Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 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 | #define REF_PAGES 0 |
| 17 | |
| 18 | void hfs_bnode_read(struct hfs_bnode *node, void *buf, |
| 19 | int off, int len) |
| 20 | { |
| 21 | struct page *page; |
| 22 | |
| 23 | off += node->page_offset; |
| 24 | page = node->page[0]; |
| 25 | |
| 26 | memcpy(buf, kmap(page) + off, len); |
| 27 | kunmap(page); |
| 28 | } |
| 29 | |
| 30 | u16 hfs_bnode_read_u16(struct hfs_bnode *node, int off) |
| 31 | { |
| 32 | __be16 data; |
| 33 | // optimize later... |
| 34 | hfs_bnode_read(node, &data, off, 2); |
| 35 | return be16_to_cpu(data); |
| 36 | } |
| 37 | |
| 38 | u8 hfs_bnode_read_u8(struct hfs_bnode *node, int off) |
| 39 | { |
| 40 | u8 data; |
| 41 | // optimize later... |
| 42 | hfs_bnode_read(node, &data, off, 1); |
| 43 | return data; |
| 44 | } |
| 45 | |
| 46 | void hfs_bnode_read_key(struct hfs_bnode *node, void *key, int off) |
| 47 | { |
| 48 | struct hfs_btree *tree; |
| 49 | int key_len; |
| 50 | |
| 51 | tree = node->tree; |
| 52 | if (node->type == HFS_NODE_LEAF || |
| 53 | tree->attributes & HFS_TREE_VARIDXKEYS) |
| 54 | key_len = hfs_bnode_read_u8(node, off) + 1; |
| 55 | else |
| 56 | key_len = tree->max_key_len + 1; |
| 57 | |
| 58 | hfs_bnode_read(node, key, off, key_len); |
| 59 | } |
| 60 | |
| 61 | void hfs_bnode_write(struct hfs_bnode *node, void *buf, int off, int len) |
| 62 | { |
| 63 | struct page *page; |
| 64 | |
| 65 | off += node->page_offset; |
| 66 | page = node->page[0]; |
| 67 | |
| 68 | memcpy(kmap(page) + off, buf, len); |
| 69 | kunmap(page); |
| 70 | set_page_dirty(page); |
| 71 | } |
| 72 | |
| 73 | void hfs_bnode_write_u16(struct hfs_bnode *node, int off, u16 data) |
| 74 | { |
| 75 | __be16 v = cpu_to_be16(data); |
| 76 | // optimize later... |
| 77 | hfs_bnode_write(node, &v, off, 2); |
| 78 | } |
| 79 | |
| 80 | void hfs_bnode_write_u8(struct hfs_bnode *node, int off, u8 data) |
| 81 | { |
| 82 | // optimize later... |
| 83 | hfs_bnode_write(node, &data, off, 1); |
| 84 | } |
| 85 | |
| 86 | void hfs_bnode_clear(struct hfs_bnode *node, int off, int len) |
| 87 | { |
| 88 | struct page *page; |
| 89 | |
| 90 | off += node->page_offset; |
| 91 | page = node->page[0]; |
| 92 | |
| 93 | memset(kmap(page) + off, 0, len); |
| 94 | kunmap(page); |
| 95 | set_page_dirty(page); |
| 96 | } |
| 97 | |
| 98 | void hfs_bnode_copy(struct hfs_bnode *dst_node, int dst, |
| 99 | struct hfs_bnode *src_node, int src, int len) |
| 100 | { |
| 101 | struct hfs_btree *tree; |
| 102 | struct page *src_page, *dst_page; |
| 103 | |
| 104 | dprint(DBG_BNODE_MOD, "copybytes: %u,%u,%u\n", dst, src, len); |
| 105 | if (!len) |
| 106 | return; |
| 107 | tree = src_node->tree; |
| 108 | src += src_node->page_offset; |
| 109 | dst += dst_node->page_offset; |
| 110 | src_page = src_node->page[0]; |
| 111 | dst_page = dst_node->page[0]; |
| 112 | |
| 113 | memcpy(kmap(dst_page) + dst, kmap(src_page) + src, len); |
| 114 | kunmap(src_page); |
| 115 | kunmap(dst_page); |
| 116 | set_page_dirty(dst_page); |
| 117 | } |
| 118 | |
| 119 | void hfs_bnode_move(struct hfs_bnode *node, int dst, int src, int len) |
| 120 | { |
