fib_trie.c 61.8 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
 * 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
 *
 *
 * 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
42 43 44 45 46 47 48
 *
 * 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>
49 50
 */

R
Robert Olsson 已提交
51
#define VERSION "0.408"
52 53 54

#include <asm/uaccess.h>
#include <asm/system.h>
J
Jiri Slaby 已提交
55
#include <linux/bitops.h>
56 57 58 59 60 61 62 63 64
#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
65
#include <linux/inetdevice.h>
66 67 68
#include <linux/netdevice.h>
#include <linux/if_arp.h>
#include <linux/proc_fs.h>
R
Robert Olsson 已提交
69
#include <linux/rcupdate.h>
70 71 72 73
#include <linux/skbuff.h>
#include <linux/netlink.h>
#include <linux/init.h>
#include <linux/list.h>
74
#include <net/net_namespace.h>
75 76 77 78 79 80 81 82
#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
83
#define MAX_STAT_DEPTH 32
84 85 86 87 88 89 90 91

#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
92 93
#define NODE_TYPE(node) ((node)->parent & NODE_TYPE_MASK)

O
Olof Johansson 已提交
94 95
#define IS_TNODE(n) (!(n->parent & T_LEAF))
#define IS_LEAF(n) (n->parent & T_LEAF)
96 97

struct node {
O
Olof Johansson 已提交
98
	unsigned long parent;
99
	t_key key;
100 101 102
};

struct leaf {
O
Olof Johansson 已提交
103
	unsigned long parent;
104
	t_key key;
105
	struct hlist_head list;
R
Robert Olsson 已提交
106
	struct rcu_head rcu;
107 108 109 110
};

struct leaf_info {
	struct hlist_node hlist;
R
Robert Olsson 已提交
111
	struct rcu_head rcu;
112 113 114 115 116
	int plen;
	struct list_head falh;
};

struct tnode {
O
Olof Johansson 已提交
117
	unsigned long parent;
118
	t_key key;
119 120
	unsigned char pos;		/* 2log(KEYLENGTH) bits needed */
	unsigned char bits;		/* 2log(KEYLENGTH) bits needed */
121 122
	unsigned int full_children;	/* KEYLENGTH bits needed */
	unsigned int empty_children;	/* KEYLENGTH bits needed */
123 124 125
	union {
		struct rcu_head rcu;
		struct work_struct work;
J
Jarek Poplawski 已提交
126
		struct tnode *tnode_free;
127
	};
O
Olof Johansson 已提交
128
	struct node *child[0];
129 130 131 132 133 134 135 136 137
};

#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;
138
	unsigned int resize_node_skipped;
139 140 141 142 143 144 145 146 147
};
#endif

struct trie_stat {
	unsigned int totdepth;
	unsigned int maxdepth;
	unsigned int tnodes;
	unsigned int leaves;
	unsigned int nullpointers;
148
	unsigned int prefixes;
R
Robert Olsson 已提交
149
	unsigned int nodesizes[MAX_STAT_DEPTH];
150
};
151 152

struct trie {
O
Olof Johansson 已提交
153
	struct node *trie;
154 155 156 157 158 159
#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);
160 161
static void tnode_put_child_reorg(struct tnode *tn, int i, struct node *n,
				  int wasfull);
162
static struct node *resize(struct trie *t, struct tnode *tn);
163 164
static struct tnode *inflate(struct trie *t, struct tnode *tn);
static struct tnode *halve(struct trie *t, struct tnode *tn);
J
Jarek Poplawski 已提交
165 166
/* tnodes to free after resize(); protected by RTNL */
static struct tnode *tnode_free_head;
167

168
static struct kmem_cache *fn_alias_kmem __read_mostly;
169
static struct kmem_cache *trie_leaf_kmem __read_mostly;
170

S
Stephen Hemminger 已提交
171 172
static inline struct tnode *node_parent(struct node *node)
{
173 174 175 176 177 178
	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
179 180 181 182

	return rcu_dereference(ret);
}

183 184 185
/* Same as rcu_assign_pointer
 * but that macro() assumes that value is a pointer.
 */
S
Stephen Hemminger 已提交
186 187
static inline void node_set_parent(struct node *node, struct tnode *ptr)
{
188 189
	smp_wmb();
	node->parent = (unsigned long)ptr | NODE_TYPE(node);
S
Stephen Hemminger 已提交
190
}
R
Robert Olsson 已提交
191

192 193 194
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
195

196 197 198 199
	return tn->child[i];
}

static inline struct node *tnode_get_child_rcu(struct tnode *tn, unsigned int i)
200
{
201
	struct node *ret = tnode_get_child(tn, i);
202

203
	return rcu_dereference(ret);
204 205
}

S
Stephen Hemmigner 已提交
206
static inline int tnode_child_length(const struct tnode *tn)
207
{
O
Olof Johansson 已提交
208
	return 1 << tn->bits;
209 210
}

S
Stephen Hemminger 已提交
211 212 213 214 215
static inline t_key mask_pfx(t_key k, unsigned short l)
{
	return (l == 0) ? 0 : k >> (KEYLENGTH-l) << (KEYLENGTH-l);
}

216 217
static inline t_key tkey_extract_bits(t_key a, int offset, int bits)
{
O
Olof Johansson 已提交
218
	if (offset < KEYLENGTH)
219
		return ((t_key)(a << offset)) >> (KEYLENGTH - bits);
O
Olof Johansson 已提交
220
	else
221 222 223 224 225
		return 0;
}

static inline int tkey_equals(t_key a, t_key b)
{
226
	return a == b;
227 228 229 230
}

static inline int tkey_sub_equals(t_key a, int offset, int bits, t_key b)
{
231 232
	if (bits == 0 || offset >= KEYLENGTH)
		return 1;
O
Olof Johansson 已提交
233 234
	bits = bits > KEYLENGTH ? KEYLENGTH : bits;
	return ((a ^ b) << offset) >> (KEYLENGTH - bits) == 0;
235
}
236 237 238 239 240 241

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

242 243 244
	if (!diff)
		return 0;
	while ((diff << i) >> (KEYLENGTH-1) == 0)
245 246 247 248 249
		i++;
	return i;
}

/*
250 251
  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
252 253 254 255
  all of the bits in that key are significant.

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

256 257 258 259 260
  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
261 262
  correct key path.

263 264 265 266 267
  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
268 269
  call to tkey_sub_equals() in trie_insert().

270
  if n is an internal node - a 'tnode' here, the various parts of its key
271 272
  have many different meanings.

273
  Example:
274 275 276
  _________________________________________________________________
  | i | i | i | i | i | i | i | N | N | N | S | S | S | S | S | C |
  -----------------------------------------------------------------
277
    0   1   2   3   4   5   6   7   8   9  10  11  12  13  14  15
278 279 280 281 282 283 284 285 286

  _________________________________________________________________
  | 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
287
  n->bits = 4
288

289 290
  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
291 292 293
  not use them for anything.

  The bits from (tp->pos) to (tp->pos + tp->bits - 1) - "N", above - are the
294
  index into the parent's child array. That is, they will be used to find
295 296 297 298 299
  'n' among tp's children.

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

300
  All the bits we have seen so far are significant to the node n. The rest
301 302
  of the bits are really not needed or indeed known in n->key.

303
  The bits from (n->pos) to (n->pos + n->bits - 1) - "C" - are the index into
304
  n's child array, and will of course be different for each child.
305

306

307 308 309 310 311
  The rest of the bits, from (n->pos + n->bits) onward, are completely unknown
  at this point.

*/

S
Stephen Hemminger 已提交
312
static inline void check_tnode(const struct tnode *tn)
313
{
S
Stephen Hemminger 已提交
314
	WARN_ON(tn && tn->pos+tn->bits > 32);
315 316
}

317 318 319 320
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;
321

R
Robert Olsson 已提交
322 323

static void __alias_free_mem(struct rcu_head *head)
324
{
R
Robert Olsson 已提交
325 326
	struct fib_alias *fa = container_of(head, struct fib_alias, rcu);
	kmem_cache_free(fn_alias_kmem, fa);
327 328
}

R
Robert Olsson 已提交
329
static inline void alias_free_mem_rcu(struct fib_alias *fa)
330
{
R
Robert Olsson 已提交
331 332
	call_rcu(&fa->rcu, __alias_free_mem);
}
O
Olof Johansson 已提交
333

R
Robert Olsson 已提交
334 335
static void __leaf_free_rcu(struct rcu_head *head)
{
336 337
	struct leaf *l = container_of(head, struct leaf, rcu);
	kmem_cache_free(trie_leaf_kmem, l);
R
Robert Olsson 已提交
338
}
O
Olof Johansson 已提交
339

340 341 342 343 344
static inline void free_leaf(struct leaf *l)
{
	call_rcu_bh(&l->rcu, __leaf_free_rcu);
}

R
Robert Olsson 已提交
345
static void __leaf_info_free_rcu(struct rcu_head *head)
346
{
R
Robert Olsson 已提交
347
	kfree(container_of(head, struct leaf_info, rcu));
348 349
}

R
Robert Olsson 已提交
350
static inline void free_leaf_info(struct leaf_info *leaf)
351
{
R
Robert Olsson 已提交
352
	call_rcu(&leaf->rcu, __leaf_info_free_rcu);
353 354
}

355
static struct tnode *tnode_alloc(size_t size)
356
{
R
Robert Olsson 已提交
357
	if (size <= PAGE_SIZE)
358
		return kzalloc(size, GFP_KERNEL);
359 360 361
	else
		return __vmalloc(size, GFP_KERNEL | __GFP_ZERO, PAGE_KERNEL);
}
R
Robert Olsson 已提交
362

363 364 365 366
static void __tnode_vfree(struct work_struct *arg)
{
	struct tnode *tn = container_of(arg, struct tnode, work);
	vfree(tn);
367 368
}

