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/config.h>
16 #include <linux/string.h>
17 #include <linux/net.h>
18 #include <linux/socket.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>
35 #include <net/dn_route.h>
36 #include <net/dn_fib.h>
37 #include <net/dn_neigh.h>
38 #include <net/dn_dev.h>
42 struct dn_zone *dz_next;
43 struct dn_fib_node **dz_hash;
47 #define DZ_HASHMASK(dz) ((dz)->dz_hashmask)
50 #define DZ_MASK(dz) ((dz)->dz_mask)
55 struct dn_zone *dh_zones[17];
56 struct dn_zone *dh_zone_list;
59 #define dz_key_0(key) ((key).datum = 0)
60 #define dz_prefix(key,dz) ((key).datum)
62 #define for_nexthops(fi) { int nhsel; const struct dn_fib_nh *nh;\
63 for(nhsel = 0, nh = (fi)->fib_nh; nhsel < (fi)->fib_nhs; nh++, nhsel++)
65 #define endfor_nexthops(fi) }
67 #define DN_MAX_DIVISOR 1024
69 #define DN_S_ACCESSED 2
71 #define DN_FIB_SCAN(f, fp) \
72 for( ; ((f) = *(fp)) != NULL; (fp) = &(f)->fn_next)
74 #define DN_FIB_SCAN_KEY(f, fp, key) \
75 for( ; ((f) = *(fp)) != NULL && dn_key_eq((f)->fn_key, (key)); (fp) = &(f)->fn_next)
77 #define RT_TABLE_MIN 1
79 static DEFINE_RWLOCK(dn_fib_tables_lock);
80 struct dn_fib_table *dn_fib_tables[RT_TABLE_MAX + 1];
82 static kmem_cache_t *dn_hash_kmem __read_mostly;
83 static int dn_fib_hash_zombies;
85 static inline dn_fib_idx_t dn_hash(dn_fib_key_t key, struct dn_zone *dz)
87 u16 h = ntohs(key.datum)>>(16 - dz->dz_order);
91 return *(dn_fib_idx_t *)&h;
94 static inline dn_fib_key_t dz_key(u16 dst, struct dn_zone *dz)
97 k.datum = dst & DZ_MASK(dz);
101 static inline struct dn_fib_node **dn_chain_p(dn_fib_key_t key, struct dn_zone *dz)
103 return &dz->dz_hash[dn_hash(key, dz).datum];
106 static inline struct dn_fib_node *dz_chain(dn_fib_key_t key, struct dn_zone *dz)
108 return dz->dz_hash[dn_hash(key, dz).datum];
111 static inline int dn_key_eq(dn_fib_key_t a, dn_fib_key_t b)
113 return a.datum == b.datum;
116 static inline int dn_key_leq(dn_fib_key_t a, dn_fib_key_t b)
118 return a.datum <= b.datum;
121 static inline void dn_rebuild_zone(struct dn_zone *dz,
122 struct dn_fib_node **old_ht,
126 struct dn_fib_node *f, **fp, *next;
128 for(i = 0; i < old_divisor; i++) {
129 for(f = old_ht[i]; f; f = f->fn_next) {
131 for(fp = dn_chain_p(f->fn_key, dz);
132 *fp && dn_key_leq((*fp)->fn_key, f->fn_key);
133 fp = &(*fp)->fn_next)
141 static void dn_rehash_zone(struct dn_zone *dz)
143 struct dn_fib_node **ht, **old_ht;
144 int old_divisor, new_divisor;
147 old_divisor = dz->dz_divisor;
149 switch(old_divisor) {
155 printk(KERN_DEBUG "DECnet: dn_rehash_zone: BUG! %d\n", old_divisor);
158 new_hashmask = 0x3FF;
162 ht = kmalloc(new_divisor*sizeof(struct dn_fib_node*), GFP_KERNEL);
167 memset(ht, 0, new_divisor*sizeof(struct dn_fib_node *));
