blob: 2e6ba03c08da58c6d851ab90045a0e29fd7cfdad [file] [log] [blame]
Willy Tarreauc2186022009-10-26 19:48:54 +01001/*
2 * Elastic Binary Trees - generic macros and structures.
Willy Tarreauf3bfede2011-07-25 11:38:17 +02003 * Version 6.0.6
4 * (C) 2002-2011 - Willy Tarreau <w@1wt.eu>
Willy Tarreauc2186022009-10-26 19:48:54 +01005 *
Willy Tarreauf3bfede2011-07-25 11:38:17 +02006 * This library is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation, version 2.1
9 * exclusively.
Willy Tarreauc2186022009-10-26 19:48:54 +010010 *
Willy Tarreauf3bfede2011-07-25 11:38:17 +020011 * This library is distributed in the hope that it will be useful,
Willy Tarreauc2186022009-10-26 19:48:54 +010012 * but WITHOUT ANY WARRANTY; without even the implied warranty of
Willy Tarreauf3bfede2011-07-25 11:38:17 +020013 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
Willy Tarreauc2186022009-10-26 19:48:54 +010015 *
Willy Tarreauf3bfede2011-07-25 11:38:17 +020016 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
Willy Tarreauc2186022009-10-26 19:48:54 +010019 */
20
21
22
23/*
24 General idea:
25 -------------
26 In a radix binary tree, we may have up to 2N-1 nodes for N keys if all of
27 them are leaves. If we find a way to differentiate intermediate nodes (later
28 called "nodes") and final nodes (later called "leaves"), and we associate
29 them by two, it is possible to build sort of a self-contained radix tree with
30 intermediate nodes always present. It will not be as cheap as the ultree for
31 optimal cases as shown below, but the optimal case almost never happens :
32
33 Eg, to store 8, 10, 12, 13, 14 :
34
35 ultree this theorical tree
36
37 8 8
38 / \ / \
39 10 12 10 12
40 / \ / \
41 13 14 12 14
42 / \
43 12 13
44
45 Note that on real-world tests (with a scheduler), is was verified that the
46 case with data on an intermediate node never happens. This is because the
47 data spectrum is too large for such coincidences to happen. It would require
48 for instance that a task has its expiration time at an exact second, with
49 other tasks sharing that second. This is too rare to try to optimize for it.
50
51 What is interesting is that the node will only be added above the leaf when
52 necessary, which implies that it will always remain somewhere above it. So
53 both the leaf and the node can share the exact value of the leaf, because
54 when going down the node, the bit mask will be applied to comparisons. So we
55 are tempted to have one single key shared between the node and the leaf.
56
57 The bit only serves the nodes, and the dups only serve the leaves. So we can
58 put a lot of information in common. This results in one single entity with
59 two branch pointers and two parent pointers, one for the node part, and one
60 for the leaf part :
61
62 node's leaf's
63 parent parent
64 | |
65 [node] [leaf]
66 / \
67 left right
68 branch branch
69
70 The node may very well refer to its leaf counterpart in one of its branches,
71 indicating that its own leaf is just below it :
72
73 node's
74 parent
75 |
76 [node]
77 / \
78 left [leaf]
79 branch
80
81 Adding keys in such a tree simply consists in inserting nodes between
82 other nodes and/or leaves :
83
84 [root]
85 |
86 [node2]
87 / \
88 [leaf1] [node3]
89 / \
90 [leaf2] [leaf3]
91
92 On this diagram, we notice that [node2] and [leaf2] have been pulled away
93 from each other due to the insertion of [node3], just as if there would be
94 an elastic between both parts. This elastic-like behaviour gave its name to
95 the tree : "Elastic Binary Tree", or "EBtree". The entity which associates a
96 node part and a leaf part will be called an "EB node".
97
98 We also notice on the diagram that there is a root entity required to attach
99 the tree. It only contains two branches and there is nothing above it. This
100 is an "EB root". Some will note that [leaf1] has no [node1]. One property of
101 the EBtree is that all nodes have their branches filled, and that if a node
102 has only one branch, it does not need to exist. Here, [leaf1] was added
103 below [root] and did not need any node.
104
105 An EB node contains :
106 - a pointer to the node's parent (node_p)
107 - a pointer to the leaf's parent (leaf_p)
108 - two branches pointing to lower nodes or leaves (branches)
109 - a bit position (bit)
110 - an optional key.
111
112 The key here is optional because it's used only during insertion, in order
113 to classify the nodes. Nothing else in the tree structure requires knowledge
114 of the key. This makes it possible to write type-agnostic primitives for
115 everything, and type-specific insertion primitives. This has led to consider
116 two types of EB nodes. The type-agnostic ones will serve as a header for the
117 other ones, and will simply be called "struct eb_node". The other ones will
118 have their type indicated in the structure name. Eg: "struct eb32_node" for
119 nodes carrying 32 bit keys.
120
121 We will also node that the two branches in a node serve exactly the same
122 purpose as an EB root. For this reason, a "struct eb_root" will be used as
123 well inside the struct eb_node. In order to ease pointer manipulation and
124 ROOT detection when walking upwards, all the pointers inside an eb_node will
125 point to the eb_root part of the referenced EB nodes, relying on the same
126 principle as the linked lists in Linux.