| 121 | struct page *page; |
| 122 | void *ptr; |
| 123 | |
| 124 | dprint(DBG_BNODE_MOD, "movebytes: %u,%u,%u\n", dst, src, len); |
| 125 | if (!len) |
| 126 | return; |
| 127 | src += node->page_offset; |
| 128 | dst += node->page_offset; |
| 129 | page = node->page[0]; |
| 130 | ptr = kmap(page); |
| 131 | memmove(ptr + dst, ptr + src, len); |
| 132 | kunmap(page); |
| 133 | set_page_dirty(page); |
| 134 | } |
| 135 | |
| 136 | void hfs_bnode_dump(struct hfs_bnode *node) |
| 137 | { |
| 138 | struct hfs_bnode_desc desc; |
| 139 | __be32 cnid; |
| 140 | int i, off, key_off; |
| 141 | |
| 142 | dprint(DBG_BNODE_MOD, "bnode: %d\n", node->this); |
| 143 | hfs_bnode_read(node, &desc, 0, sizeof(desc)); |
| 144 | dprint(DBG_BNODE_MOD, "%d, %d, %d, %d, %d\n", |
| 145 | be32_to_cpu(desc.next), be32_to_cpu(desc.prev), |
| 146 | desc.type, desc.height, be16_to_cpu(desc.num_recs)); |
| 147 | |
| 148 | off = node->tree->node_size - 2; |
| 149 | for (i = be16_to_cpu(desc.num_recs); i >= 0; off -= 2, i--) { |
| 150 | key_off = hfs_bnode_read_u16(node, off); |
| 151 | dprint(DBG_BNODE_MOD, " %d", key_off); |
| 152 | if (i && node->type == HFS_NODE_INDEX) { |
| 153 | int tmp; |
| 154 | |
| 155 | if (node->tree->attributes & HFS_TREE_VARIDXKEYS) |
| 156 | tmp = (hfs_bnode_read_u8(node, key_off) | 1) + 1; |
| 157 | else |
| 158 | tmp = node->tree->max_key_len + 1; |
| 159 | dprint(DBG_BNODE_MOD, " (%d,%d", tmp, hfs_bnode_read_u8(node, key_off)); |
| 160 | hfs_bnode_read(node, &cnid, key_off + tmp, 4); |
| 161 | dprint(DBG_BNODE_MOD, ",%d)", be32_to_cpu(cnid)); |
| 162 | } else if (i && node->type == HFS_NODE_LEAF) { |
| 163 | int tmp; |
| 164 | |
| 165 | tmp = hfs_bnode_read_u8(node, key_off); |
| 166 | dprint(DBG_BNODE_MOD, " (%d)", tmp); |
| 167 | } |
| 168 | } |
| 169 | dprint(DBG_BNODE_MOD, "\n"); |
| 170 | } |
| 171 | |
| 172 | void hfs_bnode_unlink(struct hfs_bnode *node) |
| 173 | { |
| 174 | struct hfs_btree *tree; |
| 175 | struct hfs_bnode *tmp; |
| 176 | __be32 cnid; |
| 177 | |
| 178 | tree = node->tree; |
| 179 | if (node->prev) { |
| 180 | tmp = hfs_bnode_find(tree, node->prev); |
| 181 | if (IS_ERR(tmp)) |
| 182 | return; |
| 183 | tmp->next = node->next; |
| 184 | cnid = cpu_to_be32(tmp->next); |
| 185 | hfs_bnode_write(tmp, &cnid, offsetof(struct hfs_bnode_desc, next), 4); |
| 186 | hfs_bnode_put(tmp); |
| 187 | } else if (node->type == HFS_NODE_LEAF) |
| 188 | tree->leaf_head = node->next; |
| 189 | |
| 190 | if (node->next) { |
| 191 | tmp = hfs_bnode_find(tree, node->next); |
| 192 | if (IS_ERR(tmp)) |
| 193 | return; |
| 194 | tmp->prev = node->prev; |
| 195 | cnid = cpu_to_be32(tmp->prev); |
| 196 | hfs_bnode_write(tmp, &cnid, offsetof(struct hfs_bnode_desc, prev), 4); |
| 197 | hfs_bnode_put(tmp); |
| 198 | } else if (node->type == HFS_NODE_LEAF) |
| 199 | tree->leaf_tail = node->prev; |
| 200 | |
| 201 | // move down? |
| 202 | if (!node->prev && !node->next) { |
| 203 | printk("hfs_btree_del_level\n"); |
| 204 | } |
| 205 | if (!node->parent) { |
| 206 | tree->root = 0; |
| 207 | tree->depth = 0; |
| 208 | } |
| 209 | set_bit(HFS_BNODE_DELETED, &node->flags); |
| 210 | } |
| 211 | |
| 212 | static inline int hfs_bnode_hash(u32 num) |