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 = kmalloc(sizeof(struct dn_zone), GFP_KERNEL);
192 memset(dz, 0, sizeof(struct dn_zone));
195 dz->dz_hashmask = 0x0F;
201 dz->dz_hash = kmalloc(dz->dz_divisor*sizeof(struct dn_fib_node *), GFP_KERNEL);
208 memset(dz->dz_hash, 0, dz->dz_divisor*sizeof(struct dn_fib_node*));
210 dz->dz_mask = dnet_make_mask(z);
212 for(i = z + 1; i <= 16; i++)
213 if (table->dh_zones[i])
216 write_lock_bh(&dn_fib_tables_lock);
218 dz->dz_next = table->dh_zone_list;
219 table->dh_zone_list = dz;
221 dz->dz_next = table->dh_zones[i]->dz_next;
222 table->dh_zones[i]->dz_next = dz;
224 table->dh_zones[z] = dz;
225 write_unlock_bh(&dn_fib_tables_lock);
230 static int dn_fib_nh_match(struct rtmsg *r, struct nlmsghdr *nlh, struct dn_kern_rta *rta, struct dn_fib_info *fi)
232 struct rtnexthop *nhp;
235 if (rta->rta_priority && *rta->rta_priority != fi->fib_priority)
238 if (rta->rta_oif || rta->rta_gw) {
239 if ((!rta->rta_oif || *rta->rta_oif == fi->fib_nh->nh_oif) &&
240 (!rta->rta_gw || memcmp(rta->rta_gw, &fi->fib_nh->nh_gw, 2) == 0))
245 if (rta->rta_mp == NULL)
248 nhp = RTA_DATA(rta->rta_mp);
249 nhlen = RTA_PAYLOAD(rta->rta_mp);
252 int attrlen = nhlen - sizeof(struct rtnexthop);
255 if (attrlen < 0 || (nhlen -= nhp->rtnh_len) < 0)
257 if (nhp->rtnh_ifindex && nhp->rtnh_ifindex != nh->nh_oif)
260 gw = dn_fib_get_attr16(RTNH_DATA(nhp), attrlen, RTA_GATEWAY);
262 if (gw && gw != nh->nh_gw)
265 nhp = RTNH_NEXT(nhp);
266 } endfor_nexthops(fi);
271 static int dn_fib_dump_info(struct sk_buff *skb, u32 pid, u32 seq, int event,
272 u8 tb_id, u8 type, u8 scope, void *dst, int dst_len,
273 struct dn_fib_info *fi, unsigned int flags)
276 struct nlmsghdr *nlh;
277 unsigned char *b = skb->tail;
279 nlh = NLMSG_NEW(skb, pid, seq, event, sizeof(*rtm), flags);
280 rtm = NLMSG_DATA(nlh);
281 rtm->rtm_family = AF_DECnet;
282 rtm->rtm_dst_len = dst_len;
283 rtm->rtm_src_len = 0;
285 rtm->rtm_table = tb_id;
286 rtm->rtm_flags = fi->fib_flags;
287 rtm->rtm_scope = scope;
288 rtm->rtm_type = type;
289 if (rtm->rtm_dst_len)
290 RTA_PUT(skb, RTA_DST, 2, dst);
291 rtm->rtm_protocol = fi->fib_protocol;
292 if (fi->fib_priority)
293 RTA_PUT(skb, RTA_PRIORITY, 4, &fi->fib_priority);
294 if (rtnetlink_put_metrics(skb, fi->fib_metrics) < 0)
296 if (fi->fib_nhs == 1) {
297 if (fi->fib_nh->nh_gw)
298 RTA_PUT(skb, RTA_GATEWAY, 2, &fi->fib_nh->nh_gw);
299 if (fi->fib_nh->nh_oif)
300 RTA_PUT(skb, RTA_OIF, sizeof(int), &fi->fib_nh->nh_oif);
302 if (fi->fib_nhs > 1) {
303 struct rtnexthop *nhp;
304 struct rtattr *mp_head;
305 if (skb_tailroom(skb) <= RTA_SPACE(0))
307 mp_head = (struct rtattr *)skb_put(skb, RTA_SPACE(0));
310 if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4))
312 nhp = (struct rtnexthop *)skb_put(skb, RTA_ALIGN(sizeof(*nhp)));
