xref: /linux/fs/erofs/super.c (revision 1553a1c48281243359a9529a10ddb551f3b967ab)
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 "xattr.h"
14 
15 #define CREATE_TRACE_POINTS
16 #include <trace/events/erofs.h>
17 
18 static struct kmem_cache *erofs_inode_cachep __read_mostly;
19 
20 void _erofs_err(struct super_block *sb, const char *func, const char *fmt, ...)
21 {
22 	struct va_format vaf;
23 	va_list args;
24 
25 	va_start(args, fmt);
26 
27 	vaf.fmt = fmt;
28 	vaf.va = &args;
29 
30 	if (sb)
31 		pr_err("(device %s): %s: %pV", sb->s_id, func, &vaf);
32 	else
33 		pr_err("%s: %pV", func, &vaf);
34 	va_end(args);
35 }
36 
37 void _erofs_info(struct super_block *sb, const char *func, const char *fmt, ...)
38 {
39 	struct va_format vaf;
40 	va_list args;
41 
42 	va_start(args, fmt);
43 
44 	vaf.fmt = fmt;
45 	vaf.va = &args;
46 
47 	if (sb)
48 		pr_info("(device %s): %pV", sb->s_id, &vaf);
49 	else
50 		pr_info("%pV", &vaf);
51 	va_end(args);
52 }
53 
54 static int erofs_superblock_csum_verify(struct super_block *sb, void *sbdata)
55 {
56 	size_t len = 1 << EROFS_SB(sb)->blkszbits;
57 	struct erofs_super_block *dsb;
58 	u32 expected_crc, crc;
59 
60 	if (len > EROFS_SUPER_OFFSET)
61 		len -= EROFS_SUPER_OFFSET;
62 
63 	dsb = kmemdup(sbdata + EROFS_SUPER_OFFSET, len, GFP_KERNEL);
64 	if (!dsb)
65 		return -ENOMEM;
66 
67 	expected_crc = le32_to_cpu(dsb->checksum);
68 	dsb->checksum = 0;
69 	/* to allow for x86 boot sectors and other oddities. */
70 	crc = crc32c(~0, dsb, len);
71 	kfree(dsb);
72 
73 	if (crc != expected_crc) {
74 		erofs_err(sb, "invalid checksum 0x%08x, 0x%08x expected",
75 			  crc, expected_crc);
76 		return -EBADMSG;
77 	}
78 	return 0;
79 }
80 
81 static void erofs_inode_init_once(void *ptr)
82 {
83 	struct erofs_inode *vi = ptr;
84 
85 	inode_init_once(&vi->vfs_inode);
86 }
87 
88 static struct inode *erofs_alloc_inode(struct super_block *sb)
89 {
90 	struct erofs_inode *vi =
91 		alloc_inode_sb(sb, erofs_inode_cachep, GFP_KERNEL);
92 
93 	if (!vi)
94 		return NULL;
95 
96 	/* zero out everything except vfs_inode */
97 	memset(vi, 0, offsetof(struct erofs_inode, vfs_inode));
98 	return &vi->vfs_inode;
99 }
100 
101 static void erofs_free_inode(struct inode *inode)
102 {
103 	struct erofs_inode *vi = EROFS_I(inode);
104 
105 	if (inode->i_op == &erofs_fast_symlink_iops)
106 		kfree(inode->i_link);
107 	kfree(vi->xattr_shared_xattrs);
108 	kmem_cache_free(erofs_inode_cachep, vi);
109 }
110 
111 static bool check_layout_compatibility(struct super_block *sb,
112 				       struct erofs_super_block *dsb)
113 {
114 	const unsigned int feature = le32_to_cpu(dsb->feature_incompat);
115 
116 	EROFS_SB(sb)->feature_incompat = feature;
117 
118 	/* check if current kernel meets all mandatory requirements */
119 	if (feature & (~EROFS_ALL_FEATURE_INCOMPAT)) {
120 		erofs_err(sb, "unidentified incompatible feature %x, please upgrade kernel",
121 			   feature & ~EROFS_ALL_FEATURE_INCOMPAT);
122 		return false;
123 	}
124 	return true;
125 }
126 
127 /* read variable-sized metadata, offset will be aligned by 4-byte */
128 void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf,
129 			  erofs_off_t *offset, int *lengthp)
130 {
131 	u8 *buffer, *ptr;
132 	int len, i, cnt;
133 
134 	*offset = round_up(*offset, 4);
135 	ptr = erofs_bread(buf, erofs_blknr(sb, *offset), EROFS_KMAP);
136 	if (IS_ERR(ptr))
137 		return ptr;
138 
139 	len = le16_to_cpu(*(__le16 *)&ptr[erofs_blkoff(sb, *offset)]);
