Merge with http://kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
[linux-2.6] / fs / hfsplus / bnode.c
1 /*
2  *  linux/fs/hfsplus/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/string.h>
12 #include <linux/slab.h>
13 #include <linux/pagemap.h>
14 #include <linux/fs.h>
15 #include <linux/swap.h>
16
17 #include "hfsplus_fs.h"
18 #include "hfsplus_raw.h"
19
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)
22 {
23         struct page **pagep;
24         int l;
25
26         off += node->page_offset;
27         pagep = node->page + (off >> PAGE_CACHE_SHIFT);
28         off &= ~PAGE_CACHE_MASK;
29
30         l = min(len, (int)PAGE_CACHE_SIZE - off);
31         memcpy(buf, kmap(*pagep) + off, l);
32         kunmap(*pagep);
33
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);
39         }
40 }
41
42 u16 hfs_bnode_read_u16(struct hfs_bnode *node, int off)
43 {
44         __be16 data;
45         // optimize later...
46         hfs_bnode_read(node, &data, off, 2);
47         return be16_to_cpu(data);
48 }
49
50 u8 hfs_bnode_read_u8(struct hfs_bnode *node, int off)
51 {
52         u8 data;
53         // optimize later...
54         hfs_bnode_read(node, &data, off, 1);
55         return data;
56 }
57
58 void hfs_bnode_read_key(struct hfs_bnode *node, void *key, int off)
59 {
60         struct hfs_btree *tree;
61         int key_len;
62
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;
69
70         hfs_bnode_read(node, key, off, key_len);
71 }
72
73 void hfs_bnode_write(struct hfs_bnode *node, void *buf, int off, int len)
74 {
75         struct page **pagep;
76         int l;
77
78         off += node->page_offset;
79         pagep = node->page + (off >> PAGE_CACHE_SHIFT);
80         off &= ~PAGE_CACHE_MASK;
81
82         l = min(len, (int)PAGE_CACHE_SIZE - off);
83         memcpy(kmap(*pagep) + off, buf, l);
84         set_page_dirty(*pagep);
85         kunmap(*pagep);
86
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);
93         }
94 }
95
96 void hfs_bnode_write_u16(struct hfs_bnode *node, int off, u16 data)
97 {
98         __be16 v = cpu_to_be16(data);
99         // optimize later...
100         hfs_bnode_write(node, &v, off, 2);
101 }
102
103 void hfs_bnode_clear(struct hfs_bnode *node, int off, int len)
104 {
105         struct page **pagep;
106         int l;
107
108         off += node->page_offset;
109         pagep = node->page + (off >> PAGE_CACHE_SHIFT);
110         off &= ~PAGE_CACHE_MASK;
111
112         l = min(len, (int)PAGE_CACHE_SIZE - off);
113         memset(kmap(*pagep) + off, 0, l);
114         set_page_dirty(*pagep);
115         kunmap(*pagep);
116
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);
122         }
123 }
124
125 void hfs_bnode_copy(struct hfs_bnode *dst_node, int dst,
126                     struct hfs_bnode *src_node, int src, int len)
127 {
128         struct hfs_btree *tree;
129         struct page **src_page, **dst_page;
130         int l;
131
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;
142
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);
149
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);
156                 }
157         } else {
158                 void *src_ptr, *dst_ptr;
159
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;
171                         }
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));
182         }
183 }
184
185 void hfs_bnode_move(struct hfs_bnode *node, int dst, int src, int len)
186 {
187         struct page **src_page, **dst_page;
188         int l;
189
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;
202
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--;
213                         }
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;
221
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;
233                                 }
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));
244                 }
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;
250
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);
257
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);
264                         }
265                 } else {
266                         void *src_ptr, *dst_ptr;
267
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;
279                                 }
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));
290                 }
291         }
292 }
293
294 void hfs_bnode_dump(struct hfs_bnode *node)
295 {
296         struct hfs_bnode_desc desc;
297         __be32 cnid;
298         int i, off, key_off;
299
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));
305
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;
312
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;
322
323                         tmp = hfs_bnode_read_u16(node, key_off);
324                         dprint(DBG_BNODE_MOD, " (%d)", tmp);
325                 }
326         }
327         dprint(DBG_BNODE_MOD, "\n");
328 }
329
330 void hfs_bnode_unlink(struct hfs_bnode *node)
331 {
332         struct hfs_btree *tree;
333         struct hfs_bnode *tmp;
334         __be32 cnid;
335
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;
347
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;
358
359         // move down?
