[PATCH] mm: clean up pagecache allocation
[pv_ops_mirror.git] / fs / hfsplus / bnode.c
blob29da6574ba77941b22c15b4e959cd4c84b8bbee1
1 /*
2 * linux/fs/hfsplus/bnode.c
4 * Copyright (C) 2001
5 * Brad Boyer (flar@allandria.com)
6 * (C) 2003 Ardis Technologies <roman@ardistech.com>
8 * Handle basic btree node operations
9 */
11 #include <linux/string.h>
12 #include <linux/slab.h>
13 #include <linux/pagemap.h>
14 #include <linux/fs.h>
15 #include <linux/swap.h>
17 #include "hfsplus_fs.h"
18 #include "hfsplus_raw.h"
20 /* Copy a specified range of bytes from the raw data of a node */
21 void hfs_bnode_read(struct hfs_bnode *node, void *buf, int off, int len)
23 struct page **pagep;
24 int l;
26 off += node->page_offset;
27 pagep = node->page + (off >> PAGE_CACHE_SHIFT);
28 off &= ~PAGE_CACHE_MASK;
30 l = min(len, (int)PAGE_CACHE_SIZE - off);
31 memcpy(buf, kmap(*pagep) + off, l);
32 kunmap(*pagep);
34 while ((len -= l) != 0) {
35 buf += l;
36 l = min(len, (int)PAGE_CACHE_SIZE);
37 memcpy(buf, kmap(*++pagep), l);
38 kunmap(*pagep);
42 u16 hfs_bnode_read_u16(struct hfs_bnode *node, int off)
44 __be16 data;
45 // optimize later...
46 hfs_bnode_read(node, &data, off, 2);
47 return be16_to_cpu(data);
50 u8 hfs_bnode_read_u8(struct hfs_bnode *node, int off)
52 u8 data;
53 // optimize later...
54 hfs_bnode_read(node, &data, off, 1);
55 return data;
58 void hfs_bnode_read_key(struct hfs_bnode *node, void *key, int off)
60 struct hfs_btree *tree;
61 int key_len;
63 tree = node->tree;
64 if (node->type == HFS_NODE_LEAF ||
65 tree->attributes & HFS_TREE_VARIDXKEYS)
66 key_len = hfs_bnode_read_u16(node, off) + 2;
67 else
68 key_len = tree->max_key_len + 2;
70 hfs_bnode_read(node, key, off, key_len);
73 void hfs_bnode_write(struct hfs_bnode *node, void *buf, int off, int len)
75 struct page **pagep;
76 int l;
78 off += node->page_offset;
79 pagep = node->page + (off >> PAGE_CACHE_SHIFT);
80 off &= ~PAGE_CACHE_MASK;
82 l = min(len, (int)PAGE_CACHE_SIZE - off);
83 memcpy(kmap(*pagep) + off, buf, l);
84 set_page_dirty(*pagep);
85 kunmap(*pagep);
87 while ((len -= l) != 0) {
88 buf += l;
89 l = min(len, (int)PAGE_CACHE_SIZE);
90 memcpy(kmap(*++pagep), buf, l);
91 set_page_dirty(*pagep);
92 kunmap(*pagep);
96 void hfs_bnode_write_u16(struct hfs_bnode *node, int off, u16 data)
98 __be16 v = cpu_to_be16(data);
99 // optimize later...
