xref: /linux/fs/squashfs/file.c (revision 0d456bad36d42d16022be045c8a53ddbb59ee478)
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@squashfs.org.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 
51 #include "squashfs_fs.h"
52 #include "squashfs_fs_sb.h"
53 #include "squashfs_fs_i.h"
54 #include "squashfs.h"
55 
56 /*
57  * Locate cache slot in range [offset, index] for specified inode.  If
58  * there's more than one return the slot closest to index.
59  */
60 static struct meta_index *locate_meta_index(struct inode *inode, int offset,
61 				int index)
62 {
63 	struct meta_index *meta = NULL;
64 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
65 	int i;
66 
67 	mutex_lock(&msblk->meta_index_mutex);
68 
69 	TRACE("locate_meta_index: index %d, offset %d\n", index, offset);
70 
71 	if (msblk->meta_index == NULL)
72 		goto not_allocated;
73 
74 	for (i = 0; i < SQUASHFS_META_SLOTS; i++) {
75 		if (msblk->meta_index[i].inode_number == inode->i_ino &&
76 				msblk->meta_index[i].offset >= offset &&
77 				msblk->meta_index[i].offset <= index &&
78 				msblk->meta_index[i].locked == 0) {
79 			TRACE("locate_meta_index: entry %d, offset %d\n", i,
80 					msblk->meta_index[i].offset);
81 			meta = &msblk->meta_index[i];
82 			offset = meta->offset;
83 		}
84 	}
85 
86 	if (meta)
87 		meta->locked = 1;
88 
89 not_allocated:
90 	mutex_unlock(&msblk->meta_index_mutex);
91 
92 	return meta;
93 }
94 
95 
96 /*
97  * Find and initialise an empty cache slot for index offset.
98  */
99 static struct meta_index *empty_meta_index(struct inode *inode, int offset,
100 				int skip)
101 {
102 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
103 	struct meta_index *meta = NULL;
104 	int i;
105 
106 	mutex_lock(&msblk->meta_index_mutex);
107 
108 	TRACE("empty_meta_index: offset %d, skip %d\n", offset, skip);
109 
110 	if (msblk->meta_index == NULL) {
111 		/*
112 		 * First time cache index has been used, allocate and
113 		 * initialise.  The cache index could be allocated at
114 		 * mount time but doing it here means it is allocated only
115 		 * if a 'large' file is read.
116 		 */
117 		msblk->meta_index = kcalloc(SQUASHFS_META_SLOTS,
118 			sizeof(*(msblk->meta_index)), GFP_KERNEL);
119 		if (msblk->meta_index == NULL) {
120 			ERROR("Failed to allocate meta_index\n");
121 			goto failed;
122 		}
123 		for (i = 0; i < SQUASHFS_META_SLOTS; i++) {
124 			msblk->meta_index[i].inode_number = 0;
125 			msblk->meta_index[i].locked = 0;
126 		}
127 		msblk->next_meta_index = 0;
128 	}
129 
130 	for (i = SQUASHFS_META_SLOTS; i &&
131 			msblk->meta_index[msblk->next_meta_index].locked; i--)
132 		msblk->next_meta_index = (msblk->next_meta_index + 1) %
133 			SQUASHFS_META_SLOTS;
134 
135 	if (i == 0) {
136 		TRACE("empty_meta_index: failed!\n");
137 		goto failed;
138 	}
139 
140 	TRACE("empty_meta_index: returned meta entry %d, %p\n",
141 			msblk->next_meta_index,
142 			&msblk->meta_index[msblk->next_meta_index]);
143 
144 	meta = &msblk->meta_index[msblk->next_meta_index];
145 	msblk->next_meta_index = (msblk->next_meta_index + 1) %
146 			SQUASHFS_META_SLOTS;
147 
148 	meta->inode_number = inode->i_ino;
149 	meta->offset = offset;
150 	meta->skip = skip;
151 	meta->entries = 0;
152 	meta->locked = 1;
153 
154 failed:
155 	mutex_unlock(&msblk->meta_index_mutex);
156 	return meta;
157 }
158 
159 
160 static void release_meta_index(struct inode *inode, struct meta_index *meta)
161 {
162 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
163 	mutex_lock(&msblk->meta_index_mutex);
164 	meta->locked = 0;
165 	mutex_unlock(&msblk->meta_index_mutex);
166 }
167 
168 
169 /*
170  * Read the next n blocks from the block list, starting from
171  * metadata block <start_block, offset>.