127
128 Another important point to note, is that when walking inside a tree, it is
129 very convenient to know where a node is attached in its parent, and what
130 type of branch it has below it (leaf or node). In order to simplify the
131 operations and to speed up the processing, it was decided in this specific
132 implementation to use the lowest bit from the pointer to designate the side
133 of the upper pointers (left/right) and the type of a branch (leaf/node).
134 This practise is not mandatory by design, but an implementation-specific
135 optimisation permitted on all platforms on which data must be aligned. All
136 known 32 bit platforms align their integers and pointers to 32 bits, leaving
137 the two lower bits unused. So, we say that the pointers are "tagged". And
138 since they designate pointers to root parts, we simply call them
139 "tagged root pointers", or "eb_troot" in the code.
140
141 Duplicate keys are stored in a special manner. When inserting a key, if
142 the same one is found, then an incremental binary tree is built at this
143 place from these keys. This ensures that no special case has to be written
144 to handle duplicates when walking through the tree or when deleting entries.
145 It also guarantees that duplicates will be walked in the exact same order
146 they were inserted. This is very important when trying to achieve fair
147 processing distribution for instance.
148
149 Algorithmic complexity can be derived from 3 variables :
150 - the number of possible different keys in the tree : P
151 - the number of entries in the tree : N
152 - the number of duplicates for one key : D
153
154 Note that this tree is deliberately NOT balanced. For this reason, the worst
155 case may happen with a small tree (eg: 32 distinct keys of one bit). BUT,
156 the operations required to manage such data are so much cheap that they make
157 it worth using it even under such conditions. For instance, a balanced tree
158 may require only 6 levels to store those 32 keys when this tree will
159 require 32. But if per-level operations are 5 times cheaper, it wins.
160
161 Minimal, Maximal and Average times are specified in number of operations.
162 Minimal is given for best condition, Maximal for worst condition, and the
163 average is reported for a tree containing random keys. An operation
164 generally consists in jumping from one node to the other.
165
166 Complexity :
167 - lookup : min=1, max=log(P), avg=log(N)
168 - insertion from root : min=1, max=log(P), avg=log(N)
169 - insertion of dups : min=1, max=log(D), avg=log(D)/2 after lookup
170 - deletion : min=1, max=1, avg=1
171 - prev/next : min=1, max=log(P), avg=2 :
172 N/2 nodes need 1 hop => 1*N/2
173 N/4 nodes need 2 hops => 2*N/4
174 N/8 nodes need 3 hops => 3*N/8
175 ...
176 N/x nodes need log(x) hops => log2(x)*N/x
177 Total cost for all N nodes : sum[i=1..N](log2(i)*N/i) = N*sum[i=1..N](log2(i)/i)
178 Average cost across N nodes = total / N = sum[i=1..N](log2(i)/i) = 2
179
180 This design is currently limited to only two branches per node. Most of the
181 tree descent algorithm would be compatible with more branches (eg: 4, to cut
182 the height in half), but this would probably require more complex operations
183 and the deletion algorithm would be problematic.
184
185 Useful properties :
186 - a node is always added above the leaf it is tied to, and never can get
187 below nor in another branch. This implies that leaves directly attached
188 to the root do not use their node part, which is indicated by a NULL
189 value in node_p. This also enhances the cache efficiency when walking
190 down the tree, because when the leaf is reached, its node part will
191 already have been visited (unless it's the first leaf in the tree).
192
193 - pointers to lower nodes or leaves are stored in "branch" pointers. Only
194 the root node may have a NULL in either branch, it is not possible for
195 other branches. Since the nodes are attached to the left branch of the
196 root, it is not possible to see a NULL left branch when walking up a
197 tree. Thus, an empty tree is immediately identified by a NULL left
198 branch at the root. Conversely, the one and only way to identify the
199 root node is to check that it right branch is NULL. Note that the
200 NULL pointer may have a few low-order bits set.
201
202 - a node connected to its own leaf will have branch[0|1] pointing to
203 itself, and leaf_p pointing to itself.
204
205 - a node can never have node_p pointing to itself.
206
207 - a node is linked in a tree if and only if it has a non-null leaf_p.
208
209 - a node can never have both branches equal, except for the root which can
210 have them both NULL.
211
212 - deletion only applies to leaves. When a leaf is deleted, its parent must
213 be released too (unless it's the root), and its sibling must attach to
214 the grand-parent, replacing the parent. Also, when a leaf is deleted,
215 the node tied to this leaf will be removed and must be released too. If
216 this node is different from the leaf's parent, the freshly released
217 leaf's parent will be used to replace the node which must go. A released
218 node will never be used anymore, so there's no point in tracking it.
219
220 - the bit index in a node indicates the bit position in the key which is
221 represented by the branches. That means that a node with (bit == 0) is
222 just above two leaves. Negative bit values are used to build a duplicate
223 tree. The first node above two identical leaves gets (bit == -1). This
224 value logarithmically decreases as the duplicate tree grows. During
225 duplicate insertion, a node is inserted above the highest bit value (the
226 lowest absolute value) in the tree during the right-sided walk. If bit
227 -1 is not encountered (highest < -1), we insert above last leaf.
