File: /usr/src/linux/net/ipv4/fib_hash.c
1 /*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * IPv4 FIB: lookup engine and maintenance routines.
7 *
8 * Version: $Id: fib_hash.c,v 1.12 1999/08/31 07:03:27 davem Exp $
9 *
10 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
11 *
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
16 */
17
18 #include <linux/config.h>
19 #include <asm/uaccess.h>
20 #include <asm/system.h>
21 #include <asm/bitops.h>
22 #include <linux/types.h>
23 #include <linux/kernel.h>
24 #include <linux/sched.h>
25 #include <linux/mm.h>
26 #include <linux/string.h>
27 #include <linux/socket.h>
28 #include <linux/sockios.h>
29 #include <linux/errno.h>
30 #include <linux/in.h>
31 #include <linux/inet.h>
32 #include <linux/netdevice.h>
33 #include <linux/if_arp.h>
34 #include <linux/proc_fs.h>
35 #include <linux/skbuff.h>
36 #include <linux/netlink.h>
37 #include <linux/init.h>
38
39 #include <net/ip.h>
40 #include <net/protocol.h>
41 #include <net/route.h>
42 #include <net/tcp.h>
43 #include <net/sock.h>
44 #include <net/ip_fib.h>
45
46 #define FTprint(a...)
47 /*
48 printk(KERN_DEBUG a)
49 */
50
51 static kmem_cache_t * fn_hash_kmem;
52
53 /*
54 These bizarre types are just to force strict type checking.
55 When I reversed order of bytes and changed to natural mask lengths,
56 I forgot to make fixes in several places. Now I am lazy to return
57 it back.
58 */
59
60 typedef struct {
61 u32 datum;
62 } fn_key_t;
63
64 typedef struct {
65 u32 datum;
66 } fn_hash_idx_t;
67
68 struct fib_node
69 {
70 struct fib_node *fn_next;
71 struct fib_info *fn_info;
72 #define FIB_INFO(f) ((f)->fn_info)
73 fn_key_t fn_key;
74 u8 fn_tos;
75 u8 fn_type;
76 u8 fn_scope;
77 u8 fn_state;
78 };
79
80 #define FN_S_ZOMBIE 1
81 #define FN_S_ACCESSED 2
82
83 static int fib_hash_zombies;
84
85 struct fn_zone
86 {
87 struct fn_zone *fz_next; /* Next not empty zone */
88 struct fib_node **fz_hash; /* Hash table pointer */
89 int fz_nent; /* Number of entries */
90
91 int fz_divisor; /* Hash divisor */
92 u32 fz_hashmask; /* (1<<fz_divisor) - 1 */
93 #define FZ_HASHMASK(fz) ((fz)->fz_hashmask)
94
95 int fz_order; /* Zone order */
96 u32 fz_mask;
97 #define FZ_MASK(fz) ((fz)->fz_mask)
98 };
99
100 /* NOTE. On fast computers evaluation of fz_hashmask and fz_mask
101 can be cheaper than memory lookup, so that FZ_* macros are used.