172  */
173 static long long read_indexes(struct super_block *sb, int n,
174 				u64 *start_block, int *offset)
175 {
176 	int err, i;
177 	long long block = 0;
178 	__le32 *blist = kmalloc(PAGE_CACHE_SIZE, GFP_KERNEL);
179 
180 	if (blist == NULL) {
181 		ERROR("read_indexes: Failed to allocate block_list\n");
182 		return -ENOMEM;
183 	}
184 
185 	while (n) {
186 		int blocks = min_t(int, n, PAGE_CACHE_SIZE >> 2);
187 
188 		err = squashfs_read_metadata(sb, blist, start_block,
189 				offset, blocks << 2);
190 		if (err < 0) {
191 			ERROR("read_indexes: reading block [%llx:%x]\n",
192 				*start_block, *offset);
193 			goto failure;
194 		}
195 
196 		for (i = 0; i < blocks; i++) {
197 			int size = le32_to_cpu(blist[i]);
198 			block += SQUASHFS_COMPRESSED_SIZE_BLOCK(size);
199 		}
200 		n -= blocks;
201 	}
202 
203 	kfree(blist);
204 	return block;
205 
206 failure:
207 	kfree(blist);
208 	return err;
209 }
210 
211 
212 /*
213  * Each cache index slot has SQUASHFS_META_ENTRIES, each of which
214  * can cache one index -> datablock/blocklist-block mapping.  We wish
215  * to distribute these over the length of the file, entry[0] maps index x,
216  * entry[1] maps index x + skip, entry[2] maps index x + 2 * skip, and so on.
217  * The larger the file, the greater the skip factor.  The skip factor is
218  * limited to the size of the metadata cache (SQUASHFS_CACHED_BLKS) to ensure
219  * the number of metadata blocks that need to be read fits into the cache.
220  * If the skip factor is limited in this way then the file will use multiple
221  * slots.
222  */
223 static inline int calculate_skip(int blocks)
224 {
225 	int skip = blocks / ((SQUASHFS_META_ENTRIES + 1)
226 		 * SQUASHFS_META_INDEXES);
227 	return min(SQUASHFS_CACHED_BLKS - 1, skip + 1);
228 }
229 
230 
231 /*
232  * Search and grow the index cache for the specified inode, returning the
233  * on-disk locations of the datablock and block list metadata block
234  * <index_block, index_offset> for index (scaled to nearest cache index).
235  */
236 static int fill_meta_index(struct inode *inode, int index,
237 		u64 *index_block, int *index_offset, u64 *data_block)
238 {
239 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
240 	int skip = calculate_skip(i_size_read(inode) >> msblk->block_log);
241 	int offset = 0;
242 	struct meta_index *meta;
243 	struct meta_entry *meta_entry;
244 	u64 cur_index_block = squashfs_i(inode)->block_list_start;
245 	int cur_offset = squashfs_i(inode)->offset;
246 	u64 cur_data_block = squashfs_i(inode)->start;
247 	int err, i;
248 
249 	/*
250 	 * Scale index to cache index (cache slot entry)
251 	 */
252 	index /= SQUASHFS_META_INDEXES * skip;
253 
254 	while (offset < index) {
255 		meta = locate_meta_index(inode, offset + 1, index);
256 
257 		if (meta == NULL) {
258 			meta = empty_meta_index(inode, offset + 1, skip);
259 			if (meta == NULL)
260 				goto all_done;
261 		} else {
262 			offset = index < meta->offset + meta->entries ? index :
263 				meta->offset + meta->entries - 1;
264 			meta_entry = &meta->meta_entry[offset - meta->offset];
265 			cur_index_block = meta_entry->index_block +
266 				msblk->inode_table;
267 			cur_offset = meta_entry->offset;
268 			cur_data_block = meta_entry->data_block;
269 			TRACE("get_meta_index: offset %d, meta->offset %d, "
270 				"meta->entries %d\n", offset, meta->offset,
271 				meta->entries);
272 			TRACE("get_meta_index: index_block 0x%llx, offset 0x%x"
273 				" data_block 0x%llx\n", cur_index_block,
274 				cur_offset, cur_data_block);
275 		}
276 
277 		/*
278 		 * If necessary grow cache slot by reading block list.  Cache
279 		 * slot is extended up to index or to the end of the slot, in
280 		 * which case further slots will be used.