| 213 | { |
| 214 | num = (num >> 16) + num; |
| 215 | num += num >> 8; |
| 216 | return num & (NODE_HASH_SIZE - 1); |
| 217 | } |
| 218 | |
| 219 | struct hfs_bnode *hfs_bnode_findhash(struct hfs_btree *tree, u32 cnid) |
| 220 | { |
| 221 | struct hfs_bnode *node; |
| 222 | |
| 223 | if (cnid >= tree->node_count) { |
| 224 | printk("HFS: request for non-existent node %d in B*Tree\n", cnid); |
| 225 | return NULL; |
| 226 | } |
| 227 | |
| 228 | for (node = tree->node_hash[hfs_bnode_hash(cnid)]; |
| 229 | node; node = node->next_hash) { |
| 230 | if (node->this == cnid) { |
| 231 | return node; |
| 232 | } |
| 233 | } |
| 234 | return NULL; |
| 235 | } |
| 236 | |
| 237 | static struct hfs_bnode *__hfs_bnode_create(struct hfs_btree *tree, u32 cnid) |
| 238 | { |
| 239 | struct super_block *sb; |
| 240 | struct hfs_bnode *node, *node2; |
| 241 | struct address_space *mapping; |
| 242 | struct page *page; |
| 243 | int size, block, i, hash; |
| 244 | loff_t off; |
| 245 | |
| 246 | if (cnid >= tree->node_count) { |
| 247 | printk("HFS: request for non-existent node %d in B*Tree\n", cnid); |
| 248 | return NULL; |
| 249 | } |
| 250 | |
| 251 | sb = tree->inode->i_sb; |
| 252 | size = sizeof(struct hfs_bnode) + tree->pages_per_bnode * |
| 253 | sizeof(struct page *); |
| 254 | node = kmalloc(size, GFP_KERNEL); |
| 255 | if (!node) |
| 256 | return NULL; |
| 257 | memset(node, 0, size); |
| 258 | node->tree = tree; |
| 259 | node->this = cnid; |
| 260 | set_bit(HFS_BNODE_NEW, &node->flags); |
| 261 | atomic_set(&node->refcnt, 1); |
| 262 | dprint(DBG_BNODE_REFS, "new_node(%d:%d): 1\n", |
| 263 | node->tree->cnid, node->this); |
| 264 | init_waitqueue_head(&node->lock_wq); |
| 265 | spin_lock(&tree->hash_lock); |
| 266 | node2 = hfs_bnode_findhash(tree, cnid); |
| 267 | if (!node2) { |
| 268 | hash = hfs_bnode_hash(cnid); |
| 269 | node->next_hash = tree->node_hash[hash]; |
| 270 | tree->node_hash[hash] = node; |
| 271 | tree->node_hash_cnt++; |
| 272 | } else { |
| 273 | spin_unlock(&tree->hash_lock); |
| 274 | kfree(node); |
| 275 | wait_event(node2->lock_wq, !test_bit(HFS_BNODE_NEW, &node2->flags)); |
| 276 | return node2; |
| 277 | } |
| 278 | spin_unlock(&tree->hash_lock); |
| 279 | |
| 280 | mapping = tree->inode->i_mapping; |
| 281 | off = (loff_t)cnid * tree->node_size; |
| 282 | block = off >> PAGE_CACHE_SHIFT; |
| 283 | node->page_offset = off & ~PAGE_CACHE_MASK; |
| 284 | for (i = 0; i < tree->pages_per_bnode; i++) { |
| 285 | page = read_cache_page(mapping, block++, (filler_t *)mapping->a_ops->readpage, NULL); |
| 286 | if (IS_ERR(page)) |
| 287 | goto fail; |
| 288 | if (PageError(page)) { |
| 289 | page_cache_release(page); |
| 290 | goto fail; |
| 291 | } |
| 292 | #if !REF_PAGES |
| 293 | page_cache_release(page); |
| 294 | #endif |
| 295 | node->page[i] = page; |
| 296 | } |
| 297 | |
| 298 | return node; |
| 299 | fail: |
| 300 | set_bit(HFS_BNODE_ERROR, &node->flags); |
| 301 | return node; |
| 302 | } |
| 303 | |
| 304 | void hfs_bnode_unhash(struct hfs_bnode *node) |
| 305 | { |
| 306 | struct hfs_bnode **p; |
| 307 | |
| 308 | dprint(DBG_BNODE_REFS, "remove_node(%d:%d): %d\n", |