102 */
103
104 struct fn_hash
105 {
106 struct fn_zone *fn_zones[33];
107 struct fn_zone *fn_zone_list;
108 };
109
110 static __inline__ fn_hash_idx_t fn_hash(fn_key_t key, struct fn_zone *fz)
111 {
112 u32 h = ntohl(key.datum)>>(32 - fz->fz_order);
113 h ^= (h>>20);
114 h ^= (h>>10);
115 h ^= (h>>5);
116 h &= FZ_HASHMASK(fz);
117 return *(fn_hash_idx_t*)&h;
118 }
119
120 #define fz_key_0(key) ((key).datum = 0)
121 #define fz_prefix(key,fz) ((key).datum)
122
123 static __inline__ fn_key_t fz_key(u32 dst, struct fn_zone *fz)
124 {
125 fn_key_t k;
126 k.datum = dst & FZ_MASK(fz);
127 return k;
128 }
129
130 static __inline__ struct fib_node ** fz_chain_p(fn_key_t key, struct fn_zone *fz)
131 {
132 return &fz->fz_hash[fn_hash(key, fz).datum];
133 }
134
135 static __inline__ struct fib_node * fz_chain(fn_key_t key, struct fn_zone *fz)
136 {
137 return fz->fz_hash[fn_hash(key, fz).datum];
138 }
139
140 extern __inline__ int fn_key_eq(fn_key_t a, fn_key_t b)
141 {
142 return a.datum == b.datum;
143 }
144
145 extern __inline__ int fn_key_leq(fn_key_t a, fn_key_t b)
146 {
147 return a.datum <= b.datum;
148 }
149
150 static rwlock_t fib_hash_lock = RW_LOCK_UNLOCKED;
151
152 #define FZ_MAX_DIVISOR 1024
153
154 #ifdef CONFIG_IP_ROUTE_LARGE_TABLES
155
156 /* The fib hash lock must be held when this is called. */
157 static __inline__ void fn_rebuild_zone(struct fn_zone *fz,
158 struct fib_node **old_ht,
159 int old_divisor)
160 {
161 int i;
162 struct fib_node *f, **fp, *next;
163
164 for (i=0; i<old_divisor; i++) {
165 for (f=old_ht[i]; f; f=next) {
166 next = f->fn_next;
167 for (fp = fz_chain_p(f->fn_key, fz);
168 *fp && fn_key_leq((*fp)->fn_key, f->fn_key);
169 fp = &(*fp)->fn_next)
170 /* NONE */;
171 f->fn_next = *fp;
172 *fp = f;
173 }
174 }
175 }
176
177 static void fn_rehash_zone(struct fn_zone *fz)
178 {
179 struct fib_node **ht, **old_ht;
180 int old_divisor, new_divisor;
181 u32 new_hashmask;
182
183 old_divisor = fz->fz_divisor;
184
185 switch (old_divisor) {
186 case 16:
187 new_divisor = 256;
188 new_hashmask = 0xFF;
189 break;
190 case 256:
191 new_divisor = 1024;
192 new_hashmask = 0x3FF;
193 break;
194 default:
195 printk(KERN_CRIT "route.c: bad divisor %d!\n", old_divisor);
196 return;
197 }
198 #if RT_CACHE_DEBUG >= 2
199 printk("fn_rehash_zone: hash for zone %d grows from %d\n", fz->fz_order, old_divisor);
200 #endif
201
202 ht = kmalloc(new_divisor*sizeof(struct fib_node*), GFP_KERNEL);
203
204 if (ht) {
205 memset(ht, 0, new_divisor*sizeof(struct fib_node*));
206 write_lock_bh(&fib_hash_lock);
207 old_ht = fz->fz_hash;
208 fz->fz_hash = ht;
209 fz->fz_hashmask = new_hashmask;
210 fz->fz_divisor = new_divisor;
211 fn_rebuild_zone(fz, old_ht, old_divisor);
212 write_unlock_bh(&fib_hash_lock);
213 kfree(old_ht);
214 }
215 }
216 #endif /* CONFIG_IP_ROUTE_LARGE_TABLES */
217
218 static void fn_free_node(struct fib_node * f)
219 {
220 fib_release_info(FIB_INFO(f));