140 	if (!len)
141 		len = U16_MAX + 1;
142 	buffer = kmalloc(len, GFP_KERNEL);
143 	if (!buffer)
144 		return ERR_PTR(-ENOMEM);
145 	*offset += sizeof(__le16);
146 	*lengthp = len;
147 
148 	for (i = 0; i < len; i += cnt) {
149 		cnt = min_t(int, sb->s_blocksize - erofs_blkoff(sb, *offset),
150 			    len - i);
151 		ptr = erofs_bread(buf, erofs_blknr(sb, *offset), EROFS_KMAP);
152 		if (IS_ERR(ptr)) {
153 			kfree(buffer);
154 			return ptr;
155 		}
156 		memcpy(buffer + i, ptr + erofs_blkoff(sb, *offset), cnt);
157 		*offset += cnt;
158 	}
159 	return buffer;
160 }
161 
162 #ifndef CONFIG_EROFS_FS_ZIP
163 static int z_erofs_parse_cfgs(struct super_block *sb,
164 			      struct erofs_super_block *dsb)
165 {
166 	if (!dsb->u1.available_compr_algs)
167 		return 0;
168 
169 	erofs_err(sb, "compression disabled, unable to mount compressed EROFS");
170 	return -EOPNOTSUPP;
171 }
172 #endif
173 
174 static int erofs_init_device(struct erofs_buf *buf, struct super_block *sb,
175 			     struct erofs_device_info *dif, erofs_off_t *pos)
176 {
177 	struct erofs_sb_info *sbi = EROFS_SB(sb);
178 	struct erofs_fscache *fscache;
179 	struct erofs_deviceslot *dis;
180 	struct file *bdev_file;
181 	void *ptr;
182 
183 	ptr = erofs_read_metabuf(buf, sb, erofs_blknr(sb, *pos), EROFS_KMAP);
184 	if (IS_ERR(ptr))
185 		return PTR_ERR(ptr);
186 	dis = ptr + erofs_blkoff(sb, *pos);
187 
188 	if (!sbi->devs->flatdev && !dif->path) {
189 		if (!dis->tag[0]) {
190 			erofs_err(sb, "empty device tag @ pos %llu", *pos);
191 			return -EINVAL;
192 		}
193 		dif->path = kmemdup_nul(dis->tag, sizeof(dis->tag), GFP_KERNEL);
194 		if (!dif->path)
195 			return -ENOMEM;
196 	}
197 
198 	if (erofs_is_fscache_mode(sb)) {
199 		fscache = erofs_fscache_register_cookie(sb, dif->path, 0);
200 		if (IS_ERR(fscache))
201 			return PTR_ERR(fscache);
202 		dif->fscache = fscache;
203 	} else if (!sbi->devs->flatdev) {
204 		bdev_file = bdev_file_open_by_path(dif->path, BLK_OPEN_READ,
205 						sb->s_type, NULL);
206 		if (IS_ERR(bdev_file))
207 			return PTR_ERR(bdev_file);
208 		dif->bdev_file = bdev_file;
209 		dif->dax_dev = fs_dax_get_by_bdev(file_bdev(bdev_file),
210 				&dif->dax_part_off, NULL, NULL);
211 	}
212 
213 	dif->blocks = le32_to_cpu(dis->blocks);
214 	dif->mapped_blkaddr = le32_to_cpu(dis->mapped_blkaddr);
215 	sbi->total_blocks += dif->blocks;
216 	*pos += EROFS_DEVT_SLOT_SIZE;
217 	return 0;
218 }
219 
220 static int erofs_scan_devices(struct super_block *sb,
221 			      struct erofs_super_block *dsb)
222 {
223 	struct erofs_sb_info *sbi = EROFS_SB(sb);
224 	unsigned int ondisk_extradevs;
225 	erofs_off_t pos;
226 	struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
227 	struct erofs_device_info *dif;
228 	int id, err = 0;
229 
230 	sbi->total_blocks = sbi->primarydevice_blocks;
231 	if (!erofs_sb_has_device_table(sbi))
232 		ondisk_extradevs = 0;
233 	else
234 		ondisk_extradevs = le16_to_cpu(dsb->extra_devices);
235 
236 	if (sbi->devs->extra_devices &&
237 	    ondisk_extradevs != sbi->devs->extra_devices) {
238 		erofs_err(sb, "extra devices don't match (ondisk %u, given %u)",
239 			  ondisk_extradevs, sbi->devs->extra_devices);
240 		return -EINVAL;
241 	}
242 	if (!ondisk_extradevs)
243 		return 0;
244 
245 	if (!sbi->devs->extra_devices && !erofs_is_fscache_mode(sb))
246 		sbi->devs->flatdev = true;
247 
248 	sbi->device_id_mask = roundup_pow_of_two(ondisk_extradevs + 1) - 1;
