5 * Brad Boyer (flar@allandria.com)
6 * (C) 2003 Ardis Technologies <roman@ardistech.com>
8 * Handle basic btree node operations
11 #include <linux/pagemap.h>
12 #include <linux/slab.h>
13 #include <linux/swap.h>
17 void hfs_bnode_read(struct hfs_bnode *node, void *buf, int off, int len)
25 off += node->page_offset;
26 pagenum = off >> PAGE_SHIFT;
27 off &= ~PAGE_MASK; /* compute page offset for the first page */
29 for (bytes_read = 0; bytes_read < len; bytes_read += bytes_to_read) {
30 if (pagenum >= node->tree->pages_per_bnode)
32 page = node->page[pagenum];
33 bytes_to_read = min_t(int, len - bytes_read, PAGE_SIZE - off);
35 vaddr = kmap_atomic(page);
36 memcpy(buf + bytes_read, vaddr + off, bytes_to_read);
40 off = 0; /* page offset only applies to the first page */
44 u16 hfs_bnode_read_u16(struct hfs_bnode *node, int off)
48 hfs_bnode_read(node, &data, off, 2);
49 return be16_to_cpu(data);
52 u8 hfs_bnode_read_u8(struct hfs_bnode *node, int off)
56 hfs_bnode_read(node, &data, off, 1);
60 void hfs_bnode_read_key(struct hfs_bnode *node, void *key, int off)
62 struct hfs_btree *tree;
66 if (node->type == HFS_NODE_LEAF ||
67 tree->attributes & HFS_TREE_VARIDXKEYS)
68 key_len = hfs_bnode_read_u8(node, off) + 1;
70 key_len = tree->max_key_len + 1;
72 hfs_bnode_read(node, key, off, key_len);
75 void hfs_bnode_write(struct hfs_bnode *node, void *buf, int off, int len)
79 off += node->page_offset;
82 memcpy(kmap(page) + off, buf, len);
87 void hfs_bnode_write_u16(struct hfs_bnode *node, int off, u16 data)
89 __be16 v = cpu_to_be16(data);
91 hfs_bnode_write(node, &v, off, 2);
94 void hfs_bnode_write_u8(struct hfs_bnode *node, int off, u8 data)
97 hfs_bnode_write(node, &data, off, 1);
100 void hfs_bnode_clear(struct hfs_bnode *node, int off, int len)
104 off += node->page_offset;
105 page = node->page[0];
107 memset(kmap(page) + off, 0, len);
109 set_page_dirty(page);
112 void hfs_bnode_copy(struct hfs_bnode *dst_node, int dst,
113 struct hfs_bnode *src_node, int src, int len)
115 struct hfs_btree *tree;
116 struct page *src_page, *dst_page;
118 hfs_dbg(BNODE_MOD, "copybytes: %u,%u,%u\n", dst, src, len);
121 tree = src_node->tree;
122 src += src_node->page_offset;
123 dst += dst_node->page_offset;
124 src_page = src_node->page[0];
125 dst_page = dst_node->page[0];
127 memcpy(kmap(dst_page) + dst, kmap(src_page) + src, len);
130 set_page_dirty(dst_page);
133 void hfs_bnode_move(struct hfs_bnode *node, int dst, int src, int len)
138 hfs_dbg(BNODE_MOD, "movebytes: %u,%u,%u\n", dst, src, len);
141 src += node->page_offset;
142 dst += node->page_offset;
143 page = node->page[0];
145 memmove(ptr + dst, ptr + src, len);
147 set_page_dirty(page);
150 void hfs_bnode_dump(struct hfs_bnode *node)
152 struct hfs_bnode_desc desc;
156 hfs_dbg(BNODE_MOD, "bnode: %d\n", node->this);
157 hfs_bnode_read(node, &desc, 0, sizeof(desc));
158 hfs_dbg(BNODE_MOD, "%d, %d, %d, %d, %d\n",
159 be32_to_cpu(desc.next), be32_to_cpu(desc.prev),
160 desc.type, desc.height, be16_to_cpu(desc.num_recs));
162 off = node->tree->node_size - 2;
163 for (i = be16_to_cpu(desc.num_recs); i >= 0; off -= 2, i--) {
164 key_off = hfs_bnode_read_u16(node, off);
165 hfs_dbg_cont(BNODE_MOD, " %d", key_off);
166 if (i && node->type == HFS_NODE_INDEX) {
169 if (node->tree->attributes & HFS_TREE_VARIDXKEYS)
170 tmp = (hfs_bnode_read_u8(node, key_off) | 1) + 1;
172 tmp = node->tree->max_key_len + 1;
