fib_trie.c 61.7 KB
Newer Older
1 2 3 4 5 6 7 8 9
/*
 *   This program is free software; you can redistribute it and/or
 *   modify it under the terms of the GNU General Public License
 *   as published by the Free Software Foundation; either version
 *   2 of the License, or (at your option) any later version.
 *
 *   Robert Olsson <robert.olsson@its.uu.se> Uppsala Universitet
 *     & Swedish University of Agricultural Sciences.
 *
10
 *   Jens Laas <jens.laas@data.slu.se> Swedish University of
11
 *     Agricultural Sciences.
12
 *
13 14 15
 *   Hans Liss <hans.liss@its.uu.se>  Uppsala Universitet
 *
 * This work is based on the LPC-trie which is originally descibed in:
16
 *
17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43
 * An experimental study of compression methods for dynamic tries
 * Stefan Nilsson and Matti Tikkanen. Algorithmica, 33(1):19-33, 2002.
 * http://www.nada.kth.se/~snilsson/public/papers/dyntrie2/
 *
 *
 * IP-address lookup using LC-tries. Stefan Nilsson and Gunnar Karlsson
 * IEEE Journal on Selected Areas in Communications, 17(6):1083-1092, June 1999
 *
 * Version:	$Id: fib_trie.c,v 1.3 2005/06/08 14:20:01 robert Exp $
 *
 *
 * Code from fib_hash has been reused which includes the following header:
 *
 *
 * INET		An implementation of the TCP/IP protocol suite for the LINUX
 *		operating system.  INET is implemented using the  BSD Socket
 *		interface as the means of communication with the user level.
 *
 *		IPv4 FIB: lookup engine and maintenance routines.
 *
 *
 * Authors:	Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
 *
 *		This program is free software; you can redistribute it and/or
 *		modify it under the terms of the GNU General Public License
 *		as published by the Free Software Foundation; either version
 *		2 of the License, or (at your option) any later version.
R
Robert Olsson 已提交
44 45 46 47 48 49 50
 *
 * Substantial contributions to this work comes from:
 *
 *		David S. Miller, <davem@davemloft.net>
 *		Stephen Hemminger <shemminger@osdl.org>
 *		Paul E. McKenney <paulmck@us.ibm.com>
 *		Patrick McHardy <kaber@trash.net>
51 52
 */

R
Robert Olsson 已提交
53
#define VERSION "0.408"
54 55 56

#include <asm/uaccess.h>
#include <asm/system.h>
J
Jiri Slaby 已提交
57
#include <linux/bitops.h>
58 59 60 61 62 63 64 65 66
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/string.h>
#include <linux/socket.h>
#include <linux/sockios.h>
#include <linux/errno.h>
#include <linux/in.h>
#include <linux/inet.h>
S
Stephen Hemminger 已提交
67
#include <linux/inetdevice.h>
68 69 70
#include <linux/netdevice.h>
#include <linux/if_arp.h>
#include <linux/proc_fs.h>
R
Robert Olsson 已提交
71
#include <linux/rcupdate.h>
72 73 74 75
#include <linux/skbuff.h>
#include <linux/netlink.h>
#include <linux/init.h>
#include <linux/list.h>
76
#include <net/net_namespace.h>
77 78 79 80 81 82 83 84
#include <net/ip.h>
#include <net/protocol.h>
#include <net/route.h>
#include <net/tcp.h>
#include <net/sock.h>
#include <net/ip_fib.h>
#include "fib_lookup.h"

R
Robert Olsson 已提交
85
#define MAX_STAT_DEPTH 32
86 87 88 89 90 91 92 93

#define KEYLENGTH (8*sizeof(t_key))

typedef unsigned int t_key;

#define T_TNODE 0
#define T_LEAF  1
#define NODE_TYPE_MASK	0x1UL
R
Robert Olsson 已提交
94 95
#define NODE_TYPE(node) ((node)->parent & NODE_TYPE_MASK)

O
Olof Johansson 已提交
96 97
#define IS_TNODE(n) (!(n->parent & T_LEAF))
#define IS_LEAF(n) (n->parent & T_LEAF)
98 99

struct node {
O
Olof Johansson 已提交
100
	unsigned long parent;
101
	t_key key;
102 103 104
};

struct leaf {
O
Olof Johansson 已提交
105
	unsigned long parent;
106
	t_key key;
107
	struct hlist_head list;
R
Robert Olsson 已提交
108
	struct rcu_head rcu;
109 110 111 112
};

struct leaf_info {
	struct hlist_node hlist;
R
Robert Olsson 已提交
113
	struct rcu_head rcu;
114 115 116 117 118
	int plen;
	struct list_head falh;
};

struct tnode {
O
Olof Johansson 已提交
119
	unsigned long parent;
120
	t_key key;
121 122
	unsigned char pos;		/* 2log(KEYLENGTH) bits needed */
	unsigned char bits;		/* 2log(KEYLENGTH) bits needed */
123 124
	unsigned int full_children;	/* KEYLENGTH bits needed */
	unsigned int empty_children;	/* KEYLENGTH bits needed */
125 126 127 128
	union {
		struct rcu_head rcu;
		struct work_struct work;
	};
O
Olof Johansson 已提交
129
	struct node *child[0];
130 131 132 133 134 135 136 137 138
};

#ifdef CONFIG_IP_FIB_TRIE_STATS
struct trie_use_stats {
	unsigned int gets;
	unsigned int backtrack;
	unsigned int semantic_match_passed;
	unsigned int semantic_match_miss;
	unsigned int null_node_hit;
139
	unsigned int resize_node_skipped;
140 141 142 143 144 145 146 147 148
};
#endif

struct trie_stat {
	unsigned int totdepth;
	unsigned int maxdepth;
	unsigned int tnodes;
	unsigned int leaves;
	unsigned int nullpointers;
149
	unsigned int prefixes;
R
Robert Olsson 已提交
150
	unsigned int nodesizes[MAX_STAT_DEPTH];
151
};
152 153

struct trie {
O
Olof Johansson 已提交
154
	struct node *trie;
155 156 157 158 159 160
#ifdef CONFIG_IP_FIB_TRIE_STATS
	struct trie_use_stats stats;
#endif
};

static void put_child(struct trie *t, struct tnode *tn, int i, struct node *n);
161 162
static void tnode_put_child_reorg(struct tnode *tn, int i, struct node *n,
				  int wasfull);
163
static struct node *resize(struct trie *t, struct tnode *tn);
164 165
static struct tnode *inflate(struct trie *t, struct tnode *tn);
static struct tnode *halve(struct trie *t, struct tnode *tn);
166

167
static struct kmem_cache *fn_alias_kmem __read_mostly;
168
static struct kmem_cache *trie_leaf_kmem __read_mostly;
169

S
Stephen Hemminger 已提交
170 171
static inline struct tnode *node_parent(struct node *node)
{
172 173 174 175 176 177
	return (struct tnode *)(node->parent & ~NODE_TYPE_MASK);
}

static inline struct tnode *node_parent_rcu(struct node *node)
{
	struct tnode *ret = node_parent(node);
S
Stephen Hemminger 已提交
178 179 180 181

	return rcu_dereference(ret);
}

182 183 184
/* Same as rcu_assign_pointer
 * but that macro() assumes that value is a pointer.
 */
S
Stephen Hemminger 已提交
185 186
static inline void node_set_parent(struct node *node, struct tnode *ptr)
{
187 188
	smp_wmb();
	node->parent = (unsigned long)ptr | NODE_TYPE(node);
S
Stephen Hemminger 已提交
189
}
R
Robert Olsson 已提交
190

191 192 193
static inline struct node *tnode_get_child(struct tnode *tn, unsigned int i)
{
	BUG_ON(i >= 1U << tn->bits);
R
Robert Olsson 已提交
194

195 196 197 198
	return tn->child[i];
}

static inline struct node *tnode_get_child_rcu(struct tnode *tn, unsigned int i)
199
{
200
	struct node *ret = tnode_get_child(tn, i);
201

202
	return rcu_dereference(ret);
203 204
}

S
Stephen Hemmigner 已提交
205
static inline int tnode_child_length(const struct tnode *tn)
206
{
O
Olof Johansson 已提交
207
	return 1 << tn->bits;
208 209
}

S
Stephen Hemminger 已提交
210 211 212 213 214
static inline t_key mask_pfx(t_key k, unsigned short l)
{
	return (l == 0) ? 0 : k >> (KEYLENGTH-l) << (KEYLENGTH-l);
}

215 216
static inline t_key tkey_extract_bits(t_key a, int offset, int bits)
{
O
Olof Johansson 已提交
217
	if (offset < KEYLENGTH)
218
		return ((t_key)(a << offset)) >> (KEYLENGTH - bits);
O
Olof Johansson 已提交
219
	else
220 221 222 223 224
		return 0;
}

static inline int tkey_equals(t_key a, t_key b)
{
225
	return a == b;
226 227 228 229
}

static inline int tkey_sub_equals(t_key a, int offset, int bits, t_key b)
{
230 231
	if (bits == 0 || offset >= KEYLENGTH)
		return 1;
O
Olof Johansson 已提交
232 233
	bits = bits > KEYLENGTH ? KEYLENGTH : bits;
	return ((a ^ b) << offset) >> (KEYLENGTH - bits) == 0;
234
}
235 236 237 238 239 240

static inline int tkey_mismatch(t_key a, int offset, t_key b)
{
	t_key diff = a ^ b;
	int i = offset;

241 242 243
	if (!diff)
		return 0;
	while ((diff << i) >> (KEYLENGTH-1) == 0)
244 245 246 247 248
		i++;
	return i;
}

/*
249 250
  To understand this stuff, an understanding of keys and all their bits is
  necessary. Every node in the trie has a key associated with it, but not
251 252 253 254
  all of the bits in that key are significant.

  Consider a node 'n' and its parent 'tp'.

255 256 257 258 259
  If n is a leaf, every bit in its key is significant. Its presence is
  necessitated by path compression, since during a tree traversal (when
  searching for a leaf - unless we are doing an insertion) we will completely
  ignore all skipped bits we encounter. Thus we need to verify, at the end of
  a potentially successful search, that we have indeed been walking the
260 261
  correct key path.

262 263 264 265 266
  Note that we can never "miss" the correct key in the tree if present by
  following the wrong path. Path compression ensures that segments of the key
  that are the same for all keys with a given prefix are skipped, but the
  skipped part *is* identical for each node in the subtrie below the skipped
  bit! trie_insert() in this implementation takes care of that - note the
267 268
  call to tkey_sub_equals() in trie_insert().

269
  if n is an internal node - a 'tnode' here, the various parts of its key
270 271
  have many different meanings.

272
  Example:
273 274 275
  _________________________________________________________________
  | i | i | i | i | i | i | i | N | N | N | S | S | S | S | S | C |
  -----------------------------------------------------------------
276
    0   1   2   3   4   5   6   7   8   9  10  11  12  13  14  15
277 278 279 280 281 282 283 284 285

  _________________________________________________________________
  | C | C | C | u | u | u | u | u | u | u | u | u | u | u | u | u |
  -----------------------------------------------------------------
   16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31

  tp->pos = 7
  tp->bits = 3
  n->pos = 15
O
Olof Johansson 已提交
286
  n->bits = 4
287

288 289
  First, let's just ignore the bits that come before the parent tp, that is
  the bits from 0 to (tp->pos-1). They are *known* but at this point we do
290 291 292
  not use them for anything.

  The bits from (tp->pos) to (tp->pos + tp->bits - 1) - "N", above - are the
293
  index into the parent's child array. That is, they will be used to find
294 295 296 297 298
  'n' among tp's children.

  The bits from (tp->pos + tp->bits) to (n->pos - 1) - "S" - are skipped bits
  for the node n.

299
  All the bits we have seen so far are significant to the node n. The rest
300 301
  of the bits are really not needed or indeed known in n->key.

302
  The bits from (n->pos) to (n->pos + n->bits - 1) - "C" - are the index into
303
  n's child array, and will of course be different for each child.
304

305

306 307 308 309 310
  The rest of the bits, from (n->pos + n->bits) onward, are completely unknown
  at this point.

