xref: /openbmc/linux/fs/erofs/super.c (revision 83d3c4f2)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * Copyright (C) 2017-2018 HUAWEI, Inc.
4  *             https://www.huawei.com/
5  */
6 #include <linux/module.h>
7 #include <linux/buffer_head.h>
8 #include <linux/statfs.h>
9 #include <linux/parser.h>
10 #include <linux/seq_file.h>
11 #include <linux/crc32c.h>
12 #include <linux/fs_context.h>
13 #include <linux/fs_parser.h>
14 #include <linux/dax.h>
15 #include "xattr.h"
16 
17 #define CREATE_TRACE_POINTS
18 #include <trace/events/erofs.h>
19 
20 static struct kmem_cache *erofs_inode_cachep __read_mostly;
21 
22 void _erofs_err(struct super_block *sb, const char *function,
23 		const char *fmt, ...)
24 {
25 	struct va_format vaf;
26 	va_list args;
27 
28 	va_start(args, fmt);
29 
30 	vaf.fmt = fmt;
31 	vaf.va = &args;
32 
33 	pr_err("(device %s): %s: %pV", sb->s_id, function, &vaf);
34 	va_end(args);
35 }
36 
37 void _erofs_info(struct super_block *sb, const char *function,
38 		 const char *fmt, ...)
39 {
40 	struct va_format vaf;
41 	va_list args;
42 
43 	va_start(args, fmt);
44 
45 	vaf.fmt = fmt;
46 	vaf.va = &args;
47 
48 	pr_info("(device %s): %pV", sb->s_id, &vaf);
49 	va_end(args);
50 }
51 
52 static int erofs_superblock_csum_verify(struct super_block *sb, void *sbdata)
53 {
54 	struct erofs_super_block *dsb;
55 	u32 expected_crc, crc;
56 
57 	dsb = kmemdup(sbdata + EROFS_SUPER_OFFSET,
58 		      EROFS_BLKSIZ - EROFS_SUPER_OFFSET, GFP_KERNEL);
59 	if (!dsb)
60 		return -ENOMEM;
61 
62 	expected_crc = le32_to_cpu(dsb->checksum);
63 	dsb->checksum = 0;
64 	/* to allow for x86 boot sectors and other oddities. */
65 	crc = crc32c(~0, dsb, EROFS_BLKSIZ - EROFS_SUPER_OFFSET);
66 	kfree(dsb);
67 
68 	if (crc != expected_crc) {
69 		erofs_err(sb, "invalid checksum 0x%08x, 0x%08x expected",
70 			  crc, expected_crc);
71 		return -EBADMSG;
72 	}
73 	return 0;
74 }
75 
76 static void erofs_inode_init_once(void *ptr)
77 {
78 	struct erofs_inode *vi = ptr;
79 
80 	inode_init_once(&vi->vfs_inode);
81 }
82 
83 static struct inode *erofs_alloc_inode(struct super_block *sb)
84 {
85 	struct erofs_inode *vi =
86 		kmem_cache_alloc(erofs_inode_cachep, GFP_KERNEL);
87 
88 	if (!vi)
89 		return NULL;
90 
91 	/* zero out everything except vfs_inode */
92 	memset(vi, 0, offsetof(struct erofs_inode, vfs_inode));
93 	return &vi->vfs_inode;
94 }
95 
96 static void erofs_free_inode(struct inode *inode)
97 {
98 	struct erofs_inode *vi = EROFS_I(inode);
99 
100 	/* be careful of RCU symlink path */
101 	if (inode->i_op == &erofs_fast_symlink_iops)
102 		kfree(inode->i_link);
103 	kfree(vi->xattr_shared_xattrs);
104 
105 	kmem_cache_free(erofs_inode_cachep, vi);
106 }
107 
108 static bool check_layout_compatibility(struct super_block *sb,
109 				       struct erofs_super_block *dsb)
110 {
111 	const unsigned int feature = le32_to_cpu(dsb->feature_incompat);
112 
113 	EROFS_SB(sb)->feature_incompat = feature;
114 
115 	/* check if current kernel meets all mandatory requirements */
116 	if (feature & (~EROFS_ALL_FEATURE_INCOMPAT)) {
117 		erofs_err(sb,
118 			  "unidentified incompatible feature %x, please upgrade kernel version",
119 			   feature & ~EROFS_ALL_FEATURE_INCOMPAT);
