]> git.proxmox.com Git - mirror_ubuntu-eoan-kernel.git/blob - net/ipv4/fib_trie.c
Merge tag 'clk-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/clk/linux
[mirror_ubuntu-eoan-kernel.git] / net / ipv4 / fib_trie.c
1 /*
2 * This program is free software; you can redistribute it and/or
3 * modify it under the terms of the GNU General Public License
4 * as published by the Free Software Foundation; either version
5 * 2 of the License, or (at your option) any later version.
6 *
7 * Robert Olsson <robert.olsson@its.uu.se> Uppsala Universitet
8 * & Swedish University of Agricultural Sciences.
9 *
10 * Jens Laas <jens.laas@data.slu.se> Swedish University of
11 * Agricultural Sciences.
12 *
13 * Hans Liss <hans.liss@its.uu.se> Uppsala Universitet
14 *
15 * This work is based on the LPC-trie which is originally described in:
16 *
17 * An experimental study of compression methods for dynamic tries
18 * Stefan Nilsson and Matti Tikkanen. Algorithmica, 33(1):19-33, 2002.
19 * http://www.csc.kth.se/~snilsson/software/dyntrie2/
20 *
21 *
22 * IP-address lookup using LC-tries. Stefan Nilsson and Gunnar Karlsson
23 * IEEE Journal on Selected Areas in Communications, 17(6):1083-1092, June 1999
24 *
25 *
26 * Code from fib_hash has been reused which includes the following header:
27 *
28 *
29 * INET An implementation of the TCP/IP protocol suite for the LINUX
30 * operating system. INET is implemented using the BSD Socket
31 * interface as the means of communication with the user level.
32 *
33 * IPv4 FIB: lookup engine and maintenance routines.
34 *
35 *
36 * Authors: Alexey Kuznetsov, <kuznet@ms2.inr.ac.ru>
37 *
38 * This program is free software; you can redistribute it and/or
39 * modify it under the terms of the GNU General Public License
40 * as published by the Free Software Foundation; either version
41 * 2 of the License, or (at your option) any later version.
42 *
43 * Substantial contributions to this work comes from:
44 *
45 * David S. Miller, <davem@davemloft.net>
46 * Stephen Hemminger <shemminger@osdl.org>
47 * Paul E. McKenney <paulmck@us.ibm.com>
48 * Patrick McHardy <kaber@trash.net>
49 */
50
51 #define VERSION "0.409"
52
53 #include <linux/cache.h>
54 #include <linux/uaccess.h>
55 #include <linux/bitops.h>
56 #include <linux/types.h>
57 #include <linux/kernel.h>
58 #include <linux/mm.h>
59 #include <linux/string.h>
60 #include <linux/socket.h>
61 #include <linux/sockios.h>
62 #include <linux/errno.h>
63 #include <linux/in.h>
64 #include <linux/inet.h>
65 #include <linux/inetdevice.h>
66 #include <linux/netdevice.h>
67 #include <linux/if_arp.h>
68 #include <linux/proc_fs.h>
69 #include <linux/rcupdate.h>
70 #include <linux/skbuff.h>
71 #include <linux/netlink.h>
72 #include <linux/init.h>
73 #include <linux/list.h>
74 #include <linux/slab.h>
75 #include <linux/export.h>
76 #include <linux/vmalloc.h>
77 #include <linux/notifier.h>
78 #include <net/net_namespace.h>
79 #include <net/ip.h>
80 #include <net/protocol.h>
81 #include <net/route.h>
82 #include <net/tcp.h>
83 #include <net/sock.h>
84 #include <net/ip_fib.h>
85 #include <net/fib_notifier.h>
86 #include <trace/events/fib.h>
87 #include "fib_lookup.h"
88
89 static int call_fib_entry_notifier(struct notifier_block *nb, struct net *net,
90 enum fib_event_type event_type, u32 dst,
91 int dst_len, struct fib_alias *fa)
92 {
93 struct fib_entry_notifier_info info = {
94 .dst = dst,
95 .dst_len = dst_len,
96 .fi = fa->fa_info,
97 .tos = fa->fa_tos,
98 .type = fa->fa_type,
99 .tb_id = fa->tb_id,
100 };
101 return call_fib4_notifier(nb, net, event_type, &info.info);
102 }
103
104 static int call_fib_entry_notifiers(struct net *net,
105 enum fib_event_type event_type, u32 dst,
106 int dst_len, struct fib_alias *fa,
107 struct netlink_ext_ack *extack)
108 {
109 struct fib_entry_notifier_info info = {
110 .info.extack = extack,
111 .dst = dst,
112 .dst_len = dst_len,
113 .fi = fa->fa_info,
114 .tos = fa->fa_tos,
115 .type = fa->fa_type,
116 .tb_id = fa->tb_id,
117 };
118 return call_fib4_notifiers(net, event_type, &info.info);
119 }
120
121 #define MAX_STAT_DEPTH 32
122
123 #define KEYLENGTH (8*sizeof(t_key))
124 #define KEY_MAX ((t_key)~0)
125
126 typedef unsigned int t_key;
127
128 #define IS_TRIE(n) ((n)->pos >= KEYLENGTH)
129 #define IS_TNODE(n) ((n)->bits)
130 #define IS_LEAF(n) (!(n)->bits)
131
132 struct key_vector {
133 t_key key;
134 unsigned char pos; /* 2log(KEYLENGTH) bits needed */
135 unsigned char bits; /* 2log(KEYLENGTH) bits needed */
136 unsigned char slen;
137 union {
138 /* This list pointer if valid if (pos | bits) == 0 (LEAF) */
139 struct hlist_head leaf;
140 /* This array is valid if (pos | bits) > 0 (TNODE) */
141 struct key_vector __rcu *tnode[0];
142 };
143 };
144
145 struct tnode {
146 struct rcu_head rcu;
147 t_key empty_children; /* KEYLENGTH bits needed */
148 t_key full_children; /* KEYLENGTH bits needed */
149 struct key_vector __rcu *parent;
150 struct key_vector kv[1];
151 #define tn_bits kv[0].bits
152 };
153
154 #define TNODE_SIZE(n) offsetof(struct tnode, kv[0].tnode[n])
155 #define LEAF_SIZE TNODE_SIZE(1)
156
157 #ifdef CONFIG_IP_FIB_TRIE_STATS
158 struct trie_use_stats {
159 unsigned int gets;
160 unsigned int backtrack;
161 unsigned int semantic_match_passed;
162 unsigned int semantic_match_miss;
163 unsigned int null_node_hit;
164 unsigned int resize_node_skipped;
165 };
166 #endif
167
168 struct trie_stat {
169 unsigned int totdepth;
170 unsigned int maxdepth;
171 unsigned int tnodes;
172 unsigned int leaves;
173 unsigned int nullpointers;
174 unsigned int prefixes;
175 unsigned int nodesizes[MAX_STAT_DEPTH];
176 };
177
178 struct trie {
179 struct key_vector kv[1];
180 #ifdef CONFIG_IP_FIB_TRIE_STATS
181 struct trie_use_stats __percpu *stats;
182 #endif
183 };
184
185 static struct key_vector *resize(struct trie *t, struct key_vector *tn);
186 static unsigned int tnode_free_size;
187
188 /*
189 * synchronize_rcu after call_rcu for outstanding dirty memory; it should be
190 * especially useful before resizing the root node with PREEMPT_NONE configs;
191 * the value was obtained experimentally, aiming to avoid visible slowdown.
192 */
193 unsigned int sysctl_fib_sync_mem = 512 * 1024;
194 unsigned int sysctl_fib_sync_mem_min = 64 * 1024;
195 unsigned int sysctl_fib_sync_mem_max = 64 * 1024 * 1024;
196
197 static struct kmem_cache *fn_alias_kmem __ro_after_init;
198 static struct kmem_cache *trie_leaf_kmem __ro_after_init;
199
200 static inline struct tnode *tn_info(struct key_vector *kv)
201 {
202 return container_of(kv, struct tnode, kv[0]);
203 }
204
205 /* caller must hold RTNL */
206 #define node_parent(tn) rtnl_dereference(tn_info(tn)->parent)
207 #define get_child(tn, i) rtnl_dereference((tn)->tnode[i])
208
209 /* caller must hold RCU read lock or RTNL */
210 #define node_parent_rcu(tn) rcu_dereference_rtnl(tn_info(tn)->parent)
211 #define get_child_rcu(tn, i) rcu_dereference_rtnl((tn)->tnode[i])
212
213 /* wrapper for rcu_assign_pointer */
214 static inline void node_set_parent(struct key_vector *n, struct key_vector *tp)
215 {
216 if (n)
217 rcu_assign_pointer(tn_info(n)->parent, tp);
218 }
219
220 #define NODE_INIT_PARENT(n, p) RCU_INIT_POINTER(tn_info(n)->parent, p)
221
222 /* This provides us with the number of children in this node, in the case of a
223 * leaf this will return 0 meaning none of the children are accessible.
224 */
225 static inline unsigned long child_length(const struct key_vector *tn)
226 {
227 return (1ul << tn->bits) & ~(1ul);
228 }
229
230 #define get_cindex(key, kv) (((key) ^ (kv)->key) >> (kv)->pos)
231
232 static inline unsigned long get_index(t_key key, struct key_vector *kv)
233 {
234 unsigned long index = key ^ kv->key;
235
236 if ((BITS_PER_LONG <= KEYLENGTH) && (KEYLENGTH == kv->pos))
237 return 0;
238
239 return index >> kv->pos;
240 }
241
242 /* To understand this stuff, an understanding of keys and all their bits is
243 * necessary. Every node in the trie has a key associated with it, but not
244 * all of the bits in that key are significant.
245 *
246 * Consider a node 'n' and its parent 'tp'.
247 *
248 * If n is a leaf, every bit in its key is significant. Its presence is
249 * necessitated by path compression, since during a tree traversal (when
250 * searching for a leaf - unless we are doing an insertion) we will completely
251 * ignore all skipped bits we encounter. Thus we need to verify, at the end of
252 * a potentially successful search, that we have indeed been walking the
253 * correct key path.
254 *
255 * Note that we can never "miss" the correct key in the tree if present by
256 * following the wrong path. Path compression ensures that segments of the key
257 * that are the same for all keys with a given prefix are skipped, but the
258 * skipped part *is* identical for each node in the subtrie below the skipped
259 * bit! trie_insert() in this implementation takes care of that.
260 *
261 * if n is an internal node - a 'tnode' here, the various parts of its key
262 * have many different meanings.
263 *
264 * Example:
265 * _________________________________________________________________
266 * | i | i | i | i | i | i | i | N | N | N | S | S | S | S | S | C |
267 * -----------------------------------------------------------------
268 * 31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16
269 *
270 * _________________________________________________________________
271 * | C | C | C | u | u | u | u | u | u | u | u | u | u | u | u | u |
272 * -----------------------------------------------------------------
273 * 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
274 *
275 * tp->pos = 22
276 * tp->bits = 3
277 * n->pos = 13
278 * n->bits = 4
279 *
280 * First, let's just ignore the bits that come before the parent tp, that is
281 * the bits from (tp->pos + tp->bits) to 31. They are *known* but at this
282 * point we do not use them for anything.
283 *
284 * The bits from (tp->pos) to (tp->pos + tp->bits - 1) - "N", above - are the
285 * index into the parent's child array. That is, they will be used to find
286 * 'n' among tp's children.
287 *
288 * The bits from (n->pos + n->bits) to (tp->pos - 1) - "S" - are skipped bits
289 * for the node n.
290 *
291 * All the bits we have seen so far are significant to the node n. The rest
292 * of the bits are really not needed or indeed known in n->key.
293 *
294 * The bits from (n->pos) to (n->pos + n->bits - 1) - "C" - are the index into
295 * n's child array, and will of course be different for each child.
296 *
297 * The rest of the bits, from 0 to (n->pos -1) - "u" - are completely unknown
298 * at this point.