*/

S
Stephen Hemminger 已提交
311
static inline void check_tnode(const struct tnode *tn)
312
{
S
Stephen Hemminger 已提交
313
	WARN_ON(tn && tn->pos+tn->bits > 32);
314 315
}

316 317 318 319
static const int halve_threshold = 25;
static const int inflate_threshold = 50;
static const int halve_threshold_root = 8;
static const int inflate_threshold_root = 15;
320

R
Robert Olsson 已提交
321 322

static void __alias_free_mem(struct rcu_head *head)
323
{
R
Robert Olsson 已提交
324 325
	struct fib_alias *fa = container_of(head, struct fib_alias, rcu);
	kmem_cache_free(fn_alias_kmem, fa);
326 327
}

R
Robert Olsson 已提交
328
static inline void alias_free_mem_rcu(struct fib_alias *fa)
329
{
R
Robert Olsson 已提交
330 331
	call_rcu(&fa->rcu, __alias_free_mem);
}
O
Olof Johansson 已提交
332

R
Robert Olsson 已提交
333 334
static void __leaf_free_rcu(struct rcu_head *head)
{
335 336
	struct leaf *l = container_of(head, struct leaf, rcu);
	kmem_cache_free(trie_leaf_kmem, l);
R
Robert Olsson 已提交
337
}
O
Olof Johansson 已提交
338

339 340 341 342 343
static inline void free_leaf(struct leaf *l)
{
	call_rcu_bh(&l->rcu, __leaf_free_rcu);
}

R
Robert Olsson 已提交
344
static void __leaf_info_free_rcu(struct rcu_head *head)
345
{
R
Robert Olsson 已提交
346
	kfree(container_of(head, struct leaf_info, rcu));
347 348
}

R
Robert Olsson 已提交
349
static inline void free_leaf_info(struct leaf_info *leaf)
350
{
R
Robert Olsson 已提交
351
	call_rcu(&leaf->rcu, __leaf_info_free_rcu);
352 353
}

354
static struct tnode *tnode_alloc(size_t size)
355
{
R
Robert Olsson 已提交
356
	if (size <= PAGE_SIZE)
357
		return kzalloc(size, GFP_KERNEL);
358 359 360
	else
		return __vmalloc(size, GFP_KERNEL | __GFP_ZERO, PAGE_KERNEL);
}
R
Robert Olsson 已提交
361

362 363 364 365
static void __tnode_vfree(struct work_struct *arg)
{
	struct tnode *tn = container_of(arg, struct tnode, work);
	vfree(tn);
366 367
}

R
Robert Olsson 已提交
368
static void __tnode_free_rcu(struct rcu_head *head)
369
{
R
Robert Olsson 已提交
370
	struct tnode *tn = container_of(head, struct tnode, rcu);
371 372
	size_t size = sizeof(struct tnode) +
		      (sizeof(struct node *) << tn->bits);
373 374 375

	if (size <= PAGE_SIZE)
		kfree(tn);
376 377 378 379
	else {
		INIT_WORK(&tn->work, __tnode_vfree);
		schedule_work(&tn->work);
	}
380 381
}

R
Robert Olsson 已提交
382 383
static inline void tnode_free(struct tnode *tn)
{
384 385 386
	if (IS_LEAF(tn))
		free_leaf((struct leaf *) tn);
	else
R
Robert Olsson 已提交
387
		call_rcu(&tn->rcu, __tnode_free_rcu);
R
Robert Olsson 已提交
388 389 390 391
}

static struct leaf *leaf_new(void)
{
392
	struct leaf *l = kmem_cache_alloc(trie_leaf_kmem, GFP_KERNEL);
R
Robert Olsson 已提交
393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409
	if (l) {
		l->parent = T_LEAF;
		INIT_HLIST_HEAD(&l->list);
	}
	return l;
}

static struct leaf_info *leaf_info_new(int plen)
{
	struct leaf_info *li = kmalloc(sizeof(struct leaf_info),  GFP_KERNEL);
	if (li) {
		li->plen = plen;
		INIT_LIST_HEAD(&li->falh);
	}
	return li;
}

410
static struct tnode *tnode_new(t_key key, int pos, int bits)
411
{
412
	size_t sz = sizeof(struct tnode) + (sizeof(struct node *) << bits);
413
	struct tnode *tn = tnode_alloc(sz);
414

O
Olof Johansson 已提交
415
	if (tn) {
R
Robert Olsson 已提交
416
		tn->parent = T_TNODE;
417 418 419 420 421 422
		tn->pos = pos;
		tn->bits = bits;
		tn->key = key;
		tn->full_children = 0;
		tn->empty_children = 1<<bits;
	}
423

424 425
	pr_debug("AT %p s=%u %lu\n", tn, (unsigned int) sizeof(struct tnode),
		 (unsigned long) (sizeof(struct node) << bits));
426 427 428 429 430 431 432 433
	return tn;
}

/*
 * Check whether a tnode 'n' is "full", i.e. it is an internal node
 * and no bits are skipped. See discussion in dyntree paper p. 6
 */

S
Stephen Hemmigner 已提交
434
static inline int tnode_full(const struct tnode *tn, const struct node *n)
435
{
436
	if (n == NULL || IS_LEAF(n))
437 438 439 440 441
		return 0;

	return ((struct tnode *) n)->pos == tn->pos + tn->bits;
}

442 443
static inline void put_child(struct trie *t, struct tnode *tn, int i,
			     struct node *n)
444 445 446 447
{
	tnode_put_child_reorg(tn, i, n, -1);
}

448
 /*
449 450 451 452
  * Add a child at position i overwriting the old value.
  * Update the value of full_children and empty_children.
  */

453 454
static void tnode_put_child_reorg(struct tnode *tn, int i, struct node *n,
				  int wasfull)
455
{
R
Robert Olsson 已提交
456
	struct node *chi = tn->child[i];
457 458
	int isfull;

S
Stephen Hemminger 已提交
459 460
	BUG_ON(i >= 1<<tn->bits);

461 462 463 464 465
	/* update emptyChildren */
	if (n == NULL && chi != NULL)
		tn->empty_children++;
	else if (n != NULL && chi == NULL)
		tn->empty_children--;
466

467
	/* update fullChildren */
O
Olof Johansson 已提交
468
	if (wasfull == -1)
469 470 471
		wasfull = tnode_full(tn, chi);

	isfull = tnode_full(tn, n);
472
	if (wasfull && !isfull)
473
		tn->full_children--;
474
	else if (!wasfull && isfull)
475
		tn->full_children++;
O
Olof Johansson 已提交
476

477
	if (n)
S
Stephen Hemminger 已提交
478
		node_set_parent(n, tn);
479

R
Robert Olsson 已提交
480
	rcu_assign_pointer(tn->child[i], n);
481 482
}

483
static struct node *resize(struct trie *t, struct tnode *tn)
484 485
{
	int i;
486
	int err = 0;
487
	struct tnode *old_tn;
488 489
	int inflate_threshold_use;
	int halve_threshold_use;
R
Robert Olsson 已提交
490
	int max_resize;
491

492
	if (!tn)
493 494
		return NULL;

S
Stephen Hemminger 已提交
495 496
	pr_debug("In tnode_resize %p inflate_threshold=%d threshold=%d\n",
		 tn, inflate_threshold, halve_threshold);
497 498 499 500 501 502 503 504 505

	/* No children */
	if (tn->empty_children == tnode_child_length(tn)) {
		tnode_free(tn);
		return NULL;
	}
	/* One child */
	if (tn->empty_children == tnode_child_length(tn) - 1)
		for (i = 0; i < tnode_child_length(tn); i++) {
O
Olof Johansson 已提交
506
			struct node *n;
507

O
Olof Johansson 已提交
508
			n = tn->child[i];
R
Robert Olsson 已提交
509
			if (!n)
O
Olof Johansson 已提交
510 511 512
				continue;

			/* compress one level */
S
Stephen Hemminger 已提交
513
			node_set_parent(n, NULL);
O
Olof Johansson 已提交
514 515
			tnode_free(tn);
			return n;
516
		}
517
	/*
518 519 520 521 522
	 * Double as long as the resulting node has a number of
	 * nonempty nodes that are above the threshold.
	 */

	/*
523 524
	 * From "Implementing a dynamic compressed trie" by Stefan Nilsson of
	 * the Helsinki University of Technology and Matti Tikkanen of Nokia
525
	 * Telecommunications, page 6:
526
	 * "A node is doubled if the ratio of non-empty children to all
527 528
	 * children in the *doubled* node is at least 'high'."
	 *
529 530 531 532 533
	 * 'high' in this instance is the variable 'inflate_threshold'. It
	 * is expressed as a percentage, so we multiply it with
	 * tnode_child_length() and instead of multiplying by 2 (since the
	 * child array will be doubled by inflate()) and multiplying
	 * the left-hand side by 100 (to handle the percentage thing) we
534
	 * multiply the left-hand side by 50.
535 536 537 538
	 *
	 * The left-hand side may look a bit weird: tnode_child_length(tn)
	 * - tn->empty_children is of course the number of non-null children
	 * in the current node. tn->full_children is the number of "full"
539
	 * children, that is non-null tnodes with a skip value of 0.
540
	 * All of those will be doubled in the resulting inflated tnode, so
541
	 * we just count them one extra time here.
542
	 *
543
	 * A clearer way to write this would be:
544
	 *
545
	 * to_be_doubled = tn->full_children;
546
	 * not_to_be_doubled = tnode_child_length(tn) - tn->empty_children -
547 548 549 550
	 *     tn->full_children;
	 *
	 * new_child_length = tnode_child_length(tn) * 2;
	 *
551
	 * new_fill_factor = 100 * (not_to_be_doubled + 2*to_be_doubled) /
552 553
	 *      new_child_length;
	 * if (new_fill_factor >= inflate_threshold)
554 555 556
	 *
	 * ...and so on, tho it would mess up the while () loop.
	 *
557 558 559
	 * anyway,
	 * 100 * (not_to_be_doubled + 2*to_be_doubled) / new_child_length >=
	 *      inflate_threshold
560
	 *
561 562 563
	 * avoid a division:
	 * 100 * (not_to_be_doubled + 2*to_be_doubled) >=
	 *      inflate_threshold * new_child_length
564
	 *
565
	 * expand not_to_be_doubled and to_be_doubled, and shorten:
566
	 * 100 * (tnode_child_length(tn) - tn->empty_children +
O
Olof Johansson 已提交
567
	 *    tn->full_children) >= inflate_threshold * new_child_length
568
	 *
569
	 * expand new_child_length:
570
	 * 100 * (tnode_child_length(tn) - tn->empty_children +
O
Olof Johansson 已提交
571
	 *    tn->full_children) >=
572
	 *      inflate_threshold * tnode_child_length(tn) * 2
573
	 *
574
	 * shorten again:
575
	 * 50 * (tn->full_children + tnode_child_length(tn) -
O
Olof Johansson 已提交
576
	 *    tn->empty_children) >= inflate_threshold *
577
	 *    tnode_child_length(tn)
578
	 *
579 580 581
	 */

	check_tnode(tn);
582

583 584
	/* Keep root node larger  */

S
Stephen Hemminger 已提交
585
	if (!tn->parent)
586
		inflate_threshold_use = inflate_threshold_root;
587
	else
588 589
		inflate_threshold_use = inflate_threshold;

590
	err = 0;
R
Robert Olsson 已提交
591 592
	max_resize = 10;
	while ((tn->full_children > 0 &&  max_resize-- &&
593 594 595
		50 * (tn->full_children + tnode_child_length(tn)
		      - tn->empty_children)
		>= inflate_threshold_use * tnode_child_length(tn))) {
596

597 598
		old_tn = tn;
		tn = inflate(t, tn);
599

600 601
		if (IS_ERR(tn)) {
			tn = old_tn;
602 603 604 605 606
#ifdef CONFIG_IP_FIB_TRIE_STATS
			t->stats.resize_node_skipped++;
#endif
			break;
		}
607 608
	}

R
Robert Olsson 已提交
609 610
	if (max_resize < 0) {
		if (!tn->parent)
611 612 613
			pr_warning("Fix inflate_threshold_root."
				   " Now=%d size=%d bits\n",
				   inflate_threshold_root, tn->bits);
R
Robert Olsson 已提交
614
		else
615 616 617
			pr_warning("Fix inflate_threshold."
				   " Now=%d size=%d bits\n",
				   inflate_threshold, tn->bits);
R
Robert Olsson 已提交
618 619
	}

620 621 622 623 624 625
	check_tnode(tn);

	/*
	 * Halve as long as the number of empty children in this
	 * node is above threshold.
	 */
626

627 628 629

	/* Keep root node larger  */

S
Stephen Hemminger 已提交
630
	if (!tn->parent)
631
		halve_threshold_use = halve_threshold_root;
632
	else
633 634
		halve_threshold_use = halve_threshold;

635
	err = 0;
R
Robert Olsson 已提交
636 637
	max_resize = 10;
	while (tn->bits > 1 &&  max_resize-- &&
638
	       100 * (tnode_child_length(tn) - tn->empty_children) <
639
	       halve_threshold_use * tnode_child_length(tn)) {
640

641 642 643 644
		old_tn = tn;
		tn = halve(t, tn);
		if (IS_ERR(tn)) {
			tn = old_tn;
645 646 647 648 649 650
#ifdef CONFIG_IP_FIB_TRIE_STATS
			t->stats.resize_node_skipped++;
#endif
			break;
		}
	}
651

R
Robert Olsson 已提交
652 653
	if (max_resize < 0) {
		if (!tn->parent)
654 655 656
			pr_warning("Fix halve_threshold_root."
				   " Now=%d size=%d bits\n",
				   halve_threshold_root, tn->bits);
R
Robert Olsson 已提交
657
		else
658 659 660
			pr_warning("Fix halve_threshold."
				   " Now=%d size=%d bits\n",
				   halve_threshold, tn->bits);
R
Robert Olsson 已提交
661
	}
662