120 		return false;
121 	}
122 	return true;
123 }
124 
125 #ifdef CONFIG_EROFS_FS_ZIP
126 /* read variable-sized metadata, offset will be aligned by 4-byte */
127 static void *erofs_read_metadata(struct super_block *sb, struct page **pagep,
128 				 erofs_off_t *offset, int *lengthp)
129 {
130 	struct page *page = *pagep;
131 	u8 *buffer, *ptr;
132 	int len, i, cnt;
133 	erofs_blk_t blk;
134 
135 	*offset = round_up(*offset, 4);
136 	blk = erofs_blknr(*offset);
137 
138 	if (!page || page->index != blk) {
139 		if (page) {
140 			unlock_page(page);
141 			put_page(page);
142 		}
143 		page = erofs_get_meta_page(sb, blk);
144 		if (IS_ERR(page))
145 			goto err_nullpage;
146 	}
147 
148 	ptr = kmap(page);
149 	len = le16_to_cpu(*(__le16 *)&ptr[erofs_blkoff(*offset)]);
150 	if (!len)
151 		len = U16_MAX + 1;
152 	buffer = kmalloc(len, GFP_KERNEL);
153 	if (!buffer) {
154 		buffer = ERR_PTR(-ENOMEM);
155 		goto out;
156 	}
157 	*offset += sizeof(__le16);
158 	*lengthp = len;
159 
160 	for (i = 0; i < len; i += cnt) {
161 		cnt = min(EROFS_BLKSIZ - (int)erofs_blkoff(*offset), len - i);
162 		blk = erofs_blknr(*offset);
163 
164 		if (!page || page->index != blk) {
165 			if (page) {
166 				kunmap(page);
167 				unlock_page(page);
168 				put_page(page);
169 			}
170 			page = erofs_get_meta_page(sb, blk);
171 			if (IS_ERR(page)) {
172 				kfree(buffer);
173 				goto err_nullpage;
174 			}
175 			ptr = kmap(page);
176 		}
177 		memcpy(buffer + i, ptr + erofs_blkoff(*offset), cnt);
178 		*offset += cnt;
179 	}
180 out:
181 	kunmap(page);
182 	*pagep = page;
183 	return buffer;
184 err_nullpage:
185 	*pagep = NULL;
186 	return page;
187 }
188 
189 static int erofs_load_compr_cfgs(struct super_block *sb,
190 				 struct erofs_super_block *dsb)
191 {
192 	struct erofs_sb_info *sbi;
193 	struct page *page;
194 	unsigned int algs, alg;
195 	erofs_off_t offset;
196 	int size, ret;
197 
198 	sbi = EROFS_SB(sb);
199 	sbi->available_compr_algs = le16_to_cpu(dsb->u1.available_compr_algs);
200 
201 	if (sbi->available_compr_algs & ~Z_EROFS_ALL_COMPR_ALGS) {
202 		erofs_err(sb, "try to load compressed fs with unsupported algorithms %x",
203 			  sbi->available_compr_algs & ~Z_EROFS_ALL_COMPR_ALGS);
204 		return -EINVAL;
205 	}
206 
207 	offset = EROFS_SUPER_OFFSET + sbi->sb_size;
208 	page = NULL;
209 	alg = 0;
210 	ret = 0;
211 
212 	for (algs = sbi->available_compr_algs; algs; algs >>= 1, ++alg) {
213 		void *data;
214 
215 		if (!(algs & 1))
216 			continue;
217 
218 		data = erofs_read_metadata(sb, &page, &offset, &size);
219 		if (IS_ERR(data)) {
220 			ret = PTR_ERR(data);
221 			goto err;
222 		}
223 
224 		switch (alg) {
225 		case Z_EROFS_COMPRESSION_LZ4:
226 			ret = z_erofs_load_lz4_config(sb, dsb, data, size);
227 			break;
228 		default:
229 			DBG_BUGON(1);
230 			ret = -EFAULT;
231 		}
232 		kfree(data);
233 		if (ret)
234 			goto err;
235 	}
236 err:
237 	if (page) {
238 		unlock_page(page);
239 		put_page(page);
240 	}
241 	return ret;
242 }
243 #else
244 static int erofs_load_compr_cfgs(struct super_block *sb,
245 				 struct erofs_super_block *dsb)
246 {
247 	if (dsb->u1.available_compr_algs) {
248 		erofs_err(sb, "try to load compressed fs when compression is disabled");
