Phillip Lougher | 1701aec | 2009-01-05 08:46:24 +0000 | [diff] [blame] | 1 | /* |
| 2 | * Squashfs - a compressed read only filesystem for Linux |
| 3 | * |
| 4 | * Copyright (c) 2002, 2003, 2004, 2005, 2006, 2007, 2008 |
| 5 | * Phillip Lougher <phillip@lougher.demon.co.uk> |
| 6 | * |
| 7 | * This program is free software; you can redistribute it and/or |
| 8 | * modify it under the terms of the GNU General Public License |
| 9 | * as published by the Free Software Foundation; either version 2, |
| 10 | * or (at your option) any later version. |
| 11 | * |
| 12 | * This program is distributed in the hope that it will be useful, |
| 13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 15 | * GNU General Public License for more details. |
| 16 | * |
| 17 | * You should have received a copy of the GNU General Public License |
| 18 | * along with this program; if not, write to the Free Software |
| 19 | * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. |
| 20 | * |
| 21 | * file.c |
| 22 | */ |
| 23 | |
| 24 | /* |
| 25 | * This file contains code for handling regular files. A regular file |
| 26 | * consists of a sequence of contiguous compressed blocks, and/or a |
| 27 | * compressed fragment block (tail-end packed block). The compressed size |
| 28 | * of each datablock is stored in a block list contained within the |
| 29 | * file inode (itself stored in one or more compressed metadata blocks). |
| 30 | * |
| 31 | * To speed up access to datablocks when reading 'large' files (256 Mbytes or |
| 32 | * larger), the code implements an index cache that caches the mapping from |
| 33 | * block index to datablock location on disk. |
| 34 | * |
| 35 | * The index cache allows Squashfs to handle large files (up to 1.75 TiB) while |
| 36 | * retaining a simple and space-efficient block list on disk. The cache |
| 37 | * is split into slots, caching up to eight 224 GiB files (128 KiB blocks). |
| 38 | * Larger files use multiple slots, with 1.75 TiB files using all 8 slots. |
| 39 | * The index cache is designed to be memory efficient, and by default uses |
| 40 | * 16 KiB. |
| 41 | */ |
| 42 | |
| 43 | #include <linux/fs.h> |
| 44 | #include <linux/vfs.h> |
| 45 | #include <linux/kernel.h> |
| 46 | #include <linux/slab.h> |
| 47 | #include <linux/string.h> |
| 48 | #include <linux/pagemap.h> |
| 49 | #include <linux/mutex.h> |
| 50 | #include <linux/zlib.h> |
| 51 | |
| 52 | #include "squashfs_fs.h" |
| 53 | #include "squashfs_fs_sb.h" |
| 54 | #include "squashfs_fs_i.h" |
| 55 | #include "squashfs.h" |
| 56 | |
| 57 | /* |
| 58 | * Locate cache slot in range [offset, index] for specified inode. If |
| 59 | * there's more than one return the slot closest to index. |
| 60 | */ |
| 61 | static struct meta_index *locate_meta_index(struct inode *inode, int offset, |
| 62 | int index) |
| 63 | { |
| 64 | struct meta_index *meta = NULL; |
| 65 | struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info; |
| 66 | int i; |
| 67 | |
| 68 | mutex_lock(&msblk->meta_index_mutex); |
| 69 | |
| 70 | TRACE("locate_meta_index: index %d, offset %d\n", index, offset); |
| 71 | |
| 72 | if (msblk->meta_index == NULL) |
| 73 | goto not_allocated; |
| 74 | |
| 75 | for (i = 0; i < SQUASHFS_META_SLOTS; i++) { |
| 76 | if (msblk->meta_index[i].inode_number == inode->i_ino && |
| 77 | msblk->meta_index[i].offset >= offset && |
| 78 | msblk->meta_index[i].offset <= index && |
| 79 | msblk->meta_index[i].locked == 0) { |
| 80 | TRACE("locate_meta_index: entry %d, offset %d\n", i, |
| 81 | msblk->meta_index[i].offset); |
| 82 | meta = &msblk->meta_index[i]; |
| 83 | offset = meta->offset; |
| 84 | } |
| 85 | } |
