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>
31 #include <net/netlink.h>
34 #include <net/fib_rules.h>
36 #include <net/dn_route.h>
37 #include <net/dn_fib.h>
38 #include <net/dn_neigh.h>
39 #include <net/dn_dev.h>
43 struct dn_zone *dz_next;
44 struct dn_fib_node **dz_hash;
48 #define DZ_HASHMASK(dz) ((dz)->dz_hashmask)
51 #define DZ_MASK(dz) ((dz)->dz_mask)
56 struct dn_zone *dh_zones[17];
57 struct dn_zone *dh_zone_list;
60 #define dz_key_0(key) ((key).datum = 0)
61 #define dz_prefix(key,dz) ((key).datum)
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 = dn_ntohs(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,
127 struct dn_fib_node *f, **fp, *next;
129 for(i = 0; i < old_divisor; i++) {
130 for(f = old_ht[i]; f; f = f->fn_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", old_divisor);
159 new_hashmask = 0x3FF;
163 ht = kcalloc(new_divisor, sizeof(struct dn_fib_node*), GFP_KERNEL);
167 write_lock_bh(&dn_fib_tables_lock);
168 old_ht = dz->dz_hash;
170 dz->dz_hashmask = new_hashmask;
171 dz->dz_divisor = new_divisor;
172 dn_rebuild_zone(dz, old_ht, old_divisor);
173 write_unlock_bh(&dn_fib_tables_lock);
177 static void dn_free_node(struct dn_fib_node *f)
179 dn_fib_release_info(DN_FIB_INFO(f));
180 kmem_cache_free(dn_hash_kmem, f);
184 static struct dn_zone *dn_new_zone(struct dn_hash *table, int z)
187 struct dn_zone *dz = kzalloc(sizeof(struct dn_zone), GFP_KERNEL);
193 dz->dz_hashmask = 0x0F;
199 dz->dz_hash = kcalloc(dz->dz_divisor, sizeof(struct dn_fib_node *), GFP_KERNEL);
206 dz->dz_mask = dnet_make_mask(z);
208 for(i = z + 1; i <= 16; i++)
209 if (table->dh_zones[i])
212 write_lock_bh(&dn_fib_tables_lock);
214 dz->dz_next = table->dh_zone_list;
215 table->dh_zone_list = dz;
217 dz->dz_next = table->dh_zones[i]->dz_next;
218 table->dh_zones[i]->dz_next = dz;
220 table->dh_zones[z] = dz;
221 write_unlock_bh(&dn_fib_tables_lock);
226 static int dn_fib_nh_match(struct rtmsg *r, struct nlmsghdr *nlh, struct dn_kern_rta *rta, struct dn_fib_info *fi)
228 struct rtnexthop *nhp;
231 if (rta->rta_priority && *rta->rta_priority != fi->fib_priority)
234 if (rta->rta_oif || rta->rta_gw) {
235 if ((!rta->rta_oif || *rta->rta_oif == fi->fib_nh->nh_oif) &&
236 (!rta->rta_gw || memcmp(rta->rta_gw, &fi->fib_nh->nh_gw, 2) == 0))
241 if (rta->rta_mp == NULL)
244 nhp = RTA_DATA(rta->rta_mp);
245 nhlen = RTA_PAYLOAD(rta->rta_mp);
248 int attrlen = nhlen - sizeof(struct rtnexthop);
251 if (attrlen < 0 || (nhlen -= nhp->rtnh_len) < 0)
253 if (nhp->rtnh_ifindex && nhp->rtnh_ifindex != nh->nh_oif)
256 gw = dn_fib_get_attr16(RTNH_DATA(nhp), attrlen, RTA_GATEWAY);
258 if (gw && gw != nh->nh_gw)
261 nhp = RTNH_NEXT(nhp);
262 } endfor_nexthops(fi);
267 static inline size_t dn_fib_nlmsg_size(struct dn_fib_info *fi)