R
Robert Olsson 已提交
369
static void __tnode_free_rcu(struct rcu_head *head)
370
{
R
Robert Olsson 已提交
371
	struct tnode *tn = container_of(head, struct tnode, rcu);
372 373
	size_t size = sizeof(struct tnode) +
		      (sizeof(struct node *) << tn->bits);
374 375 376

	if (size <= PAGE_SIZE)
		kfree(tn);
377 378 379 380
	else {
		INIT_WORK(&tn->work, __tnode_vfree);
		schedule_work(&tn->work);
	}
381 382
}

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

J
Jarek Poplawski 已提交
391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413
static void tnode_free_safe(struct tnode *tn)
{
	BUG_ON(IS_LEAF(tn));

	if (node_parent((struct node *) tn)) {
		tn->tnode_free = tnode_free_head;
		tnode_free_head = tn;
	} else {
		tnode_free(tn);
	}
}

static void tnode_free_flush(void)
{
	struct tnode *tn;

	while ((tn = tnode_free_head)) {
		tnode_free_head = tn->tnode_free;
		tn->tnode_free = NULL;
		tnode_free(tn);
	}
}

R
Robert Olsson 已提交
414 415
static struct leaf *leaf_new(void)
{
416
	struct leaf *l = kmem_cache_alloc(trie_leaf_kmem, GFP_KERNEL);
R
Robert Olsson 已提交
417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433
	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;
}

434
static struct tnode *tnode_new(t_key key, int pos, int bits)
435
{
436
	size_t sz = sizeof(struct tnode) + (sizeof(struct node *) << bits);
437
	struct tnode *tn = tnode_alloc(sz);
438

O
Olof Johansson 已提交
439
	if (tn) {
R
Robert Olsson 已提交
440
		tn->parent = T_TNODE;
441 442 443 444 445 446
		tn->pos = pos;
		tn->bits = bits;
		tn->key = key;
		tn->full_children = 0;
		tn->empty_children = 1<<bits;
	}
447

448 449
	pr_debug("AT %p s=%u %lu\n", tn, (unsigned int) sizeof(struct tnode),
		 (unsigned long) (sizeof(struct node) << bits));
450 451 452 453 454 455 456 457
	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
458
static inline int tnode_full(const struct tnode *tn, const struct node *n)
459
{
460
	if (n == NULL || IS_LEAF(n))
461 462 463 464 465
		return 0;

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

466 467
static inline void put_child(struct trie *t, struct tnode *tn, int i,
			     struct node *n)
468 469 470 471
{
	tnode_put_child_reorg(tn, i, n, -1);
}

472
 /*
473 474 475 476
  * Add a child at position i overwriting the old value.
  * Update the value of full_children and empty_children.
  */

477 478
static void tnode_put_child_reorg(struct tnode *tn, int i, struct node *n,
				  int wasfull)
479
{
R
Robert Olsson 已提交
480
	struct node *chi = tn->child[i];
481 482
	int isfull;

S
Stephen Hemminger 已提交
483 484
	BUG_ON(i >= 1<<tn->bits);

485 486 487 488 489
	/* update emptyChildren */
	if (n == NULL && chi != NULL)
		tn->empty_children++;
	else if (n != NULL && chi == NULL)
		tn->empty_children--;
490

491
	/* update fullChildren */
O
Olof Johansson 已提交
492
	if (wasfull == -1)
493 494 495
		wasfull = tnode_full(tn, chi);

	isfull = tnode_full(tn, n);
496
	if (wasfull && !isfull)
497
		tn->full_children--;
498
	else if (!wasfull && isfull)
499
		tn->full_children++;
O
Olof Johansson 已提交
500

501
	if (n)
S
Stephen Hemminger 已提交
502
		node_set_parent(n, tn);
503

R
Robert Olsson 已提交
504
	rcu_assign_pointer(tn->child[i], n);
505 506
}

507
static struct node *resize(struct trie *t, struct tnode *tn)
508 509
{
	int i;
510
	int err = 0;
511
	struct tnode *old_tn;
512 513
	int inflate_threshold_use;
	int halve_threshold_use;
R
Robert Olsson 已提交
514
	int max_resize;
515

516
	if (!tn)
517 518
		return NULL;

S
Stephen Hemminger 已提交
519 520
	pr_debug("In tnode_resize %p inflate_threshold=%d threshold=%d\n",
		 tn, inflate_threshold, halve_threshold);
521 522 523

	/* No children */
	if (tn->empty_children == tnode_child_length(tn)) {
J
Jarek Poplawski 已提交
524
		tnode_free_safe(tn);
525 526 527 528 529
		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
530
			struct node *n;
531

O
Olof Johansson 已提交
532
			n = tn->child[i];
R
Robert Olsson 已提交
533
			if (!n)
O
Olof Johansson 已提交
534 535 536
				continue;

			/* compress one level */
S
Stephen Hemminger 已提交
537
			node_set_parent(n, NULL);
J
Jarek Poplawski 已提交
538
			tnode_free_safe(tn);
O
Olof Johansson 已提交
539
			return n;
540
		}
541
	/*
542 543 544 545 546
	 * Double as long as the resulting node has a number of
	 * nonempty nodes that are above the threshold.
	 */

	/*
547 548
	 * From "Implementing a dynamic compressed trie" by Stefan Nilsson of
	 * the Helsinki University of Technology and Matti Tikkanen of Nokia
549
	 * Telecommunications, page 6:
550
	 * "A node is doubled if the ratio of non-empty children to all
551 552
	 * children in the *doubled* node is at least 'high'."
	 *
553 554 555 556 557
	 * '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
558
	 * multiply the left-hand side by 50.
559 560 561 562
	 *
	 * 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"
563
	 * children, that is non-null tnodes with a skip value of 0.
564
	 * All of those will be doubled in the resulting inflated tnode, so
565
	 * we just count them one extra time here.
566
	 *
567
	 * A clearer way to write this would be:
568
	 *
569
	 * to_be_doubled = tn->full_children;
570
	 * not_to_be_doubled = tnode_child_length(tn) - tn->empty_children -
571 572 573 574
	 *     tn->full_children;
	 *
	 * new_child_length = tnode_child_length(tn) * 2;
	 *
575
	 * new_fill_factor = 100 * (not_to_be_doubled + 2*to_be_doubled) /
576 577
	 *      new_child_length;
	 * if (new_fill_factor >= inflate_threshold)
578 579 580
	 *
	 * ...and so on, tho it would mess up the while () loop.
	 *
581 582 583
	 * anyway,
	 * 100 * (not_to_be_doubled + 2*to_be_doubled) / new_child_length >=
	 *      inflate_threshold
584
	 *
585 586 587
	 * avoid a division:
	 * 100 * (not_to_be_doubled + 2*to_be_doubled) >=
	 *      inflate_threshold * new_child_length
588
	 *
589
	 * expand not_to_be_doubled and to_be_doubled, and shorten:
590
	 * 100 * (tnode_child_length(tn) - tn->empty_children +
O
Olof Johansson 已提交
591
	 *    tn->full_children) >= inflate_threshold * new_child_length
592
	 *
593
	 * expand new_child_length:
594
	 * 100 * (tnode_child_length(tn) - tn->empty_children +
O
Olof Johansson 已提交
595
	 *    tn->full_children) >=
596
	 *      inflate_threshold * tnode_child_length(tn) * 2
597
	 *
598
	 * shorten again:
599
	 * 50 * (tn->full_children + tnode_child_length(tn) -
O
Olof Johansson 已提交
600
	 *    tn->empty_children) >= inflate_threshold *
601
	 *    tnode_child_length(tn)
602
	 *
603 604 605
	 */

	check_tnode(tn);
606

607 608
	/* Keep root node larger  */

S
Stephen Hemminger 已提交
609
	if (!tn->parent)
610
		inflate_threshold_use = inflate_threshold_root;
611
	else
612 613
		inflate_threshold_use = inflate_threshold;

614
	err = 0;
R
Robert Olsson 已提交
615 616
	max_resize = 10;
	while ((tn->full_children > 0 &&  max_resize-- &&
617 618 619
		50 * (tn->full_children + tnode_child_length(tn)
		      - tn->empty_children)
		>= inflate_threshold_use * tnode_child_length(tn))) {
620

621 622
		old_tn = tn;
		tn = inflate(t, tn);
623

624 625
		if (IS_ERR(tn)) {
			tn = old_tn;
626 627 628 629 630
#ifdef CONFIG_IP_FIB_TRIE_STATS
			t->stats.resize_node_skipped++;
#endif
			break;
		}
631 632
	}

R
Robert Olsson 已提交
633 634
	if (max_resize < 0) {
		if (!tn->parent)
635 636 637
			pr_warning("Fix inflate_threshold_root."
				   " Now=%d size=%d bits\n",
				   inflate_threshold_root, tn->bits);
R
Robert Olsson 已提交
638
		else
639 640 641
			pr_warning("Fix inflate_threshold."
				   " Now=%d size=%d bits\n",
				   inflate_threshold, tn->bits);
R
Robert Olsson 已提交
642 643
	}

644 645 646 647 648 649
	check_tnode(tn);

	/*
	 * Halve as long as the number of empty children in this
	 * node is above threshold.
	 */
650

651 652 653

	/* Keep root node larger  */

S
Stephen Hemminger 已提交
654
	if (!tn->parent)
655
		halve_threshold_use = halve_threshold_root;
656
	else
657 658
		halve_threshold_use = halve_threshold;

659
	err = 0;
R
Robert Olsson 已提交
660 661
	max_resize = 10;
	while (tn->bits > 1 &&  max_resize-- &&
662
	       100 * (tnode_child_length(tn) - tn->empty_children) <
663
	       halve_threshold_use * tnode_child_length(tn)) {
664