| 86 | |
| 87 | if (meta) |
| 88 | meta->locked = 1; |
| 89 | |
| 90 | not_allocated: |
| 91 | mutex_unlock(&msblk->meta_index_mutex); |
| 92 | |
| 93 | return meta; |
| 94 | } |
| 95 | |
| 96 | |
| 97 | /* |
| 98 | * Find and initialise an empty cache slot for index offset. |
| 99 | */ |
| 100 | static struct meta_index *empty_meta_index(struct inode *inode, int offset, |
| 101 | int skip) |
| 102 | { |
| 103 | struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info; |
| 104 | struct meta_index *meta = NULL; |
| 105 | int i; |
| 106 | |
| 107 | mutex_lock(&msblk->meta_index_mutex); |
| 108 | |
| 109 | TRACE("empty_meta_index: offset %d, skip %d\n", offset, skip); |
| 110 | |
| 111 | if (msblk->meta_index == NULL) { |
| 112 | /* |
| 113 | * First time cache index has been used, allocate and |
| 114 | * initialise. The cache index could be allocated at |
| 115 | * mount time but doing it here means it is allocated only |
| 116 | * if a 'large' file is read. |
| 117 | */ |
| 118 | msblk->meta_index = kcalloc(SQUASHFS_META_SLOTS, |
| 119 | sizeof(*(msblk->meta_index)), GFP_KERNEL); |
| 120 | if (msblk->meta_index == NULL) { |
| 121 | ERROR("Failed to allocate meta_index\n"); |
| 122 | goto failed; |
| 123 | } |
| 124 | for (i = 0; i < SQUASHFS_META_SLOTS; i++) { |
| 125 | msblk->meta_index[i].inode_number = 0; |
| 126 | msblk->meta_index[i].locked = 0; |
| 127 | } |
| 128 | msblk->next_meta_index = 0; |
| 129 | } |
| 130 | |
| 131 | for (i = SQUASHFS_META_SLOTS; i && |
| 132 | msblk->meta_index[msblk->next_meta_index].locked; i--) |
| 133 | msblk->next_meta_index = (msblk->next_meta_index + 1) % |
| 134 | SQUASHFS_META_SLOTS; |
| 135 | |
| 136 | if (i == 0) { |
| 137 | TRACE("empty_meta_index: failed!\n"); |
| 138 | goto failed; |
| 139 | } |
| 140 | |
| 141 | TRACE("empty_meta_index: returned meta entry %d, %p\n", |
| 142 | msblk->next_meta_index, |
| 143 | &msblk->meta_index[msblk->next_meta_index]); |
| 144 | |
| 145 | meta = &msblk->meta_index[msblk->next_meta_index]; |
| 146 | msblk->next_meta_index = (msblk->next_meta_index + 1) % |
| 147 | SQUASHFS_META_SLOTS; |
| 148 | |
| 149 | meta->inode_number = inode->i_ino; |
| 150 | meta->offset = offset; |
| 151 | meta->skip = skip; |
| 152 | meta->entries = 0; |
| 153 | meta->locked = 1; |
| 154 | |
| 155 | failed: |
| 156 | mutex_unlock(&msblk->meta_index_mutex); |
| 157 | return meta; |
| 158 | } |
| 159 | |
| 160 | |
| 161 | static void release_meta_index(struct inode *inode, struct meta_index *meta) |
| 162 | { |
| 163 | struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info; |
| 164 | mutex_lock(&msblk->meta_index_mutex); |
| 165 | meta->locked = 0; |
| 166 | mutex_unlock(&msblk->meta_index_mutex); |
| 167 | } |
| 168 | |
| 169 | |
| 170 | /* |
| 171 | * Read the next n blocks from the block list, starting from |
| 172 | * metadata block <start_block, offset>. |
| 173 | */ |
| 174 | static long long read_indexes(struct super_block *sb, int n, |
| 175 | u64 *start_block, int *offset) |
| 176 | { |
| 177 | int err, i; |
| 178 | long long block = 0; |
| 179 | __le32 *blist = kmalloc(PAGE_CACHE_SIZE, GFP_KERNEL); |
| 180 | |
| 181 | if (blist == NULL) { |
| 182 | ERROR("read_indexes: Failed to allocate block_list\n"); |
| 183 | return -ENOMEM; |
| 184 | } |
| 185 | |
| 186 | while (n) { |
| 187 | int blocks = min_t(int, n, PAGE_CACHE_SIZE >> 2); |
| 188 | |
| 189 | err = squashfs_read_metadata(sb, blist, start_block, |
| 190 | offset, blocks << 2); |
| 191 | if (err < 0) { |
| 192 | ERROR("read_indexes: reading block [%llx:%x]\n", |
| 193 | *start_block, *offset); |
| 194 | goto failure; |
| 195 | } |
| 196 | |
| 197 | for (i = 0; i < blocks; i++) { |
