netns xfrm: fix "ip xfrm state|policy count" misreport
[safe/jmp/linux-2.6] / net / decnet / dn_table.c
1 /*
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.
5  *
6  *              DECnet Routing Forwarding Information Base (Routing Tables)
7  *
8  * Author:      Steve Whitehouse <SteveW@ACM.org>
9  *              Mostly copied from the IPv4 routing code
10  *
11  *
12  * Changes:
13  *
14  */
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>
32 #include <net/dst.h>
33 #include <net/flow.h>
34 #include <net/fib_rules.h>
35 #include <net/dn.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>
40
41 struct dn_zone
42 {
43         struct dn_zone          *dz_next;
44         struct dn_fib_node      **dz_hash;
45         int                     dz_nent;
46         int                     dz_divisor;
47         u32                     dz_hashmask;
48 #define DZ_HASHMASK(dz) ((dz)->dz_hashmask)
49         int                     dz_order;
50         __le16                  dz_mask;
51 #define DZ_MASK(dz)     ((dz)->dz_mask)
52 };
53
54 struct dn_hash
55 {
56         struct dn_zone  *dh_zones[17];
57         struct dn_zone  *dh_zone_list;
58 };
59
60 #define dz_key_0(key)           ((key).datum = 0)
61 #define dz_prefix(key,dz)       ((key).datum)
62
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++)
65
66 #define endfor_nexthops(fi) }
67
68 #define DN_MAX_DIVISOR 1024
69 #define DN_S_ZOMBIE 1
70 #define DN_S_ACCESSED 2
71
72 #define DN_FIB_SCAN(f, fp) \
73 for( ; ((f) = *(fp)) != NULL; (fp) = &(f)->fn_next)
74
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)
77
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);
82
83 static struct kmem_cache *dn_hash_kmem __read_mostly;
84 static int dn_fib_hash_zombies;
85
86 static inline dn_fib_idx_t dn_hash(dn_fib_key_t key, struct dn_zone *dz)
87 {
88         u16 h = le16_to_cpu(key.datum)>>(16 - dz->dz_order);
89         h ^= (h >> 10);
90         h ^= (h >> 6);
91         h &= DZ_HASHMASK(dz);
92         return *(dn_fib_idx_t *)&h;
93 }
94
95 static inline dn_fib_key_t dz_key(__le16 dst, struct dn_zone *dz)
96 {
97         dn_fib_key_t k;
98         k.datum = dst & DZ_MASK(dz);
99         return k;
100 }
101
102 static inline struct dn_fib_node **dn_chain_p(dn_fib_key_t key, struct dn_zone *dz)
103 {
104         return &dz->dz_hash[dn_hash(key, dz).datum];
105 }
106
107 static inline struct dn_fib_node *dz_chain(dn_fib_key_t key, struct dn_zone *dz)
108 {
109         return dz->dz_hash[dn_hash(key, dz).datum];
110 }
111
112 static inline int dn_key_eq(dn_fib_key_t a, dn_fib_key_t b)
113 {
114         return a.datum == b.datum;
115 }
116
117 static inline int dn_key_leq(dn_fib_key_t a, dn_fib_key_t b)
118 {
119         return a.datum <= b.datum;
120 }
121
122 static inline void dn_rebuild_zone(struct dn_zone *dz,
123                                    struct dn_fib_node **old_ht,
124                                    int old_divisor)
125 {
126         int i;
127         struct dn_fib_node *f, **fp, *next;
128
129         for(i = 0; i < old_divisor; i++) {
130                 for(f = old_ht[i]; f; f = f->fn_next) {
131                         next = 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)
135                                 /* NOTHING */;
136                         f->fn_next = *fp;
137                         *fp = f;
138                 }
139         }
140 }
141
142 static void dn_rehash_zone(struct dn_zone *dz)
143 {
144         struct dn_fib_node **ht, **old_ht;
145         int old_divisor, new_divisor;
146         u32 new_hashmask;
147
148         old_divisor = dz->dz_divisor;
149
150         switch(old_divisor) {
151                 case 16:
152                         new_divisor = 256;
153                         new_hashmask = 0xFF;
154                         break;
155                 default:
156                         printk(KERN_DEBUG "DECnet: dn_rehash_zone: BUG! %d\n", old_divisor);
157                 case 256:
158                         new_divisor = 1024;
159                         new_hashmask = 0x3FF;