249 	pos = le16_to_cpu(dsb->devt_slotoff) * EROFS_DEVT_SLOT_SIZE;
250 	down_read(&sbi->devs->rwsem);
251 	if (sbi->devs->extra_devices) {
252 		idr_for_each_entry(&sbi->devs->tree, dif, id) {
253 			err = erofs_init_device(&buf, sb, dif, &pos);
254 			if (err)
255 				break;
256 		}
257 	} else {
258 		for (id = 0; id < ondisk_extradevs; id++) {
259 			dif = kzalloc(sizeof(*dif), GFP_KERNEL);
260 			if (!dif) {
261 				err = -ENOMEM;
262 				break;
263 			}
264 
265 			err = idr_alloc(&sbi->devs->tree, dif, 0, 0, GFP_KERNEL);
266 			if (err < 0) {
267 				kfree(dif);
268 				break;
269 			}
270 			++sbi->devs->extra_devices;
271 
272 			err = erofs_init_device(&buf, sb, dif, &pos);
273 			if (err)
274 				break;
275 		}
276 	}
277 	up_read(&sbi->devs->rwsem);
278 	erofs_put_metabuf(&buf);
279 	return err;
280 }
281 
282 static int erofs_read_superblock(struct super_block *sb)
283 {
284 	struct erofs_sb_info *sbi;
285 	struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
286 	struct erofs_super_block *dsb;
287 	void *data;
288 	int ret;
289 
290 	data = erofs_read_metabuf(&buf, sb, 0, EROFS_KMAP);
291 	if (IS_ERR(data)) {
292 		erofs_err(sb, "cannot read erofs superblock");
293 		return PTR_ERR(data);
294 	}
295 
296 	sbi = EROFS_SB(sb);
297 	dsb = (struct erofs_super_block *)(data + EROFS_SUPER_OFFSET);
298 
299 	ret = -EINVAL;
300 	if (le32_to_cpu(dsb->magic) != EROFS_SUPER_MAGIC_V1) {
301 		erofs_err(sb, "cannot find valid erofs superblock");
302 		goto out;
303 	}
304 
305 	sbi->blkszbits  = dsb->blkszbits;
306 	if (sbi->blkszbits < 9 || sbi->blkszbits > PAGE_SHIFT) {
307 		erofs_err(sb, "blkszbits %u isn't supported", sbi->blkszbits);
308 		goto out;
309 	}
310 	if (dsb->dirblkbits) {
311 		erofs_err(sb, "dirblkbits %u isn't supported", dsb->dirblkbits);
312 		goto out;
313 	}
314 
315 	sbi->feature_compat = le32_to_cpu(dsb->feature_compat);
316 	if (erofs_sb_has_sb_chksum(sbi)) {
317 		ret = erofs_superblock_csum_verify(sb, data);
318 		if (ret)
319 			goto out;
320 	}
321 
322 	ret = -EINVAL;
323 	if (!check_layout_compatibility(sb, dsb))
324 		goto out;
325 
326 	sbi->sb_size = 128 + dsb->sb_extslots * EROFS_SB_EXTSLOT_SIZE;
327 	if (sbi->sb_size > PAGE_SIZE - EROFS_SUPER_OFFSET) {
328 		erofs_err(sb, "invalid sb_extslots %u (more than a fs block)",
329 			  sbi->sb_size);
330 		goto out;
331 	}
332 	sbi->primarydevice_blocks = le32_to_cpu(dsb->blocks);
333 	sbi->meta_blkaddr = le32_to_cpu(dsb->meta_blkaddr);
334 #ifdef CONFIG_EROFS_FS_XATTR
335 	sbi->xattr_blkaddr = le32_to_cpu(dsb->xattr_blkaddr);
336 	sbi->xattr_prefix_start = le32_to_cpu(dsb->xattr_prefix_start);
337 	sbi->xattr_prefix_count = dsb->xattr_prefix_count;
338 	sbi->xattr_filter_reserved = dsb->xattr_filter_reserved;
339 #endif
340 	sbi->islotbits = ilog2(sizeof(struct erofs_inode_compact));
341 	sbi->root_nid = le16_to_cpu(dsb->root_nid);
342 	sbi->packed_nid = le64_to_cpu(dsb->packed_nid);
343 	sbi->inos = le64_to_cpu(dsb->inos);
344 
345 	sbi->build_time = le64_to_cpu(dsb->build_time);
346 	sbi->build_time_nsec = le32_to_cpu(dsb->build_time_nsec);
347 
348 	memcpy(&sb->s_uuid, dsb->uuid, sizeof(dsb->uuid));
349 
350 	ret = strscpy(sbi->volume_name, dsb->volume_name,
351 		      sizeof(dsb->volume_name));
352 	if (ret < 0) {	/* -E2BIG */
353 		erofs_err(sb, "bad volume name without NIL terminator");
354 		ret = -EFSCORRUPTED;
355 		goto out;
356 	}
357 
358 	/* parse on-disk compression configurations */
359 	ret = z_erofs_parse_cfgs(sb, dsb);