663 664 665
	/* Only one child remains */
	if (tn->empty_children == tnode_child_length(tn) - 1)
		for (i = 0; i < tnode_child_length(tn); i++) {
O
Olof Johansson 已提交
666
			struct node *n;
667

O
Olof Johansson 已提交
668
			n = tn->child[i];
R
Robert Olsson 已提交
669
			if (!n)
O
Olof Johansson 已提交
670 671 672 673
				continue;

			/* compress one level */

S
Stephen Hemminger 已提交
674
			node_set_parent(n, NULL);
O
Olof Johansson 已提交
675 676
			tnode_free(tn);
			return n;
677 678 679 680 681
		}

	return (struct node *) tn;
}

682
static struct tnode *inflate(struct trie *t, struct tnode *tn)
683 684 685 686 687
{
	struct tnode *oldtnode = tn;
	int olen = tnode_child_length(tn);
	int i;

S
Stephen Hemminger 已提交
688
	pr_debug("In inflate\n");
689 690 691

	tn = tnode_new(oldtnode->key, oldtnode->pos, oldtnode->bits + 1);

S
Stephen Hemminger 已提交
692
	if (!tn)
693
		return ERR_PTR(-ENOMEM);
694 695

	/*
696 697 698
	 * Preallocate and store tnodes before the actual work so we
	 * don't get into an inconsistent state if memory allocation
	 * fails. In case of failure we return the oldnode and  inflate
699 700
	 * of tnode is ignored.
	 */
O
Olof Johansson 已提交
701 702

	for (i = 0; i < olen; i++) {
703
		struct tnode *inode;
704

705
		inode = (struct tnode *) tnode_get_child(oldtnode, i);
706 707 708 709 710
		if (inode &&
		    IS_TNODE(inode) &&
		    inode->pos == oldtnode->pos + oldtnode->bits &&
		    inode->bits > 1) {
			struct tnode *left, *right;
S
Stephen Hemminger 已提交
711
			t_key m = ~0U << (KEYLENGTH - 1) >> inode->pos;
712

713 714
			left = tnode_new(inode->key&(~m), inode->pos + 1,
					 inode->bits - 1);
715 716
			if (!left)
				goto nomem;
O
Olof Johansson 已提交
717

718 719 720
			right = tnode_new(inode->key|m, inode->pos + 1,
					  inode->bits - 1);

721
			if (!right) {
722 723
				tnode_free(left);
				goto nomem;
724
			}
725 726 727 728 729 730

			put_child(t, tn, 2*i, (struct node *) left);
			put_child(t, tn, 2*i+1, (struct node *) right);
		}
	}

O
Olof Johansson 已提交
731
	for (i = 0; i < olen; i++) {
732
		struct tnode *inode;
733
		struct node *node = tnode_get_child(oldtnode, i);
O
Olof Johansson 已提交
734 735
		struct tnode *left, *right;
		int size, j;
736

737 738 739 740 741 742
		/* An empty child */
		if (node == NULL)
			continue;

		/* A leaf or an internal node with skipped bits */

743
		if (IS_LEAF(node) || ((struct tnode *) node)->pos >
744
		   tn->pos + tn->bits - 1) {
745 746 747
			if (tkey_extract_bits(node->key,
					      oldtnode->pos + oldtnode->bits,
					      1) == 0)
748 749 750 751 752 753 754 755 756 757 758 759 760 761
				put_child(t, tn, 2*i, node);
			else
				put_child(t, tn, 2*i+1, node);
			continue;
		}

		/* An internal node with two children */
		inode = (struct tnode *) node;

		if (inode->bits == 1) {
			put_child(t, tn, 2*i, inode->child[0]);
			put_child(t, tn, 2*i+1, inode->child[1]);

			tnode_free(inode);
O
Olof Johansson 已提交
762
			continue;
763 764
		}

O
Olof Johansson 已提交
765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782
		/* An internal node with more than two children */

		/* We will replace this node 'inode' with two new
		 * ones, 'left' and 'right', each with half of the
		 * original children. The two new nodes will have
		 * a position one bit further down the key and this
		 * means that the "significant" part of their keys
		 * (see the discussion near the top of this file)
		 * will differ by one bit, which will be "0" in
		 * left's key and "1" in right's key. Since we are
		 * moving the key position by one step, the bit that
		 * we are moving away from - the bit at position
		 * (inode->pos) - is the one that will differ between
		 * left and right. So... we synthesize that bit in the
		 * two  new keys.
		 * The mask 'm' below will be a single "one" bit at
		 * the position (inode->pos)
		 */
783

O
Olof Johansson 已提交
784 785 786
		/* Use the old key, but set the new significant
		 *   bit to zero.
		 */
787

O
Olof Johansson 已提交
788 789
		left = (struct tnode *) tnode_get_child(tn, 2*i);
		put_child(t, tn, 2*i, NULL);
790

O
Olof Johansson 已提交
791
		BUG_ON(!left);
792

O
Olof Johansson 已提交
793 794
		right = (struct tnode *) tnode_get_child(tn, 2*i+1);
		put_child(t, tn, 2*i+1, NULL);
795

O
Olof Johansson 已提交
796
		BUG_ON(!right);
797

O
Olof Johansson 已提交
798 799 800 801
		size = tnode_child_length(left);
		for (j = 0; j < size; j++) {
			put_child(t, left, j, inode->child[j]);
			put_child(t, right, j, inode->child[j + size]);
802
		}
O
Olof Johansson 已提交
803 804 805 806
		put_child(t, tn, 2*i, resize(t, left));
		put_child(t, tn, 2*i+1, resize(t, right));

		tnode_free(inode);
807 808 809
	}
	tnode_free(oldtnode);
	return tn;
810 811 812 813 814
nomem:
	{
		int size = tnode_child_length(tn);
		int j;

S
Stephen Hemminger 已提交
815
		for (j = 0; j < size; j++)
816 817 818 819
			if (tn->child[j])
				tnode_free((struct tnode *)tn->child[j]);

		tnode_free(tn);
S
Stephen Hemminger 已提交
820

821 822
		return ERR_PTR(-ENOMEM);
	}
823 824
}

825
static struct tnode *halve(struct trie *t, struct tnode *tn)
826 827 828 829 830 831
{
	struct tnode *oldtnode = tn;
	struct node *left, *right;
	int i;
	int olen = tnode_child_length(tn);

S
Stephen Hemminger 已提交
832
	pr_debug("In halve\n");
833 834

	tn = tnode_new(oldtnode->key, oldtnode->pos, oldtnode->bits - 1);
835

836 837
	if (!tn)
		return ERR_PTR(-ENOMEM);
838 839

	/*
840 841 842
	 * Preallocate and store tnodes before the actual work so we
	 * don't get into an inconsistent state if memory allocation
	 * fails. In case of failure we return the oldnode and halve
843 844 845
	 * of tnode is ignored.
	 */

O
Olof Johansson 已提交
846
	for (i = 0; i < olen; i += 2) {
847 848
		left = tnode_get_child(oldtnode, i);
		right = tnode_get_child(oldtnode, i+1);
849

850
		/* Two nonempty children */
S
Stephen Hemminger 已提交
851
		if (left && right) {
852
			struct tnode *newn;
S
Stephen Hemminger 已提交
853

854
			newn = tnode_new(left->key, tn->pos + tn->bits, 1);
S
Stephen Hemminger 已提交
855 856

			if (!newn)
857
				goto nomem;
S
Stephen Hemminger 已提交
858

859
			put_child(t, tn, i/2, (struct node *)newn);
860 861 862
		}

	}
863

O
Olof Johansson 已提交
864 865 866
	for (i = 0; i < olen; i += 2) {
		struct tnode *newBinNode;

867 868
		left = tnode_get_child(oldtnode, i);
		right = tnode_get_child(oldtnode, i+1);
869

870 871 872 873 874
		/* At least one of the children is empty */
		if (left == NULL) {
			if (right == NULL)    /* Both are empty */
				continue;
			put_child(t, tn, i/2, right);
O
Olof Johansson 已提交
875
			continue;
S
Stephen Hemminger 已提交
876
		}
O
Olof Johansson 已提交
877 878

		if (right == NULL) {
879
			put_child(t, tn, i/2, left);
O
Olof Johansson 已提交
880 881
			continue;
		}
882

883
		/* Two nonempty children */
O
Olof Johansson 已提交
884 885 886 887 888
		newBinNode = (struct tnode *) tnode_get_child(tn, i/2);
		put_child(t, tn, i/2, NULL);
		put_child(t, newBinNode, 0, left);
		put_child(t, newBinNode, 1, right);
		put_child(t, tn, i/2, resize(t, newBinNode));
889 890 891
	}
	tnode_free(oldtnode);
	return tn;
892 893 894 895 896
nomem:
	{
		int size = tnode_child_length(tn);
		int j;

S
Stephen Hemminger 已提交
897
		for (j = 0; j < size; j++)
898 899 900 901
			if (tn->child[j])
				tnode_free((struct tnode *)tn->child[j]);

		tnode_free(tn);
S
Stephen Hemminger 已提交
902

903 904
		return ERR_PTR(-ENOMEM);
	}
905 906
}

R
Robert Olsson 已提交
907
/* readside must use rcu_read_lock currently dump routines
R
Robert Olsson 已提交
908 909
 via get_fa_head and dump */

R
Robert Olsson 已提交
910
static struct leaf_info *find_leaf_info(struct leaf *l, int plen)
911
{
R
Robert Olsson 已提交
912
	struct hlist_head *head = &l->list;
913 914 915
	struct hlist_node *node;
	struct leaf_info *li;

R
Robert Olsson 已提交
916
	hlist_for_each_entry_rcu(li, node, head, hlist)
917
		if (li->plen == plen)
918
			return li;
O
Olof Johansson 已提交
919

920 921 922
	return NULL;
}

923
static inline struct list_head *get_fa_head(struct leaf *l, int plen)
924
{
R
Robert Olsson 已提交
925
	struct leaf_info *li = find_leaf_info(l, plen);
926

O
Olof Johansson 已提交
927 928
	if (!li)
		return NULL;
929

O
Olof Johansson 已提交
930
	return &li->falh;
931 932 933 934
}

static void insert_leaf_info(struct hlist_head *head, struct leaf_info *new)
{
935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951
	struct leaf_info *li = NULL, *last = NULL;
	struct hlist_node *node;

	if (hlist_empty(head)) {
		hlist_add_head_rcu(&new->hlist, head);
	} else {
		hlist_for_each_entry(li, node, head, hlist) {
			if (new->plen > li->plen)
				break;

			last = li;
		}
		if (last)
			hlist_add_after_rcu(&last->hlist, &new->hlist);
		else
			hlist_add_before_rcu(&new->hlist, &li->hlist);
	}
952 953
}

R
Robert Olsson 已提交
954 955
/* rcu_read_lock needs to be hold by caller from readside */

956 957 958 959 960 961 962 963
static struct leaf *
fib_find_node(struct trie *t, u32 key)
{
	int pos;
	struct tnode *tn;
	struct node *n;

	pos = 0;
R
Robert Olsson 已提交
964
	n = rcu_dereference(t->trie);
965 966 967

	while (n != NULL &&  NODE_TYPE(n) == T_TNODE) {
		tn = (struct tnode *) n;
O
Olof Johansson 已提交
968

969
		check_tnode(tn);
O
Olof Johansson 已提交
970

971
		if (tkey_sub_equals(tn->key, pos, tn->pos-pos, key)) {
O
Olof Johansson 已提交
972
			pos = tn->pos + tn->bits;
973 974 975 976
			n = tnode_get_child_rcu(tn,
						tkey_extract_bits(key,
								  tn->pos,
								  tn->bits));
O
Olof Johansson 已提交
977
		} else
978 979 980 981
			break;
	}
	/* Case we have found a leaf. Compare prefixes */

O
Olof Johansson 已提交
982 983 984
	if (n != NULL && IS_LEAF(n) && tkey_equals(key, n->key))
		return (struct leaf *)n;

985 986 987 988 989 990
	return NULL;
}

static struct node *trie_rebalance(struct trie *t, struct tnode *tn)
{
	int wasfull;
S
Stephen Hemminger 已提交
991 992
	t_key cindex, key = tn->key;
	struct tnode *tp;
993

S
Stephen Hemminger 已提交
994
	while (tn != NULL && (tp = node_parent((struct node *)tn)) != NULL) {
995 996
		cindex = tkey_extract_bits(key, tp->pos, tp->bits);
		wasfull = tnode_full(tp, tnode_get_child(tp, cindex));
997 998 999 1000
		tn = (struct tnode *) resize(t, (struct tnode *)tn);

		tnode_put_child_reorg((struct tnode *)tp, cindex,
				      (struct node *)tn, wasfull);
O
Olof Johansson 已提交
1001