| 198 | int size = le32_to_cpu(blist[i]); |
| 199 | block += SQUASHFS_COMPRESSED_SIZE_BLOCK(size); |
| 200 | } |
| 201 | n -= blocks; |
| 202 | } |
| 203 | |
| 204 | kfree(blist); |
| 205 | return block; |
| 206 | |
| 207 | failure: |
| 208 | kfree(blist); |
| 209 | return err; |
| 210 | } |
| 211 | |
| 212 | |
| 213 | /* |
| 214 | * Each cache index slot has SQUASHFS_META_ENTRIES, each of which |
| 215 | * can cache one index -> datablock/blocklist-block mapping. We wish |
| 216 | * to distribute these over the length of the file, entry[0] maps index x, |
| 217 | * entry[1] maps index x + skip, entry[2] maps index x + 2 * skip, and so on. |
| 218 | * The larger the file, the greater the skip factor. The skip factor is |
| 219 | * limited to the size of the metadata cache (SQUASHFS_CACHED_BLKS) to ensure |
| 220 | * the number of metadata blocks that need to be read fits into the cache. |
| 221 | * If the skip factor is limited in this way then the file will use multiple |
| 222 | * slots. |
| 223 | */ |
| 224 | static inline int calculate_skip(int blocks) |
| 225 | { |
| 226 | int skip = blocks / ((SQUASHFS_META_ENTRIES + 1) |
| 227 | * SQUASHFS_META_INDEXES); |
| 228 | return min(SQUASHFS_CACHED_BLKS - 1, skip + 1); |
| 229 | } |
| 230 | |
| 231 | |
| 232 | /* |
| 233 | * Search and grow the index cache for the specified inode, returning the |
| 234 | * on-disk locations of the datablock and block list metadata block |
| 235 | * <index_block, index_offset> for index (scaled to nearest cache index). |
| 236 | */ |
| 237 | static int fill_meta_index(struct inode *inode, int index, |
| 238 | u64 *index_block, int *index_offset, u64 *data_block) |
| 239 | { |
| 240 | struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info; |
| 241 | int skip = calculate_skip(i_size_read(inode) >> msblk->block_log); |
| 242 | int offset = 0; |
| 243 | struct meta_index *meta; |
| 244 | struct meta_entry *meta_entry; |
| 245 | u64 cur_index_block = squashfs_i(inode)->block_list_start; |
| 246 | int cur_offset = squashfs_i(inode)->offset; |
| 247 | u64 cur_data_block = squashfs_i(inode)->start; |
| 248 | int err, i; |
| 249 | |
| 250 | /* |
| 251 | * Scale index to cache index (cache slot entry) |
| 252 | */ |
| 253 | index /= SQUASHFS_META_INDEXES * skip; |
| 254 | |
| 255 | while (offset < index) { |
| 256 | meta = locate_meta_index(inode, offset + 1, index); |
| 257 | |
| 258 | if (meta == NULL) { |
| 259 | meta = empty_meta_index(inode, offset + 1, skip); |
| 260 | if (meta == NULL) |
| 261 | goto all_done; |
| 262 | } else { |
| 263 | offset = index < meta->offset + meta->entries ? index : |
| 264 | meta->offset + meta->entries - 1; |
| 265 | meta_entry = &meta->meta_entry[offset - meta->offset]; |
| 266 | cur_index_block = meta_entry->index_block + |
| 267 | msblk->inode_table; |
| 268 | cur_offset = meta_entry->offset; |
| 269 | cur_data_block = meta_entry->data_block; |
| 270 | TRACE("get_meta_index: offset %d, meta->offset %d, " |
| 271 | "meta->entries %d\n", offset, meta->offset, |
| 272 | meta->entries); |
| 273 | TRACE("get_meta_index: index_block 0x%llx, offset 0x%x" |
| 274 | " data_block 0x%llx\n", cur_index_block, |
| 275 | cur_offset, cur_data_block); |
| 276 | } |
| 277 | |
| 278 | /* |
| 279 | * If necessary grow cache slot by reading block list. Cache |
| 280 | * slot is extended up to index or to the end of the slot, in |
| 281 | * which case further slots will be used. |
| 282 | */ |
| 283 | for (i = meta->offset + meta->entries; i <= index && |
| 284 | i < meta->offset + SQUASHFS_META_ENTRIES; i++) { |
| 285 | int blocks = skip * SQUASHFS_META_INDEXES; |