360 	if (ret < 0)
361 		goto out;
362 
363 	/* handle multiple devices */
364 	ret = erofs_scan_devices(sb, dsb);
365 
366 	if (erofs_is_fscache_mode(sb))
367 		erofs_info(sb, "EXPERIMENTAL fscache-based on-demand read feature in use. Use at your own risk!");
368 out:
369 	erofs_put_metabuf(&buf);
370 	return ret;
371 }
372 
373 static void erofs_default_options(struct erofs_fs_context *ctx)
374 {
375 #ifdef CONFIG_EROFS_FS_ZIP
376 	ctx->opt.cache_strategy = EROFS_ZIP_CACHE_READAROUND;
377 	ctx->opt.max_sync_decompress_pages = 3;
378 	ctx->opt.sync_decompress = EROFS_SYNC_DECOMPRESS_AUTO;
379 #endif
380 #ifdef CONFIG_EROFS_FS_XATTR
381 	set_opt(&ctx->opt, XATTR_USER);
382 #endif
383 #ifdef CONFIG_EROFS_FS_POSIX_ACL
384 	set_opt(&ctx->opt, POSIX_ACL);
385 #endif
386 }
387 
388 enum {
389 	Opt_user_xattr,
390 	Opt_acl,
391 	Opt_cache_strategy,
392 	Opt_dax,
393 	Opt_dax_enum,
394 	Opt_device,
395 	Opt_fsid,
396 	Opt_domain_id,
397 	Opt_err
398 };
399 
400 static const struct constant_table erofs_param_cache_strategy[] = {
401 	{"disabled",	EROFS_ZIP_CACHE_DISABLED},
402 	{"readahead",	EROFS_ZIP_CACHE_READAHEAD},
403 	{"readaround",	EROFS_ZIP_CACHE_READAROUND},
404 	{}
405 };
406 
407 static const struct constant_table erofs_dax_param_enums[] = {
408 	{"always",	EROFS_MOUNT_DAX_ALWAYS},
409 	{"never",	EROFS_MOUNT_DAX_NEVER},
410 	{}
411 };
412 
413 static const struct fs_parameter_spec erofs_fs_parameters[] = {
414 	fsparam_flag_no("user_xattr",	Opt_user_xattr),
415 	fsparam_flag_no("acl",		Opt_acl),
416 	fsparam_enum("cache_strategy",	Opt_cache_strategy,
417 		     erofs_param_cache_strategy),
418 	fsparam_flag("dax",             Opt_dax),
419 	fsparam_enum("dax",		Opt_dax_enum, erofs_dax_param_enums),
420 	fsparam_string("device",	Opt_device),
421 	fsparam_string("fsid",		Opt_fsid),
422 	fsparam_string("domain_id",	Opt_domain_id),
423 	{}
424 };
425 
426 static bool erofs_fc_set_dax_mode(struct fs_context *fc, unsigned int mode)
427 {
428 #ifdef CONFIG_FS_DAX
429 	struct erofs_fs_context *ctx = fc->fs_private;
430 
431 	switch (mode) {
432 	case EROFS_MOUNT_DAX_ALWAYS:
433 		warnfc(fc, "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
434 		set_opt(&ctx->opt, DAX_ALWAYS);
435 		clear_opt(&ctx->opt, DAX_NEVER);
436 		return true;
437 	case EROFS_MOUNT_DAX_NEVER:
438 		set_opt(&ctx->opt, DAX_NEVER);
439 		clear_opt(&ctx->opt, DAX_ALWAYS);
440 		return true;
441 	default:
442 		DBG_BUGON(1);
443 		return false;
444 	}
445 #else
446 	errorfc(fc, "dax options not supported");
447 	return false;
448 #endif
449 }
450 
451 static int erofs_fc_parse_param(struct fs_context *fc,
452 				struct fs_parameter *param)
453 {
454 	struct erofs_fs_context *ctx = fc->fs_private;
455 	struct fs_parse_result result;
456 	struct erofs_device_info *dif;
457 	int opt, ret;
458 
459 	opt = fs_parse(fc, erofs_fs_parameters, param, &result);
460 	if (opt < 0)
461 		return opt;
462 
463 	switch (opt) {
464 	case Opt_user_xattr:
465 #ifdef CONFIG_EROFS_FS_XATTR
466 		if (result.boolean)
467 			set_opt(&ctx->opt, XATTR_USER);
468 		else
469 			clear_opt(&ctx->opt, XATTR_USER);
470 #else
471 		errorfc(fc, "{,no}user_xattr options not supported");
472 #endif
473 		break;
474 	case Opt_acl:
475 #ifdef CONFIG_EROFS_FS_POSIX_ACL
476 		if (result.boolean)
477 			set_opt(&ctx->opt, POSIX_ACL);
478 		else
479 			clear_opt(&ctx->opt, POSIX_ACL);
480 #else
481 		errorfc(fc, "{,no}acl options not supported");
482 #endif
483 		break;