100 hfs_bnode_write(node, &v, off, 2);
103 void hfs_bnode_clear(struct hfs_bnode *node, int off, int len)
105 struct page **pagep;
106 int l;
108 off += node->page_offset;
109 pagep = node->page + (off >> PAGE_CACHE_SHIFT);
110 off &= ~PAGE_CACHE_MASK;
112 l = min(len, (int)PAGE_CACHE_SIZE - off);
113 memset(kmap(*pagep) + off, 0, l);
114 set_page_dirty(*pagep);
115 kunmap(*pagep);
117 while ((len -= l) != 0) {
118 l = min(len, (int)PAGE_CACHE_SIZE);
119 memset(kmap(*++pagep), 0, l);
120 set_page_dirty(*pagep);
121 kunmap(*pagep);
125 void hfs_bnode_copy(struct hfs_bnode *dst_node, int dst,
126 struct hfs_bnode *src_node, int src, int len)
128 struct hfs_btree *tree;
129 struct page **src_page, **dst_page;
130 int l;
132 dprint(DBG_BNODE_MOD, "copybytes: %u,%u,%u\n", dst, src, len);
133 if (!len)
134 return;
135 tree = src_node->tree;
136 src += src_node->page_offset;
137 dst += dst_node->page_offset;
138 src_page = src_node->page + (src >> PAGE_CACHE_SHIFT);
139 src &= ~PAGE_CACHE_MASK;
140 dst_page = dst_node->page + (dst >> PAGE_CACHE_SHIFT);
141 dst &= ~PAGE_CACHE_MASK;
143 if (src == dst) {
144 l = min(len, (int)PAGE_CACHE_SIZE - src);
145 memcpy(kmap(*dst_page) + src, kmap(*src_page) + src, l);
146 kunmap(*src_page);
147 set_page_dirty(*dst_page);
148 kunmap(*dst_page);
150 while ((len -= l) != 0) {
151 l = min(len, (int)PAGE_CACHE_SIZE);
152 memcpy(kmap(*++dst_page), kmap(*++src_page), l);
153 kunmap(*src_page);
154 set_page_dirty(*dst_page);
155 kunmap(*dst_page);
157 } else {
158 void *src_ptr, *dst_ptr;
160 do {
161 src_ptr = kmap(*src_page) + src;
162 dst_ptr = kmap(*dst_page) + dst;
163 if (PAGE_CACHE_SIZE - src < PAGE_CACHE_SIZE - dst) {
164 l = PAGE_CACHE_SIZE - src;
165 src = 0;
166 dst += l;
167 } else {
168 l = PAGE_CACHE_SIZE - dst;
169 src += l;
170 dst = 0;
172 l = min(len, l);
173 memcpy(dst_ptr, src_ptr, l);
174 kunmap(*src_page);
175 set_page_dirty(*dst_page);
176 kunmap(*dst_page);
177 if (!dst)
178 dst_page++;
179 else
180 src_page++;
181 } while ((len -= l));
185 void hfs_bnode_move(struct hfs_bnode *node, int dst, int src, int len)
187 struct page **src_page, **dst_page;
188 int l;
190 dprint(DBG_BNODE_MOD, "movebytes: %u,%u,%u\n", dst, src, len);
191 if (!len)
192 return;
193 src += node->page_offset;
194 dst += node->page_offset;
195 if (dst > src) {
196 src += len - 1;
197 src_page = node->page + (src >> PAGE_CACHE_SHIFT);
198 src = (src & ~PAGE_CACHE_MASK) + 1;
199 dst += len - 1;
200 dst_page = node->page + (dst >> PAGE_CACHE_SHIFT);
201 dst = (dst & ~PAGE_CACHE_MASK) + 1;
203 if (src == dst) {
204 while (src < len) {
205 memmove(kmap(*dst_page), kmap(*src_page), src);
206 kunmap(*src_page);
207 set_page_dirty(*dst_page);
208 kunmap(*dst_page);
209 len -= src;
210 src = PAGE_CACHE_SIZE;
211 src_page--;
212 dst_page--;
214 src -= len;
215 memmove(kmap(*dst_page) + src, kmap(*src_page) + src, len);
216 kunmap(*src_page);
217 set_page_dirty(*dst_page);
218 kunmap(*dst_page);
219 } else {
220 void *src_ptr, *dst_ptr;
222 do {
223 src_ptr = kmap(*src_page) + src;
224 dst_ptr = kmap(*dst_page) + dst;
225 if (src < dst) {
226 l = src;
227 src = PAGE_CACHE_SIZE;
228 dst -= l;
229 } else {
230 l = dst;
231 src -= l;
232 dst = PAGE_CACHE_SIZE;
234 l = min(len, l);