221 kmem_cache_free(fn_hash_kmem, f);
222 }
223
224
225 static struct fn_zone *
226 fn_new_zone(struct fn_hash *table, int z)
227 {
228 int i;
229 struct fn_zone *fz = kmalloc(sizeof(struct fn_zone), GFP_KERNEL);
230 if (!fz)
231 return NULL;
232
233 memset(fz, 0, sizeof(struct fn_zone));
234 if (z) {
235 fz->fz_divisor = 16;
236 fz->fz_hashmask = 0xF;
237 } else {
238 fz->fz_divisor = 1;
239 fz->fz_hashmask = 0;
240 }
241 fz->fz_hash = kmalloc(fz->fz_divisor*sizeof(struct fib_node*), GFP_KERNEL);
242 if (!fz->fz_hash) {
243 kfree(fz);
244 return NULL;
245 }
246 memset(fz->fz_hash, 0, fz->fz_divisor*sizeof(struct fib_node*));
247 fz->fz_order = z;
248 fz->fz_mask = inet_make_mask(z);
249
250 /* Find the first not empty zone with more specific mask */
251 for (i=z+1; i<=32; i++)
252 if (table->fn_zones[i])
253 break;
254 write_lock_bh(&fib_hash_lock);
255 if (i>32) {
256 /* No more specific masks, we are the first. */
257 fz->fz_next = table->fn_zone_list;
258 table->fn_zone_list = fz;
259 } else {
260 fz->fz_next = table->fn_zones[i]->fz_next;
261 table->fn_zones[i]->fz_next = fz;
262 }
263 table->fn_zones[z] = fz;
264 write_unlock_bh(&fib_hash_lock);
265 return fz;
266 }
267
268 static int
269 fn_hash_lookup(struct fib_table *tb, const struct rt_key *key, struct fib_result *res)
270 {
271 int err;
272 struct fn_zone *fz;
273 struct fn_hash *t = (struct fn_hash*)tb->tb_data;
274
275 read_lock(&fib_hash_lock);
276 for (fz = t->fn_zone_list; fz; fz = fz->fz_next) {
277 struct fib_node *f;
278 fn_key_t k = fz_key(key->dst, fz);
279
280 for (f = fz_chain(k, fz); f; f = f->fn_next) {
281 if (!fn_key_eq(k, f->fn_key)) {
282 if (fn_key_leq(k, f->fn_key))
283 break;
284 else
285 continue;
286 }
287 #ifdef CONFIG_IP_ROUTE_TOS
288 if (f->fn_tos && f->fn_tos != key->tos)
289 continue;
290 #endif
291 f->fn_state |= FN_S_ACCESSED;
292
293 if (f->fn_state&FN_S_ZOMBIE)
294 continue;
295 if (f->fn_scope < key->scope)
296 continue;
297
298 err = fib_semantic_match(f->fn_type, FIB_INFO(f), key, res);
299 if (err == 0) {
300 res->type = f->fn_type;
301 res->scope = f->fn_scope;
302 res->prefixlen = fz->fz_order;
303 goto out;
304 }
305 if (err < 0)
306 goto out;
307 }
308 }
309 err = 1;
310 out:
311 read_unlock(&fib_hash_lock);
312 return err;
313 }
314
315 static int fn_hash_last_dflt=-1;
316
317 static int fib_detect_death(struct fib_info *fi, int order,
318 struct fib_info **last_resort, int *last_idx)
319 {
320 struct neighbour *n;
321 int state = NUD_NONE;
322
323 n = neigh_lookup(&arp_tbl, &fi->fib_nh[0].nh_gw, fi->fib_dev);
324 if (n) {
325 state = n->nud_state;
326 neigh_release(n);
327 }
328 if (state==NUD_REACHABLE)
329 return 0;
330 if ((state&NUD_VALID) && order != fn_hash_last_dflt)
331 return 0;
332 if ((state&NUD_VALID) ||
333 (*last_idx<0 && order > fn_hash_last_dflt)) {
334 *last_resort = fi;
335 *last_idx = order;
336 }
337 return 1;
338 }
339
340 static void
