1 =================================
2 Red-black Trees (rbtree) in Linux
3 =================================
6 :Date: January 18, 2007
7 :Author: Rob Landley <rob@landley.net>
9 What are red-black trees, and what are they for?
10 ------------------------------------------------
12 Red-black trees are a type of self-balancing binary search tree, used for
13 storing sortable key/value data pairs. This differs from radix trees (which
14 are used to efficiently store sparse arrays and thus use long integer indexes
15 to insert/access/delete nodes) and hash tables (which are not kept sorted to
16 be easily traversed in order, and must be tuned for a specific size and
17 hash function where rbtrees scale gracefully storing arbitrary keys).
19 Red-black trees are similar to AVL trees, but provide faster real-time bounded
20 worst case performance for insertion and deletion (at most two rotations and
21 three rotations, respectively, to balance the tree), with slightly slower
22 (but still O(log n)) lookup time.
24 To quote Linux Weekly News:
26 There are a number of red-black trees in use in the kernel.
27 The deadline and CFQ I/O schedulers employ rbtrees to
28 track requests; the packet CD/DVD driver does the same.
29 The high-resolution timer code uses an rbtree to organize outstanding
30 timer requests. The ext3 filesystem tracks directory entries in a
31 red-black tree. Virtual memory areas (VMAs) are tracked with red-black
32 trees, as are epoll file descriptors, cryptographic keys, and network
33 packets in the "hierarchical token bucket" scheduler.
35 This document covers use of the Linux rbtree implementation. For more
36 information on the nature and implementation of Red Black Trees, see:
38 Linux Weekly News article on red-black trees
39 http://lwn.net/Articles/184495/
41 Wikipedia entry on red-black trees
42 http://en.wikipedia.org/wiki/Red-black_tree
44 Linux implementation of red-black trees
45 ---------------------------------------
47 Linux's rbtree implementation lives in the file "lib/rbtree.c". To use it,
48 "#include <linux/rbtree.h>".
50 The Linux rbtree implementation is optimized for speed, and thus has one
51 less layer of indirection (and better cache locality) than more traditional
52 tree implementations. Instead of using pointers to separate rb_node and data
53 structures, each instance of struct rb_node is embedded in the data structure
54 it organizes. And instead of using a comparison callback function pointer,
55 users are expected to write their own tree search and insert functions
56 which call the provided rbtree functions. Locking is also left up to the
57 user of the rbtree code.
62 Data nodes in an rbtree tree are structures containing a struct rb_node member::
69 When dealing with a pointer to the embedded struct rb_node, the containing data
70 structure may be accessed with the standard container_of() macro. In addition,
71 individual members may be accessed directly via rb_entry(node, type, member).
73 At the root of each rbtree is an rb_root structure, which is initialized to be
76 struct rb_root mytree = RB_ROOT;
78 Searching for a value in an rbtree
79 ----------------------------------
81 Writing a search function for your tree is fairly straightforward: start at the
82 root, compare each value, and follow the left or right branch as necessary.
86 struct mytype *my_search(struct rb_root *root, char *string)
88 struct rb_node *node = root->rb_node;
91 struct mytype *data = container_of(node, struct mytype, node);
94 result = strcmp(string, data->keystring);
99 node = node->rb_right;
106 Inserting data into an rbtree
107 -----------------------------
109 Inserting data in the tree involves first searching for the place to insert the
110 new node, then inserting the node and rebalancing ("recoloring") the tree.
112 The search for insertion differs from the previous search by finding the
113 location of the pointer on which to graft the new node. The new node also
114 needs a link to its parent node for rebalancing purposes.
118 int my_insert(struct rb_root *root, struct mytype *data)
120 struct rb_node **new = &(root->rb_node), *parent = NULL;
122 /* Figure out where to put new node */
124 struct mytype *this = container_of(*new, struct mytype, node);
125 int result = strcmp(data->keystring, this->keystring);
129 new = &((*new)->rb_left);
131 new = &((*new)->rb_right);
136 /* Add new node and rebalance tree. */
137 rb_link_node(&data->node, parent, new);
138 rb_insert_color(&data->node, root);
143 Removing or replacing existing data in an rbtree
144 ------------------------------------------------
146 To remove an existing node from a tree, call::
148 void rb_erase(struct rb_node *victim, struct rb_root *tree);
152 struct mytype *data = mysearch(&mytree, "walrus");
155 rb_erase(&data->node, &mytree);
159 To replace an existing node in a tree with a new one with the same key, call::
161 void rb_replace_node(struct rb_node *old, struct rb_node *new,
162 struct rb_root *tree);
164 Replacing a node this way does not re-sort the tree: If the new node doesn't
165 have the same key as the old node, the rbtree will probably become corrupted.
