xref: /linux/fs/erofs/super.c (revision 7f71507851fc7764b36a3221839607d3a45c2025)
1 // SPDX-License-Identifier: GPL-2.0-only
2 /*
3  * Copyright (C) 2017-2018 HUAWEI, Inc.
4  *             https://www.huawei.com/
5  * Copyright (C) 2021, Alibaba Cloud
6  */
7 #include <linux/statfs.h>
8 #include <linux/seq_file.h>
9 #include <linux/crc32c.h>
10 #include <linux/fs_context.h>
11 #include <linux/fs_parser.h>
12 #include <linux/exportfs.h>
13 #include <linux/backing-dev.h>
14 #include "xattr.h"
15 
16 #define CREATE_TRACE_POINTS
17 #include <trace/events/erofs.h>
18 
19 static struct kmem_cache *erofs_inode_cachep __read_mostly;
20 
21 void _erofs_printk(struct super_block *sb, const char *fmt, ...)
22 {
23 	struct va_format vaf;
24 	va_list args;
25 	int level;
26 
27 	va_start(args, fmt);
28 
29 	level = printk_get_level(fmt);
30 	vaf.fmt = printk_skip_level(fmt);
31 	vaf.va = &args;
32 	if (sb)
33 		printk("%c%cerofs (device %s): %pV",
34 				KERN_SOH_ASCII, level, sb->s_id, &vaf);
35 	else
36 		printk("%c%cerofs: %pV", KERN_SOH_ASCII, level, &vaf);
37 	va_end(args);
38 }
39 
40 static int erofs_superblock_csum_verify(struct super_block *sb, void *sbdata)
41 {
42 	size_t len = 1 << EROFS_SB(sb)->blkszbits;
43 	struct erofs_super_block *dsb;
44 	u32 expected_crc, crc;
45 
46 	if (len > EROFS_SUPER_OFFSET)
47 		len -= EROFS_SUPER_OFFSET;
48 
49 	dsb = kmemdup(sbdata + EROFS_SUPER_OFFSET, len, GFP_KERNEL);
50 	if (!dsb)
51 		return -ENOMEM;
52 
53 	expected_crc = le32_to_cpu(dsb->checksum);
54 	dsb->checksum = 0;
55 	/* to allow for x86 boot sectors and other oddities. */
56 	crc = crc32c(~0, dsb, len);
57 	kfree(dsb);
58 
59 	if (crc != expected_crc) {
60 		erofs_err(sb, "invalid checksum 0x%08x, 0x%08x expected",
61 			  crc, expected_crc);
62 		return -EBADMSG;
63 	}
64 	return 0;
65 }
66 
67 static void erofs_inode_init_once(void *ptr)
68 {
69 	struct erofs_inode *vi = ptr;
70 
71 	inode_init_once(&vi->vfs_inode);
72 }
73 
74 static struct inode *erofs_alloc_inode(struct super_block *sb)
75 {
76 	struct erofs_inode *vi =
77 		alloc_inode_sb(sb, erofs_inode_cachep, GFP_KERNEL);
78 
79 	if (!vi)
80 		return NULL;
81 
82 	/* zero out everything except vfs_inode */
83 	memset(vi, 0, offsetof(struct erofs_inode, vfs_inode));
84 	return &vi->vfs_inode;
85 }
86 
87 static void erofs_free_inode(struct inode *inode)
88 {
89 	struct erofs_inode *vi = EROFS_I(inode);
90 
91 	if (inode->i_op == &erofs_fast_symlink_iops)
92 		kfree(inode->i_link);
93 	kfree(vi->xattr_shared_xattrs);
94 	kmem_cache_free(erofs_inode_cachep, vi);
95 }
96 
97 /* read variable-sized metadata, offset will be aligned by 4-byte */
98 void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf,
99 			  erofs_off_t *offset, int *lengthp)
100 {
101 	u8 *buffer, *ptr;
102 	int len, i, cnt;
103 
104 	*offset = round_up(*offset, 4);
105 	ptr = erofs_bread(buf, *offset, EROFS_KMAP);
106 	if (IS_ERR(ptr))
107 		return ptr;
108 
109 	len = le16_to_cpu(*(__le16 *)ptr);
110 	if (!len)
111 		len = U16_MAX + 1;
112 	buffer = kmalloc(len, GFP_KERNEL);
113 	if (!buffer)
114 		return ERR_PTR(-ENOMEM);
115 	*offset += sizeof(__le16);
116 	*lengthp = len;
117 
118 	for (i = 0; i < len; i += cnt) {
119 		cnt = min_t(int, sb->s_blocksize - erofs_blkoff(sb, *offset),
120 			    len - i);
121 		ptr = erofs_bread(buf, *offset, EROFS_KMAP);
122 		if (IS_ERR(ptr)) {