228 Otherwise, we insert above the node with the highest value which was not
229 equal to the one of its parent + 1.
230
231 - the "eb_next" primitive walks from left to right, which means from lower
232 to higher keys. It returns duplicates in the order they were inserted.
233 The "eb_first" primitive returns the left-most entry.
234
235 - the "eb_prev" primitive walks from right to left, which means from
236 higher to lower keys. It returns duplicates in the opposite order they
237 were inserted. The "eb_last" primitive returns the right-most entry.
238
239 - a tree which has 1 in the lower bit of its root's right branch is a
240 tree with unique nodes. This means that when a node is inserted with
241 a key which already exists will not be inserted, and the previous
242 entry will be returned.
243
244 */
245
246#ifndef _EBTREE_H
247#define _EBTREE_H
248
249#include <stdlib.h>
Willy Tarreaucc05fba2009-10-27 21:40:18 +0100250#include "compiler.h"
Willy Tarreauc2186022009-10-26 19:48:54 +0100251
Willy Tarreau3a932442010-05-09 19:29:23 +0200252static inline int flsnz8_generic(unsigned int x)
253{
254 int ret = 0;
255 if (x >> 4) { x >>= 4; ret += 4; }
256 return ret + ((0xFFFFAA50U >> (x << 1)) & 3) + 1;
257}
258
Willy Tarreauc2186022009-10-26 19:48:54 +0100259/* Note: we never need to run fls on null keys, so we can optimize the fls
260 * function by removing a conditional jump.
261 */
Willy Tarreau3a932442010-05-09 19:29:23 +0200262#if defined(__i386__) || defined(__x86_64__)
263/* this code is similar on 32 and 64 bit */
Willy Tarreauc2186022009-10-26 19:48:54 +0100264static inline int flsnz(int x)
265{
266 int r;
267 __asm__("bsrl %1,%0\n"
268 : "=r" (r) : "rm" (x));
269 return r+1;
270}
Willy Tarreau3a932442010-05-09 19:29:23 +0200271
272static inline int flsnz8(unsigned char x)
273{
274 int r;
275 __asm__("movzbl %%al, %%eax\n"
276 "bsrl %%eax,%0\n"
277 : "=r" (r) : "a" (x));
278 return r+1;
279}
280
Willy Tarreauc2186022009-10-26 19:48:54 +0100281#else
282// returns 1 to 32 for 1<<0 to 1<<31. Undefined for 0.
283#define flsnz(___a) ({ \
284 register int ___x, ___bits = 0; \
285 ___x = (___a); \
286 if (___x & 0xffff0000) { ___x &= 0xffff0000; ___bits += 16;} \
287 if (___x & 0xff00ff00) { ___x &= 0xff00ff00; ___bits += 8;} \
288 if (___x & 0xf0f0f0f0) { ___x &= 0xf0f0f0f0; ___bits += 4;} \
289 if (___x & 0xcccccccc) { ___x &= 0xcccccccc; ___bits += 2;} \
290 if (___x & 0xaaaaaaaa) { ___x &= 0xaaaaaaaa; ___bits += 1;} \
291 ___bits + 1; \
292 })
Willy Tarreau3a932442010-05-09 19:29:23 +0200293
294static inline int flsnz8(unsigned int x)
295{
296 return flsnz8_generic(x);
297}
298
299
Willy Tarreauc2186022009-10-26 19:48:54 +0100300#endif
301
302static inline int fls64(unsigned long long x)
303{
304 unsigned int h;
305 unsigned int bits = 32;
306
307 h = x >> 32;
308 if (!h) {
309 h = x;
310 bits = 0;
311 }
312 return flsnz(h) + bits;
313}
314
315#define fls_auto(x) ((sizeof(x) > 4) ? fls64(x) : flsnz(x))
316
317/* Linux-like "container_of". It returns a pointer to the structure of type
318 * <type> which has its member <name> stored at address <ptr>.
319 */
320#ifndef container_of
321#define container_of(ptr, type, name) ((type *)(((void *)(ptr)) - ((long)&((type *)0)->name)))
322#endif
323
Willy Tarreau2b570202013-05-07 15:58:28 +0200324/* returns a pointer to the structure of type <type> which has its member <name>
325 * stored at address <ptr>, unless <ptr> is 0, in which case 0 is returned.
326 */
327#ifndef container_of_safe
328#define container_of_safe(ptr, type, name) \
329 ({ void *__p = (ptr); \
330 __p ? (type *)(__p - ((long)&((type *)0)->name)) : (type *)0; \
331 })
332#endif
333
Willy Tarreauc2186022009-10-26 19:48:54 +0100334/* Number of bits per node, and number of leaves per node */
335#define EB_NODE_BITS 1
336#define EB_NODE_BRANCHES (1 << EB_NODE_BITS)
337#define EB_NODE_BRANCH_MASK (EB_NODE_BRANCHES - 1)
338
339/* Be careful not to tweak those values. The walking code is optimized for NULL
340 * detection on the assumption that the following values are intact.