269 size_t payload = NLMSG_ALIGN(sizeof(struct rtmsg))
270 + nla_total_size(4) /* RTA_TABLE */
271 + nla_total_size(2) /* RTA_DST */
272 + nla_total_size(4); /* RTA_PRIORITY */
274 /* space for nested metrics */
275 payload += nla_total_size((RTAX_MAX * nla_total_size(4)));
278 /* Also handles the special case fib_nhs == 1 */
280 /* each nexthop is packed in an attribute */
281 size_t nhsize = nla_total_size(sizeof(struct rtnexthop));
283 /* may contain a gateway attribute */
284 nhsize += nla_total_size(4);
286 /* all nexthops are packed in a nested attribute */
287 payload += nla_total_size(fi->fib_nhs * nhsize);
293 static int dn_fib_dump_info(struct sk_buff *skb, u32 pid, u32 seq, int event,
294 u32 tb_id, u8 type, u8 scope, void *dst, int dst_len,
295 struct dn_fib_info *fi, unsigned int flags)
298 struct nlmsghdr *nlh;
299 unsigned char *b = skb_tail_pointer(skb);
301 nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*rtm), flags);
302 rtm = NLMSG_DATA(nlh);
303 rtm->rtm_family = AF_DECnet;
304 rtm->rtm_dst_len = dst_len;
305 rtm->rtm_src_len = 0;
307 rtm->rtm_table = tb_id;
308 RTA_PUT_U32(skb, RTA_TABLE, tb_id);
309 rtm->rtm_flags = fi->fib_flags;
310 rtm->rtm_scope = scope;
311 rtm->rtm_type = type;
312 if (rtm->rtm_dst_len)
313 RTA_PUT(skb, RTA_DST, 2, dst);
314 rtm->rtm_protocol = fi->fib_protocol;
315 if (fi->fib_priority)
316 RTA_PUT(skb, RTA_PRIORITY, 4, &fi->fib_priority);
317 if (rtnetlink_put_metrics(skb, fi->fib_metrics) < 0)
319 if (fi->fib_nhs == 1) {
320 if (fi->fib_nh->nh_gw)
321 RTA_PUT(skb, RTA_GATEWAY, 2, &fi->fib_nh->nh_gw);
322 if (fi->fib_nh->nh_oif)
323 RTA_PUT(skb, RTA_OIF, sizeof(int), &fi->fib_nh->nh_oif);
325 if (fi->fib_nhs > 1) {
326 struct rtnexthop *nhp;
327 struct rtattr *mp_head;
328 if (skb_tailroom(skb) <= RTA_SPACE(0))
330 mp_head = (struct rtattr *)skb_put(skb, RTA_SPACE(0));
333 if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4))
335 nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp)));
336 nhp->rtnh_flags = nh->nh_flags & 0xFF;
337 nhp->rtnh_hops = nh->nh_weight - 1;
338 nhp->rtnh_ifindex = nh->nh_oif;
340 RTA_PUT(skb, RTA_GATEWAY, 2, &nh->nh_gw);
341 nhp->rtnh_len = skb_tail_pointer(skb) - (unsigned char *)nhp;
342 } endfor_nexthops(fi);
343 mp_head->rta_type = RTA_MULTIPATH;
344 mp_head->rta_len = skb_tail_pointer(skb) - (u8 *)mp_head;
347 nlh->nlmsg_len = skb_tail_pointer(skb) - b;
358 static void dn_rtmsg_fib(int event, struct dn_fib_node *f, int z, u32 tb_id,
359 struct nlmsghdr *nlh, struct netlink_skb_parms *req)
362 u32 pid = req ? req->pid : 0;
365 skb = nlmsg_new(dn_fib_nlmsg_size(DN_FIB_INFO(f)), GFP_KERNEL);
369 err = dn_fib_dump_info(skb, pid, nlh->nlmsg_seq, event, tb_id,
370 f->fn_type, f->fn_scope, &f->fn_key, z,