313 nhp->rtnh_flags = nh->nh_flags & 0xFF;
314 nhp->rtnh_hops = nh->nh_weight - 1;
315 nhp->rtnh_ifindex = nh->nh_oif;
317 RTA_PUT(skb, RTA_GATEWAY, 2, &nh->nh_gw);
318 nhp->rtnh_len = skb->tail - (unsigned char *)nhp;
319 } endfor_nexthops(fi);
320 mp_head->rta_type = RTA_MULTIPATH;
321 mp_head->rta_len = skb->tail - (u8*)mp_head;
324 nlh->nlmsg_len = skb->tail - b;
330 skb_trim(skb, b - skb->data);
335 static void dn_rtmsg_fib(int event, struct dn_fib_node *f, int z, int tb_id,
336 struct nlmsghdr *nlh, struct netlink_skb_parms *req)
339 u32 pid = req ? req->pid : 0;
340 int size = NLMSG_SPACE(sizeof(struct rtmsg) + 256);
342 skb = alloc_skb(size, GFP_KERNEL);
346 if (dn_fib_dump_info(skb, pid, nlh->nlmsg_seq, event, tb_id,
347 f->fn_type, f->fn_scope, &f->fn_key, z,
348 DN_FIB_INFO(f), 0) < 0) {
352 NETLINK_CB(skb).dst_group = RTNLGRP_DECnet_ROUTE;
353 if (nlh->nlmsg_flags & NLM_F_ECHO)
354 atomic_inc(&skb->users);
355 netlink_broadcast(rtnl, skb, pid, RTNLGRP_DECnet_ROUTE, GFP_KERNEL);
356 if (nlh->nlmsg_flags & NLM_F_ECHO)
357 netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
360 static __inline__ int dn_hash_dump_bucket(struct sk_buff *skb,
361 struct netlink_callback *cb,
362 struct dn_fib_table *tb,
364 struct dn_fib_node *f)
369 for(i = 0; f; i++, f = f->fn_next) {
372 if (f->fn_state & DN_S_ZOMBIE)
374 if (dn_fib_dump_info(skb, NETLINK_CB(cb->skb).pid,
378 (f->fn_state & DN_S_ZOMBIE) ? 0 : f->fn_type,
379 f->fn_scope, &f->fn_key, dz->dz_order,
380 f->fn_info, NLM_F_MULTI) < 0) {
389 static __inline__ int dn_hash_dump_zone(struct sk_buff *skb,
390 struct netlink_callback *cb,
391 struct dn_fib_table *tb,
397 for(h = 0; h < dz->dz_divisor; h++) {
401 memset(&cb->args[3], 0, sizeof(cb->args) - 3*sizeof(cb->args[0]));
402 if (dz->dz_hash == NULL || dz->dz_hash[h] == NULL)
404 if (dn_hash_dump_bucket(skb, cb, tb, dz, dz->dz_hash[h]) < 0) {
413 static int dn_fib_table_dump(struct dn_fib_table *tb, struct sk_buff *skb,
414 struct netlink_callback *cb)
418 struct dn_hash *table = (struct dn_hash *)tb->data;
421 read_lock(&dn_fib_tables_lock);
422 for(dz = table->dh_zone_list, m = 0; dz; dz = dz->dz_next, m++) {
426 memset(&cb->args[2], 0, sizeof(cb->args) - 2*sizeof(cb->args[0]));
428 if (dn_hash_dump_zone(skb, cb, tb, dz) < 0) {
430 read_unlock(&dn_fib_tables_lock);
434 read_unlock(&dn_fib_tables_lock);
440 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)
442 struct dn_hash *table = (struct dn_hash *)tb->data;
443 struct dn_fib_node *new_f, *f, **fp, **del_fp;
445 struct dn_fib_info *fi;
446 int z = r->rtm_dst_len;
447 int type = r->rtm_type;
454 dz = table->dh_zones[z];
455 if (!dz && !(dz = dn_new_zone(table, z)))
461 memcpy(&dst, rta->rta_dst, 2);
462 if (dst & ~DZ_MASK(dz))
