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/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 <linux/atomic.h>
28 #include <asm/uaccess.h>
29 #include <linux/route.h> /* RTF_xxx */
30 #include <net/neighbour.h>
31 #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 nlattr *attrs[], struct dn_fib_info *fi)
229 struct rtnexthop *nhp;
232 if (attrs[RTA_PRIORITY] &&
233 nla_get_u32(attrs[RTA_PRIORITY]) != fi->fib_priority)
236 if (attrs[RTA_OIF] || attrs[RTA_GATEWAY]) {
237 if ((!attrs[RTA_OIF] || nla_get_u32(attrs[RTA_OIF]) == fi->fib_nh->nh_oif) &&
238 (!attrs[RTA_GATEWAY] || nla_get_le16(attrs[RTA_GATEWAY]) != fi->fib_nh->nh_gw))
243 if (!attrs[RTA_MULTIPATH])
246 nhp = nla_data(attrs[RTA_MULTIPATH]);
247 nhlen = nla_len(attrs[RTA_MULTIPATH]);
250 int attrlen = nhlen - sizeof(struct rtnexthop);
253 if (attrlen < 0 || (nhlen -= nhp->rtnh_len) < 0)
255 if (nhp->rtnh_ifindex && nhp->rtnh_ifindex != nh->nh_oif)
258 struct nlattr *gw_attr;
260 gw_attr = nla_find((struct nlattr *) (nhp + 1), attrlen, RTA_GATEWAY);
261 gw = gw_attr ? nla_get_le16(gw_attr) : 0;
263 if (gw && gw != nh->nh_gw)
266 nhp = RTNH_NEXT(nhp);
267 } endfor_nexthops(fi);
272 static inline size_t dn_fib_nlmsg_size(struct dn_fib_info *fi)
274 size_t payload = NLMSG_ALIGN(sizeof(struct rtmsg))
275 + nla_total_size(4) /* RTA_TABLE */
276 + nla_total_size(2) /* RTA_DST */
277 + nla_total_size(4) /* RTA_PRIORITY */
278 + nla_total_size(TCP_CA_NAME_MAX); /* RTAX_CC_ALGO */
280 /* space for nested metrics */
281 payload += nla_total_size((RTAX_MAX * nla_total_size(4)));
284 /* Also handles the special case fib_nhs == 1 */
286 /* each nexthop is packed in an attribute */
287 size_t nhsize = nla_total_size(sizeof(struct rtnexthop));
289 /* may contain a gateway attribute */
290 nhsize += nla_total_size(4);
292 /* all nexthops are packed in a nested attribute */
293 payload += nla_total_size(fi->fib_nhs * nhsize);
299 static int dn_fib_dump_info(struct sk_buff *skb, u32 portid, u32 seq, int event,
300 u32 tb_id, u8 type, u8 scope, void *dst, int dst_len,
301 struct dn_fib_info *fi, unsigned int flags)
304 struct nlmsghdr *nlh;
306 nlh = nlmsg_put(skb, portid, seq, event, sizeof(*rtm), flags);
310 rtm = nlmsg_data(nlh);
311 rtm->rtm_family = AF_DECnet;
312 rtm->rtm_dst_len = dst_len;
313 rtm->rtm_src_len = 0;
315 rtm->rtm_table = tb_id;
316 rtm->rtm_flags = fi->fib_flags;
317 rtm->rtm_scope = scope;
318 rtm->rtm_type = type;
319 rtm->rtm_protocol = fi->fib_protocol;
321 if (nla_put_u32(skb, RTA_TABLE, tb_id) < 0)
324 if (rtm->rtm_dst_len &&
325 nla_put(skb, RTA_DST, 2, dst) < 0)
328 if (fi->fib_priority &&
329 nla_put_u32(skb, RTA_PRIORITY, fi->fib_priority) < 0)
332 if (rtnetlink_put_metrics(skb, fi->fib_metrics) < 0)
335 if (fi->fib_nhs == 1) {
336 if (fi->fib_nh->nh_gw &&
337 nla_put_le16(skb, RTA_GATEWAY, fi->fib_nh->nh_gw) < 0)
340 if (fi->fib_nh->nh_oif &&
341 nla_put_u32(skb, RTA_OIF, fi->fib_nh->nh_oif) < 0)
345 if (fi->fib_nhs > 1) {
346 struct rtnexthop *nhp;
347 struct nlattr *mp_head;
349 if (!(mp_head = nla_nest_start(skb, RTA_MULTIPATH)))