665 666 667 668
		old_tn = tn;
		tn = halve(t, tn);
		if (IS_ERR(tn)) {
			tn = old_tn;
669 670 671 672 673 674
#ifdef CONFIG_IP_FIB_TRIE_STATS
			t->stats.resize_node_skipped++;
#endif
			break;
		}
	}
675

R
Robert Olsson 已提交
676 677
	if (max_resize < 0) {
		if (!tn->parent)
678 679 680
			pr_warning("Fix halve_threshold_root."
				   " Now=%d size=%d bits\n",
				   halve_threshold_root, tn->bits);
R
Robert Olsson 已提交
681
		else
682 683 684
			pr_warning("Fix halve_threshold."
				   " Now=%d size=%d bits\n",
				   halve_threshold, tn->bits);
R
Robert Olsson 已提交
685
	}
686

687 688 689
	/* 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
690
			struct node *n;
691

O
Olof Johansson 已提交
692
			n = tn->child[i];
R
Robert Olsson 已提交
693
			if (!n)
O
Olof Johansson 已提交
694 695 696 697
				continue;

			/* compress one level */

S
Stephen Hemminger 已提交
698
			node_set_parent(n, NULL);
J
Jarek Poplawski 已提交
699
			tnode_free_safe(tn);
O
Olof Johansson 已提交
700
			return n;
701 702 703 704 705
		}

	return (struct node *) tn;
}

706
static struct tnode *inflate(struct trie *t, struct tnode *tn)
707 708 709 710 711
{
	struct tnode *oldtnode = tn;
	int olen = tnode_child_length(tn);
	int i;

S
Stephen Hemminger 已提交
712
	pr_debug("In inflate\n");
713 714 715

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

S
Stephen Hemminger 已提交
716
	if (!tn)
717
		return ERR_PTR(-ENOMEM);
718 719

	/*
720 721 722
	 * 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
723 724
	 * of tnode is ignored.
	 */
O
Olof Johansson 已提交
725 726

	for (i = 0; i < olen; i++) {
727
		struct tnode *inode;
728

729
		inode = (struct tnode *) tnode_get_child(oldtnode, i);
730 731 732 733 734
		if (inode &&
		    IS_TNODE(inode) &&
		    inode->pos == oldtnode->pos + oldtnode->bits &&
		    inode->bits > 1) {
			struct tnode *left, *right;
S
Stephen Hemminger 已提交
735
			t_key m = ~0U << (KEYLENGTH - 1) >> inode->pos;
736

737 738
			left = tnode_new(inode->key&(~m), inode->pos + 1,
					 inode->bits - 1);
739 740
			if (!left)
				goto nomem;
O
Olof Johansson 已提交
741

742 743 744
			right = tnode_new(inode->key|m, inode->pos + 1,
					  inode->bits - 1);

745
			if (!right) {
746 747
				tnode_free(left);
				goto nomem;
748
			}
749 750 751 752 753 754

			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
755
	for (i = 0; i < olen; i++) {
756
		struct tnode *inode;
757
		struct node *node = tnode_get_child(oldtnode, i);
O
Olof Johansson 已提交
758 759
		struct tnode *left, *right;
		int size, j;
760

761 762 763 764 765 766
		/* An empty child */
		if (node == NULL)
			continue;

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

767
		if (IS_LEAF(node) || ((struct tnode *) node)->pos >
768
		   tn->pos + tn->bits - 1) {
769 770 771
			if (tkey_extract_bits(node->key,
					      oldtnode->pos + oldtnode->bits,
					      1) == 0)
772 773 774 775 776 777 778 779 780 781 782 783 784
				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]);

J
Jarek Poplawski 已提交
785
			tnode_free_safe(inode);
O
Olof Johansson 已提交
786
			continue;
787 788
		}

O
Olof Johansson 已提交
789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806
		/* 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)
		 */
807

O
Olof Johansson 已提交
808 809 810
		/* Use the old key, but set the new significant
		 *   bit to zero.
		 */
811

O
Olof Johansson 已提交
812 813
		left = (struct tnode *) tnode_get_child(tn, 2*i);
		put_child(t, tn, 2*i, NULL);
814

O
Olof Johansson 已提交
815
		BUG_ON(!left);
816

O
Olof Johansson 已提交
817 818
		right = (struct tnode *) tnode_get_child(tn, 2*i+1);
		put_child(t, tn, 2*i+1, NULL);
819

O
Olof Johansson 已提交
820
		BUG_ON(!right);
821

O
Olof Johansson 已提交
822 823 824 825
		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]);
826
		}
O
Olof Johansson 已提交
827 828 829
		put_child(t, tn, 2*i, resize(t, left));
		put_child(t, tn, 2*i+1, resize(t, right));

J
Jarek Poplawski 已提交
830
		tnode_free_safe(inode);
831
	}
J
Jarek Poplawski 已提交
832
	tnode_free_safe(oldtnode);
833
	return tn;
834 835 836 837 838
nomem:
	{
		int size = tnode_child_length(tn);
		int j;

S
Stephen Hemminger 已提交
839
		for (j = 0; j < size; j++)
840 841 842 843
			if (tn->child[j])
				tnode_free((struct tnode *)tn->child[j]);

		tnode_free(tn);
S
Stephen Hemminger 已提交
844

845 846
		return ERR_PTR(-ENOMEM);
	}
847 848
}

849
static struct tnode *halve(struct trie *t, struct tnode *tn)
850 851 852 853 854 855
{
	struct tnode *oldtnode = tn;
	struct node *left, *right;
	int i;
	int olen = tnode_child_length(tn);

S
Stephen Hemminger 已提交
856
	pr_debug("In halve\n");
857 858

	tn = tnode_new(oldtnode->key, oldtnode->pos, oldtnode->bits - 1);
859

860 861
	if (!tn)
		return ERR_PTR(-ENOMEM);
862 863

	/*
864 865 866
	 * 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
867 868 869
	 * of tnode is ignored.
	 */

O
Olof Johansson 已提交
870
	for (i = 0; i < olen; i += 2) {
871 872
		left = tnode_get_child(oldtnode, i);
		right = tnode_get_child(oldtnode, i+1);
873

874
		/* Two nonempty children */
S
Stephen Hemminger 已提交
875
		if (left && right) {
876
			struct tnode *newn;
S
Stephen Hemminger 已提交
877

878
			newn = tnode_new(left->key, tn->pos + tn->bits, 1);
S
Stephen Hemminger 已提交
879 880

			if (!newn)
881
				goto nomem;
S
Stephen Hemminger 已提交
882

883
			put_child(t, tn, i/2, (struct node *)newn);
884 885 886
		}

	}
887

O
Olof Johansson 已提交
888 889 890
	for (i = 0; i < olen; i += 2) {
		struct tnode *newBinNode;

891 892
		left = tnode_get_child(oldtnode, i);
		right = tnode_get_child(oldtnode, i+1);
893

894 895 896 897 898
		/* 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
899
			continue;
S
Stephen Hemminger 已提交
900
		}
O
Olof Johansson 已提交
901 902

		if (right == NULL) {
903
			put_child(t, tn, i/2, left);
O
Olof Johansson 已提交
904 905
			continue;
		}
906

907
		/* Two nonempty children */
O
Olof Johansson 已提交
908 909 910 911 912
		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));
913
	}
J
Jarek Poplawski 已提交
914
	tnode_free_safe(oldtnode);
915
	return tn;
916 917 918 919 920
nomem:
	{
		int size = tnode_child_length(tn);
		int j;

S
Stephen Hemminger 已提交
921
		for (j = 0; j < size; j++)
922 923 924 925
			if (tn->child[j])
				tnode_free((struct tnode *)tn->child[j]);

		tnode_free(tn);
S
Stephen Hemminger 已提交
926

927 928
		return ERR_PTR(-ENOMEM);
	}
929 930
}

R
Robert Olsson 已提交
931
/* readside must use rcu_read_lock currently dump routines
R
Robert Olsson 已提交
932 933
 via get_fa_head and dump */

R
Robert Olsson 已提交
934
static struct leaf_info *find_leaf_info(struct leaf *l, int plen)
935
{
R
Robert Olsson 已提交
936
	struct hlist_head *head = &l->list;
937 938 939
	struct hlist_node *node;
	struct leaf_info *li;

R
Robert Olsson 已提交
940
	hlist_for_each_entry_rcu(li, node, head, hlist)
941
		if (li->plen == plen)
942
			return li;
O
Olof Johansson 已提交
943

944 945 946
	return NULL;
}

947
static inline struct list_head *get_fa_head(struct leaf *l, int plen)
948
{
R
Robert Olsson 已提交
949
	struct leaf_info *li = find_leaf_info(l, plen);
950

O
Olof Johansson 已提交
951 952
	if (!li)
		return NULL;
953

O
Olof Johansson 已提交
954
	return &li->falh;
955 956 957 958
}

static void insert_leaf_info(struct hlist_head *head, struct leaf_info *new)
{
959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975
	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);
	}
976 977
}

R
Robert Olsson 已提交
978 979
/* rcu_read_lock needs to be hold by caller from readside */

980 981 982 983 984 985 986 987
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
988
	n = rcu_dereference(t->trie);
989 990 991

	while (n != NULL &&  NODE_TYPE(n) == T_TNODE) {
		tn = (struct tnode *) n;
O
Olof Johansson 已提交
992

993
		check_tnode(tn);
O
Olof Johansson 已提交
994

995
		if (tkey_sub_equals(tn->key, pos, tn->pos-pos, key)) {
O
Olof Johansson 已提交
996
			pos = tn->pos + tn->bits;
997 998 999 1000
			n = tnode_get_child_rcu(tn,
						tkey_extract_bits(key,
								  tn->pos,
								  tn->bits));
O
Olof Johansson 已提交
1001
		} else
1002 1003 1004 1005
			break;
	}
	/* Case we have found a leaf. Compare prefixes */