373 /* -EMSGSIZE implies BUG in dn_fib_nlmsg_size() */
374 WARN_ON(err == -EMSGSIZE);
378 err = rtnl_notify(skb, &init_net, pid, RTNLGRP_DECnet_ROUTE, nlh, GFP_KERNEL);
381 rtnl_set_sk_err(&init_net, RTNLGRP_DECnet_ROUTE, err);
384 static __inline__ int dn_hash_dump_bucket(struct sk_buff *skb,
385 struct netlink_callback *cb,
386 struct dn_fib_table *tb,
388 struct dn_fib_node *f)
393 for(i = 0; f; i++, f = f->fn_next) {
396 if (f->fn_state & DN_S_ZOMBIE)
398 if (dn_fib_dump_info(skb, NETLINK_CB(cb->skb).pid,
402 (f->fn_state & DN_S_ZOMBIE) ? 0 : f->fn_type,
403 f->fn_scope, &f->fn_key, dz->dz_order,
404 f->fn_info, NLM_F_MULTI) < 0) {
413 static __inline__ int dn_hash_dump_zone(struct sk_buff *skb,
414 struct netlink_callback *cb,
415 struct dn_fib_table *tb,
421 for(h = 0; h < dz->dz_divisor; h++) {
425 memset(&cb->args[4], 0, sizeof(cb->args) - 4*sizeof(cb->args[0]));
426 if (dz->dz_hash == NULL || dz->dz_hash[h] == NULL)
428 if (dn_hash_dump_bucket(skb, cb, tb, dz, dz->dz_hash[h]) < 0) {
437 static int dn_fib_table_dump(struct dn_fib_table *tb, struct sk_buff *skb,
438 struct netlink_callback *cb)
442 struct dn_hash *table = (struct dn_hash *)tb->data;
445 read_lock(&dn_fib_tables_lock);
446 for(dz = table->dh_zone_list, m = 0; dz; dz = dz->dz_next, m++) {
450 memset(&cb->args[3], 0, sizeof(cb->args) - 3*sizeof(cb->args[0]));
452 if (dn_hash_dump_zone(skb, cb, tb, dz) < 0) {
454 read_unlock(&dn_fib_tables_lock);
458 read_unlock(&dn_fib_tables_lock);
464 int dn_fib_dump(struct sk_buff *skb, struct netlink_callback *cb)
466 struct net *net = skb->sk->sk_net;
468 unsigned int e = 0, s_e;
469 struct dn_fib_table *tb;
470 struct hlist_node *node;
473 if (net != &init_net)
476 if (NLMSG_PAYLOAD(cb->nlh, 0) >= sizeof(struct rtmsg) &&
477 ((struct rtmsg *)NLMSG_DATA(cb->nlh))->rtm_flags&RTM_F_CLONED)
478 return dn_cache_dump(skb, cb);
483 for (h = s_h; h < DN_FIB_TABLE_HASHSZ; h++, s_h = 0) {
485 hlist_for_each_entry(tb, node, &dn_fib_table_hash[h], hlist) {
489 memset(&cb->args[2], 0, sizeof(cb->args) -
490 2 * sizeof(cb->args[0]));
491 if (tb->dump(tb, skb, cb) < 0)
505 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)
507 struct dn_hash *table = (struct dn_hash *)tb->data;
508 struct dn_fib_node *new_f, *f, **fp, **del_fp;
510 struct dn_fib_info *fi;
511 int z = r->rtm_dst_len;
512 int type = r->rtm_type;
519 dz = table->dh_zones[z];
520 if (!dz && !(dz = dn_new_zone(table, z)))
526 memcpy(&dst, rta->rta_dst, 2);
527 if (dst & ~DZ_MASK(dz))
529 key = dz_key(dst, dz);
532 if ((fi = dn_fib_create_info(r, rta, n, &err)) == NULL)
535 if (dz->dz_nent > (dz->dz_divisor << 2) &&
536 dz->dz_divisor > DN_MAX_DIVISOR &&
537 (z==16 || (1<<z) > dz->dz_divisor))
540 fp = dn_chain_p(key, dz);