160                         break;
161         }
162
163         ht = kcalloc(new_divisor, sizeof(struct dn_fib_node*), GFP_KERNEL);
164         if (ht == NULL)
165                 return;
166
167         write_lock_bh(&dn_fib_tables_lock);
168         old_ht = dz->dz_hash;
169         dz->dz_hash = ht;
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);
174         kfree(old_ht);
175 }
176
177 static void dn_free_node(struct dn_fib_node *f)
178 {
179         dn_fib_release_info(DN_FIB_INFO(f));
180         kmem_cache_free(dn_hash_kmem, f);
181 }
182
183
184 static struct dn_zone *dn_new_zone(struct dn_hash *table, int z)
185 {
186         int i;
187         struct dn_zone *dz = kzalloc(sizeof(struct dn_zone), GFP_KERNEL);
188         if (!dz)
189                 return NULL;
190
191         if (z) {
192                 dz->dz_divisor = 16;
193                 dz->dz_hashmask = 0x0F;
194         } else {
195                 dz->dz_divisor = 1;
196                 dz->dz_hashmask = 0;
197         }
198
199         dz->dz_hash = kcalloc(dz->dz_divisor, sizeof(struct dn_fib_node *), GFP_KERNEL);
200         if (!dz->dz_hash) {
201                 kfree(dz);
202                 return NULL;
203         }
204
205         dz->dz_order = z;
206         dz->dz_mask = dnet_make_mask(z);
207
208         for(i = z + 1; i <= 16; i++)
209                 if (table->dh_zones[i])
210                         break;
211
212         write_lock_bh(&dn_fib_tables_lock);
213         if (i>16) {
214                 dz->dz_next = table->dh_zone_list;
215                 table->dh_zone_list = dz;
216         } else {
217                 dz->dz_next = table->dh_zones[i]->dz_next;
218                 table->dh_zones[i]->dz_next = dz;
219         }
220         table->dh_zones[z] = dz;
221         write_unlock_bh(&dn_fib_tables_lock);
222         return dz;
223 }
224
225
226 static int dn_fib_nh_match(struct rtmsg *r, struct nlmsghdr *nlh, struct dn_kern_rta *rta, struct dn_fib_info *fi)
227 {
228         struct rtnexthop *nhp;
229         int nhlen;
230
231         if (rta->rta_priority && *rta->rta_priority != fi->fib_priority)
232                 return 1;
233
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))
237                         return 0;
238                 return 1;
239         }
240
241         if (rta->rta_mp == NULL)
242                 return 0;
243
244         nhp = RTA_DATA(rta->rta_mp);
245         nhlen = RTA_PAYLOAD(rta->rta_mp);
246
247         for_nexthops(fi) {
248                 int attrlen = nhlen - sizeof(struct rtnexthop);
249                 __le16 gw;
250
251                 if (attrlen < 0 || (nhlen -= nhp->rtnh_len) < 0)
252                         return -EINVAL;
253                 if (nhp->rtnh_ifindex && nhp->rtnh_ifindex != nh->nh_oif)
254                         return 1;
255                 if (attrlen) {
256                         gw = dn_fib_get_attr16(RTNH_DATA(nhp), attrlen, RTA_GATEWAY);
257
258                         if (gw && gw != nh->nh_gw)
259                                 return 1;
260                 }
261                 nhp = RTNH_NEXT(nhp);
262         } endfor_nexthops(fi);
263
264         return 0;
265 }
266
267 static inline size_t dn_fib_nlmsg_size(struct dn_fib_info *fi)
268 {
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 */
273
274         /* space for nested metrics */
275         payload += nla_total_size((RTAX_MAX * nla_total_size(4)));
276
277         if (fi->fib_nhs) {
278                 /* Also handles the special case fib_nhs == 1 */
279
280                 /* each nexthop is packed in an attribute */
281                 size_t nhsize = nla_total_size(sizeof(struct rtnexthop));
282
283                 /* may contain a gateway attribute */
284                 nhsize += nla_total_size(4);
285
286                 /* all nexthops are packed in a nested attribute */
287                 payload += nla_total_size(fi->fib_nhs * nhsize);
288         }
289
290         return payload;
291 }
292
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)
