xref: /linux/fs/squashfs/file.c (revision 3c4fc7bf4c9e66fe71abcbf93f62f4ddb89b7f15)
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 #include "page_actor.h"
43 
44 /*
45  * Locate cache slot in range [offset, index] for specified inode.  If
46  * there's more than one return the slot closest to index.
47  */
48 static struct meta_index *locate_meta_index(struct inode *inode, int offset,
49 				int index)
50 {
51 	struct meta_index *meta = NULL;
52 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
53 	int i;
54 
55 	mutex_lock(&msblk->meta_index_mutex);
56 
57 	TRACE("locate_meta_index: index %d, offset %d\n", index, offset);
58 
59 	if (msblk->meta_index == NULL)
60 		goto not_allocated;
61 
62 	for (i = 0; i < SQUASHFS_META_SLOTS; i++) {
63 		if (msblk->meta_index[i].inode_number == inode->i_ino &&
64 				msblk->meta_index[i].offset >= offset &&
65 				msblk->meta_index[i].offset <= index &&
66 				msblk->meta_index[i].locked == 0) {
67 			TRACE("locate_meta_index: entry %d, offset %d\n", i,
68 					msblk->meta_index[i].offset);
69 			meta = &msblk->meta_index[i];
70 			offset = meta->offset;
71 		}
72 	}
73 
74 	if (meta)
75 		meta->locked = 1;
76 
77 not_allocated:
78 	mutex_unlock(&msblk->meta_index_mutex);
79 
80 	return meta;
81 }
82 
83 
84 /*
85  * Find and initialise an empty cache slot for index offset.
86  */
87 static struct meta_index *empty_meta_index(struct inode *inode, int offset,
88 				int skip)
89 {
90 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
91 	struct meta_index *meta = NULL;
92 	int i;
93 
94 	mutex_lock(&msblk->meta_index_mutex);
95 
96 	TRACE("empty_meta_index: offset %d, skip %d\n", offset, skip);
97 
98 	if (msblk->meta_index == NULL) {
99 		/*
100 		 * First time cache index has been used, allocate and
101 		 * initialise.  The cache index could be allocated at
102 		 * mount time but doing it here means it is allocated only
103 		 * if a 'large' file is read.
104 		 */
105 		msblk->meta_index = kcalloc(SQUASHFS_META_SLOTS,
106 			sizeof(*(msblk->meta_index)), GFP_KERNEL);
107 		if (msblk->meta_index == NULL) {
108 			ERROR("Failed to allocate meta_index\n");
109 			goto failed;
110 		}
111 		for (i = 0; i < SQUASHFS_META_SLOTS; i++) {
112 			msblk->meta_index[i].inode_number = 0;
113 			msblk->meta_index[i].locked = 0;
114 		}
115 		msblk->next_meta_index = 0;
116 	}
117 
118 	for (i = SQUASHFS_META_SLOTS; i &&
119 			msblk->meta_index[msblk->next_meta_index].locked; i--)
120 		msblk->next_meta_index = (msblk->next_meta_index + 1) %
121 			SQUASHFS_META_SLOTS;
122 
123 	if (i == 0) {
124 		TRACE("empty_meta_index: failed!\n");
125 		goto failed;
126 	}
127 
128 	TRACE("empty_meta_index: returned meta entry %d, %p\n",
129 			msblk->next_meta_index,
130 			&msblk->meta_index[msblk->next_meta_index]);
131 
132 	meta = &msblk->meta_index[msblk->next_meta_index];
133 	msblk->next_meta_index = (msblk->next_meta_index + 1) %
134 			SQUASHFS_META_SLOTS;
135 
136 	meta->inode_number = inode->i_ino;
137 	meta->offset = offset;
138 	meta->skip = skip;
139 	meta->entries = 0;
140 	meta->locked = 1;
141 
142 failed:
143 	mutex_unlock(&msblk->meta_index_mutex);
144 	return meta;
145 }
146 
147 
148 static void release_meta_index(struct inode *inode, struct meta_index *meta)
149 {
150 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
151 	mutex_lock(&msblk->meta_index_mutex);
152 	meta->locked = 0;
153 	mutex_unlock(&msblk->meta_index_mutex);
154 }
155 
156 
157 /*
158  * Read the next n blocks from the block list, starting from
159  * metadata block <start_block, offset>.