O
Olof Johansson 已提交
1006 1007 1008
	if (n != NULL && IS_LEAF(n) && tkey_equals(key, n->key))
		return (struct leaf *)n;

1009 1010 1011 1012 1013 1014
	return NULL;
}

static struct node *trie_rebalance(struct trie *t, struct tnode *tn)
{
	int wasfull;
R
Robert Olsson 已提交
1015
	t_key cindex, key;
S
Stephen Hemminger 已提交
1016
	struct tnode *tp;
1017

R
Robert Olsson 已提交
1018 1019
	key = tn->key;

S
Stephen Hemminger 已提交
1020
	while (tn != NULL && (tp = node_parent((struct node *)tn)) != NULL) {
1021 1022
		cindex = tkey_extract_bits(key, tp->pos, tp->bits);
		wasfull = tnode_full(tp, tnode_get_child(tp, cindex));
1023 1024 1025 1026
		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
1027

S
Stephen Hemminger 已提交
1028
		tp = node_parent((struct node *) tn);
J
Jarek Poplawski 已提交
1029
		tnode_free_flush();
S
Stephen Hemminger 已提交
1030
		if (!tp)
1031
			break;
S
Stephen Hemminger 已提交
1032
		tn = tp;
1033
	}
S
Stephen Hemminger 已提交
1034

1035
	/* Handle last (top) tnode */
J
Jarek Poplawski 已提交
1036
	if (IS_TNODE(tn)) {
1037
		tn = (struct tnode *)resize(t, (struct tnode *)tn);
J
Jarek Poplawski 已提交
1038 1039
		tnode_free_flush();
	}
1040

1041
	return (struct node *)tn;
1042 1043
}

R
Robert Olsson 已提交
1044 1045
/* only used from updater-side */

1046
static struct list_head *fib_insert_node(struct trie *t, u32 key, int plen)
1047 1048 1049 1050 1051 1052
{
	int pos, newpos;
	struct tnode *tp = NULL, *tn = NULL;
	struct node *n;
	struct leaf *l;
	int missbit;
1053
	struct list_head *fa_head = NULL;
1054 1055 1056 1057
	struct leaf_info *li;
	t_key cindex;

	pos = 0;
1058
	n = t->trie;
1059

1060 1061
	/* 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,
1062
	 * and we should just put our new leaf in that.
1063 1064
	 * 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
1065 1066
	 * not be the parent's 'pos'+'bits'!
	 *
1067
	 * If it does match the current key, get pos/bits from it, extract
1068 1069 1070 1071
	 * 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.
	 *
1072 1073 1074
	 * 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.
1075 1076 1077 1078 1079
	 * 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
1080

1081
		check_tnode(tn);
O
Olof Johansson 已提交
1082

1083
		if (tkey_sub_equals(tn->key, pos, tn->pos-pos, key)) {
1084
			tp = tn;
O
Olof Johansson 已提交
1085
			pos = tn->pos + tn->bits;
1086 1087 1088 1089
			n = tnode_get_child(tn,
					    tkey_extract_bits(key,
							      tn->pos,
							      tn->bits));
1090

S
Stephen Hemminger 已提交
1091
			BUG_ON(n && node_parent(n) != tn);
O
Olof Johansson 已提交
1092
		} else
1093 1094 1095 1096 1097 1098
			break;
	}

	/*
	 * n  ----> NULL, LEAF or TNODE
	 *
1099
	 * tp is n's (parent) ----> NULL or TNODE
1100 1101
	 */

O
Olof Johansson 已提交
1102
	BUG_ON(tp && IS_LEAF(tp));
1103 1104 1105

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

1106
	if (n != NULL && IS_LEAF(n) && tkey_equals(key, n->key)) {
1107
		l = (struct leaf *) n;
1108
		li = leaf_info_new(plen);
O
Olof Johansson 已提交
1109

1110 1111
		if (!li)
			return NULL;
1112 1113 1114 1115 1116 1117 1118

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

1119 1120
	if (!l)
		return NULL;
1121 1122 1123 1124

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

1125
	if (!li) {
1126
		free_leaf(l);
1127
		return NULL;
1128
	}
1129 1130 1131 1132 1133

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

	if (t->trie && n == NULL) {
O
Olof Johansson 已提交
1134
		/* Case 2: n is NULL, and will just insert a new leaf */
1135

S
Stephen Hemminger 已提交
1136
		node_set_parent((struct node *)l, tp);
1137

O
Olof Johansson 已提交
1138 1139 1140 1141
		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. */
1142 1143
		/*
		 *  Add a new tnode here
1144 1145 1146 1147
		 *  first tnode need some special handling
		 */

		if (tp)
O
Olof Johansson 已提交
1148
			pos = tp->pos+tp->bits;
1149
		else
O
Olof Johansson 已提交
1150 1151
			pos = 0;

1152
		if (n) {
1153 1154
			newpos = tkey_mismatch(key, pos, n->key);
			tn = tnode_new(n->key, newpos, 1);
O
Olof Johansson 已提交
1155
		} else {
1156
			newpos = 0;
1157
			tn = tnode_new(key, newpos, 1); /* First tnode */
1158 1159
		}

1160
		if (!tn) {
1161
			free_leaf_info(li);
1162
			free_leaf(l);
1163
			return NULL;
O
Olof Johansson 已提交
1164 1165
		}

S
Stephen Hemminger 已提交
1166
		node_set_parent((struct node *)tn, tp);
1167

O
Olof Johansson 已提交
1168
		missbit = tkey_extract_bits(key, newpos, 1);
1169 1170 1171
		put_child(t, tn, missbit, (struct node *)l);
		put_child(t, tn, 1-missbit, n);

1172
		if (tp) {
1173
			cindex = tkey_extract_bits(key, tp->pos, tp->bits);
1174 1175
			put_child(t, (struct tnode *)tp, cindex,
				  (struct node *)tn);
O
Olof Johansson 已提交
1176
		} else {
1177
			rcu_assign_pointer(t->trie, (struct node *)tn);
1178 1179 1180
			tp = tn;
		}
	}
O
Olof Johansson 已提交
1181 1182

	if (tp && tp->pos + tp->bits > 32)
1183 1184 1185
		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
1186

1187
	/* Rebalance the trie */
R
Robert Olsson 已提交
1188 1189

	rcu_assign_pointer(t->trie, trie_rebalance(t, tp));
1190
done:
1191 1192 1193
	return fa_head;
}

1194 1195 1196
/*
 * Caller must hold RTNL.
 */
1197
static int fn_trie_insert(struct fib_table *tb, struct fib_config *cfg)
1198 1199 1200
{
	struct trie *t = (struct trie *) tb->tb_data;
	struct fib_alias *fa, *new_fa;
1201
	struct list_head *fa_head = NULL;
1202
	struct fib_info *fi;
1203 1204
	int plen = cfg->fc_dst_len;
	u8 tos = cfg->fc_tos;
1205 1206 1207 1208 1209 1210 1211
	u32 key, mask;
	int err;
	struct leaf *l;

	if (plen > 32)
		return -EINVAL;

1212
	key = ntohl(cfg->fc_dst);
1213

1214
	pr_debug("Insert table=%u %08x/%d\n", tb->tb_id, key, plen);
1215

O
Olof Johansson 已提交
1216
	mask = ntohl(inet_make_mask(plen));
1217

1218
	if (key & ~mask)
1219 1220 1221 1222
		return -EINVAL;

	key = key & mask;

1223 1224 1225
	fi = fib_create_info(cfg);
	if (IS_ERR(fi)) {
		err = PTR_ERR(fi);
1226
		goto err;
1227
	}
1228 1229

	l = fib_find_node(t, key);
1230
	fa = NULL;
1231

1232
	if (l) {
1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247
		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.
	 */

1248 1249 1250
	if (fa && fa->fa_tos == tos &&
	    fa->fa_info->fib_priority == fi->fib_priority) {
		struct fib_alias *fa_first, *fa_match;
1251 1252

		err = -EEXIST;
1253
		if (cfg->fc_nlflags & NLM_F_EXCL)
1254 1255
			goto out;

1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276
		/* 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;
			}
		}

1277
		if (cfg->fc_nlflags & NLM_F_REPLACE) {
1278 1279 1280
			struct fib_info *fi_drop;
			u8 state;

1281 1282 1283 1284
			fa = fa_first;
			if (fa_match) {
				if (fa == fa_match)
					err = 0;
1285
				goto out;
1286
			}
R
Robert Olsson 已提交
1287
			err = -ENOBUFS;
1288
			new_fa = kmem_cache_alloc(fn_alias_kmem, GFP_KERNEL);
R
Robert Olsson 已提交
1289 1290
			if (new_fa == NULL)
				goto out;
1291 1292

			fi_drop = fa->fa_info;
R
Robert Olsson 已提交
1293 1294
			new_fa->fa_tos = fa->fa_tos;
			new_fa->fa_info = fi;
1295 1296
			new_fa->fa_type = cfg->fc_type;
			new_fa->fa_scope = cfg->fc_scope;
1297
			state = fa->fa_state;
1298
			new_fa->fa_state = state & ~FA_S_ACCESSED;
1299

R
Robert Olsson 已提交
1300 1301
			list_replace_rcu(&fa->fa_list, &new_fa->fa_list);
			alias_free_mem_rcu(fa);
1302 1303 1304

			fib_release_info(fi_drop);
			if (state & FA_S_ACCESSED)
1305
				rt_cache_flush(cfg->fc_nlinfo.nl_net, -1);
1306 1307
			rtmsg_fib(RTM_NEWROUTE, htonl(key), new_fa, plen,
				tb->tb_id, &cfg->fc_nlinfo, NLM_F_REPLACE);
1308

O
Olof Johansson 已提交
1309
			goto succeeded;
1310 1311 1312 1313 1314
		}
		/* Error if we find a perfect match which
		 * uses the same scope, type, and nexthop
		 * information.
		 */
1315 1316
		if (fa_match)
			goto out;
1317