341 */
342#define EB_LEFT 0
343#define EB_RGHT 1
344#define EB_LEAF 0
345#define EB_NODE 1
346
347/* Tags to set in root->b[EB_RGHT] :
348 * - EB_NORMAL is a normal tree which stores duplicate keys.
349 * - EB_UNIQUE is a tree which stores unique keys.
350 */
351#define EB_NORMAL 0
352#define EB_UNIQUE 1
353
354/* This is the same as an eb_node pointer, except that the lower bit embeds
355 * a tag. See eb_dotag()/eb_untag()/eb_gettag(). This tag has two meanings :
356 * - 0=left, 1=right to designate the parent's branch for leaf_p/node_p
357 * - 0=link, 1=leaf to designate the branch's type for branch[]
358 */
359typedef void eb_troot_t;
360
361/* The eb_root connects the node which contains it, to two nodes below it, one
362 * of which may be the same node. At the top of the tree, we use an eb_root
363 * too, which always has its right branch NULL (+/1 low-order bits).
364 */
365struct eb_root {
366 eb_troot_t *b[EB_NODE_BRANCHES]; /* left and right branches */
367};
368
369/* The eb_node contains the two parts, one for the leaf, which always exists,
370 * and one for the node, which remains unused in the very first node inserted
371 * into the tree. This structure is 20 bytes per node on 32-bit machines. Do
372 * not change the order, benchmarks have shown that it's optimal this way.
373 */
374struct eb_node {
375 struct eb_root branches; /* branches, must be at the beginning */
376 eb_troot_t *node_p; /* link node's parent */
377 eb_troot_t *leaf_p; /* leaf node's parent */
Willy Tarreau3a932442010-05-09 19:29:23 +0200378 short int bit; /* link's bit position. */
Willy Tarreau22c0a932011-07-25 12:22:44 +0200379 short unsigned int pfx; /* data prefix length, always related to leaf */
Willy Tarreauf82eb022013-12-07 01:09:04 +0100380} __attribute__((packed));
Willy Tarreauc2186022009-10-26 19:48:54 +0100381
382/* Return the structure of type <type> whose member <member> points to <ptr> */
383#define eb_entry(ptr, type, member) container_of(ptr, type, member)
384
385/* The root of a tree is an eb_root initialized with both pointers NULL.
386 * During its life, only the left pointer will change. The right one will
387 * always remain NULL, which is the way we detect it.
388 */
389#define EB_ROOT \
390 (struct eb_root) { \
391 .b = {[0] = NULL, [1] = NULL }, \
392 }
393
394#define EB_ROOT_UNIQUE \
395 (struct eb_root) { \
396 .b = {[0] = NULL, [1] = (void *)1 }, \
397 }
398
399#define EB_TREE_HEAD(name) \
400 struct eb_root name = EB_ROOT
401
402
403/***************************************\
404 * Private functions. Not for end-user *
405\***************************************/
406
407/* Converts a root pointer to its equivalent eb_troot_t pointer,
408 * ready to be stored in ->branch[], leaf_p or node_p. NULL is not
409 * conserved. To be used with EB_LEAF, EB_NODE, EB_LEFT or EB_RGHT in <tag>.
410 */
411static inline eb_troot_t *eb_dotag(const struct eb_root *root, const int tag)
412{
413 return (eb_troot_t *)((void *)root + tag);
414}
415
416/* Converts an eb_troot_t pointer pointer to its equivalent eb_root pointer,
417 * for use with pointers from ->branch[], leaf_p or node_p. NULL is conserved
418 * as long as the tree is not corrupted. To be used with EB_LEAF, EB_NODE,
419 * EB_LEFT or EB_RGHT in <tag>.
420 */
421static inline struct eb_root *eb_untag(const eb_troot_t *troot, const int tag)
422{
423 return (struct eb_root *)((void *)troot - tag);
424}
425
426/* returns the tag associated with an eb_troot_t pointer */
427static inline int eb_gettag(eb_troot_t *troot)
428{
429 return (unsigned long)troot & 1;
430}
431
432/* Converts a root pointer to its equivalent eb_troot_t pointer and clears the
433 * tag, no matter what its value was.
434 */
435static inline struct eb_root *eb_clrtag(const eb_troot_t *troot)
436{
437 return (struct eb_root *)((unsigned long)troot & ~1UL);
438}
439
440/* Returns a pointer to the eb_node holding <root> */
441static inline struct eb_node *eb_root_to_node(struct eb_root *root)
442{
443 return container_of(root, struct eb_node, branches);
444}
445
446/* Walks down starting at root pointer <start>, and always walking on side
447 * <side>. It either returns the node hosting the first leaf on that side,
448 * or NULL if no leaf is found. <start> may either be NULL or a branch pointer.
449 * The pointer to the leaf (or NULL) is returned.