543 if (dn_key_leq(key, f->fn_key))
549 if (f && (f->fn_state & DN_S_ZOMBIE) &&
550 dn_key_eq(f->fn_key, key)) {
557 DN_FIB_SCAN_KEY(f, fp, key) {
558 if (fi->fib_priority <= DN_FIB_INFO(f)->fib_priority)
562 if (f && dn_key_eq(f->fn_key, key) &&
563 fi->fib_priority == DN_FIB_INFO(f)->fib_priority) {
564 struct dn_fib_node **ins_fp;
567 if (n->nlmsg_flags & NLM_F_EXCL)
570 if (n->nlmsg_flags & NLM_F_REPLACE) {
580 DN_FIB_SCAN_KEY(f, fp, key) {
581 if (fi->fib_priority != DN_FIB_INFO(f)->fib_priority)
583 if (f->fn_type == type && f->fn_scope == r->rtm_scope
584 && DN_FIB_INFO(f) == fi)
588 if (!(n->nlmsg_flags & NLM_F_APPEND)) {
596 if (!(n->nlmsg_flags & NLM_F_CREATE))
601 new_f = kmem_cache_zalloc(dn_hash_kmem, GFP_KERNEL);
606 new_f->fn_type = type;
607 new_f->fn_scope = r->rtm_scope;
608 DN_FIB_INFO(new_f) = fi;
611 write_lock_bh(&dn_fib_tables_lock);
613 write_unlock_bh(&dn_fib_tables_lock);
618 write_lock_bh(&dn_fib_tables_lock);
619 *del_fp = f->fn_next;
620 write_unlock_bh(&dn_fib_tables_lock);
622 if (!(f->fn_state & DN_S_ZOMBIE))
623 dn_rtmsg_fib(RTM_DELROUTE, f, z, tb->n, n, req);
624 if (f->fn_state & DN_S_ACCESSED)
625 dn_rt_cache_flush(-1);
629 dn_rt_cache_flush(-1);
632 dn_rtmsg_fib(RTM_NEWROUTE, new_f, z, tb->n, n, req);
636 dn_fib_release_info(fi);
641 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)
643 struct dn_hash *table = (struct dn_hash*)tb->data;
644 struct dn_fib_node **fp, **del_fp, *f;
645 int z = r->rtm_dst_len;
654 if ((dz = table->dh_zones[z]) == NULL)
660 memcpy(&dst, rta->rta_dst, 2);
661 if (dst & ~DZ_MASK(dz))
663 key = dz_key(dst, dz);
666 fp = dn_chain_p(key, dz);
669 if (dn_key_eq(f->fn_key, key))
671 if (dn_key_leq(key, f->fn_key))
677 DN_FIB_SCAN_KEY(f, fp, key) {
678 struct dn_fib_info *fi = DN_FIB_INFO(f);
680 if (f->fn_state & DN_S_ZOMBIE)
685 if (del_fp == NULL &&
686 (!r->rtm_type || f->fn_type == r->rtm_type) &&
687 (r->rtm_scope == RT_SCOPE_NOWHERE || f->fn_scope == r->rtm_scope) &&
689 fi->fib_protocol == r->rtm_protocol) &&
690 dn_fib_nh_match(r, n, rta, fi) == 0)
696 dn_rtmsg_fib(RTM_DELROUTE, f, z, tb->n, n, req);
699 write_lock_bh(&dn_fib_tables_lock);
700 *del_fp = f->fn_next;
701 write_unlock_bh(&dn_fib_tables_lock);
703 if (f->fn_state & DN_S_ACCESSED)
704 dn_rt_cache_flush(-1);
708 f->fn_state |= DN_S_ZOMBIE;
709 if (f->fn_state & DN_S_ACCESSED) {
710 f->fn_state &= ~DN_S_ACCESSED;
711 dn_rt_cache_flush(-1);
713 if (++dn_fib_hash_zombies > 128)
723 static inline int dn_flush_list(struct dn_fib_node **fp, int z, struct dn_hash *table)
726 struct dn_fib_node *f;
728 while((f = *fp) != NULL) {
729 struct dn_fib_info *fi = DN_FIB_INFO(f);
731 if (fi && ((f->fn_state & DN_S_ZOMBIE) || (fi->fib_flags & RTNH_F_DEAD))) {
732 write_lock_bh(&dn_fib_tables_lock);