S
Stephen Hemminger 已提交
1002 1003
		tp = node_parent((struct node *) tn);
		if (!tp)
1004
			break;
S
Stephen Hemminger 已提交
1005
		tn = tp;
1006
	}
S
Stephen Hemminger 已提交
1007

1008
	/* Handle last (top) tnode */
1009
	if (IS_TNODE(tn))
1010
		tn = (struct tnode *)resize(t, (struct tnode *)tn);
1011

1012
	return (struct node *)tn;
1013 1014
}

R
Robert Olsson 已提交
1015 1016
/* only used from updater-side */

1017
static struct list_head *fib_insert_node(struct trie *t, u32 key, int plen)
1018 1019 1020 1021 1022 1023
{
	int pos, newpos;
	struct tnode *tp = NULL, *tn = NULL;
	struct node *n;
	struct leaf *l;
	int missbit;
1024
	struct list_head *fa_head = NULL;
1025 1026 1027 1028
	struct leaf_info *li;
	t_key cindex;

	pos = 0;
1029
	n = t->trie;
1030

1031 1032
	/* If we point to NULL, stop. Either the tree is empty and we should
	 * just put a new leaf in if, or we have reached an empty child slot,
1033
	 * and we should just put our new leaf in that.
1034 1035
	 * If we point to a T_TNODE, check if it matches our key. Note that
	 * a T_TNODE might be skipping any number of bits - its 'pos' need
1036 1037
	 * not be the parent's 'pos'+'bits'!
	 *
1038
	 * If it does match the current key, get pos/bits from it, extract
1039 1040 1041 1042
	 * the index from our key, push the T_TNODE and walk the tree.
	 *
	 * If it doesn't, we have to replace it with a new T_TNODE.
	 *
1043 1044 1045
	 * If we point to a T_LEAF, it might or might not have the same key
	 * as we do. If it does, just change the value, update the T_LEAF's
	 * value, and return it.
1046 1047 1048 1049 1050
	 * If it doesn't, we need to replace it with a T_TNODE.
	 */

	while (n != NULL &&  NODE_TYPE(n) == T_TNODE) {
		tn = (struct tnode *) n;
O
Olof Johansson 已提交
1051

1052
		check_tnode(tn);
O
Olof Johansson 已提交
1053

1054
		if (tkey_sub_equals(tn->key, pos, tn->pos-pos, key)) {
1055
			tp = tn;
O
Olof Johansson 已提交
1056
			pos = tn->pos + tn->bits;
1057 1058 1059 1060
			n = tnode_get_child(tn,
					    tkey_extract_bits(key,
							      tn->pos,
							      tn->bits));
1061

S
Stephen Hemminger 已提交
1062
			BUG_ON(n && node_parent(n) != tn);
O
Olof Johansson 已提交
1063
		} else
1064 1065 1066 1067 1068 1069
			break;
	}

	/*
	 * n  ----> NULL, LEAF or TNODE
	 *
1070
	 * tp is n's (parent) ----> NULL or TNODE
1071 1072
	 */

O
Olof Johansson 已提交
1073
	BUG_ON(tp && IS_LEAF(tp));
1074 1075 1076

	/* Case 1: n is a leaf. Compare prefixes */

1077
	if (n != NULL && IS_LEAF(n) && tkey_equals(key, n->key)) {
1078
		l = (struct leaf *) n;
1079
		li = leaf_info_new(plen);
O
Olof Johansson 已提交
1080

1081 1082
		if (!li)
			return NULL;
1083 1084 1085 1086 1087 1088 1089

		fa_head = &li->falh;
		insert_leaf_info(&l->list, li);
		goto done;
	}
	l = leaf_new();

1090 1091
	if (!l)
		return NULL;
1092 1093 1094 1095

	l->key = key;
	li = leaf_info_new(plen);

1096
	if (!li) {
1097
		free_leaf(l);
1098
		return NULL;
1099
	}
1100 1101 1102 1103 1104

	fa_head = &li->falh;
	insert_leaf_info(&l->list, li);

	if (t->trie && n == NULL) {
O
Olof Johansson 已提交
1105
		/* Case 2: n is NULL, and will just insert a new leaf */
1106

S
Stephen Hemminger 已提交
1107
		node_set_parent((struct node *)l, tp);
1108

O
Olof Johansson 已提交
1109 1110 1111 1112
		cindex = tkey_extract_bits(key, tp->pos, tp->bits);
		put_child(t, (struct tnode *)tp, cindex, (struct node *)l);
	} else {
		/* Case 3: n is a LEAF or a TNODE and the key doesn't match. */
1113 1114
		/*
		 *  Add a new tnode here
1115 1116 1117 1118
		 *  first tnode need some special handling
		 */

		if (tp)
O
Olof Johansson 已提交
1119
			pos = tp->pos+tp->bits;
1120
		else
O
Olof Johansson 已提交
1121 1122
			pos = 0;

1123
		if (n) {
1124 1125
			newpos = tkey_mismatch(key, pos, n->key);
			tn = tnode_new(n->key, newpos, 1);
O
Olof Johansson 已提交
1126
		} else {
1127
			newpos = 0;
1128
			tn = tnode_new(key, newpos, 1); /* First tnode */
1129 1130
		}

1131
		if (!tn) {
1132
			free_leaf_info(li);
1133
			free_leaf(l);
1134
			return NULL;
O
Olof Johansson 已提交
1135 1136
		}

S
Stephen Hemminger 已提交
1137
		node_set_parent((struct node *)tn, tp);
1138

O
Olof Johansson 已提交
1139
		missbit = tkey_extract_bits(key, newpos, 1);
1140 1141 1142
		put_child(t, tn, missbit, (struct node *)l);
		put_child(t, tn, 1-missbit, n);

1143
		if (tp) {
1144
			cindex = tkey_extract_bits(key, tp->pos, tp->bits);
1145 1146
			put_child(t, (struct tnode *)tp, cindex,
				  (struct node *)tn);
O
Olof Johansson 已提交
1147
		} else {
1148
			rcu_assign_pointer(t->trie, (struct node *)tn);
1149 1150 1151
			tp = tn;
		}
	}
O
Olof Johansson 已提交
1152 1153

	if (tp && tp->pos + tp->bits > 32)
1154 1155 1156
		pr_warning("fib_trie"
			   " tp=%p pos=%d, bits=%d, key=%0x plen=%d\n",
			   tp, tp->pos, tp->bits, key, plen);
O
Olof Johansson 已提交
1157

1158
	/* Rebalance the trie */
R
Robert Olsson 已提交
1159 1160

	rcu_assign_pointer(t->trie, trie_rebalance(t, tp));
1161
done:
1162 1163 1164
	return fa_head;
}

1165 1166 1167
/*
 * Caller must hold RTNL.
 */
1168
static int fn_trie_insert(struct fib_table *tb, struct fib_config *cfg)
1169 1170 1171
{
	struct trie *t = (struct trie *) tb->tb_data;
	struct fib_alias *fa, *new_fa;
1172
	struct list_head *fa_head = NULL;
1173
	struct fib_info *fi;
1174 1175
	int plen = cfg->fc_dst_len;
	u8 tos = cfg->fc_tos;
1176 1177 1178 1179 1180 1181 1182
	u32 key, mask;
	int err;
	struct leaf *l;

	if (plen > 32)
		return -EINVAL;

1183
	key = ntohl(cfg->fc_dst);
1184

1185
	pr_debug("Insert table=%u %08x/%d\n", tb->tb_id, key, plen);
1186

O
Olof Johansson 已提交
1187
	mask = ntohl(inet_make_mask(plen));
1188

1189
	if (key & ~mask)
1190 1191 1192 1193
		return -EINVAL;

	key = key & mask;

1194 1195 1196
	fi = fib_create_info(cfg);
	if (IS_ERR(fi)) {
		err = PTR_ERR(fi);
1197
		goto err;
1198
	}
1199 1200

	l = fib_find_node(t, key);
1201
	fa = NULL;
1202

1203
	if (l) {
1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 1218
		fa_head = get_fa_head(l, plen);
		fa = fib_find_alias(fa_head, tos, fi->fib_priority);
	}

	/* Now fa, if non-NULL, points to the first fib alias
	 * with the same keys [prefix,tos,priority], if such key already
	 * exists or to the node before which we will insert new one.
	 *
	 * If fa is NULL, we will need to allocate a new one and
	 * insert to the head of f.
	 *
	 * If f is NULL, no fib node matched the destination key
	 * and we need to allocate a new one of those as well.
	 */

1219 1220 1221
	if (fa && fa->fa_tos == tos &&
	    fa->fa_info->fib_priority == fi->fib_priority) {
		struct fib_alias *fa_first, *fa_match;
1222 1223

		err = -EEXIST;
1224
		if (cfg->fc_nlflags & NLM_F_EXCL)
1225 1226
			goto out;

1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247
		/* We have 2 goals:
		 * 1. Find exact match for type, scope, fib_info to avoid
		 * duplicate routes
		 * 2. Find next 'fa' (or head), NLM_F_APPEND inserts before it
		 */
		fa_match = NULL;
		fa_first = fa;
		fa = list_entry(fa->fa_list.prev, struct fib_alias, fa_list);
		list_for_each_entry_continue(fa, fa_head, fa_list) {
			if (fa->fa_tos != tos)
				break;
			if (fa->fa_info->fib_priority != fi->fib_priority)
				break;
			if (fa->fa_type == cfg->fc_type &&
			    fa->fa_scope == cfg->fc_scope &&
			    fa->fa_info == fi) {
				fa_match = fa;
				break;
			}
		}

1248
		if (cfg->fc_nlflags & NLM_F_REPLACE) {
1249 1250 1251
			struct fib_info *fi_drop;
			u8 state;

1252 1253 1254 1255
			fa = fa_first;
			if (fa_match) {
				if (fa == fa_match)
					err = 0;
1256
				goto out;
1257
			}
R
Robert Olsson 已提交
1258
			err = -ENOBUFS;
1259
			new_fa = kmem_cache_alloc(fn_alias_kmem, GFP_KERNEL);
R
Robert Olsson 已提交
1260 1261
			if (new_fa == NULL)
				goto out;
1262 1263

			fi_drop = fa->fa_info;
R
Robert Olsson 已提交
1264 1265
			new_fa->fa_tos = fa->fa_tos;
			new_fa->fa_info = fi;
1266 1267
			new_fa->fa_type = cfg->fc_type;
			new_fa->fa_scope = cfg->fc_scope;
1268
			state = fa->fa_state;
1269
			new_fa->fa_state = state & ~FA_S_ACCESSED;
1270

R
Robert Olsson 已提交
1271 1272
			list_replace_rcu(&fa->fa_list, &new_fa->fa_list);
			alias_free_mem_rcu(fa);
1273 1274 1275

			fib_release_info(fi_drop);
			if (state & FA_S_ACCESSED)
O
Olof Johansson 已提交
1276
				rt_cache_flush(-1);
1277 1278
			rtmsg_fib(RTM_NEWROUTE, htonl(key), new_fa, plen,
				tb->tb_id, &cfg->fc_nlinfo, NLM_F_REPLACE);
1279

O
Olof Johansson 已提交
1280
			goto succeeded;
1281 1282 1283 1284 1285
		}
		/* Error if we find a perfect match which
		 * uses the same scope, type, and nexthop
		 * information.
		 */
1286 1287
		if (fa_match)
			goto out;
1288

1289
		if (!(cfg->fc_nlflags & NLM_F_APPEND))
1290
			fa = fa_first;
1291 1292
	}
	err = -ENOENT;
1293
	if (!(cfg->fc_nlflags & NLM_F_CREATE))
1294 1295 1296
		goto out;

	err = -ENOBUFS;
1297
	new_fa = kmem_cache_alloc(fn_alias_kmem, GFP_KERNEL);
1298 1299 1300 1301 1302
	if (new_fa == NULL)
		goto out;

	new_fa->fa_info = fi;
	new_fa->fa_tos = tos;
1303 1304
	new_fa->fa_type = cfg->fc_type;
	new_fa->fa_scope = cfg->fc_scope;
1305 1306 1307 1308 1309
	new_fa->fa_state = 0;
	/*
	 * Insert new entry to the list.
	 */

1310
	if (!fa_head) {
1311 1312 1313
		fa_head = fib_insert_node(t, key, plen);
		if (unlikely(!fa_head)) {
			err = -ENOMEM;
1314
			goto out_free_new_fa;
1315
		}
1316
	}
1317

R
Robert Olsson 已提交
1318 1319
	list_add_tail_rcu(&new_fa->fa_list,
			  (fa ? &fa->fa_list : fa_head));
1320 1321

	rt_cache_flush(-1);
1322
	rtmsg_fib(RTM_NEWROUTE, htonl(key), new_fa, plen, tb->tb_id,
1323
		  &cfg->fc_nlinfo, 0);
1324 1325
succeeded:
	return 0;
1326 1327 1328

out_free_new_fa:
	kmem_cache_free(fn_alias_kmem, new_fa);
1329 1330
out:
	fib_release_info(fi);
O
Olof Johansson 已提交
1331
err:
1332 1333 1334
	return err;
}