450 */
451static inline struct eb_node *eb_walk_down(eb_troot_t *start, unsigned int side)
452{
453 /* A NULL pointer on an empty tree root will be returned as-is */
454 while (eb_gettag(start) == EB_NODE)
455 start = (eb_untag(start, EB_NODE))->b[side];
456 /* NULL is left untouched (root==eb_node, EB_LEAF==0) */
457 return eb_root_to_node(eb_untag(start, EB_LEAF));
458}
459
460/* This function is used to build a tree of duplicates by adding a new node to
461 * a subtree of at least 2 entries. It will probably never be needed inlined,
462 * and it is not for end-user.
463 */
464static forceinline struct eb_node *
465__eb_insert_dup(struct eb_node *sub, struct eb_node *new)
466{
467 struct eb_node *head = sub;
468
Willy Tarreau655c84a2011-09-19 20:36:45 +0200469 eb_troot_t *new_left = eb_dotag(&new->branches, EB_LEFT);
470 eb_troot_t *new_rght = eb_dotag(&new->branches, EB_RGHT);
471 eb_troot_t *new_leaf = eb_dotag(&new->branches, EB_LEAF);
Willy Tarreauc2186022009-10-26 19:48:54 +0100472
473 /* first, identify the deepest hole on the right branch */
474 while (eb_gettag(head->branches.b[EB_RGHT]) != EB_LEAF) {
475 struct eb_node *last = head;
476 head = container_of(eb_untag(head->branches.b[EB_RGHT], EB_NODE),
477 struct eb_node, branches);
478 if (head->bit > last->bit + 1)
479 sub = head; /* there's a hole here */
480 }
481
482 /* Here we have a leaf attached to (head)->b[EB_RGHT] */
483 if (head->bit < -1) {
484 /* A hole exists just before the leaf, we insert there */
485 new->bit = -1;
486 sub = container_of(eb_untag(head->branches.b[EB_RGHT], EB_LEAF),
487 struct eb_node, branches);
488 head->branches.b[EB_RGHT] = eb_dotag(&new->branches, EB_NODE);
489
490 new->node_p = sub->leaf_p;
491 new->leaf_p = new_rght;
492 sub->leaf_p = new_left;
493 new->branches.b[EB_LEFT] = eb_dotag(&sub->branches, EB_LEAF);
494 new->branches.b[EB_RGHT] = new_leaf;
495 return new;
496 } else {
497 int side;
498 /* No hole was found before a leaf. We have to insert above
499 * <sub>. Note that we cannot be certain that <sub> is attached
500 * to the right of its parent, as this is only true if <sub>
501 * is inside the dup tree, not at the head.
502 */
503 new->bit = sub->bit - 1; /* install at the lowest level */
504 side = eb_gettag(sub->node_p);
505 head = container_of(eb_untag(sub->node_p, side), struct eb_node, branches);
506 head->branches.b[side] = eb_dotag(&new->branches, EB_NODE);
507
508 new->node_p = sub->node_p;
509 new->leaf_p = new_rght;
510 sub->node_p = new_left;
511 new->branches.b[EB_LEFT] = eb_dotag(&sub->branches, EB_NODE);
512 new->branches.b[EB_RGHT] = new_leaf;
513 return new;
514 }
515}
516
517
518/**************************************\
519 * Public functions, for the end-user *
520\**************************************/
521
Willy Tarreaufdc10182010-05-16 21:13:24 +0200522/* Return non-zero if the tree is empty, otherwise zero */
Willy Tarreau43be3402019-10-02 15:21:58 +0200523static inline int eb_is_empty(const struct eb_root *root)
Willy Tarreaufdc10182010-05-16 21:13:24 +0200524{
525 return !root->b[EB_LEFT];
526}
527
Willy Tarreau2b570202013-05-07 15:58:28 +0200528/* Return non-zero if the node is a duplicate, otherwise zero */
Willy Tarreau43be3402019-10-02 15:21:58 +0200529static inline int eb_is_dup(const struct eb_node *node)
Willy Tarreau2b570202013-05-07 15:58:28 +0200530{
531 return node->bit < 0;
532}
533
Willy Tarreauc2186022009-10-26 19:48:54 +0100534/* Return the first leaf in the tree starting at <root>, or NULL if none */
535static inline struct eb_node *eb_first(struct eb_root *root)
536{
537 return eb_walk_down(root->b[0], EB_LEFT);
538}
539
540/* Return the last leaf in the tree starting at <root>, or NULL if none */
541static inline struct eb_node *eb_last(struct eb_root *root)
542{
543 return eb_walk_down(root->b[0], EB_RGHT);
544}
545
546/* Return previous leaf node before an existing leaf node, or NULL if none. */
547static inline struct eb_node *eb_prev(struct eb_node *node)
548{
549 eb_troot_t *t = node->leaf_p;
550
551 while (eb_gettag(t) == EB_LEFT) {
552 /* Walking up from left branch. We must ensure that we never
553 * walk beyond root.