299 */
300
301 static const int halve_threshold = 25;
302 static const int inflate_threshold = 50;
303 static const int halve_threshold_root = 15;
304 static const int inflate_threshold_root = 30;
305
306 static void __alias_free_mem(struct rcu_head *head)
307 {
308 struct fib_alias *fa = container_of(head, struct fib_alias, rcu);
309 kmem_cache_free(fn_alias_kmem, fa);
310 }
311
312 static inline void alias_free_mem_rcu(struct fib_alias *fa)
313 {
314 call_rcu(&fa->rcu, __alias_free_mem);
315 }
316
317 #define TNODE_KMALLOC_MAX \
318 ilog2((PAGE_SIZE - TNODE_SIZE(0)) / sizeof(struct key_vector *))
319 #define TNODE_VMALLOC_MAX \
320 ilog2((SIZE_MAX - TNODE_SIZE(0)) / sizeof(struct key_vector *))
321
322 static void __node_free_rcu(struct rcu_head *head)
323 {
324 struct tnode *n = container_of(head, struct tnode, rcu);
325
326 if (!n->tn_bits)
327 kmem_cache_free(trie_leaf_kmem, n);
328 else
329 kvfree(n);
330 }
331
332 #define node_free(n) call_rcu(&tn_info(n)->rcu, __node_free_rcu)
333
334 static struct tnode *tnode_alloc(int bits)
335 {
336 size_t size;
337
338 /* verify bits is within bounds */
339 if (bits > TNODE_VMALLOC_MAX)
340 return NULL;
341
342 /* determine size and verify it is non-zero and didn't overflow */
343 size = TNODE_SIZE(1ul << bits);
344
345 if (size <= PAGE_SIZE)
346 return kzalloc(size, GFP_KERNEL);
347 else
348 return vzalloc(size);
349 }
350
351 static inline void empty_child_inc(struct key_vector *n)
352 {
353 ++tn_info(n)->empty_children ? : ++tn_info(n)->full_children;
354 }
355
356 static inline void empty_child_dec(struct key_vector *n)
357 {
358 tn_info(n)->empty_children-- ? : tn_info(n)->full_children--;
359 }
360
361 static struct key_vector *leaf_new(t_key key, struct fib_alias *fa)
362 {
363 struct key_vector *l;
364 struct tnode *kv;
365
366 kv = kmem_cache_alloc(trie_leaf_kmem, GFP_KERNEL);
367 if (!kv)
368 return NULL;
369
370 /* initialize key vector */
371 l = kv->kv;
372 l->key = key;
373 l->pos = 0;
374 l->bits = 0;
375 l->slen = fa->fa_slen;
376
377 /* link leaf to fib alias */
378 INIT_HLIST_HEAD(&l->leaf);
379 hlist_add_head(&fa->fa_list, &l->leaf);
380
381 return l;
382 }
383
384 static struct key_vector *tnode_new(t_key key, int pos, int bits)
385 {
386 unsigned int shift = pos + bits;
387 struct key_vector *tn;
388 struct tnode *tnode;
389
390 /* verify bits and pos their msb bits clear and values are valid */
391 BUG_ON(!bits || (shift > KEYLENGTH));
392
393 tnode = tnode_alloc(bits);
394 if (!tnode)
395 return NULL;
396
397 pr_debug("AT %p s=%zu %zu\n", tnode, TNODE_SIZE(0),
398 sizeof(struct key_vector *) << bits);
399
400 if (bits == KEYLENGTH)
401 tnode->full_children = 1;
402 else
403 tnode->empty_children = 1ul << bits;
404
405 tn = tnode->kv;
406 tn->key = (shift < KEYLENGTH) ? (key >> shift) << shift : 0;
407 tn->pos = pos;
408 tn->bits = bits;
409 tn->slen = pos;
410
411 return tn;
412 }
413
414 /* Check whether a tnode 'n' is "full", i.e. it is an internal node
415 * and no bits are skipped. See discussion in dyntree paper p. 6
416 */
417 static inline int tnode_full(struct key_vector *tn, struct key_vector *n)
418 {
419 return n && ((n->pos + n->bits) == tn->pos) && IS_TNODE(n);
420 }
421
422 /* Add a child at position i overwriting the old value.
423 * Update the value of full_children and empty_children.
424 */
425 static void put_child(struct key_vector *tn, unsigned long i,
426 struct key_vector *n)
427 {
428 struct key_vector *chi = get_child(tn, i);
429 int isfull, wasfull;
430
431 BUG_ON(i >= child_length(tn));
432
433 /* update emptyChildren, overflow into fullChildren */
434 if (!n && chi)
435 empty_child_inc(tn);
436 if (n && !chi)
437 empty_child_dec(tn);
438
439 /* update fullChildren */
440 wasfull = tnode_full(tn, chi);
441 isfull = tnode_full(tn, n);
442
443 if (wasfull && !isfull)
444 tn_info(tn)->full_children--;
445 else if (!wasfull && isfull)
446 tn_info(tn)->full_children++;
447
448 if (n && (tn->slen < n->slen))
449 tn->slen = n->slen;
450
451 rcu_assign_pointer(tn->tnode[i], n);
452 }
453
454 static void update_children(struct key_vector *tn)
455 {
456 unsigned long i;
457
458 /* update all of the child parent pointers */
459 for (i = child_length(tn); i;) {
460 struct key_vector *inode = get_child(tn, --i);
461
462 if (!inode)
463 continue;
464
465 /* Either update the children of a tnode that
466 * already belongs to us or update the child
467 * to point to ourselves.
468 */
469 if (node_parent(inode) == tn)
470 update_children(inode);
471 else
472 node_set_parent(inode, tn);
473 }
474 }
475
476 static inline void put_child_root(struct key_vector *tp, t_key key,
477 struct key_vector *n)
478 {
479 if (IS_TRIE(tp))
480 rcu_assign_pointer(tp->tnode[0], n);
481 else
482 put_child(tp, get_index(key, tp), n);
483 }
484
485 static inline void tnode_free_init(struct key_vector *tn)
486 {
487 tn_info(tn)->rcu.next = NULL;
488 }
489
490 static inline void tnode_free_append(struct key_vector *tn,
491 struct key_vector *n)
492 {
493 tn_info(n)->rcu.next = tn_info(tn)->rcu.next;
494 tn_info(tn)->rcu.next = &tn_info(n)->rcu;
495 }
496
497 static void tnode_free(struct key_vector *tn)
498 {
499 struct callback_head *head = &tn_info(tn)->rcu;
500
501 while (head) {
502 head = head->next;
503 tnode_free_size += TNODE_SIZE(1ul << tn->bits);
504 node_free(tn);
505
506 tn = container_of(head, struct tnode, rcu)->kv;
507 }
508
509 if (tnode_free_size >= sysctl_fib_sync_mem) {
510 tnode_free_size = 0;
511 synchronize_rcu();
512 }
513 }
514
515 static struct key_vector *replace(struct trie *t,
516 struct key_vector *oldtnode,
517 struct key_vector *tn)
518 {
519 struct key_vector *tp = node_parent(oldtnode);
520 unsigned long i;
521
522 /* setup the parent pointer out of and back into this node */
523 NODE_INIT_PARENT(tn, tp);
524 put_child_root(tp, tn->key, tn);
525
526 /* update all of the child parent pointers */
527 update_children(tn);
528
529 /* all pointers should be clean so we are done */
530 tnode_free(oldtnode);
531
532 /* resize children now that oldtnode is freed */
533 for (i = child_length(tn); i;) {
534 struct key_vector *inode = get_child(tn, --i);
535
536 /* resize child node */
537 if (tnode_full(tn, inode))
538 tn = resize(t, inode);
539 }
540
541 return tp;
542 }
543
544 static struct key_vector *inflate(struct trie *t,
545 struct key_vector *oldtnode)
546 {
547 struct key_vector *tn;
548 unsigned long i;
549 t_key m;
550
551 pr_debug("In inflate\n");
552
553 tn = tnode_new(oldtnode->key, oldtnode->pos - 1, oldtnode->bits + 1);
554 if (!tn)
555 goto notnode;
556
557 /* prepare oldtnode to be freed */
558 tnode_free_init(oldtnode);
559
560 /* Assemble all of the pointers in our cluster, in this case that
561 * represents all of the pointers out of our allocated nodes that
562 * point to existing tnodes and the links between our allocated
563 * nodes.
564 */
565 for (i = child_length(oldtnode), m = 1u << tn->pos; i;) {
566 struct key_vector *inode = get_child(oldtnode, --i);
567 struct key_vector *node0, *node1;
568 unsigned long j, k;
569
570 /* An empty child */
571 if (!inode)
572 continue;
573
574 /* A leaf or an internal node with skipped bits */
575 if (!tnode_full(oldtnode, inode)) {
576 put_child(tn, get_index(inode->key, tn), inode);
577 continue;
578 }
579
580 /* drop the node in the old tnode free list */
581 tnode_free_append(oldtnode, inode);
582
583 /* An internal node with two children */
584 if (inode->bits == 1) {
585 put_child(tn, 2 * i + 1, get_child(inode, 1));
586 put_child(tn, 2 * i, get_child(inode, 0));
587 continue;
588 }
589
590 /* We will replace this node 'inode' with two new
591 * ones, 'node0' and 'node1', each with half of the
592 * original children. The two new nodes will have
593 * a position one bit further down the key and this
594 * means that the "significant" part of their keys
595 * (see the discussion near the top of this file)
596 * will differ by one bit, which will be "0" in
597 * node0's key and "1" in node1's key. Since we are
598 * moving the key position by one step, the bit that
599 * we are moving away from - the bit at position
600 * (tn->pos) - is the one that will differ between
601 * node0 and node1. So... we synthesize that bit in the
602 * two new keys.
603 */
604 node1 = tnode_new(inode->key | m, inode->pos, inode->bits - 1);
605 if (!node1)
606 goto nomem;
607 node0 = tnode_new(inode->key, inode->pos, inode->bits - 1);
608
609 tnode_free_append(tn, node1);
610 if (!node0)
611 goto nomem;
612 tnode_free_append(tn, node0);
613
614 /* populate child pointers in new nodes */
615 for (k = child_length(inode), j = k / 2; j;) {
616 put_child(node1, --j, get_child(inode, --k));
617 put_child(node0, j, get_child(inode, j));
618 put_child(node1, --j, get_child(inode, --k));
619 put_child(node0, j, get_child(inode, j));
620 }
621
622 /* link new nodes to parent */
623 NODE_INIT_PARENT(node1, tn);
624 NODE_INIT_PARENT(node0, tn);
625
626 /* link parent to nodes */
627 put_child(tn, 2 * i + 1, node1);
628 put_child(tn, 2 * i, node0);
629 }
630
631 /* setup the parent pointers into and out of this node */
632 return replace(t, oldtnode, tn);
633 nomem:
634 /* all pointers should be clean so we are done */
635 tnode_free(tn);
636 notnode:
637 return NULL;
638 }
639
640 static struct key_vector *halve(struct trie *t,
641 struct key_vector *oldtnode)
642 {
643 struct key_vector *tn;
644 unsigned long i;
645
646 pr_debug("In halve\n");
647
648 tn = tnode_new(oldtnode->key, oldtnode->pos + 1, oldtnode->bits - 1);
649 if (!tn)
650 goto notnode;
651
652 /* prepare oldtnode to be freed */
653 tnode_free_init(oldtnode);
654
655 /* Assemble all of the pointers in our cluster, in this case that
656 * represents all of the pointers out of our allocated nodes that
657 * point to existing tnodes and the links between our allocated
658 * nodes.