| 309 | node->tree->cnid, node->this, atomic_read(&node->refcnt)); |
| 310 | for (p = &node->tree->node_hash[hfs_bnode_hash(node->this)]; |
| 311 | *p && *p != node; p = &(*p)->next_hash) |
| 312 | ; |
| 313 | if (!*p) |
| 314 | BUG(); |
| 315 | *p = node->next_hash; |
| 316 | node->tree->node_hash_cnt--; |
| 317 | } |
| 318 | |
| 319 | /* Load a particular node out of a tree */ |
| 320 | struct hfs_bnode *hfs_bnode_find(struct hfs_btree *tree, u32 num) |
| 321 | { |
| 322 | struct hfs_bnode *node; |
| 323 | struct hfs_bnode_desc *desc; |
| 324 | int i, rec_off, off, next_off; |
| 325 | int entry_size, key_size; |
| 326 | |
| 327 | spin_lock(&tree->hash_lock); |
| 328 | node = hfs_bnode_findhash(tree, num); |
| 329 | if (node) { |
| 330 | hfs_bnode_get(node); |
| 331 | spin_unlock(&tree->hash_lock); |
| 332 | wait_event(node->lock_wq, !test_bit(HFS_BNODE_NEW, &node->flags)); |
| 333 | if (test_bit(HFS_BNODE_ERROR, &node->flags)) |
| 334 | goto node_error; |
| 335 | return node; |
| 336 | } |
| 337 | spin_unlock(&tree->hash_lock); |
| 338 | node = __hfs_bnode_create(tree, num); |
| 339 | if (!node) |
| 340 | return ERR_PTR(-ENOMEM); |
| 341 | if (test_bit(HFS_BNODE_ERROR, &node->flags)) |
| 342 | goto node_error; |
| 343 | if (!test_bit(HFS_BNODE_NEW, &node->flags)) |
| 344 | return node; |
| 345 | |
| 346 | desc = (struct hfs_bnode_desc *)(kmap(node->page[0]) + node->page_offset); |
| 347 | node->prev = be32_to_cpu(desc->prev); |
| 348 | node->next = be32_to_cpu(desc->next); |
| 349 | node->num_recs = be16_to_cpu(desc->num_recs); |
| 350 | node->type = desc->type; |
| 351 | node->height = desc->height; |
| 352 | kunmap(node->page[0]); |
| 353 | |
| 354 | switch (node->type) { |
| 355 | case HFS_NODE_HEADER: |
| 356 | case HFS_NODE_MAP: |
| 357 | if (node->height != 0) |
| 358 | goto node_error; |
| 359 | break; |
| 360 | case HFS_NODE_LEAF: |
| 361 | if (node->height != 1) |
| 362 | goto node_error; |
| 363 | break; |
| 364 | case HFS_NODE_INDEX: |
| 365 | if (node->height <= 1 || node->height > tree->depth) |
| 366 | goto node_error; |
| 367 | break; |
| 368 | default: |
| 369 | goto node_error; |
| 370 | } |
| 371 | |
| 372 | rec_off = tree->node_size - 2; |
| 373 | off = hfs_bnode_read_u16(node, rec_off); |
| 374 | if (off != sizeof(struct hfs_bnode_desc)) |
| 375 | goto node_error; |
| 376 | for (i = 1; i <= node->num_recs; off = next_off, i++) { |
| 377 | rec_off -= 2; |
| 378 | next_off = hfs_bnode_read_u16(node, rec_off); |
| 379 | if (next_off <= off || |
| 380 | next_off > tree->node_size || |
| 381 | next_off & 1) |
| 382 | goto node_error; |
| 383 | entry_size = next_off - off; |
| 384 | if (node->type != HFS_NODE_INDEX && |
| 385 | node->type != HFS_NODE_LEAF) |
| 386 | continue; |
| 387 | key_size = hfs_bnode_read_u8(node, off) + 1; |
| 388 | if (key_size >= entry_size /*|| key_size & 1*/) |
| 389 | goto node_error; |
| 390 | } |
| 391 | clear_bit(HFS_BNODE_NEW, &node->flags); |
| 392 | wake_up(&node->lock_wq); |
| 393 | return node; |
| 394 | |
| 395 | node_error: |
| 396 | set_bit(HFS_BNODE_ERROR, &node->flags); |
| 397 | clear_bit(HFS_BNODE_NEW, &node->flags); |
| 398 | wake_up(&node->lock_wq); |
| 399 | hfs_bnode_put(node); |
| 400 | return ERR_PTR(-EIO); |