484 	case Opt_cache_strategy:
485 #ifdef CONFIG_EROFS_FS_ZIP
486 		ctx->opt.cache_strategy = result.uint_32;
487 #else
488 		errorfc(fc, "compression not supported, cache_strategy ignored");
489 #endif
490 		break;
491 	case Opt_dax:
492 		if (!erofs_fc_set_dax_mode(fc, EROFS_MOUNT_DAX_ALWAYS))
493 			return -EINVAL;
494 		break;
495 	case Opt_dax_enum:
496 		if (!erofs_fc_set_dax_mode(fc, result.uint_32))
497 			return -EINVAL;
498 		break;
499 	case Opt_device:
500 		dif = kzalloc(sizeof(*dif), GFP_KERNEL);
501 		if (!dif)
502 			return -ENOMEM;
503 		dif->path = kstrdup(param->string, GFP_KERNEL);
504 		if (!dif->path) {
505 			kfree(dif);
506 			return -ENOMEM;
507 		}
508 		down_write(&ctx->devs->rwsem);
509 		ret = idr_alloc(&ctx->devs->tree, dif, 0, 0, GFP_KERNEL);
510 		up_write(&ctx->devs->rwsem);
511 		if (ret < 0) {
512 			kfree(dif->path);
513 			kfree(dif);
514 			return ret;
515 		}
516 		++ctx->devs->extra_devices;
517 		break;
518 #ifdef CONFIG_EROFS_FS_ONDEMAND
519 	case Opt_fsid:
520 		kfree(ctx->fsid);
521 		ctx->fsid = kstrdup(param->string, GFP_KERNEL);
522 		if (!ctx->fsid)
523 			return -ENOMEM;
524 		break;
525 	case Opt_domain_id:
526 		kfree(ctx->domain_id);
527 		ctx->domain_id = kstrdup(param->string, GFP_KERNEL);
528 		if (!ctx->domain_id)
529 			return -ENOMEM;
530 		break;
531 #else
532 	case Opt_fsid:
533 	case Opt_domain_id:
534 		errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
535 		break;
536 #endif
537 	default:
538 		return -ENOPARAM;
539 	}
540 	return 0;
541 }
542 
543 static struct inode *erofs_nfs_get_inode(struct super_block *sb,
544 					 u64 ino, u32 generation)
545 {
546 	return erofs_iget(sb, ino);
547 }
548 
549 static struct dentry *erofs_fh_to_dentry(struct super_block *sb,
550 		struct fid *fid, int fh_len, int fh_type)
551 {
552 	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
553 				    erofs_nfs_get_inode);
554 }
555 
556 static struct dentry *erofs_fh_to_parent(struct super_block *sb,
557 		struct fid *fid, int fh_len, int fh_type)
558 {
559 	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
560 				    erofs_nfs_get_inode);
561 }
562 
563 static struct dentry *erofs_get_parent(struct dentry *child)
564 {
565 	erofs_nid_t nid;
566 	unsigned int d_type;
567 	int err;
568 
569 	err = erofs_namei(d_inode(child), &dotdot_name, &nid, &d_type);
570 	if (err)
571 		return ERR_PTR(err);
572 	return d_obtain_alias(erofs_iget(child->d_sb, nid));
573 }
574 
575 static const struct export_operations erofs_export_ops = {
576 	.encode_fh = generic_encode_ino32_fh,
577 	.fh_to_dentry = erofs_fh_to_dentry,
578 	.fh_to_parent = erofs_fh_to_parent,
579 	.get_parent = erofs_get_parent,
580 };
581 
582 static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
583 {
584 	struct inode *inode;
585 	struct erofs_sb_info *sbi;
586 	struct erofs_fs_context *ctx = fc->fs_private;
587 	int err;
588 
589 	sb->s_magic = EROFS_SUPER_MAGIC;
590 	sb->s_flags |= SB_RDONLY | SB_NOATIME;
591 	sb->s_maxbytes = MAX_LFS_FILESIZE;
592 	sb->s_op = &erofs_sops;
593 
594 	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
595 	if (!sbi)
596 		return -ENOMEM;
597 
598 	sb->s_fs_info = sbi;
599 	sbi->opt = ctx->opt;
600 	sbi->devs = ctx->devs;
601 	ctx->devs = NULL;
602 	sbi->fsid = ctx->fsid;
603 	ctx->fsid = NULL;
604 	sbi->domain_id = ctx->domain_id;
605 	ctx->domain_id = NULL;
606 
607 	sbi->blkszbits = PAGE_SHIFT;
608 	if (erofs_is_fscache_mode(sb)) {
