Merge branch 'for-2.6.27' of git://git.marvell.com/mv643xx_eth into upstream-fixes
[linux-2.6] / fs / hfs / bnode.c
1 /*
2  *  linux/fs/hfs/bnode.c
3  *
4  * Copyright (C) 2001
5  * Brad Boyer (flar@allandria.com)
6  * (C) 2003 Ardis Technologies <roman@ardistech.com>
7  *
8  * Handle basic btree node operations
9  */
10
11 #include <linux/pagemap.h>
12 #include <linux/swap.h>
13
14 #include "btree.h"
15
16 void hfs_bnode_read(struct hfs_bnode *node, void *buf,
17                 int off, int len)
18 {
19         struct page *page;
20
21         off += node->page_offset;
22         page = node->page[0];
23
24         memcpy(buf, kmap(page) + off, len);
25         kunmap(page);
26 }
27
28 u16 hfs_bnode_read_u16(struct hfs_bnode *node, int off)
29 {
30         __be16 data;
31         // optimize later...
32         hfs_bnode_read(node, &data, off, 2);
33         return be16_to_cpu(data);
34 }
35
36 u8 hfs_bnode_read_u8(struct hfs_bnode *node, int off)
37 {
38         u8 data;
39         // optimize later...
40         hfs_bnode_read(node, &data, off, 1);
41         return data;
42 }
43
44 void hfs_bnode_read_key(struct hfs_bnode *node, void *key, int off)
45 {
46         struct hfs_btree *tree;
47         int key_len;
48
49         tree = node->tree;
50         if (node->type == HFS_NODE_LEAF ||
51             tree->attributes & HFS_TREE_VARIDXKEYS)
52                 key_len = hfs_bnode_read_u8(node, off) + 1;
53         else
54                 key_len = tree->max_key_len + 1;
55
56         hfs_bnode_read(node, key, off, key_len);
57 }
58
59 void hfs_bnode_write(struct hfs_bnode *node, void *buf, int off, int len)
60 {
61         struct page *page;
62
63         off += node->page_offset;
64         page = node->page[0];
65
66         memcpy(kmap(page) + off, buf, len);
67         kunmap(page);
68         set_page_dirty(page);
69 }
70
71 void hfs_bnode_write_u16(struct hfs_bnode *node, int off, u16 data)
72 {
73         __be16 v = cpu_to_be16(data);
74         // optimize later...
75         hfs_bnode_write(node, &v, off, 2);
76 }
77
78 void hfs_bnode_write_u8(struct hfs_bnode *node, int off, u8 data)
79 {
80         // optimize later...
81         hfs_bnode_write(node, &data, off, 1);
82 }
83
84 void hfs_bnode_clear(struct hfs_bnode *node, int off, int len)
85 {
86         struct page *page;
87
88         off += node->page_offset;
89         page = node->page[0];
90
91         memset(kmap(page) + off, 0, len);
92         kunmap(page);
93         set_page_dirty(page);
94 }
95
96 void hfs_bnode_copy(struct hfs_bnode *dst_node, int dst,
97                 struct hfs_bnode *src_node, int src, int len)
98 {
99         struct hfs_btree *tree;
100         struct page *src_page, *dst_page;
101
102         dprint(DBG_BNODE_MOD, "copybytes: %u,%u,%u\n", dst, src, len);
103         if (!len)
104                 return;
105         tree = src_node->tree;
106         src += src_node->page_offset;
107         dst += dst_node->page_offset;
108         src_page = src_node->page[0];
109         dst_page = dst_node->page[0];
110
111         memcpy(kmap(dst_page) + dst, kmap(src_page) + src, len);
112         kunmap(src_page);
113         kunmap(dst_page);
114         set_page_dirty(dst_page);
115 }
116
117 void hfs_bnode_move(struct hfs_bnode *node, int dst, int src, int len)
118 {
119         struct page *page;
120         void *ptr;
121
122         dprint(DBG_BNODE_MOD, "movebytes: %u,%u,%u\n", dst, src, len);
123         if (!len)
124                 return;
125         src += node->page_offset;
126         dst += node->page_offset;
127         page = node->page[0];
128         ptr = kmap(page);
129         memmove(ptr + dst, ptr + src, len);
130         kunmap(page);
131         set_page_dirty(page);
132 }
133
134 void hfs_bnode_dump(struct hfs_bnode *node)
135 {
136         struct hfs_bnode_desc desc;
137         __be32 cnid;
138         int i, off, key_off;
139
140         dprint(DBG_BNODE_MOD, "bnode: %d\n", node->this);
141         hfs_bnode_read(node, &desc, 0, sizeof(desc));
142         dprint(DBG_BNODE_MOD, "%d, %d, %d, %d, %d\n",
143                 be32_to_cpu(desc.next), be32_to_cpu(desc.prev),
144                 desc.type, desc.height, be16_to_cpu(desc.num_recs));