| 401 | } |
| 402 | |
| 403 | void hfs_bnode_free(struct hfs_bnode *node) |
| 404 | { |
| 405 | //int i; |
| 406 | |
| 407 | //for (i = 0; i < node->tree->pages_per_bnode; i++) |
| 408 | // if (node->page[i]) |
| 409 | // page_cache_release(node->page[i]); |
| 410 | kfree(node); |
| 411 | } |
| 412 | |
| 413 | struct hfs_bnode *hfs_bnode_create(struct hfs_btree *tree, u32 num) |
| 414 | { |
| 415 | struct hfs_bnode *node; |
| 416 | struct page **pagep; |
| 417 | int i; |
| 418 | |
| 419 | spin_lock(&tree->hash_lock); |
| 420 | node = hfs_bnode_findhash(tree, num); |
| 421 | spin_unlock(&tree->hash_lock); |
| 422 | if (node) |
| 423 | BUG(); |
| 424 | node = __hfs_bnode_create(tree, num); |
| 425 | if (!node) |
| 426 | return ERR_PTR(-ENOMEM); |
| 427 | if (test_bit(HFS_BNODE_ERROR, &node->flags)) { |
| 428 | hfs_bnode_put(node); |
| 429 | return ERR_PTR(-EIO); |
| 430 | } |
| 431 | |
| 432 | pagep = node->page; |
| 433 | memset(kmap(*pagep) + node->page_offset, 0, |
| 434 | min((int)PAGE_CACHE_SIZE, (int)tree->node_size)); |
| 435 | set_page_dirty(*pagep); |
| 436 | kunmap(*pagep); |
| 437 | for (i = 1; i < tree->pages_per_bnode; i++) { |
| 438 | memset(kmap(*++pagep), 0, PAGE_CACHE_SIZE); |
| 439 | set_page_dirty(*pagep); |
| 440 | kunmap(*pagep); |
| 441 | } |
| 442 | clear_bit(HFS_BNODE_NEW, &node->flags); |
| 443 | wake_up(&node->lock_wq); |
| 444 | |
| 445 | return node; |
| 446 | } |
| 447 | |
| 448 | void hfs_bnode_get(struct hfs_bnode *node) |
| 449 | { |
| 450 | if (node) { |
| 451 | atomic_inc(&node->refcnt); |
| 452 | #if REF_PAGES |
| 453 | { |
| 454 | int i; |
| 455 | for (i = 0; i < node->tree->pages_per_bnode; i++) |
| 456 | get_page(node->page[i]); |
| 457 | } |
| 458 | #endif |
| 459 | dprint(DBG_BNODE_REFS, "get_node(%d:%d): %d\n", |
| 460 | node->tree->cnid, node->this, atomic_read(&node->refcnt)); |
| 461 | } |
| 462 | } |
| 463 | |
| 464 | /* Dispose of resources used by a node */ |
| 465 | void hfs_bnode_put(struct hfs_bnode *node) |
| 466 | { |
| 467 | if (node) { |
| 468 | struct hfs_btree *tree = node->tree; |
| 469 | int i; |
| 470 | |
| 471 | dprint(DBG_BNODE_REFS, "put_node(%d:%d): %d\n", |
| 472 | node->tree->cnid, node->this, atomic_read(&node->refcnt)); |
| 473 | if (!atomic_read(&node->refcnt)) |
| 474 | BUG(); |
| 475 | if (!atomic_dec_and_lock(&node->refcnt, &tree->hash_lock)) { |
| 476 | #if REF_PAGES |
| 477 | for (i = 0; i < tree->pages_per_bnode; i++) |
| 478 | put_page(node->page[i]); |
| 479 | #endif |
| 480 | return; |
| 481 | } |
| 482 | for (i = 0; i < tree->pages_per_bnode; i++) { |
Roman Zippel | 74f9c9c | 2005-08-01 21:11:41 -0700 | [diff] [blame] | 483 | if (!node->page[i]) |
| 484 | continue; |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 485 | mark_page_accessed(node->page[i]); |
| 486 | #if REF_PAGES |
| 487 | put_page(node->page[i]); |
| 488 | #endif |
| 489 | } |
| 490 | |
| 491 | if (test_bit(HFS_BNODE_DELETED, &node->flags)) { |
| 492 | hfs_bnode_unhash(node); |
| 493 | spin_unlock(&tree->hash_lock); |
| 494 | hfs_bmap_free(node); |
| 495 | hfs_bnode_free(node); |
| 496 | return; |
| 497 | } |
| 498 | spin_unlock(&tree->hash_lock); |
| 499 | } |
| 500 | } |