123 			kfree(buffer);
124 			return ptr;
125 		}
126 		memcpy(buffer + i, ptr, cnt);
127 		*offset += cnt;
128 	}
129 	return buffer;
130 }
131 
132 #ifndef CONFIG_EROFS_FS_ZIP
133 static int z_erofs_parse_cfgs(struct super_block *sb,
134 			      struct erofs_super_block *dsb)
135 {
136 	if (!dsb->u1.available_compr_algs)
137 		return 0;
138 
139 	erofs_err(sb, "compression disabled, unable to mount compressed EROFS");
140 	return -EOPNOTSUPP;
141 }
142 #endif
143 
144 static int erofs_init_device(struct erofs_buf *buf, struct super_block *sb,
145 			     struct erofs_device_info *dif, erofs_off_t *pos)
146 {
147 	struct erofs_sb_info *sbi = EROFS_SB(sb);
148 	struct erofs_fscache *fscache;
149 	struct erofs_deviceslot *dis;
150 	struct file *file;
151 
152 	dis = erofs_read_metabuf(buf, sb, *pos, EROFS_KMAP);
153 	if (IS_ERR(dis))
154 		return PTR_ERR(dis);
155 
156 	if (!sbi->devs->flatdev && !dif->path) {
157 		if (!dis->tag[0]) {
158 			erofs_err(sb, "empty device tag @ pos %llu", *pos);
159 			return -EINVAL;
160 		}
161 		dif->path = kmemdup_nul(dis->tag, sizeof(dis->tag), GFP_KERNEL);
162 		if (!dif->path)
163 			return -ENOMEM;
164 	}
165 
166 	if (erofs_is_fscache_mode(sb)) {
167 		fscache = erofs_fscache_register_cookie(sb, dif->path, 0);
168 		if (IS_ERR(fscache))
169 			return PTR_ERR(fscache);
170 		dif->fscache = fscache;
171 	} else if (!sbi->devs->flatdev) {
172 		file = erofs_is_fileio_mode(sbi) ?
173 				filp_open(dif->path, O_RDONLY | O_LARGEFILE, 0) :
174 				bdev_file_open_by_path(dif->path,
175 						BLK_OPEN_READ, sb->s_type, NULL);
176 		if (IS_ERR(file))
177 			return PTR_ERR(file);
178 
179 		if (!erofs_is_fileio_mode(sbi)) {
180 			dif->dax_dev = fs_dax_get_by_bdev(file_bdev(file),
181 					&dif->dax_part_off, NULL, NULL);
182 		} else if (!S_ISREG(file_inode(file)->i_mode)) {
183 			fput(file);
184 			return -EINVAL;
185 		}
186 		dif->file = file;
187 	}
188 
189 	dif->blocks = le32_to_cpu(dis->blocks);
190 	dif->mapped_blkaddr = le32_to_cpu(dis->mapped_blkaddr);
191 	sbi->total_blocks += dif->blocks;
192 	*pos += EROFS_DEVT_SLOT_SIZE;
193 	return 0;
194 }
195 
196 static int erofs_scan_devices(struct super_block *sb,
197 			      struct erofs_super_block *dsb)
198 {
199 	struct erofs_sb_info *sbi = EROFS_SB(sb);
200 	unsigned int ondisk_extradevs;
201 	erofs_off_t pos;
202 	struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
203 	struct erofs_device_info *dif;
204 	int id, err = 0;
205 
206 	sbi->total_blocks = sbi->primarydevice_blocks;
207 	if (!erofs_sb_has_device_table(sbi))
208 		ondisk_extradevs = 0;
209 	else
210 		ondisk_extradevs = le16_to_cpu(dsb->extra_devices);
211 
212 	if (sbi->devs->extra_devices &&
213 	    ondisk_extradevs != sbi->devs->extra_devices) {
214 		erofs_err(sb, "extra devices don't match (ondisk %u, given %u)",
215 			  ondisk_extradevs, sbi->devs->extra_devices);
216 		return -EINVAL;
217 	}
218 	if (!ondisk_extradevs)
219 		return 0;
220 
221 	if (!sbi->devs->extra_devices && !erofs_is_fscache_mode(sb))
222 		sbi->devs->flatdev = true;
223 
224 	sbi->device_id_mask = roundup_pow_of_two(ondisk_extradevs + 1) - 1;
225 	pos = le16_to_cpu(dsb->devt_slotoff) * EROFS_DEVT_SLOT_SIZE;
226 	down_read(&sbi->devs->rwsem);
227 	if (sbi->devs->extra_devices) {
228 		idr_for_each_entry(&sbi->devs->tree, dif, id) {
229 			err = erofs_init_device(&buf, sb, dif, &pos);