609 		sb->s_blocksize = PAGE_SIZE;
610 		sb->s_blocksize_bits = PAGE_SHIFT;
611 
612 		err = erofs_fscache_register_fs(sb);
613 		if (err)
614 			return err;
615 
616 		err = super_setup_bdi(sb);
617 		if (err)
618 			return err;
619 	} else {
620 		if (!sb_set_blocksize(sb, PAGE_SIZE)) {
621 			errorfc(fc, "failed to set initial blksize");
622 			return -EINVAL;
623 		}
624 
625 		sbi->dax_dev = fs_dax_get_by_bdev(sb->s_bdev,
626 						  &sbi->dax_part_off,
627 						  NULL, NULL);
628 	}
629 
630 	err = erofs_read_superblock(sb);
631 	if (err)
632 		return err;
633 
634 	if (sb->s_blocksize_bits != sbi->blkszbits) {
635 		if (erofs_is_fscache_mode(sb)) {
636 			errorfc(fc, "unsupported blksize for fscache mode");
637 			return -EINVAL;
638 		}
639 		if (!sb_set_blocksize(sb, 1 << sbi->blkszbits)) {
640 			errorfc(fc, "failed to set erofs blksize");
641 			return -EINVAL;
642 		}
643 	}
644 
645 	if (test_opt(&sbi->opt, DAX_ALWAYS)) {
646 		if (!sbi->dax_dev) {
647 			errorfc(fc, "DAX unsupported by block device. Turning off DAX.");
648 			clear_opt(&sbi->opt, DAX_ALWAYS);
649 		} else if (sbi->blkszbits != PAGE_SHIFT) {
650 			errorfc(fc, "unsupported blocksize for DAX");
651 			clear_opt(&sbi->opt, DAX_ALWAYS);
652 		}
653 	}
654 
655 	sb->s_time_gran = 1;
656 	sb->s_xattr = erofs_xattr_handlers;
657 	sb->s_export_op = &erofs_export_ops;
658 
659 	if (test_opt(&sbi->opt, POSIX_ACL))
660 		sb->s_flags |= SB_POSIXACL;
661 	else
662 		sb->s_flags &= ~SB_POSIXACL;
663 
664 #ifdef CONFIG_EROFS_FS_ZIP
665 	xa_init(&sbi->managed_pslots);
666 #endif
667 
668 	inode = erofs_iget(sb, sbi->root_nid);
669 	if (IS_ERR(inode))
670 		return PTR_ERR(inode);
671 
672 	if (!S_ISDIR(inode->i_mode)) {
673 		erofs_err(sb, "rootino(nid %llu) is not a directory(i_mode %o)",
674 			  sbi->root_nid, inode->i_mode);
675 		iput(inode);
676 		return -EINVAL;
677 	}
678 
679 	sb->s_root = d_make_root(inode);
680 	if (!sb->s_root)
681 		return -ENOMEM;
682 
683 	erofs_shrinker_register(sb);
684 	if (erofs_sb_has_fragments(sbi) && sbi->packed_nid) {
685 		sbi->packed_inode = erofs_iget(sb, sbi->packed_nid);
686 		if (IS_ERR(sbi->packed_inode)) {
687 			err = PTR_ERR(sbi->packed_inode);
688 			sbi->packed_inode = NULL;
689 			return err;
690 		}
691 	}
692 	err = erofs_init_managed_cache(sb);
693 	if (err)
694 		return err;
695 
696 	err = erofs_xattr_prefixes_init(sb);
697 	if (err)
698 		return err;
699 
700 	err = erofs_register_sysfs(sb);
701 	if (err)
702 		return err;
703 
704 	erofs_info(sb, "mounted with root inode @ nid %llu.", sbi->root_nid);
705 	return 0;
706 }
707 
708 static int erofs_fc_get_tree(struct fs_context *fc)
709 {
710 	struct erofs_fs_context *ctx = fc->fs_private;
711 
712 	if (IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && ctx->fsid)
713 		return get_tree_nodev(fc, erofs_fc_fill_super);
714 
715 	return get_tree_bdev(fc, erofs_fc_fill_super);
716 }
717 
718 static int erofs_fc_reconfigure(struct fs_context *fc)
719 {
720 	struct super_block *sb = fc->root->d_sb;
721 	struct erofs_sb_info *sbi = EROFS_SB(sb);
722 	struct erofs_fs_context *ctx = fc->fs_private;
723 
724 	DBG_BUGON(!sb_rdonly(sb));
725 
726 	if (ctx->fsid || ctx->domain_id)
727 		erofs_info(sb, "ignoring reconfiguration for fsid|domain_id.");
728 
729 	if (test_opt(&ctx->opt, POSIX_ACL))
730 		fc->sb_flags |= SB_POSIXACL;
731 	else
732 		fc->sb_flags &= ~SB_POSIXACL;
733 
734 	sbi->opt = ctx->opt;
735 
736 	fc->sb_flags |= SB_RDONLY;