659 */
660 for (i = child_length(oldtnode); i;) {
661 struct key_vector *node1 = get_child(oldtnode, --i);
662 struct key_vector *node0 = get_child(oldtnode, --i);
663 struct key_vector *inode;
664
665 /* At least one of the children is empty */
666 if (!node1 || !node0) {
667 put_child(tn, i / 2, node1 ? : node0);
668 continue;
669 }
670
671 /* Two nonempty children */
672 inode = tnode_new(node0->key, oldtnode->pos, 1);
673 if (!inode)
674 goto nomem;
675 tnode_free_append(tn, inode);
676
677 /* initialize pointers out of node */
678 put_child(inode, 1, node1);
679 put_child(inode, 0, node0);
680 NODE_INIT_PARENT(inode, tn);
681
682 /* link parent to node */
683 put_child(tn, i / 2, inode);
684 }
685
686 /* setup the parent pointers into and out of this node */
687 return replace(t, oldtnode, tn);
688 nomem:
689 /* all pointers should be clean so we are done */
690 tnode_free(tn);
691 notnode:
692 return NULL;
693 }
694
695 static struct key_vector *collapse(struct trie *t,
696 struct key_vector *oldtnode)
697 {
698 struct key_vector *n, *tp;
699 unsigned long i;
700
701 /* scan the tnode looking for that one child that might still exist */
702 for (n = NULL, i = child_length(oldtnode); !n && i;)
703 n = get_child(oldtnode, --i);
704
705 /* compress one level */
706 tp = node_parent(oldtnode);
707 put_child_root(tp, oldtnode->key, n);
708 node_set_parent(n, tp);
709
710 /* drop dead node */
711 node_free(oldtnode);
712
713 return tp;
714 }
715
716 static unsigned char update_suffix(struct key_vector *tn)
717 {
718 unsigned char slen = tn->pos;
719 unsigned long stride, i;
720 unsigned char slen_max;
721
722 /* only vector 0 can have a suffix length greater than or equal to
723 * tn->pos + tn->bits, the second highest node will have a suffix
724 * length at most of tn->pos + tn->bits - 1
725 */
726 slen_max = min_t(unsigned char, tn->pos + tn->bits - 1, tn->slen);
727
728 /* search though the list of children looking for nodes that might
729 * have a suffix greater than the one we currently have. This is
730 * why we start with a stride of 2 since a stride of 1 would
731 * represent the nodes with suffix length equal to tn->pos
732 */
733 for (i = 0, stride = 0x2ul ; i < child_length(tn); i += stride) {
734 struct key_vector *n = get_child(tn, i);
735
736 if (!n || (n->slen <= slen))
737 continue;
738
739 /* update stride and slen based on new value */
740 stride <<= (n->slen - slen);
741 slen = n->slen;
742 i &= ~(stride - 1);
743
744 /* stop searching if we have hit the maximum possible value */
745 if (slen >= slen_max)
746 break;
747 }
748
749 tn->slen = slen;
750
751 return slen;
752 }
753
754 /* From "Implementing a dynamic compressed trie" by Stefan Nilsson of
755 * the Helsinki University of Technology and Matti Tikkanen of Nokia
756 * Telecommunications, page 6:
757 * "A node is doubled if the ratio of non-empty children to all
758 * children in the *doubled* node is at least 'high'."
759 *
760 * 'high' in this instance is the variable 'inflate_threshold'. It
761 * is expressed as a percentage, so we multiply it with
762 * child_length() and instead of multiplying by 2 (since the
763 * child array will be doubled by inflate()) and multiplying
764 * the left-hand side by 100 (to handle the percentage thing) we
765 * multiply the left-hand side by 50.
766 *
767 * The left-hand side may look a bit weird: child_length(tn)
768 * - tn->empty_children is of course the number of non-null children
769 * in the current node. tn->full_children is the number of "full"
770 * children, that is non-null tnodes with a skip value of 0.
771 * All of those will be doubled in the resulting inflated tnode, so
772 * we just count them one extra time here.
773 *
774 * A clearer way to write this would be:
775 *
776 * to_be_doubled = tn->full_children;
777 * not_to_be_doubled = child_length(tn) - tn->empty_children -
778 * tn->full_children;
779 *
780 * new_child_length = child_length(tn) * 2;
781 *
782 * new_fill_factor = 100 * (not_to_be_doubled + 2*to_be_doubled) /
783 * new_child_length;
784 * if (new_fill_factor >= inflate_threshold)
785 *
786 * ...and so on, tho it would mess up the while () loop.
787 *
788 * anyway,
789 * 100 * (not_to_be_doubled + 2*to_be_doubled) / new_child_length >=
790 * inflate_threshold
791 *
792 * avoid a division:
793 * 100 * (not_to_be_doubled + 2*to_be_doubled) >=
794 * inflate_threshold * new_child_length
795 *
796 * expand not_to_be_doubled and to_be_doubled, and shorten:
797 * 100 * (child_length(tn) - tn->empty_children +
798 * tn->full_children) >= inflate_threshold * new_child_length
799 *
800 * expand new_child_length:
801 * 100 * (child_length(tn) - tn->empty_children +
802 * tn->full_children) >=
803 * inflate_threshold * child_length(tn) * 2
804 *
805 * shorten again:
806 * 50 * (tn->full_children + child_length(tn) -
807 * tn->empty_children) >= inflate_threshold *
808 * child_length(tn)
809 *
810 */
811 static inline bool should_inflate(struct key_vector *tp, struct key_vector *tn)
812 {
813 unsigned long used = child_length(tn);
814 unsigned long threshold = used;
815
816 /* Keep root node larger */
817 threshold *= IS_TRIE(tp) ? inflate_threshold_root : inflate_threshold;
818 used -= tn_info(tn)->empty_children;
819 used += tn_info(tn)->full_children;
820
821 /* if bits == KEYLENGTH then pos = 0, and will fail below */
822
823 return (used > 1) && tn->pos && ((50 * used) >= threshold);
824 }
825
826 static inline bool should_halve(struct key_vector *tp, struct key_vector *tn)
827 {
828 unsigned long used = child_length(tn);
829 unsigned long threshold = used;
830
831 /* Keep root node larger */
832 threshold *= IS_TRIE(tp) ? halve_threshold_root : halve_threshold;
833 used -= tn_info(tn)->empty_children;
834
835 /* if bits == KEYLENGTH then used = 100% on wrap, and will fail below */
836
837 return (used > 1) && (tn->bits > 1) && ((100 * used) < threshold);
838 }
839
840 static inline bool should_collapse(struct key_vector *tn)
841 {
842 unsigned long used = child_length(tn);
843
844 used -= tn_info(tn)->empty_children;
845
846 /* account for bits == KEYLENGTH case */
847 if ((tn->bits == KEYLENGTH) && tn_info(tn)->full_children)
848 used -= KEY_MAX;
849
850 /* One child or none, time to drop us from the trie */
851 return used < 2;
852 }
853
854 #define MAX_WORK 10
855 static struct key_vector *resize(struct trie *t, struct key_vector *tn)
856 {
857 #ifdef CONFIG_IP_FIB_TRIE_STATS
858 struct trie_use_stats __percpu *stats = t->stats;
859 #endif
860 struct key_vector *tp = node_parent(tn);
861 unsigned long cindex = get_index(tn->key, tp);
862 int max_work = MAX_WORK;
863
864 pr_debug("In tnode_resize %p inflate_threshold=%d threshold=%d\n",
865 tn, inflate_threshold, halve_threshold);
866
867 /* track the tnode via the pointer from the parent instead of
868 * doing it ourselves. This way we can let RCU fully do its
869 * thing without us interfering
870 */
871 BUG_ON(tn != get_child(tp, cindex));
872
873 /* Double as long as the resulting node has a number of
874 * nonempty nodes that are above the threshold.
875 */
876 while (should_inflate(tp, tn) && max_work) {
877 tp = inflate(t, tn);
878 if (!tp) {
879 #ifdef CONFIG_IP_FIB_TRIE_STATS
880 this_cpu_inc(stats->resize_node_skipped);
881 #endif
882 break;
883 }
884
885 max_work--;
886 tn = get_child(tp, cindex);
887 }
888
889 /* update parent in case inflate failed */
890 tp = node_parent(tn);
891
892 /* Return if at least one inflate is run */
893 if (max_work != MAX_WORK)
894 return tp;
895
896 /* Halve as long as the number of empty children in this
897 * node is above threshold.
898 */
899 while (should_halve(tp, tn) && max_work) {
900 tp = halve(t, tn);
901 if (!tp) {
902 #ifdef CONFIG_IP_FIB_TRIE_STATS
903 this_cpu_inc(stats->resize_node_skipped);
904 #endif
905 break;
906 }
907
908 max_work--;
909 tn = get_child(tp, cindex);
910 }
911
912 /* Only one child remains */
913 if (should_collapse(tn))
914 return collapse(t, tn);
915
916 /* update parent in case halve failed */
917 return node_parent(tn);
918 }
919
920 static void node_pull_suffix(struct key_vector *tn, unsigned char slen)
921 {
922 unsigned char node_slen = tn->slen;
923
924 while ((node_slen > tn->pos) && (node_slen > slen)) {
925 slen = update_suffix(tn);
926 if (node_slen == slen)
927 break;
928
929 tn = node_parent(tn);
930 node_slen = tn->slen;
931 }
932 }
933
934 static void node_push_suffix(struct key_vector *tn, unsigned char slen)
935 {
936 while (tn->slen < slen) {
937 tn->slen = slen;
938 tn = node_parent(tn);
939 }
940 }
941
942 /* rcu_read_lock needs to be hold by caller from readside */
943 static struct key_vector *fib_find_node(struct trie *t,
944 struct key_vector **tp, u32 key)
945 {
946 struct key_vector *pn, *n = t->kv;
947 unsigned long index = 0;
948
949 do {
950 pn = n;
951 n = get_child_rcu(n, index);
952
953 if (!n)
954 break;
955
956 index = get_cindex(key, n);
957
958 /* This bit of code is a bit tricky but it combines multiple
959 * checks into a single check. The prefix consists of the
960 * prefix plus zeros for the bits in the cindex. The index
961 * is the difference between the key and this value. From
962 * this we can actually derive several pieces of data.
963 * if (index >= (1ul << bits))
964 * we have a mismatch in skip bits and failed
965 * else
966 * we know the value is cindex
967 *
968 * This check is safe even if bits == KEYLENGTH due to the
969 * fact that we can only allocate a node with 32 bits if a
970 * long is greater than 32 bits.
971 */
972 if (index >= (1ul << n->bits)) {
973 n = NULL;
974 break;
975 }
976
977 /* keep searching until we find a perfect match leaf or NULL */
978 } while (IS_TNODE(n));
979
980 *tp = pn;
981
982 return n;
983 }
984
985 /* Return the first fib alias matching TOS with
986 * priority less than or equal to PRIO.
987 */
988 static struct fib_alias *fib_find_alias(struct hlist_head *fah, u8 slen,
989 u8 tos, u32 prio, u32 tb_id)
990 {
991 struct fib_alias *fa;
992
993 if (!fah)
994 return NULL;
995
996 hlist_for_each_entry(fa, fah, fa_list) {
997 if (fa->fa_slen < slen)
998 continue;
999 if (fa->fa_slen != slen)
1000 break;
1001 if (fa->tb_id > tb_id)
1002 continue;
1003 if (fa->tb_id != tb_id)
1004 break;
1005 if (fa->fa_tos > tos)
1006 continue;
1007 if (fa->fa_info->fib_priority >= prio || fa->fa_tos < tos)
1008 return fa;
1009 }
1010
1011 return NULL;
1012 }
1013
1014 static void trie_rebalance(struct trie *t, struct key_vector *tn)
1015 {
1016 while (!IS_TRIE(tn))
1017 tn = resize(t, tn);
1018 }
1019
1020 static int fib_insert_node(struct trie *t, struct key_vector *tp,
1021 struct fib_alias *new, t_key key)
1022 {
1023 struct key_vector *n, *l;
1024
1025 l = leaf_new(key, new);
1026 if (!l)
1027 goto noleaf;
1028
1029 /* retrieve child from parent node */
1030 n = get_child(tp, get_index(key, tp));
1031
1032 /* Case 2: n is a LEAF or a TNODE and the key doesn't match.