R
Robert Olsson 已提交
1335
/* should be called with rcu_read_lock */
1336 1337 1338
static int check_leaf(struct trie *t, struct leaf *l,
		      t_key key,  const struct flowi *flp,
		      struct fib_result *res)
1339 1340 1341 1342
{
	struct leaf_info *li;
	struct hlist_head *hhead = &l->list;
	struct hlist_node *node;
1343

R
Robert Olsson 已提交
1344
	hlist_for_each_entry_rcu(li, node, hhead, hlist) {
1345 1346 1347 1348
		int err;
		int plen = li->plen;
		__be32 mask = inet_make_mask(plen);

1349
		if (l->key != (key & ntohl(mask)))
1350 1351
			continue;

1352 1353 1354
		err = fib_semantic_match(&li->falh, flp, res,
					 htonl(l->key), mask, plen);

1355
#ifdef CONFIG_IP_FIB_TRIE_STATS
1356
		if (err <= 0)
1357
			t->stats.semantic_match_passed++;
1358 1359
		else
			t->stats.semantic_match_miss++;
1360
#endif
1361 1362
		if (err <= 0)
			return plen;
1363
	}
1364 1365

	return -1;
1366 1367
}

1368 1369
static int fn_trie_lookup(struct fib_table *tb, const struct flowi *flp,
			  struct fib_result *res)
1370 1371 1372 1373 1374 1375
{
	struct trie *t = (struct trie *) tb->tb_data;
	int plen, ret = 0;
	struct node *n;
	struct tnode *pn;
	int pos, bits;
O
Olof Johansson 已提交
1376
	t_key key = ntohl(flp->fl4_dst);
1377 1378 1379
	int chopped_off;
	t_key cindex = 0;
	int current_prefix_length = KEYLENGTH;
O
Olof Johansson 已提交
1380 1381 1382 1383
	struct tnode *cn;
	t_key node_prefix, key_prefix, pref_mismatch;
	int mp;

R
Robert Olsson 已提交
1384
	rcu_read_lock();
O
Olof Johansson 已提交
1385

R
Robert Olsson 已提交
1386
	n = rcu_dereference(t->trie);
1387
	if (!n)
1388 1389 1390 1391 1392 1393 1394 1395
		goto failed;

#ifdef CONFIG_IP_FIB_TRIE_STATS
	t->stats.gets++;
#endif

	/* Just a leaf? */
	if (IS_LEAF(n)) {
1396 1397 1398 1399 1400
		plen = check_leaf(t, (struct leaf *)n, key, flp, res);
		if (plen < 0)
			goto failed;
		ret = 0;
		goto found;
1401
	}
1402

1403 1404
	pn = (struct tnode *) n;
	chopped_off = 0;
1405

O
Olof Johansson 已提交
1406
	while (pn) {
1407 1408 1409
		pos = pn->pos;
		bits = pn->bits;

1410
		if (!chopped_off)
S
Stephen Hemminger 已提交
1411 1412
			cindex = tkey_extract_bits(mask_pfx(key, current_prefix_length),
						   pos, bits);
1413 1414 1415 1416 1417 1418 1419 1420 1421 1422

		n = tnode_get_child(pn, cindex);

		if (n == NULL) {
#ifdef CONFIG_IP_FIB_TRIE_STATS
			t->stats.null_node_hit++;
#endif
			goto backtrace;
		}

O
Olof Johansson 已提交
1423
		if (IS_LEAF(n)) {
1424 1425
			plen = check_leaf(t, (struct leaf *)n, key, flp, res);
			if (plen < 0)
O
Olof Johansson 已提交
1426
				goto backtrace;
1427 1428 1429

			ret = 0;
			goto found;
O
Olof Johansson 已提交
1430 1431 1432
		}

		cn = (struct tnode *)n;
1433

O
Olof Johansson 已提交
1434 1435 1436 1437 1438 1439 1440 1441 1442 1443 1444 1445 1446 1447 1448
		/*
		 * It's a tnode, and we can do some extra checks here if we
		 * like, to avoid descending into a dead-end branch.
		 * This tnode is in the parent's child array at index
		 * key[p_pos..p_pos+p_bits] but potentially with some bits
		 * chopped off, so in reality the index may be just a
		 * subprefix, padded with zero at the end.
		 * We can also take a look at any skipped bits in this
		 * tnode - everything up to p_pos is supposed to be ok,
		 * and the non-chopped bits of the index (se previous
		 * paragraph) are also guaranteed ok, but the rest is
		 * considered unknown.
		 *
		 * The skipped bits are key[pos+bits..cn->pos].
		 */
1449

O
Olof Johansson 已提交
1450 1451 1452 1453 1454 1455 1456 1457 1458
		/* If current_prefix_length < pos+bits, we are already doing
		 * actual prefix  matching, which means everything from
		 * pos+(bits-chopped_off) onward must be zero along some
		 * branch of this subtree - otherwise there is *no* valid
		 * prefix present. Here we can only check the skipped
		 * bits. Remember, since we have already indexed into the
		 * parent's child array, we know that the bits we chopped of
		 * *are* zero.
		 */
1459

1460 1461
		/* NOTA BENE: Checking only skipped bits
		   for the new node here */
1462

O
Olof Johansson 已提交
1463 1464
		if (current_prefix_length < pos+bits) {
			if (tkey_extract_bits(cn->key, current_prefix_length,
1465 1466
						cn->pos - current_prefix_length)
			    || !(cn->child[0]))
O
Olof Johansson 已提交
1467 1468
				goto backtrace;
		}
1469

O
Olof Johansson 已提交
1470 1471 1472 1473 1474 1475 1476 1477 1478
		/*
		 * If chopped_off=0, the index is fully validated and we
		 * only need to look at the skipped bits for this, the new,
		 * tnode. What we actually want to do is to find out if
		 * these skipped bits match our key perfectly, or if we will
		 * have to count on finding a matching prefix further down,
		 * because if we do, we would like to have some way of
		 * verifying the existence of such a prefix at this point.
		 */
1479

O
Olof Johansson 已提交
1480 1481 1482 1483 1484 1485 1486 1487
		/* The only thing we can do at this point is to verify that
		 * any such matching prefix can indeed be a prefix to our
		 * key, and if the bits in the node we are inspecting that
		 * do not match our key are not ZERO, this cannot be true.
		 * Thus, find out where there is a mismatch (before cn->pos)
		 * and verify that all the mismatching bits are zero in the
		 * new tnode's key.
		 */
1488

1489 1490 1491 1492 1493 1494 1495 1496 1497 1498 1499
		/*
		 * Note: We aren't very concerned about the piece of
		 * the key that precede pn->pos+pn->bits, since these
		 * have already been checked. The bits after cn->pos
		 * aren't checked since these are by definition
		 * "unknown" at this point. Thus, what we want to see
		 * is if we are about to enter the "prefix matching"
		 * state, and in that case verify that the skipped
		 * bits that will prevail throughout this subtree are
		 * zero, as they have to be if we are to find a
		 * matching prefix.
O
Olof Johansson 已提交
1500 1501
		 */

S
Stephen Hemminger 已提交
1502 1503
		node_prefix = mask_pfx(cn->key, cn->pos);
		key_prefix = mask_pfx(key, cn->pos);
O
Olof Johansson 已提交
1504 1505 1506
		pref_mismatch = key_prefix^node_prefix;
		mp = 0;

1507 1508 1509 1510
		/*
		 * In short: If skipped bits in this node do not match
		 * the search key, enter the "prefix matching"
		 * state.directly.
O
Olof Johansson 已提交
1511 1512 1513 1514
		 */
		if (pref_mismatch) {
			while (!(pref_mismatch & (1<<(KEYLENGTH-1)))) {
				mp++;
1515
				pref_mismatch = pref_mismatch << 1;
O
Olof Johansson 已提交
1516 1517 1518 1519 1520 1521 1522 1523
			}
			key_prefix = tkey_extract_bits(cn->key, mp, cn->pos-mp);

			if (key_prefix != 0)
				goto backtrace;

			if (current_prefix_length >= cn->pos)
				current_prefix_length = mp;
1524
		}
1525

O
Olof Johansson 已提交
1526 1527 1528 1529
		pn = (struct tnode *)n; /* Descend */
		chopped_off = 0;
		continue;

1530 1531 1532 1533
backtrace:
		chopped_off++;

		/* As zero don't change the child key (cindex) */
1534 1535
		while ((chopped_off <= pn->bits)
		       && !(cindex & (1<<(chopped_off-1))))
1536 1537 1538 1539
			chopped_off++;

		/* Decrease current_... with bits chopped off */
		if (current_prefix_length > pn->pos + pn->bits - chopped_off)
1540 1541
			current_prefix_length = pn->pos + pn->bits
				- chopped_off;
O
Olof Johansson 已提交
1542

1543
		/*
1544
		 * Either we do the actual chop off according or if we have
1545 1546 1547
		 * chopped off all bits in this tnode walk up to our parent.
		 */

O
Olof Johansson 已提交
1548
		if (chopped_off <= pn->bits) {
1549
			cindex &= ~(1 << (chopped_off-1));
O
Olof Johansson 已提交
1550
		} else {
S
Stephen Hemminger 已提交
1551 1552
			struct tnode *parent = node_parent((struct node *) pn);
			if (!parent)
1553
				goto failed;
O
Olof Johansson 已提交
1554

1555
			/* Get Child's index */
S
Stephen Hemminger 已提交
1556 1557
			cindex = tkey_extract_bits(pn->key, parent->pos, parent->bits);
			pn = parent;
1558 1559 1560 1561 1562 1563
			chopped_off = 0;

#ifdef CONFIG_IP_FIB_TRIE_STATS
			t->stats.backtrack++;
#endif
			goto backtrace;
1564
		}
1565 1566
	}
failed:
1567
	ret = 1;
1568
found:
R
Robert Olsson 已提交
1569
	rcu_read_unlock();
1570 1571 1572
	return ret;
}

1573 1574 1575 1576
/*
 * Remove the leaf and return parent.
 */
static void trie_leaf_remove(struct trie *t, struct leaf *l)
1577
{
1578
	struct tnode *tp = node_parent((struct node *) l);
1579

1580
	pr_debug("entering trie_leaf_remove(%p)\n", l);
1581

1582
	if (tp) {
1583
		t_key cindex = tkey_extract_bits(l->key, tp->pos, tp->bits);
1584
		put_child(t, (struct tnode *)tp, cindex, NULL);
R
Robert Olsson 已提交
1585
		rcu_assign_pointer(t->trie, trie_rebalance(t, tp));
O
Olof Johansson 已提交
1586
	} else
R
Robert Olsson 已提交
1587
		rcu_assign_pointer(t->trie, NULL);
1588

1589
	free_leaf(l);
1590 1591
}

1592 1593 1594
/*
 * Caller must hold RTNL.
 */
1595
static int fn_trie_delete(struct fib_table *tb, struct fib_config *cfg)
1596 1597 1598
{
	struct trie *t = (struct trie *) tb->tb_data;
	u32 key, mask;
1599 1600
	int plen = cfg->fc_dst_len;
	u8 tos = cfg->fc_tos;
1601 1602 1603
	struct fib_alias *fa, *fa_to_delete;
	struct list_head *fa_head;
	struct leaf *l;
O
Olof Johansson 已提交
1604 1605
	struct leaf_info *li;

1606
	if (plen > 32)
1607 1608
		return -EINVAL;

1609
	key = ntohl(cfg->fc_dst);
O
Olof Johansson 已提交
1610
	mask = ntohl(inet_make_mask(plen));
1611

1612
	if (key & ~mask)
1613 1614 1615 1616 1617
		return -EINVAL;

	key = key & mask;
	l = fib_find_node(t, key);

1618
	if (!l)
1619 1620 1621 1622 1623 1624 1625 1626
		return -ESRCH;

	fa_head = get_fa_head(l, plen);
	fa = fib_find_alias(fa_head, tos, 0);

	if (!fa)
		return -ESRCH;

S
Stephen Hemminger 已提交
1627
	pr_debug("Deleting %08x/%d tos=%d t=%p\n", key, plen, tos, t);
1628 1629

	fa_to_delete = NULL;
1630 1631
	fa = list_entry(fa->fa_list.prev, struct fib_alias, fa_list);
	list_for_each_entry_continue(fa, fa_head, fa_list) {
1632 1633 1634 1635 1636
		struct fib_info *fi = fa->fa_info;

		if (fa->fa_tos != tos)
			break;

1637 1638 1639 1640 1641 1642
		if ((!cfg->fc_type || fa->fa_type == cfg->fc_type) &&
		    (cfg->fc_scope == RT_SCOPE_NOWHERE ||
		     fa->fa_scope == cfg->fc_scope) &&
		    (!cfg->fc_protocol ||
		     fi->fib_protocol == cfg->fc_protocol) &&
		    fib_nh_match(cfg, fi) == 0) {
1643 1644 1645 1646 1647
			fa_to_delete = fa;
			break;
		}
	}