734 write_unlock_bh(&dn_fib_tables_lock);
746 static int dn_fib_table_flush(struct dn_fib_table *tb)
748 struct dn_hash *table = (struct dn_hash *)tb->data;
752 dn_fib_hash_zombies = 0;
753 for(dz = table->dh_zone_list; dz; dz = dz->dz_next) {
756 for(i = dz->dz_divisor-1; i >= 0; i--)
757 tmp += dn_flush_list(&dz->dz_hash[i], dz->dz_order, table);
765 static int dn_fib_table_lookup(struct dn_fib_table *tb, const struct flowi *flp, struct dn_fib_res *res)
769 struct dn_hash *t = (struct dn_hash *)tb->data;
771 read_lock(&dn_fib_tables_lock);
772 for(dz = t->dh_zone_list; dz; dz = dz->dz_next) {
773 struct dn_fib_node *f;
774 dn_fib_key_t k = dz_key(flp->fld_dst, dz);
776 for(f = dz_chain(k, dz); f; f = f->fn_next) {
777 if (!dn_key_eq(k, f->fn_key)) {
778 if (dn_key_leq(k, f->fn_key))
784 f->fn_state |= DN_S_ACCESSED;
786 if (f->fn_state&DN_S_ZOMBIE)
789 if (f->fn_scope < flp->fld_scope)
792 err = dn_fib_semantic_match(f->fn_type, DN_FIB_INFO(f), flp, res);
795 res->type = f->fn_type;
796 res->scope = f->fn_scope;
797 res->prefixlen = dz->dz_order;
806 read_unlock(&dn_fib_tables_lock);
811 struct dn_fib_table *dn_fib_get_table(u32 n, int create)
813 struct dn_fib_table *t;
814 struct hlist_node *node;
817 if (n < RT_TABLE_MIN)
820 if (n > RT_TABLE_MAX)
823 h = n & (DN_FIB_TABLE_HASHSZ - 1);
825 hlist_for_each_entry_rcu(t, node, &dn_fib_table_hash[h], hlist) {
836 if (in_interrupt() && net_ratelimit()) {
837 printk(KERN_DEBUG "DECnet: BUG! Attempt to create routing table from interrupt\n");
841 t = kzalloc(sizeof(struct dn_fib_table) + sizeof(struct dn_hash),
847 t->insert = dn_fib_table_insert;
848 t->delete = dn_fib_table_delete;
849 t->lookup = dn_fib_table_lookup;
850 t->flush = dn_fib_table_flush;
851 t->dump = dn_fib_table_dump;
852 hlist_add_head_rcu(&t->hlist, &dn_fib_table_hash[h]);
857 struct dn_fib_table *dn_fib_empty_table(void)
861 for(id = RT_TABLE_MIN; id <= RT_TABLE_MAX; id++)
862 if (dn_fib_get_table(id, 0) == NULL)
863 return dn_fib_get_table(id, 1);
867 void dn_fib_flush(void)
870 struct dn_fib_table *tb;
871 struct hlist_node *node;
874 for (h = 0; h < DN_FIB_TABLE_HASHSZ; h++) {
875 hlist_for_each_entry(tb, node, &dn_fib_table_hash[h], hlist)
876 flushed += tb->flush(tb);
880 dn_rt_cache_flush(-1);
883 void __init dn_fib_table_init(void)
885 dn_hash_kmem = kmem_cache_create("dn_fib_info_cache",
886 sizeof(struct dn_fib_info),
887 0, SLAB_HWCACHE_ALIGN,
891 void __exit dn_fib_table_cleanup(void)
893 struct dn_fib_table *t;
894 struct hlist_node *node, *next;
897 write_lock(&dn_fib_tables_lock);
898 for (h = 0; h < DN_FIB_TABLE_HASHSZ; h++) {
899 hlist_for_each_entry_safe(t, node, next, &dn_fib_table_hash[h],
901 hlist_del(&t->hlist);
905 write_unlock(&dn_fib_tables_lock);