464 key = dz_key(dst, dz);
467 if ((fi = dn_fib_create_info(r, rta, n, &err)) == NULL)
470 if (dz->dz_nent > (dz->dz_divisor << 2) &&
471 dz->dz_divisor > DN_MAX_DIVISOR &&
472 (z==16 || (1<<z) > dz->dz_divisor))
475 fp = dn_chain_p(key, dz);
478 if (dn_key_leq(key, f->fn_key))
484 if (f && (f->fn_state & DN_S_ZOMBIE) &&
485 dn_key_eq(f->fn_key, key)) {
492 DN_FIB_SCAN_KEY(f, fp, key) {
493 if (fi->fib_priority <= DN_FIB_INFO(f)->fib_priority)
497 if (f && dn_key_eq(f->fn_key, key) &&
498 fi->fib_priority == DN_FIB_INFO(f)->fib_priority) {
499 struct dn_fib_node **ins_fp;
502 if (n->nlmsg_flags & NLM_F_EXCL)
505 if (n->nlmsg_flags & NLM_F_REPLACE) {
515 DN_FIB_SCAN_KEY(f, fp, key) {
516 if (fi->fib_priority != DN_FIB_INFO(f)->fib_priority)
518 if (f->fn_type == type && f->fn_scope == r->rtm_scope
519 && DN_FIB_INFO(f) == fi)
523 if (!(n->nlmsg_flags & NLM_F_APPEND)) {
531 if (!(n->nlmsg_flags & NLM_F_CREATE))
536 new_f = kmem_cache_alloc(dn_hash_kmem, SLAB_KERNEL);
540 memset(new_f, 0, sizeof(struct dn_fib_node));
543 new_f->fn_type = type;
544 new_f->fn_scope = r->rtm_scope;
545 DN_FIB_INFO(new_f) = fi;
548 write_lock_bh(&dn_fib_tables_lock);
550 write_unlock_bh(&dn_fib_tables_lock);
555 write_lock_bh(&dn_fib_tables_lock);
556 *del_fp = f->fn_next;
557 write_unlock_bh(&dn_fib_tables_lock);
559 if (!(f->fn_state & DN_S_ZOMBIE))
560 dn_rtmsg_fib(RTM_DELROUTE, f, z, tb->n, n, req);
561 if (f->fn_state & DN_S_ACCESSED)
562 dn_rt_cache_flush(-1);
566 dn_rt_cache_flush(-1);
569 dn_rtmsg_fib(RTM_NEWROUTE, new_f, z, tb->n, n, req);
573 dn_fib_release_info(fi);
578 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)
580 struct dn_hash *table = (struct dn_hash*)tb->data;
581 struct dn_fib_node **fp, **del_fp, *f;
582 int z = r->rtm_dst_len;
591 if ((dz = table->dh_zones[z]) == NULL)
597 memcpy(&dst, rta->rta_dst, 2);
598 if (dst & ~DZ_MASK(dz))
600 key = dz_key(dst, dz);
603 fp = dn_chain_p(key, dz);
606 if (dn_key_eq(f->fn_key, key))
608 if (dn_key_leq(key, f->fn_key))
614 DN_FIB_SCAN_KEY(f, fp, key) {
615 struct dn_fib_info *fi = DN_FIB_INFO(f);
617 if (f->fn_state & DN_S_ZOMBIE)
622 if (del_fp == NULL &&
623 (!r->rtm_type || f->fn_type == r->rtm_type) &&
624 (r->rtm_scope == RT_SCOPE_NOWHERE || f->fn_scope == r->rtm_scope) &&
626 fi->fib_protocol == r->rtm_protocol) &&
627 dn_fib_nh_match(r, n, rta, fi) == 0)
633 dn_rtmsg_fib(RTM_DELROUTE, f, z, tb->n, n, req);
636 write_lock_bh(&dn_fib_tables_lock);
637 *del_fp = f->fn_next;
638 write_unlock_bh(&dn_fib_tables_lock);
640 if (f->fn_state & DN_S_ACCESSED)
641 dn_rt_cache_flush(-1);
645 f->fn_state |= DN_S_ZOMBIE;
646 if (f->fn_state & DN_S_ACCESSED) {
647 f->fn_state &= ~DN_S_ACCESSED;
648 dn_rt_cache_flush(-1);
650 if (++dn_fib_hash_zombies > 128)