O
Olof Johansson 已提交
1648 1649
	if (!fa_to_delete)
		return -ESRCH;
1650

O
Olof Johansson 已提交
1651
	fa = fa_to_delete;
1652
	rtmsg_fib(RTM_DELROUTE, htonl(key), fa, plen, tb->tb_id,
1653
		  &cfg->fc_nlinfo, 0);
O
Olof Johansson 已提交
1654 1655

	l = fib_find_node(t, key);
R
Robert Olsson 已提交
1656
	li = find_leaf_info(l, plen);
1657

R
Robert Olsson 已提交
1658
	list_del_rcu(&fa->fa_list);
1659

O
Olof Johansson 已提交
1660
	if (list_empty(fa_head)) {
R
Robert Olsson 已提交
1661
		hlist_del_rcu(&li->hlist);
O
Olof Johansson 已提交
1662
		free_leaf_info(li);
R
Robert Olsson 已提交
1663
	}
1664

O
Olof Johansson 已提交
1665
	if (hlist_empty(&l->list))
1666
		trie_leaf_remove(t, l);
1667

O
Olof Johansson 已提交
1668 1669
	if (fa->fa_state & FA_S_ACCESSED)
		rt_cache_flush(-1);
1670

R
Robert Olsson 已提交
1671 1672
	fib_release_info(fa->fa_info);
	alias_free_mem_rcu(fa);
O
Olof Johansson 已提交
1673
	return 0;
1674 1675
}

1676
static int trie_flush_list(struct list_head *head)
1677 1678 1679 1680 1681 1682 1683
{
	struct fib_alias *fa, *fa_node;
	int found = 0;

	list_for_each_entry_safe(fa, fa_node, head, fa_list) {
		struct fib_info *fi = fa->fa_info;

R
Robert Olsson 已提交
1684 1685 1686 1687
		if (fi && (fi->fib_flags & RTNH_F_DEAD)) {
			list_del_rcu(&fa->fa_list);
			fib_release_info(fa->fa_info);
			alias_free_mem_rcu(fa);
1688 1689 1690 1691 1692 1693
			found++;
		}
	}
	return found;
}

1694
static int trie_flush_leaf(struct leaf *l)
1695 1696 1697 1698 1699 1700 1701
{
	int found = 0;
	struct hlist_head *lih = &l->list;
	struct hlist_node *node, *tmp;
	struct leaf_info *li = NULL;

	hlist_for_each_entry_safe(li, node, tmp, lih, hlist) {
1702
		found += trie_flush_list(&li->falh);
1703 1704

		if (list_empty(&li->falh)) {
R
Robert Olsson 已提交
1705
			hlist_del_rcu(&li->hlist);
1706 1707 1708 1709 1710 1711
			free_leaf_info(li);
		}
	}
	return found;
}

1712 1713 1714 1715 1716
/*
 * Scan for the next right leaf starting at node p->child[idx]
 * Since we have back pointer, no recursion necessary.
 */
static struct leaf *leaf_walk_rcu(struct tnode *p, struct node *c)
1717
{
1718 1719
	do {
		t_key idx;
1720 1721

		if (c)
1722
			idx = tkey_extract_bits(c->key, p->pos, p->bits) + 1;
1723
		else
1724
			idx = 0;
R
Robert Olsson 已提交
1725

1726 1727
		while (idx < 1u << p->bits) {
			c = tnode_get_child_rcu(p, idx++);
R
Robert Olsson 已提交
1728
			if (!c)
O
Olof Johansson 已提交
1729 1730
				continue;

1731 1732 1733
			if (IS_LEAF(c)) {
				prefetch(p->child[idx]);
				return (struct leaf *) c;
1734
			}
1735 1736 1737 1738

			/* Rescan start scanning in new node */
			p = (struct tnode *) c;
			idx = 0;
1739
		}
1740 1741

		/* Node empty, walk back up to parent */
O
Olof Johansson 已提交
1742
		c = (struct node *) p;
1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756 1757 1758 1759 1760 1761 1762 1763 1764 1765 1766 1767 1768 1769
	} while ( (p = node_parent_rcu(c)) != NULL);

	return NULL; /* Root of trie */
}

static struct leaf *trie_firstleaf(struct trie *t)
{
	struct tnode *n = (struct tnode *) rcu_dereference(t->trie);

	if (!n)
		return NULL;

	if (IS_LEAF(n))          /* trie is just a leaf */
		return (struct leaf *) n;

	return leaf_walk_rcu(n, NULL);
}

static struct leaf *trie_nextleaf(struct leaf *l)
{
	struct node *c = (struct node *) l;
	struct tnode *p = node_parent(c);

	if (!p)
		return NULL;	/* trie with just one leaf */

	return leaf_walk_rcu(p, c);
1770 1771
}

1772 1773 1774 1775
static struct leaf *trie_leafindex(struct trie *t, int index)
{
	struct leaf *l = trie_firstleaf(t);

S
Stephen Hemminger 已提交
1776
	while (l && index-- > 0)
1777
		l = trie_nextleaf(l);
S
Stephen Hemminger 已提交
1778

1779 1780 1781 1782
	return l;
}


1783 1784 1785
/*
 * Caller must hold RTNL.
 */
1786 1787 1788
static int fn_trie_flush(struct fib_table *tb)
{
	struct trie *t = (struct trie *) tb->tb_data;
1789
	struct leaf *l, *ll = NULL;
1790
	int found = 0;
1791

1792
	for (l = trie_firstleaf(t); l; l = trie_nextleaf(l)) {
1793
		found += trie_flush_leaf(l);
1794 1795

		if (ll && hlist_empty(&ll->list))
1796
			trie_leaf_remove(t, ll);
1797 1798 1799 1800
		ll = l;
	}

	if (ll && hlist_empty(&ll->list))
1801
		trie_leaf_remove(t, ll);
1802

S
Stephen Hemminger 已提交
1803
	pr_debug("trie_flush found=%d\n", found);
1804 1805 1806
	return found;
}

1807 1808 1809
static void fn_trie_select_default(struct fib_table *tb,
				   const struct flowi *flp,
				   struct fib_result *res)
1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822
{
	struct trie *t = (struct trie *) tb->tb_data;
	int order, last_idx;
	struct fib_info *fi = NULL;
	struct fib_info *last_resort;
	struct fib_alias *fa = NULL;
	struct list_head *fa_head;
	struct leaf *l;

	last_idx = -1;
	last_resort = NULL;
	order = -1;

R
Robert Olsson 已提交
1823
	rcu_read_lock();
1824

1825
	l = fib_find_node(t, 0);
1826
	if (!l)
1827 1828 1829
		goto out;

	fa_head = get_fa_head(l, 0);
1830
	if (!fa_head)
1831 1832
		goto out;

1833
	if (list_empty(fa_head))
1834 1835
		goto out;

R
Robert Olsson 已提交
1836
	list_for_each_entry_rcu(fa, fa_head, fa_list) {
1837
		struct fib_info *next_fi = fa->fa_info;
O
Olof Johansson 已提交
1838

1839 1840 1841
		if (fa->fa_scope != res->scope ||
		    fa->fa_type != RTN_UNICAST)
			continue;
O
Olof Johansson 已提交
1842

1843 1844 1845 1846 1847 1848
		if (next_fi->fib_priority > res->fi->fib_priority)
			break;
		if (!next_fi->fib_nh[0].nh_gw ||
		    next_fi->fib_nh[0].nh_scope != RT_SCOPE_LINK)
			continue;
		fa->fa_state |= FA_S_ACCESSED;
O
Olof Johansson 已提交
1849

1850 1851 1852 1853
		if (fi == NULL) {
			if (next_fi != res->fi)
				break;
		} else if (!fib_detect_death(fi, order, &last_resort,
1854
					     &last_idx, tb->tb_default)) {
1855
			fib_result_assign(res, fi);
1856
			tb->tb_default = order;
1857 1858 1859 1860 1861 1862
			goto out;
		}
		fi = next_fi;
		order++;
	}
	if (order <= 0 || fi == NULL) {
1863
		tb->tb_default = -1;
1864 1865 1866
		goto out;
	}

1867 1868
	if (!fib_detect_death(fi, order, &last_resort, &last_idx,
				tb->tb_default)) {
1869
		fib_result_assign(res, fi);
1870
		tb->tb_default = order;
1871 1872
		goto out;
	}
1873 1874
	if (last_idx >= 0)
		fib_result_assign(res, last_resort);
1875 1876
	tb->tb_default = last_idx;
out:
R
Robert Olsson 已提交
1877
	rcu_read_unlock();
1878 1879
}

1880 1881
static int fn_trie_dump_fa(t_key key, int plen, struct list_head *fah,
			   struct fib_table *tb,
1882 1883 1884 1885
			   struct sk_buff *skb, struct netlink_callback *cb)
{
	int i, s_i;
	struct fib_alias *fa;
A
Al Viro 已提交
1886
	__be32 xkey = htonl(key);
1887

1888
	s_i = cb->args[5];
1889 1890
	i = 0;

R
Robert Olsson 已提交
1891 1892 1893
	/* rcu_read_lock is hold by caller */

	list_for_each_entry_rcu(fa, fah, fa_list) {
1894 1895 1896 1897 1898 1899 1900 1901 1902 1903 1904
		if (i < s_i) {
			i++;
			continue;
		}

		if (fib_dump_info(skb, NETLINK_CB(cb->skb).pid,
				  cb->nlh->nlmsg_seq,
				  RTM_NEWROUTE,
				  tb->tb_id,
				  fa->fa_type,
				  fa->fa_scope,
1905
				  xkey,
1906 1907
				  plen,
				  fa->fa_tos,
1908
				  fa->fa_info, NLM_F_MULTI) < 0) {
1909
			cb->args[5] = i;
1910
			return -1;
O
Olof Johansson 已提交
1911
		}
1912 1913
		i++;
	}
1914
	cb->args[5] = i;
1915 1916 1917
	return skb->len;
}

1918 1919
static int fn_trie_dump_leaf(struct leaf *l, struct fib_table *tb,
			struct sk_buff *skb, struct netlink_callback *cb)
1920
{
1921 1922 1923
	struct leaf_info *li;
	struct hlist_node *node;
	int i, s_i;
1924

1925
	s_i = cb->args[4];
1926
	i = 0;
1927

1928 1929 1930 1931
	/* rcu_read_lock is hold by caller */
	hlist_for_each_entry_rcu(li, node, &l->list, hlist) {
		if (i < s_i) {
			i++;
1932
			continue;
1933
		}
O
Olof Johansson 已提交
1934

1935
		if (i > s_i)
1936
			cb->args[5] = 0;
1937

1938
		if (list_empty(&li->falh))
1939 1940
			continue;

1941
		if (fn_trie_dump_fa(l->key, li->plen, &li->falh, tb, skb, cb) < 0) {
1942
			cb->args[4] = i;
1943 1944
			return -1;
		}
1945
		i++;
1946
	}
1947

1948
	cb->args[4] = i;
1949 1950 1951
	return skb->len;
}

1952 1953
static int fn_trie_dump(struct fib_table *tb, struct sk_buff *skb,
			struct netlink_callback *cb)
1954
{
1955
	struct leaf *l;
1956
	struct trie *t = (struct trie *) tb->tb_data;
1957
	t_key key = cb->args[2];
1958
	int count = cb->args[3];
1959

R
Robert Olsson 已提交
1960
	rcu_read_lock();
1961 1962 1963
	/* Dump starting at last key.
	 * Note: 0.0.0.0/0 (ie default) is first key.
	 */
1964
	if (count == 0)
1965 1966
		l = trie_firstleaf(t);
	else {
1967 1968 1969
		/* Normally, continue from last key, but if that is missing
		 * fallback to using slow rescan
		 */
1970
		l = fib_find_node(t, key);
1971 1972
		if (!l)
			l = trie_leafindex(t, count);
1973
	}
1974

1975 1976
	while (l) {
		cb->args[2] = l->key;
1977
		if (fn_trie_dump_leaf(l, tb, skb, cb) < 0) {
1978
			cb->args[3] = count;
1979 1980
			rcu_read_unlock();
			return -1;
1981
		}
1982

1983
		++count;
1984
		l = trie_nextleaf(l);
1985 1986
		memset(&cb->args[4], 0,
		       sizeof(cb->args) - 4*sizeof(cb->args[0]));
1987
	}
1988
	cb->args[3] = count;
R
Robert Olsson 已提交
1989
	rcu_read_unlock();
1990

1991 1992 1993
	return skb->len;
}

1994 1995
void __init fib_hash_init(void)
{
1996 1997
	fn_alias_kmem = kmem_cache_create("ip_fib_alias",
					  sizeof(struct fib_alias),
1998 1999 2000 2001 2002 2003
					  0, SLAB_PANIC, NULL);

	trie_leaf_kmem = kmem_cache_create("ip_fib_trie",
					   max(sizeof(struct leaf),
					       sizeof(struct leaf_info)),
					   0, SLAB_PANIC, NULL);
2004
}
2005