145
146         off = node->tree->node_size - 2;
147         for (i = be16_to_cpu(desc.num_recs); i >= 0; off -= 2, i--) {
148                 key_off = hfs_bnode_read_u16(node, off);
149                 dprint(DBG_BNODE_MOD, " %d", key_off);
150                 if (i && node->type == HFS_NODE_INDEX) {
151                         int tmp;
152
153                         if (node->tree->attributes & HFS_TREE_VARIDXKEYS)
154                                 tmp = (hfs_bnode_read_u8(node, key_off) | 1) + 1;
155                         else
156                                 tmp = node->tree->max_key_len + 1;
157                         dprint(DBG_BNODE_MOD, " (%d,%d", tmp, hfs_bnode_read_u8(node, key_off));
158                         hfs_bnode_read(node, &cnid, key_off + tmp, 4);
159                         dprint(DBG_BNODE_MOD, ",%d)", be32_to_cpu(cnid));
160                 } else if (i && node->type == HFS_NODE_LEAF) {
161                         int tmp;
162
163                         tmp = hfs_bnode_read_u8(node, key_off);
164                         dprint(DBG_BNODE_MOD, " (%d)", tmp);
165                 }
166         }
167         dprint(DBG_BNODE_MOD, "\n");
168 }
169
170 void hfs_bnode_unlink(struct hfs_bnode *node)
171 {
172         struct hfs_btree *tree;
173         struct hfs_bnode *tmp;
174         __be32 cnid;
175
176         tree = node->tree;
177         if (node->prev) {
178                 tmp = hfs_bnode_find(tree, node->prev);
179                 if (IS_ERR(tmp))
180                         return;
181                 tmp->next = node->next;
182                 cnid = cpu_to_be32(tmp->next);
183                 hfs_bnode_write(tmp, &cnid, offsetof(struct hfs_bnode_desc, next), 4);
184                 hfs_bnode_put(tmp);
185         } else if (node->type == HFS_NODE_LEAF)
186                 tree->leaf_head = node->next;
187
188         if (node->next) {
189                 tmp = hfs_bnode_find(tree, node->next);
190                 if (IS_ERR(tmp))
191                         return;
192                 tmp->prev = node->prev;
193                 cnid = cpu_to_be32(tmp->prev);
194                 hfs_bnode_write(tmp, &cnid, offsetof(struct hfs_bnode_desc, prev), 4);
195                 hfs_bnode_put(tmp);
196         } else if (node->type == HFS_NODE_LEAF)
197                 tree->leaf_tail = node->prev;
198
199         // move down?
200         if (!node->prev && !node->next) {
201                 printk(KERN_DEBUG "hfs_btree_del_level\n");
202         }
203         if (!node->parent) {
204                 tree->root = 0;
205                 tree->depth = 0;
206         }
207         set_bit(HFS_BNODE_DELETED, &node->flags);
208 }
209
210 static inline int hfs_bnode_hash(u32 num)
211 {
212         num = (num >> 16) + num;
213         num += num >> 8;
214         return num & (NODE_HASH_SIZE - 1);
215 }
216
217 struct hfs_bnode *hfs_bnode_findhash(struct hfs_btree *tree, u32 cnid)
218 {
219         struct hfs_bnode *node;
220
221         if (cnid >= tree->node_count) {
222                 printk(KERN_ERR "hfs: request for non-existent node %d in B*Tree\n", cnid);
223                 return NULL;
224         }
225
226         for (node = tree->node_hash[hfs_bnode_hash(cnid)];
227              node; node = node->next_hash) {
228                 if (node->this == cnid) {
229                         return node;
230                 }
231         }
232         return NULL;
233 }
234
235 static struct hfs_bnode *__hfs_bnode_create(struct hfs_btree *tree, u32 cnid)
236 {
237         struct super_block *sb;
238         struct hfs_bnode *node, *node2;
239         struct address_space *mapping;
240         struct page *page;
241         int size, block, i, hash;
242         loff_t off;
243
244         if (cnid >= tree->node_count) {
245                 printk(KERN_ERR "hfs: request for non-existent node %d in B*Tree\n", cnid);
246                 return NULL;
247         }
248
249         sb = tree->inode->i_sb;
250         size = sizeof(struct hfs_bnode) + tree->pages_per_bnode *
251                 sizeof(struct page *);
252         node = kzalloc(size, GFP_KERNEL);
253         if (!node)
254                 return NULL;
255         node->tree = tree;