167 Iterating through the elements stored in an rbtree (in sort order)
168 ------------------------------------------------------------------
170 Four functions are provided for iterating through an rbtree's contents in
171 sorted order. These work on arbitrary trees, and should not need to be
172 modified or wrapped (except for locking purposes)::
174 struct rb_node *rb_first(struct rb_root *tree);
175 struct rb_node *rb_last(struct rb_root *tree);
176 struct rb_node *rb_next(struct rb_node *node);
177 struct rb_node *rb_prev(struct rb_node *node);
179 To start iterating, call rb_first() or rb_last() with a pointer to the root
180 of the tree, which will return a pointer to the node structure contained in
181 the first or last element in the tree. To continue, fetch the next or previous
182 node by calling rb_next() or rb_prev() on the current node. This will return
183 NULL when there are no more nodes left.
185 The iterator functions return a pointer to the embedded struct rb_node, from
186 which the containing data structure may be accessed with the container_of()
187 macro, and individual members may be accessed directly via
188 rb_entry(node, type, member).
192 struct rb_node *node;
193 for (node = rb_first(&mytree); node; node = rb_next(node))
194 printk("key=%s\n", rb_entry(node, struct mytype, node)->keystring);
196 Support for Augmented rbtrees
197 -----------------------------
199 Augmented rbtree is an rbtree with "some" additional data stored in
200 each node, where the additional data for node N must be a function of
201 the contents of all nodes in the subtree rooted at N. This data can
202 be used to augment some new functionality to rbtree. Augmented rbtree
203 is an optional feature built on top of basic rbtree infrastructure.
204 An rbtree user who wants this feature will have to call the augmentation
205 functions with the user provided augmentation callback when inserting
208 C files implementing augmented rbtree manipulation must include
209 <linux/rbtree_augmented.h> instead of <linux/rbtree.h>. Note that
210 linux/rbtree_augmented.h exposes some rbtree implementations details
211 you are not expected to rely on; please stick to the documented APIs
212 there and do not include <linux/rbtree_augmented.h> from header files
213 either so as to minimize chances of your users accidentally relying on
214 such implementation details.
216 On insertion, the user must update the augmented information on the path
217 leading to the inserted node, then call rb_link_node() as usual and
218 rb_augment_inserted() instead of the usual rb_insert_color() call.
219 If rb_augment_inserted() rebalances the rbtree, it will callback into
220 a user provided function to update the augmented information on the
223 When erasing a node, the user must call rb_erase_augmented() instead of
224 rb_erase(). rb_erase_augmented() calls back into user provided functions
225 to updated the augmented information on affected subtrees.
227 In both cases, the callbacks are provided through struct rb_augment_callbacks.
228 3 callbacks must be defined:
230 - A propagation callback, which updates the augmented value for a given
231 node and its ancestors, up to a given stop point (or NULL to update
232 all the way to the root).
234 - A copy callback, which copies the augmented value for a given subtree
235 to a newly assigned subtree root.
237 - A tree rotation callback, which copies the augmented value for a given
238 subtree to a newly assigned subtree root AND recomputes the augmented
239 information for the former subtree root.
241 The compiled code for rb_erase_augmented() may inline the propagation and
242 copy callbacks, which results in a large function, so each augmented rbtree
243 user should have a single rb_erase_augmented() call site in order to limit
250 Interval tree is an example of augmented rb tree. Reference -
251 "Introduction to Algorithms" by Cormen, Leiserson, Rivest and Stein.
252 More details about interval trees:
254 Classical rbtree has a single key and it cannot be directly used to store
255 interval ranges like [lo:hi] and do a quick lookup for any overlap with a new
256 lo:hi or to find whether there is an exact match for a new lo:hi.