230 			if (err)
231 				break;
232 		}
233 	} else {
234 		for (id = 0; id < ondisk_extradevs; id++) {
235 			dif = kzalloc(sizeof(*dif), GFP_KERNEL);
236 			if (!dif) {
237 				err = -ENOMEM;
238 				break;
239 			}
240 
241 			err = idr_alloc(&sbi->devs->tree, dif, 0, 0, GFP_KERNEL);
242 			if (err < 0) {
243 				kfree(dif);
244 				break;
245 			}
246 			++sbi->devs->extra_devices;
247 
248 			err = erofs_init_device(&buf, sb, dif, &pos);
249 			if (err)
250 				break;
251 		}
252 	}
253 	up_read(&sbi->devs->rwsem);
254 	erofs_put_metabuf(&buf);
255 	return err;
256 }
257 
258 static int erofs_read_superblock(struct super_block *sb)
259 {
260 	struct erofs_sb_info *sbi = EROFS_SB(sb);
261 	struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
262 	struct erofs_super_block *dsb;
263 	void *data;
264 	int ret;
265 
266 	data = erofs_read_metabuf(&buf, sb, 0, EROFS_KMAP);
267 	if (IS_ERR(data)) {
268 		erofs_err(sb, "cannot read erofs superblock");
269 		return PTR_ERR(data);
270 	}
271 
272 	dsb = (struct erofs_super_block *)(data + EROFS_SUPER_OFFSET);
273 	ret = -EINVAL;
274 	if (le32_to_cpu(dsb->magic) != EROFS_SUPER_MAGIC_V1) {
275 		erofs_err(sb, "cannot find valid erofs superblock");
276 		goto out;
277 	}
278 
279 	sbi->blkszbits  = dsb->blkszbits;
280 	if (sbi->blkszbits < 9 || sbi->blkszbits > PAGE_SHIFT) {
281 		erofs_err(sb, "blkszbits %u isn't supported", sbi->blkszbits);
282 		goto out;
283 	}
284 	if (dsb->dirblkbits) {
285 		erofs_err(sb, "dirblkbits %u isn't supported", dsb->dirblkbits);
286 		goto out;
287 	}
288 
289 	sbi->feature_compat = le32_to_cpu(dsb->feature_compat);
290 	if (erofs_sb_has_sb_chksum(sbi)) {
291 		ret = erofs_superblock_csum_verify(sb, data);
292 		if (ret)
293 			goto out;
294 	}
295 
296 	ret = -EINVAL;
297 	sbi->feature_incompat = le32_to_cpu(dsb->feature_incompat);
298 	if (sbi->feature_incompat & ~EROFS_ALL_FEATURE_INCOMPAT) {
299 		erofs_err(sb, "unidentified incompatible feature %x, please upgrade kernel",
300 			  sbi->feature_incompat & ~EROFS_ALL_FEATURE_INCOMPAT);
301 		goto out;
302 	}
303 
304 	sbi->sb_size = 128 + dsb->sb_extslots * EROFS_SB_EXTSLOT_SIZE;
305 	if (sbi->sb_size > PAGE_SIZE - EROFS_SUPER_OFFSET) {
306 		erofs_err(sb, "invalid sb_extslots %u (more than a fs block)",
307 			  sbi->sb_size);
308 		goto out;
309 	}
310 	sbi->primarydevice_blocks = le32_to_cpu(dsb->blocks);
311 	sbi->meta_blkaddr = le32_to_cpu(dsb->meta_blkaddr);
312 #ifdef CONFIG_EROFS_FS_XATTR
313 	sbi->xattr_blkaddr = le32_to_cpu(dsb->xattr_blkaddr);
314 	sbi->xattr_prefix_start = le32_to_cpu(dsb->xattr_prefix_start);
315 	sbi->xattr_prefix_count = dsb->xattr_prefix_count;
316 	sbi->xattr_filter_reserved = dsb->xattr_filter_reserved;
317 #endif
318 	sbi->islotbits = ilog2(sizeof(struct erofs_inode_compact));
319 	sbi->root_nid = le16_to_cpu(dsb->root_nid);
320 	sbi->packed_nid = le64_to_cpu(dsb->packed_nid);
321 	sbi->inos = le64_to_cpu(dsb->inos);
322 
323 	sbi->build_time = le64_to_cpu(dsb->build_time);
324 	sbi->build_time_nsec = le32_to_cpu(dsb->build_time_nsec);
325 
326 	super_set_uuid(sb, (void *)dsb->uuid, sizeof(dsb->uuid));
327 
328 	ret = strscpy(sbi->volume_name, dsb->volume_name,
329 		      sizeof(dsb->volume_name));
330 	if (ret < 0) {	/* -E2BIG */
331 		erofs_err(sb, "bad volume name without NIL terminator");
332 		ret = -EFSCORRUPTED;
333 		goto out;
334 	}