353 if (!(nhp = nla_reserve_nohdr(skb, sizeof(*nhp))))
356 nhp->rtnh_flags = nh->nh_flags & 0xFF;
357 nhp->rtnh_hops = nh->nh_weight - 1;
358 nhp->rtnh_ifindex = nh->nh_oif;
361 nla_put_le16(skb, RTA_GATEWAY, nh->nh_gw) < 0)
364 nhp->rtnh_len = skb_tail_pointer(skb) - (unsigned char *)nhp;
365 } endfor_nexthops(fi);
367 nla_nest_end(skb, mp_head);
374 nlmsg_cancel(skb, nlh);
379 static void dn_rtmsg_fib(int event, struct dn_fib_node *f, int z, u32 tb_id,
380 struct nlmsghdr *nlh, struct netlink_skb_parms *req)
383 u32 portid = req ? req->portid : 0;
386 skb = nlmsg_new(dn_fib_nlmsg_size(DN_FIB_INFO(f)), GFP_KERNEL);
390 err = dn_fib_dump_info(skb, portid, nlh->nlmsg_seq, event, tb_id,
391 f->fn_type, f->fn_scope, &f->fn_key, z,
394 /* -EMSGSIZE implies BUG in dn_fib_nlmsg_size() */
395 WARN_ON(err == -EMSGSIZE);
399 rtnl_notify(skb, &init_net, portid, RTNLGRP_DECnet_ROUTE, nlh, GFP_KERNEL);
403 rtnl_set_sk_err(&init_net, RTNLGRP_DECnet_ROUTE, err);
406 static __inline__ int dn_hash_dump_bucket(struct sk_buff *skb,
407 struct netlink_callback *cb,
408 struct dn_fib_table *tb,
410 struct dn_fib_node *f)
415 for(i = 0; f; i++, f = f->fn_next) {
418 if (f->fn_state & DN_S_ZOMBIE)
420 if (dn_fib_dump_info(skb, NETLINK_CB(cb->skb).portid,
424 (f->fn_state & DN_S_ZOMBIE) ? 0 : f->fn_type,
425 f->fn_scope, &f->fn_key, dz->dz_order,
426 f->fn_info, NLM_F_MULTI) < 0) {
435 static __inline__ int dn_hash_dump_zone(struct sk_buff *skb,
436 struct netlink_callback *cb,
437 struct dn_fib_table *tb,
443 for(h = 0; h < dz->dz_divisor; h++) {
447 memset(&cb->args[4], 0, sizeof(cb->args) - 4*sizeof(cb->args[0]));
448 if (dz->dz_hash == NULL || dz->dz_hash[h] == NULL)
450 if (dn_hash_dump_bucket(skb, cb, tb, dz, dz->dz_hash[h]) < 0) {
459 static int dn_fib_table_dump(struct dn_fib_table *tb, struct sk_buff *skb,
460 struct netlink_callback *cb)
464 struct dn_hash *table = (struct dn_hash *)tb->data;
467 read_lock(&dn_fib_tables_lock);
468 for(dz = table->dh_zone_list, m = 0; dz; dz = dz->dz_next, m++) {
472 memset(&cb->args[3], 0, sizeof(cb->args) - 3*sizeof(cb->args[0]));
474 if (dn_hash_dump_zone(skb, cb, tb, dz) < 0) {
476 read_unlock(&dn_fib_tables_lock);
480 read_unlock(&dn_fib_tables_lock);
486 int dn_fib_dump(struct sk_buff *skb, struct netlink_callback *cb)
488 struct net *net = sock_net(skb->sk);
490 unsigned int e = 0, s_e;
491 struct dn_fib_table *tb;
494 if (!net_eq(net, &init_net))
497 if (nlmsg_len(cb->nlh) >= sizeof(struct rtmsg) &&
498 ((struct rtmsg *)nlmsg_data(cb->nlh))->rtm_flags&RTM_F_CLONED)
499 return dn_cache_dump(skb, cb);
504 for (h = s_h; h < DN_FIB_TABLE_HASHSZ; h++, s_h = 0) {
506 hlist_for_each_entry(tb, &dn_fib_table_hash[h], hlist) {
510 memset(&cb->args[2], 0, sizeof(cb->args) -
511 2 * sizeof(cb->args[0]));
512 if (tb->dump(tb, skb, cb) < 0)
526 static int dn_fib_table_insert(struct dn_fib_table *tb, struct rtmsg *r, struct nlattr *attrs[],
527 struct nlmsghdr *n, struct netlink_skb_parms *req)
529 struct dn_hash *table = (struct dn_hash *)tb->data;