341 fn_hash_select_default(struct fib_table *tb, const struct rt_key *key, struct fib_result *res)
342 {
343 int order, last_idx;
344 struct fib_node *f;
345 struct fib_info *fi = NULL;
346 struct fib_info *last_resort;
347 struct fn_hash *t = (struct fn_hash*)tb->tb_data;
348 struct fn_zone *fz = t->fn_zones[0];
349
350 if (fz == NULL)
351 return;
352
353 last_idx = -1;
354 last_resort = NULL;
355 order = -1;
356
357 read_lock(&fib_hash_lock);
358 for (f = fz->fz_hash[0]; f; f = f->fn_next) {
359 struct fib_info *next_fi = FIB_INFO(f);
360
361 if ((f->fn_state&FN_S_ZOMBIE) ||
362 f->fn_scope != res->scope ||
363 f->fn_type != RTN_UNICAST)
364 continue;
365
366 if (next_fi->fib_priority > res->fi->fib_priority)
367 break;
368 if (!next_fi->fib_nh[0].nh_gw || next_fi->fib_nh[0].nh_scope != RT_SCOPE_LINK)
369 continue;
370 f->fn_state |= FN_S_ACCESSED;
371
372 if (fi == NULL) {
373 if (next_fi != res->fi)
374 break;
375 } else if (!fib_detect_death(fi, order, &last_resort, &last_idx)) {
376 if (res->fi)
377 fib_info_put(res->fi);
378 res->fi = fi;
379 atomic_inc(&fi->fib_clntref);
380 fn_hash_last_dflt = order;
381 goto out;
382 }
383 fi = next_fi;
384 order++;
385 }
386
387 if (order<=0 || fi==NULL) {
388 fn_hash_last_dflt = -1;
389 goto out;
390 }
391
392 if (!fib_detect_death(fi, order, &last_resort, &last_idx)) {
393 if (res->fi)
394 fib_info_put(res->fi);
395 res->fi = fi;
396 atomic_inc(&fi->fib_clntref);
397 fn_hash_last_dflt = order;
398 goto out;
399 }
400
401 if (last_idx >= 0) {
402 if (res->fi)
403 fib_info_put(res->fi);
404 res->fi = last_resort;
405 if (last_resort)
406 atomic_inc(&last_resort->fib_clntref);
407 }
408 fn_hash_last_dflt = last_idx;
409 out:
410 read_unlock(&fib_hash_lock);
411 }
412
413 #define FIB_SCAN(f, fp) \
414 for ( ; ((f) = *(fp)) != NULL; (fp) = &(f)->fn_next)
415
416 #define FIB_SCAN_KEY(f, fp, key) \
417 for ( ; ((f) = *(fp)) != NULL && fn_key_eq((f)->fn_key, (key)); (fp) = &(f)->fn_next)
418
419 #ifndef CONFIG_IP_ROUTE_TOS
420 #define FIB_SCAN_TOS(f, fp, key, tos) FIB_SCAN_KEY(f, fp, key)
421 #else
422 #define FIB_SCAN_TOS(f, fp, key, tos) \
423 for ( ; ((f) = *(fp)) != NULL && fn_key_eq((f)->fn_key, (key)) && \
424 (f)->fn_tos == (tos) ; (fp) = &(f)->fn_next)
425 #endif
426
427
428 #ifdef CONFIG_RTNETLINK
429 static void rtmsg_fib(int, struct fib_node*, int, int,
430 struct nlmsghdr *n,
431 struct netlink_skb_parms *);
432 #else
433 #define rtmsg_fib(a, b, c, d, e, f)
434 #endif
435
436
437 static int
438 fn_hash_insert(struct fib_table *tb, struct rtmsg *r, struct kern_rta *rta,
439 struct nlmsghdr *n, struct netlink_skb_parms *req)
440 {
441 struct fn_hash *table = (struct fn_hash*)tb->tb_data;
442 struct fib_node *new_f, *f, **fp, **del_fp;
443 struct fn_zone *fz;
444 struct fib_info *fi;
445
446 int z = r->rtm_dst_len;
447 int type = r->rtm_type;
448 #ifdef CONFIG_IP_ROUTE_TOS
449 u8 tos = r->rtm_tos;
450 #endif
451 fn_key_t key;
452 int err;