335 
336 	/* parse on-disk compression configurations */
337 	ret = z_erofs_parse_cfgs(sb, dsb);
338 	if (ret < 0)
339 		goto out;
340 
341 	/* handle multiple devices */
342 	ret = erofs_scan_devices(sb, dsb);
343 
344 	if (erofs_is_fscache_mode(sb))
345 		erofs_info(sb, "[deprecated] fscache-based on-demand read feature in use. Use at your own risk!");
346 out:
347 	erofs_put_metabuf(&buf);
348 	return ret;
349 }
350 
351 static void erofs_default_options(struct erofs_sb_info *sbi)
352 {
353 #ifdef CONFIG_EROFS_FS_ZIP
354 	sbi->opt.cache_strategy = EROFS_ZIP_CACHE_READAROUND;
355 	sbi->opt.max_sync_decompress_pages = 3;
356 	sbi->opt.sync_decompress = EROFS_SYNC_DECOMPRESS_AUTO;
357 #endif
358 #ifdef CONFIG_EROFS_FS_XATTR
359 	set_opt(&sbi->opt, XATTR_USER);
360 #endif
361 #ifdef CONFIG_EROFS_FS_POSIX_ACL
362 	set_opt(&sbi->opt, POSIX_ACL);
363 #endif
364 }
365 
366 enum {
367 	Opt_user_xattr,
368 	Opt_acl,
369 	Opt_cache_strategy,
370 	Opt_dax,
371 	Opt_dax_enum,
372 	Opt_device,
373 	Opt_fsid,
374 	Opt_domain_id,
375 	Opt_err
376 };
377 
378 static const struct constant_table erofs_param_cache_strategy[] = {
379 	{"disabled",	EROFS_ZIP_CACHE_DISABLED},
380 	{"readahead",	EROFS_ZIP_CACHE_READAHEAD},
381 	{"readaround",	EROFS_ZIP_CACHE_READAROUND},
382 	{}
383 };
384 
385 static const struct constant_table erofs_dax_param_enums[] = {
386 	{"always",	EROFS_MOUNT_DAX_ALWAYS},
387 	{"never",	EROFS_MOUNT_DAX_NEVER},
388 	{}
389 };
390 
391 static const struct fs_parameter_spec erofs_fs_parameters[] = {
392 	fsparam_flag_no("user_xattr",	Opt_user_xattr),
393 	fsparam_flag_no("acl",		Opt_acl),
394 	fsparam_enum("cache_strategy",	Opt_cache_strategy,
395 		     erofs_param_cache_strategy),
396 	fsparam_flag("dax",             Opt_dax),
397 	fsparam_enum("dax",		Opt_dax_enum, erofs_dax_param_enums),
398 	fsparam_string("device",	Opt_device),
399 	fsparam_string("fsid",		Opt_fsid),
400 	fsparam_string("domain_id",	Opt_domain_id),
401 	{}
402 };
403 
404 static bool erofs_fc_set_dax_mode(struct fs_context *fc, unsigned int mode)
405 {
406 #ifdef CONFIG_FS_DAX
407 	struct erofs_sb_info *sbi = fc->s_fs_info;
408 
409 	switch (mode) {
410 	case EROFS_MOUNT_DAX_ALWAYS:
411 		set_opt(&sbi->opt, DAX_ALWAYS);
412 		clear_opt(&sbi->opt, DAX_NEVER);
413 		return true;
414 	case EROFS_MOUNT_DAX_NEVER:
415 		set_opt(&sbi->opt, DAX_NEVER);
416 		clear_opt(&sbi->opt, DAX_ALWAYS);
417 		return true;
418 	default:
419 		DBG_BUGON(1);
420 		return false;
421 	}
422 #else
423 	errorfc(fc, "dax options not supported");
424 	return false;
425 #endif
426 }
427 
428 static int erofs_fc_parse_param(struct fs_context *fc,
429 				struct fs_parameter *param)
430 {
431 	struct erofs_sb_info *sbi = fc->s_fs_info;
432 	struct fs_parse_result result;
433 	struct erofs_device_info *dif;
434 	int opt, ret;
435 
436 	opt = fs_parse(fc, erofs_fs_parameters, param, &result);
437 	if (opt < 0)
438 		return opt;
439 
440 	switch (opt) {
441 	case Opt_user_xattr:
442 #ifdef CONFIG_EROFS_FS_XATTR
443 		if (result.boolean)
444 			set_opt(&sbi->opt, XATTR_USER);
445 		else
446 			clear_opt(&sbi->opt, XATTR_USER);
447 #else
448 		errorfc(fc, "{,no}user_xattr options not supported");
449 #endif
450 		break;
451 	case Opt_acl:
452 #ifdef CONFIG_EROFS_FS_POSIX_ACL
453 		if (result.boolean)
454 			set_opt(&sbi->opt, POSIX_ACL);
455 		else
456 			clear_opt(&sbi->opt, POSIX_ACL);