737 	return 0;
738 }
739 
740 static int erofs_release_device_info(int id, void *ptr, void *data)
741 {
742 	struct erofs_device_info *dif = ptr;
743 
744 	fs_put_dax(dif->dax_dev, NULL);
745 	if (dif->bdev_file)
746 		fput(dif->bdev_file);
747 	erofs_fscache_unregister_cookie(dif->fscache);
748 	dif->fscache = NULL;
749 	kfree(dif->path);
750 	kfree(dif);
751 	return 0;
752 }
753 
754 static void erofs_free_dev_context(struct erofs_dev_context *devs)
755 {
756 	if (!devs)
757 		return;
758 	idr_for_each(&devs->tree, &erofs_release_device_info, NULL);
759 	idr_destroy(&devs->tree);
760 	kfree(devs);
761 }
762 
763 static void erofs_fc_free(struct fs_context *fc)
764 {
765 	struct erofs_fs_context *ctx = fc->fs_private;
766 
767 	erofs_free_dev_context(ctx->devs);
768 	kfree(ctx->fsid);
769 	kfree(ctx->domain_id);
770 	kfree(ctx);
771 }
772 
773 static const struct fs_context_operations erofs_context_ops = {
774 	.parse_param	= erofs_fc_parse_param,
775 	.get_tree       = erofs_fc_get_tree,
776 	.reconfigure    = erofs_fc_reconfigure,
777 	.free		= erofs_fc_free,
778 };
779 
780 static int erofs_init_fs_context(struct fs_context *fc)
781 {
782 	struct erofs_fs_context *ctx;
783 
784 	ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
785 	if (!ctx)
786 		return -ENOMEM;
787 	ctx->devs = kzalloc(sizeof(struct erofs_dev_context), GFP_KERNEL);
788 	if (!ctx->devs) {
789 		kfree(ctx);
790 		return -ENOMEM;
791 	}
792 	fc->fs_private = ctx;
793 
794 	idr_init(&ctx->devs->tree);
795 	init_rwsem(&ctx->devs->rwsem);
796 	erofs_default_options(ctx);
797 	fc->ops = &erofs_context_ops;
798 	return 0;
799 }
800 
801 static void erofs_kill_sb(struct super_block *sb)
802 {
803 	struct erofs_sb_info *sbi;
804 
805 	if (erofs_is_fscache_mode(sb))
806 		kill_anon_super(sb);
807 	else
808 		kill_block_super(sb);
809 
810 	sbi = EROFS_SB(sb);
811 	if (!sbi)
812 		return;
813 
814 	erofs_free_dev_context(sbi->devs);
815 	fs_put_dax(sbi->dax_dev, NULL);
816 	erofs_fscache_unregister_fs(sb);
817 	kfree(sbi->fsid);
818 	kfree(sbi->domain_id);
819 	kfree(sbi);
820 	sb->s_fs_info = NULL;
821 }
822 
823 static void erofs_put_super(struct super_block *sb)
824 {
825 	struct erofs_sb_info *const sbi = EROFS_SB(sb);
826 
827 	DBG_BUGON(!sbi);
828 
829 	erofs_unregister_sysfs(sb);
830 	erofs_shrinker_unregister(sb);
831 	erofs_xattr_prefixes_cleanup(sb);
832 #ifdef CONFIG_EROFS_FS_ZIP
833 	iput(sbi->managed_cache);
834 	sbi->managed_cache = NULL;
835 #endif
836 	iput(sbi->packed_inode);
837 	sbi->packed_inode = NULL;
838 	erofs_free_dev_context(sbi->devs);
839 	sbi->devs = NULL;
840 	erofs_fscache_unregister_fs(sb);
841 }
842 
843 static struct file_system_type erofs_fs_type = {
844 	.owner          = THIS_MODULE,
845 	.name           = "erofs",
846 	.init_fs_context = erofs_init_fs_context,
847 	.kill_sb        = erofs_kill_sb,
848 	.fs_flags       = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
849 };
850 MODULE_ALIAS_FS("erofs");
851 
852 static int __init erofs_module_init(void)
853 {
854 	int err;
855 
856 	erofs_check_ondisk_layout_definitions();
857 
858 	erofs_inode_cachep = kmem_cache_create("erofs_inode",
859 			sizeof(struct erofs_inode), 0,
860 			SLAB_RECLAIM_ACCOUNT | SLAB_ACCOUNT,
861 			erofs_inode_init_once);
862 	if (!erofs_inode_cachep)
863 		return -ENOMEM;
864 
865 	err = erofs_init_shrinker();
866 	if (err)
867 		goto shrinker_err;
868 
869 	err = z_erofs_lzma_init();
870 	if (err)
871 		goto lzma_err;