249 		return -EINVAL;
250 	}
251 	return 0;
252 }
253 #endif
254 
255 static int erofs_init_devices(struct super_block *sb,
256 			      struct erofs_super_block *dsb)
257 {
258 	struct erofs_sb_info *sbi = EROFS_SB(sb);
259 	unsigned int ondisk_extradevs;
260 	erofs_off_t pos;
261 	struct page *page = NULL;
262 	struct erofs_device_info *dif;
263 	struct erofs_deviceslot *dis;
264 	void *ptr;
265 	int id, err = 0;
266 
267 	sbi->total_blocks = sbi->primarydevice_blocks;
268 	if (!erofs_sb_has_device_table(sbi))
269 		ondisk_extradevs = 0;
270 	else
271 		ondisk_extradevs = le16_to_cpu(dsb->extra_devices);
272 
273 	if (ondisk_extradevs != sbi->devs->extra_devices) {
274 		erofs_err(sb, "extra devices don't match (ondisk %u, given %u)",
275 			  ondisk_extradevs, sbi->devs->extra_devices);
276 		return -EINVAL;
277 	}
278 	if (!ondisk_extradevs)
279 		return 0;
280 
281 	sbi->device_id_mask = roundup_pow_of_two(ondisk_extradevs + 1) - 1;
282 	pos = le16_to_cpu(dsb->devt_slotoff) * EROFS_DEVT_SLOT_SIZE;
283 	down_read(&sbi->devs->rwsem);
284 	idr_for_each_entry(&sbi->devs->tree, dif, id) {
285 		erofs_blk_t blk = erofs_blknr(pos);
286 		struct block_device *bdev;
287 
288 		if (!page || page->index != blk) {
289 			if (page) {
290 				kunmap(page);
291 				unlock_page(page);
292 				put_page(page);
293 			}
294 
295 			page = erofs_get_meta_page(sb, blk);
296 			if (IS_ERR(page)) {
297 				up_read(&sbi->devs->rwsem);
298 				return PTR_ERR(page);
299 			}
300 			ptr = kmap(page);
301 		}
302 		dis = ptr + erofs_blkoff(pos);
303 
304 		bdev = blkdev_get_by_path(dif->path,
305 					  FMODE_READ | FMODE_EXCL,
306 					  sb->s_type);
307 		if (IS_ERR(bdev)) {
308 			err = PTR_ERR(bdev);
309 			goto err_out;
310 		}
311 		dif->bdev = bdev;
312 		dif->dax_dev = fs_dax_get_by_bdev(bdev);
313 		dif->blocks = le32_to_cpu(dis->blocks);
314 		dif->mapped_blkaddr = le32_to_cpu(dis->mapped_blkaddr);
315 		sbi->total_blocks += dif->blocks;
316 		pos += EROFS_DEVT_SLOT_SIZE;
317 	}
318 err_out:
319 	up_read(&sbi->devs->rwsem);
320 	if (page) {
321 		kunmap(page);
322 		unlock_page(page);
323 		put_page(page);
324 	}
325 	return err;
326 }
327 
328 static int erofs_read_superblock(struct super_block *sb)
329 {
330 	struct erofs_sb_info *sbi;
331 	struct page *page;
332 	struct erofs_super_block *dsb;
333 	unsigned int blkszbits;
334 	void *data;
335 	int ret;
336 
337 	page = read_mapping_page(sb->s_bdev->bd_inode->i_mapping, 0, NULL);
338 	if (IS_ERR(page)) {
339 		erofs_err(sb, "cannot read erofs superblock");
340 		return PTR_ERR(page);
341 	}
342 
343 	sbi = EROFS_SB(sb);
344 
345 	data = kmap(page);
346 	dsb = (struct erofs_super_block *)(data + EROFS_SUPER_OFFSET);
347 
348 	ret = -EINVAL;
349 	if (le32_to_cpu(dsb->magic) != EROFS_SUPER_MAGIC_V1) {
350 		erofs_err(sb, "cannot find valid erofs superblock");
351 		goto out;
352 	}
353 
354 	sbi->feature_compat = le32_to_cpu(dsb->feature_compat);
355 	if (erofs_sb_has_sb_chksum(sbi)) {
356 		ret = erofs_superblock_csum_verify(sb, data);
357 		if (ret)
358 			goto out;
359 	}
360 
361 	ret = -EINVAL;
362 	blkszbits = dsb->blkszbits;
363 	/* 9(512 bytes) + LOG_SECTORS_PER_BLOCK == LOG_BLOCK_SIZE */
364 	if (blkszbits != LOG_BLOCK_SIZE) {