457 #else
458 		errorfc(fc, "{,no}acl options not supported");
459 #endif
460 		break;
461 	case Opt_cache_strategy:
462 #ifdef CONFIG_EROFS_FS_ZIP
463 		sbi->opt.cache_strategy = result.uint_32;
464 #else
465 		errorfc(fc, "compression not supported, cache_strategy ignored");
466 #endif
467 		break;
468 	case Opt_dax:
469 		if (!erofs_fc_set_dax_mode(fc, EROFS_MOUNT_DAX_ALWAYS))
470 			return -EINVAL;
471 		break;
472 	case Opt_dax_enum:
473 		if (!erofs_fc_set_dax_mode(fc, result.uint_32))
474 			return -EINVAL;
475 		break;
476 	case Opt_device:
477 		dif = kzalloc(sizeof(*dif), GFP_KERNEL);
478 		if (!dif)
479 			return -ENOMEM;
480 		dif->path = kstrdup(param->string, GFP_KERNEL);
481 		if (!dif->path) {
482 			kfree(dif);
483 			return -ENOMEM;
484 		}
485 		down_write(&sbi->devs->rwsem);
486 		ret = idr_alloc(&sbi->devs->tree, dif, 0, 0, GFP_KERNEL);
487 		up_write(&sbi->devs->rwsem);
488 		if (ret < 0) {
489 			kfree(dif->path);
490 			kfree(dif);
491 			return ret;
492 		}
493 		++sbi->devs->extra_devices;
494 		break;
495 #ifdef CONFIG_EROFS_FS_ONDEMAND
496 	case Opt_fsid:
497 		kfree(sbi->fsid);
498 		sbi->fsid = kstrdup(param->string, GFP_KERNEL);
499 		if (!sbi->fsid)
500 			return -ENOMEM;
501 		break;
502 	case Opt_domain_id:
503 		kfree(sbi->domain_id);
504 		sbi->domain_id = kstrdup(param->string, GFP_KERNEL);
505 		if (!sbi->domain_id)
506 			return -ENOMEM;
507 		break;
508 #else
509 	case Opt_fsid:
510 	case Opt_domain_id:
511 		errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
512 		break;
513 #endif
514 	default:
515 		return -ENOPARAM;
516 	}
517 	return 0;
518 }
519 
520 static struct inode *erofs_nfs_get_inode(struct super_block *sb,
521 					 u64 ino, u32 generation)
522 {
523 	return erofs_iget(sb, ino);
524 }
525 
526 static struct dentry *erofs_fh_to_dentry(struct super_block *sb,
527 		struct fid *fid, int fh_len, int fh_type)
528 {
529 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
530 				    erofs_nfs_get_inode);
531 }
532 
533 static struct dentry *erofs_fh_to_parent(struct super_block *sb,
534 		struct fid *fid, int fh_len, int fh_type)
535 {
536 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
537 				    erofs_nfs_get_inode);
538 }
539 
540 static struct dentry *erofs_get_parent(struct dentry *child)
541 {
542 	erofs_nid_t nid;
543 	unsigned int d_type;
544 	int err;
545 
546 	err = erofs_namei(d_inode(child), &dotdot_name, &nid, &d_type);
547 	if (err)
548 		return ERR_PTR(err);
549 	return d_obtain_alias(erofs_iget(child->d_sb, nid));
550 }
551 
552 static const struct export_operations erofs_export_ops = {
553 	.encode_fh = generic_encode_ino32_fh,
554 	.fh_to_dentry = erofs_fh_to_dentry,
555 	.fh_to_parent = erofs_fh_to_parent,
556 	.get_parent = erofs_get_parent,
557 };
558 
559 static void erofs_set_sysfs_name(struct super_block *sb)
560 {
561 	struct erofs_sb_info *sbi = EROFS_SB(sb);
562 
563 	if (sbi->domain_id)
564 		super_set_sysfs_name_generic(sb, "%s,%s", sbi->domain_id,
565 					     sbi->fsid);
566 	else if (sbi->fsid)
567 		super_set_sysfs_name_generic(sb, "%s", sbi->fsid);
568 	else if (erofs_is_fileio_mode(sbi))
569 		super_set_sysfs_name_generic(sb, "%s",
570 					     bdi_dev_name(sb->s_bdi));
571 	else
572 		super_set_sysfs_name_id(sb);
573 }
574 
575 static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
576 {