1318
		if (!(cfg->fc_nlflags & NLM_F_APPEND))
1319
			fa = fa_first;
1320 1321
	}
	err = -ENOENT;
1322
	if (!(cfg->fc_nlflags & NLM_F_CREATE))
1323 1324 1325
		goto out;

	err = -ENOBUFS;
1326
	new_fa = kmem_cache_alloc(fn_alias_kmem, GFP_KERNEL);
1327 1328 1329 1330 1331
	if (new_fa == NULL)
		goto out;

	new_fa->fa_info = fi;
	new_fa->fa_tos = tos;
1332 1333
	new_fa->fa_type = cfg->fc_type;
	new_fa->fa_scope = cfg->fc_scope;
1334 1335 1336 1337 1338
	new_fa->fa_state = 0;
	/*
	 * Insert new entry to the list.
	 */

1339
	if (!fa_head) {
1340 1341 1342
		fa_head = fib_insert_node(t, key, plen);
		if (unlikely(!fa_head)) {
			err = -ENOMEM;
1343
			goto out_free_new_fa;
1344
		}
1345
	}
1346

R
Robert Olsson 已提交
1347 1348
	list_add_tail_rcu(&new_fa->fa_list,
			  (fa ? &fa->fa_list : fa_head));
1349

1350
	rt_cache_flush(cfg->fc_nlinfo.nl_net, -1);
1351
	rtmsg_fib(RTM_NEWROUTE, htonl(key), new_fa, plen, tb->tb_id,
1352
		  &cfg->fc_nlinfo, 0);
1353 1354
succeeded:
	return 0;
1355 1356 1357

out_free_new_fa:
	kmem_cache_free(fn_alias_kmem, new_fa);
1358 1359
out:
	fib_release_info(fi);
O
Olof Johansson 已提交
1360
err:
1361 1362 1363
	return err;
}

R
Robert Olsson 已提交
1364
/* should be called with rcu_read_lock */
1365 1366 1367
static int check_leaf(struct trie *t, struct leaf *l,
		      t_key key,  const struct flowi *flp,
		      struct fib_result *res)
1368 1369 1370 1371
{
	struct leaf_info *li;
	struct hlist_head *hhead = &l->list;
	struct hlist_node *node;
1372

R
Robert Olsson 已提交
1373
	hlist_for_each_entry_rcu(li, node, hhead, hlist) {
1374 1375 1376 1377
		int err;
		int plen = li->plen;
		__be32 mask = inet_make_mask(plen);

1378
		if (l->key != (key & ntohl(mask)))
1379 1380
			continue;

1381
		err = fib_semantic_match(&li->falh, flp, res, plen);
1382

1383
#ifdef CONFIG_IP_FIB_TRIE_STATS
1384
		if (err <= 0)
1385
			t->stats.semantic_match_passed++;
1386 1387
		else
			t->stats.semantic_match_miss++;
1388
#endif
1389
		if (err <= 0)
1390
			return err;
1391
	}
1392

1393
	return 1;
1394 1395
}

1396 1397
static int fn_trie_lookup(struct fib_table *tb, const struct flowi *flp,
			  struct fib_result *res)
1398 1399
{
	struct trie *t = (struct trie *) tb->tb_data;
1400
	int ret;
1401 1402 1403
	struct node *n;
	struct tnode *pn;
	int pos, bits;
O
Olof Johansson 已提交
1404
	t_key key = ntohl(flp->fl4_dst);
1405 1406 1407
	int chopped_off;
	t_key cindex = 0;
	int current_prefix_length = KEYLENGTH;
O
Olof Johansson 已提交
1408 1409 1410 1411
	struct tnode *cn;
	t_key node_prefix, key_prefix, pref_mismatch;
	int mp;

R
Robert Olsson 已提交
1412
	rcu_read_lock();
O
Olof Johansson 已提交
1413

R
Robert Olsson 已提交
1414
	n = rcu_dereference(t->trie);
1415
	if (!n)
1416 1417 1418 1419 1420 1421 1422 1423
		goto failed;

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

	/* Just a leaf? */
	if (IS_LEAF(n)) {
1424
		ret = check_leaf(t, (struct leaf *)n, key, flp, res);
1425
		goto found;
1426
	}
1427

1428 1429
	pn = (struct tnode *) n;
	chopped_off = 0;
1430

O
Olof Johansson 已提交
1431
	while (pn) {
1432 1433 1434
		pos = pn->pos;
		bits = pn->bits;

1435
		if (!chopped_off)
S
Stephen Hemminger 已提交
1436 1437
			cindex = tkey_extract_bits(mask_pfx(key, current_prefix_length),
						   pos, bits);
1438 1439 1440 1441 1442 1443 1444 1445 1446 1447

		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
1448
		if (IS_LEAF(n)) {
1449 1450
			ret = check_leaf(t, (struct leaf *)n, key, flp, res);
			if (ret > 0)
O
Olof Johansson 已提交
1451
				goto backtrace;
1452
			goto found;
O
Olof Johansson 已提交
1453 1454 1455
		}

		cn = (struct tnode *)n;
1456

O
Olof Johansson 已提交
1457 1458 1459 1460 1461 1462 1463 1464 1465 1466 1467 1468 1469 1470 1471
		/*
		 * 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].
		 */
1472

O
Olof Johansson 已提交
1473 1474 1475 1476 1477 1478 1479 1480 1481
		/* 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.
		 */
1482

1483 1484
		/* NOTA BENE: Checking only skipped bits
		   for the new node here */
1485

O
Olof Johansson 已提交
1486 1487
		if (current_prefix_length < pos+bits) {
			if (tkey_extract_bits(cn->key, current_prefix_length,
1488 1489
						cn->pos - current_prefix_length)
			    || !(cn->child[0]))
O
Olof Johansson 已提交
1490 1491
				goto backtrace;
		}
1492

O
Olof Johansson 已提交
1493 1494 1495 1496 1497 1498 1499 1500 1501
		/*
		 * 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.
		 */
1502

O
Olof Johansson 已提交
1503 1504 1505 1506 1507 1508 1509 1510
		/* 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.
		 */
1511

1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522
		/*
		 * 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
1523 1524
		 */

S
Stephen Hemminger 已提交
1525 1526
		node_prefix = mask_pfx(cn->key, cn->pos);
		key_prefix = mask_pfx(key, cn->pos);
O
Olof Johansson 已提交
1527 1528 1529
		pref_mismatch = key_prefix^node_prefix;
		mp = 0;

1530 1531 1532 1533
		/*
		 * 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
1534 1535 1536 1537
		 */
		if (pref_mismatch) {
			while (!(pref_mismatch & (1<<(KEYLENGTH-1)))) {
				mp++;
1538
				pref_mismatch = pref_mismatch << 1;
O
Olof Johansson 已提交
1539 1540 1541 1542 1543 1544 1545 1546
			}
			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;
1547
		}
1548

O
Olof Johansson 已提交
1549 1550 1551 1552
		pn = (struct tnode *)n; /* Descend */
		chopped_off = 0;
		continue;

1553 1554 1555 1556
backtrace:
		chopped_off++;

		/* As zero don't change the child key (cindex) */
1557 1558
		while ((chopped_off <= pn->bits)
		       && !(cindex & (1<<(chopped_off-1))))
1559 1560 1561 1562
			chopped_off++;

		/* Decrease current_... with bits chopped off */
		if (current_prefix_length > pn->pos + pn->bits - chopped_off)
1563 1564
			current_prefix_length = pn->pos + pn->bits
				- chopped_off;
O
Olof Johansson 已提交
1565

1566
		/*
1567
		 * Either we do the actual chop off according or if we have
1568 1569 1570
		 * chopped off all bits in this tnode walk up to our parent.
		 */

O
Olof Johansson 已提交
1571
		if (chopped_off <= pn->bits) {
1572
			cindex &= ~(1 << (chopped_off-1));
O
Olof Johansson 已提交
1573
		} else {
S
Stephen Hemminger 已提交
1574 1575
			struct tnode *parent = node_parent((struct node *) pn);
			if (!parent)
1576
				goto failed;
O
Olof Johansson 已提交
1577

1578
			/* Get Child's index */
S
Stephen Hemminger 已提交
1579 1580
			cindex = tkey_extract_bits(pn->key, parent->pos, parent->bits);
			pn = parent;
1581 1582 1583 1584 1585 1586
			chopped_off = 0;

#ifdef CONFIG_IP_FIB_TRIE_STATS
			t->stats.backtrack++;
#endif
			goto backtrace;
1587
		}
1588 1589
	}
failed:
1590
	ret = 1;
1591
found:
R
Robert Olsson 已提交
1592
	rcu_read_unlock();
1593 1594 1595
	return ret;
}

1596 1597 1598 1599
/*
 * Remove the leaf and return parent.
 */
static void trie_leaf_remove(struct trie *t, struct leaf *l)
1600
{
1601
	struct tnode *tp = node_parent((struct node *) l);
1602

1603
	pr_debug("entering trie_leaf_remove(%p)\n", l);
1604

1605
	if (tp) {
1606
		t_key cindex = tkey_extract_bits(l->key, tp->pos, tp->bits);
1607
		put_child(t, (struct tnode *)tp, cindex, NULL);
R
Robert Olsson 已提交
1608
		rcu_assign_pointer(t->trie, trie_rebalance(t, tp));
O
Olof Johansson 已提交
1609
	} else
R
Robert Olsson 已提交
1610
		rcu_assign_pointer(t->trie, NULL);
1611