365 		erofs_err(sb, "blkszbits %u isn't supported on this platform",
366 			  blkszbits);
367 		goto out;
368 	}
369 
370 	if (!check_layout_compatibility(sb, dsb))
371 		goto out;
372 
373 	sbi->sb_size = 128 + dsb->sb_extslots * EROFS_SB_EXTSLOT_SIZE;
374 	if (sbi->sb_size > EROFS_BLKSIZ) {
375 		erofs_err(sb, "invalid sb_extslots %u (more than a fs block)",
376 			  sbi->sb_size);
377 		goto out;
378 	}
379 	sbi->primarydevice_blocks = le32_to_cpu(dsb->blocks);
380 	sbi->meta_blkaddr = le32_to_cpu(dsb->meta_blkaddr);
381 #ifdef CONFIG_EROFS_FS_XATTR
382 	sbi->xattr_blkaddr = le32_to_cpu(dsb->xattr_blkaddr);
383 #endif
384 	sbi->islotbits = ilog2(sizeof(struct erofs_inode_compact));
385 	sbi->root_nid = le16_to_cpu(dsb->root_nid);
386 	sbi->inos = le64_to_cpu(dsb->inos);
387 
388 	sbi->build_time = le64_to_cpu(dsb->build_time);
389 	sbi->build_time_nsec = le32_to_cpu(dsb->build_time_nsec);
390 
391 	memcpy(&sb->s_uuid, dsb->uuid, sizeof(dsb->uuid));
392 
393 	ret = strscpy(sbi->volume_name, dsb->volume_name,
394 		      sizeof(dsb->volume_name));
395 	if (ret < 0) {	/* -E2BIG */
396 		erofs_err(sb, "bad volume name without NIL terminator");
397 		ret = -EFSCORRUPTED;
398 		goto out;
399 	}
400 
401 	/* parse on-disk compression configurations */
402 	if (erofs_sb_has_compr_cfgs(sbi))
403 		ret = erofs_load_compr_cfgs(sb, dsb);
404 	else
405 		ret = z_erofs_load_lz4_config(sb, dsb, NULL, 0);
406 	if (ret < 0)
407 		goto out;
408 
409 	/* handle multiple devices */
410 	ret = erofs_init_devices(sb, dsb);
411 out:
412 	kunmap(page);
413 	put_page(page);
414 	return ret;
415 }
416 
417 /* set up default EROFS parameters */
418 static void erofs_default_options(struct erofs_fs_context *ctx)
419 {
420 #ifdef CONFIG_EROFS_FS_ZIP
421 	ctx->opt.cache_strategy = EROFS_ZIP_CACHE_READAROUND;
422 	ctx->opt.max_sync_decompress_pages = 3;
423 	ctx->opt.readahead_sync_decompress = false;
424 #endif
425 #ifdef CONFIG_EROFS_FS_XATTR
426 	set_opt(&ctx->opt, XATTR_USER);
427 #endif
428 #ifdef CONFIG_EROFS_FS_POSIX_ACL
429 	set_opt(&ctx->opt, POSIX_ACL);
430 #endif
431 }
432 
433 enum {
434 	Opt_user_xattr,
435 	Opt_acl,
436 	Opt_cache_strategy,
437 	Opt_dax,
438 	Opt_dax_enum,
439 	Opt_device,
440 	Opt_err
441 };
442 
443 static const struct constant_table erofs_param_cache_strategy[] = {
444 	{"disabled",	EROFS_ZIP_CACHE_DISABLED},
445 	{"readahead",	EROFS_ZIP_CACHE_READAHEAD},
446 	{"readaround",	EROFS_ZIP_CACHE_READAROUND},
447 	{}
448 };
449 
450 static const struct constant_table erofs_dax_param_enums[] = {
451 	{"always",	EROFS_MOUNT_DAX_ALWAYS},
452 	{"never",	EROFS_MOUNT_DAX_NEVER},
453 	{}
454 };
455 
456 static const struct fs_parameter_spec erofs_fs_parameters[] = {
457 	fsparam_flag_no("user_xattr",	Opt_user_xattr),
458 	fsparam_flag_no("acl",		Opt_acl),
459 	fsparam_enum("cache_strategy",	Opt_cache_strategy,
460 		     erofs_param_cache_strategy),
461 	fsparam_flag("dax",             Opt_dax),
462 	fsparam_enum("dax",		Opt_dax_enum, erofs_dax_param_enums),
463 	fsparam_string("device",	Opt_device),
464 	{}
465 };
466 
467 static bool erofs_fc_set_dax_mode(struct fs_context *fc, unsigned int mode)
468 {
469 #ifdef CONFIG_FS_DAX
470 	struct erofs_fs_context *ctx = fc->fs_private;