577 	struct inode *inode;
578 	struct erofs_sb_info *sbi = EROFS_SB(sb);
579 	int err;
580 
581 	sb->s_magic = EROFS_SUPER_MAGIC;
582 	sb->s_flags |= SB_RDONLY | SB_NOATIME;
583 	sb->s_maxbytes = MAX_LFS_FILESIZE;
584 	sb->s_op = &erofs_sops;
585 
586 	sbi->blkszbits = PAGE_SHIFT;
587 	if (!sb->s_bdev) {
588 		sb->s_blocksize = PAGE_SIZE;
589 		sb->s_blocksize_bits = PAGE_SHIFT;
590 
591 		if (erofs_is_fscache_mode(sb)) {
592 			err = erofs_fscache_register_fs(sb);
593 			if (err)
594 				return err;
595 		}
596 		err = super_setup_bdi(sb);
597 		if (err)
598 			return err;
599 	} else {
600 		if (!sb_set_blocksize(sb, PAGE_SIZE)) {
601 			errorfc(fc, "failed to set initial blksize");
602 			return -EINVAL;
603 		}
604 
605 		sbi->dax_dev = fs_dax_get_by_bdev(sb->s_bdev,
606 						  &sbi->dax_part_off,
607 						  NULL, NULL);
608 	}
609 
610 	err = erofs_read_superblock(sb);
611 	if (err)
612 		return err;
613 
614 	if (sb->s_blocksize_bits != sbi->blkszbits) {
615 		if (erofs_is_fscache_mode(sb)) {
616 			errorfc(fc, "unsupported blksize for fscache mode");
617 			return -EINVAL;
618 		}
619 
620 		if (erofs_is_fileio_mode(sbi)) {
621 			sb->s_blocksize = 1 << sbi->blkszbits;
622 			sb->s_blocksize_bits = sbi->blkszbits;
623 		} else if (!sb_set_blocksize(sb, 1 << sbi->blkszbits)) {
624 			errorfc(fc, "failed to set erofs blksize");
625 			return -EINVAL;
626 		}
627 	}
628 
629 	if (test_opt(&sbi->opt, DAX_ALWAYS)) {
630 		if (!sbi->dax_dev) {
631 			errorfc(fc, "DAX unsupported by block device. Turning off DAX.");
632 			clear_opt(&sbi->opt, DAX_ALWAYS);
633 		} else if (sbi->blkszbits != PAGE_SHIFT) {
634 			errorfc(fc, "unsupported blocksize for DAX");
635 			clear_opt(&sbi->opt, DAX_ALWAYS);
636 		}
637 	}
638 
639 	sb->s_time_gran = 1;
640 	sb->s_xattr = erofs_xattr_handlers;
641 	sb->s_export_op = &erofs_export_ops;
642 
643 	if (test_opt(&sbi->opt, POSIX_ACL))
644 		sb->s_flags |= SB_POSIXACL;
645 	else
646 		sb->s_flags &= ~SB_POSIXACL;
647 
648 #ifdef CONFIG_EROFS_FS_ZIP
649 	xa_init(&sbi->managed_pslots);
650 #endif
651 
652 	inode = erofs_iget(sb, sbi->root_nid);
653 	if (IS_ERR(inode))
654 		return PTR_ERR(inode);
655 
656 	if (!S_ISDIR(inode->i_mode)) {
657 		erofs_err(sb, "rootino(nid %llu) is not a directory(i_mode %o)",
658 			  sbi->root_nid, inode->i_mode);
659 		iput(inode);
660 		return -EINVAL;
661 	}
662 
663 	sb->s_root = d_make_root(inode);
664 	if (!sb->s_root)
665 		return -ENOMEM;
666 
667 	erofs_shrinker_register(sb);
668 	if (erofs_sb_has_fragments(sbi) && sbi->packed_nid) {
669 		sbi->packed_inode = erofs_iget(sb, sbi->packed_nid);
670 		if (IS_ERR(sbi->packed_inode)) {
671 			err = PTR_ERR(sbi->packed_inode);
672 			sbi->packed_inode = NULL;
673 			return err;
674 		}
675 	}
676 	err = erofs_init_managed_cache(sb);
677 	if (err)
678 		return err;
679 
680 	err = erofs_xattr_prefixes_init(sb);
681 	if (err)
682 		return err;
683 
684 	erofs_set_sysfs_name(sb);
685 	err = erofs_register_sysfs(sb);
686 	if (err)
687 		return err;
688 
689 	erofs_info(sb, "mounted with root inode @ nid %llu.", sbi->root_nid);
690 	return 0;
691 }
692 
693 static int erofs_fc_get_tree(struct fs_context *fc)
694 {
695 	struct erofs_sb_info *sbi = fc->s_fs_info;
696 	int ret;
697 
698 	if (IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid)
699 		return get_tree_nodev(fc, erofs_fc_fill_super);