| 286 | long long res = read_indexes(inode->i_sb, blocks, |
| 287 | &cur_index_block, &cur_offset); |
| 288 | |
| 289 | if (res < 0) { |
| 290 | if (meta->entries == 0) |
| 291 | /* |
| 292 | * Don't leave an empty slot on read |
| 293 | * error allocated to this inode... |
| 294 | */ |
| 295 | meta->inode_number = 0; |
| 296 | err = res; |
| 297 | goto failed; |
| 298 | } |
| 299 | |
| 300 | cur_data_block += res; |
| 301 | meta_entry = &meta->meta_entry[i - meta->offset]; |
| 302 | meta_entry->index_block = cur_index_block - |
| 303 | msblk->inode_table; |
| 304 | meta_entry->offset = cur_offset; |
| 305 | meta_entry->data_block = cur_data_block; |
| 306 | meta->entries++; |
| 307 | offset++; |
| 308 | } |
| 309 | |
| 310 | TRACE("get_meta_index: meta->offset %d, meta->entries %d\n", |
| 311 | meta->offset, meta->entries); |
| 312 | |
| 313 | release_meta_index(inode, meta); |
| 314 | } |
| 315 | |
| 316 | all_done: |
| 317 | *index_block = cur_index_block; |
| 318 | *index_offset = cur_offset; |
| 319 | *data_block = cur_data_block; |
| 320 | |
| 321 | /* |
| 322 | * Scale cache index (cache slot entry) to index |
| 323 | */ |
| 324 | return offset * SQUASHFS_META_INDEXES * skip; |
| 325 | |
| 326 | failed: |
| 327 | release_meta_index(inode, meta); |
| 328 | return err; |
| 329 | } |
| 330 | |
| 331 | |
| 332 | /* |
| 333 | * Get the on-disk location and compressed size of the datablock |
| 334 | * specified by index. Fill_meta_index() does most of the work. |
| 335 | */ |
| 336 | static int read_blocklist(struct inode *inode, int index, u64 *block) |
| 337 | { |
| 338 | u64 start; |
| 339 | long long blks; |
| 340 | int offset; |
| 341 | __le32 size; |
| 342 | int res = fill_meta_index(inode, index, &start, &offset, block); |
| 343 | |
| 344 | TRACE("read_blocklist: res %d, index %d, start 0x%llx, offset" |
| 345 | " 0x%x, block 0x%llx\n", res, index, start, offset, |
| 346 | *block); |
| 347 | |
| 348 | if (res < 0) |
| 349 | return res; |
| 350 | |
| 351 | /* |
| 352 | * res contains the index of the mapping returned by fill_meta_index(), |
| 353 | * this will likely be less than the desired index (because the |
| 354 | * meta_index cache works at a higher granularity). Read any |
| 355 | * extra block indexes needed. |
| 356 | */ |
| 357 | if (res < index) { |
| 358 | blks = read_indexes(inode->i_sb, index - res, &start, &offset); |
| 359 | if (blks < 0) |
| 360 | return (int) blks; |
| 361 | *block += blks; |
| 362 | } |
| 363 | |
| 364 | /* |
| 365 | * Read length of block specified by index. |
| 366 | */ |
| 367 | res = squashfs_read_metadata(inode->i_sb, &size, &start, &offset, |
| 368 | sizeof(size)); |
| 369 | if (res < 0) |
| 370 | return res; |
| 371 | return le32_to_cpu(size); |
| 372 | } |
| 373 | |
| 374 | |
| 375 | static int squashfs_readpage(struct file *file, struct page *page) |
| 376 | { |
| 377 | struct inode *inode = page->mapping->host; |
| 378 | struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info; |
| 379 | int bytes, i, offset = 0, sparse = 0; |
| 380 | struct squashfs_cache_entry *buffer = NULL; |
| 381 | void *pageaddr; |
| 382 | |
| 383 | int mask = (1 << (msblk->block_log - PAGE_CACHE_SHIFT)) - 1; |
| 384 | int index = page->index >> (msblk->block_log - PAGE_CACHE_SHIFT); |
| 385 | int start_index = page->index & ~mask; |
| 386 | int end_index = start_index | mask; |
| 387 | int file_end = i_size_read(inode) >> msblk->block_log; |
| 388 | |
| 389 | TRACE("Entered squashfs_readpage, page index %lx, start block %llx\n", |
| 390 | page->index, squashfs_i(inode)->start); |
| 391 | |
| 392 | if (page->index >= ((i_size_read(inode) + PAGE_CACHE_SIZE - 1) >> |