281 		 */
282 		for (i = meta->offset + meta->entries; i <= index &&
283 				i < meta->offset + SQUASHFS_META_ENTRIES; i++) {
284 			int blocks = skip * SQUASHFS_META_INDEXES;
285 			long long res = read_indexes(inode->i_sb, blocks,
286 					&cur_index_block, &cur_offset);
287 
288 			if (res < 0) {
289 				if (meta->entries == 0)
290 					/*
291 					 * Don't leave an empty slot on read
292 					 * error allocated to this inode...
293 					 */
294 					meta->inode_number = 0;
295 				err = res;
296 				goto failed;
297 			}
298 
299 			cur_data_block += res;
300 			meta_entry = &meta->meta_entry[i - meta->offset];
301 			meta_entry->index_block = cur_index_block -
302 				msblk->inode_table;
303 			meta_entry->offset = cur_offset;
304 			meta_entry->data_block = cur_data_block;
305 			meta->entries++;
306 			offset++;
307 		}
308 
309 		TRACE("get_meta_index: meta->offset %d, meta->entries %d\n",
310 				meta->offset, meta->entries);
311 
312 		release_meta_index(inode, meta);
313 	}
314 
315 all_done:
316 	*index_block = cur_index_block;
317 	*index_offset = cur_offset;
318 	*data_block = cur_data_block;
319 
320 	/*
321 	 * Scale cache index (cache slot entry) to index
322 	 */
323 	return offset * SQUASHFS_META_INDEXES * skip;
324 
325 failed:
326 	release_meta_index(inode, meta);
327 	return err;
328 }
329 
330 
331 /*
332  * Get the on-disk location and compressed size of the datablock
333  * specified by index.  Fill_meta_index() does most of the work.
334  */
335 static int read_blocklist(struct inode *inode, int index, u64 *block)
336 {
337 	u64 start;
338 	long long blks;
339 	int offset;
340 	__le32 size;
341 	int res = fill_meta_index(inode, index, &start, &offset, block);
342 
343 	TRACE("read_blocklist: res %d, index %d, start 0x%llx, offset"
344 		       " 0x%x, block 0x%llx\n", res, index, start, offset,
345 			*block);
346 
347 	if (res < 0)
348 		return res;
349 
350 	/*
351 	 * res contains the index of the mapping returned by fill_meta_index(),
352 	 * this will likely be less than the desired index (because the
353 	 * meta_index cache works at a higher granularity).  Read any
354 	 * extra block indexes needed.
355 	 */
356 	if (res < index) {
357 		blks = read_indexes(inode->i_sb, index - res, &start, &offset);
358 		if (blks < 0)
359 			return (int) blks;
360 		*block += blks;
361 	}
362 
363 	/*
364 	 * Read length of block specified by index.
365 	 */
366 	res = squashfs_read_metadata(inode->i_sb, &size, &start, &offset,
367 			sizeof(size));
368 	if (res < 0)
369 		return res;
370 	return le32_to_cpu(size);
371 }
372 
373 
374 static int squashfs_readpage(struct file *file, struct page *page)
375 {
376 	struct inode *inode = page->mapping->host;
377 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
378 	int bytes, i, offset = 0, sparse = 0;
379 	struct squashfs_cache_entry *buffer = NULL;
380 	void *pageaddr;
381 
382 	int mask = (1 << (msblk->block_log - PAGE_CACHE_SHIFT)) - 1;
383 	int index = page->index >> (msblk->block_log - PAGE_CACHE_SHIFT);
384 	int start_index = page->index & ~mask;
385 	int end_index = start_index | mask;
386 	int file_end = i_size_read(inode) >> msblk->block_log;
387 
388 	TRACE("Entered squashfs_readpage, page index %lx, start block %llx\n",
389 				page->index, squashfs_i(inode)->start);
390 
391 	if (page->index >= ((i_size_read(inode) + PAGE_CACHE_SIZE - 1) >>
392 					PAGE_CACHE_SHIFT))
393 		goto out;
394 
395 	if (index < file_end || squashfs_i(inode)->fragment_block ==
396 					SQUASHFS_INVALID_BLK) {
397 		/*
398 		 * Reading a datablock from disk.  Need to read block list
399 		 * to get location and block size.