453
454 FTprint("tb(%d)_insert: %d %08x/%d %d %08x\n", tb->tb_id, r->rtm_type, rta->rta_dst ?
455 *(u32*)rta->rta_dst : 0, z, rta->rta_oif ? *rta->rta_oif : -1,
456 rta->rta_prefsrc ? *(u32*)rta->rta_prefsrc : 0);
457 if (z > 32)
458 return -EINVAL;
459 fz = table->fn_zones[z];
460 if (!fz && !(fz = fn_new_zone(table, z)))
461 return -ENOBUFS;
462
463 fz_key_0(key);
464 if (rta->rta_dst) {
465 u32 dst;
466 memcpy(&dst, rta->rta_dst, 4);
467 if (dst & ~FZ_MASK(fz))
468 return -EINVAL;
469 key = fz_key(dst, fz);
470 }
471
472 if ((fi = fib_create_info(r, rta, n, &err)) == NULL)
473 return err;
474
475 #ifdef CONFIG_IP_ROUTE_LARGE_TABLES
476 if (fz->fz_nent > (fz->fz_divisor<<2) &&
477 fz->fz_divisor < FZ_MAX_DIVISOR &&
478 (z==32 || (1<<z) > fz->fz_divisor))
479 fn_rehash_zone(fz);
480 #endif
481
482 fp = fz_chain_p(key, fz);
483
484
485 /*
486 * Scan list to find the first route with the same destination
487 */
488 FIB_SCAN(f, fp) {
489 if (fn_key_leq(key,f->fn_key))
490 break;
491 }
492
493 #ifdef CONFIG_IP_ROUTE_TOS
494 /*
495 * Find route with the same destination and tos.
496 */
497 FIB_SCAN_KEY(f, fp, key) {
498 if (f->fn_tos <= tos)
499 break;
500 }
501 #endif
502
503 del_fp = NULL;
504
505 if (f && (f->fn_state&FN_S_ZOMBIE) &&
506 #ifdef CONFIG_IP_ROUTE_TOS
507 f->fn_tos == tos &&
508 #endif
509 fn_key_eq(f->fn_key, key)) {
510 del_fp = fp;
511 fp = &f->fn_next;
512 f = *fp;
513 goto create;
514 }
515
516 FIB_SCAN_TOS(f, fp, key, tos) {
517 if (fi->fib_priority <= FIB_INFO(f)->fib_priority)
518 break;
519 }
520
521 /* Now f==*fp points to the first node with the same
522 keys [prefix,tos,priority], if such key already
523 exists or to the node, before which we will insert new one.
524 */
525
526 if (f &&
527 #ifdef CONFIG_IP_ROUTE_TOS
528 f->fn_tos == tos &&
529 #endif
530 fn_key_eq(f->fn_key, key) &&
531 fi->fib_priority == FIB_INFO(f)->fib_priority) {
532 struct fib_node **ins_fp;
533
534 err = -EEXIST;
535 if (n->nlmsg_flags&NLM_F_EXCL)
536 goto out;
537
538 if (n->nlmsg_flags&NLM_F_REPLACE) {
539 del_fp = fp;
540 fp = &f->fn_next;
541 f = *fp;
542 goto replace;
543 }
544
545 ins_fp = fp;
546 err = -EEXIST;
547
548 FIB_SCAN_TOS(f, fp, key, tos) {
549 if (fi->fib_priority != FIB_INFO(f)->fib_priority)
550 break;
551 if (f->fn_type == type && f->fn_scope == r->rtm_scope
552 && FIB_INFO(f) == fi)
553 goto out;
554 }
555
556 if (!(n->nlmsg_flags&NLM_F_APPEND)) {
557 fp = ins_fp;
558 f = *fp;
559 }
560 }
561
562 create:
563 err = -ENOENT;
564 if (!(n->nlmsg_flags&NLM_F_CREATE))
565 goto out;
566
567 replace:
568 err = -ENOBUFS;
569 new_f = kmem_cache_alloc(fn_hash_kmem, SLAB_KERNEL);
570 if (new_f == NULL)
571 goto out;
572
573 memset(new_f, 0, sizeof(struct fib_node));
574
575 new_f->fn_key = key;
576 #ifdef CONFIG_IP_ROUTE_TOS
577 new_f->fn_tos = tos;
578 #endif
579 new_f->fn_type = type;
580 new_f->fn_scope = r->rtm_scope;
581 FIB_INFO(new_f) = fi;
582
583 /*
584 * Insert new entry to the list.