554 */
555 if (unlikely(eb_clrtag((eb_untag(t, EB_LEFT))->b[EB_RGHT]) == NULL))
556 return NULL;
557 t = (eb_root_to_node(eb_untag(t, EB_LEFT)))->node_p;
558 }
559 /* Note that <t> cannot be NULL at this stage */
560 t = (eb_untag(t, EB_RGHT))->b[EB_LEFT];
561 return eb_walk_down(t, EB_RGHT);
562}
563
564/* Return next leaf node after an existing leaf node, or NULL if none. */
565static inline struct eb_node *eb_next(struct eb_node *node)
566{
567 eb_troot_t *t = node->leaf_p;
568
569 while (eb_gettag(t) != EB_LEFT)
570 /* Walking up from right branch, so we cannot be below root */
571 t = (eb_root_to_node(eb_untag(t, EB_RGHT)))->node_p;
572
573 /* Note that <t> cannot be NULL at this stage */
Willy Tarreau2b570202013-05-07 15:58:28 +0200574 t = (eb_untag(t, EB_LEFT))->b[EB_RGHT];
575 if (eb_clrtag(t) == NULL)
576 return NULL;
577 return eb_walk_down(t, EB_LEFT);
578}
579
580/* Return previous leaf node within a duplicate sub-tree, or NULL if none. */
581static inline struct eb_node *eb_prev_dup(struct eb_node *node)
582{
583 eb_troot_t *t = node->leaf_p;
584
585 while (eb_gettag(t) == EB_LEFT) {
586 /* Walking up from left branch. We must ensure that we never
587 * walk beyond root.
588 */
589 if (unlikely(eb_clrtag((eb_untag(t, EB_LEFT))->b[EB_RGHT]) == NULL))
590 return NULL;
591 /* if the current node leaves a dup tree, quit */
592 if ((eb_root_to_node(eb_untag(t, EB_LEFT)))->bit >= 0)
593 return NULL;
594 t = (eb_root_to_node(eb_untag(t, EB_LEFT)))->node_p;
595 }
596 /* Note that <t> cannot be NULL at this stage */
597 if ((eb_root_to_node(eb_untag(t, EB_RGHT)))->bit >= 0)
598 return NULL;
599 t = (eb_untag(t, EB_RGHT))->b[EB_LEFT];
600 return eb_walk_down(t, EB_RGHT);
601}
602
603/* Return next leaf node within a duplicate sub-tree, or NULL if none. */
604static inline struct eb_node *eb_next_dup(struct eb_node *node)
605{
606 eb_troot_t *t = node->leaf_p;
607
608 while (eb_gettag(t) != EB_LEFT) {
609 /* Walking up from right branch, so we cannot be below root */
610 /* if the current node leaves a dup tree, quit */
611 if ((eb_root_to_node(eb_untag(t, EB_RGHT)))->bit >= 0)
612 return NULL;
613 t = (eb_root_to_node(eb_untag(t, EB_RGHT)))->node_p;
614 }
615
616 /* Note that <t> cannot be NULL at this stage */
617 if ((eb_root_to_node(eb_untag(t, EB_LEFT)))->bit >= 0)
618 return NULL;
Willy Tarreauc2186022009-10-26 19:48:54 +0100619 t = (eb_untag(t, EB_LEFT))->b[EB_RGHT];
620 if (eb_clrtag(t) == NULL)
621 return NULL;
622 return eb_walk_down(t, EB_LEFT);
623}
624
625/* Return previous leaf node before an existing leaf node, skipping duplicates,
626 * or NULL if none. */
627static inline struct eb_node *eb_prev_unique(struct eb_node *node)
628{
629 eb_troot_t *t = node->leaf_p;
630
631 while (1) {
632 if (eb_gettag(t) != EB_LEFT) {
633 node = eb_root_to_node(eb_untag(t, EB_RGHT));
634 /* if we're right and not in duplicates, stop here */
635 if (node->bit >= 0)
636 break;
637 t = node->node_p;
638 }
639 else {
640 /* Walking up from left branch. We must ensure that we never
641 * walk beyond root.
642 */
643 if (unlikely(eb_clrtag((eb_untag(t, EB_LEFT))->b[EB_RGHT]) == NULL))
644 return NULL;
645 t = (eb_root_to_node(eb_untag(t, EB_LEFT)))->node_p;
646 }
647 }
648 /* Note that <t> cannot be NULL at this stage */
649 t = (eb_untag(t, EB_RGHT))->b[EB_LEFT];
650 return eb_walk_down(t, EB_RGHT);
651}
652
653/* Return next leaf node after an existing leaf node, skipping duplicates, or
654 * NULL if none.
655 */
656static inline struct eb_node *eb_next_unique(struct eb_node *node)
657{
658 eb_troot_t *t = node->leaf_p;
659
660 while (1) {
661 if (eb_gettag(t) == EB_LEFT) {
662 if (unlikely(eb_clrtag((eb_untag(t, EB_LEFT))->b[EB_RGHT]) == NULL))
663 return NULL; /* we reached root */
664 node = eb_root_to_node(eb_untag(t, EB_LEFT));
665 /* if we're left and not in duplicates, stop here */
666 if (node->bit >= 0)
667 break;
668 t = node->node_p;
669 }
670 else {
671 /* Walking up from right branch, so we cannot be below root */
672 t = (eb_root_to_node(eb_untag(t, EB_RGHT)))->node_p;
673 }
674 }
675
676 /* Note that <t> cannot be NULL at this stage */
677 t = (eb_untag(t, EB_LEFT))->b[EB_RGHT];
678 if (eb_clrtag(t) == NULL)
679 return NULL;
680 return eb_walk_down(t, EB_LEFT);
681}
682
683
684/* Removes a leaf node from the tree if it was still in it. Marks the node
685 * as unlinked.