872 
873 	err = z_erofs_deflate_init();
874 	if (err)
875 		goto deflate_err;
876 
877 	erofs_pcpubuf_init();
878 	err = z_erofs_init_zip_subsystem();
879 	if (err)
880 		goto zip_err;
881 
882 	err = erofs_init_sysfs();
883 	if (err)
884 		goto sysfs_err;
885 
886 	err = register_filesystem(&erofs_fs_type);
887 	if (err)
888 		goto fs_err;
889 
890 	return 0;
891 
892 fs_err:
893 	erofs_exit_sysfs();
894 sysfs_err:
895 	z_erofs_exit_zip_subsystem();
896 zip_err:
897 	z_erofs_deflate_exit();
898 deflate_err:
899 	z_erofs_lzma_exit();
900 lzma_err:
901 	erofs_exit_shrinker();
902 shrinker_err:
903 	kmem_cache_destroy(erofs_inode_cachep);
904 	return err;
905 }
906 
907 static void __exit erofs_module_exit(void)
908 {
909 	unregister_filesystem(&erofs_fs_type);
910 
911 	/* Ensure all RCU free inodes / pclusters are safe to be destroyed. */
912 	rcu_barrier();
913 
914 	erofs_exit_sysfs();
915 	z_erofs_exit_zip_subsystem();
916 	z_erofs_deflate_exit();
917 	z_erofs_lzma_exit();
918 	erofs_exit_shrinker();
919 	kmem_cache_destroy(erofs_inode_cachep);
920 	erofs_pcpubuf_exit();
921 }
922 
923 static int erofs_statfs(struct dentry *dentry, struct kstatfs *buf)
924 {
925 	struct super_block *sb = dentry->d_sb;
926 	struct erofs_sb_info *sbi = EROFS_SB(sb);
927 	u64 id = 0;
928 
929 	if (!erofs_is_fscache_mode(sb))
930 		id = huge_encode_dev(sb->s_bdev->bd_dev);
931 
932 	buf->f_type = sb->s_magic;
933 	buf->f_bsize = sb->s_blocksize;
934 	buf->f_blocks = sbi->total_blocks;
935 	buf->f_bfree = buf->f_bavail = 0;
936 
937 	buf->f_files = ULLONG_MAX;
938 	buf->f_ffree = ULLONG_MAX - sbi->inos;
939 
940 	buf->f_namelen = EROFS_NAME_LEN;
941 
942 	buf->f_fsid    = u64_to_fsid(id);
943 	return 0;
944 }
945 
946 static int erofs_show_options(struct seq_file *seq, struct dentry *root)
947 {
948 	struct erofs_sb_info *sbi = EROFS_SB(root->d_sb);
949 	struct erofs_mount_opts *opt = &sbi->opt;
950 
951 #ifdef CONFIG_EROFS_FS_XATTR
952 	if (test_opt(opt, XATTR_USER))
953 		seq_puts(seq, ",user_xattr");
954 	else
955 		seq_puts(seq, ",nouser_xattr");
956 #endif
957 #ifdef CONFIG_EROFS_FS_POSIX_ACL
958 	if (test_opt(opt, POSIX_ACL))
959 		seq_puts(seq, ",acl");
960 	else
961 		seq_puts(seq, ",noacl");
962 #endif
963 #ifdef CONFIG_EROFS_FS_ZIP
964 	if (opt->cache_strategy == EROFS_ZIP_CACHE_DISABLED)
965 		seq_puts(seq, ",cache_strategy=disabled");
966 	else if (opt->cache_strategy == EROFS_ZIP_CACHE_READAHEAD)
967 		seq_puts(seq, ",cache_strategy=readahead");
968 	else if (opt->cache_strategy == EROFS_ZIP_CACHE_READAROUND)
969 		seq_puts(seq, ",cache_strategy=readaround");
970 #endif
971 	if (test_opt(opt, DAX_ALWAYS))
972 		seq_puts(seq, ",dax=always");
973 	if (test_opt(opt, DAX_NEVER))
974 		seq_puts(seq, ",dax=never");
975 #ifdef CONFIG_EROFS_FS_ONDEMAND
976 	if (sbi->fsid)
977 		seq_printf(seq, ",fsid=%s", sbi->fsid);
978 	if (sbi->domain_id)
979 		seq_printf(seq, ",domain_id=%s", sbi->domain_id);
980 #endif
981 	return 0;
982 }
983 
984 const struct super_operations erofs_sops = {
985 	.put_super = erofs_put_super,
986 	.alloc_inode = erofs_alloc_inode,
987 	.free_inode = erofs_free_inode,
988 	.statfs = erofs_statfs,
989 	.show_options = erofs_show_options,
990 };
991 
992 module_init(erofs_module_init);
993 module_exit(erofs_module_exit);
994 
995 MODULE_DESCRIPTION("Enhanced ROM File System");
996 MODULE_AUTHOR("Gao Xiang, Chao Yu, Miao Xie, CONSUMER BG, HUAWEI Inc.");
997 MODULE_LICENSE("GPL");
998