1033 *
1034 * Add a new tnode here
1035 * first tnode need some special handling
1036 * leaves us in position for handling as case 3
1037 */
1038 if (n) {
1039 struct key_vector *tn;
1040
1041 tn = tnode_new(key, __fls(key ^ n->key), 1);
1042 if (!tn)
1043 goto notnode;
1044
1045 /* initialize routes out of node */
1046 NODE_INIT_PARENT(tn, tp);
1047 put_child(tn, get_index(key, tn) ^ 1, n);
1048
1049 /* start adding routes into the node */
1050 put_child_root(tp, key, tn);
1051 node_set_parent(n, tn);
1052
1053 /* parent now has a NULL spot where the leaf can go */
1054 tp = tn;
1055 }
1056
1057 /* Case 3: n is NULL, and will just insert a new leaf */
1058 node_push_suffix(tp, new->fa_slen);
1059 NODE_INIT_PARENT(l, tp);
1060 put_child_root(tp, key, l);
1061 trie_rebalance(t, tp);
1062
1063 return 0;
1064 notnode:
1065 node_free(l);
1066 noleaf:
1067 return -ENOMEM;
1068 }
1069
1070 /* fib notifier for ADD is sent before calling fib_insert_alias with
1071 * the expectation that the only possible failure ENOMEM
1072 */
1073 static int fib_insert_alias(struct trie *t, struct key_vector *tp,
1074 struct key_vector *l, struct fib_alias *new,
1075 struct fib_alias *fa, t_key key)
1076 {
1077 if (!l)
1078 return fib_insert_node(t, tp, new, key);
1079
1080 if (fa) {
1081 hlist_add_before_rcu(&new->fa_list, &fa->fa_list);
1082 } else {
1083 struct fib_alias *last;
1084
1085 hlist_for_each_entry(last, &l->leaf, fa_list) {
1086 if (new->fa_slen < last->fa_slen)
1087 break;
1088 if ((new->fa_slen == last->fa_slen) &&
1089 (new->tb_id > last->tb_id))
1090 break;
1091 fa = last;
1092 }
1093
1094 if (fa)
1095 hlist_add_behind_rcu(&new->fa_list, &fa->fa_list);
1096 else
1097 hlist_add_head_rcu(&new->fa_list, &l->leaf);
1098 }
1099
1100 /* if we added to the tail node then we need to update slen */
1101 if (l->slen < new->fa_slen) {
1102 l->slen = new->fa_slen;
1103 node_push_suffix(tp, new->fa_slen);
1104 }
1105
1106 return 0;
1107 }
1108
1109 static bool fib_valid_key_len(u32 key, u8 plen, struct netlink_ext_ack *extack)
1110 {
1111 if (plen > KEYLENGTH) {
1112 NL_SET_ERR_MSG(extack, "Invalid prefix length");
1113 return false;
1114 }
1115
1116 if ((plen < KEYLENGTH) && (key << plen)) {
1117 NL_SET_ERR_MSG(extack,
1118 "Invalid prefix for given prefix length");
1119 return false;
1120 }
1121
1122 return true;
1123 }
1124
1125 /* Caller must hold RTNL. */
1126 int fib_table_insert(struct net *net, struct fib_table *tb,
1127 struct fib_config *cfg, struct netlink_ext_ack *extack)
1128 {
1129 enum fib_event_type event = FIB_EVENT_ENTRY_ADD;
1130 struct trie *t = (struct trie *)tb->tb_data;
1131 struct fib_alias *fa, *new_fa;
1132 struct key_vector *l, *tp;
1133 u16 nlflags = NLM_F_EXCL;
1134 struct fib_info *fi;
1135 u8 plen = cfg->fc_dst_len;
1136 u8 slen = KEYLENGTH - plen;
1137 u8 tos = cfg->fc_tos;
1138 u32 key;
1139 int err;
1140
1141 key = ntohl(cfg->fc_dst);
1142
1143 if (!fib_valid_key_len(key, plen, extack))
1144 return -EINVAL;
1145
1146 pr_debug("Insert table=%u %08x/%d\n", tb->tb_id, key, plen);
1147
1148 fi = fib_create_info(cfg, extack);
1149 if (IS_ERR(fi)) {
1150 err = PTR_ERR(fi);
1151 goto err;
1152 }
1153
1154 l = fib_find_node(t, &tp, key);
1155 fa = l ? fib_find_alias(&l->leaf, slen, tos, fi->fib_priority,
1156 tb->tb_id) : NULL;
1157
1158 /* Now fa, if non-NULL, points to the first fib alias
1159 * with the same keys [prefix,tos,priority], if such key already
1160 * exists or to the node before which we will insert new one.
1161 *
1162 * If fa is NULL, we will need to allocate a new one and
1163 * insert to the tail of the section matching the suffix length
1164 * of the new alias.
1165 */
1166
1167 if (fa && fa->fa_tos == tos &&
1168 fa->fa_info->fib_priority == fi->fib_priority) {
1169 struct fib_alias *fa_first, *fa_match;
1170
1171 err = -EEXIST;
1172 if (cfg->fc_nlflags & NLM_F_EXCL)
1173 goto out;
1174
1175 nlflags &= ~NLM_F_EXCL;
1176
1177 /* We have 2 goals:
1178 * 1. Find exact match for type, scope, fib_info to avoid
1179 * duplicate routes
1180 * 2. Find next 'fa' (or head), NLM_F_APPEND inserts before it
1181 */
1182 fa_match = NULL;
1183 fa_first = fa;
1184 hlist_for_each_entry_from(fa, fa_list) {
1185 if ((fa->fa_slen != slen) ||
1186 (fa->tb_id != tb->tb_id) ||
1187 (fa->fa_tos != tos))
1188 break;
1189 if (fa->fa_info->fib_priority != fi->fib_priority)
1190 break;
1191 if (fa->fa_type == cfg->fc_type &&
1192 fa->fa_info == fi) {
1193 fa_match = fa;
1194 break;
1195 }
1196 }
1197
1198 if (cfg->fc_nlflags & NLM_F_REPLACE) {
1199 struct fib_info *fi_drop;
1200 u8 state;
1201
1202 nlflags |= NLM_F_REPLACE;
1203 fa = fa_first;
1204 if (fa_match) {
1205 if (fa == fa_match)
1206 err = 0;
1207 goto out;
1208 }
1209 err = -ENOBUFS;
1210 new_fa = kmem_cache_alloc(fn_alias_kmem, GFP_KERNEL);
1211 if (!new_fa)
1212 goto out;
1213
1214 fi_drop = fa->fa_info;
1215 new_fa->fa_tos = fa->fa_tos;
1216 new_fa->fa_info = fi;
1217 new_fa->fa_type = cfg->fc_type;
1218 state = fa->fa_state;
1219 new_fa->fa_state = state & ~FA_S_ACCESSED;
1220 new_fa->fa_slen = fa->fa_slen;
1221 new_fa->tb_id = tb->tb_id;
1222 new_fa->fa_default = -1;
1223
1224 err = call_fib_entry_notifiers(net,
1225 FIB_EVENT_ENTRY_REPLACE,
1226 key, plen, new_fa,
1227 extack);
1228 if (err)
1229 goto out_free_new_fa;
1230
1231 rtmsg_fib(RTM_NEWROUTE, htonl(key), new_fa, plen,
1232 tb->tb_id, &cfg->fc_nlinfo, nlflags);
1233
1234 hlist_replace_rcu(&fa->fa_list, &new_fa->fa_list);
1235
1236 alias_free_mem_rcu(fa);
1237
1238 fib_release_info(fi_drop);
1239 if (state & FA_S_ACCESSED)
1240 rt_cache_flush(cfg->fc_nlinfo.nl_net);
1241
1242 goto succeeded;
1243 }
1244 /* Error if we find a perfect match which
1245 * uses the same scope, type, and nexthop
1246 * information.
1247 */
1248 if (fa_match)
1249 goto out;
1250
1251 if (cfg->fc_nlflags & NLM_F_APPEND) {
1252 event = FIB_EVENT_ENTRY_APPEND;
1253 nlflags |= NLM_F_APPEND;
1254 } else {
1255 fa = fa_first;
1256 }
1257 }
1258 err = -ENOENT;
1259 if (!(cfg->fc_nlflags & NLM_F_CREATE))
1260 goto out;
1261
1262 nlflags |= NLM_F_CREATE;
1263 err = -ENOBUFS;
1264 new_fa = kmem_cache_alloc(fn_alias_kmem, GFP_KERNEL);
1265 if (!new_fa)
1266 goto out;
1267
1268 new_fa->fa_info = fi;
1269 new_fa->fa_tos = tos;
1270 new_fa->fa_type = cfg->fc_type;
1271 new_fa->fa_state = 0;
1272 new_fa->fa_slen = slen;
1273 new_fa->tb_id = tb->tb_id;
1274 new_fa->fa_default = -1;
1275
1276 err = call_fib_entry_notifiers(net, event, key, plen, new_fa, extack);
1277 if (err)
1278 goto out_free_new_fa;
1279
1280 /* Insert new entry to the list. */
1281 err = fib_insert_alias(t, tp, l, new_fa, fa, key);
1282 if (err)
1283 goto out_fib_notif;
1284
1285 if (!plen)
1286 tb->tb_num_default++;
1287
1288 rt_cache_flush(cfg->fc_nlinfo.nl_net);
1289 rtmsg_fib(RTM_NEWROUTE, htonl(key), new_fa, plen, new_fa->tb_id,
1290 &cfg->fc_nlinfo, nlflags);
1291 succeeded:
1292 return 0;
1293
1294 out_fib_notif:
1295 /* notifier was sent that entry would be added to trie, but
1296 * the add failed and need to recover. Only failure for
1297 * fib_insert_alias is ENOMEM.
1298 */
1299 NL_SET_ERR_MSG(extack, "Failed to insert route into trie");
1300 call_fib_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, key,
1301 plen, new_fa, NULL);
1302 out_free_new_fa:
1303 kmem_cache_free(fn_alias_kmem, new_fa);
1304 out:
1305 fib_release_info(fi);
1306 err:
1307 return err;
1308 }
1309
1310 static inline t_key prefix_mismatch(t_key key, struct key_vector *n)
1311 {
1312 t_key prefix = n->key;
1313
1314 return (key ^ prefix) & (prefix | -prefix);
1315 }
1316
1317 /* should be called with rcu_read_lock */
1318 int fib_table_lookup(struct fib_table *tb, const struct flowi4 *flp,
1319 struct fib_result *res, int fib_flags)
1320 {
1321 struct trie *t = (struct trie *) tb->tb_data;
1322 #ifdef CONFIG_IP_FIB_TRIE_STATS
1323 struct trie_use_stats __percpu *stats = t->stats;
1324 #endif
1325 const t_key key = ntohl(flp->daddr);
1326 struct key_vector *n, *pn;
1327 struct fib_alias *fa;
1328 unsigned long index;
1329 t_key cindex;
1330
1331 pn = t->kv;
1332 cindex = 0;
1333
1334 n = get_child_rcu(pn, cindex);
1335 if (!n) {
1336 trace_fib_table_lookup(tb->tb_id, flp, NULL, -EAGAIN);
1337 return -EAGAIN;
1338 }
1339
1340 #ifdef CONFIG_IP_FIB_TRIE_STATS
1341 this_cpu_inc(stats->gets);
1342 #endif
1343
1344 /* Step 1: Travel to the longest prefix match in the trie */
1345 for (;;) {
1346 index = get_cindex(key, n);
1347
1348 /* This bit of code is a bit tricky but it combines multiple
1349 * checks into a single check. The prefix consists of the
1350 * prefix plus zeros for the "bits" in the prefix. The index
1351 * is the difference between the key and this value. From
1352 * this we can actually derive several pieces of data.
1353 * if (index >= (1ul << bits))
1354 * we have a mismatch in skip bits and failed
1355 * else
1356 * we know the value is cindex
1357 *
1358 * This check is safe even if bits == KEYLENGTH due to the
1359 * fact that we can only allocate a node with 32 bits if a
1360 * long is greater than 32 bits.
1361 */
1362 if (index >= (1ul << n->bits))
1363 break;
1364
1365 /* we have found a leaf. Prefixes have already been compared */
1366 if (IS_LEAF(n))
1367 goto found;
1368
1369 /* only record pn and cindex if we are going to be chopping
1370 * bits later. Otherwise we are just wasting cycles.
1371 */
1372 if (n->slen > n->pos) {
1373 pn = n;
1374 cindex = index;
1375 }
1376
1377 n = get_child_rcu(n, index);
1378 if (unlikely(!n))
1379 goto backtrace;
1380 }
1381
1382 /* Step 2: Sort out leaves and begin backtracing for longest prefix */
1383 for (;;) {
1384 /* record the pointer where our next node pointer is stored */
1385 struct key_vector __rcu **cptr = n->tnode;
1386
1387 /* This test verifies that none of the bits that differ
1388 * between the key and the prefix exist in the region of
1389 * the lsb and higher in the prefix.