160  */
161 static long long read_indexes(struct super_block *sb, int n,
162 				u64 *start_block, int *offset)
163 {
164 	int err, i;
165 	long long block = 0;
166 	__le32 *blist = kmalloc(PAGE_SIZE, GFP_KERNEL);
167 
168 	if (blist == NULL) {
169 		ERROR("read_indexes: Failed to allocate block_list\n");
170 		return -ENOMEM;
171 	}
172 
173 	while (n) {
174 		int blocks = min_t(int, n, PAGE_SIZE >> 2);
175 
176 		err = squashfs_read_metadata(sb, blist, start_block,
177 				offset, blocks << 2);
178 		if (err < 0) {
179 			ERROR("read_indexes: reading block [%llx:%x]\n",
180 				*start_block, *offset);
181 			goto failure;
182 		}
183 
184 		for (i = 0; i < blocks; i++) {
185 			int size = squashfs_block_size(blist[i]);
186 			if (size < 0) {
187 				err = size;
188 				goto failure;
189 			}
190 			block += SQUASHFS_COMPRESSED_SIZE_BLOCK(size);
191 		}
192 		n -= blocks;
193 	}
194 
195 	kfree(blist);
196 	return block;
197 
198 failure:
199 	kfree(blist);
200 	return err;
201 }
202 
203 
204 /*
205  * Each cache index slot has SQUASHFS_META_ENTRIES, each of which
206  * can cache one index -> datablock/blocklist-block mapping.  We wish
207  * to distribute these over the length of the file, entry[0] maps index x,
208  * entry[1] maps index x + skip, entry[2] maps index x + 2 * skip, and so on.
209  * The larger the file, the greater the skip factor.  The skip factor is
210  * limited to the size of the metadata cache (SQUASHFS_CACHED_BLKS) to ensure
211  * the number of metadata blocks that need to be read fits into the cache.
212  * If the skip factor is limited in this way then the file will use multiple
213  * slots.
214  */
215 static inline int calculate_skip(u64 blocks)
216 {
217 	u64 skip = blocks / ((SQUASHFS_META_ENTRIES + 1)
218 		 * SQUASHFS_META_INDEXES);
219 	return min((u64) SQUASHFS_CACHED_BLKS - 1, skip + 1);
220 }
221 
222 
223 /*
224  * Search and grow the index cache for the specified inode, returning the
225  * on-disk locations of the datablock and block list metadata block
226  * <index_block, index_offset> for index (scaled to nearest cache index).
227  */
228 static int fill_meta_index(struct inode *inode, int index,
229 		u64 *index_block, int *index_offset, u64 *data_block)
230 {
231 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
232 	int skip = calculate_skip(i_size_read(inode) >> msblk->block_log);
233 	int offset = 0;
234 	struct meta_index *meta;
235 	struct meta_entry *meta_entry;
236 	u64 cur_index_block = squashfs_i(inode)->block_list_start;
237 	int cur_offset = squashfs_i(inode)->offset;
238 	u64 cur_data_block = squashfs_i(inode)->start;
239 	int err, i;
240 
241 	/*
242 	 * Scale index to cache index (cache slot entry)
243 	 */
244 	index /= SQUASHFS_META_INDEXES * skip;
245 
246 	while (offset < index) {
247 		meta = locate_meta_index(inode, offset + 1, index);
248 
249 		if (meta == NULL) {
250 			meta = empty_meta_index(inode, offset + 1, skip);
251 			if (meta == NULL)
252 				goto all_done;
253 		} else {
254 			offset = index < meta->offset + meta->entries ? index :
255 				meta->offset + meta->entries - 1;
256 			meta_entry = &meta->meta_entry[offset - meta->offset];
257 			cur_index_block = meta_entry->index_block +
258 				msblk->inode_table;
259 			cur_offset = meta_entry->offset;
260 			cur_data_block = meta_entry->data_block;
261 			TRACE("get_meta_index: offset %d, meta->offset %d, "
262 				"meta->entries %d\n", offset, meta->offset,
263 				meta->entries);
264 			TRACE("get_meta_index: index_block 0x%llx, offset 0x%x"
265 				" data_block 0x%llx\n", cur_index_block,
266 				cur_offset, cur_data_block);
267 		}
268 
269 		/*
270 		 * If necessary grow cache slot by reading block list.  Cache
271 		 * slot is extended up to index or to the end of the slot, in
272 		 * which case further slots will be used.
