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
18static struct kmem_cache *erofs_inode_cachep __read_mostly;
19
20void _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
37void _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
54static 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
81static void erofs_inode_init_once(void *ptr)
82{
83	struct erofs_inode *vi = ptr;
84
85	inode_init_once(&vi->vfs_inode);
86}
87
88static 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
101static 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
111static 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 */
128void *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
163static 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
174static 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
220static 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
282static 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!");
368out:
369	erofs_put_metabuf(&buf);
370	return ret;
371}
372
373static void erofs_default_options(struct erofs_sb_info *sbi)
374{
375#ifdef CONFIG_EROFS_FS_ZIP
376	sbi->opt.cache_strategy = EROFS_ZIP_CACHE_READAROUND;
377	sbi->opt.max_sync_decompress_pages = 3;
378	sbi->opt.sync_decompress = EROFS_SYNC_DECOMPRESS_AUTO;
379#endif
380#ifdef CONFIG_EROFS_FS_XATTR
381	set_opt(&sbi->opt, XATTR_USER);
382#endif
383#ifdef CONFIG_EROFS_FS_POSIX_ACL
384	set_opt(&sbi->opt, POSIX_ACL);
385#endif
386}
387
388enum {
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
400static 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
407static const struct constant_table erofs_dax_param_enums[] = {
408	{"always",	EROFS_MOUNT_DAX_ALWAYS},
409	{"never",	EROFS_MOUNT_DAX_NEVER},
410	{}
411};
412
413static 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
426static bool erofs_fc_set_dax_mode(struct fs_context *fc, unsigned int mode)
427{
428#ifdef CONFIG_FS_DAX
429	struct erofs_sb_info *sbi = fc->s_fs_info;
430
431	switch (mode) {
432	case EROFS_MOUNT_DAX_ALWAYS:
433		set_opt(&sbi->opt, DAX_ALWAYS);
434		clear_opt(&sbi->opt, DAX_NEVER);
435		return true;
436	case EROFS_MOUNT_DAX_NEVER:
437		set_opt(&sbi->opt, DAX_NEVER);
438		clear_opt(&sbi->opt, DAX_ALWAYS);
439		return true;
440	default:
441		DBG_BUGON(1);
442		return false;
443	}
444#else
445	errorfc(fc, "dax options not supported");
446	return false;
447#endif
448}
449
450static int erofs_fc_parse_param(struct fs_context *fc,
451				struct fs_parameter *param)
452{
453	struct erofs_sb_info *sbi = fc->s_fs_info;
454	struct fs_parse_result result;
455	struct erofs_device_info *dif;
456	int opt, ret;
457
458	opt = fs_parse(fc, erofs_fs_parameters, param, &result);
459	if (opt < 0)
460		return opt;
461
462	switch (opt) {
463	case Opt_user_xattr:
464#ifdef CONFIG_EROFS_FS_XATTR
465		if (result.boolean)
466			set_opt(&sbi->opt, XATTR_USER);
467		else
468			clear_opt(&sbi->opt, XATTR_USER);
469#else
470		errorfc(fc, "{,no}user_xattr options not supported");
471#endif
472		break;
473	case Opt_acl:
474#ifdef CONFIG_EROFS_FS_POSIX_ACL
475		if (result.boolean)
476			set_opt(&sbi->opt, POSIX_ACL);
477		else
478			clear_opt(&sbi->opt, POSIX_ACL);
479#else
480		errorfc(fc, "{,no}acl options not supported");
481#endif
482		break;
483	case Opt_cache_strategy:
484#ifdef CONFIG_EROFS_FS_ZIP
485		sbi->opt.cache_strategy = result.uint_32;
486#else
