xref: /openbmc/linux/fs/squashfs/file.c (revision 4419617e)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3  * Squashfs - a compressed read only filesystem for Linux
4  *
5  * Copyright (c) 2002, 2003, 2004, 2005, 2006, 2007, 2008
6  * Phillip Lougher <phillip@squashfs.org.uk>
7  *
8  * file.c
9  */
10 
11 /*
12  * This file contains code for handling regular files.  A regular file
13  * consists of a sequence of contiguous compressed blocks, and/or a
14  * compressed fragment block (tail-end packed block).   The compressed size
15  * of each datablock is stored in a block list contained within the
16  * file inode (itself stored in one or more compressed metadata blocks).
17  *
18  * To speed up access to datablocks when reading 'large' files (256 Mbytes or
19  * larger), the code implements an index cache that caches the mapping from
20  * block index to datablock location on disk.
21  *
22  * The index cache allows Squashfs to handle large files (up to 1.75 TiB) while
23  * retaining a simple and space-efficient block list on disk.  The cache
24  * is split into slots, caching up to eight 224 GiB files (128 KiB blocks).
25  * Larger files use multiple slots, with 1.75 TiB files using all 8 slots.
26  * The index cache is designed to be memory efficient, and by default uses
27  * 16 KiB.
28  */
29 
30 #include <linux/fs.h>
31 #include <linux/vfs.h>
32 #include <linux/kernel.h>
33 #include <linux/slab.h>
34 #include <linux/string.h>
35 #include <linux/pagemap.h>
36 #include <linux/mutex.h>
37 
38 #include "squashfs_fs.h"
39 #include "squashfs_fs_sb.h"
40 #include "squashfs_fs_i.h"
41 #include "squashfs.h"
42 
43 /*
44  * Locate cache slot in range [offset, index] for specified inode.  If
45  * there's more than one return the slot closest to index.
46  */
47 static struct meta_index *locate_meta_index(struct inode *inode, int offset,
48 				int index)
49 {
50 	struct meta_index *meta = NULL;
51 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
52 	int i;
53 
54 	mutex_lock(&msblk->meta_index_mutex);
55 
56 	TRACE("locate_meta_index: index %d, offset %d\n", index, offset);
57 
58 	if (msblk->meta_index == NULL)
59 		goto not_allocated;
60 
61 	for (i = 0; i < SQUASHFS_META_SLOTS; i++) {
62 		if (msblk->meta_index[i].inode_number == inode->i_ino &&
63 				msblk->meta_index[i].offset >= offset &&
64 				msblk->meta_index[i].offset <= index &&
65 				msblk->meta_index[i].locked == 0) {
66 			TRACE("locate_meta_index: entry %d, offset %d\n", i,
67 					msblk->meta_index[i].offset);
68 			meta = &msblk->meta_index[i];
69 			offset = meta->offset;
70 		}
71 	}
72 
73 	if (meta)
74 		meta->locked = 1;
75 
76 not_allocated:
77 	mutex_unlock(&msblk->meta_index_mutex);
78 
79 	return meta;
80 }
81 
82 
83 /*
84  * Find and initialise an empty cache slot for index offset.
85  */
86 static struct meta_index *empty_meta_index(struct inode *inode, int offset,
87 				int skip)
88 {
89 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
90 	struct meta_index *meta = NULL;
91 	int i;
92 
93 	mutex_lock(&msblk->meta_index_mutex);
94 
95 	TRACE("empty_meta_index: offset %d, skip %d\n", offset, skip);
96 
97 	if (msblk->meta_index == NULL) {
98 		/*
99 		 * First time cache index has been used, allocate and
100 		 * initialise.  The cache index could be allocated at
101 		 * mount time but doing it here means it is allocated only
102 		 * if a 'large' file is read.