1612
	free_leaf(l);
1613 1614
}

1615 1616 1617
/*
 * Caller must hold RTNL.
 */
1618
static int fn_trie_delete(struct fib_table *tb, struct fib_config *cfg)
1619 1620 1621
{
	struct trie *t = (struct trie *) tb->tb_data;
	u32 key, mask;
1622 1623
	int plen = cfg->fc_dst_len;
	u8 tos = cfg->fc_tos;
1624 1625 1626
	struct fib_alias *fa, *fa_to_delete;
	struct list_head *fa_head;
	struct leaf *l;
O
Olof Johansson 已提交
1627 1628
	struct leaf_info *li;

1629
	if (plen > 32)
1630 1631
		return -EINVAL;

1632
	key = ntohl(cfg->fc_dst);
O
Olof Johansson 已提交
1633
	mask = ntohl(inet_make_mask(plen));
1634

1635
	if (key & ~mask)
1636 1637 1638 1639 1640
		return -EINVAL;

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

1641
	if (!l)
1642 1643 1644 1645 1646 1647 1648 1649
		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
1650
	pr_debug("Deleting %08x/%d tos=%d t=%p\n", key, plen, tos, t);
1651 1652

	fa_to_delete = NULL;
1653 1654
	fa = list_entry(fa->fa_list.prev, struct fib_alias, fa_list);
	list_for_each_entry_continue(fa, fa_head, fa_list) {
1655 1656 1657 1658 1659
		struct fib_info *fi = fa->fa_info;

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

1660 1661 1662 1663 1664 1665
		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) {
1666 1667 1668 1669 1670
			fa_to_delete = fa;
			break;
		}
	}

O
Olof Johansson 已提交
1671 1672
	if (!fa_to_delete)
		return -ESRCH;
1673

O
Olof Johansson 已提交
1674
	fa = fa_to_delete;
1675
	rtmsg_fib(RTM_DELROUTE, htonl(key), fa, plen, tb->tb_id,
1676
		  &cfg->fc_nlinfo, 0);
O
Olof Johansson 已提交
1677 1678

	l = fib_find_node(t, key);
R
Robert Olsson 已提交
1679
	li = find_leaf_info(l, plen);
1680

R
Robert Olsson 已提交
1681
	list_del_rcu(&fa->fa_list);
1682

O
Olof Johansson 已提交
1683
	if (list_empty(fa_head)) {
R
Robert Olsson 已提交
1684
		hlist_del_rcu(&li->hlist);
O
Olof Johansson 已提交
1685
		free_leaf_info(li);
R
Robert Olsson 已提交
1686
	}
1687

O
Olof Johansson 已提交
1688
	if (hlist_empty(&l->list))
1689
		trie_leaf_remove(t, l);
1690

O
Olof Johansson 已提交
1691
	if (fa->fa_state & FA_S_ACCESSED)
1692
		rt_cache_flush(cfg->fc_nlinfo.nl_net, -1);
1693

R
Robert Olsson 已提交
1694 1695
	fib_release_info(fa->fa_info);
	alias_free_mem_rcu(fa);
O
Olof Johansson 已提交
1696
	return 0;
1697 1698
}

1699
static int trie_flush_list(struct list_head *head)
1700 1701 1702 1703 1704 1705 1706
{
	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
1707 1708 1709 1710
		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);
1711 1712 1713 1714 1715 1716
			found++;
		}
	}
	return found;
}

1717
static int trie_flush_leaf(struct leaf *l)
1718 1719 1720 1721 1722 1723 1724
{
	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) {
1725
		found += trie_flush_list(&li->falh);
1726 1727

		if (list_empty(&li->falh)) {
R
Robert Olsson 已提交
1728
			hlist_del_rcu(&li->hlist);
1729 1730 1731 1732 1733 1734
			free_leaf_info(li);
		}
	}
	return found;
}

1735 1736 1737 1738 1739
/*
 * 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)
1740
{
1741 1742
	do {
		t_key idx;
1743 1744

		if (c)
1745
			idx = tkey_extract_bits(c->key, p->pos, p->bits) + 1;
1746
		else
1747
			idx = 0;
R
Robert Olsson 已提交
1748

1749 1750
		while (idx < 1u << p->bits) {
			c = tnode_get_child_rcu(p, idx++);
R
Robert Olsson 已提交
1751
			if (!c)
O
Olof Johansson 已提交
1752 1753
				continue;

1754 1755 1756
			if (IS_LEAF(c)) {
				prefetch(p->child[idx]);
				return (struct leaf *) c;
1757
			}
1758 1759 1760 1761

			/* Rescan start scanning in new node */
			p = (struct tnode *) c;
			idx = 0;
1762
		}
1763 1764

		/* Node empty, walk back up to parent */
O
Olof Johansson 已提交
1765
		c = (struct node *) p;
1766 1767 1768 1769 1770 1771 1772 1773 1774 1775 1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 1790 1791 1792
	} 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);
1793 1794
}

1795 1796 1797 1798
static struct leaf *trie_leafindex(struct trie *t, int index)
{
	struct leaf *l = trie_firstleaf(t);

S
Stephen Hemminger 已提交
1799
	while (l && index-- > 0)
1800
		l = trie_nextleaf(l);
S
Stephen Hemminger 已提交
1801

1802 1803 1804 1805
	return l;
}


1806 1807 1808
/*
 * Caller must hold RTNL.
 */
1809 1810 1811
static int fn_trie_flush(struct fib_table *tb)
{
	struct trie *t = (struct trie *) tb->tb_data;
1812
	struct leaf *l, *ll = NULL;
1813
	int found = 0;
1814

1815
	for (l = trie_firstleaf(t); l; l = trie_nextleaf(l)) {
1816
		found += trie_flush_leaf(l);
1817 1818

		if (ll && hlist_empty(&ll->list))
1819
			trie_leaf_remove(t, ll);
1820 1821 1822 1823
		ll = l;
	}

	if (ll && hlist_empty(&ll->list))
1824
		trie_leaf_remove(t, ll);
1825

S
Stephen Hemminger 已提交
1826
	pr_debug("trie_flush found=%d\n", found);
1827 1828 1829
	return found;
}

1830 1831 1832
static void fn_trie_select_default(struct fib_table *tb,
				   const struct flowi *flp,
				   struct fib_result *res)
1833 1834 1835 1836 1837 1838 1839 1840 1841 1842 1843 1844 1845
{
	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
1846
	rcu_read_lock();
1847

1848
	l = fib_find_node(t, 0);
1849
	if (!l)
1850 1851 1852
		goto out;

	fa_head = get_fa_head(l, 0);
1853
	if (!fa_head)
1854 1855
		goto out;

1856
	if (list_empty(fa_head))
1857 1858
		goto out;

R
Robert Olsson 已提交
1859
	list_for_each_entry_rcu(fa, fa_head, fa_list) {
1860
		struct fib_info *next_fi = fa->fa_info;
O
Olof Johansson 已提交
1861

1862 1863 1864
		if (fa->fa_scope != res->scope ||
		    fa->fa_type != RTN_UNICAST)
			continue;
O
Olof Johansson 已提交
1865

1866 1867 1868 1869 1870 1871
		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
1872

1873 1874 1875 1876
		if (fi == NULL) {
			if (next_fi != res->fi)
				break;
		} else if (!fib_detect_death(fi, order, &last_resort,
1877
					     &last_idx, tb->tb_default)) {
1878
			fib_result_assign(res, fi);
1879
			tb->tb_default = order;
1880 1881 1882 1883 1884 1885
			goto out;
		}
		fi = next_fi;
		order++;
	}
	if (order <= 0 || fi == NULL) {
1886
		tb->tb_default = -1;
1887 1888 1889
		goto out;
	}

1890 1891
	if (!fib_detect_death(fi, order, &last_resort, &last_idx,
				tb->tb_default)) {
1892
		fib_result_assign(res, fi);
1893
		tb->tb_default = order;
1894 1895
		goto out;
	}
1896 1897
	if (last_idx >= 0)
		fib_result_assign(res, last_resort);
1898 1899
	tb->tb_default = last_idx;
out:
R
Robert Olsson 已提交
1900
	rcu_read_unlock();
1901 1902
}

1903 1904
static int fn_trie_dump_fa(t_key key, int plen, struct list_head *fah,
			   struct fib_table *tb,
1905 1906 1907 1908
			   struct sk_buff *skb, struct netlink_callback *cb)
{
	int i, s_i;
	struct fib_alias *fa;
A
Al Viro 已提交
1909
	__be32 xkey = htonl(key);
1910

1911
	s_i = cb->args[5];
1912 1913
	i = 0;

R
Robert Olsson 已提交
1914 1915 1916
	/* rcu_read_lock is hold by caller */

	list_for_each_entry_rcu(fa, fah, fa_list) {
1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927
		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,
1928
				  xkey,
1929 1930
				  plen,
				  fa->fa_tos,
1931
				  fa->fa_info, NLM_F_MULTI) < 0) {
1932
			cb->args[5] = i;
1933
			return -1;
O
Olof Johansson 已提交
1934
		}
1935 1936
		i++;
	}
1937
	cb->args[5] = i;
1938 1939 1940
	return skb->len;
}

1941 1942
static int fn_trie_dump_leaf(struct leaf *l, struct fib_table *tb,
			struct sk_buff *skb, struct netlink_callback *cb)
1943
{
1944 1945 1946
	struct leaf_info *li;
	struct hlist_node *node;
	int i, s_i;
1947

1948
	s_i = cb->args[4];
1949
	i = 0;
1950

1951 1952 1953 1954
	/* rcu_read_lock is hold by caller */
	hlist_for_each_entry_rcu(li, node, &l->list, hlist) {
		if (i < s_i) {
			i++;
1955
			continue;
1956
		}
O
Olof Johansson 已提交
1957

1958
		if (i > s_i)
1959
			cb->args[5] = 0;
1960

1961
		if (list_empty(&li->falh))
1962 1963
			continue;

1964
		if (fn_trie_dump_fa(l->key, li->plen, &li->falh, tb, skb, cb) < 0) {
1965
			cb->args[4] = i;
1966 1967
			return -1;
		}
1968
		i++;
1969
	}
1970