2006 2007 2008

/* Fix more generic FIB names for init later */
struct fib_table *fib_hash_table(u32 id)
2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
{
	struct fib_table *tb;
	struct trie *t;

	tb = kmalloc(sizeof(struct fib_table) + sizeof(struct trie),
		     GFP_KERNEL);
	if (tb == NULL)
		return NULL;

	tb->tb_id = id;
2019
	tb->tb_default = -1;
2020 2021 2022 2023 2024 2025 2026 2027
	tb->tb_lookup = fn_trie_lookup;
	tb->tb_insert = fn_trie_insert;
	tb->tb_delete = fn_trie_delete;
	tb->tb_flush = fn_trie_flush;
	tb->tb_select_default = fn_trie_select_default;
	tb->tb_dump = fn_trie_dump;

	t = (struct trie *) tb->tb_data;
2028
	memset(t, 0, sizeof(*t));
2029 2030

	if (id == RT_TABLE_LOCAL)
2031
		pr_info("IPv4 FIB: Using LC-trie version %s\n", VERSION);
2032 2033 2034 2035

	return tb;
}

2036 2037 2038
#ifdef CONFIG_PROC_FS
/* Depth first Trie walk iterator */
struct fib_trie_iter {
2039
	struct seq_net_private p;
2040
	struct fib_table *tb;
2041 2042 2043 2044
	struct tnode *tnode;
	unsigned index;
	unsigned depth;
};
2045

2046
static struct node *fib_trie_get_next(struct fib_trie_iter *iter)
2047
{
2048 2049 2050
	struct tnode *tn = iter->tnode;
	unsigned cindex = iter->index;
	struct tnode *p;
2051

2052 2053 2054 2055
	/* A single entry routing table */
	if (!tn)
		return NULL;

2056 2057 2058 2059
	pr_debug("get_next iter={node=%p index=%d depth=%d}\n",
		 iter->tnode, iter->index, iter->depth);
rescan:
	while (cindex < (1<<tn->bits)) {
2060
		struct node *n = tnode_get_child_rcu(tn, cindex);
2061

2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072 2073
		if (n) {
			if (IS_LEAF(n)) {
				iter->tnode = tn;
				iter->index = cindex + 1;
			} else {
				/* push down one level */
				iter->tnode = (struct tnode *) n;
				iter->index = 0;
				++iter->depth;
			}
			return n;
		}
2074

2075 2076
		++cindex;
	}
O
Olof Johansson 已提交
2077

2078
	/* Current node exhausted, pop back up */
2079
	p = node_parent_rcu((struct node *)tn);
2080 2081 2082 2083 2084
	if (p) {
		cindex = tkey_extract_bits(tn->key, p->pos, p->bits)+1;
		tn = p;
		--iter->depth;
		goto rescan;
2085
	}
2086 2087 2088

	/* got root? */
	return NULL;
2089 2090
}

2091 2092
static struct node *fib_trie_get_first(struct fib_trie_iter *iter,
				       struct trie *t)
2093
{
2094
	struct node *n;
2095

S
Stephen Hemminger 已提交
2096
	if (!t)
2097 2098 2099
		return NULL;

	n = rcu_dereference(t->trie);
2100
	if (!n)
2101
		return NULL;
2102

2103 2104 2105 2106 2107 2108 2109 2110
	if (IS_TNODE(n)) {
		iter->tnode = (struct tnode *) n;
		iter->index = 0;
		iter->depth = 1;
	} else {
		iter->tnode = NULL;
		iter->index = 0;
		iter->depth = 0;
O
Olof Johansson 已提交
2111
	}
2112 2113

	return n;
2114
}
O
Olof Johansson 已提交
2115

2116 2117 2118 2119
static void trie_collect_stats(struct trie *t, struct trie_stat *s)
{
	struct node *n;
	struct fib_trie_iter iter;
O
Olof Johansson 已提交
2120

2121
	memset(s, 0, sizeof(*s));
O
Olof Johansson 已提交
2122

2123
	rcu_read_lock();
2124
	for (n = fib_trie_get_first(&iter, t); n; n = fib_trie_get_next(&iter)) {
2125
		if (IS_LEAF(n)) {
2126 2127 2128 2129
			struct leaf *l = (struct leaf *)n;
			struct leaf_info *li;
			struct hlist_node *tmp;

2130 2131 2132 2133
			s->leaves++;
			s->totdepth += iter.depth;
			if (iter.depth > s->maxdepth)
				s->maxdepth = iter.depth;
2134 2135 2136

			hlist_for_each_entry_rcu(li, tmp, &l->list, hlist)
				++s->prefixes;
2137 2138 2139 2140 2141
		} else {
			const struct tnode *tn = (const struct tnode *) n;
			int i;

			s->tnodes++;
S
Stephen Hemminger 已提交
2142
			if (tn->bits < MAX_STAT_DEPTH)
R
Robert Olsson 已提交
2143 2144
				s->nodesizes[tn->bits]++;

2145 2146 2147
			for (i = 0; i < (1<<tn->bits); i++)
				if (!tn->child[i])
					s->nullpointers++;
2148 2149
		}
	}
R
Robert Olsson 已提交
2150
	rcu_read_unlock();
2151 2152
}

2153 2154 2155 2156
/*
 *	This outputs /proc/net/fib_triestats
 */
static void trie_show_stats(struct seq_file *seq, struct trie_stat *stat)
2157
{
2158
	unsigned i, max, pointers, bytes, avdepth;
2159

2160 2161 2162 2163
	if (stat->leaves)
		avdepth = stat->totdepth*100 / stat->leaves;
	else
		avdepth = 0;
O
Olof Johansson 已提交
2164

2165 2166
	seq_printf(seq, "\tAver depth:     %u.%02d\n",
		   avdepth / 100, avdepth % 100);
2167
	seq_printf(seq, "\tMax depth:      %u\n", stat->maxdepth);
O
Olof Johansson 已提交
2168

2169 2170
	seq_printf(seq, "\tLeaves:         %u\n", stat->leaves);
	bytes = sizeof(struct leaf) * stat->leaves;
2171 2172 2173 2174

	seq_printf(seq, "\tPrefixes:       %u\n", stat->prefixes);
	bytes += sizeof(struct leaf_info) * stat->prefixes;

2175
	seq_printf(seq, "\tInternal nodes: %u\n\t", stat->tnodes);
2176
	bytes += sizeof(struct tnode) * stat->tnodes;
2177

R
Robert Olsson 已提交
2178 2179
	max = MAX_STAT_DEPTH;
	while (max > 0 && stat->nodesizes[max-1] == 0)
2180
		max--;
2181

2182 2183 2184
	pointers = 0;
	for (i = 1; i <= max; i++)
		if (stat->nodesizes[i] != 0) {
2185
			seq_printf(seq, "  %u: %u",  i, stat->nodesizes[i]);
2186 2187 2188
			pointers += (1<<i) * stat->nodesizes[i];
		}
	seq_putc(seq, '\n');
2189
	seq_printf(seq, "\tPointers: %u\n", pointers);
R
Robert Olsson 已提交
2190

2191
	bytes += sizeof(struct node *) * pointers;
2192 2193
	seq_printf(seq, "Null ptrs: %u\n", stat->nullpointers);
	seq_printf(seq, "Total size: %u  kB\n", (bytes + 1023) / 1024);
2194
}
R
Robert Olsson 已提交
2195

2196
#ifdef CONFIG_IP_FIB_TRIE_STATS
2197 2198 2199 2200
static void trie_show_usage(struct seq_file *seq,
			    const struct trie_use_stats *stats)
{
	seq_printf(seq, "\nCounters:\n---------\n");
2201 2202 2203 2204 2205 2206 2207 2208 2209
	seq_printf(seq, "gets = %u\n", stats->gets);
	seq_printf(seq, "backtracks = %u\n", stats->backtrack);
	seq_printf(seq, "semantic match passed = %u\n",
		   stats->semantic_match_passed);
	seq_printf(seq, "semantic match miss = %u\n",
		   stats->semantic_match_miss);
	seq_printf(seq, "null node hit= %u\n", stats->null_node_hit);
	seq_printf(seq, "skipped node resize = %u\n\n",
		   stats->resize_node_skipped);
2210
}
2211 2212
#endif /*  CONFIG_IP_FIB_TRIE_STATS */

2213
static void fib_table_print(struct seq_file *seq, struct fib_table *tb)
2214
{
2215 2216 2217 2218 2219 2220
	if (tb->tb_id == RT_TABLE_LOCAL)
		seq_puts(seq, "Local:\n");
	else if (tb->tb_id == RT_TABLE_MAIN)
		seq_puts(seq, "Main:\n");
	else
		seq_printf(seq, "Id %d:\n", tb->tb_id);
2221
}
2222

2223

2224 2225
static int fib_triestat_seq_show(struct seq_file *seq, void *v)
{
2226
	struct net *net = (struct net *)seq->private;
2227
	unsigned int h;
2228

2229
	seq_printf(seq,
2230 2231
		   "Basic info: size of leaf:"
		   " %Zd bytes, size of tnode: %Zd bytes.\n",
2232 2233
		   sizeof(struct leaf), sizeof(struct tnode));

2234 2235 2236 2237 2238 2239 2240 2241
	for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
		struct hlist_head *head = &net->ipv4.fib_table_hash[h];
		struct hlist_node *node;
		struct fib_table *tb;

		hlist_for_each_entry_rcu(tb, node, head, tb_hlist) {
			struct trie *t = (struct trie *) tb->tb_data;
			struct trie_stat stat;
2242

2243 2244 2245 2246 2247 2248 2249 2250 2251 2252 2253 2254
			if (!t)
				continue;

			fib_table_print(seq, tb);

			trie_collect_stats(t, &stat);
			trie_show_stats(seq, &stat);
#ifdef CONFIG_IP_FIB_TRIE_STATS
			trie_show_usage(seq, &t->stats);
#endif
		}
	}
2255

2256
	return 0;
2257 2258
}

2259
static int fib_triestat_seq_open(struct inode *inode, struct file *file)
2260
{
2261 2262 2263 2264 2265 2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277 2278 2279
	int err;
	struct net *net;

	net = get_proc_net(inode);
	if (net == NULL)
		return -ENXIO;
	err = single_open(file, fib_triestat_seq_show, net);
	if (err < 0) {
		put_net(net);
		return err;
	}
	return 0;
}

static int fib_triestat_seq_release(struct inode *ino, struct file *f)
{
	struct seq_file *seq = f->private_data;
	put_net(seq->private);
	return single_release(ino, f);
2280 2281
}

2282
static const struct file_operations fib_triestat_fops = {
2283 2284 2285 2286
	.owner	= THIS_MODULE,
	.open	= fib_triestat_seq_open,
	.read	= seq_read,
	.llseek	= seq_lseek,
2287
	.release = fib_triestat_seq_release,
2288 2289
};

2290
static struct node *fib_trie_get_idx(struct seq_file *seq, loff_t pos)
2291
{
2292 2293
	struct fib_trie_iter *iter = seq->private;
	struct net *net = seq_file_net(seq);
2294
	loff_t idx = 0;
2295
	unsigned int h;
2296

2297 2298 2299 2300
	for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
		struct hlist_head *head = &net->ipv4.fib_table_hash[h];
		struct hlist_node *node;
		struct fib_table *tb;
2301

2302 2303 2304 2305 2306 2307 2308 2309 2310 2311 2312
		hlist_for_each_entry_rcu(tb, node, head, tb_hlist) {
			struct node *n;

			for (n = fib_trie_get_first(iter,
						    (struct trie *) tb->tb_data);
			     n; n = fib_trie_get_next(iter))
				if (pos == idx++) {
					iter->tb = tb;
					return n;
				}
		}
2313
	}
2314

2315 2316 2317
	return NULL;
}

2318
static void *fib_trie_seq_start(struct seq_file *seq, loff_t *pos)
2319
	__acquires(RCU)
2320
{
2321
	rcu_read_lock();
2322
	return fib_trie_get_idx(seq, *pos);
2323 2324
}

2325
static void *fib_trie_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2326
{
2327
	struct fib_trie_iter *iter = seq->private;
2328
	struct net *net = seq_file_net(seq);
2329 2330 2331 2332
	struct fib_table *tb = iter->tb;
	struct hlist_node *tb_node;
	unsigned int h;
	struct node *n;
2333

2334
	++*pos;
2335 2336 2337 2338
	/* next node in same table */
	n = fib_trie_get_next(iter);
	if (n)
		return n;
2339

2340 2341 2342 2343 2344 2345 2346 2347
	/* walk rest of this hash chain */
	h = tb->tb_id & (FIB_TABLE_HASHSZ - 1);
	while ( (tb_node = rcu_dereference(tb->tb_hlist.next)) ) {
		tb = hlist_entry(tb_node, struct fib_table, tb_hlist);
		n = fib_trie_get_first(iter, (struct trie *) tb->tb_data);
		if (n)
			goto found;
	}
2348