103 		 */
104 		msblk->meta_index = kcalloc(SQUASHFS_META_SLOTS,
105 			sizeof(*(msblk->meta_index)), GFP_KERNEL);
106 		if (msblk->meta_index == NULL) {
107 			ERROR("Failed to allocate meta_index\n");
108 			goto failed;
109 		}
110 		for (i = 0; i < SQUASHFS_META_SLOTS; i++) {
111 			msblk->meta_index[i].inode_number = 0;
112 			msblk->meta_index[i].locked = 0;
113 		}
114 		msblk->next_meta_index = 0;
115 	}
116 
117 	for (i = SQUASHFS_META_SLOTS; i &&
118 			msblk->meta_index[msblk->next_meta_index].locked; i--)
119 		msblk->next_meta_index = (msblk->next_meta_index + 1) %
120 			SQUASHFS_META_SLOTS;
121 
122 	if (i == 0) {
123 		TRACE("empty_meta_index: failed!\n");
124 		goto failed;
125 	}
126 
127 	TRACE("empty_meta_index: returned meta entry %d, %p\n",
128 			msblk->next_meta_index,
129 			&msblk->meta_index[msblk->next_meta_index]);
130 
131 	meta = &msblk->meta_index[msblk->next_meta_index];
132 	msblk->next_meta_index = (msblk->next_meta_index + 1) %
133 			SQUASHFS_META_SLOTS;
134 
135 	meta->inode_number = inode->i_ino;
136 	meta->offset = offset;
137 	meta->skip = skip;
138 	meta->entries = 0;
139 	meta->locked = 1;
140 
141 failed:
142 	mutex_unlock(&msblk->meta_index_mutex);
143 	return meta;
144 }
145 
146 
147 static void release_meta_index(struct inode *inode, struct meta_index *meta)
148 {
149 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
150 	mutex_lock(&msblk->meta_index_mutex);
151 	meta->locked = 0;
152 	mutex_unlock(&msblk->meta_index_mutex);
153 }
154 
155 
156 /*
157  * Read the next n blocks from the block list, starting from
158  * metadata block <start_block, offset>.
159  */
160 static long long read_indexes(struct super_block *sb, int n,
161 				u64 *start_block, int *offset)
162 {
163 	int err, i;
164 	long long block = 0;
165 	__le32 *blist = kmalloc(PAGE_SIZE, GFP_KERNEL);
166 
167 	if (blist == NULL) {
168 		ERROR("read_indexes: Failed to allocate block_list\n");
169 		return -ENOMEM;
170 	}
171 
172 	while (n) {
173 		int blocks = min_t(int, n, PAGE_SIZE >> 2);
174 
175 		err = squashfs_read_metadata(sb, blist, start_block,
176 				offset, blocks << 2);
177 		if (err < 0) {
178 			ERROR("read_indexes: reading block [%llx:%x]\n",
179 				*start_block, *offset);
180 			goto failure;
181 		}
182 
183 		for (i = 0; i < blocks; i++) {
184 			int size = squashfs_block_size(blist[i]);
185 			if (size < 0) {
186 				err = size;
187 				goto failure;
188 			}
189 			block += SQUASHFS_COMPRESSED_SIZE_BLOCK(size);
190 		}
191 		n -= blocks;
192 	}
193 
194 	kfree(blist);
195 	return block;
196 
197 failure:
198 	kfree(blist);
199 	return err;
200 }
201 
202 
203 /*
204  * Each cache index slot has SQUASHFS_META_ENTRIES, each of which
205  * can cache one index -> datablock/blocklist-block mapping.  We wish
206  * to distribute these over the length of the file, entry[0] maps index x,
207  * entry[1] maps index x + skip, entry[2] maps index x + 2 * skip, and so on.
208  * The larger the file, the greater the skip factor.  The skip factor is
209  * limited to the size of the metadata cache (SQUASHFS_CACHED_BLKS) to ensure
210  * the number of metadata blocks that need to be read fits into the cache.
211  * If the skip factor is limited in this way then the file will use multiple
212  * slots.
213  */
214 static inline int calculate_skip(int blocks)
215 {
216 	int skip = blocks / ((SQUASHFS_META_ENTRIES + 1)
217 		 * SQUASHFS_META_INDEXES);
218 	return min(SQUASHFS_CACHED_BLKS - 1, skip + 1);
219 }
220 
221 
222 /*
223  * Search and grow the index cache for the specified inode, returning the
224  * on-disk locations of the datablock and block list metadata block
225  * <index_block, index_offset> for index (scaled to nearest cache index).