1390 */
1391 if (unlikely(prefix_mismatch(key, n)) || (n->slen == n->pos))
1392 goto backtrace;
1393
1394 /* exit out and process leaf */
1395 if (unlikely(IS_LEAF(n)))
1396 break;
1397
1398 /* Don't bother recording parent info. Since we are in
1399 * prefix match mode we will have to come back to wherever
1400 * we started this traversal anyway
1401 */
1402
1403 while ((n = rcu_dereference(*cptr)) == NULL) {
1404 backtrace:
1405 #ifdef CONFIG_IP_FIB_TRIE_STATS
1406 if (!n)
1407 this_cpu_inc(stats->null_node_hit);
1408 #endif
1409 /* If we are at cindex 0 there are no more bits for
1410 * us to strip at this level so we must ascend back
1411 * up one level to see if there are any more bits to
1412 * be stripped there.
1413 */
1414 while (!cindex) {
1415 t_key pkey = pn->key;
1416
1417 /* If we don't have a parent then there is
1418 * nothing for us to do as we do not have any
1419 * further nodes to parse.
1420 */
1421 if (IS_TRIE(pn)) {
1422 trace_fib_table_lookup(tb->tb_id, flp,
1423 NULL, -EAGAIN);
1424 return -EAGAIN;
1425 }
1426 #ifdef CONFIG_IP_FIB_TRIE_STATS
1427 this_cpu_inc(stats->backtrack);
1428 #endif
1429 /* Get Child's index */
1430 pn = node_parent_rcu(pn);
1431 cindex = get_index(pkey, pn);
1432 }
1433
1434 /* strip the least significant bit from the cindex */
1435 cindex &= cindex - 1;
1436
1437 /* grab pointer for next child node */
1438 cptr = &pn->tnode[cindex];
1439 }
1440 }
1441
1442 found:
1443 /* this line carries forward the xor from earlier in the function */
1444 index = key ^ n->key;
1445
1446 /* Step 3: Process the leaf, if that fails fall back to backtracing */
1447 hlist_for_each_entry_rcu(fa, &n->leaf, fa_list) {
1448 struct fib_info *fi = fa->fa_info;
1449 int nhsel, err;
1450
1451 if ((BITS_PER_LONG > KEYLENGTH) || (fa->fa_slen < KEYLENGTH)) {
1452 if (index >= (1ul << fa->fa_slen))
1453 continue;
1454 }
1455 if (fa->fa_tos && fa->fa_tos != flp->flowi4_tos)
1456 continue;
1457 if (fi->fib_dead)
1458 continue;
1459 if (fa->fa_info->fib_scope < flp->flowi4_scope)
1460 continue;
1461 fib_alias_accessed(fa);
1462 err = fib_props[fa->fa_type].error;
1463 if (unlikely(err < 0)) {
1464 #ifdef CONFIG_IP_FIB_TRIE_STATS
1465 this_cpu_inc(stats->semantic_match_passed);
1466 #endif
1467 trace_fib_table_lookup(tb->tb_id, flp, NULL, err);
1468 return err;
1469 }
1470 if (fi->fib_flags & RTNH_F_DEAD)
1471 continue;
1472 for (nhsel = 0; nhsel < fi->fib_nhs; nhsel++) {
1473 struct fib_nh_common *nhc = fib_info_nhc(fi, nhsel);
1474
1475 if (nhc->nhc_flags & RTNH_F_DEAD)
1476 continue;
1477 if (ip_ignore_linkdown(nhc->nhc_dev) &&
1478 nhc->nhc_flags & RTNH_F_LINKDOWN &&
1479 !(fib_flags & FIB_LOOKUP_IGNORE_LINKSTATE))
1480 continue;
1481 if (!(flp->flowi4_flags & FLOWI_FLAG_SKIP_NH_OIF)) {
1482 if (flp->flowi4_oif &&
1483 flp->flowi4_oif != nhc->nhc_oif)
1484 continue;
1485 }
1486
1487 if (!(fib_flags & FIB_LOOKUP_NOREF))
1488 refcount_inc(&fi->fib_clntref);
1489
1490 res->prefix = htonl(n->key);
1491 res->prefixlen = KEYLENGTH - fa->fa_slen;
1492 res->nh_sel = nhsel;
1493 res->nhc = nhc;
1494 res->type = fa->fa_type;
1495 res->scope = fi->fib_scope;
1496 res->fi = fi;
1497 res->table = tb;
1498 res->fa_head = &n->leaf;
1499 #ifdef CONFIG_IP_FIB_TRIE_STATS
1500 this_cpu_inc(stats->semantic_match_passed);
1501 #endif
1502 trace_fib_table_lookup(tb->tb_id, flp, nhc, err);
1503
1504 return err;
1505 }
1506 }
1507 #ifdef CONFIG_IP_FIB_TRIE_STATS
1508 this_cpu_inc(stats->semantic_match_miss);
1509 #endif
1510 goto backtrace;
1511 }
1512 EXPORT_SYMBOL_GPL(fib_table_lookup);
1513
1514 static void fib_remove_alias(struct trie *t, struct key_vector *tp,
1515 struct key_vector *l, struct fib_alias *old)
1516 {
1517 /* record the location of the previous list_info entry */
1518 struct hlist_node **pprev = old->fa_list.pprev;
1519 struct fib_alias *fa = hlist_entry(pprev, typeof(*fa), fa_list.next);
1520
1521 /* remove the fib_alias from the list */
1522 hlist_del_rcu(&old->fa_list);
1523
1524 /* if we emptied the list this leaf will be freed and we can sort
1525 * out parent suffix lengths as a part of trie_rebalance
1526 */
1527 if (hlist_empty(&l->leaf)) {
1528 if (tp->slen == l->slen)
1529 node_pull_suffix(tp, tp->pos);
1530 put_child_root(tp, l->key, NULL);
1531 node_free(l);
1532 trie_rebalance(t, tp);
1533 return;
1534 }
1535
1536 /* only access fa if it is pointing at the last valid hlist_node */
1537 if (*pprev)
1538 return;
1539
1540 /* update the trie with the latest suffix length */
1541 l->slen = fa->fa_slen;
1542 node_pull_suffix(tp, fa->fa_slen);
1543 }
1544
1545 /* Caller must hold RTNL. */
1546 int fib_table_delete(struct net *net, struct fib_table *tb,
1547 struct fib_config *cfg, struct netlink_ext_ack *extack)
1548 {
1549 struct trie *t = (struct trie *) tb->tb_data;
1550 struct fib_alias *fa, *fa_to_delete;
1551 struct key_vector *l, *tp;
1552 u8 plen = cfg->fc_dst_len;
1553 u8 slen = KEYLENGTH - plen;
1554 u8 tos = cfg->fc_tos;
1555 u32 key;
1556
1557 key = ntohl(cfg->fc_dst);
1558
1559 if (!fib_valid_key_len(key, plen, extack))
1560 return -EINVAL;
1561
1562 l = fib_find_node(t, &tp, key);
1563 if (!l)
1564 return -ESRCH;
1565
1566 fa = fib_find_alias(&l->leaf, slen, tos, 0, tb->tb_id);
1567 if (!fa)
1568 return -ESRCH;
1569
1570 pr_debug("Deleting %08x/%d tos=%d t=%p\n", key, plen, tos, t);
1571
1572 fa_to_delete = NULL;
1573 hlist_for_each_entry_from(fa, fa_list) {
1574 struct fib_info *fi = fa->fa_info;
1575
1576 if ((fa->fa_slen != slen) ||
1577 (fa->tb_id != tb->tb_id) ||
1578 (fa->fa_tos != tos))
1579 break;
1580
1581 if ((!cfg->fc_type || fa->fa_type == cfg->fc_type) &&
1582 (cfg->fc_scope == RT_SCOPE_NOWHERE ||
1583 fa->fa_info->fib_scope == cfg->fc_scope) &&
1584 (!cfg->fc_prefsrc ||
1585 fi->fib_prefsrc == cfg->fc_prefsrc) &&
1586 (!cfg->fc_protocol ||
1587 fi->fib_protocol == cfg->fc_protocol) &&
1588 fib_nh_match(cfg, fi, extack) == 0 &&
1589 fib_metrics_match(cfg, fi)) {
1590 fa_to_delete = fa;
1591 break;
1592 }
1593 }
1594
1595 if (!fa_to_delete)
1596 return -ESRCH;
1597
1598 call_fib_entry_notifiers(net, FIB_EVENT_ENTRY_DEL, key, plen,
1599 fa_to_delete, extack);
1600 rtmsg_fib(RTM_DELROUTE, htonl(key), fa_to_delete, plen, tb->tb_id,
1601 &cfg->fc_nlinfo, 0);
1602
1603 if (!plen)
1604 tb->tb_num_default--;
1605
1606 fib_remove_alias(t, tp, l, fa_to_delete);
1607
1608 if (fa_to_delete->fa_state & FA_S_ACCESSED)
1609 rt_cache_flush(cfg->fc_nlinfo.nl_net);
1610
1611 fib_release_info(fa_to_delete->fa_info);
1612 alias_free_mem_rcu(fa_to_delete);
1613 return 0;
1614 }
1615
1616 /* Scan for the next leaf starting at the provided key value */
1617 static struct key_vector *leaf_walk_rcu(struct key_vector **tn, t_key key)
1618 {
1619 struct key_vector *pn, *n = *tn;
1620 unsigned long cindex;
1621
1622 /* this loop is meant to try and find the key in the trie */
1623 do {
1624 /* record parent and next child index */
1625 pn = n;
1626 cindex = (key > pn->key) ? get_index(key, pn) : 0;
1627
1628 if (cindex >> pn->bits)
1629 break;
1630
1631 /* descend into the next child */
1632 n = get_child_rcu(pn, cindex++);
1633 if (!n)
1634 break;
1635
1636 /* guarantee forward progress on the keys */
1637 if (IS_LEAF(n) && (n->key >= key))
1638 goto found;
1639 } while (IS_TNODE(n));
1640
1641 /* this loop will search for the next leaf with a greater key */
1642 while (!IS_TRIE(pn)) {
1643 /* if we exhausted the parent node we will need to climb */
1644 if (cindex >= (1ul << pn->bits)) {
1645 t_key pkey = pn->key;
1646
1647 pn = node_parent_rcu(pn);
1648 cindex = get_index(pkey, pn) + 1;
1649 continue;
1650 }
1651
1652 /* grab the next available node */
1653 n = get_child_rcu(pn, cindex++);
1654 if (!n)
1655 continue;
1656
1657 /* no need to compare keys since we bumped the index */
1658 if (IS_LEAF(n))
1659 goto found;
1660
1661 /* Rescan start scanning in new node */
1662 pn = n;
1663 cindex = 0;
1664 }
1665
1666 *tn = pn;
1667 return NULL; /* Root of trie */
1668 found:
1669 /* if we are at the limit for keys just return NULL for the tnode */
1670 *tn = pn;
1671 return n;
1672 }
1673
1674 static void fib_trie_free(struct fib_table *tb)
1675 {
1676 struct trie *t = (struct trie *)tb->tb_data;
1677 struct key_vector *pn = t->kv;
1678 unsigned long cindex = 1;
1679 struct hlist_node *tmp;
1680 struct fib_alias *fa;
1681
1682 /* walk trie in reverse order and free everything */
1683 for (;;) {
1684 struct key_vector *n;
1685
1686 if (!(cindex--)) {
1687 t_key pkey = pn->key;
1688
1689 if (IS_TRIE(pn))
1690 break;
1691
1692 n = pn;
1693 pn = node_parent(pn);
1694
1695 /* drop emptied tnode */
1696 put_child_root(pn, n->key, NULL);
1697 node_free(n);
1698
1699 cindex = get_index(pkey, pn);
1700
1701 continue;
1702 }
1703
1704 /* grab the next available node */
1705 n = get_child(pn, cindex);
1706 if (!n)
1707 continue;
1708
1709 if (IS_TNODE(n)) {
1710 /* record pn and cindex for leaf walking */