686 */
687static forceinline void __eb_delete(struct eb_node *node)
688{
689 __label__ delete_unlink;
690 unsigned int pside, gpside, sibtype;
691 struct eb_node *parent;
692 struct eb_root *gparent;
693
694 if (!node->leaf_p)
695 return;
696
697 /* we need the parent, our side, and the grand parent */
698 pside = eb_gettag(node->leaf_p);
699 parent = eb_root_to_node(eb_untag(node->leaf_p, pside));
700
701 /* We likely have to release the parent link, unless it's the root,
702 * in which case we only set our branch to NULL. Note that we can
703 * only be attached to the root by its left branch.
704 */
705
706 if (eb_clrtag(parent->branches.b[EB_RGHT]) == NULL) {
707 /* we're just below the root, it's trivial. */
708 parent->branches.b[EB_LEFT] = NULL;
709 goto delete_unlink;
710 }
711
712 /* To release our parent, we have to identify our sibling, and reparent
713 * it directly to/from the grand parent. Note that the sibling can
714 * either be a link or a leaf.
715 */
716
717 gpside = eb_gettag(parent->node_p);
718 gparent = eb_untag(parent->node_p, gpside);
719
720 gparent->b[gpside] = parent->branches.b[!pside];
721 sibtype = eb_gettag(gparent->b[gpside]);
722
723 if (sibtype == EB_LEAF) {
724 eb_root_to_node(eb_untag(gparent->b[gpside], EB_LEAF))->leaf_p =
725 eb_dotag(gparent, gpside);
726 } else {
727 eb_root_to_node(eb_untag(gparent->b[gpside], EB_NODE))->node_p =
728 eb_dotag(gparent, gpside);
729 }
730 /* Mark the parent unused. Note that we do not check if the parent is
731 * our own node, but that's not a problem because if it is, it will be
732 * marked unused at the same time, which we'll use below to know we can
733 * safely remove it.
734 */
735 parent->node_p = NULL;
736
737 /* The parent node has been detached, and is currently unused. It may
738 * belong to another node, so we cannot remove it that way. Also, our
739 * own node part might still be used. so we can use this spare node
740 * to replace ours if needed.
741 */
742
743 /* If our link part is unused, we can safely exit now */
744 if (!node->node_p)
745 goto delete_unlink;
746
747 /* From now on, <node> and <parent> are necessarily different, and the
748 * <node>'s node part is in use. By definition, <parent> is at least
749 * below <node>, so keeping its key for the bit string is OK.
750 */
751
752 parent->node_p = node->node_p;
753 parent->branches = node->branches;
754 parent->bit = node->bit;
755
756 /* We must now update the new node's parent... */
757 gpside = eb_gettag(parent->node_p);
758 gparent = eb_untag(parent->node_p, gpside);
759 gparent->b[gpside] = eb_dotag(&parent->branches, EB_NODE);
760
761 /* ... and its branches */
762 for (pside = 0; pside <= 1; pside++) {
763 if (eb_gettag(parent->branches.b[pside]) == EB_NODE) {
764 eb_root_to_node(eb_untag(parent->branches.b[pside], EB_NODE))->node_p =
765 eb_dotag(&parent->branches, pside);
766 } else {
767 eb_root_to_node(eb_untag(parent->branches.b[pside], EB_LEAF))->leaf_p =
768 eb_dotag(&parent->branches, pside);
769 }
770 }
771 delete_unlink:
772 /* Now the node has been completely unlinked */
773 node->leaf_p = NULL;
774 return; /* tree is not empty yet */
775}
776
777/* Compare blocks <a> and <b> byte-to-byte, from bit <ignore> to bit <len-1>.
778 * Return the number of equal bits between strings, assuming that the first
779 * <ignore> bits are already identical. It is possible to return slightly more
780 * than <len> bits if <len> does not stop on a byte boundary and we find exact
781 * bytes. Note that parts or all of <ignore> bits may be rechecked. It is only
782 * passed here as a hint to speed up the check.
783 */
Willy Tarreau3a932442010-05-09 19:29:23 +0200784static forceinline int equal_bits(const unsigned char *a,
785 const unsigned char *b,
786 int ignore, int len)
Willy Tarreauc2186022009-10-26 19:48:54 +0100787{
Willy Tarreau3a932442010-05-09 19:29:23 +0200788 for (ignore >>= 3, a += ignore, b += ignore, ignore <<= 3;
789 ignore < len; ) {
790 unsigned char c;
Willy Tarreauc2186022009-10-26 19:48:54 +0100791
Willy Tarreau3a932442010-05-09 19:29:23 +0200792 a++; b++;
793 ignore += 8;
794 c = b[-1] ^ a[-1];
795
796 if (c) {
797 /* OK now we know that old and new differ at byte <ptr> and that <c> holds
798 * the bit differences. We have to find what bit is differing and report
799 * it as the number of identical bits. Note that low bit numbers are
800 * assigned to high positions in the byte, as we compare them as strings.