273 		 */
274 		for (i = meta->offset + meta->entries; i <= index &&
275 				i < meta->offset + SQUASHFS_META_ENTRIES; i++) {
276 			int blocks = skip * SQUASHFS_META_INDEXES;
277 			long long res = read_indexes(inode->i_sb, blocks,
278 					&cur_index_block, &cur_offset);
279 
280 			if (res < 0) {
281 				if (meta->entries == 0)
282 					/*
283 					 * Don't leave an empty slot on read
284 					 * error allocated to this inode...
285 					 */
286 					meta->inode_number = 0;
287 				err = res;
288 				goto failed;
289 			}
290 
291 			cur_data_block += res;
292 			meta_entry = &meta->meta_entry[i - meta->offset];
293 			meta_entry->index_block = cur_index_block -
294 				msblk->inode_table;
295 			meta_entry->offset = cur_offset;
296 			meta_entry->data_block = cur_data_block;
297 			meta->entries++;
298 			offset++;
299 		}
300 
301 		TRACE("get_meta_index: meta->offset %d, meta->entries %d\n",
302 				meta->offset, meta->entries);
303 
304 		release_meta_index(inode, meta);
305 	}
306 
307 all_done:
308 	*index_block = cur_index_block;
309 	*index_offset = cur_offset;
310 	*data_block = cur_data_block;
311 
312 	/*
313 	 * Scale cache index (cache slot entry) to index
314 	 */
315 	return offset * SQUASHFS_META_INDEXES * skip;
316 
317 failed:
318 	release_meta_index(inode, meta);
319 	return err;
320 }
321 
322 
323 /*
324  * Get the on-disk location and compressed size of the datablock
325  * specified by index.  Fill_meta_index() does most of the work.
326  */
327 static int read_blocklist(struct inode *inode, int index, u64 *block)
328 {
329 	u64 start;
330 	long long blks;
331 	int offset;
332 	__le32 size;
333 	int res = fill_meta_index(inode, index, &start, &offset, block);
334 
335 	TRACE("read_blocklist: res %d, index %d, start 0x%llx, offset"
336 		       " 0x%x, block 0x%llx\n", res, index, start, offset,
337 			*block);
338 
339 	if (res < 0)
340 		return res;
341 
342 	/*
343 	 * res contains the index of the mapping returned by fill_meta_index(),
344 	 * this will likely be less than the desired index (because the
345 	 * meta_index cache works at a higher granularity).  Read any
346 	 * extra block indexes needed.
347 	 */
348 	if (res < index) {
349 		blks = read_indexes(inode->i_sb, index - res, &start, &offset);
350 		if (blks < 0)
351 			return (int) blks;
352 		*block += blks;
353 	}
354 
355 	/*
356 	 * Read length of block specified by index.
357 	 */
358 	res = squashfs_read_metadata(inode->i_sb, &size, &start, &offset,
359 			sizeof(size));
360 	if (res < 0)
361 		return res;
362 	return squashfs_block_size(size);
363 }
364 
365 void squashfs_fill_page(struct page *page, struct squashfs_cache_entry *buffer, int offset, int avail)
366 {
367 	int copied;
368 	void *pageaddr;
369 
370 	pageaddr = kmap_atomic(page);
371 	copied = squashfs_copy_data(pageaddr, buffer, offset, avail);
372 	memset(pageaddr + copied, 0, PAGE_SIZE - copied);
373 	kunmap_atomic(pageaddr);
374 
375 	flush_dcache_page(page);
376 	if (copied == avail)
377 		SetPageUptodate(page);
378 	else
379 		SetPageError(page);
380 }
381 
382 /* Copy data into page cache  */
383 void squashfs_copy_cache(struct page *page, struct squashfs_cache_entry *buffer,
384 	int bytes, int offset)
385 {
386 	struct inode *inode = page->mapping->host;
387 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
388 	int i, mask = (1 << (msblk->block_log - PAGE_SHIFT)) - 1;
389 	int start_index = page->index & ~mask, end_index = start_index | mask;
390 
391 	/*
392 	 * Loop copying datablock into pages.  As the datablock likely covers
393 	 * many PAGE_SIZE pages (default block size is 128 KiB) explicitly
394 	 * grab the pages from the page cache, except for the page that we've
395 	 * been called to fill.