| 393 | PAGE_CACHE_SHIFT)) |
| 394 | goto out; |
| 395 | |
| 396 | if (index < file_end || squashfs_i(inode)->fragment_block == |
| 397 | SQUASHFS_INVALID_BLK) { |
| 398 | /* |
| 399 | * Reading a datablock from disk. Need to read block list |
| 400 | * to get location and block size. |
| 401 | */ |
| 402 | u64 block = 0; |
| 403 | int bsize = read_blocklist(inode, index, &block); |
| 404 | if (bsize < 0) |
| 405 | goto error_out; |
| 406 | |
| 407 | if (bsize == 0) { /* hole */ |
| 408 | bytes = index == file_end ? |
| 409 | (i_size_read(inode) & (msblk->block_size - 1)) : |
| 410 | msblk->block_size; |
| 411 | sparse = 1; |
| 412 | } else { |
| 413 | /* |
| 414 | * Read and decompress datablock. |
| 415 | */ |
| 416 | buffer = squashfs_get_datablock(inode->i_sb, |
| 417 | block, bsize); |
| 418 | if (buffer->error) { |
| 419 | ERROR("Unable to read page, block %llx, size %x" |
| 420 | "\n", block, bsize); |
| 421 | squashfs_cache_put(buffer); |
| 422 | goto error_out; |
| 423 | } |
| 424 | bytes = buffer->length; |
| 425 | } |
| 426 | } else { |
| 427 | /* |
| 428 | * Datablock is stored inside a fragment (tail-end packed |
| 429 | * block). |
| 430 | */ |
| 431 | buffer = squashfs_get_fragment(inode->i_sb, |
| 432 | squashfs_i(inode)->fragment_block, |
| 433 | squashfs_i(inode)->fragment_size); |
| 434 | |
| 435 | if (buffer->error) { |
| 436 | ERROR("Unable to read page, block %llx, size %x\n", |
| 437 | squashfs_i(inode)->fragment_block, |
| 438 | squashfs_i(inode)->fragment_size); |
| 439 | squashfs_cache_put(buffer); |
| 440 | goto error_out; |
| 441 | } |
| 442 | bytes = i_size_read(inode) & (msblk->block_size - 1); |
| 443 | offset = squashfs_i(inode)->fragment_offset; |
| 444 | } |
| 445 | |
| 446 | /* |
| 447 | * Loop copying datablock into pages. As the datablock likely covers |
| 448 | * many PAGE_CACHE_SIZE pages (default block size is 128 KiB) explicitly |
| 449 | * grab the pages from the page cache, except for the page that we've |
| 450 | * been called to fill. |
| 451 | */ |
| 452 | for (i = start_index; i <= end_index && bytes > 0; i++, |
| 453 | bytes -= PAGE_CACHE_SIZE, offset += PAGE_CACHE_SIZE) { |
| 454 | struct page *push_page; |
| 455 | int avail = sparse ? 0 : min_t(int, bytes, PAGE_CACHE_SIZE); |
| 456 | |
| 457 | TRACE("bytes %d, i %d, available_bytes %d\n", bytes, i, avail); |
| 458 | |
| 459 | push_page = (i == page->index) ? page : |
| 460 | grab_cache_page_nowait(page->mapping, i); |
| 461 | |
| 462 | if (!push_page) |
| 463 | continue; |
| 464 | |
| 465 | if (PageUptodate(push_page)) |
| 466 | goto skip_page; |
| 467 | |
| 468 | pageaddr = kmap_atomic(push_page, KM_USER0); |
| 469 | squashfs_copy_data(pageaddr, buffer, offset, avail); |
| 470 | memset(pageaddr + avail, 0, PAGE_CACHE_SIZE - avail); |
| 471 | kunmap_atomic(pageaddr, KM_USER0); |
| 472 | flush_dcache_page(push_page); |
| 473 | SetPageUptodate(push_page); |
| 474 | skip_page: |
| 475 | unlock_page(push_page); |
| 476 | if (i != page->index) |
| 477 | page_cache_release(push_page); |
| 478 | } |
| 479 | |
| 480 | if (!sparse) |
| 481 | squashfs_cache_put(buffer); |
| 482 | |
| 483 | return 0; |
| 484 | |
| 485 | error_out: |
| 486 | SetPageError(page); |
| 487 | out: |
| 488 | pageaddr = kmap_atomic(page, KM_USER0); |
| 489 | memset(pageaddr, 0, PAGE_CACHE_SIZE); |
| 490 | kunmap_atomic(pageaddr, KM_USER0); |
| 491 | flush_dcache_page(page); |
| 492 | if (!PageError(page)) |
| 493 | SetPageUptodate(page); |
| 494 | unlock_page(page); |
| 495 | |
| 496 | return 0; |
| 497 | } |
| 498 | |
| 499 | |
| 500 | const struct address_space_operations squashfs_aops = { |
| 501 | .readpage = squashfs_readpage |
| 502 | }; |