173 hfs_dbg_cont(BNODE_MOD, " (%d,%d",
174 tmp, hfs_bnode_read_u8(node, key_off));
175 hfs_bnode_read(node, &cnid, key_off + tmp, 4);
176 hfs_dbg_cont(BNODE_MOD, ",%d)", be32_to_cpu(cnid));
177 } else if (i && node->type == HFS_NODE_LEAF) {
180 tmp = hfs_bnode_read_u8(node, key_off);
181 hfs_dbg_cont(BNODE_MOD, " (%d)", tmp);
184 hfs_dbg_cont(BNODE_MOD, "\n");
187 void hfs_bnode_unlink(struct hfs_bnode *node)
189 struct hfs_btree *tree;
190 struct hfs_bnode *tmp;
195 tmp = hfs_bnode_find(tree, node->prev);
198 tmp->next = node->next;
199 cnid = cpu_to_be32(tmp->next);
200 hfs_bnode_write(tmp, &cnid, offsetof(struct hfs_bnode_desc, next), 4);
202 } else if (node->type == HFS_NODE_LEAF)
203 tree->leaf_head = node->next;
206 tmp = hfs_bnode_find(tree, node->next);
209 tmp->prev = node->prev;
210 cnid = cpu_to_be32(tmp->prev);
211 hfs_bnode_write(tmp, &cnid, offsetof(struct hfs_bnode_desc, prev), 4);
213 } else if (node->type == HFS_NODE_LEAF)
214 tree->leaf_tail = node->prev;
217 if (!node->prev && !node->next) {
218 printk(KERN_DEBUG "hfs_btree_del_level\n");
224 set_bit(HFS_BNODE_DELETED, &node->flags);
227 static inline int hfs_bnode_hash(u32 num)
229 num = (num >> 16) + num;
231 return num & (NODE_HASH_SIZE - 1);
234 struct hfs_bnode *hfs_bnode_findhash(struct hfs_btree *tree, u32 cnid)
236 struct hfs_bnode *node;
238 if (cnid >= tree->node_count) {
239 pr_err("request for non-existent node %d in B*Tree\n", cnid);
243 for (node = tree->node_hash[hfs_bnode_hash(cnid)];
244 node; node = node->next_hash) {
245 if (node->this == cnid) {
252 static struct hfs_bnode *__hfs_bnode_create(struct hfs_btree *tree, u32 cnid)
254 struct super_block *sb;
255 struct hfs_bnode *node, *node2;
256 struct address_space *mapping;
258 int size, block, i, hash;
261 if (cnid >= tree->node_count) {
262 pr_err("request for non-existent node %d in B*Tree\n", cnid);
266 sb = tree->inode->i_sb;
267 size = sizeof(struct hfs_bnode) + tree->pages_per_bnode *
268 sizeof(struct page *);
269 node = kzalloc(size, GFP_KERNEL);
274 set_bit(HFS_BNODE_NEW, &node->flags);
275 atomic_set(&node->refcnt, 1);
276 hfs_dbg(BNODE_REFS, "new_node(%d:%d): 1\n",
277 node->tree->cnid, node->this);
278 init_waitqueue_head(&node->lock_wq);
279 spin_lock(&tree->hash_lock);
280 node2 = hfs_bnode_findhash(tree, cnid);
282 hash = hfs_bnode_hash(cnid);
283 node->next_hash = tree->node_hash[hash];
284 tree->node_hash[hash] = node;
285 tree->node_hash_cnt++;
287 spin_unlock(&tree->hash_lock);
289 wait_event(node2->lock_wq, !test_bit(HFS_BNODE_NEW, &node2->flags));
292 spin_unlock(&tree->hash_lock);
294 mapping = tree->inode->i_mapping;
295 off = (loff_t)cnid * tree->node_size;
296 block = off >> PAGE_SHIFT;
297 node->page_offset = off & ~PAGE_MASK;
298 for (i = 0; i < tree->pages_per_bnode; i++) {
299 page = read_mapping_page(mapping, block++, NULL);
302 if (PageError(page)) {
306 node->page[i] = page;
311 set_bit(HFS_BNODE_ERROR, &node->flags);
315 void hfs_bnode_unhash(struct hfs_bnode *node)
317 struct hfs_bnode **p;
319 hfs_dbg(BNODE_REFS, "remove_node(%d:%d): %d\n",
320 node->tree->cnid, node->this, atomic_read(&node->refcnt));
321 for (p = &node->tree->node_hash[hfs_bnode_hash(node->this)];
322 *p && *p != node; p = &(*p)->next_hash)
325 *p = node->next_hash;
326 node->tree->node_hash_cnt--;
329 /* Load a particular node out of a tree */
330 struct hfs_bnode *hfs_bnode_find(struct hfs_btree *tree, u32 num)