256         node->this = cnid;
257         set_bit(HFS_BNODE_NEW, &node->flags);
258         atomic_set(&node->refcnt, 1);
259         dprint(DBG_BNODE_REFS, "new_node(%d:%d): 1\n",
260                node->tree->cnid, node->this);
261         init_waitqueue_head(&node->lock_wq);
262         spin_lock(&tree->hash_lock);
263         node2 = hfs_bnode_findhash(tree, cnid);
264         if (!node2) {
265                 hash = hfs_bnode_hash(cnid);
266                 node->next_hash = tree->node_hash[hash];
267                 tree->node_hash[hash] = node;
268                 tree->node_hash_cnt++;
269         } else {
270                 spin_unlock(&tree->hash_lock);
271                 kfree(node);
272                 wait_event(node2->lock_wq, !test_bit(HFS_BNODE_NEW, &node2->flags));
273                 return node2;
274         }
275         spin_unlock(&tree->hash_lock);
276
277         mapping = tree->inode->i_mapping;
278         off = (loff_t)cnid * tree->node_size;
279         block = off >> PAGE_CACHE_SHIFT;
280         node->page_offset = off & ~PAGE_CACHE_MASK;
281         for (i = 0; i < tree->pages_per_bnode; i++) {
282                 page = read_mapping_page(mapping, block++, NULL);
283                 if (IS_ERR(page))
284                         goto fail;
285                 if (PageError(page)) {
286                         page_cache_release(page);
287                         goto fail;
288                 }
289                 page_cache_release(page);
290                 node->page[i] = page;
291         }
292
293         return node;
294 fail:
295         set_bit(HFS_BNODE_ERROR, &node->flags);
296         return node;
297 }
298
299 void hfs_bnode_unhash(struct hfs_bnode *node)
300 {
301         struct hfs_bnode **p;
302
303         dprint(DBG_BNODE_REFS, "remove_node(%d:%d): %d\n",
304                 node->tree->cnid, node->this, atomic_read(&node->refcnt));
305         for (p = &node->tree->node_hash[hfs_bnode_hash(node->this)];
306              *p && *p != node; p = &(*p)->next_hash)
307                 ;
308         BUG_ON(!*p);
309         *p = node->next_hash;
310         node->tree->node_hash_cnt--;
311 }
312
313 /* Load a particular node out of a tree */
314 struct hfs_bnode *hfs_bnode_find(struct hfs_btree *tree, u32 num)
315 {
316         struct hfs_bnode *node;
317         struct hfs_bnode_desc *desc;
318         int i, rec_off, off, next_off;
319         int entry_size, key_size;
320
321         spin_lock(&tree->hash_lock);
322         node = hfs_bnode_findhash(tree, num);
323         if (node) {
324                 hfs_bnode_get(node);
325                 spin_unlock(&tree->hash_lock);
326                 wait_event(node->lock_wq, !test_bit(HFS_BNODE_NEW, &node->flags));
327                 if (test_bit(HFS_BNODE_ERROR, &node->flags))
328                         goto node_error;
329                 return node;
330         }
331         spin_unlock(&tree->hash_lock);
332         node = __hfs_bnode_create(tree, num);
333         if (!node)
334                 return ERR_PTR(-ENOMEM);
335         if (test_bit(HFS_BNODE_ERROR, &node->flags))
336                 goto node_error;
337         if (!test_bit(HFS_BNODE_NEW, &node->flags))
338                 return node;
339
340         desc = (struct hfs_bnode_desc *)(kmap(node->page[0]) + node->page_offset);
341         node->prev = be32_to_cpu(desc->prev);
342         node->next = be32_to_cpu(desc->next);
343         node->num_recs = be16_to_cpu(desc->num_recs);
344         node->type = desc->type;
345         node->height = desc->height;
346         kunmap(node->page[0]);
347
348         switch (node->type) {
349         case HFS_NODE_HEADER:
350         case HFS_NODE_MAP:
351                 if (node->height != 0)
352                         goto node_error;
353                 break;
354         case HFS_NODE_LEAF:
355                 if (node->height != 1)
356                         goto node_error;
357                 break;
358         case HFS_NODE_INDEX:
359                 if (node->height <= 1 || node->height > tree->depth)
360                         goto node_error;
361                 break;
362         default:
363                 goto node_error;