1711 pn = n;
1712 cindex = 1ul << n->bits;
1713
1714 continue;
1715 }
1716
1717 hlist_for_each_entry_safe(fa, tmp, &n->leaf, fa_list) {
1718 hlist_del_rcu(&fa->fa_list);
1719 alias_free_mem_rcu(fa);
1720 }
1721
1722 put_child_root(pn, n->key, NULL);
1723 node_free(n);
1724 }
1725
1726 #ifdef CONFIG_IP_FIB_TRIE_STATS
1727 free_percpu(t->stats);
1728 #endif
1729 kfree(tb);
1730 }
1731
1732 struct fib_table *fib_trie_unmerge(struct fib_table *oldtb)
1733 {
1734 struct trie *ot = (struct trie *)oldtb->tb_data;
1735 struct key_vector *l, *tp = ot->kv;
1736 struct fib_table *local_tb;
1737 struct fib_alias *fa;
1738 struct trie *lt;
1739 t_key key = 0;
1740
1741 if (oldtb->tb_data == oldtb->__data)
1742 return oldtb;
1743
1744 local_tb = fib_trie_table(RT_TABLE_LOCAL, NULL);
1745 if (!local_tb)
1746 return NULL;
1747
1748 lt = (struct trie *)local_tb->tb_data;
1749
1750 while ((l = leaf_walk_rcu(&tp, key)) != NULL) {
1751 struct key_vector *local_l = NULL, *local_tp;
1752
1753 hlist_for_each_entry_rcu(fa, &l->leaf, fa_list) {
1754 struct fib_alias *new_fa;
1755
1756 if (local_tb->tb_id != fa->tb_id)
1757 continue;
1758
1759 /* clone fa for new local table */
1760 new_fa = kmem_cache_alloc(fn_alias_kmem, GFP_KERNEL);
1761 if (!new_fa)
1762 goto out;
1763
1764 memcpy(new_fa, fa, sizeof(*fa));
1765
1766 /* insert clone into table */
1767 if (!local_l)
1768 local_l = fib_find_node(lt, &local_tp, l->key);
1769
1770 if (fib_insert_alias(lt, local_tp, local_l, new_fa,
1771 NULL, l->key)) {
1772 kmem_cache_free(fn_alias_kmem, new_fa);
1773 goto out;
1774 }
1775 }
1776
1777 /* stop loop if key wrapped back to 0 */
1778 key = l->key + 1;
1779 if (key < l->key)
1780 break;
1781 }
1782
1783 return local_tb;
1784 out:
1785 fib_trie_free(local_tb);
1786
1787 return NULL;
1788 }
1789
1790 /* Caller must hold RTNL */
1791 void fib_table_flush_external(struct fib_table *tb)
1792 {
1793 struct trie *t = (struct trie *)tb->tb_data;
1794 struct key_vector *pn = t->kv;
1795 unsigned long cindex = 1;
1796 struct hlist_node *tmp;
1797 struct fib_alias *fa;
1798
1799 /* walk trie in reverse order */
1800 for (;;) {
1801 unsigned char slen = 0;
1802 struct key_vector *n;
1803
1804 if (!(cindex--)) {
1805 t_key pkey = pn->key;
1806
1807 /* cannot resize the trie vector */
1808 if (IS_TRIE(pn))
1809 break;
1810
1811 /* update the suffix to address pulled leaves */
1812 if (pn->slen > pn->pos)
1813 update_suffix(pn);
1814
1815 /* resize completed node */
1816 pn = resize(t, pn);
1817 cindex = get_index(pkey, pn);
1818
1819 continue;
1820 }
1821
1822 /* grab the next available node */
1823 n = get_child(pn, cindex);
1824 if (!n)
1825 continue;
1826
1827 if (IS_TNODE(n)) {
1828 /* record pn and cindex for leaf walking */
1829 pn = n;
1830 cindex = 1ul << n->bits;
1831
1832 continue;
1833 }
1834
1835 hlist_for_each_entry_safe(fa, tmp, &n->leaf, fa_list) {
1836 /* if alias was cloned to local then we just
1837 * need to remove the local copy from main
1838 */
1839 if (tb->tb_id != fa->tb_id) {
1840 hlist_del_rcu(&fa->fa_list);
1841 alias_free_mem_rcu(fa);
1842 continue;
1843 }
1844
1845 /* record local slen */
1846 slen = fa->fa_slen;
1847 }
1848
1849 /* update leaf slen */
1850 n->slen = slen;
1851
1852 if (hlist_empty(&n->leaf)) {
1853 put_child_root(pn, n->key, NULL);
1854 node_free(n);
1855 }
1856 }
1857 }
1858
1859 /* Caller must hold RTNL. */
1860 int fib_table_flush(struct net *net, struct fib_table *tb, bool flush_all)
1861 {
1862 struct trie *t = (struct trie *)tb->tb_data;
1863 struct key_vector *pn = t->kv;
1864 unsigned long cindex = 1;
1865 struct hlist_node *tmp;
1866 struct fib_alias *fa;
1867 int found = 0;
1868
1869 /* walk trie in reverse order */
1870 for (;;) {
1871 unsigned char slen = 0;
1872 struct key_vector *n;
1873
1874 if (!(cindex--)) {
1875 t_key pkey = pn->key;
1876
1877 /* cannot resize the trie vector */
1878 if (IS_TRIE(pn))
1879 break;
1880
1881 /* update the suffix to address pulled leaves */
1882 if (pn->slen > pn->pos)
1883 update_suffix(pn);
1884
1885 /* resize completed node */
1886 pn = resize(t, pn);
1887 cindex = get_index(pkey, pn);
1888
1889 continue;
1890 }
1891
1892 /* grab the next available node */
1893 n = get_child(pn, cindex);
1894 if (!n)
1895 continue;
1896
1897 if (IS_TNODE(n)) {
1898 /* record pn and cindex for leaf walking */
1899 pn = n;
1900 cindex = 1ul << n->bits;
1901
1902 continue;
1903 }
1904
1905 hlist_for_each_entry_safe(fa, tmp, &n->leaf, fa_list) {
1906 struct fib_info *fi = fa->fa_info;
1907
1908 if (!fi || tb->tb_id != fa->tb_id ||
1909 (!(fi->fib_flags & RTNH_F_DEAD) &&
1910 !fib_props[fa->fa_type].error)) {
1911 slen = fa->fa_slen;
1912 continue;
1913 }
1914
1915 /* Do not flush error routes if network namespace is
1916 * not being dismantled
1917 */
1918 if (!flush_all && fib_props[fa->fa_type].error) {
1919 slen = fa->fa_slen;
1920 continue;
1921 }
1922
1923 call_fib_entry_notifiers(net, FIB_EVENT_ENTRY_DEL,
1924 n->key,
1925 KEYLENGTH - fa->fa_slen, fa,
1926 NULL);
1927 hlist_del_rcu(&fa->fa_list);
1928 fib_release_info(fa->fa_info);
1929 alias_free_mem_rcu(fa);
1930 found++;
1931 }
1932
1933 /* update leaf slen */
1934 n->slen = slen;
1935
1936 if (hlist_empty(&n->leaf)) {
1937 put_child_root(pn, n->key, NULL);
1938 node_free(n);
1939 }
1940 }
1941
1942 pr_debug("trie_flush found=%d\n", found);
1943 return found;
1944 }
1945
1946 static void fib_leaf_notify(struct net *net, struct key_vector *l,
1947 struct fib_table *tb, struct notifier_block *nb)
1948 {
1949 struct fib_alias *fa;
1950
1951 hlist_for_each_entry_rcu(fa, &l->leaf, fa_list) {
1952 struct fib_info *fi = fa->fa_info;
1953
1954 if (!fi)
1955 continue;
1956
1957 /* local and main table can share the same trie,
1958 * so don't notify twice for the same entry.
1959 */
1960 if (tb->tb_id != fa->tb_id)
1961 continue;
1962
1963 call_fib_entry_notifier(nb, net, FIB_EVENT_ENTRY_ADD, l->key,
1964 KEYLENGTH - fa->fa_slen, fa);
1965 }
1966 }
1967
1968 static void fib_table_notify(struct net *net, struct fib_table *tb,
1969 struct notifier_block *nb)
1970 {
1971 struct trie *t = (struct trie *)tb->tb_data;
1972 struct key_vector *l, *tp = t->kv;
1973 t_key key = 0;
1974
1975 while ((l = leaf_walk_rcu(&tp, key)) != NULL) {
1976 fib_leaf_notify(net, l, tb, nb);
1977
1978 key = l->key + 1;
1979 /* stop in case of wrap around */
1980 if (key < l->key)
1981 break;
1982 }
1983 }
1984
1985 void fib_notify(struct net *net, struct notifier_block *nb)
1986 {
1987 unsigned int h;
1988
1989 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
1990 struct hlist_head *head = &net->ipv4.fib_table_hash[h];
1991 struct fib_table *tb;
1992
1993 hlist_for_each_entry_rcu(tb, head, tb_hlist)
1994 fib_table_notify(net, tb, nb);
1995 }
1996 }
1997
1998 static void __trie_free_rcu(struct rcu_head *head)
1999 {
2000 struct fib_table *tb = container_of(head, struct fib_table, rcu);
2001 #ifdef CONFIG_IP_FIB_TRIE_STATS
2002 struct trie *t = (struct trie *)tb->tb_data;
2003
2004 if (tb->tb_data == tb->__data)
2005 free_percpu(t->stats);
2006 #endif /* CONFIG_IP_FIB_TRIE_STATS */
2007 kfree(tb);
2008 }
2009
2010 void fib_free_table(struct fib_table *tb)
2011 {
2012 call_rcu(&tb->rcu, __trie_free_rcu);
2013 }
2014
2015 static int fn_trie_dump_leaf(struct key_vector *l, struct fib_table *tb,
2016 struct sk_buff *skb, struct netlink_callback *cb,
2017 struct fib_dump_filter *filter)
2018 {
2019 unsigned int flags = NLM_F_MULTI;
2020 __be32 xkey = htonl(l->key);
2021 struct fib_alias *fa;
2022 int i, s_i;
2023
2024 if (filter->filter_set)
2025 flags |= NLM_F_DUMP_FILTERED;
2026
2027 s_i = cb->args[4];
2028 i = 0;
2029
2030 /* rcu_read_lock is hold by caller */
2031 hlist_for_each_entry_rcu(fa, &l->leaf, fa_list) {
2032 int err;
2033
2034 if (i < s_i)
2035 goto next;
2036
2037 if (tb->tb_id != fa->tb_id)
2038 goto next;
2039
2040 if (filter->filter_set) {
2041 if (filter->rt_type && fa->fa_type != filter->rt_type)
2042 goto next;
2043
2044 if ((filter->protocol &&
2045 fa->fa_info->fib_protocol != filter->protocol))
2046 goto next;
2047
2048 if (filter->dev &&
2049 !fib_info_nh_uses_dev(fa->fa_info, filter->dev))
2050 goto next;
2051 }
2052
2053 err = fib_dump_info(skb, NETLINK_CB(cb->skb).portid,
2054 cb->nlh->nlmsg_seq, RTM_NEWROUTE,
2055 tb->tb_id, fa->fa_type,
2056 xkey, KEYLENGTH - fa->fa_slen,
2057 fa->fa_tos, fa->fa_info, flags);
2058 if (err < 0) {
2059 cb->args[4] = i;
2060 return err;
2061 }
2062 next:
2063 i++;
2064 }
2065
2066 cb->args[4] = i;
2067 return skb->len;
2068 }
2069
2070 /* rcu_read_lock needs to be hold by caller from readside */
2071 int fib_table_dump(struct fib_table *tb, struct sk_buff *skb,
2072 struct netlink_callback *cb, struct fib_dump_filter *filter)
2073 {
2074 struct trie *t = (struct trie *)tb->tb_data;
2075 struct key_vector *l, *tp = t->kv;
2076 /* Dump starting at last key.
2077 * Note: 0.0.0.0/0 (ie default) is first key.