471 
472 	switch (mode) {
473 	case EROFS_MOUNT_DAX_ALWAYS:
474 		warnfc(fc, "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
475 		set_opt(&ctx->opt, DAX_ALWAYS);
476 		clear_opt(&ctx->opt, DAX_NEVER);
477 		return true;
478 	case EROFS_MOUNT_DAX_NEVER:
479 		set_opt(&ctx->opt, DAX_NEVER);
480 		clear_opt(&ctx->opt, DAX_ALWAYS);
481 		return true;
482 	default:
483 		DBG_BUGON(1);
484 		return false;
485 	}
486 #else
487 	errorfc(fc, "dax options not supported");
488 	return false;
489 #endif
490 }
491 
492 static int erofs_fc_parse_param(struct fs_context *fc,
493 				struct fs_parameter *param)
494 {
495 	struct erofs_fs_context *ctx = fc->fs_private;
496 	struct fs_parse_result result;
497 	struct erofs_device_info *dif;
498 	int opt, ret;
499 
500 	opt = fs_parse(fc, erofs_fs_parameters, param, &result);
501 	if (opt < 0)
502 		return opt;
503 
504 	switch (opt) {
505 	case Opt_user_xattr:
506 #ifdef CONFIG_EROFS_FS_XATTR
507 		if (result.boolean)
508 			set_opt(&ctx->opt, XATTR_USER);
509 		else
510 			clear_opt(&ctx->opt, XATTR_USER);
511 #else
512 		errorfc(fc, "{,no}user_xattr options not supported");
513 #endif
514 		break;
515 	case Opt_acl:
516 #ifdef CONFIG_EROFS_FS_POSIX_ACL
517 		if (result.boolean)
518 			set_opt(&ctx->opt, POSIX_ACL);
519 		else
520 			clear_opt(&ctx->opt, POSIX_ACL);
521 #else
522 		errorfc(fc, "{,no}acl options not supported");
523 #endif
524 		break;
525 	case Opt_cache_strategy:
526 #ifdef CONFIG_EROFS_FS_ZIP
527 		ctx->opt.cache_strategy = result.uint_32;
528 #else
529 		errorfc(fc, "compression not supported, cache_strategy ignored");
530 #endif
531 		break;
532 	case Opt_dax:
533 		if (!erofs_fc_set_dax_mode(fc, EROFS_MOUNT_DAX_ALWAYS))
534 			return -EINVAL;
535 		break;
536 	case Opt_dax_enum:
537 		if (!erofs_fc_set_dax_mode(fc, result.uint_32))
538 			return -EINVAL;
539 		break;
540 	case Opt_device:
541 		dif = kzalloc(sizeof(*dif), GFP_KERNEL);
542 		if (!dif)
543 			return -ENOMEM;
544 		dif->path = kstrdup(param->string, GFP_KERNEL);
545 		if (!dif->path) {
546 			kfree(dif);
547 			return -ENOMEM;
548 		}
549 		down_write(&ctx->devs->rwsem);
550 		ret = idr_alloc(&ctx->devs->tree, dif, 0, 0, GFP_KERNEL);
551 		up_write(&ctx->devs->rwsem);
552 		if (ret < 0) {
553 			kfree(dif->path);
554 			kfree(dif);
555 			return ret;
556 		}
557 		++ctx->devs->extra_devices;
558 		break;
559 	default:
560 		return -ENOPARAM;
561 	}
562 	return 0;
563 }
564 
565 #ifdef CONFIG_EROFS_FS_ZIP
566 static const struct address_space_operations managed_cache_aops;
567 
568 static int erofs_managed_cache_releasepage(struct page *page, gfp_t gfp_mask)
569 {
570 	int ret = 1;	/* 0 - busy */
571 	struct address_space *const mapping = page->mapping;
572 
573 	DBG_BUGON(!PageLocked(page));
574 	DBG_BUGON(mapping->a_ops != &managed_cache_aops);
575 
576 	if (PagePrivate(page))
577 		ret = erofs_try_to_free_cached_page(page);
578 
579 	return ret;
580 }
581 
582 static void erofs_managed_cache_invalidatepage(struct page *page,
583 					       unsigned int offset,
584 					       unsigned int length)
585 {
586 	const unsigned int stop = length + offset;
587 
588 	DBG_BUGON(!PageLocked(page));
589 