487		errorfc(fc, "compression not supported, cache_strategy ignored");
488#endif
489		break;
490	case Opt_dax:
491		if (!erofs_fc_set_dax_mode(fc, EROFS_MOUNT_DAX_ALWAYS))
492			return -EINVAL;
493		break;
494	case Opt_dax_enum:
495		if (!erofs_fc_set_dax_mode(fc, result.uint_32))
496			return -EINVAL;
497		break;
498	case Opt_device:
499		dif = kzalloc(sizeof(*dif), GFP_KERNEL);
500		if (!dif)
501			return -ENOMEM;
502		dif->path = kstrdup(param->string, GFP_KERNEL);
503		if (!dif->path) {
504			kfree(dif);
505			return -ENOMEM;
506		}
507		down_write(&sbi->devs->rwsem);
508		ret = idr_alloc(&sbi->devs->tree, dif, 0, 0, GFP_KERNEL);
509		up_write(&sbi->devs->rwsem);
510		if (ret < 0) {
511			kfree(dif->path);
512			kfree(dif);
513			return ret;
514		}
515		++sbi->devs->extra_devices;
516		break;
517#ifdef CONFIG_EROFS_FS_ONDEMAND
518	case Opt_fsid:
519		kfree(sbi->fsid);
520		sbi->fsid = kstrdup(param->string, GFP_KERNEL);
521		if (!sbi->fsid)
522			return -ENOMEM;
523		break;
524	case Opt_domain_id:
525		kfree(sbi->domain_id);
526		sbi->domain_id = kstrdup(param->string, GFP_KERNEL);
527		if (!sbi->domain_id)
528			return -ENOMEM;
529		break;
530#else
531	case Opt_fsid:
532	case Opt_domain_id:
533		errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
534		break;
535#endif
536	default:
537		return -ENOPARAM;
538	}
539	return 0;
540}
541
542static struct inode *erofs_nfs_get_inode(struct super_block *sb,
543					 u64 ino, u32 generation)
544{
545	return erofs_iget(sb, ino);
546}
547
548static struct dentry *erofs_fh_to_dentry(struct super_block *sb,
549		struct fid *fid, int fh_len, int fh_type)
550{
551	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
552				    erofs_nfs_get_inode);
553}
554
555static struct dentry *erofs_fh_to_parent(struct super_block *sb,
556		struct fid *fid, int fh_len, int fh_type)
557{
558	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
559				    erofs_nfs_get_inode);
560}
561
562static struct dentry *erofs_get_parent(struct dentry *child)
563{
564	erofs_nid_t nid;
565	unsigned int d_type;
566	int err;
567
568	err = erofs_namei(d_inode(child), &dotdot_name, &nid, &d_type);
569	if (err)
570		return ERR_PTR(err);
571	return d_obtain_alias(erofs_iget(child->d_sb, nid));
572}
573
574static const struct export_operations erofs_export_ops = {
575	.encode_fh = generic_encode_ino32_fh,
576	.fh_to_dentry = erofs_fh_to_dentry,
577	.fh_to_parent = erofs_fh_to_parent,
578	.get_parent = erofs_get_parent,
579};
580
581static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
582{
583	struct inode *inode;
584	struct erofs_sb_info *sbi = EROFS_SB(sb);
585	int err;
586
587	sb->s_magic = EROFS_SUPER_MAGIC;
588	sb->s_flags |= SB_RDONLY | SB_NOATIME;
589	sb->s_maxbytes = MAX_LFS_FILESIZE;
590	sb->s_op = &erofs_sops;
591
592	sbi->blkszbits = PAGE_SHIFT;
593	if (erofs_is_fscache_mode(sb)) {
594		sb->s_blocksize = PAGE_SIZE;
595		sb->s_blocksize_bits = PAGE_SHIFT;
596
597		err = erofs_fscache_register_fs(sb);
598		if (err)
599			return err;
600
601		err = super_setup_bdi(sb);
602		if (err)
603			return err;
604	} else {
605		if (!sb_set_blocksize(sb, PAGE_SIZE)) {
606			errorfc(fc, "failed to set initial blksize");
607			return -EINVAL;
608		}
609
610		sbi->dax_dev = fs_dax_get_by_bdev(sb->s_bdev,
611						  &sbi->dax_part_off,
612						  NULL, NULL);
613	}
614
615	err = erofs_read_superblock(sb);
616	if (err)