585 */
586
587 new_f->fn_next = f;
588 write_lock_bh(&fib_hash_lock);
589 *fp = new_f;
590 write_unlock_bh(&fib_hash_lock);
591 fz->fz_nent++;
592
593 if (del_fp) {
594 f = *del_fp;
595 /* Unlink replaced node */
596 write_lock_bh(&fib_hash_lock);
597 *del_fp = f->fn_next;
598 write_unlock_bh(&fib_hash_lock);
599
600 if (!(f->fn_state&FN_S_ZOMBIE))
601 rtmsg_fib(RTM_DELROUTE, f, z, tb->tb_id, n, req);
602 if (f->fn_state&FN_S_ACCESSED)
603 rt_cache_flush(-1);
604 fn_free_node(f);
605 fz->fz_nent--;
606 } else {
607 rt_cache_flush(-1);
608 }
609 rtmsg_fib(RTM_NEWROUTE, new_f, z, tb->tb_id, n, req);
610 return 0;
611
612 out:
613 fib_release_info(fi);
614 return err;
615 }
616
617
618 static int
619 fn_hash_delete(struct fib_table *tb, struct rtmsg *r, struct kern_rta *rta,
620 struct nlmsghdr *n, struct netlink_skb_parms *req)
621 {
622 struct fn_hash *table = (struct fn_hash*)tb->tb_data;
623 struct fib_node **fp, **del_fp, *f;
624 int z = r->rtm_dst_len;
625 struct fn_zone *fz;
626 fn_key_t key;
627 int matched;
628 #ifdef CONFIG_IP_ROUTE_TOS
629 u8 tos = r->rtm_tos;
630 #endif
631
632 FTprint("tb(%d)_delete: %d %08x/%d %d\n", tb->tb_id, r->rtm_type, rta->rta_dst ?
633 *(u32*)rta->rta_dst : 0, z, rta->rta_oif ? *rta->rta_oif : -1);
634 if (z > 32)
635 return -EINVAL;
636 if ((fz = table->fn_zones[z]) == NULL)
637 return -ESRCH;
638
639 fz_key_0(key);
640 if (rta->rta_dst) {
641 u32 dst;
642 memcpy(&dst, rta->rta_dst, 4);
643 if (dst & ~FZ_MASK(fz))
644 return -EINVAL;
645 key = fz_key(dst, fz);
646 }
647
648 fp = fz_chain_p(key, fz);
649
650
651 FIB_SCAN(f, fp) {
652 if (fn_key_eq(f->fn_key, key))
653 break;
654 if (fn_key_leq(key, f->fn_key)) {
655 return -ESRCH;
656 }
657 }
658 #ifdef CONFIG_IP_ROUTE_TOS
659 FIB_SCAN_KEY(f, fp, key) {
660 if (f->fn_tos == tos)
661 break;
662 }
663 #endif
664
665 matched = 0;
666 del_fp = NULL;
667 FIB_SCAN_TOS(f, fp, key, tos) {
668 struct fib_info * fi = FIB_INFO(f);
669
670 if (f->fn_state&FN_S_ZOMBIE) {
671 return -ESRCH;
672 }
673 matched++;
674
675 if (del_fp == NULL &&
676 (!r->rtm_type || f->fn_type == r->rtm_type) &&
677 (r->rtm_scope == RT_SCOPE_NOWHERE || f->fn_scope == r->rtm_scope) &&
678 (!r->rtm_protocol || fi->fib_protocol == r->rtm_protocol) &&
679 fib_nh_match(r, n, rta, fi) == 0)
680 del_fp = fp;
681 }
682
683 if (del_fp) {
684 f = *del_fp;
685 rtmsg_fib(RTM_DELROUTE, f, z, tb->tb_id, n, req);
686
687 if (matched != 1) {
688 write_lock_bh(&fib_hash_lock);
689 *del_fp = f->fn_next;
690 write_unlock_bh(&fib_hash_lock);
691
692 if (f->fn_state&FN_S_ACCESSED)
693 rt_cache_flush(-1);
694 fn_free_node(f);
695 fz->fz_nent--;
696 } else {
697 f->fn_state |= FN_S_ZOMBIE;
698 if (f->fn_state&FN_S_ACCESSED) {
699 f->fn_state &= ~FN_S_ACCESSED;
700 rt_cache_flush(-1);
701 }
702 if (++fib_hash_zombies > 128)
703 fib_flush();
704 }
705