1971
	cb->args[4] = i;
1972 1973 1974
	return skb->len;
}

1975 1976
static int fn_trie_dump(struct fib_table *tb, struct sk_buff *skb,
			struct netlink_callback *cb)
1977
{
1978
	struct leaf *l;
1979
	struct trie *t = (struct trie *) tb->tb_data;
1980
	t_key key = cb->args[2];
1981
	int count = cb->args[3];
1982

R
Robert Olsson 已提交
1983
	rcu_read_lock();
1984 1985 1986
	/* Dump starting at last key.
	 * Note: 0.0.0.0/0 (ie default) is first key.
	 */
1987
	if (count == 0)
1988 1989
		l = trie_firstleaf(t);
	else {
1990 1991 1992
		/* Normally, continue from last key, but if that is missing
		 * fallback to using slow rescan
		 */
1993
		l = fib_find_node(t, key);
1994 1995
		if (!l)
			l = trie_leafindex(t, count);
1996
	}
1997

1998 1999
	while (l) {
		cb->args[2] = l->key;
2000
		if (fn_trie_dump_leaf(l, tb, skb, cb) < 0) {
2001
			cb->args[3] = count;
2002 2003
			rcu_read_unlock();
			return -1;
2004
		}
2005

2006
		++count;
2007
		l = trie_nextleaf(l);
2008 2009
		memset(&cb->args[4], 0,
		       sizeof(cb->args) - 4*sizeof(cb->args[0]));
2010
	}
2011
	cb->args[3] = count;
R
Robert Olsson 已提交
2012
	rcu_read_unlock();
2013

2014 2015 2016
	return skb->len;
}

2017 2018
void __init fib_hash_init(void)
{
2019 2020
	fn_alias_kmem = kmem_cache_create("ip_fib_alias",
					  sizeof(struct fib_alias),
2021 2022 2023 2024 2025 2026
					  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);
2027
}
2028

2029 2030 2031

/* Fix more generic FIB names for init later */
struct fib_table *fib_hash_table(u32 id)
2032 2033 2034 2035 2036 2037 2038 2039 2040 2041
{
	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;
2042
	tb->tb_default = -1;
2043 2044 2045 2046 2047 2048 2049 2050
	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;
2051
	memset(t, 0, sizeof(*t));
2052 2053

	if (id == RT_TABLE_LOCAL)
2054
		pr_info("IPv4 FIB: Using LC-trie version %s\n", VERSION);
2055 2056 2057 2058

	return tb;
}

2059 2060 2061
#ifdef CONFIG_PROC_FS
/* Depth first Trie walk iterator */
struct fib_trie_iter {
2062
	struct seq_net_private p;
2063
	struct fib_table *tb;
2064 2065 2066 2067
	struct tnode *tnode;
	unsigned index;
	unsigned depth;
};
2068

2069
static struct node *fib_trie_get_next(struct fib_trie_iter *iter)
2070
{
2071 2072 2073
	struct tnode *tn = iter->tnode;
	unsigned cindex = iter->index;
	struct tnode *p;
2074

2075 2076 2077 2078
	/* A single entry routing table */
	if (!tn)
		return NULL;

2079 2080 2081 2082
	pr_debug("get_next iter={node=%p index=%d depth=%d}\n",
		 iter->tnode, iter->index, iter->depth);
rescan:
	while (cindex < (1<<tn->bits)) {
2083
		struct node *n = tnode_get_child_rcu(tn, cindex);
2084

2085 2086 2087 2088 2089 2090 2091 2092 2093 2094 2095 2096
		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;
		}
2097

2098 2099
		++cindex;
	}
O
Olof Johansson 已提交
2100

2101
	/* Current node exhausted, pop back up */
2102
	p = node_parent_rcu((struct node *)tn);
2103 2104 2105 2106 2107
	if (p) {
		cindex = tkey_extract_bits(tn->key, p->pos, p->bits)+1;
		tn = p;
		--iter->depth;
		goto rescan;
2108
	}
2109 2110 2111

	/* got root? */
	return NULL;
2112 2113
}

2114 2115
static struct node *fib_trie_get_first(struct fib_trie_iter *iter,
				       struct trie *t)
2116
{
2117
	struct node *n;
2118

S
Stephen Hemminger 已提交
2119
	if (!t)
2120 2121 2122
		return NULL;

	n = rcu_dereference(t->trie);
2123
	if (!n)
2124
		return NULL;
2125

2126 2127 2128 2129 2130 2131 2132 2133
	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
2134
	}
2135 2136

	return n;
2137
}
O
Olof Johansson 已提交
2138

2139 2140 2141 2142
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
2143

2144
	memset(s, 0, sizeof(*s));
O
Olof Johansson 已提交
2145

2146
	rcu_read_lock();
2147
	for (n = fib_trie_get_first(&iter, t); n; n = fib_trie_get_next(&iter)) {
2148
		if (IS_LEAF(n)) {
2149 2150 2151 2152
			struct leaf *l = (struct leaf *)n;
			struct leaf_info *li;
			struct hlist_node *tmp;

2153 2154 2155 2156
			s->leaves++;
			s->totdepth += iter.depth;
			if (iter.depth > s->maxdepth)
				s->maxdepth = iter.depth;
2157 2158 2159

			hlist_for_each_entry_rcu(li, tmp, &l->list, hlist)
				++s->prefixes;
2160 2161 2162 2163 2164
		} else {
			const struct tnode *tn = (const struct tnode *) n;
			int i;

			s->tnodes++;
S
Stephen Hemminger 已提交
2165
			if (tn->bits < MAX_STAT_DEPTH)
R
Robert Olsson 已提交
2166 2167
				s->nodesizes[tn->bits]++;

2168 2169 2170
			for (i = 0; i < (1<<tn->bits); i++)
				if (!tn->child[i])
					s->nullpointers++;
2171 2172
		}
	}
R
Robert Olsson 已提交
2173
	rcu_read_unlock();
2174 2175
}

2176 2177 2178 2179
/*
 *	This outputs /proc/net/fib_triestats
 */
static void trie_show_stats(struct seq_file *seq, struct trie_stat *stat)
2180
{
2181
	unsigned i, max, pointers, bytes, avdepth;
2182

2183 2184 2185 2186
	if (stat->leaves)
		avdepth = stat->totdepth*100 / stat->leaves;
	else
		avdepth = 0;
O
Olof Johansson 已提交
2187

2188 2189
	seq_printf(seq, "\tAver depth:     %u.%02d\n",
		   avdepth / 100, avdepth % 100);
2190
	seq_printf(seq, "\tMax depth:      %u\n", stat->maxdepth);
O
Olof Johansson 已提交
2191

2192 2193
	seq_printf(seq, "\tLeaves:         %u\n", stat->leaves);
	bytes = sizeof(struct leaf) * stat->leaves;
2194 2195 2196 2197

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

2198
	seq_printf(seq, "\tInternal nodes: %u\n\t", stat->tnodes);
2199
	bytes += sizeof(struct tnode) * stat->tnodes;
2200

R
Robert Olsson 已提交
2201 2202
	max = MAX_STAT_DEPTH;
	while (max > 0 && stat->nodesizes[max-1] == 0)
2203
		max--;
2204

2205 2206 2207
	pointers = 0;
	for (i = 1; i <= max; i++)
		if (stat->nodesizes[i] != 0) {
2208
			seq_printf(seq, "  %u: %u",  i, stat->nodesizes[i]);
2209 2210 2211
			pointers += (1<<i) * stat->nodesizes[i];
		}
	seq_putc(seq, '\n');
2212
	seq_printf(seq, "\tPointers: %u\n", pointers);
R
Robert Olsson 已提交
2213

2214
	bytes += sizeof(struct node *) * pointers;
2215 2216
	seq_printf(seq, "Null ptrs: %u\n", stat->nullpointers);
	seq_printf(seq, "Total size: %u  kB\n", (bytes + 1023) / 1024);
2217
}
R
Robert Olsson 已提交
2218

2219
#ifdef CONFIG_IP_FIB_TRIE_STATS
2220 2221 2222 2223
static void trie_show_usage(struct seq_file *seq,
			    const struct trie_use_stats *stats)
{
	seq_printf(seq, "\nCounters:\n---------\n");
2224 2225 2226 2227 2228 2229 2230 2231 2232
	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);
2233
}
2234 2235
#endif /*  CONFIG_IP_FIB_TRIE_STATS */

2236
static void fib_table_print(struct seq_file *seq, struct fib_table *tb)
2237
{
2238 2239 2240 2241 2242 2243
	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);
2244
}
2245

2246

2247 2248
static int fib_triestat_seq_show(struct seq_file *seq, void *v)
{
2249
	struct net *net = (struct net *)seq->private;
2250
	unsigned int h;
2251

2252
	seq_printf(seq,
2253 2254
		   "Basic info: size of leaf:"
		   " %Zd bytes, size of tnode: %Zd bytes.\n",
2255 2256
		   sizeof(struct leaf), sizeof(struct tnode));

2257 2258 2259 2260 2261 2262 2263 2264
	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;
2265

2266 2267 2268 2269 2270 2271 2272 2273 2274 2275 2276 2277
			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
		}
	}
2278

2279
	return 0;
2280 2281
}

2282
static int fib_triestat_seq_open(struct inode *inode, struct file *file)
2283
{
2284
	return single_open_net(inode, file, fib_triestat_seq_show);
2285 2286
}

2287
static const struct file_operations fib_triestat_fops = {
2288 2289 2290 2291
	.owner	= THIS_MODULE,
	.open	= fib_triestat_seq_open,
	.read	= seq_read,
	.llseek	= seq_lseek,
2292
	.release = single_release_net,
2293 2294
};