700 
701 	ret = get_tree_bdev_flags(fc, erofs_fc_fill_super,
702 		IS_ENABLED(CONFIG_EROFS_FS_BACKED_BY_FILE) ?
703 			GET_TREE_BDEV_QUIET_LOOKUP : 0);
704 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
705 	if (ret == -ENOTBLK) {
706 		if (!fc->source)
707 			return invalf(fc, "No source specified");
708 		sbi->fdev = filp_open(fc->source, O_RDONLY | O_LARGEFILE, 0);
709 		if (IS_ERR(sbi->fdev))
710 			return PTR_ERR(sbi->fdev);
711 
712 		if (S_ISREG(file_inode(sbi->fdev)->i_mode) &&
713 		    sbi->fdev->f_mapping->a_ops->read_folio)
714 			return get_tree_nodev(fc, erofs_fc_fill_super);
715 		fput(sbi->fdev);
716 	}
717 #endif
718 	return ret;
719 }
720 
721 static int erofs_fc_reconfigure(struct fs_context *fc)
722 {
723 	struct super_block *sb = fc->root->d_sb;
724 	struct erofs_sb_info *sbi = EROFS_SB(sb);
725 	struct erofs_sb_info *new_sbi = fc->s_fs_info;
726 
727 	DBG_BUGON(!sb_rdonly(sb));
728 
729 	if (new_sbi->fsid || new_sbi->domain_id)
730 		erofs_info(sb, "ignoring reconfiguration for fsid|domain_id.");
731 
732 	if (test_opt(&new_sbi->opt, POSIX_ACL))
733 		fc->sb_flags |= SB_POSIXACL;
734 	else
735 		fc->sb_flags &= ~SB_POSIXACL;
736 
737 	sbi->opt = new_sbi->opt;
738 
739 	fc->sb_flags |= SB_RDONLY;
740 	return 0;
741 }
742 
743 static int erofs_release_device_info(int id, void *ptr, void *data)
744 {
745 	struct erofs_device_info *dif = ptr;
746 
747 	fs_put_dax(dif->dax_dev, NULL);
748 	if (dif->file)
749 		fput(dif->file);
750 	erofs_fscache_unregister_cookie(dif->fscache);
751 	dif->fscache = NULL;
752 	kfree(dif->path);
753 	kfree(dif);
754 	return 0;
755 }
756 
757 static void erofs_free_dev_context(struct erofs_dev_context *devs)
758 {
759 	if (!devs)
760 		return;
761 	idr_for_each(&devs->tree, &erofs_release_device_info, NULL);
762 	idr_destroy(&devs->tree);
763 	kfree(devs);
764 }
765 
766 static void erofs_fc_free(struct fs_context *fc)
767 {
768 	struct erofs_sb_info *sbi = fc->s_fs_info;
769 
770 	if (!sbi)
771 		return;
772 
773 	erofs_free_dev_context(sbi->devs);
774 	kfree(sbi->fsid);
775 	kfree(sbi->domain_id);
776 	kfree(sbi);
777 }
778 
779 static const struct fs_context_operations erofs_context_ops = {
780 	.parse_param	= erofs_fc_parse_param,
781 	.get_tree       = erofs_fc_get_tree,
782 	.reconfigure    = erofs_fc_reconfigure,
783 	.free		= erofs_fc_free,
784 };
785 
786 static int erofs_init_fs_context(struct fs_context *fc)
787 {
788 	struct erofs_sb_info *sbi;
789 
790 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
791 	if (!sbi)
792 		return -ENOMEM;
793 
794 	sbi->devs = kzalloc(sizeof(struct erofs_dev_context), GFP_KERNEL);
795 	if (!sbi->devs) {
796 		kfree(sbi);
797 		return -ENOMEM;
798 	}
799 	fc->s_fs_info = sbi;
800 
801 	idr_init(&sbi->devs->tree);
802 	init_rwsem(&sbi->devs->rwsem);
803 	erofs_default_options(sbi);
804 	fc->ops = &erofs_context_ops;
805 	return 0;
806 }
807 
808 static void erofs_kill_sb(struct super_block *sb)
809 {
810 	struct erofs_sb_info *sbi = EROFS_SB(sb);
811 
812 	if ((IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid) || sbi->fdev)
813 		kill_anon_super(sb);
814 	else
815 		kill_block_super(sb);
816 
817 	erofs_free_dev_context(sbi->devs);
818 	fs_put_dax(sbi->dax_dev, NULL);
819 	erofs_fscache_unregister_fs(sb);
820 	kfree(sbi->fsid);
821 	kfree(sbi->domain_id);
822 	if (sbi->fdev)
823 		fput(sbi->fdev);
824 	kfree(sbi);
825 	sb->s_fs_info = NULL;
826 }
827 
828 static void erofs_put_super(struct super_block *sb)
829 {
830 	struct erofs_sb_info *const sbi = EROFS_SB(sb);
831 
832 	DBG_BUGON(!sbi);
833 
834 	erofs_unregister_sysfs(sb);
835 	erofs_shrinker_unregister(sb);
836 	erofs_xattr_prefixes_cleanup(sb);
837 #ifdef CONFIG_EROFS_FS_ZIP
838 	iput(sbi->managed_cache);
839 	sbi->managed_cache = NULL;
840 #endif
841 	iput(sbi->packed_inode);