296 {
297         struct rtmsg *rtm;
298         struct nlmsghdr *nlh;
299         unsigned char *b = skb_tail_pointer(skb);
300
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;
306         rtm->rtm_tos = 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)
318                 goto rtattr_failure;
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);
324         }
325         if (fi->fib_nhs > 1) {
326                 struct rtnexthop *nhp;
327                 struct rtattr *mp_head;
328                 if (skb_tailroom(skb) <= RTA_SPACE(0))
329                         goto rtattr_failure;
330                 mp_head = (struct rtattr *)skb_put(skb, RTA_SPACE(0));
331
332                 for_nexthops(fi) {
333                         if (skb_tailroom(skb) < RTA_ALIGN(RTA_ALIGN(sizeof(*nhp)) + 4))
334                                 goto rtattr_failure;
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;
339                         if (nh->nh_gw)
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;
345         }
346
347         nlh->nlmsg_len = skb_tail_pointer(skb) - b;
348         return skb->len;
349
350
351 nlmsg_failure:
352 rtattr_failure:
353         nlmsg_trim(skb, b);
354         return -EMSGSIZE;
355 }
356
357
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)
360 {
361         struct sk_buff *skb;
362         u32 pid = req ? req->pid : 0;
363         int err = -ENOBUFS;
364
365         skb = nlmsg_new(dn_fib_nlmsg_size(DN_FIB_INFO(f)), GFP_KERNEL);
366         if (skb == NULL)
367                 goto errout;
368
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,
371                                DN_FIB_INFO(f), 0);
372         if (err < 0) {
373                 /* -EMSGSIZE implies BUG in dn_fib_nlmsg_size() */
374                 WARN_ON(err == -EMSGSIZE);
375                 kfree_skb(skb);
376                 goto errout;
377         }
378         rtnl_notify(skb, &init_net, pid, RTNLGRP_DECnet_ROUTE, nlh, GFP_KERNEL);
379         return;
380 errout:
381         if (err < 0)
382                 rtnl_set_sk_err(&init_net, RTNLGRP_DECnet_ROUTE, err);
383 }
384
385 static __inline__ int dn_hash_dump_bucket(struct sk_buff *skb,
386                                 struct netlink_callback *cb,
387                                 struct dn_fib_table *tb,
388                                 struct dn_zone *dz,
389                                 struct dn_fib_node *f)
390 {
391         int i, s_i;
392
393         s_i = cb->args[4];
394         for(i = 0; f; i++, f = f->fn_next) {
395                 if (i < s_i)
396                         continue;
397                 if (f->fn_state & DN_S_ZOMBIE)
398                         continue;
399                 if (dn_fib_dump_info(skb, NETLINK_CB(cb->skb).pid,
400                                 cb->nlh->nlmsg_seq,
401                                 RTM_NEWROUTE,
402                                 tb->n,
403                                 (f->fn_state & DN_S_ZOMBIE) ? 0 : f->fn_type,
404                                 f->fn_scope, &f->fn_key, dz->dz_order,
405                                 f->fn_info, NLM_F_MULTI) < 0) {
406                         cb->args[4] = i;
407                         return -1;
408                 }
409         }
410         cb->args[4] = i;
411         return skb->len;
412 }
413
414 static __inline__ int dn_hash_dump_zone(struct sk_buff *skb,
415                                 struct netlink_callback *cb,
416                                 struct dn_fib_table *tb,
417                                 struct dn_zone *dz)
418 {
419         int h, s_h;
420
421         s_h = cb->args[3];
422         for(h = 0; h < dz->dz_divisor; h++) {
423                 if (h < s_h)
424                         continue;
425                 if (h > s_h)
426                         memset(&cb->args[4], 0, sizeof(cb->args) - 4*sizeof(cb->args[0]));
427                 if (dz->dz_hash == NULL || dz->dz_hash[h] == NULL)
428                         continue;
429                 if (dn_hash_dump_bucket(skb, cb, tb, dz, dz->dz_hash[h]) < 0) {