235 memmove(dst_ptr - l, src_ptr - l, l);
236 kunmap(*src_page);
237 set_page_dirty(*dst_page);
238 kunmap(*dst_page);
239 if (dst == PAGE_CACHE_SIZE)
240 dst_page--;
241 else
242 src_page--;
243 } while ((len -= l));
245 } else {
246 src_page = node->page + (src >> PAGE_CACHE_SHIFT);
247 src &= ~PAGE_CACHE_MASK;
248 dst_page = node->page + (dst >> PAGE_CACHE_SHIFT);
249 dst &= ~PAGE_CACHE_MASK;
251 if (src == dst) {
252 l = min(len, (int)PAGE_CACHE_SIZE - src);
253 memmove(kmap(*dst_page) + src, kmap(*src_page) + src, l);
254 kunmap(*src_page);
255 set_page_dirty(*dst_page);
256 kunmap(*dst_page);
258 while ((len -= l) != 0) {
259 l = min(len, (int)PAGE_CACHE_SIZE);
260 memmove(kmap(*++dst_page), kmap(*++src_page), l);
261 kunmap(*src_page);
262 set_page_dirty(*dst_page);
263 kunmap(*dst_page);
265 } else {
266 void *src_ptr, *dst_ptr;
268 do {
269 src_ptr = kmap(*src_page) + src;
270 dst_ptr = kmap(*dst_page) + dst;
271 if (PAGE_CACHE_SIZE - src < PAGE_CACHE_SIZE - dst) {
272 l = PAGE_CACHE_SIZE - src;
273 src = 0;
274 dst += l;
275 } else {
276 l = PAGE_CACHE_SIZE - dst;
277 src += l;
278 dst = 0;
280 l = min(len, l);
281 memmove(dst_ptr, src_ptr, l);
282 kunmap(*src_page);
283 set_page_dirty(*dst_page);
284 kunmap(*dst_page);
285 if (!dst)
286 dst_page++;
287 else
288 src_page++;
289 } while ((len -= l));
294 void hfs_bnode_dump(struct hfs_bnode *node)
296 struct hfs_bnode_desc desc;
297 __be32 cnid;
298 int i, off, key_off;
300 dprint(DBG_BNODE_MOD, "bnode: %d\n", node->this);
301 hfs_bnode_read(node, &desc, 0, sizeof(desc));
302 dprint(DBG_BNODE_MOD, "%d, %d, %d, %d, %d\n",
303 be32_to_cpu(desc.next), be32_to_cpu(desc.prev),
304 desc.type, desc.height, be16_to_cpu(desc.num_recs));
306 off = node->tree->node_size - 2;
307 for (i = be16_to_cpu(desc.num_recs); i >= 0; off -= 2, i--) {
308 key_off = hfs_bnode_read_u16(node, off);
309 dprint(DBG_BNODE_MOD, " %d", key_off);
310 if (i && node->type == HFS_NODE_INDEX) {
311 int tmp;
313 if (node->tree->attributes & HFS_TREE_VARIDXKEYS)
314 tmp = hfs_bnode_read_u16(node, key_off) + 2;
315 else
316 tmp = node->tree->max_key_len + 2;
317 dprint(DBG_BNODE_MOD, " (%d", tmp);
318 hfs_bnode_read(node, &cnid, key_off + tmp, 4);
319 dprint(DBG_BNODE_MOD, ",%d)", be32_to_cpu(cnid));
320 } else if (i && node->type == HFS_NODE_LEAF) {
321 int tmp;
323 tmp = hfs_bnode_read_u16(node, key_off);
324 dprint(DBG_BNODE_MOD, " (%d)", tmp);
327 dprint(DBG_BNODE_MOD, "\n");
330 void hfs_bnode_unlink(struct hfs_bnode *node)
332 struct hfs_btree *tree;
333 struct hfs_bnode *tmp;
334 __be32 cnid;
336 tree = node->tree;
337 if (node->prev) {
338 tmp = hfs_bnode_find(tree, node->prev);
339 if (IS_ERR(tmp))
340 return;
341 tmp->next = node->next;
342 cnid = cpu_to_be32(tmp->next);
343 hfs_bnode_write(tmp, &cnid, offsetof(struct hfs_bnode_desc, next), 4);
344 hfs_bnode_put(tmp);
345 } else if (node->type == HFS_NODE_LEAF)
346 tree->leaf_head = node->next;
348 if (node->next) {
349 tmp = hfs_bnode_find(tree, node->next);
350 if (IS_ERR(tmp))
351 return;
352 tmp->prev = node->prev;
353 cnid = cpu_to_be32(tmp->prev);
354 hfs_bnode_write(tmp, &cnid, offsetof(struct hfs_bnode_desc, prev), 4);
355 hfs_bnode_put(tmp);
356 } else if (node->type == HFS_NODE_LEAF)
357 tree->leaf_tail = node->prev;
359 // move down?