226  */
227 static int fill_meta_index(struct inode *inode, int index,
228 		u64 *index_block, int *index_offset, u64 *data_block)
229 {
230 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
231 	int skip = calculate_skip(i_size_read(inode) >> msblk->block_log);
232 	int offset = 0;
233 	struct meta_index *meta;
234 	struct meta_entry *meta_entry;
235 	u64 cur_index_block = squashfs_i(inode)->block_list_start;
236 	int cur_offset = squashfs_i(inode)->offset;
237 	u64 cur_data_block = squashfs_i(inode)->start;
238 	int err, i;
239 
240 	/*
241 	 * Scale index to cache index (cache slot entry)
242 	 */
243 	index /= SQUASHFS_META_INDEXES * skip;
244 
245 	while (offset < index) {
246 		meta = locate_meta_index(inode, offset + 1, index);
247 
248 		if (meta == NULL) {
249 			meta = empty_meta_index(inode, offset + 1, skip);
250 			if (meta == NULL)
251 				goto all_done;
252 		} else {
253 			offset = index < meta->offset + meta->entries ? index :
254 				meta->offset + meta->entries - 1;
255 			meta_entry = &meta->meta_entry[offset - meta->offset];
256 			cur_index_block = meta_entry->index_block +
257 				msblk->inode_table;
258 			cur_offset = meta_entry->offset;
259 			cur_data_block = meta_entry->data_block;
260 			TRACE("get_meta_index: offset %d, meta->offset %d, "
261 				"meta->entries %d\n", offset, meta->offset,
262 				meta->entries);
263 			TRACE("get_meta_index: index_block 0x%llx, offset 0x%x"
264 				" data_block 0x%llx\n", cur_index_block,
265 				cur_offset, cur_data_block);
266 		}
267 
268 		/*
269 		 * If necessary grow cache slot by reading block list.  Cache
270 		 * slot is extended up to index or to the end of the slot, in
271 		 * which case further slots will be used.
272 		 */
273 		for (i = meta->offset + meta->entries; i <= index &&
274 				i < meta->offset + SQUASHFS_META_ENTRIES; i++) {
275 			int blocks = skip * SQUASHFS_META_INDEXES;
276 			long long res = read_indexes(inode->i_sb, blocks,
277 					&cur_index_block, &cur_offset);
278 
279 			if (res < 0) {
280 				if (meta->entries == 0)
281 					/*
282 					 * Don't leave an empty slot on read
283 					 * error allocated to this inode...
284 					 */
285 					meta->inode_number = 0;
286 				err = res;
287 				goto failed;
288 			}
289 
290 			cur_data_block += res;
291 			meta_entry = &meta->meta_entry[i - meta->offset];
292 			meta_entry->index_block = cur_index_block -
293 				msblk->inode_table;
294 			meta_entry->offset = cur_offset;
295 			meta_entry->data_block = cur_data_block;
296 			meta->entries++;
297 			offset++;
298 		}
299 
300 		TRACE("get_meta_index: meta->offset %d, meta->entries %d\n",
301 				meta->offset, meta->entries);
302 
303 		release_meta_index(inode, meta);
304 	}
305 
306 all_done:
307 	*index_block = cur_index_block;
308 	*index_offset = cur_offset;
309 	*data_block = cur_data_block;
310 
311 	/*
312 	 * Scale cache index (cache slot entry) to index
313 	 */
314 	return offset * SQUASHFS_META_INDEXES * skip;
315 
316 failed:
317 	release_meta_index(inode, meta);
318 	return err;
319 }
320 
321 
322 /*
323  * Get the on-disk location and compressed size of the datablock
324  * specified by index.  Fill_meta_index() does most of the work.
325  */
326 static int read_blocklist(struct inode *inode, int index, u64 *block)
327 {
328 	u64 start;
329 	long long blks;
330 	int offset;
331 	__le32 size;
332 	int res = fill_meta_index(inode, index, &start, &offset, block);
333 
334 	TRACE("read_blocklist: res %d, index %d, start 0x%llx, offset"
335 		       " 0x%x, block 0x%llx\n", res, index, start, offset,
336 			*block);
337 
338 	if (res < 0)
339 		return res;
340 
341 	/*
342 	 * res contains the index of the mapping returned by fill_meta_index(),
343 	 * this will likely be less than the desired index (because the
344 	 * meta_index cache works at a higher granularity).  Read any
345 	 * extra block indexes needed.
346 	 */
347 	if (res < index) {
348 		blks = read_indexes(inode->i_sb, index - res, &start, &offset);
349 		if (blks < 0)
350 			return (int) blks;
351 		*block += blks;
352 	}
353 
354 	/*
355 	 * Read length of block specified by index.
356 	 */
357 	res = squashfs_read_metadata(inode->i_sb, &size, &start, &offset,
358 			sizeof(size));
359 	if (res < 0)
360 		return res;
361 	return squashfs_block_size(size);
362 }
363 
364 void squashfs_fill_page(struct page *page, struct squashfs_cache_entry *buffer, int offset, int avail)
365 {
366 	int copied;
367 	void *pageaddr;
368 
369 	pageaddr = kmap_atomic(page);
370 	copied = squashfs_copy_data(pageaddr, buffer, offset, avail);
371 	memset(pageaddr + copied, 0, PAGE_SIZE - copied);
372 	kunmap_atomic(pageaddr);
373 
374 	flush_dcache_page(page);
375 	if (copied == avail)
376 		SetPageUptodate(page);
377 	else
378 		SetPageError(page);
379 }
380 
381 /* Copy data into page cache  */
382 void squashfs_copy_cache(struct page *page, struct squashfs_cache_entry *buffer,
383 	int bytes, int offset)
384 {
385 	struct inode *inode = page->mapping->host;
386 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
387 	int i, mask = (1 << (msblk->block_log - PAGE_SHIFT)) - 1;
388 	int start_index = page->index & ~mask, end_index = start_index | mask;
389 
390 	/*
391 	 * Loop copying datablock into pages.  As the datablock likely covers
392 	 * many PAGE_SIZE pages (default block size is 128 KiB) explicitly
393 	 * grab the pages from the page cache, except for the page that we've
394 	 * been called to fill.
