
894 void __exit dn_fib_table_cleanup(void)
896 struct dn_fib_table *t;
897 struct hlist_node *node, *next;
900 write_lock(&dn_fib_tables_lock);
901 for (h = 0; h < DN_FIB_TABLE_HASHSZ; h++) {
902 hlist_for_each_entry_safe(t, node, next, &dn_fib_table_hash[h],
904 hlist_del(&t->hlist);
908 write_unlock(&dn_fib_tables_lock);