360 if (!node->prev && !node->next) {
361 printk(KERN_DEBUG "hfs_btree_del_level\n");
363 if (!node->parent) {
364 tree->root = 0;
365 tree->depth = 0;
367 set_bit(HFS_BNODE_DELETED, &node->flags);
370 static inline int hfs_bnode_hash(u32 num)
372 num = (num >> 16) + num;
373 num += num >> 8;
374 return num & (NODE_HASH_SIZE - 1);
377 struct hfs_bnode *hfs_bnode_findhash(struct hfs_btree *tree, u32 cnid)
379 struct hfs_bnode *node;
381 if (cnid >= tree->node_count) {
382 printk(KERN_ERR "hfs: request for non-existent node %d in B*Tree\n", cnid);
383 return NULL;
386 for (node = tree->node_hash[hfs_bnode_hash(cnid)];
387 node; node = node->next_hash) {
388 if (node->this == cnid) {
389 return node;
392 return NULL;
395 static struct hfs_bnode *__hfs_bnode_create(struct hfs_btree *tree, u32 cnid)
397 struct super_block *sb;
398 struct hfs_bnode *node, *node2;
399 struct address_space *mapping;
400 struct page *page;
401 int size, block, i, hash;
402 loff_t off;
404 if (cnid >= tree->node_count) {
405 printk(KERN_ERR "hfs: request for non-existent node %d in B*Tree\n", cnid);
406 return NULL;
409 sb = tree->inode->i_sb;
410 size = sizeof(struct hfs_bnode) + tree->pages_per_bnode *
411 sizeof(struct page *);
412 node = kzalloc(size, GFP_KERNEL);
413 if (!node)
414 return NULL;
415 node->tree = tree;
416 node->this = cnid;
417 set_bit(HFS_BNODE_NEW, &node->flags);
418 atomic_set(&node->refcnt, 1);
419 dprint(DBG_BNODE_REFS, "new_node(%d:%d): 1\n",
420 node->tree->cnid, node->this);
421 init_waitqueue_head(&node->lock_wq);
422 spin_lock(&tree->hash_lock);
423 node2 = hfs_bnode_findhash(tree, cnid);
424 if (!node2) {
425 hash = hfs_bnode_hash(cnid);
426 node->next_hash = tree->node_hash[hash];
427 tree->node_hash[hash] = node;
428 tree->node_hash_cnt++;
429 } else {
430 spin_unlock(&tree->hash_lock);
431 kfree(node);
432 wait_event(node2->lock_wq, !test_bit(HFS_BNODE_NEW, &node2->flags));
433 return node2;
435 spin_unlock(&tree->hash_lock);
437 mapping = tree->inode->i_mapping;
438 off = (loff_t)cnid << tree->node_size_shift;
439 block = off >> PAGE_CACHE_SHIFT;
440 node->page_offset = off & ~PAGE_CACHE_MASK;
441 for (i = 0; i < tree->pages_per_bnode; block++, i++) {
442 page = read_mapping_page(mapping, block, NULL);
443 if (IS_ERR(page))
444 goto fail;
445 if (PageError(page)) {
446 page_cache_release(page);
447 goto fail;
449 page_cache_release(page);
450 node->page[i] = page;
453 return node;
454 fail:
455 set_bit(HFS_BNODE_ERROR, &node->flags);
456 return node;
459 void hfs_bnode_unhash(struct hfs_bnode *node)
461 struct hfs_bnode **p;
463 dprint(DBG_BNODE_REFS, "remove_node(%d:%d): %d\n",
464 node->tree->cnid, node->this, atomic_read(&node->refcnt));
465 for (p = &node->tree->node_hash[hfs_bnode_hash(node->this)];
466 *p && *p != node; p = &(*p)->next_hash)
468 BUG_ON(!*p);
469 *p = node->next_hash;
470 node->tree->node_hash_cnt--;
473 /* Load a particular node out of a tree */
474 struct hfs_bnode *hfs_bnode_find(struct hfs_btree *tree, u32 num)
476 struct hfs_bnode *node;
477 struct hfs_bnode_desc *desc;
478 int i, rec_off, off, next_off;
479 int entry_size, key_size;
481 spin_lock(&tree->hash_lock);
482 node = hfs_bnode_findhash(tree, num);
483 if (node) {
484 hfs_bnode_get(node);
485 spin_unlock(&tree->hash_lock);
486 wait_event(node->lock_wq, !test_bit(HFS_BNODE_NEW, &node->flags));
487 if (test_bit(HFS_BNODE_ERROR, &node->flags))
488 goto node_error;
489 return node;
491 spin_unlock(&tree->hash_lock);
492 node = __hfs_bnode_create(tree, num);
493 if (!node)
494 return ERR_PTR(-ENOMEM);
495 if (test_bit(HFS_BNODE_ERROR, &node->flags))
496 goto node_error;
497 if (!test_bit(HFS_BNODE_NEW, &node->flags))