332 struct hfs_bnode *node;
333 struct hfs_bnode_desc *desc;
334 int i, rec_off, off, next_off;
335 int entry_size, key_size;
337 spin_lock(&tree->hash_lock);
338 node = hfs_bnode_findhash(tree, num);
341 spin_unlock(&tree->hash_lock);
342 wait_event(node->lock_wq, !test_bit(HFS_BNODE_NEW, &node->flags));
343 if (test_bit(HFS_BNODE_ERROR, &node->flags))
347 spin_unlock(&tree->hash_lock);
348 node = __hfs_bnode_create(tree, num);
350 return ERR_PTR(-ENOMEM);
351 if (test_bit(HFS_BNODE_ERROR, &node->flags))
353 if (!test_bit(HFS_BNODE_NEW, &node->flags))
356 desc = (struct hfs_bnode_desc *)(kmap(node->page[0]) + node->page_offset);
357 node->prev = be32_to_cpu(desc->prev);
358 node->next = be32_to_cpu(desc->next);
359 node->num_recs = be16_to_cpu(desc->num_recs);
360 node->type = desc->type;
361 node->height = desc->height;
362 kunmap(node->page[0]);
364 switch (node->type) {
365 case HFS_NODE_HEADER:
367 if (node->height != 0)
371 if (node->height != 1)
375 if (node->height <= 1 || node->height > tree->depth)
382 rec_off = tree->node_size - 2;
383 off = hfs_bnode_read_u16(node, rec_off);
384 if (off != sizeof(struct hfs_bnode_desc))
386 for (i = 1; i <= node->num_recs; off = next_off, i++) {
388 next_off = hfs_bnode_read_u16(node, rec_off);
389 if (next_off <= off ||
390 next_off > tree->node_size ||
393 entry_size = next_off - off;
394 if (node->type != HFS_NODE_INDEX &&
395 node->type != HFS_NODE_LEAF)
397 key_size = hfs_bnode_read_u8(node, off) + 1;
398 if (key_size >= entry_size /*|| key_size & 1*/)
401 clear_bit(HFS_BNODE_NEW, &node->flags);
402 wake_up(&node->lock_wq);
406 set_bit(HFS_BNODE_ERROR, &node->flags);
407 clear_bit(HFS_BNODE_NEW, &node->flags);
408 wake_up(&node->lock_wq);
410 return ERR_PTR(-EIO);
413 void hfs_bnode_free(struct hfs_bnode *node)
417 for (i = 0; i < node->tree->pages_per_bnode; i++)
419 put_page(node->page[i]);
423 struct hfs_bnode *hfs_bnode_create(struct hfs_btree *tree, u32 num)
425 struct hfs_bnode *node;
429 spin_lock(&tree->hash_lock);
430 node = hfs_bnode_findhash(tree, num);
431 spin_unlock(&tree->hash_lock);
433 pr_crit("new node %u already hashed?\n", num);
437 node = __hfs_bnode_create(tree, num);
439 return ERR_PTR(-ENOMEM);
440 if (test_bit(HFS_BNODE_ERROR, &node->flags)) {
442 return ERR_PTR(-EIO);
446 memset(kmap(*pagep) + node->page_offset, 0,
447 min((int)PAGE_SIZE, (int)tree->node_size));
448 set_page_dirty(*pagep);
450 for (i = 1; i < tree->pages_per_bnode; i++) {
451 memset(kmap(*++pagep), 0, PAGE_SIZE);
452 set_page_dirty(*pagep);
455 clear_bit(HFS_BNODE_NEW, &node->flags);
456 wake_up(&node->lock_wq);
461 void hfs_bnode_get(struct hfs_bnode *node)
464 atomic_inc(&node->refcnt);
465 hfs_dbg(BNODE_REFS, "get_node(%d:%d): %d\n",
466 node->tree->cnid, node->this,
467 atomic_read(&node->refcnt));
471 /* Dispose of resources used by a node */
472 void hfs_bnode_put(struct hfs_bnode *node)
475 struct hfs_btree *tree = node->tree;
478 hfs_dbg(BNODE_REFS, "put_node(%d:%d): %d\n",
479 node->tree->cnid, node->this,
480 atomic_read(&node->refcnt));
481 BUG_ON(!atomic_read(&node->refcnt));
482 if (!atomic_dec_and_lock(&node->refcnt, &tree->hash_lock))
484 for (i = 0; i < tree->pages_per_bnode; i++) {
487 mark_page_accessed(node->page[i]);
490 if (test_bit(HFS_BNODE_DELETED, &node->flags)) {
491 hfs_bnode_unhash(node);
492 spin_unlock(&tree->hash_lock);
494 hfs_bnode_free(node);
497 spin_unlock(&tree->hash_lock);