530 struct dn_fib_node *new_f, *f, **fp, **del_fp;
532 struct dn_fib_info *fi;
533 int z = r->rtm_dst_len;
534 int type = r->rtm_type;
541 dz = table->dh_zones[z];
542 if (!dz && !(dz = dn_new_zone(table, z)))
546 if (attrs[RTA_DST]) {
547 __le16 dst = nla_get_le16(attrs[RTA_DST]);
548 if (dst & ~DZ_MASK(dz))
550 key = dz_key(dst, dz);
553 if ((fi = dn_fib_create_info(r, attrs, n, &err)) == NULL)
556 if (dz->dz_nent > (dz->dz_divisor << 2) &&
557 dz->dz_divisor > DN_MAX_DIVISOR &&
558 (z==16 || (1<<z) > dz->dz_divisor))
561 fp = dn_chain_p(key, dz);
564 if (dn_key_leq(key, f->fn_key))
570 if (f && (f->fn_state & DN_S_ZOMBIE) &&
571 dn_key_eq(f->fn_key, key)) {
578 DN_FIB_SCAN_KEY(f, fp, key) {
579 if (fi->fib_priority <= DN_FIB_INFO(f)->fib_priority)
583 if (f && dn_key_eq(f->fn_key, key) &&
584 fi->fib_priority == DN_FIB_INFO(f)->fib_priority) {
585 struct dn_fib_node **ins_fp;
588 if (n->nlmsg_flags & NLM_F_EXCL)
591 if (n->nlmsg_flags & NLM_F_REPLACE) {
601 DN_FIB_SCAN_KEY(f, fp, key) {
602 if (fi->fib_priority != DN_FIB_INFO(f)->fib_priority)
604 if (f->fn_type == type &&
605 f->fn_scope == r->rtm_scope &&
606 DN_FIB_INFO(f) == fi)
610 if (!(n->nlmsg_flags & NLM_F_APPEND)) {
618 if (!(n->nlmsg_flags & NLM_F_CREATE))
623 new_f = kmem_cache_zalloc(dn_hash_kmem, GFP_KERNEL);
628 new_f->fn_type = type;
629 new_f->fn_scope = r->rtm_scope;
630 DN_FIB_INFO(new_f) = fi;
633 write_lock_bh(&dn_fib_tables_lock);
635 write_unlock_bh(&dn_fib_tables_lock);
640 write_lock_bh(&dn_fib_tables_lock);
641 *del_fp = f->fn_next;
642 write_unlock_bh(&dn_fib_tables_lock);
644 if (!(f->fn_state & DN_S_ZOMBIE))
645 dn_rtmsg_fib(RTM_DELROUTE, f, z, tb->n, n, req);
646 if (f->fn_state & DN_S_ACCESSED)
647 dn_rt_cache_flush(-1);
651 dn_rt_cache_flush(-1);
654 dn_rtmsg_fib(RTM_NEWROUTE, new_f, z, tb->n, n, req);
658 dn_fib_release_info(fi);
663 static int dn_fib_table_delete(struct dn_fib_table *tb, struct rtmsg *r, struct nlattr *attrs[],
664 struct nlmsghdr *n, struct netlink_skb_parms *req)
666 struct dn_hash *table = (struct dn_hash*)tb->data;
667 struct dn_fib_node **fp, **del_fp, *f;
668 int z = r->rtm_dst_len;
677 if ((dz = table->dh_zones[z]) == NULL)
681 if (attrs[RTA_DST]) {
682 __le16 dst = nla_get_le16(attrs[RTA_DST]);
683 if (dst & ~DZ_MASK(dz))
685 key = dz_key(dst, dz);
688 fp = dn_chain_p(key, dz);
691 if (dn_key_eq(f->fn_key, key))
693 if (dn_key_leq(key, f->fn_key))
699 DN_FIB_SCAN_KEY(f, fp, key) {
700 struct dn_fib_info *fi = DN_FIB_INFO(f);
702 if (f->fn_state & DN_S_ZOMBIE)
707 if (del_fp == NULL &&
708 (!r->rtm_type || f->fn_type == r->rtm_type) &&
709 (r->rtm_scope == RT_SCOPE_NOWHERE || f->fn_scope == r->rtm_scope) &&
711 fi->fib_protocol == r->rtm_protocol) &&
712 dn_fib_nh_match(r, n, attrs, fi) == 0)
718 dn_rtmsg_fib(RTM_DELROUTE, f, z, tb->n, n, req);
721 write_lock_bh(&dn_fib_tables_lock);
722 *del_fp = f->fn_next;
723 write_unlock_bh(&dn_fib_tables_lock);
725 if (f->fn_state & DN_S_ACCESSED)
726 dn_rt_cache_flush(-1);
730 f->fn_state |= DN_S_ZOMBIE;
731 if (f->fn_state & DN_S_ACCESSED) {