842 	sbi->packed_inode = NULL;
843 	erofs_free_dev_context(sbi->devs);
844 	sbi->devs = NULL;
845 	erofs_fscache_unregister_fs(sb);
846 }
847 
848 static struct file_system_type erofs_fs_type = {
849 	.owner          = THIS_MODULE,
850 	.name           = "erofs",
851 	.init_fs_context = erofs_init_fs_context,
852 	.kill_sb        = erofs_kill_sb,
853 	.fs_flags       = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
854 };
855 MODULE_ALIAS_FS("erofs");
856 
857 static int __init erofs_module_init(void)
858 {
859 	int err;
860 
861 	erofs_check_ondisk_layout_definitions();
862 
863 	erofs_inode_cachep = kmem_cache_create("erofs_inode",
864 			sizeof(struct erofs_inode), 0,
865 			SLAB_RECLAIM_ACCOUNT | SLAB_ACCOUNT,
866 			erofs_inode_init_once);
867 	if (!erofs_inode_cachep)
868 		return -ENOMEM;
869 
870 	err = erofs_init_shrinker();
871 	if (err)
872 		goto shrinker_err;
873 
874 	err = z_erofs_init_subsystem();
875 	if (err)
876 		goto zip_err;
877 
878 	err = erofs_init_sysfs();
879 	if (err)
880 		goto sysfs_err;
881 
882 	err = register_filesystem(&erofs_fs_type);
883 	if (err)
884 		goto fs_err;
885 
886 	return 0;
887 
888 fs_err:
889 	erofs_exit_sysfs();
890 sysfs_err:
891 	z_erofs_exit_subsystem();
892 zip_err:
893 	erofs_exit_shrinker();
894 shrinker_err:
895 	kmem_cache_destroy(erofs_inode_cachep);
896 	return err;
897 }
898 
899 static void __exit erofs_module_exit(void)
900 {
901 	unregister_filesystem(&erofs_fs_type);
902 
903 	/* Ensure all RCU free inodes / pclusters are safe to be destroyed. */
904 	rcu_barrier();
905 
906 	erofs_exit_sysfs();
907 	z_erofs_exit_subsystem();
908 	erofs_exit_shrinker();
909 	kmem_cache_destroy(erofs_inode_cachep);
910 }
911 
912 static int erofs_statfs(struct dentry *dentry, struct kstatfs *buf)
913 {
914 	struct super_block *sb = dentry->d_sb;
915 	struct erofs_sb_info *sbi = EROFS_SB(sb);
916 
917 	buf->f_type = sb->s_magic;
918 	buf->f_bsize = sb->s_blocksize;
919 	buf->f_blocks = sbi->total_blocks;
920 	buf->f_bfree = buf->f_bavail = 0;
921 	buf->f_files = ULLONG_MAX;
922 	buf->f_ffree = ULLONG_MAX - sbi->inos;
923 	buf->f_namelen = EROFS_NAME_LEN;
924 
925 	if (uuid_is_null(&sb->s_uuid))
926 		buf->f_fsid = u64_to_fsid(!sb->s_bdev ? 0 :
927 				huge_encode_dev(sb->s_bdev->bd_dev));
928 	else
929 		buf->f_fsid = uuid_to_fsid(sb->s_uuid.b);
930 	return 0;
931 }
932 
933 static int erofs_show_options(struct seq_file *seq, struct dentry *root)
934 {
935 	struct erofs_sb_info *sbi = EROFS_SB(root->d_sb);
936 	struct erofs_mount_opts *opt = &sbi->opt;
937 
938 	if (IS_ENABLED(CONFIG_EROFS_FS_XATTR))
939 		seq_puts(seq, test_opt(opt, XATTR_USER) ?
940 				",user_xattr" : ",nouser_xattr");
941 	if (IS_ENABLED(CONFIG_EROFS_FS_POSIX_ACL))
942 		seq_puts(seq, test_opt(opt, POSIX_ACL) ? ",acl" : ",noacl");
943 	if (IS_ENABLED(CONFIG_EROFS_FS_ZIP))
944 		seq_printf(seq, ",cache_strategy=%s",
945 			  erofs_param_cache_strategy[opt->cache_strategy].name);
946 	if (test_opt(opt, DAX_ALWAYS))
947 		seq_puts(seq, ",dax=always");
948 	if (test_opt(opt, DAX_NEVER))
949 		seq_puts(seq, ",dax=never");
950 #ifdef CONFIG_EROFS_FS_ONDEMAND
951 	if (sbi->fsid)
952 		seq_printf(seq, ",fsid=%s", sbi->fsid);
953 	if (sbi->domain_id)
954 		seq_printf(seq, ",domain_id=%s", sbi->domain_id);
955 #endif
956 	return 0;
957 }
958 
959 const struct super_operations erofs_sops = {
960 	.put_super = erofs_put_super,
961 	.alloc_inode = erofs_alloc_inode,
962 	.free_inode = erofs_free_inode,
963 	.statfs = erofs_statfs,
964 	.show_options = erofs_show_options,
965 };
966 
967 module_init(erofs_module_init);
968 module_exit(erofs_module_exit);
969 
970 MODULE_DESCRIPTION("Enhanced ROM File System");
971 MODULE_AUTHOR("Gao Xiang, Chao Yu, Miao Xie, CONSUMER BG, HUAWEI Inc.");
972 MODULE_LICENSE("GPL");
973