430                         cb->args[3] = h;
431                         return -1;
432                 }
433         }
434         cb->args[3] = h;
435         return skb->len;
436 }
437
438 static int dn_fib_table_dump(struct dn_fib_table *tb, struct sk_buff *skb,
439                                 struct netlink_callback *cb)
440 {
441         int m, s_m;
442         struct dn_zone *dz;
443         struct dn_hash *table = (struct dn_hash *)tb->data;
444
445         s_m = cb->args[2];
446         read_lock(&dn_fib_tables_lock);
447         for(dz = table->dh_zone_list, m = 0; dz; dz = dz->dz_next, m++) {
448                 if (m < s_m)
449                         continue;
450                 if (m > s_m)
451                         memset(&cb->args[3], 0, sizeof(cb->args) - 3*sizeof(cb->args[0]));
452
453                 if (dn_hash_dump_zone(skb, cb, tb, dz) < 0) {
454                         cb->args[2] = m;
455                         read_unlock(&dn_fib_tables_lock);
456                         return -1;
457                 }
458         }
459         read_unlock(&dn_fib_tables_lock);
460         cb->args[2] = m;
461
462         return skb->len;
463 }
464
465 int dn_fib_dump(struct sk_buff *skb, struct netlink_callback *cb)
466 {
467         struct net *net = sock_net(skb->sk);
468         unsigned int h, s_h;
469         unsigned int e = 0, s_e;
470         struct dn_fib_table *tb;
471         struct hlist_node *node;
472         int dumped = 0;
473
474         if (!net_eq(net, &init_net))
475                 return 0;
476
477         if (NLMSG_PAYLOAD(cb->nlh, 0) >= sizeof(struct rtmsg) &&
478                 ((struct rtmsg *)NLMSG_DATA(cb->nlh))->rtm_flags&RTM_F_CLONED)
479                         return dn_cache_dump(skb, cb);
480
481         s_h = cb->args[0];
482         s_e = cb->args[1];
483
484         for (h = s_h; h < DN_FIB_TABLE_HASHSZ; h++, s_h = 0) {
485                 e = 0;
486                 hlist_for_each_entry(tb, node, &dn_fib_table_hash[h], hlist) {
487                         if (e < s_e)
488                                 goto next;
489                         if (dumped)
490                                 memset(&cb->args[2], 0, sizeof(cb->args) -
491                                                  2 * sizeof(cb->args[0]));
492                         if (tb->dump(tb, skb, cb) < 0)
493                                 goto out;
494                         dumped = 1;
495 next:
496                         e++;
497                 }
498         }
499 out:
500         cb->args[1] = e;
501         cb->args[0] = h;
502
503         return skb->len;
504 }
505
506 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 {
508         struct dn_hash *table = (struct dn_hash *)tb->data;
509         struct dn_fib_node *new_f, *f, **fp, **del_fp;
510         struct dn_zone *dz;
511         struct dn_fib_info *fi;
512         int z = r->rtm_dst_len;
513         int type = r->rtm_type;
514         dn_fib_key_t key;
515         int err;
516
517         if (z > 16)
518                 return -EINVAL;
519
520         dz = table->dh_zones[z];
521         if (!dz && !(dz = dn_new_zone(table, z)))
522                 return -ENOBUFS;
523
524         dz_key_0(key);
525         if (rta->rta_dst) {
526                 __le16 dst;
527                 memcpy(&dst, rta->rta_dst, 2);
528                 if (dst & ~DZ_MASK(dz))
529                         return -EINVAL;
530                 key = dz_key(dst, dz);
531         }
532
533         if ((fi = dn_fib_create_info(r, rta, n, &err)) == NULL)
534                 return err;
535
536         if (dz->dz_nent > (dz->dz_divisor << 2) &&
537                         dz->dz_divisor > DN_MAX_DIVISOR &&
538                         (z==16 || (1<<z) > dz->dz_divisor))
539                 dn_rehash_zone(dz);
540
541         fp = dn_chain_p(key, dz);
542
543         DN_FIB_SCAN(f, fp) {
544                 if (dn_key_leq(key, f->fn_key))
545                         break;
546         }
547
548         del_fp = NULL;
549