400 		 */
401 		u64 block = 0;
402 		int bsize = read_blocklist(inode, index, &block);
403 		if (bsize < 0)
404 			goto error_out;
405 
406 		if (bsize == 0) { /* hole */
407 			bytes = index == file_end ?
408 				(i_size_read(inode) & (msblk->block_size - 1)) :
409 				 msblk->block_size;
410 			sparse = 1;
411 		} else {
412 			/*
413 			 * Read and decompress datablock.
414 			 */
415 			buffer = squashfs_get_datablock(inode->i_sb,
416 								block, bsize);
417 			if (buffer->error) {
418 				ERROR("Unable to read page, block %llx, size %x"
419 					"\n", block, bsize);
420 				squashfs_cache_put(buffer);
421 				goto error_out;
422 			}
423 			bytes = buffer->length;
424 		}
425 	} else {
426 		/*
427 		 * Datablock is stored inside a fragment (tail-end packed
428 		 * block).
429 		 */
430 		buffer = squashfs_get_fragment(inode->i_sb,
431 				squashfs_i(inode)->fragment_block,
432 				squashfs_i(inode)->fragment_size);
433 
434 		if (buffer->error) {
435 			ERROR("Unable to read page, block %llx, size %x\n",
436 				squashfs_i(inode)->fragment_block,
437 				squashfs_i(inode)->fragment_size);
438 			squashfs_cache_put(buffer);
439 			goto error_out;
440 		}
441 		bytes = i_size_read(inode) & (msblk->block_size - 1);
442 		offset = squashfs_i(inode)->fragment_offset;
443 	}
444 
445 	/*
446 	 * Loop copying datablock into pages.  As the datablock likely covers
447 	 * many PAGE_CACHE_SIZE pages (default block size is 128 KiB) explicitly
448 	 * grab the pages from the page cache, except for the page that we've
449 	 * been called to fill.
450 	 */
451 	for (i = start_index; i <= end_index && bytes > 0; i++,
452 			bytes -= PAGE_CACHE_SIZE, offset += PAGE_CACHE_SIZE) {
453 		struct page *push_page;
454 		int avail = sparse ? 0 : min_t(int, bytes, PAGE_CACHE_SIZE);
455 
456 		TRACE("bytes %d, i %d, available_bytes %d\n", bytes, i, avail);
457 
458 		push_page = (i == page->index) ? page :
459 			grab_cache_page_nowait(page->mapping, i);
460 
461 		if (!push_page)
462 			continue;
463 
464 		if (PageUptodate(push_page))
465 			goto skip_page;
466 
467 		pageaddr = kmap_atomic(push_page);
468 		squashfs_copy_data(pageaddr, buffer, offset, avail);
469 		memset(pageaddr + avail, 0, PAGE_CACHE_SIZE - avail);
470 		kunmap_atomic(pageaddr);
471 		flush_dcache_page(push_page);
472 		SetPageUptodate(push_page);
473 skip_page:
474 		unlock_page(push_page);
475 		if (i != page->index)
476 			page_cache_release(push_page);
477 	}
478 
479 	if (!sparse)
480 		squashfs_cache_put(buffer);
481 
482 	return 0;
483 
484 error_out:
485 	SetPageError(page);
486 out:
487 	pageaddr = kmap_atomic(page);
488 	memset(pageaddr, 0, PAGE_CACHE_SIZE);
489 	kunmap_atomic(pageaddr);
490 	flush_dcache_page(page);
491 	if (!PageError(page))
492 		SetPageUptodate(page);
493 	unlock_page(page);
494 
495 	return 0;
496 }
497 
498 
499 const struct address_space_operations squashfs_aops = {
500 	.readpage = squashfs_readpage
501 };
502