801 */
802 ignore -= flsnz8(c);
803 break;
804 }
805 }
806 return ignore;
807}
Willy Tarreauc2186022009-10-26 19:48:54 +0100808
Willy Tarreau3a932442010-05-09 19:29:23 +0200809/* check that the two blocks <a> and <b> are equal on <len> bits. If it is known
810 * they already are on some bytes, this number of equal bytes to be skipped may
811 * be passed in <skip>. It returns 0 if they match, otherwise non-zero.
812 */
813static forceinline int check_bits(const unsigned char *a,
814 const unsigned char *b,
815 int skip,
816 int len)
817{
818 int bit, ret;
819
820 /* This uncommon construction gives the best performance on x86 because
821 * it makes heavy use multiple-index addressing and parallel instructions,
822 * and it prevents gcc from reordering the loop since it is already
823 * properly oriented. Tested to be fine with 2.95 to 4.2.
Willy Tarreauc2186022009-10-26 19:48:54 +0100824 */
Willy Tarreau3a932442010-05-09 19:29:23 +0200825 bit = ~len + (skip << 3) + 9; // = (skip << 3) + (8 - len)
826 ret = a[skip] ^ b[skip];
827 if (unlikely(bit >= 0))
828 return ret >> bit;
829 while (1) {
830 skip++;
831 if (ret)
832 return ret;
833 ret = a[skip] ^ b[skip];
834 bit += 8;
835 if (bit >= 0)
836 return ret >> bit;
837 }
Willy Tarreauc2186022009-10-26 19:48:54 +0100838}
839
Willy Tarreau3a932442010-05-09 19:29:23 +0200840
Willy Tarreauc2186022009-10-26 19:48:54 +0100841/* Compare strings <a> and <b> byte-to-byte, from bit <ignore> to the last 0.
842 * Return the number of equal bits between strings, assuming that the first
843 * <ignore> bits are already identical. Note that parts or all of <ignore> bits
844 * may be rechecked. It is only passed here as a hint to speed up the check.
845 * The caller is responsible for not passing an <ignore> value larger than any
846 * of the two strings. However, referencing any bit from the trailing zero is
Willy Tarreaub55fcf22010-10-28 22:48:29 +0200847 * permitted. Equal strings are reported as a negative number of bits, which
848 * indicates the end was reached.
Willy Tarreauc2186022009-10-26 19:48:54 +0100849 */
Willy Tarreau3a932442010-05-09 19:29:23 +0200850static forceinline int string_equal_bits(const unsigned char *a,
851 const unsigned char *b,
852 int ignore)
Willy Tarreauc2186022009-10-26 19:48:54 +0100853{
Willy Tarreau3a932442010-05-09 19:29:23 +0200854 int beg;
Willy Tarreauc2186022009-10-26 19:48:54 +0100855 unsigned char c;
856
857 beg = ignore >> 3;
858
859 /* skip known and identical bits. We stop at the first different byte
860 * or at the first zero we encounter on either side.
861 */
862 while (1) {
863 unsigned char d;
864
865 c = a[beg];
866 d = b[beg];
867 beg++;
868
869 c ^= d;
870 if (c)
871 break;
872 if (!d)
Willy Tarreaub55fcf22010-10-28 22:48:29 +0200873 return -1;
Willy Tarreauc2186022009-10-26 19:48:54 +0100874 }
Willy Tarreauc2186022009-10-26 19:48:54 +0100875 /* OK now we know that a and b differ at byte <beg>, or that both are zero.
876 * We have to find what bit is differing and report it as the number of
877 * identical bits. Note that low bit numbers are assigned to high positions
878 * in the byte, as we compare them as strings.
879 */
Willy Tarreau3a932442010-05-09 19:29:23 +0200880 return (beg << 3) - flsnz8(c);
Willy Tarreauc2186022009-10-26 19:48:54 +0100881}
882
883static forceinline int cmp_bits(const unsigned char *a, const unsigned char *b, unsigned int pos)
884{
885 unsigned int ofs;
886 unsigned char bit_a, bit_b;
887
888 ofs = pos >> 3;
889 pos = ~pos & 7;
890
891 bit_a = (a[ofs] >> pos) & 1;
892 bit_b = (b[ofs] >> pos) & 1;
893
894 return bit_a - bit_b; /* -1: a<b; 0: a=b; 1: a>b */
895}
896
897static forceinline int get_bit(const unsigned char *a, unsigned int pos)
898{
899 unsigned int ofs;
900
901 ofs = pos >> 3;
902 pos = ~pos & 7;
903 return (a[ofs] >> pos) & 1;
904}
905
906/* These functions are declared in ebtree.c */
907void eb_delete(struct eb_node *node);
908REGPRM1 struct eb_node *eb_insert_dup(struct eb_node *sub, struct eb_node *new);
909
910#endif /* _EB_TREE_H */
911
912/*
913 * Local variables:
914 * c-indent-level: 8
915 * c-basic-offset: 8
916 * End:
917 */