2349 2350 2351 2352 2353 2354 2355 2356 2357
	/* new hash chain */
	while (++h < FIB_TABLE_HASHSZ) {
		struct hlist_head *head = &net->ipv4.fib_table_hash[h];
		hlist_for_each_entry_rcu(tb, tb_node, head, tb_hlist) {
			n = fib_trie_get_first(iter, (struct trie *) tb->tb_data);
			if (n)
				goto found;
		}
	}
2358
	return NULL;
2359 2360 2361 2362

found:
	iter->tb = tb;
	return n;
2363
}
2364

2365
static void fib_trie_seq_stop(struct seq_file *seq, void *v)
2366
	__releases(RCU)
2367
{
2368 2369
	rcu_read_unlock();
}
O
Olof Johansson 已提交
2370

2371 2372 2373 2374
static void seq_indent(struct seq_file *seq, int n)
{
	while (n-- > 0) seq_puts(seq, "   ");
}
2375

2376
static inline const char *rtn_scope(char *buf, size_t len, enum rt_scope_t s)
2377
{
S
Stephen Hemminger 已提交
2378
	switch (s) {
2379 2380 2381 2382 2383 2384
	case RT_SCOPE_UNIVERSE: return "universe";
	case RT_SCOPE_SITE:	return "site";
	case RT_SCOPE_LINK:	return "link";
	case RT_SCOPE_HOST:	return "host";
	case RT_SCOPE_NOWHERE:	return "nowhere";
	default:
2385
		snprintf(buf, len, "scope=%d", s);
2386 2387 2388
		return buf;
	}
}
2389

2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403
static const char *rtn_type_names[__RTN_MAX] = {
	[RTN_UNSPEC] = "UNSPEC",
	[RTN_UNICAST] = "UNICAST",
	[RTN_LOCAL] = "LOCAL",
	[RTN_BROADCAST] = "BROADCAST",
	[RTN_ANYCAST] = "ANYCAST",
	[RTN_MULTICAST] = "MULTICAST",
	[RTN_BLACKHOLE] = "BLACKHOLE",
	[RTN_UNREACHABLE] = "UNREACHABLE",
	[RTN_PROHIBIT] = "PROHIBIT",
	[RTN_THROW] = "THROW",
	[RTN_NAT] = "NAT",
	[RTN_XRESOLVE] = "XRESOLVE",
};
2404

2405
static inline const char *rtn_type(char *buf, size_t len, unsigned t)
2406 2407 2408
{
	if (t < __RTN_MAX && rtn_type_names[t])
		return rtn_type_names[t];
2409
	snprintf(buf, len, "type %u", t);
2410
	return buf;
2411 2412
}

2413 2414
/* Pretty print the trie */
static int fib_trie_seq_show(struct seq_file *seq, void *v)
2415
{
2416 2417
	const struct fib_trie_iter *iter = seq->private;
	struct node *n = v;
2418

2419 2420
	if (!node_parent_rcu(n))
		fib_table_print(seq, iter->tb);
2421

2422 2423
	if (IS_TNODE(n)) {
		struct tnode *tn = (struct tnode *) n;
S
Stephen Hemminger 已提交
2424
		__be32 prf = htonl(mask_pfx(tn->key, tn->pos));
O
Olof Johansson 已提交
2425

2426
		seq_indent(seq, iter->depth-1);
2427
		seq_printf(seq, "  +-- " NIPQUAD_FMT "/%d %d %d %d\n",
2428
			   NIPQUAD(prf), tn->pos, tn->bits, tn->full_children,
2429
			   tn->empty_children);
2430

2431 2432
	} else {
		struct leaf *l = (struct leaf *) n;
2433 2434
		struct leaf_info *li;
		struct hlist_node *node;
A
Al Viro 已提交
2435
		__be32 val = htonl(l->key);
2436 2437

		seq_indent(seq, iter->depth);
2438
		seq_printf(seq, "  |-- " NIPQUAD_FMT "\n", NIPQUAD(val));
2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452

		hlist_for_each_entry_rcu(li, node, &l->list, hlist) {
			struct fib_alias *fa;

			list_for_each_entry_rcu(fa, &li->falh, fa_list) {
				char buf1[32], buf2[32];

				seq_indent(seq, iter->depth+1);
				seq_printf(seq, "  /%d %s %s", li->plen,
					   rtn_scope(buf1, sizeof(buf1),
						     fa->fa_scope),
					   rtn_type(buf2, sizeof(buf2),
						    fa->fa_type));
				if (fa->fa_tos)
2453
					seq_printf(seq, " tos=%d", fa->fa_tos);
2454
				seq_putc(seq, '\n');
2455 2456
			}
		}
2457
	}
2458

2459 2460 2461
	return 0;
}

2462
static const struct seq_operations fib_trie_seq_ops = {
2463 2464 2465 2466
	.start  = fib_trie_seq_start,
	.next   = fib_trie_seq_next,
	.stop   = fib_trie_seq_stop,
	.show   = fib_trie_seq_show,
2467 2468
};

2469
static int fib_trie_seq_open(struct inode *inode, struct file *file)
2470
{
2471 2472
	return seq_open_net(inode, file, &fib_trie_seq_ops,
			    sizeof(struct fib_trie_iter));
2473 2474
}

2475
static const struct file_operations fib_trie_fops = {
2476 2477 2478 2479
	.owner  = THIS_MODULE,
	.open   = fib_trie_seq_open,
	.read   = seq_read,
	.llseek = seq_lseek,
2480
	.release = seq_release_net,
2481 2482
};

2483 2484 2485 2486 2487 2488 2489 2490 2491 2492 2493 2494 2495 2496 2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510 2511 2512 2513 2514 2515 2516 2517 2518 2519 2520 2521 2522
struct fib_route_iter {
	struct seq_net_private p;
	struct trie *main_trie;
	loff_t	pos;
	t_key	key;
};

static struct leaf *fib_route_get_idx(struct fib_route_iter *iter, loff_t pos)
{
	struct leaf *l = NULL;
	struct trie *t = iter->main_trie;

	/* use cache location of last found key */
	if (iter->pos > 0 && pos >= iter->pos && (l = fib_find_node(t, iter->key)))
		pos -= iter->pos;
	else {
		iter->pos = 0;
		l = trie_firstleaf(t);
	}

	while (l && pos-- > 0) {
		iter->pos++;
		l = trie_nextleaf(l);
	}

	if (l)
		iter->key = pos;	/* remember it */
	else
		iter->pos = 0;		/* forget it */

	return l;
}

static void *fib_route_seq_start(struct seq_file *seq, loff_t *pos)
	__acquires(RCU)
{
	struct fib_route_iter *iter = seq->private;
	struct fib_table *tb;

	rcu_read_lock();
2523
	tb = fib_get_table(seq_file_net(seq), RT_TABLE_MAIN);
2524 2525 2526 2527 2528 2529 2530 2531 2532 2533 2534 2535 2536 2537 2538 2539 2540 2541 2542 2543 2544 2545 2546 2547 2548 2549 2550 2551 2552 2553 2554 2555 2556 2557 2558 2559 2560
	if (!tb)
		return NULL;

	iter->main_trie = (struct trie *) tb->tb_data;
	if (*pos == 0)
		return SEQ_START_TOKEN;
	else
		return fib_route_get_idx(iter, *pos - 1);
}

static void *fib_route_seq_next(struct seq_file *seq, void *v, loff_t *pos)
{
	struct fib_route_iter *iter = seq->private;
	struct leaf *l = v;

	++*pos;
	if (v == SEQ_START_TOKEN) {
		iter->pos = 0;
		l = trie_firstleaf(iter->main_trie);
	} else {
		iter->pos++;
		l = trie_nextleaf(l);
	}

	if (l)
		iter->key = l->key;
	else
		iter->pos = 0;
	return l;
}

static void fib_route_seq_stop(struct seq_file *seq, void *v)
	__releases(RCU)
{
	rcu_read_unlock();
}

A
Al Viro 已提交
2561
static unsigned fib_flag_trans(int type, __be32 mask, const struct fib_info *fi)
2562
{
2563 2564 2565 2566
	static unsigned type2flags[RTN_MAX + 1] = {
		[7] = RTF_REJECT, [8] = RTF_REJECT,
	};
	unsigned flags = type2flags[type];
2567

2568 2569
	if (fi && fi->fib_nh->nh_gw)
		flags |= RTF_GATEWAY;
A
Al Viro 已提交
2570
	if (mask == htonl(0xFFFFFFFF))
2571 2572 2573
		flags |= RTF_HOST;
	flags |= RTF_UP;
	return flags;
2574 2575
}

2576 2577 2578 2579 2580 2581 2582
/*
 *	This outputs /proc/net/route.
 *	The format of the file is not supposed to be changed
 * 	and needs to be same as fib_hash output to avoid breaking
 *	legacy utilities
 */
static int fib_route_seq_show(struct seq_file *seq, void *v)
2583
{
2584
	struct leaf *l = v;
2585 2586
	struct leaf_info *li;
	struct hlist_node *node;
2587

2588 2589 2590 2591 2592 2593
	if (v == SEQ_START_TOKEN) {
		seq_printf(seq, "%-127s\n", "Iface\tDestination\tGateway "
			   "\tFlags\tRefCnt\tUse\tMetric\tMask\t\tMTU"
			   "\tWindow\tIRTT");
		return 0;
	}
2594

2595
	hlist_for_each_entry_rcu(li, node, &l->list, hlist) {
2596
		struct fib_alias *fa;
A
Al Viro 已提交
2597
		__be32 mask, prefix;
O
Olof Johansson 已提交
2598

2599 2600
		mask = inet_make_mask(li->plen);
		prefix = htonl(l->key);
2601

2602
		list_for_each_entry_rcu(fa, &li->falh, fa_list) {
2603
			const struct fib_info *fi = fa->fa_info;
2604
			unsigned flags = fib_flag_trans(fa->fa_type, mask, fi);
2605
			char bf[128];
2606

2607 2608 2609
			if (fa->fa_type == RTN_BROADCAST
			    || fa->fa_type == RTN_MULTICAST)
				continue;
2610

2611 2612 2613 2614 2615 2616 2617 2618
			if (fi)
				snprintf(bf, sizeof(bf),
					 "%s\t%08X\t%08X\t%04X\t%d\t%u\t%d\t%08X\t%d\t%u\t%u",
					 fi->fib_dev ? fi->fib_dev->name : "*",
					 prefix,
					 fi->fib_nh->nh_gw, flags, 0, 0,
					 fi->fib_priority,
					 mask,
2619 2620
					 (fi->fib_advmss ?
					  fi->fib_advmss + 40 : 0),
2621 2622 2623 2624 2625 2626 2627
					 fi->fib_window,
					 fi->fib_rtt >> 3);
			else
				snprintf(bf, sizeof(bf),
					 "*\t%08X\t%08X\t%04X\t%d\t%u\t%d\t%08X\t%d\t%u\t%u",
					 prefix, 0, flags, 0, 0, 0,
					 mask, 0, 0, 0);
2628

2629 2630
			seq_printf(seq, "%-127s\n", bf);
		}
2631 2632 2633 2634 2635
	}

	return 0;
}

2636
static const struct seq_operations fib_route_seq_ops = {
2637 2638 2639
	.start  = fib_route_seq_start,
	.next   = fib_route_seq_next,
	.stop   = fib_route_seq_stop,
2640
	.show   = fib_route_seq_show,
2641 2642
};

2643
static int fib_route_seq_open(struct inode *inode, struct file *file)
2644
{
2645
	return seq_open_net(inode, file, &fib_route_seq_ops,
2646
			    sizeof(struct fib_route_iter));
2647 2648
}

2649
static const struct file_operations fib_route_fops = {
2650 2651 2652 2653
	.owner  = THIS_MODULE,
	.open   = fib_route_seq_open,
	.read   = seq_read,
	.llseek = seq_lseek,
2654
	.release = seq_release_net,
2655 2656
};

2657
int __net_init fib_proc_init(struct net *net)
2658
{
2659
	if (!proc_net_fops_create(net, "fib_trie", S_IRUGO, &fib_trie_fops))
2660 2661
		goto out1;

2662 2663
	if (!proc_net_fops_create(net, "fib_triestat", S_IRUGO,
				  &fib_triestat_fops))
2664 2665
		goto out2;

2666
	if (!proc_net_fops_create(net, "route", S_IRUGO, &fib_route_fops))
2667 2668
		goto out3;

2669
	return 0;
2670 2671

out3:
2672
	proc_net_remove(net, "fib_triestat");
2673
out2:
2674
	proc_net_remove(net, "fib_trie");
2675 2676
out1:
	return -ENOMEM;
2677 2678
}

2679
void __net_exit fib_proc_exit(struct net *net)
2680
{
2681 2682 2683
	proc_net_remove(net, "fib_trie");
	proc_net_remove(net, "fib_triestat");
	proc_net_remove(net, "route");
2684 2685 2686
}

#endif /* CONFIG_PROC_FS */