2295
static struct node *fib_trie_get_idx(struct seq_file *seq, loff_t pos)
2296
{
2297 2298
	struct fib_trie_iter *iter = seq->private;
	struct net *net = seq_file_net(seq);
2299
	loff_t idx = 0;
2300
	unsigned int h;
2301

2302 2303 2304 2305
	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;
2306

2307 2308 2309 2310 2311 2312 2313 2314 2315 2316 2317
		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;
				}
		}
2318
	}
2319

2320 2321 2322
	return NULL;
}

2323
static void *fib_trie_seq_start(struct seq_file *seq, loff_t *pos)
2324
	__acquires(RCU)
2325
{
2326
	rcu_read_lock();
2327
	return fib_trie_get_idx(seq, *pos);
2328 2329
}

2330
static void *fib_trie_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2331
{
2332
	struct fib_trie_iter *iter = seq->private;
2333
	struct net *net = seq_file_net(seq);
2334 2335 2336 2337
	struct fib_table *tb = iter->tb;
	struct hlist_node *tb_node;
	unsigned int h;
	struct node *n;
2338

2339
	++*pos;
2340 2341 2342 2343
	/* next node in same table */
	n = fib_trie_get_next(iter);
	if (n)
		return n;
2344

2345 2346 2347 2348 2349 2350 2351 2352
	/* 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;
	}
2353

2354 2355 2356 2357 2358 2359 2360 2361 2362
	/* 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;
		}
	}
2363
	return NULL;
2364 2365 2366 2367

found:
	iter->tb = tb;
	return n;
2368
}
2369

2370
static void fib_trie_seq_stop(struct seq_file *seq, void *v)
2371
	__releases(RCU)
2372
{
2373 2374
	rcu_read_unlock();
}
O
Olof Johansson 已提交
2375

2376 2377 2378 2379
static void seq_indent(struct seq_file *seq, int n)
{
	while (n-- > 0) seq_puts(seq, "   ");
}
2380

2381
static inline const char *rtn_scope(char *buf, size_t len, enum rt_scope_t s)
2382
{
S
Stephen Hemminger 已提交
2383
	switch (s) {
2384 2385 2386 2387 2388 2389
	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:
2390
		snprintf(buf, len, "scope=%d", s);
2391 2392 2393
		return buf;
	}
}
2394

2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405 2406 2407 2408
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",
};
2409

2410
static inline const char *rtn_type(char *buf, size_t len, unsigned t)
2411 2412 2413
{
	if (t < __RTN_MAX && rtn_type_names[t])
		return rtn_type_names[t];
2414
	snprintf(buf, len, "type %u", t);
2415
	return buf;
2416 2417
}

2418 2419
/* Pretty print the trie */
static int fib_trie_seq_show(struct seq_file *seq, void *v)
2420
{
2421 2422
	const struct fib_trie_iter *iter = seq->private;
	struct node *n = v;
2423

2424 2425
	if (!node_parent_rcu(n))
		fib_table_print(seq, iter->tb);
2426

2427 2428
	if (IS_TNODE(n)) {
		struct tnode *tn = (struct tnode *) n;
S
Stephen Hemminger 已提交
2429
		__be32 prf = htonl(mask_pfx(tn->key, tn->pos));
O
Olof Johansson 已提交
2430

2431
		seq_indent(seq, iter->depth-1);
2432 2433
		seq_printf(seq, "  +-- %pI4/%d %d %d %d\n",
			   &prf, tn->pos, tn->bits, tn->full_children,
2434
			   tn->empty_children);
2435

2436 2437
	} else {
		struct leaf *l = (struct leaf *) n;
2438 2439
		struct leaf_info *li;
		struct hlist_node *node;
A
Al Viro 已提交
2440
		__be32 val = htonl(l->key);
2441 2442

		seq_indent(seq, iter->depth);
2443
		seq_printf(seq, "  |-- %pI4\n", &val);
2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457

		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)
2458
					seq_printf(seq, " tos=%d", fa->fa_tos);
2459
				seq_putc(seq, '\n');
2460 2461
			}
		}
2462
	}
2463

2464 2465 2466
	return 0;
}

2467
static const struct seq_operations fib_trie_seq_ops = {
2468 2469 2470 2471
	.start  = fib_trie_seq_start,
	.next   = fib_trie_seq_next,
	.stop   = fib_trie_seq_stop,
	.show   = fib_trie_seq_show,
2472 2473
};

2474
static int fib_trie_seq_open(struct inode *inode, struct file *file)
2475
{
2476 2477
	return seq_open_net(inode, file, &fib_trie_seq_ops,
			    sizeof(struct fib_trie_iter));
2478 2479
}

2480
static const struct file_operations fib_trie_fops = {
2481 2482 2483 2484
	.owner  = THIS_MODULE,
	.open   = fib_trie_seq_open,
	.read   = seq_read,
	.llseek = seq_lseek,
2485
	.release = seq_release_net,
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 2523 2524 2525 2526 2527
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();
2528
	tb = fib_get_table(seq_file_net(seq), RT_TABLE_MAIN);
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 2561 2562 2563 2564 2565
	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
2566
static unsigned fib_flag_trans(int type, __be32 mask, const struct fib_info *fi)
2567
{
2568 2569 2570 2571
	static unsigned type2flags[RTN_MAX + 1] = {
		[7] = RTF_REJECT, [8] = RTF_REJECT,
	};
	unsigned flags = type2flags[type];
2572

2573 2574
	if (fi && fi->fib_nh->nh_gw)
		flags |= RTF_GATEWAY;
A
Al Viro 已提交
2575
	if (mask == htonl(0xFFFFFFFF))
2576 2577 2578
		flags |= RTF_HOST;
	flags |= RTF_UP;
	return flags;
2579 2580
}

2581 2582 2583 2584 2585 2586 2587
/*
 *	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)
2588
{
2589
	struct leaf *l = v;
2590 2591
	struct leaf_info *li;
	struct hlist_node *node;
2592

2593 2594 2595 2596 2597 2598
	if (v == SEQ_START_TOKEN) {
		seq_printf(seq, "%-127s\n", "Iface\tDestination\tGateway "
			   "\tFlags\tRefCnt\tUse\tMetric\tMask\t\tMTU"
			   "\tWindow\tIRTT");
		return 0;
	}
2599

2600
	hlist_for_each_entry_rcu(li, node, &l->list, hlist) {
2601
		struct fib_alias *fa;
A
Al Viro 已提交
2602
		__be32 mask, prefix;
O
Olof Johansson 已提交
2603

2604 2605
		mask = inet_make_mask(li->plen);
		prefix = htonl(l->key);
2606

2607
		list_for_each_entry_rcu(fa, &li->falh, fa_list) {
2608
			const struct fib_info *fi = fa->fa_info;
2609
			unsigned flags = fib_flag_trans(fa->fa_type, mask, fi);
2610
			int len;
2611

2612 2613 2614
			if (fa->fa_type == RTN_BROADCAST
			    || fa->fa_type == RTN_MULTICAST)
				continue;
2615

2616
			if (fi)
2617 2618 2619
				seq_printf(seq,
					 "%s\t%08X\t%08X\t%04X\t%d\t%u\t"
					 "%d\t%08X\t%d\t%u\t%u%n",
2620 2621 2622 2623 2624
					 fi->fib_dev ? fi->fib_dev->name : "*",
					 prefix,
					 fi->fib_nh->nh_gw, flags, 0, 0,
					 fi->fib_priority,
					 mask,
2625 2626
					 (fi->fib_advmss ?
					  fi->fib_advmss + 40 : 0),
2627
					 fi->fib_window,
2628
					 fi->fib_rtt >> 3, &len);
2629
			else
2630 2631 2632
				seq_printf(seq,
					 "*\t%08X\t%08X\t%04X\t%d\t%u\t"
					 "%d\t%08X\t%d\t%u\t%u%n",
2633
					 prefix, 0, flags, 0, 0, 0,
2634
					 mask, 0, 0, 0, &len);
2635

2636
			seq_printf(seq, "%*s\n", 127 - len, "");
2637
		}
2638 2639 2640 2641 2642
	}

	return 0;
}

2643
static const struct seq_operations fib_route_seq_ops = {
2644 2645 2646
	.start  = fib_route_seq_start,
	.next   = fib_route_seq_next,
	.stop   = fib_route_seq_stop,
2647
	.show   = fib_route_seq_show,
2648 2649
};

2650
static int fib_route_seq_open(struct inode *inode, struct file *file)
2651
{
2652
	return seq_open_net(inode, file, &fib_route_seq_ops,
2653
			    sizeof(struct fib_route_iter));
2654 2655
}

2656
static const struct file_operations fib_route_fops = {
2657 2658 2659 2660
	.owner  = THIS_MODULE,
	.open   = fib_route_seq_open,
	.read   = seq_read,
	.llseek = seq_lseek,
2661
	.release = seq_release_net,
2662 2663
};

2664
int __net_init fib_proc_init(struct net *net)
2665
{
2666
	if (!proc_net_fops_create(net, "fib_trie", S_IRUGO, &fib_trie_fops))
2667 2668
		goto out1;

2669 2670
	if (!proc_net_fops_create(net, "fib_triestat", S_IRUGO,
				  &fib_triestat_fops))
2671 2672
		goto out2;

2673
	if (!proc_net_fops_create(net, "route", S_IRUGO, &fib_route_fops))
2674 2675
		goto out3;

2676
	return 0;
2677 2678

out3:
2679
	proc_net_remove(net, "fib_triestat");
2680
out2:
2681
	proc_net_remove(net, "fib_trie");
2682 2683
out1:
	return -ENOMEM;
2684 2685
}

2686
void __net_exit fib_proc_exit(struct net *net)
2687
{
2688 2689 2690
	proc_net_remove(net, "fib_trie");
	proc_net_remove(net, "fib_triestat");
	proc_net_remove(net, "route");
2691 2692 2693
}

#endif /* CONFIG_PROC_FS */