396 	 */
397 	for (i = start_index; i <= end_index && bytes > 0; i++,
398 			bytes -= PAGE_SIZE, offset += PAGE_SIZE) {
399 		struct page *push_page;
400 		int avail = buffer ? min_t(int, bytes, PAGE_SIZE) : 0;
401 
402 		TRACE("bytes %d, i %d, available_bytes %d\n", bytes, i, avail);
403 
404 		push_page = (i == page->index) ? page :
405 			grab_cache_page_nowait(page->mapping, i);
406 
407 		if (!push_page)
408 			continue;
409 
410 		if (PageUptodate(push_page))
411 			goto skip_page;
412 
413 		squashfs_fill_page(push_page, buffer, offset, avail);
414 skip_page:
415 		unlock_page(push_page);
416 		if (i != page->index)
417 			put_page(push_page);
418 	}
419 }
420 
421 /* Read datablock stored packed inside a fragment (tail-end packed block) */
422 static int squashfs_readpage_fragment(struct page *page, int expected)
423 {
424 	struct inode *inode = page->mapping->host;
425 	struct squashfs_cache_entry *buffer = squashfs_get_fragment(inode->i_sb,
426 		squashfs_i(inode)->fragment_block,
427 		squashfs_i(inode)->fragment_size);
428 	int res = buffer->error;
429 
430 	if (res)
431 		ERROR("Unable to read page, block %llx, size %x\n",
432 			squashfs_i(inode)->fragment_block,
433 			squashfs_i(inode)->fragment_size);
434 	else
435 		squashfs_copy_cache(page, buffer, expected,
436 			squashfs_i(inode)->fragment_offset);
437 
438 	squashfs_cache_put(buffer);
439 	return res;
440 }
441 
442 static int squashfs_readpage_sparse(struct page *page, int expected)
443 {
444 	squashfs_copy_cache(page, NULL, expected, 0);
445 	return 0;
446 }
447 
448 static int squashfs_read_folio(struct file *file, struct folio *folio)
449 {
450 	struct page *page = &folio->page;
451 	struct inode *inode = page->mapping->host;
452 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
453 	int index = page->index >> (msblk->block_log - PAGE_SHIFT);
454 	int file_end = i_size_read(inode) >> msblk->block_log;
455 	int expected = index == file_end ?
456 			(i_size_read(inode) & (msblk->block_size - 1)) :
457 			 msblk->block_size;
458 	int res = 0;
459 	void *pageaddr;
460 
461 	TRACE("Entered squashfs_readpage, page index %lx, start block %llx\n",
462 				page->index, squashfs_i(inode)->start);
463 
464 	if (page->index >= ((i_size_read(inode) + PAGE_SIZE - 1) >>
465 					PAGE_SHIFT))
466 		goto out;
467 
468 	if (index < file_end || squashfs_i(inode)->fragment_block ==
469 					SQUASHFS_INVALID_BLK) {
470 		u64 block = 0;
471 
472 		res = read_blocklist(inode, index, &block);
473 		if (res < 0)
474 			goto error_out;
475 
476 		if (res == 0)
477 			res = squashfs_readpage_sparse(page, expected);
478 		else
479 			res = squashfs_readpage_block(page, block, res, expected);
480 	} else
481 		res = squashfs_readpage_fragment(page, expected);
482 
483 	if (!res)
484 		return 0;
485 
486 error_out:
487 	SetPageError(page);
488 out:
489 	pageaddr = kmap_atomic(page);
490 	memset(pageaddr, 0, PAGE_SIZE);
491 	kunmap_atomic(pageaddr);
492 	flush_dcache_page(page);
493 	if (res == 0)
494 		SetPageUptodate(page);
495 	unlock_page(page);
496 
497 	return res;
498 }
499 
500 static int squashfs_readahead_fragment(struct page **page,
501 	unsigned int pages, unsigned int expected)
502 {
503 	struct inode *inode = page[0]->mapping->host;
504 	struct squashfs_cache_entry *buffer = squashfs_get_fragment(inode->i_sb,
505 		squashfs_i(inode)->fragment_block,
506 		squashfs_i(inode)->fragment_size);
507 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