2078 */
2079 int count = cb->args[2];
2080 t_key key = cb->args[3];
2081
2082 while ((l = leaf_walk_rcu(&tp, key)) != NULL) {
2083 int err;
2084
2085 err = fn_trie_dump_leaf(l, tb, skb, cb, filter);
2086 if (err < 0) {
2087 cb->args[3] = key;
2088 cb->args[2] = count;
2089 return err;
2090 }
2091
2092 ++count;
2093 key = l->key + 1;
2094
2095 memset(&cb->args[4], 0,
2096 sizeof(cb->args) - 4*sizeof(cb->args[0]));
2097
2098 /* stop loop if key wrapped back to 0 */
2099 if (key < l->key)
2100 break;
2101 }
2102
2103 cb->args[3] = key;
2104 cb->args[2] = count;
2105
2106 return skb->len;
2107 }
2108
2109 void __init fib_trie_init(void)
2110 {
2111 fn_alias_kmem = kmem_cache_create("ip_fib_alias",
2112 sizeof(struct fib_alias),
2113 0, SLAB_PANIC, NULL);
2114
2115 trie_leaf_kmem = kmem_cache_create("ip_fib_trie",
2116 LEAF_SIZE,
2117 0, SLAB_PANIC, NULL);
2118 }
2119
2120 struct fib_table *fib_trie_table(u32 id, struct fib_table *alias)
2121 {
2122 struct fib_table *tb;
2123 struct trie *t;
2124 size_t sz = sizeof(*tb);
2125
2126 if (!alias)
2127 sz += sizeof(struct trie);
2128
2129 tb = kzalloc(sz, GFP_KERNEL);
2130 if (!tb)
2131 return NULL;
2132
2133 tb->tb_id = id;
2134 tb->tb_num_default = 0;
2135 tb->tb_data = (alias ? alias->__data : tb->__data);
2136
2137 if (alias)
2138 return tb;
2139
2140 t = (struct trie *) tb->tb_data;
2141 t->kv[0].pos = KEYLENGTH;
2142 t->kv[0].slen = KEYLENGTH;
2143 #ifdef CONFIG_IP_FIB_TRIE_STATS
2144 t->stats = alloc_percpu(struct trie_use_stats);
2145 if (!t->stats) {
2146 kfree(tb);
2147 tb = NULL;
2148 }
2149 #endif
2150
2151 return tb;
2152 }
2153
2154 #ifdef CONFIG_PROC_FS
2155 /* Depth first Trie walk iterator */
2156 struct fib_trie_iter {
2157 struct seq_net_private p;
2158 struct fib_table *tb;
2159 struct key_vector *tnode;
2160 unsigned int index;
2161 unsigned int depth;
2162 };
2163
2164 static struct key_vector *fib_trie_get_next(struct fib_trie_iter *iter)
2165 {
2166 unsigned long cindex = iter->index;
2167 struct key_vector *pn = iter->tnode;
2168 t_key pkey;
2169
2170 pr_debug("get_next iter={node=%p index=%d depth=%d}\n",
2171 iter->tnode, iter->index, iter->depth);
2172
2173 while (!IS_TRIE(pn)) {
2174 while (cindex < child_length(pn)) {
2175 struct key_vector *n = get_child_rcu(pn, cindex++);
2176
2177 if (!n)
2178 continue;
2179
2180 if (IS_LEAF(n)) {
2181 iter->tnode = pn;
2182 iter->index = cindex;
2183 } else {
2184 /* push down one level */
2185 iter->tnode = n;
2186 iter->index = 0;
2187 ++iter->depth;
2188 }
2189
2190 return n;
2191 }
2192
2193 /* Current node exhausted, pop back up */
2194 pkey = pn->key;
2195 pn = node_parent_rcu(pn);
2196 cindex = get_index(pkey, pn) + 1;
2197 --iter->depth;
2198 }
2199
2200 /* record root node so further searches know we are done */
2201 iter->tnode = pn;
2202 iter->index = 0;
2203
2204 return NULL;
2205 }
2206
2207 static struct key_vector *fib_trie_get_first(struct fib_trie_iter *iter,
2208 struct trie *t)
2209 {
2210 struct key_vector *n, *pn;
2211
2212 if (!t)
2213 return NULL;
2214
2215 pn = t->kv;
2216 n = rcu_dereference(pn->tnode[0]);
2217 if (!n)
2218 return NULL;
2219
2220 if (IS_TNODE(n)) {
2221 iter->tnode = n;
2222 iter->index = 0;
2223 iter->depth = 1;
2224 } else {
2225 iter->tnode = pn;
2226 iter->index = 0;
2227 iter->depth = 0;
2228 }
2229
2230 return n;
2231 }
2232
2233 static void trie_collect_stats(struct trie *t, struct trie_stat *s)
2234 {
2235 struct key_vector *n;
2236 struct fib_trie_iter iter;
2237
2238 memset(s, 0, sizeof(*s));
2239
2240 rcu_read_lock();
2241 for (n = fib_trie_get_first(&iter, t); n; n = fib_trie_get_next(&iter)) {
2242 if (IS_LEAF(n)) {
2243 struct fib_alias *fa;
2244
2245 s->leaves++;
2246 s->totdepth += iter.depth;
2247 if (iter.depth > s->maxdepth)
2248 s->maxdepth = iter.depth;
2249
2250 hlist_for_each_entry_rcu(fa, &n->leaf, fa_list)
2251 ++s->prefixes;
2252 } else {
2253 s->tnodes++;
2254 if (n->bits < MAX_STAT_DEPTH)
2255 s->nodesizes[n->bits]++;
2256 s->nullpointers += tn_info(n)->empty_children;
2257 }
2258 }
2259 rcu_read_unlock();
2260 }
2261
2262 /*
2263 * This outputs /proc/net/fib_triestats
2264 */
2265 static void trie_show_stats(struct seq_file *seq, struct trie_stat *stat)
2266 {
2267 unsigned int i, max, pointers, bytes, avdepth;
2268
2269 if (stat->leaves)
2270 avdepth = stat->totdepth*100 / stat->leaves;
2271 else
2272 avdepth = 0;
2273
2274 seq_printf(seq, "\tAver depth: %u.%02d\n",
2275 avdepth / 100, avdepth % 100);
2276 seq_printf(seq, "\tMax depth: %u\n", stat->maxdepth);
2277
2278 seq_printf(seq, "\tLeaves: %u\n", stat->leaves);
2279 bytes = LEAF_SIZE * stat->leaves;
2280
2281 seq_printf(seq, "\tPrefixes: %u\n", stat->prefixes);
2282 bytes += sizeof(struct fib_alias) * stat->prefixes;
2283
2284 seq_printf(seq, "\tInternal nodes: %u\n\t", stat->tnodes);
2285 bytes += TNODE_SIZE(0) * stat->tnodes;
2286
2287 max = MAX_STAT_DEPTH;
2288 while (max > 0 && stat->nodesizes[max-1] == 0)
2289 max--;
2290
2291 pointers = 0;
2292 for (i = 1; i < max; i++)
2293 if (stat->nodesizes[i] != 0) {
2294 seq_printf(seq, " %u: %u", i, stat->nodesizes[i]);
2295 pointers += (1<<i) * stat->nodesizes[i];
2296 }
2297 seq_putc(seq, '\n');
2298 seq_printf(seq, "\tPointers: %u\n", pointers);
2299
2300 bytes += sizeof(struct key_vector *) * pointers;
2301 seq_printf(seq, "Null ptrs: %u\n", stat->nullpointers);
2302 seq_printf(seq, "Total size: %u kB\n", (bytes + 1023) / 1024);
2303 }
2304
2305 #ifdef CONFIG_IP_FIB_TRIE_STATS
2306 static void trie_show_usage(struct seq_file *seq,
2307 const struct trie_use_stats __percpu *stats)
2308 {
2309 struct trie_use_stats s = { 0 };
2310 int cpu;
2311
2312 /* loop through all of the CPUs and gather up the stats */
2313 for_each_possible_cpu(cpu) {
2314 const struct trie_use_stats *pcpu = per_cpu_ptr(stats, cpu);
2315
2316 s.gets += pcpu->gets;
2317 s.backtrack += pcpu->backtrack;
2318 s.semantic_match_passed += pcpu->semantic_match_passed;
2319 s.semantic_match_miss += pcpu->semantic_match_miss;
2320 s.null_node_hit += pcpu->null_node_hit;
2321 s.resize_node_skipped += pcpu->resize_node_skipped;
2322 }
2323
2324 seq_printf(seq, "\nCounters:\n---------\n");
2325 seq_printf(seq, "gets = %u\n", s.gets);
2326 seq_printf(seq, "backtracks = %u\n", s.backtrack);
2327 seq_printf(seq, "semantic match passed = %u\n",
2328 s.semantic_match_passed);
2329 seq_printf(seq, "semantic match miss = %u\n", s.semantic_match_miss);
2330 seq_printf(seq, "null node hit= %u\n", s.null_node_hit);
2331 seq_printf(seq, "skipped node resize = %u\n\n", s.resize_node_skipped);
2332 }
2333 #endif /* CONFIG_IP_FIB_TRIE_STATS */
2334
2335 static void fib_table_print(struct seq_file *seq, struct fib_table *tb)
2336 {
2337 if (tb->tb_id == RT_TABLE_LOCAL)
2338 seq_puts(seq, "Local:\n");
2339 else if (tb->tb_id == RT_TABLE_MAIN)
2340 seq_puts(seq, "Main:\n");
2341 else
2342 seq_printf(seq, "Id %d:\n", tb->tb_id);
2343 }
2344
2345
2346 static int fib_triestat_seq_show(struct seq_file *seq, void *v)
2347 {
2348 struct net *net = (struct net *)seq->private;
2349 unsigned int h;
2350
2351 seq_printf(seq,
2352 "Basic info: size of leaf:"
2353 " %zd bytes, size of tnode: %zd bytes.\n",
2354 LEAF_SIZE, TNODE_SIZE(0));
2355
2356 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
2357 struct hlist_head *head = &net->ipv4.fib_table_hash[h];
2358 struct fib_table *tb;
2359
2360 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
2361 struct trie *t = (struct trie *) tb->tb_data;
2362 struct trie_stat stat;
2363
2364 if (!t)
2365 continue;
2366
2367 fib_table_print(seq, tb);
2368
2369 trie_collect_stats(t, &stat);
2370 trie_show_stats(seq, &stat);
2371 #ifdef CONFIG_IP_FIB_TRIE_STATS
2372 trie_show_usage(seq, t->stats);
2373 #endif
2374 }
2375 }
2376
2377 return 0;
2378 }
2379
2380 static struct key_vector *fib_trie_get_idx(struct seq_file *seq, loff_t pos)
2381 {
2382 struct fib_trie_iter *iter = seq->private;
2383 struct net *net = seq_file_net(seq);
2384 loff_t idx = 0;
2385 unsigned int h;
2386
2387 for (h = 0; h < FIB_TABLE_HASHSZ; h++) {
2388 struct hlist_head *head = &net->ipv4.fib_table_hash[h];
2389 struct fib_table *tb;
2390
2391 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
2392 struct key_vector *n;
2393
2394 for (n = fib_trie_get_first(iter,
2395 (struct trie *) tb->tb_data);
2396 n; n = fib_trie_get_next(iter))
2397 if (pos == idx++) {
2398 iter->tb = tb;
2399 return n;
2400 }
2401 }
2402 }
2403
2404 return NULL;
2405 }
2406
2407 static void *fib_trie_seq_start(struct seq_file *seq, loff_t *pos)
2408 __acquires(RCU)
2409 {
2410 rcu_read_lock();
2411 return fib_trie_get_idx(seq, *pos);
2412 }
2413
2414 static void *fib_trie_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2415 {
2416 struct fib_trie_iter *iter = seq->private;
2417 struct net *net = seq_file_net(seq);
2418 struct fib_table *tb = iter->tb;