498 return node;
500 desc = (struct hfs_bnode_desc *)(kmap(node->page[0]) + node->page_offset);
501 node->prev = be32_to_cpu(desc->prev);
502 node->next = be32_to_cpu(desc->next);
503 node->num_recs = be16_to_cpu(desc->num_recs);
504 node->type = desc->type;
505 node->height = desc->height;
506 kunmap(node->page[0]);
508 switch (node->type) {
509 case HFS_NODE_HEADER:
510 case HFS_NODE_MAP:
511 if (node->height != 0)
512 goto node_error;
513 break;
514 case HFS_NODE_LEAF:
515 if (node->height != 1)
516 goto node_error;
517 break;
518 case HFS_NODE_INDEX:
519 if (node->height <= 1 || node->height > tree->depth)
520 goto node_error;
521 break;
522 default:
523 goto node_error;
526 rec_off = tree->node_size - 2;
527 off = hfs_bnode_read_u16(node, rec_off);
528 if (off != sizeof(struct hfs_bnode_desc))
529 goto node_error;
530 for (i = 1; i <= node->num_recs; off = next_off, i++) {
531 rec_off -= 2;
532 next_off = hfs_bnode_read_u16(node, rec_off);
533 if (next_off <= off ||
534 next_off > tree->node_size ||
535 next_off & 1)
536 goto node_error;
537 entry_size = next_off - off;
538 if (node->type != HFS_NODE_INDEX &&
539 node->type != HFS_NODE_LEAF)
540 continue;
541 key_size = hfs_bnode_read_u16(node, off) + 2;
542 if (key_size >= entry_size || key_size & 1)
543 goto node_error;
545 clear_bit(HFS_BNODE_NEW, &node->flags);
546 wake_up(&node->lock_wq);
547 return node;
549 node_error:
550 set_bit(HFS_BNODE_ERROR, &node->flags);
551 clear_bit(HFS_BNODE_NEW, &node->flags);
552 wake_up(&node->lock_wq);
553 hfs_bnode_put(node);
554 return ERR_PTR(-EIO);
557 void hfs_bnode_free(struct hfs_bnode *node)
559 //int i;
561 //for (i = 0; i < node->tree->pages_per_bnode; i++)
562 // if (node->page[i])
563 // page_cache_release(node->page[i]);
564 kfree(node);
567 struct hfs_bnode *hfs_bnode_create(struct hfs_btree *tree, u32 num)
569 struct hfs_bnode *node;
570 struct page **pagep;
571 int i;
573 spin_lock(&tree->hash_lock);
574 node = hfs_bnode_findhash(tree, num);
575 spin_unlock(&tree->hash_lock);
576 if (node) {
577 printk(KERN_CRIT "new node %u already hashed?\n", num);
578 WARN_ON(1);
579 return node;
581 node = __hfs_bnode_create(tree, num);
582 if (!node)
583 return ERR_PTR(-ENOMEM);
584 if (test_bit(HFS_BNODE_ERROR, &node->flags)) {
585 hfs_bnode_put(node);
586 return ERR_PTR(-EIO);
589 pagep = node->page;
590 memset(kmap(*pagep) + node->page_offset, 0,
591 min((int)PAGE_CACHE_SIZE, (int)tree->node_size));
592 set_page_dirty(*pagep);
593 kunmap(*pagep);
594 for (i = 1; i < tree->pages_per_bnode; i++) {
595 memset(kmap(*++pagep), 0, PAGE_CACHE_SIZE);
596 set_page_dirty(*pagep);
597 kunmap(*pagep);
599 clear_bit(HFS_BNODE_NEW, &node->flags);
600 wake_up(&node->lock_wq);
602 return node;
605 void hfs_bnode_get(struct hfs_bnode *node)
607 if (node) {
608 atomic_inc(&node->refcnt);
609 dprint(DBG_BNODE_REFS, "get_node(%d:%d): %d\n",
610 node->tree->cnid, node->this, atomic_read(&node->refcnt));
614 /* Dispose of resources used by a node */
615 void hfs_bnode_put(struct hfs_bnode *node)
617 if (node) {
618 struct hfs_btree *tree = node->tree;
619 int i;
621 dprint(DBG_BNODE_REFS, "put_node(%d:%d): %d\n",
622 node->tree->cnid, node->this, atomic_read(&node->refcnt));
623 BUG_ON(!atomic_read(&node->refcnt));
624 if (!atomic_dec_and_lock(&node->refcnt, &tree->hash_lock))
625 return;
626 for (i = 0; i < tree->pages_per_bnode; i++) {
627 if (!node->page[i])
628 continue;
629 mark_page_accessed(node->page[i]);
632 if (test_bit(HFS_BNODE_DELETED, &node->flags)) {
633 hfs_bnode_unhash(node);
634 spin_unlock(&tree->hash_lock);
635 hfs_bmap_free(node);
636 hfs_bnode_free(node);
637 return;
639 spin_unlock(&tree->hash_lock);