590 	/* Check for potential overflow in debug mode */
591 	DBG_BUGON(stop > PAGE_SIZE || stop < length);
592 
593 	if (offset == 0 && stop == PAGE_SIZE)
594 		while (!erofs_managed_cache_releasepage(page, GFP_NOFS))
595 			cond_resched();
596 }
597 
598 static const struct address_space_operations managed_cache_aops = {
599 	.releasepage = erofs_managed_cache_releasepage,
600 	.invalidatepage = erofs_managed_cache_invalidatepage,
601 };
602 
603 static int erofs_init_managed_cache(struct super_block *sb)
604 {
605 	struct erofs_sb_info *const sbi = EROFS_SB(sb);
606 	struct inode *const inode = new_inode(sb);
607 
608 	if (!inode)
609 		return -ENOMEM;
610 
611 	set_nlink(inode, 1);
612 	inode->i_size = OFFSET_MAX;
613 
614 	inode->i_mapping->a_ops = &managed_cache_aops;
615 	mapping_set_gfp_mask(inode->i_mapping,
616 			     GFP_NOFS | __GFP_HIGHMEM | __GFP_MOVABLE);
617 	sbi->managed_cache = inode;
618 	return 0;
619 }
620 #else
621 static int erofs_init_managed_cache(struct super_block *sb) { return 0; }
622 #endif
623 
624 static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
625 {
626 	struct inode *inode;
627 	struct erofs_sb_info *sbi;
628 	struct erofs_fs_context *ctx = fc->fs_private;
629 	int err;
630 
631 	sb->s_magic = EROFS_SUPER_MAGIC;
632 
633 	if (!sb_set_blocksize(sb, EROFS_BLKSIZ)) {
634 		erofs_err(sb, "failed to set erofs blksize");
635 		return -EINVAL;
636 	}
637 
638 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
639 	if (!sbi)
640 		return -ENOMEM;
641 
642 	sb->s_fs_info = sbi;
643 	sbi->opt = ctx->opt;
644 	sbi->dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
645 	sbi->devs = ctx->devs;
646 	ctx->devs = NULL;
647 
648 	err = erofs_read_superblock(sb);
649 	if (err)
650 		return err;
651 
652 	if (test_opt(&sbi->opt, DAX_ALWAYS) &&
653 	    !dax_supported(sbi->dax_dev, sb->s_bdev, EROFS_BLKSIZ, 0, bdev_nr_sectors(sb->s_bdev))) {
654 		errorfc(fc, "DAX unsupported by block device. Turning off DAX.");
655 		clear_opt(&sbi->opt, DAX_ALWAYS);
656 	}
657 	sb->s_flags |= SB_RDONLY | SB_NOATIME;
658 	sb->s_maxbytes = MAX_LFS_FILESIZE;
659 	sb->s_time_gran = 1;
660 
661 	sb->s_op = &erofs_sops;
662 	sb->s_xattr = erofs_xattr_handlers;
663 
664 	if (test_opt(&sbi->opt, POSIX_ACL))
665 		sb->s_flags |= SB_POSIXACL;
666 	else
667 		sb->s_flags &= ~SB_POSIXACL;
668 
669 #ifdef CONFIG_EROFS_FS_ZIP
670 	xa_init(&sbi->managed_pslots);
671 #endif
672 
673 	/* get the root inode */
674 	inode = erofs_iget(sb, ROOT_NID(sbi), true);
675 	if (IS_ERR(inode))
676 		return PTR_ERR(inode);
677 
678 	if (!S_ISDIR(inode->i_mode)) {
679 		erofs_err(sb, "rootino(nid %llu) is not a directory(i_mode %o)",
680 			  ROOT_NID(sbi), inode->i_mode);
681 		iput(inode);
682 		return -EINVAL;
683 	}
684 
685 	sb->s_root = d_make_root(inode);
686 	if (!sb->s_root)
687 		return -ENOMEM;
688 
689 	erofs_shrinker_register(sb);
690 	/* sb->s_umount is already locked, SB_ACTIVE and SB_BORN are not set */
691 	err = erofs_init_managed_cache(sb);
692 	if (err)
693 		return err;
694 
695 	erofs_info(sb, "mounted with root inode @ nid %llu.", ROOT_NID(sbi));
696 	return 0;
697 }
698 
699 static int erofs_fc_get_tree(struct fs_context *fc)
700 {
701 	return get_tree_bdev(fc, erofs_fc_fill_super);