550         if (f && (f->fn_state & DN_S_ZOMBIE) &&
551                         dn_key_eq(f->fn_key, key)) {
552                 del_fp = fp;
553                 fp = &f->fn_next;
554                 f = *fp;
555                 goto create;
556         }
557
558         DN_FIB_SCAN_KEY(f, fp, key) {
559                 if (fi->fib_priority <= DN_FIB_INFO(f)->fib_priority)
560                         break;
561         }
562
563         if (f && dn_key_eq(f->fn_key, key) &&
564                         fi->fib_priority == DN_FIB_INFO(f)->fib_priority) {
565                 struct dn_fib_node **ins_fp;
566
567                 err = -EEXIST;
568                 if (n->nlmsg_flags & NLM_F_EXCL)
569                         goto out;
570
571                 if (n->nlmsg_flags & NLM_F_REPLACE) {
572                         del_fp = fp;
573                         fp = &f->fn_next;
574                         f = *fp;
575                         goto replace;
576                 }
577
578                 ins_fp = fp;
579                 err = -EEXIST;
580
581                 DN_FIB_SCAN_KEY(f, fp, key) {
582                         if (fi->fib_priority != DN_FIB_INFO(f)->fib_priority)
583                                 break;
584                         if (f->fn_type == type &&
585                             f->fn_scope == r->rtm_scope &&
586                             DN_FIB_INFO(f) == fi)
587                                 goto out;
588                 }
589
590                 if (!(n->nlmsg_flags & NLM_F_APPEND)) {
591                         fp = ins_fp;
592                         f = *fp;
593                 }
594         }
595
596 create:
597         err = -ENOENT;
598         if (!(n->nlmsg_flags & NLM_F_CREATE))
599                 goto out;
600
601 replace:
602         err = -ENOBUFS;
603         new_f = kmem_cache_zalloc(dn_hash_kmem, GFP_KERNEL);
604         if (new_f == NULL)
605                 goto out;
606
607         new_f->fn_key = key;
608         new_f->fn_type = type;
609         new_f->fn_scope = r->rtm_scope;
610         DN_FIB_INFO(new_f) = fi;
611
612         new_f->fn_next = f;
613         write_lock_bh(&dn_fib_tables_lock);
614         *fp = new_f;
615         write_unlock_bh(&dn_fib_tables_lock);
616         dz->dz_nent++;
617
618         if (del_fp) {
619                 f = *del_fp;
620                 write_lock_bh(&dn_fib_tables_lock);
621                 *del_fp = f->fn_next;
622                 write_unlock_bh(&dn_fib_tables_lock);
623
624                 if (!(f->fn_state & DN_S_ZOMBIE))
625                         dn_rtmsg_fib(RTM_DELROUTE, f, z, tb->n, n, req);
626                 if (f->fn_state & DN_S_ACCESSED)
627                         dn_rt_cache_flush(-1);
628                 dn_free_node(f);
629                 dz->dz_nent--;
630         } else {
631                 dn_rt_cache_flush(-1);
632         }
633
634         dn_rtmsg_fib(RTM_NEWROUTE, new_f, z, tb->n, n, req);
635
636         return 0;
637 out:
638         dn_fib_release_info(fi);
639         return err;
640 }
641
642
643 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)
644 {
645         struct dn_hash *table = (struct dn_hash*)tb->data;
646         struct dn_fib_node **fp, **del_fp, *f;
647         int z = r->rtm_dst_len;
648         struct dn_zone *dz;
649         dn_fib_key_t key;
650         int matched;
651
652
653         if (z > 16)
654                 return -EINVAL;
655
656         if ((dz = table->dh_zones[z]) == NULL)
657                 return -ESRCH;
658
659         dz_key_0(key);
660         if (rta->rta_dst) {
661                 __le16 dst;
662                 memcpy(&dst, rta->rta_dst, 2);
663                 if (dst & ~DZ_MASK(dz))
664                         return -EINVAL;
665                 key = dz_key(dst, dz);
666         }
667
668         fp = dn_chain_p(key, dz);
669
670         DN_FIB_SCAN(f, fp) {
671                 if (dn_key_eq(f->fn_key, key))
672                         break;
673                 if (dn_key_leq(key, f->fn_key))
674                         return -ESRCH;