395 	 */
396 	for (i = start_index; i <= end_index && bytes > 0; i++,
397 			bytes -= PAGE_SIZE, offset += PAGE_SIZE) {
398 		struct page *push_page;
399 		int avail = buffer ? min_t(int, bytes, PAGE_SIZE) : 0;
400 
401 		TRACE("bytes %d, i %d, available_bytes %d\n", bytes, i, avail);
402 
403 		push_page = (i == page->index) ? page :
404 			grab_cache_page_nowait(page->mapping, i);
405 
406 		if (!push_page)
407 			continue;
408 
409 		if (PageUptodate(push_page))
410 			goto skip_page;
411 
412 		squashfs_fill_page(push_page, buffer, offset, avail);
413 skip_page:
414 		unlock_page(push_page);
415 		if (i != page->index)
416 			put_page(push_page);
417 	}
418 }
419 
420 /* Read datablock stored packed inside a fragment (tail-end packed block) */
421 static int squashfs_readpage_fragment(struct page *page, int expected)
422 {
423 	struct inode *inode = page->mapping->host;
424 	struct squashfs_cache_entry *buffer = squashfs_get_fragment(inode->i_sb,
425 		squashfs_i(inode)->fragment_block,
426 		squashfs_i(inode)->fragment_size);
427 	int res = buffer->error;
428 
429 	if (res)
430 		ERROR("Unable to read page, block %llx, size %x\n",
431 			squashfs_i(inode)->fragment_block,
432 			squashfs_i(inode)->fragment_size);
433 	else
434 		squashfs_copy_cache(page, buffer, expected,
435 			squashfs_i(inode)->fragment_offset);
436 
437 	squashfs_cache_put(buffer);
438 	return res;
439 }
440 
441 static int squashfs_readpage_sparse(struct page *page, int expected)
442 {
443 	squashfs_copy_cache(page, NULL, expected, 0);
444 	return 0;
445 }
446 
447 static int squashfs_readpage(struct file *file, struct page *page)
448 {
449 	struct inode *inode = page->mapping->host;
450 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
451 	int index = page->index >> (msblk->block_log - PAGE_SHIFT);
452 	int file_end = i_size_read(inode) >> msblk->block_log;
453 	int expected = index == file_end ?
454 			(i_size_read(inode) & (msblk->block_size - 1)) :
455 			 msblk->block_size;
456 	int res;
457 	void *pageaddr;
458 
459 	TRACE("Entered squashfs_readpage, page index %lx, start block %llx\n",
460 				page->index, squashfs_i(inode)->start);
461 
462 	if (page->index >= ((i_size_read(inode) + PAGE_SIZE - 1) >>
463 					PAGE_SHIFT))
464 		goto out;
465 
466 	if (index < file_end || squashfs_i(inode)->fragment_block ==
467 					SQUASHFS_INVALID_BLK) {
468 		u64 block = 0;
469 		int bsize = read_blocklist(inode, index, &block);
470 		if (bsize < 0)
471 			goto error_out;
472 
473 		if (bsize == 0)
474 			res = squashfs_readpage_sparse(page, expected);
475 		else
476 			res = squashfs_readpage_block(page, block, bsize, expected);
477 	} else
478 		res = squashfs_readpage_fragment(page, expected);
479 
480 	if (!res)
481 		return 0;
482 
483 error_out:
484 	SetPageError(page);
485 out:
486 	pageaddr = kmap_atomic(page);
487 	memset(pageaddr, 0, PAGE_SIZE);
488 	kunmap_atomic(pageaddr);
489 	flush_dcache_page(page);
490 	if (!PageError(page))
491 		SetPageUptodate(page);
492 	unlock_page(page);
493 
494 	return 0;
495 }
496 
497 
498 const struct address_space_operations squashfs_aops = {
499 	.readpage = squashfs_readpage
500 };
501