732 f->fn_state &= ~DN_S_ACCESSED;
733 dn_rt_cache_flush(-1);
735 if (++dn_fib_hash_zombies > 128)
745 static inline int dn_flush_list(struct dn_fib_node **fp, int z, struct dn_hash *table)
748 struct dn_fib_node *f;
750 while((f = *fp) != NULL) {
751 struct dn_fib_info *fi = DN_FIB_INFO(f);
753 if (fi && ((f->fn_state & DN_S_ZOMBIE) || (fi->fib_flags & RTNH_F_DEAD))) {
754 write_lock_bh(&dn_fib_tables_lock);
756 write_unlock_bh(&dn_fib_tables_lock);
768 static int dn_fib_table_flush(struct dn_fib_table *tb)
770 struct dn_hash *table = (struct dn_hash *)tb->data;
774 dn_fib_hash_zombies = 0;
775 for(dz = table->dh_zone_list; dz; dz = dz->dz_next) {
778 for(i = dz->dz_divisor-1; i >= 0; i--)
779 tmp += dn_flush_list(&dz->dz_hash[i], dz->dz_order, table);
787 static int dn_fib_table_lookup(struct dn_fib_table *tb, const struct flowidn *flp, struct dn_fib_res *res)
791 struct dn_hash *t = (struct dn_hash *)tb->data;
793 read_lock(&dn_fib_tables_lock);
794 for(dz = t->dh_zone_list; dz; dz = dz->dz_next) {
795 struct dn_fib_node *f;
796 dn_fib_key_t k = dz_key(flp->daddr, dz);
798 for(f = dz_chain(k, dz); f; f = f->fn_next) {
799 if (!dn_key_eq(k, f->fn_key)) {
800 if (dn_key_leq(k, f->fn_key))
806 f->fn_state |= DN_S_ACCESSED;
808 if (f->fn_state&DN_S_ZOMBIE)
811 if (f->fn_scope < flp->flowidn_scope)
814 err = dn_fib_semantic_match(f->fn_type, DN_FIB_INFO(f), flp, res);
817 res->type = f->fn_type;
818 res->scope = f->fn_scope;
819 res->prefixlen = dz->dz_order;
828 read_unlock(&dn_fib_tables_lock);
833 struct dn_fib_table *dn_fib_get_table(u32 n, int create)
835 struct dn_fib_table *t;
838 if (n < RT_TABLE_MIN)
841 if (n > RT_TABLE_MAX)
844 h = n & (DN_FIB_TABLE_HASHSZ - 1);
846 hlist_for_each_entry_rcu(t, &dn_fib_table_hash[h], hlist) {
857 if (in_interrupt()) {
858 net_dbg_ratelimited("DECnet: BUG! Attempt to create routing table from interrupt\n");
862 t = kzalloc(sizeof(struct dn_fib_table) + sizeof(struct dn_hash),
868 t->insert = dn_fib_table_insert;
869 t->delete = dn_fib_table_delete;
870 t->lookup = dn_fib_table_lookup;
871 t->flush = dn_fib_table_flush;
872 t->dump = dn_fib_table_dump;
873 hlist_add_head_rcu(&t->hlist, &dn_fib_table_hash[h]);
878 struct dn_fib_table *dn_fib_empty_table(void)
882 for(id = RT_TABLE_MIN; id <= RT_TABLE_MAX; id++)
883 if (dn_fib_get_table(id, 0) == NULL)
884 return dn_fib_get_table(id, 1);
888 void dn_fib_flush(void)
891 struct dn_fib_table *tb;
894 for (h = 0; h < DN_FIB_TABLE_HASHSZ; h++) {
895 hlist_for_each_entry(tb, &dn_fib_table_hash[h], hlist)
896 flushed += tb->flush(tb);
900 dn_rt_cache_flush(-1);
903 void __init dn_fib_table_init(void)
905 dn_hash_kmem = kmem_cache_create("dn_fib_info_cache",
906 sizeof(struct dn_fib_info),
907 0, SLAB_HWCACHE_ALIGN,
911 void __exit dn_fib_table_cleanup(void)
913 struct dn_fib_table *t;
914 struct hlist_node *next;
917 write_lock(&dn_fib_tables_lock);
918 for (h = 0; h < DN_FIB_TABLE_HASHSZ; h++) {
919 hlist_for_each_entry_safe(t, next, &dn_fib_table_hash[h],
921 hlist_del(&t->hlist);
925 write_unlock(&dn_fib_tables_lock);