675         }
676
677         matched = 0;
678         del_fp = NULL;
679         DN_FIB_SCAN_KEY(f, fp, key) {
680                 struct dn_fib_info *fi = DN_FIB_INFO(f);
681
682                 if (f->fn_state & DN_S_ZOMBIE)
683                         return -ESRCH;
684
685                 matched++;
686
687                 if (del_fp == NULL &&
688                                 (!r->rtm_type || f->fn_type == r->rtm_type) &&
689                                 (r->rtm_scope == RT_SCOPE_NOWHERE || f->fn_scope == r->rtm_scope) &&
690                                 (!r->rtm_protocol ||
691                                         fi->fib_protocol == r->rtm_protocol) &&
692                                 dn_fib_nh_match(r, n, rta, fi) == 0)
693                         del_fp = fp;
694         }
695
696         if (del_fp) {
697                 f = *del_fp;
698                 dn_rtmsg_fib(RTM_DELROUTE, f, z, tb->n, n, req);
699
700                 if (matched != 1) {
701                         write_lock_bh(&dn_fib_tables_lock);
702                         *del_fp = f->fn_next;
703                         write_unlock_bh(&dn_fib_tables_lock);
704
705                         if (f->fn_state & DN_S_ACCESSED)
706                                 dn_rt_cache_flush(-1);
707                         dn_free_node(f);
708                         dz->dz_nent--;
709                 } else {
710                         f->fn_state |= DN_S_ZOMBIE;
711                         if (f->fn_state & DN_S_ACCESSED) {
712                                 f->fn_state &= ~DN_S_ACCESSED;
713                                 dn_rt_cache_flush(-1);
714                         }
715                         if (++dn_fib_hash_zombies > 128)
716                                 dn_fib_flush();
717                 }
718
719                 return 0;
720         }
721
722         return -ESRCH;
723 }
724
725 static inline int dn_flush_list(struct dn_fib_node **fp, int z, struct dn_hash *table)
726 {
727         int found = 0;
728         struct dn_fib_node *f;
729
730         while((f = *fp) != NULL) {
731                 struct dn_fib_info *fi = DN_FIB_INFO(f);
732
733                 if (fi && ((f->fn_state & DN_S_ZOMBIE) || (fi->fib_flags & RTNH_F_DEAD))) {
734                         write_lock_bh(&dn_fib_tables_lock);
735                         *fp = f->fn_next;
736                         write_unlock_bh(&dn_fib_tables_lock);
737
738                         dn_free_node(f);
739                         found++;
740                         continue;
741                 }
742                 fp = &f->fn_next;
743         }
744
745         return found;
746 }
747
748 static int dn_fib_table_flush(struct dn_fib_table *tb)
749 {
750         struct dn_hash *table = (struct dn_hash *)tb->data;
751         struct dn_zone *dz;
752         int found = 0;
753
754         dn_fib_hash_zombies = 0;
755         for(dz = table->dh_zone_list; dz; dz = dz->dz_next) {
756                 int i;
757                 int tmp = 0;
758                 for(i = dz->dz_divisor-1; i >= 0; i--)
759                         tmp += dn_flush_list(&dz->dz_hash[i], dz->dz_order, table);
760                 dz->dz_nent -= tmp;
761                 found += tmp;
762         }
763
764         return found;
765 }
766
767 static int dn_fib_table_lookup(struct dn_fib_table *tb, const struct flowi *flp, struct dn_fib_res *res)
768 {
769         int err;
770         struct dn_zone *dz;
771         struct dn_hash *t = (struct dn_hash *)tb->data;
772
773         read_lock(&dn_fib_tables_lock);
774         for(dz = t->dh_zone_list; dz; dz = dz->dz_next) {
775                 struct dn_fib_node *f;
776                 dn_fib_key_t k = dz_key(flp->fld_dst, dz);
777
778                 for(f = dz_chain(k, dz); f; f = f->fn_next) {
779                         if (!dn_key_eq(k, f->fn_key)) {
780                                 if (dn_key_leq(k, f->fn_key))
781                                         break;
782                                 else
783                                         continue;