508 	unsigned int n, mask = (1 << (msblk->block_log - PAGE_SHIFT)) - 1;
509 
510 	if (buffer->error)
511 		goto out;
512 
513 	expected += squashfs_i(inode)->fragment_offset;
514 
515 	for (n = 0; n < pages; n++) {
516 		unsigned int base = (page[n]->index & mask) << PAGE_SHIFT;
517 		unsigned int offset = base + squashfs_i(inode)->fragment_offset;
518 
519 		if (expected > offset) {
520 			unsigned int avail = min_t(unsigned int, expected -
521 				offset, PAGE_SIZE);
522 
523 			squashfs_fill_page(page[n], buffer, offset, avail);
524 		}
525 
526 		unlock_page(page[n]);
527 		put_page(page[n]);
528 	}
529 
530 out:
531 	squashfs_cache_put(buffer);
532 	return buffer->error;
533 }
534 
535 static void squashfs_readahead(struct readahead_control *ractl)
536 {
537 	struct inode *inode = ractl->mapping->host;
538 	struct squashfs_sb_info *msblk = inode->i_sb->s_fs_info;
539 	size_t mask = (1UL << msblk->block_log) - 1;
540 	unsigned short shift = msblk->block_log - PAGE_SHIFT;
541 	loff_t start = readahead_pos(ractl) & ~mask;
542 	size_t len = readahead_length(ractl) + readahead_pos(ractl) - start;
543 	struct squashfs_page_actor *actor;
544 	unsigned int nr_pages = 0;
545 	struct page **pages;
546 	int i, file_end = i_size_read(inode) >> msblk->block_log;
547 	unsigned int max_pages = 1UL << shift;
548 
549 	readahead_expand(ractl, start, (len | mask) + 1);
550 
551 	pages = kmalloc_array(max_pages, sizeof(void *), GFP_KERNEL);
552 	if (!pages)
553 		return;
554 
555 	for (;;) {
556 		pgoff_t index;
557 		int res, bsize;
558 		u64 block = 0;
559 		unsigned int expected;
560 
561 		nr_pages = __readahead_batch(ractl, pages, max_pages);
562 		if (!nr_pages)
563 			break;
564 
565 		if (readahead_pos(ractl) >= i_size_read(inode))
566 			goto skip_pages;
567 
568 		index = pages[0]->index >> shift;
569 		if ((pages[nr_pages - 1]->index >> shift) != index)
570 			goto skip_pages;
571 
572 		expected = index == file_end ?
573 			   (i_size_read(inode) & (msblk->block_size - 1)) :
574 			    msblk->block_size;
575 
576 		if (index == file_end && squashfs_i(inode)->fragment_block !=
577 						SQUASHFS_INVALID_BLK) {
578 			res = squashfs_readahead_fragment(pages, nr_pages,
579 							  expected);
580 			if (res)
581 				goto skip_pages;
582 			continue;
583 		}
584 
585 		bsize = read_blocklist(inode, index, &block);
586 		if (bsize == 0)
587 			goto skip_pages;
588 
589 		actor = squashfs_page_actor_init_special(msblk, pages, nr_pages,
590 							 expected);
591 		if (!actor)
592 			goto skip_pages;
593 
594 		res = squashfs_read_data(inode->i_sb, block, bsize, NULL, actor);
595 
596 		squashfs_page_actor_free(actor);
597 
598 		if (res == expected) {
599 			int bytes;
600 
601 			/* Last page (if present) may have trailing bytes not filled */
602 			bytes = res % PAGE_SIZE;
603 			if (pages[nr_pages - 1]->index == file_end && bytes)
604 				memzero_page(pages[nr_pages - 1], bytes,
605 					     PAGE_SIZE - bytes);
606 
607 			for (i = 0; i < nr_pages; i++) {
608 				flush_dcache_page(pages[i]);
609 				SetPageUptodate(pages[i]);
610 			}
611 		}
612 
613 		for (i = 0; i < nr_pages; i++) {
614 			unlock_page(pages[i]);
615 			put_page(pages[i]);
616 		}
617 	}
618 
619 	kfree(pages);
620 	return;
621 
622 skip_pages:
623 	for (i = 0; i < nr_pages; i++) {
624 		unlock_page(pages[i]);
625 		put_page(pages[i]);
626 	}
627 	kfree(pages);
628 }
629 
630 const struct address_space_operations squashfs_aops = {
631 	.read_folio = squashfs_read_folio,
632 	.readahead = squashfs_readahead
633 };
634