660 static inline int dn_flush_list(struct dn_fib_node **fp, int z, struct dn_hash *table)
663 struct dn_fib_node *f;
665 while((f = *fp) != NULL) {
666 struct dn_fib_info *fi = DN_FIB_INFO(f);
668 if (fi && ((f->fn_state & DN_S_ZOMBIE) || (fi->fib_flags & RTNH_F_DEAD))) {
669 write_lock_bh(&dn_fib_tables_lock);
671 write_unlock_bh(&dn_fib_tables_lock);
683 static int dn_fib_table_flush(struct dn_fib_table *tb)
685 struct dn_hash *table = (struct dn_hash *)tb->data;
689 dn_fib_hash_zombies = 0;
690 for(dz = table->dh_zone_list; dz; dz = dz->dz_next) {
693 for(i = dz->dz_divisor-1; i >= 0; i--)
694 tmp += dn_flush_list(&dz->dz_hash[i], dz->dz_order, table);
702 static int dn_fib_table_lookup(struct dn_fib_table *tb, const struct flowi *flp, struct dn_fib_res *res)
706 struct dn_hash *t = (struct dn_hash *)tb->data;
708 read_lock(&dn_fib_tables_lock);
709 for(dz = t->dh_zone_list; dz; dz = dz->dz_next) {
710 struct dn_fib_node *f;
711 dn_fib_key_t k = dz_key(flp->fld_dst, dz);
713 for(f = dz_chain(k, dz); f; f = f->fn_next) {
714 if (!dn_key_eq(k, f->fn_key)) {
715 if (dn_key_leq(k, f->fn_key))
721 f->fn_state |= DN_S_ACCESSED;
723 if (f->fn_state&DN_S_ZOMBIE)
726 if (f->fn_scope < flp->fld_scope)
729 err = dn_fib_semantic_match(f->fn_type, DN_FIB_INFO(f), flp, res);
732 res->type = f->fn_type;
733 res->scope = f->fn_scope;
734 res->prefixlen = dz->dz_order;
743 read_unlock(&dn_fib_tables_lock);
748 struct dn_fib_table *dn_fib_get_table(int n, int create)
750 struct dn_fib_table *t;
752 if (n < RT_TABLE_MIN)
755 if (n > RT_TABLE_MAX)
758 if (dn_fib_tables[n])
759 return dn_fib_tables[n];
764 if (in_interrupt() && net_ratelimit()) {
765 printk(KERN_DEBUG "DECnet: BUG! Attempt to create routing table from interrupt\n");
768 if ((t = kmalloc(sizeof(struct dn_fib_table) + sizeof(struct dn_hash), GFP_KERNEL)) == NULL)
771 memset(t, 0, sizeof(struct dn_fib_table));
774 t->insert = dn_fib_table_insert;
775 t->delete = dn_fib_table_delete;
776 t->lookup = dn_fib_table_lookup;
777 t->flush = dn_fib_table_flush;
778 t->dump = dn_fib_table_dump;
779 memset(t->data, 0, sizeof(struct dn_hash));
780 dn_fib_tables[n] = t;
785 static void dn_fib_del_tree(int n)
787 struct dn_fib_table *t;
789 write_lock(&dn_fib_tables_lock);
790 t = dn_fib_tables[n];
791 dn_fib_tables[n] = NULL;
792 write_unlock(&dn_fib_tables_lock);
797 struct dn_fib_table *dn_fib_empty_table(void)
801 for(id = RT_TABLE_MIN; id <= RT_TABLE_MAX; id++)
802 if (dn_fib_tables[id] == NULL)
803 return dn_fib_get_table(id, 1);
807 void __init dn_fib_table_init(void)
809 dn_hash_kmem = kmem_cache_create("dn_fib_info_cache",
810 sizeof(struct dn_fib_info),
811 0, SLAB_HWCACHE_ALIGN,
815 void __exit dn_fib_table_cleanup(void)
819 for (i = RT_TABLE_MIN; i <= RT_TABLE_MAX; ++i)