360         if (!node->prev && !node->next) {
361                 printk("hfs_btree_del_level\n");
362         }
363         if (!node->parent) {
364                 tree->root = 0;
365                 tree->depth = 0;
366         }
367         set_bit(HFS_BNODE_DELETED, &node->flags);
368 }
369
370 static inline int hfs_bnode_hash(u32 num)
371 {
372         num = (num >> 16) + num;
373         num += num >> 8;
374         return num & (NODE_HASH_SIZE - 1);
375 }
376
377 struct hfs_bnode *hfs_bnode_findhash(struct hfs_btree *tree, u32 cnid)
378 {
379         struct hfs_bnode *node;
380
381         if (cnid >= tree->node_count) {
382                 printk("HFS+-fs: request for non-existent node %d in B*Tree\n", cnid);
383                 return NULL;
384         }
385
386         for (node = tree->node_hash[hfs_bnode_hash(cnid)];
387              node; node = node->next_hash) {
388                 if (node->this == cnid) {
389                         return node;
390                 }
391         }
392         return NULL;
393 }
394
395 static struct hfs_bnode *__hfs_bnode_create(struct hfs_btree *tree, u32 cnid)
396 {
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;
403
404         if (cnid >= tree->node_count) {
405                 printk("HFS+-fs: request for non-existent node %d in B*Tree\n", cnid);
406                 return NULL;
407         }
408
409         sb = tree->inode->i_sb;
410         size = sizeof(struct hfs_bnode) + tree->pages_per_bnode *
411                 sizeof(struct page *);
412         node = kmalloc(size, GFP_KERNEL);
413         if (!node)
414                 return NULL;
415         memset(node, 0, size);
416         node->tree = tree;
417         node->this = cnid;
418         set_bit(HFS_BNODE_NEW, &node->flags);
419         atomic_set(&node->refcnt, 1);
420         dprint(DBG_BNODE_REFS, "new_node(%d:%d): 1\n",
421                node->tree->cnid, node->this);
422         init_waitqueue_head(&node->lock_wq);
423         spin_lock(&tree->hash_lock);
424         node2 = hfs_bnode_findhash(tree, cnid);
425         if (!node2) {
426                 hash = hfs_bnode_hash(cnid);
427                 node->next_hash = tree->node_hash[hash];
428                 tree->node_hash[hash] = node;
429                 tree->node_hash_cnt++;
430         } else {
431                 spin_unlock(&tree->hash_lock);
432                 kfree(node);
433                 wait_event(node2->lock_wq, !test_bit(HFS_BNODE_NEW, &node2->flags));
434                 return node2;
435         }
436         spin_unlock(&tree->hash_lock);
437
438         mapping = tree->inode->i_mapping;
439         off = (loff_t)cnid << tree->node_size_shift;
440         block = off >> PAGE_CACHE_SHIFT;
441         node->page_offset = off & ~PAGE_CACHE_MASK;
442         for (i = 0; i < tree->pages_per_bnode; block++, i++) {
443                 page = read_cache_page(mapping, block, (filler_t *)mapping->a_ops->readpage, NULL);
444                 if (IS_ERR(page))
445                         goto fail;
446                 if (PageError(page)) {
447                         page_cache_release(page);
448                         goto fail;
449                 }
450                 page_cache_release(page);
451                 node->page[i] = page;
452         }
453
454         return node;
455 fail:
456         set_bit(HFS_BNODE_ERROR, &node->flags);
457         return node;
458 }
459
460 void hfs_bnode_unhash(struct hfs_bnode *node)
461 {
462         struct hfs_bnode **p;
463
464         dprint(DBG_BNODE_REFS, "remove_node(%d:%d): %d\n",
465                 node->tree->cnid, node->this, atomic_read(&node->refcnt));
466         for (p = &node->tree->node_hash[hfs_bnode_hash(node->this)];
467              *p && *p != node; p = &(*p)->next_hash)
468                 ;
469         if (!*p)
470                 BUG();
471         *p = node->next_hash;
472         node->tree->node_hash_cnt--;
473 }
474
475 /* Load a particular node out of a tree */
476 struct hfs_bnode *hfs_bnode_find(struct hfs_btree *tree, u32 num)
477 {
478         struct hfs_bnode *node;
479         struct hfs_bnode_desc *desc;
480         int i, rec_off, off, next_off;
481         int entry_size, key_size;
482
483         spin_lock(&tree->hash_lock);
484         node = hfs_bnode_findhash(tree, num);
485         if (node) {
486                 hfs_bnode_get(node);
487                 spin_unlock(&tree->hash_lock);
488                 wait_event(node->lock_wq, !test_bit(HFS_BNODE_NEW, &node->flags));
489                 if (test_bit(HFS_BNODE_ERROR, &node->flags))
490                         goto node_error;
491                 return node;
492         }
493         spin_unlock(&tree->hash_lock);
494         node = __hfs_bnode_create(tree, num);
495         if (!node)
496                 return ERR_PTR(-ENOMEM);
497         if (test_bit(HFS_BNODE_ERROR, &node->flags))
498                 goto node_error;
499         if (!test_bit(HFS_BNODE_NEW, &node->flags))
500                 return node;
501
502         desc = (struct hfs_bnode_desc *)(kmap(node->page[0]) + node->page_offset);
503         node->prev = be32_to_cpu(desc->prev);