364         }
365
366         rec_off = tree->node_size - 2;
367         off = hfs_bnode_read_u16(node, rec_off);
368         if (off != sizeof(struct hfs_bnode_desc))
369                 goto node_error;
370         for (i = 1; i <= node->num_recs; off = next_off, i++) {
371                 rec_off -= 2;
372                 next_off = hfs_bnode_read_u16(node, rec_off);
373                 if (next_off <= off ||
374                     next_off > tree->node_size ||
375                     next_off & 1)
376                         goto node_error;
377                 entry_size = next_off - off;
378                 if (node->type != HFS_NODE_INDEX &&
379                     node->type != HFS_NODE_LEAF)
380                         continue;
381                 key_size = hfs_bnode_read_u8(node, off) + 1;
382                 if (key_size >= entry_size /*|| key_size & 1*/)
383                         goto node_error;
384         }
385         clear_bit(HFS_BNODE_NEW, &node->flags);
386         wake_up(&node->lock_wq);
387         return node;
388
389 node_error:
390         set_bit(HFS_BNODE_ERROR, &node->flags);
391         clear_bit(HFS_BNODE_NEW, &node->flags);
392         wake_up(&node->lock_wq);
393         hfs_bnode_put(node);
394         return ERR_PTR(-EIO);
395 }
396
397 void hfs_bnode_free(struct hfs_bnode *node)
398 {
399         //int i;
400
401         //for (i = 0; i < node->tree->pages_per_bnode; i++)
402         //      if (node->page[i])
403         //              page_cache_release(node->page[i]);
404         kfree(node);
405 }
406
407 struct hfs_bnode *hfs_bnode_create(struct hfs_btree *tree, u32 num)
408 {
409         struct hfs_bnode *node;
410         struct page **pagep;
411         int i;
412
413         spin_lock(&tree->hash_lock);
414         node = hfs_bnode_findhash(tree, num);
415         spin_unlock(&tree->hash_lock);
416         BUG_ON(node);
417         node = __hfs_bnode_create(tree, num);
418         if (!node)
419                 return ERR_PTR(-ENOMEM);
420         if (test_bit(HFS_BNODE_ERROR, &node->flags)) {
421                 hfs_bnode_put(node);
422                 return ERR_PTR(-EIO);
423         }
424
425         pagep = node->page;
426         memset(kmap(*pagep) + node->page_offset, 0,
427                min((int)PAGE_CACHE_SIZE, (int)tree->node_size));
428         set_page_dirty(*pagep);
429         kunmap(*pagep);
430         for (i = 1; i < tree->pages_per_bnode; i++) {
431                 memset(kmap(*++pagep), 0, PAGE_CACHE_SIZE);
432                 set_page_dirty(*pagep);
433                 kunmap(*pagep);
434         }
435         clear_bit(HFS_BNODE_NEW, &node->flags);
436         wake_up(&node->lock_wq);
437
438         return node;
439 }
440
441 void hfs_bnode_get(struct hfs_bnode *node)
442 {
443         if (node) {
444                 atomic_inc(&node->refcnt);
445                 dprint(DBG_BNODE_REFS, "get_node(%d:%d): %d\n",
446                        node->tree->cnid, node->this, atomic_read(&node->refcnt));
447         }
448 }
449
450 /* Dispose of resources used by a node */
451 void hfs_bnode_put(struct hfs_bnode *node)
452 {
453         if (node) {
454                 struct hfs_btree *tree = node->tree;
455                 int i;
456
457                 dprint(DBG_BNODE_REFS, "put_node(%d:%d): %d\n",
458                        node->tree->cnid, node->this, atomic_read(&node->refcnt));
459                 BUG_ON(!atomic_read(&node->refcnt));
460                 if (!atomic_dec_and_lock(&node->refcnt, &tree->hash_lock))
461                         return;
462                 for (i = 0; i < tree->pages_per_bnode; i++) {
463                         if (!node->page[i])
464                                 continue;
465                         mark_page_accessed(node->page[i]);
466                 }
467
468                 if (test_bit(HFS_BNODE_DELETED, &node->flags)) {
469                         hfs_bnode_unhash(node);
470                         spin_unlock(&tree->hash_lock);
471                         hfs_bmap_free(node);
472                         hfs_bnode_free(node);
473                         return;
474                 }
475                 spin_unlock(&tree->hash_lock);
476         }
477 }