2419 struct hlist_node *tb_node;
2420 unsigned int h;
2421 struct key_vector *n;
2422
2423 ++*pos;
2424 /* next node in same table */
2425 n = fib_trie_get_next(iter);
2426 if (n)
2427 return n;
2428
2429 /* walk rest of this hash chain */
2430 h = tb->tb_id & (FIB_TABLE_HASHSZ - 1);
2431 while ((tb_node = rcu_dereference(hlist_next_rcu(&tb->tb_hlist)))) {
2432 tb = hlist_entry(tb_node, struct fib_table, tb_hlist);
2433 n = fib_trie_get_first(iter, (struct trie *) tb->tb_data);
2434 if (n)
2435 goto found;
2436 }
2437
2438 /* new hash chain */
2439 while (++h < FIB_TABLE_HASHSZ) {
2440 struct hlist_head *head = &net->ipv4.fib_table_hash[h];
2441 hlist_for_each_entry_rcu(tb, head, tb_hlist) {
2442 n = fib_trie_get_first(iter, (struct trie *) tb->tb_data);
2443 if (n)
2444 goto found;
2445 }
2446 }
2447 return NULL;
2448
2449 found:
2450 iter->tb = tb;
2451 return n;
2452 }
2453
2454 static void fib_trie_seq_stop(struct seq_file *seq, void *v)
2455 __releases(RCU)
2456 {
2457 rcu_read_unlock();
2458 }
2459
2460 static void seq_indent(struct seq_file *seq, int n)
2461 {
2462 while (n-- > 0)
2463 seq_puts(seq, " ");
2464 }
2465
2466 static inline const char *rtn_scope(char *buf, size_t len, enum rt_scope_t s)
2467 {
2468 switch (s) {
2469 case RT_SCOPE_UNIVERSE: return "universe";
2470 case RT_SCOPE_SITE: return "site";
2471 case RT_SCOPE_LINK: return "link";
2472 case RT_SCOPE_HOST: return "host";
2473 case RT_SCOPE_NOWHERE: return "nowhere";
2474 default:
2475 snprintf(buf, len, "scope=%d", s);
2476 return buf;
2477 }
2478 }
2479
2480 static const char *const rtn_type_names[__RTN_MAX] = {
2481 [RTN_UNSPEC] = "UNSPEC",
2482 [RTN_UNICAST] = "UNICAST",
2483 [RTN_LOCAL] = "LOCAL",
2484 [RTN_BROADCAST] = "BROADCAST",
2485 [RTN_ANYCAST] = "ANYCAST",
2486 [RTN_MULTICAST] = "MULTICAST",
2487 [RTN_BLACKHOLE] = "BLACKHOLE",
2488 [RTN_UNREACHABLE] = "UNREACHABLE",
2489 [RTN_PROHIBIT] = "PROHIBIT",
2490 [RTN_THROW] = "THROW",
2491 [RTN_NAT] = "NAT",
2492 [RTN_XRESOLVE] = "XRESOLVE",
2493 };
2494
2495 static inline const char *rtn_type(char *buf, size_t len, unsigned int t)
2496 {
2497 if (t < __RTN_MAX && rtn_type_names[t])
2498 return rtn_type_names[t];
2499 snprintf(buf, len, "type %u", t);
2500 return buf;
2501 }
2502
2503 /* Pretty print the trie */
2504 static int fib_trie_seq_show(struct seq_file *seq, void *v)
2505 {
2506 const struct fib_trie_iter *iter = seq->private;
2507 struct key_vector *n = v;
2508
2509 if (IS_TRIE(node_parent_rcu(n)))
2510 fib_table_print(seq, iter->tb);
2511
2512 if (IS_TNODE(n)) {
2513 __be32 prf = htonl(n->key);
2514
2515 seq_indent(seq, iter->depth-1);
2516 seq_printf(seq, " +-- %pI4/%zu %u %u %u\n",
2517 &prf, KEYLENGTH - n->pos - n->bits, n->bits,
2518 tn_info(n)->full_children,
2519 tn_info(n)->empty_children);
2520 } else {
2521 __be32 val = htonl(n->key);
2522 struct fib_alias *fa;
2523
2524 seq_indent(seq, iter->depth);
2525 seq_printf(seq, " |-- %pI4\n", &val);
2526
2527 hlist_for_each_entry_rcu(fa, &n->leaf, fa_list) {
2528 char buf1[32], buf2[32];
2529
2530 seq_indent(seq, iter->depth + 1);
2531 seq_printf(seq, " /%zu %s %s",
2532 KEYLENGTH - fa->fa_slen,
2533 rtn_scope(buf1, sizeof(buf1),
2534 fa->fa_info->fib_scope),
2535 rtn_type(buf2, sizeof(buf2),
2536 fa->fa_type));
2537 if (fa->fa_tos)
2538 seq_printf(seq, " tos=%d", fa->fa_tos);
2539 seq_putc(seq, '\n');
2540 }
2541 }
2542
2543 return 0;
2544 }
2545
2546 static const struct seq_operations fib_trie_seq_ops = {
2547 .start = fib_trie_seq_start,
2548 .next = fib_trie_seq_next,
2549 .stop = fib_trie_seq_stop,
2550 .show = fib_trie_seq_show,
2551 };
2552
2553 struct fib_route_iter {
2554 struct seq_net_private p;
2555 struct fib_table *main_tb;
2556 struct key_vector *tnode;
2557 loff_t pos;
2558 t_key key;
2559 };
2560
2561 static struct key_vector *fib_route_get_idx(struct fib_route_iter *iter,
2562 loff_t pos)
2563 {
2564 struct key_vector *l, **tp = &iter->tnode;
2565 t_key key;
2566
2567 /* use cached location of previously found key */
2568 if (iter->pos > 0 && pos >= iter->pos) {
2569 key = iter->key;
2570 } else {
2571 iter->pos = 1;
2572 key = 0;
2573 }
2574
2575 pos -= iter->pos;
2576
2577 while ((l = leaf_walk_rcu(tp, key)) && (pos-- > 0)) {
2578 key = l->key + 1;
2579 iter->pos++;
2580 l = NULL;
2581
2582 /* handle unlikely case of a key wrap */
2583 if (!key)
2584 break;
2585 }
2586
2587 if (l)
2588 iter->key = l->key; /* remember it */
2589 else
2590 iter->pos = 0; /* forget it */
2591
2592 return l;
2593 }
2594
2595 static void *fib_route_seq_start(struct seq_file *seq, loff_t *pos)
2596 __acquires(RCU)
2597 {
2598 struct fib_route_iter *iter = seq->private;
2599 struct fib_table *tb;
2600 struct trie *t;
2601
2602 rcu_read_lock();
2603
2604 tb = fib_get_table(seq_file_net(seq), RT_TABLE_MAIN);
2605 if (!tb)
2606 return NULL;
2607
2608 iter->main_tb = tb;
2609 t = (struct trie *)tb->tb_data;
2610 iter->tnode = t->kv;
2611
2612 if (*pos != 0)
2613 return fib_route_get_idx(iter, *pos);
2614
2615 iter->pos = 0;
2616 iter->key = KEY_MAX;
2617
2618 return SEQ_START_TOKEN;
2619 }
2620
2621 static void *fib_route_seq_next(struct seq_file *seq, void *v, loff_t *pos)
2622 {
2623 struct fib_route_iter *iter = seq->private;
2624 struct key_vector *l = NULL;
2625 t_key key = iter->key + 1;
2626
2627 ++*pos;
2628
2629 /* only allow key of 0 for start of sequence */
2630 if ((v == SEQ_START_TOKEN) || key)
2631 l = leaf_walk_rcu(&iter->tnode, key);
2632
2633 if (l) {
2634 iter->key = l->key;
2635 iter->pos++;
2636 } else {
2637 iter->pos = 0;
2638 }
2639
2640 return l;
2641 }
2642
2643 static void fib_route_seq_stop(struct seq_file *seq, void *v)
2644 __releases(RCU)
2645 {
2646 rcu_read_unlock();
2647 }
2648
2649 static unsigned int fib_flag_trans(int type, __be32 mask, const struct fib_info *fi)
2650 {
2651 unsigned int flags = 0;
2652
2653 if (type == RTN_UNREACHABLE || type == RTN_PROHIBIT)
2654 flags = RTF_REJECT;
2655 if (fi && fi->fib_nh->fib_nh_gw4)
2656 flags |= RTF_GATEWAY;
2657 if (mask == htonl(0xFFFFFFFF))
2658 flags |= RTF_HOST;
2659 flags |= RTF_UP;
2660 return flags;
2661 }
2662
2663 /*
2664 * This outputs /proc/net/route.
2665 * The format of the file is not supposed to be changed
2666 * and needs to be same as fib_hash output to avoid breaking
2667 * legacy utilities
2668 */
2669 static int fib_route_seq_show(struct seq_file *seq, void *v)
2670 {
2671 struct fib_route_iter *iter = seq->private;
2672 struct fib_table *tb = iter->main_tb;
2673 struct fib_alias *fa;
2674 struct key_vector *l = v;
2675 __be32 prefix;
2676
2677 if (v == SEQ_START_TOKEN) {
2678 seq_printf(seq, "%-127s\n", "Iface\tDestination\tGateway "
2679 "\tFlags\tRefCnt\tUse\tMetric\tMask\t\tMTU"
2680 "\tWindow\tIRTT");
2681 return 0;
2682 }
2683
2684 prefix = htonl(l->key);
2685
2686 hlist_for_each_entry_rcu(fa, &l->leaf, fa_list) {
2687 const struct fib_info *fi = fa->fa_info;
2688 __be32 mask = inet_make_mask(KEYLENGTH - fa->fa_slen);
2689 unsigned int flags = fib_flag_trans(fa->fa_type, mask, fi);
2690
2691 if ((fa->fa_type == RTN_BROADCAST) ||
2692 (fa->fa_type == RTN_MULTICAST))
2693 continue;
2694
2695 if (fa->tb_id != tb->tb_id)
2696 continue;
2697
2698 seq_setwidth(seq, 127);
2699
2700 if (fi)
2701 seq_printf(seq,
2702 "%s\t%08X\t%08X\t%04X\t%d\t%u\t"
2703 "%d\t%08X\t%d\t%u\t%u",
2704 fi->fib_dev ? fi->fib_dev->name : "*",
2705 prefix,
2706 fi->fib_nh->fib_nh_gw4, flags, 0, 0,
2707 fi->fib_priority,
2708 mask,
2709 (fi->fib_advmss ?
2710 fi->fib_advmss + 40 : 0),
2711 fi->fib_window,
2712 fi->fib_rtt >> 3);
2713 else
2714 seq_printf(seq,
2715 "*\t%08X\t%08X\t%04X\t%d\t%u\t"
2716 "%d\t%08X\t%d\t%u\t%u",
2717 prefix, 0, flags, 0, 0, 0,
2718 mask, 0, 0, 0);
2719
2720 seq_pad(seq, '\n');
2721 }
2722
2723 return 0;
2724 }
2725
2726 static const struct seq_operations fib_route_seq_ops = {
2727 .start = fib_route_seq_start,
2728 .next = fib_route_seq_next,
2729 .stop = fib_route_seq_stop,
2730 .show = fib_route_seq_show,
2731 };
2732
2733 int __net_init fib_proc_init(struct net *net)
2734 {
2735 if (!proc_create_net("fib_trie", 0444, net->proc_net, &fib_trie_seq_ops,
2736 sizeof(struct fib_trie_iter)))
2737 goto out1;
2738
2739 if (!proc_create_net_single("fib_triestat", 0444, net->proc_net,
2740 fib_triestat_seq_show, NULL))
2741 goto out2;
2742
2743 if (!proc_create_net("route", 0444, net->proc_net, &fib_route_seq_ops,
2744 sizeof(struct fib_route_iter)))
2745 goto out3;
2746
2747 return 0;
2748
2749 out3:
2750 remove_proc_entry("fib_triestat", net->proc_net);
2751 out2:
2752 remove_proc_entry("fib_trie", net->proc_net);
2753 out1:
2754 return -ENOMEM;
2755 }
2756
2757 void __net_exit fib_proc_exit(struct net *net)
2758 {
2759 remove_proc_entry("fib_trie", net->proc_net);
2760 remove_proc_entry("fib_triestat", net->proc_net);
2761 remove_proc_entry("route", net->proc_net);
2762 }
2763
2764 #endif /* CONFIG_PROC_FS */