258 However, rbtree can be augmented to store such interval ranges in a structured
259 way making it possible to do efficient lookup and exact match.
261 This "extra information" stored in each node is the maximum hi
262 (max_hi) value among all the nodes that are its descendants. This
263 information can be maintained at each node just be looking at the node
264 and its immediate children. And this will be used in O(log n) lookup
265 for lowest match (lowest start address among all possible matches)
266 with something like::
268 struct interval_tree_node *
269 interval_tree_first_match(struct rb_root *root,
270 unsigned long start, unsigned long last)
272 struct interval_tree_node *node;
276 node = rb_entry(root->rb_node, struct interval_tree_node, rb);
279 if (node->rb.rb_left) {
280 struct interval_tree_node *left =
281 rb_entry(node->rb.rb_left,
282 struct interval_tree_node, rb);
283 if (left->__subtree_last >= start) {
285 * Some nodes in left subtree satisfy Cond2.
286 * Iterate to find the leftmost such node N.
287 * If it also satisfies Cond1, that's the match
288 * we are looking for. Otherwise, there is no
289 * matching interval as nodes to the right of N
290 * can't satisfy Cond1 either.
296 if (node->start <= last) { /* Cond1 */
297 if (node->last >= start) /* Cond2 */
298 return node; /* node is leftmost match */
299 if (node->rb.rb_right) {
300 node = rb_entry(node->rb.rb_right,
301 struct interval_tree_node, rb);
302 if (node->__subtree_last >= start)
306 return NULL; /* No match */
310 Insertion/removal are defined using the following augmented callbacks::
312 static inline unsigned long
313 compute_subtree_last(struct interval_tree_node *node)
315 unsigned long max = node->last, subtree_last;
316 if (node->rb.rb_left) {
317 subtree_last = rb_entry(node->rb.rb_left,
318 struct interval_tree_node, rb)->__subtree_last;
319 if (max < subtree_last)
322 if (node->rb.rb_right) {
323 subtree_last = rb_entry(node->rb.rb_right,
324 struct interval_tree_node, rb)->__subtree_last;
325 if (max < subtree_last)
331 static void augment_propagate(struct rb_node *rb, struct rb_node *stop)
334 struct interval_tree_node *node =
335 rb_entry(rb, struct interval_tree_node, rb);
336 unsigned long subtree_last = compute_subtree_last(node);
337 if (node->__subtree_last == subtree_last)
339 node->__subtree_last = subtree_last;
340 rb = rb_parent(&node->rb);
344 static void augment_copy(struct rb_node *rb_old, struct rb_node *rb_new)
346 struct interval_tree_node *old =
347 rb_entry(rb_old, struct interval_tree_node, rb);
348 struct interval_tree_node *new =
349 rb_entry(rb_new, struct interval_tree_node, rb);
351 new->__subtree_last = old->__subtree_last;
354 static void augment_rotate(struct rb_node *rb_old, struct rb_node *rb_new)
356 struct interval_tree_node *old =
357 rb_entry(rb_old, struct interval_tree_node, rb);
358 struct interval_tree_node *new =
359 rb_entry(rb_new, struct interval_tree_node, rb);
361 new->__subtree_last = old->__subtree_last;
362 old->__subtree_last = compute_subtree_last(old);
365 static const struct rb_augment_callbacks augment_callbacks = {
366 augment_propagate, augment_copy, augment_rotate
369 void interval_tree_insert(struct interval_tree_node *node,
370 struct rb_root *root)
372 struct rb_node **link = &root->rb_node, *rb_parent = NULL;
373 unsigned long start = node->start, last = node->last;
374 struct interval_tree_node *parent;
378 parent = rb_entry(rb_parent, struct interval_tree_node, rb);
379 if (parent->__subtree_last < last)
380 parent->__subtree_last = last;
381 if (start < parent->start)
382 link = &parent->rb.rb_left;
384 link = &parent->rb.rb_right;
387 node->__subtree_last = last;
388 rb_link_node(&node->rb, rb_parent, link);
389 rb_insert_augmented(&node->rb, root, &augment_callbacks);
392 void interval_tree_remove(struct interval_tree_node *node,
393 struct rb_root *root)
395 rb_erase_augmented(&node->rb, root, &augment_callbacks);