504         node->next = be32_to_cpu(desc->next);
505         node->num_recs = be16_to_cpu(desc->num_recs);
506         node->type = desc->type;
507         node->height = desc->height;
508         kunmap(node->page[0]);
509
510         switch (node->type) {
511         case HFS_NODE_HEADER:
512         case HFS_NODE_MAP:
513                 if (node->height != 0)
514                         goto node_error;
515                 break;
516         case HFS_NODE_LEAF:
517                 if (node->height != 1)
518                         goto node_error;
519                 break;
520         case HFS_NODE_INDEX:
521                 if (node->height <= 1 || node->height > tree->depth)
522                         goto node_error;
523                 break;
524         default:
525                 goto node_error;
526         }
527
528         rec_off = tree->node_size - 2;
529         off = hfs_bnode_read_u16(node, rec_off);
530         if (off != sizeof(struct hfs_bnode_desc))
531                 goto node_error;
532         for (i = 1; i <= node->num_recs; off = next_off, i++) {
533                 rec_off -= 2;
534                 next_off = hfs_bnode_read_u16(node, rec_off);
535                 if (next_off <= off ||
536                     next_off > tree->node_size ||
537                     next_off & 1)
538                         goto node_error;
539                 entry_size = next_off - off;
540                 if (node->type != HFS_NODE_INDEX &&
541                     node->type != HFS_NODE_LEAF)
542                         continue;
543                 key_size = hfs_bnode_read_u16(node, off) + 2;
544                 if (key_size >= entry_size || key_size & 1)
545                         goto node_error;
546         }
547         clear_bit(HFS_BNODE_NEW, &node->flags);
548         wake_up(&node->lock_wq);
549         return node;
550
551 node_error:
552         set_bit(HFS_BNODE_ERROR, &node->flags);
553         clear_bit(HFS_BNODE_NEW, &node->flags);
554         wake_up(&node->lock_wq);
555         hfs_bnode_put(node);
556         return ERR_PTR(-EIO);
557 }
558
559 void hfs_bnode_free(struct hfs_bnode *node)
560 {
561         //int i;
562
563         //for (i = 0; i < node->tree->pages_per_bnode; i++)
564         //      if (node->page[i])
565         //              page_cache_release(node->page[i]);
566         kfree(node);
567 }
568
569 struct hfs_bnode *hfs_bnode_create(struct hfs_btree *tree, u32 num)
570 {
571         struct hfs_bnode *node;
572         struct page **pagep;
573         int i;
574
575         spin_lock(&tree->hash_lock);
576         node = hfs_bnode_findhash(tree, num);
577         spin_unlock(&tree->hash_lock);
578         if (node) {
579                 printk("new node %u already hashed?\n", num);
580                 BUG();
581         }
582         node = __hfs_bnode_create(tree, num);
583         if (!node)
584                 return ERR_PTR(-ENOMEM);
585         if (test_bit(HFS_BNODE_ERROR, &node->flags)) {
586                 hfs_bnode_put(node);
587                 return ERR_PTR(-EIO);
588         }
589
590         pagep = node->page;
591         memset(kmap(*pagep) + node->page_offset, 0,
592                min((int)PAGE_CACHE_SIZE, (int)tree->node_size));
593         set_page_dirty(*pagep);
594         kunmap(*pagep);
595         for (i = 1; i < tree->pages_per_bnode; i++) {
596                 memset(kmap(*++pagep), 0, PAGE_CACHE_SIZE);
597                 set_page_dirty(*pagep);
598                 kunmap(*pagep);
599         }
600         clear_bit(HFS_BNODE_NEW, &node->flags);
601         wake_up(&node->lock_wq);
602
603         return node;
604 }
605
606 void hfs_bnode_get(struct hfs_bnode *node)
607 {
608         if (node) {
609                 atomic_inc(&node->refcnt);
610                 dprint(DBG_BNODE_REFS, "get_node(%d:%d): %d\n",
611                        node->tree->cnid, node->this, atomic_read(&node->refcnt));
612         }
613 }
614
615 /* Dispose of resources used by a node */
616 void hfs_bnode_put(struct hfs_bnode *node)
617 {
618         if (node) {
619                 struct hfs_btree *tree = node->tree;
620                 int i;
621
622                 dprint(DBG_BNODE_REFS, "put_node(%d:%d): %d\n",
623                        node->tree->cnid, node->this, atomic_read(&node->refcnt));
624                 if (!atomic_read(&node->refcnt))
625                         BUG();
626                 if (!atomic_dec_and_lock(&node->refcnt, &tree->hash_lock))
627                         return;
628                 for (i = 0; i < tree->pages_per_bnode; i++) {
629                         if (!node->page[i])
630                                 continue;
631                         mark_page_accessed(node->page[i]);
632                 }
633
634                 if (test_bit(HFS_BNODE_DELETED, &node->flags)) {
635                         hfs_bnode_unhash(node);
636                         spin_unlock(&tree->hash_lock);
637                         hfs_bmap_free(node);
638                         hfs_bnode_free(node);
639                         return;
640                 }
641                 spin_unlock(&tree->hash_lock);
642         }
643 }
644