702 }
703 
704 static int erofs_fc_reconfigure(struct fs_context *fc)
705 {
706 	struct super_block *sb = fc->root->d_sb;
707 	struct erofs_sb_info *sbi = EROFS_SB(sb);
708 	struct erofs_fs_context *ctx = fc->fs_private;
709 
710 	DBG_BUGON(!sb_rdonly(sb));
711 
712 	if (test_opt(&ctx->opt, POSIX_ACL))
713 		fc->sb_flags |= SB_POSIXACL;
714 	else
715 		fc->sb_flags &= ~SB_POSIXACL;
716 
717 	sbi->opt = ctx->opt;
718 
719 	fc->sb_flags |= SB_RDONLY;
720 	return 0;
721 }
722 
723 static int erofs_release_device_info(int id, void *ptr, void *data)
724 {
725 	struct erofs_device_info *dif = ptr;
726 
727 	fs_put_dax(dif->dax_dev);
728 	if (dif->bdev)
729 		blkdev_put(dif->bdev, FMODE_READ | FMODE_EXCL);
730 	kfree(dif->path);
731 	kfree(dif);
732 	return 0;
733 }
734 
735 static void erofs_free_dev_context(struct erofs_dev_context *devs)
736 {
737 	if (!devs)
738 		return;
739 	idr_for_each(&devs->tree, &erofs_release_device_info, NULL);
740 	idr_destroy(&devs->tree);
741 	kfree(devs);
742 }
743 
744 static void erofs_fc_free(struct fs_context *fc)
745 {
746 	struct erofs_fs_context *ctx = fc->fs_private;
747 
748 	erofs_free_dev_context(ctx->devs);
749 	kfree(ctx);
750 }
751 
752 static const struct fs_context_operations erofs_context_ops = {
753 	.parse_param	= erofs_fc_parse_param,
754 	.get_tree       = erofs_fc_get_tree,
755 	.reconfigure    = erofs_fc_reconfigure,
756 	.free		= erofs_fc_free,
757 };
758 
759 static int erofs_init_fs_context(struct fs_context *fc)
760 {
761 	struct erofs_fs_context *ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
762 
763 	if (!ctx)
764 		return -ENOMEM;
765 	ctx->devs = kzalloc(sizeof(struct erofs_dev_context), GFP_KERNEL);
766 	if (!ctx->devs) {
767 		kfree(ctx);
768 		return -ENOMEM;
769 	}
770 	fc->fs_private = ctx;
771 
772 	idr_init(&ctx->devs->tree);
773 	init_rwsem(&ctx->devs->rwsem);
774 	erofs_default_options(ctx);
775 	fc->ops = &erofs_context_ops;
776 	return 0;
777 }
778 
779 /*
780  * could be triggered after deactivate_locked_super()
781  * is called, thus including umount and failed to initialize.
782  */
783 static void erofs_kill_sb(struct super_block *sb)
784 {
785 	struct erofs_sb_info *sbi;
786 
787 	WARN_ON(sb->s_magic != EROFS_SUPER_MAGIC);
788 
789 	kill_block_super(sb);
790 
791 	sbi = EROFS_SB(sb);
792 	if (!sbi)
793 		return;
794 
795 	erofs_free_dev_context(sbi->devs);
796 	fs_put_dax(sbi->dax_dev);
797 	kfree(sbi);
798 	sb->s_fs_info = NULL;
799 }
800 
801 /* called when ->s_root is non-NULL */
802 static void erofs_put_super(struct super_block *sb)
803 {
804 	struct erofs_sb_info *const sbi = EROFS_SB(sb);
805 
806 	DBG_BUGON(!sbi);
807 
808 	erofs_shrinker_unregister(sb);
809 #ifdef CONFIG_EROFS_FS_ZIP
810 	iput(sbi->managed_cache);
811 	sbi->managed_cache = NULL;
812 #endif
813 }
814 
815 static struct file_system_type erofs_fs_type = {
816 	.owner          = THIS_MODULE,
817 	.name           = "erofs",
818 	.init_fs_context = erofs_init_fs_context,
819 	.kill_sb        = erofs_kill_sb,
820 	.fs_flags       = FS_REQUIRES_DEV,
821 };
822 MODULE_ALIAS_FS("erofs");
823 
824 static int __init erofs_module_init(void)
825 {
826 	int err;
827 
828 	erofs_check_ondisk_layout_definitions();
829 