784                         }
785
786                         f->fn_state |= DN_S_ACCESSED;
787
788                         if (f->fn_state&DN_S_ZOMBIE)
789                                 continue;
790
791                         if (f->fn_scope < flp->fld_scope)
792                                 continue;
793
794                         err = dn_fib_semantic_match(f->fn_type, DN_FIB_INFO(f), flp, res);
795
796                         if (err == 0) {
797                                 res->type = f->fn_type;
798                                 res->scope = f->fn_scope;
799                                 res->prefixlen = dz->dz_order;
800                                 goto out;
801                         }
802                         if (err < 0)
803                                 goto out;
804                 }
805         }
806         err = 1;
807 out:
808         read_unlock(&dn_fib_tables_lock);
809         return err;
810 }
811
812
813 struct dn_fib_table *dn_fib_get_table(u32 n, int create)
814 {
815         struct dn_fib_table *t;
816         struct hlist_node *node;
817         unsigned int h;
818
819         if (n < RT_TABLE_MIN)
820                 return NULL;
821
822         if (n > RT_TABLE_MAX)
823                 return NULL;
824
825         h = n & (DN_FIB_TABLE_HASHSZ - 1);
826         rcu_read_lock();
827         hlist_for_each_entry_rcu(t, node, &dn_fib_table_hash[h], hlist) {
828                 if (t->n == n) {
829                         rcu_read_unlock();
830                         return t;
831                 }
832         }
833         rcu_read_unlock();
834
835         if (!create)
836                 return NULL;
837
838         if (in_interrupt() && net_ratelimit()) {
839                 printk(KERN_DEBUG "DECnet: BUG! Attempt to create routing table from interrupt\n");
840                 return NULL;
841         }
842
843         t = kzalloc(sizeof(struct dn_fib_table) + sizeof(struct dn_hash),
844                     GFP_KERNEL);
845         if (t == NULL)
846                 return NULL;
847
848         t->n = n;
849         t->insert = dn_fib_table_insert;
850         t->delete = dn_fib_table_delete;
851         t->lookup = dn_fib_table_lookup;
852         t->flush  = dn_fib_table_flush;
853         t->dump = dn_fib_table_dump;
854         hlist_add_head_rcu(&t->hlist, &dn_fib_table_hash[h]);
855
856         return t;
857 }
858
859 struct dn_fib_table *dn_fib_empty_table(void)
860 {
861         u32 id;
862
863         for(id = RT_TABLE_MIN; id <= RT_TABLE_MAX; id++)
864                 if (dn_fib_get_table(id, 0) == NULL)
865                         return dn_fib_get_table(id, 1);
866         return NULL;
867 }
868
869 void dn_fib_flush(void)
870 {
871         int flushed = 0;
872         struct dn_fib_table *tb;
873         struct hlist_node *node;
874         unsigned int h;
875
876         for (h = 0; h < DN_FIB_TABLE_HASHSZ; h++) {
877                 hlist_for_each_entry(tb, node, &dn_fib_table_hash[h], hlist)
878                         flushed += tb->flush(tb);
879         }
880
881         if (flushed)
882                 dn_rt_cache_flush(-1);
883 }
884
885 void __init dn_fib_table_init(void)
886 {
887         dn_hash_kmem = kmem_cache_create("dn_fib_info_cache",
888                                         sizeof(struct dn_fib_info),
889                                         0, SLAB_HWCACHE_ALIGN,
890                                         NULL);
891 }
892
893 void __exit dn_fib_table_cleanup(void)
894 {
895         struct dn_fib_table *t;
896         struct hlist_node *node, *next;
897         unsigned int h;
898
899         write_lock(&dn_fib_tables_lock);
900         for (h = 0; h < DN_FIB_TABLE_HASHSZ; h++) {
901                 hlist_for_each_entry_safe(t, node, next, &dn_fib_table_hash[h],
902                                           hlist) {
903                         hlist_del(&t->hlist);
904                         kfree(t);
905                 }
906         }
907         write_unlock(&dn_fib_tables_lock);
908 }