706 return 0;
707 }
708 return -ESRCH;
709 }
710
711 extern __inline__ int
712 fn_flush_list(struct fib_node ** fp, int z, struct fn_hash *table)
713 {
714 int found = 0;
715 struct fib_node *f;
716
717 while ((f = *fp) != NULL) {
718 struct fib_info *fi = FIB_INFO(f);
719
720 if (fi && ((f->fn_state&FN_S_ZOMBIE) || (fi->fib_flags&RTNH_F_DEAD))) {
721 write_lock_bh(&fib_hash_lock);
722 *fp = f->fn_next;
723 write_unlock_bh(&fib_hash_lock);
724
725 fn_free_node(f);
726 found++;
727 continue;
728 }
729 fp = &f->fn_next;
730 }
731 return found;
732 }
733
734 static int fn_hash_flush(struct fib_table *tb)
735 {
736 struct fn_hash *table = (struct fn_hash*)tb->tb_data;
737 struct fn_zone *fz;
738 int found = 0;
739
740 fib_hash_zombies = 0;
741 for (fz = table->fn_zone_list; fz; fz = fz->fz_next) {
742 int i;
743 int tmp = 0;
744 for (i=fz->fz_divisor-1; i>=0; i--)
745 tmp += fn_flush_list(&fz->fz_hash[i], fz->fz_order, table);
746 fz->fz_nent -= tmp;
747 found += tmp;
748 }
749 return found;
750 }
751
752
753 #ifdef CONFIG_PROC_FS
754
755 static int fn_hash_get_info(struct fib_table *tb, char *buffer, int first, int count)
756 {
757 struct fn_hash *table = (struct fn_hash*)tb->tb_data;
758 struct fn_zone *fz;
759 int pos = 0;
760 int n = 0;
761
762 read_lock(&fib_hash_lock);
763 for (fz=table->fn_zone_list; fz; fz = fz->fz_next) {
764 int i;
765 struct fib_node *f;
766 int maxslot = fz->fz_divisor;
767 struct fib_node **fp = fz->fz_hash;
768
769 if (fz->fz_nent == 0)
770 continue;
771
772 if (pos + fz->fz_nent <= first) {
773 pos += fz->fz_nent;
774 continue;
775 }
776
777 for (i=0; i < maxslot; i++, fp++) {
778 for (f = *fp; f; f = f->fn_next) {
779 if (++pos <= first)
780 continue;
781 fib_node_get_info(f->fn_type,
782 f->fn_state&FN_S_ZOMBIE,
783 FIB_INFO(f),
784 fz_prefix(f->fn_key, fz),
785 FZ_MASK(fz), buffer);
786 buffer += 128;
787 if (++n >= count)
788 goto out;
789 }
790 }
791 }
792 out:
793 read_unlock(&fib_hash_lock);
794 return n;
795 }
796 #endif
797
798
799 #ifdef CONFIG_RTNETLINK
800
801 extern __inline__ int
802 fn_hash_dump_bucket(struct sk_buff *skb, struct netlink_callback *cb,
803 struct fib_table *tb,
804 struct fn_zone *fz,
805 struct fib_node *f)
806 {
807 int i, s_i;
808
809 s_i = cb->args[3];
810 for (i=0; f; i++, f=f->fn_next) {
811 if (i < s_i) continue;
812 if (f->fn_state&FN_S_ZOMBIE) continue;
813 if (fib_dump_info(skb, NETLINK_CB(cb->skb).pid, cb->nlh->nlmsg_seq,
814 RTM_NEWROUTE,
815 tb->tb_id, (f->fn_state&FN_S_ZOMBIE) ? 0 : f->fn_type, f->fn_scope,
816 &f->fn_key, fz->fz_order, f->fn_tos,
817 f->fn_info) < 0) {
818 cb->args[3] = i;
819 return -1;
820 }
821 }
822 cb->args[3] = i;
823 return skb->len;
824 }
825
826 extern __inline__ int
827 fn_hash_dump_zone(struct sk_buff *skb, struct netlink_callback *cb,
828 struct fib_table *tb,
829 struct fn_zone *fz)