830 	erofs_inode_cachep = kmem_cache_create("erofs_inode",
831 					       sizeof(struct erofs_inode), 0,
832 					       SLAB_RECLAIM_ACCOUNT,
833 					       erofs_inode_init_once);
834 	if (!erofs_inode_cachep) {
835 		err = -ENOMEM;
836 		goto icache_err;
837 	}
838 
839 	err = erofs_init_shrinker();
840 	if (err)
841 		goto shrinker_err;
842 
843 	erofs_pcpubuf_init();
844 	err = z_erofs_init_zip_subsystem();
845 	if (err)
846 		goto zip_err;
847 
848 	err = register_filesystem(&erofs_fs_type);
849 	if (err)
850 		goto fs_err;
851 
852 	return 0;
853 
854 fs_err:
855 	z_erofs_exit_zip_subsystem();
856 zip_err:
857 	erofs_exit_shrinker();
858 shrinker_err:
859 	kmem_cache_destroy(erofs_inode_cachep);
860 icache_err:
861 	return err;
862 }
863 
864 static void __exit erofs_module_exit(void)
865 {
866 	unregister_filesystem(&erofs_fs_type);
867 	z_erofs_exit_zip_subsystem();
868 	erofs_exit_shrinker();
869 
870 	/* Ensure all RCU free inodes are safe before cache is destroyed. */
871 	rcu_barrier();
872 	kmem_cache_destroy(erofs_inode_cachep);
873 	erofs_pcpubuf_exit();
874 }
875 
876 /* get filesystem statistics */
877 static int erofs_statfs(struct dentry *dentry, struct kstatfs *buf)
878 {
879 	struct super_block *sb = dentry->d_sb;
880 	struct erofs_sb_info *sbi = EROFS_SB(sb);
881 	u64 id = huge_encode_dev(sb->s_bdev->bd_dev);
882 
883 	buf->f_type = sb->s_magic;
884 	buf->f_bsize = EROFS_BLKSIZ;
885 	buf->f_blocks = sbi->total_blocks;
886 	buf->f_bfree = buf->f_bavail = 0;
887 
888 	buf->f_files = ULLONG_MAX;
889 	buf->f_ffree = ULLONG_MAX - sbi->inos;
890 
891 	buf->f_namelen = EROFS_NAME_LEN;
892 
893 	buf->f_fsid    = u64_to_fsid(id);
894 	return 0;
895 }
896 
897 static int erofs_show_options(struct seq_file *seq, struct dentry *root)
898 {
899 	struct erofs_sb_info *sbi = EROFS_SB(root->d_sb);
900 	struct erofs_mount_opts *opt = &sbi->opt;
901 
902 #ifdef CONFIG_EROFS_FS_XATTR
903 	if (test_opt(opt, XATTR_USER))
904 		seq_puts(seq, ",user_xattr");
905 	else
906 		seq_puts(seq, ",nouser_xattr");
907 #endif
908 #ifdef CONFIG_EROFS_FS_POSIX_ACL
909 	if (test_opt(opt, POSIX_ACL))
910 		seq_puts(seq, ",acl");
911 	else
912 		seq_puts(seq, ",noacl");
913 #endif
914 #ifdef CONFIG_EROFS_FS_ZIP
915 	if (opt->cache_strategy == EROFS_ZIP_CACHE_DISABLED)
916 		seq_puts(seq, ",cache_strategy=disabled");
917 	else if (opt->cache_strategy == EROFS_ZIP_CACHE_READAHEAD)
918 		seq_puts(seq, ",cache_strategy=readahead");
919 	else if (opt->cache_strategy == EROFS_ZIP_CACHE_READAROUND)
920 		seq_puts(seq, ",cache_strategy=readaround");
921 #endif
922 	if (test_opt(opt, DAX_ALWAYS))
923 		seq_puts(seq, ",dax=always");
924 	if (test_opt(opt, DAX_NEVER))
925 		seq_puts(seq, ",dax=never");
926 	return 0;
927 }
928 
929 const struct super_operations erofs_sops = {
930 	.put_super = erofs_put_super,
931 	.alloc_inode = erofs_alloc_inode,
932 	.free_inode = erofs_free_inode,
933 	.statfs = erofs_statfs,
934 	.show_options = erofs_show_options,
935 };
936 
937 module_init(erofs_module_init);
938 module_exit(erofs_module_exit);
939 
940 MODULE_DESCRIPTION("Enhanced ROM File System");
941 MODULE_AUTHOR("Gao Xiang, Chao Yu, Miao Xie, CONSUMER BG, HUAWEI Inc.");
942 MODULE_LICENSE("GPL");
943