830 {
831 int h, s_h;
832
833 s_h = cb->args[2];
834 for (h=0; h < fz->fz_divisor; h++) {
835 if (h < s_h) continue;
836 if (h > s_h)
837 memset(&cb->args[3], 0, sizeof(cb->args) - 3*sizeof(cb->args[0]));
838 if (fz->fz_hash == NULL || fz->fz_hash[h] == NULL)
839 continue;
840 if (fn_hash_dump_bucket(skb, cb, tb, fz, fz->fz_hash[h]) < 0) {
841 cb->args[2] = h;
842 return -1;
843 }
844 }
845 cb->args[2] = h;
846 return skb->len;
847 }
848
849 static int fn_hash_dump(struct fib_table *tb, struct sk_buff *skb, struct netlink_callback *cb)
850 {
851 int m, s_m;
852 struct fn_zone *fz;
853 struct fn_hash *table = (struct fn_hash*)tb->tb_data;
854
855 s_m = cb->args[1];
856 read_lock(&fib_hash_lock);
857 for (fz = table->fn_zone_list, m=0; fz; fz = fz->fz_next, m++) {
858 if (m < s_m) continue;
859 if (m > s_m)
860 memset(&cb->args[2], 0, sizeof(cb->args) - 2*sizeof(cb->args[0]));
861 if (fn_hash_dump_zone(skb, cb, tb, fz) < 0) {
862 cb->args[1] = m;
863 read_unlock(&fib_hash_lock);
864 return -1;
865 }
866 }
867 read_unlock(&fib_hash_lock);
868 cb->args[1] = m;
869 return skb->len;
870 }
871
872 static void rtmsg_fib(int event, struct fib_node* f, int z, int tb_id,
873 struct nlmsghdr *n, struct netlink_skb_parms *req)
874 {
875 struct sk_buff *skb;
876 u32 pid = req ? req->pid : 0;
877 int size = NLMSG_SPACE(sizeof(struct rtmsg)+256);
878
879 skb = alloc_skb(size, GFP_KERNEL);
880 if (!skb)
881 return;
882
883 if (fib_dump_info(skb, pid, n->nlmsg_seq, event, tb_id,
884 f->fn_type, f->fn_scope, &f->fn_key, z, f->fn_tos,
885 FIB_INFO(f)) < 0) {
886 kfree_skb(skb);
887 return;
888 }
889 NETLINK_CB(skb).dst_groups = RTMGRP_IPV4_ROUTE;
890 if (n->nlmsg_flags&NLM_F_ECHO)
891 atomic_inc(&skb->users);
892 netlink_broadcast(rtnl, skb, pid, RTMGRP_IPV4_ROUTE, GFP_KERNEL);
893 if (n->nlmsg_flags&NLM_F_ECHO)
894 netlink_unicast(rtnl, skb, pid, MSG_DONTWAIT);
895 }
896
897 #endif /* CONFIG_RTNETLINK */
898
899 #ifdef CONFIG_IP_MULTIPLE_TABLES
900 struct fib_table * fib_hash_init(int id)
901 #else
902 struct fib_table * __init fib_hash_init(int id)
903 #endif
904 {
905 struct fib_table *tb;
906
907 if (fn_hash_kmem == NULL)
908 fn_hash_kmem = kmem_cache_create("ip_fib_hash",
909 sizeof(struct fib_node),
910 0, SLAB_HWCACHE_ALIGN,
911 NULL, NULL);
912
913 tb = kmalloc(sizeof(struct fib_table) + sizeof(struct fn_hash), GFP_KERNEL);
914 if (tb == NULL)
915 return NULL;
916
917 tb->tb_id = id;
918 tb->tb_lookup = fn_hash_lookup;
919 tb->tb_insert = fn_hash_insert;
920 tb->tb_delete = fn_hash_delete;
921 tb->tb_flush = fn_hash_flush;
922 tb->tb_select_default = fn_hash_select_default;
923 #ifdef CONFIG_RTNETLINK
924 tb->tb_dump = fn_hash_dump;
925 #endif
926 #ifdef CONFIG_PROC_FS
927 tb->tb_get_info = fn_hash_get_info;
928 #endif
929 memset(tb->tb_data, 0, sizeof(struct fn_hash));
930 return tb;
931 }
932