617		return err;
618
619	if (sb->s_blocksize_bits != sbi->blkszbits) {
620		if (erofs_is_fscache_mode(sb)) {
621			errorfc(fc, "unsupported blksize for fscache mode");
622			return -EINVAL;
623		}
624		if (!sb_set_blocksize(sb, 1 << sbi->blkszbits)) {
625			errorfc(fc, "failed to set erofs blksize");
626			return -EINVAL;
627		}
628	}
629
630	if (test_opt(&sbi->opt, DAX_ALWAYS)) {
631		if (!sbi->dax_dev) {
632			errorfc(fc, "DAX unsupported by block device. Turning off DAX.");
633			clear_opt(&sbi->opt, DAX_ALWAYS);
634		} else if (sbi->blkszbits != PAGE_SHIFT) {
635			errorfc(fc, "unsupported blocksize for DAX");
636			clear_opt(&sbi->opt, DAX_ALWAYS);
637		}
638	}
639
640	sb->s_time_gran = 1;
641	sb->s_xattr = erofs_xattr_handlers;
642	sb->s_export_op = &erofs_export_ops;
643
644	if (test_opt(&sbi->opt, POSIX_ACL))
645		sb->s_flags |= SB_POSIXACL;
646	else
647		sb->s_flags &= ~SB_POSIXACL;
648
649#ifdef CONFIG_EROFS_FS_ZIP
650	xa_init(&sbi->managed_pslots);
651#endif
652
653	inode = erofs_iget(sb, sbi->root_nid);
654	if (IS_ERR(inode))
655		return PTR_ERR(inode);
656
657	if (!S_ISDIR(inode->i_mode)) {
658		erofs_err(sb, "rootino(nid %llu) is not a directory(i_mode %o)",
659			  sbi->root_nid, inode->i_mode);
660		iput(inode);
661		return -EINVAL;
662	}
663
664	sb->s_root = d_make_root(inode);
665	if (!sb->s_root)
666		return -ENOMEM;
667
668	erofs_shrinker_register(sb);
669	if (erofs_sb_has_fragments(sbi) && sbi->packed_nid) {
670		sbi->packed_inode = erofs_iget(sb, sbi->packed_nid);
671		if (IS_ERR(sbi->packed_inode)) {
672			err = PTR_ERR(sbi->packed_inode);
673			sbi->packed_inode = NULL;
674			return err;
675		}
676	}
677	err = erofs_init_managed_cache(sb);
678	if (err)
679		return err;
680
681	err = erofs_xattr_prefixes_init(sb);
682	if (err)
683		return err;
684
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
693static int erofs_fc_get_tree(struct fs_context *fc)
694{
695	struct erofs_sb_info *sbi = fc->s_fs_info;
696
697	if (IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid)
698		return get_tree_nodev(fc, erofs_fc_fill_super);
699
700	return get_tree_bdev(fc, erofs_fc_fill_super);
701}
702
703static int erofs_fc_reconfigure(struct fs_context *fc)
704{
705	struct super_block *sb = fc->root->d_sb;
706	struct erofs_sb_info *sbi = EROFS_SB(sb);
707	struct erofs_sb_info *new_sbi = fc->s_fs_info;
708
709	DBG_BUGON(!sb_rdonly(sb));
710
711	if (new_sbi->fsid || new_sbi->domain_id)
712		erofs_info(sb, "ignoring reconfiguration for fsid|domain_id.");
713
714	if (test_opt(&new_sbi->opt, POSIX_ACL))
715		fc->sb_flags |= SB_POSIXACL;
716	else
717		fc->sb_flags &= ~SB_POSIXACL;
718
719	sbi->opt = new_sbi->opt;
720
721	fc->sb_flags |= SB_RDONLY;
722	return 0;
723}
724
725static int erofs_release_device_info(int id, void *ptr, void *data)
726{
727	struct erofs_device_info *dif = ptr;
728
729	fs_put_dax(dif->dax_dev, NULL);
730	if (dif->bdev_file)
731		fput(dif->bdev_file);
732	erofs_fscache_unregister_cookie(dif->fscache);
733	dif->fscache = NULL;
734	kfree(dif->path);
735	kfree(dif);
736	return 0;
737}
738
739static void erofs_free_dev_context(struct erofs_dev_context *devs)
740{
741	if (!devs)
742		return;
743	idr_for_each(&devs->tree, &erofs_release_device_info, NULL);
744	idr_destroy(&devs->tree);
745	kfree(devs);
746}
747
748static void erofs_fc_free(struct fs_context *fc)
749{
750	struct erofs_sb_info *sbi = fc->s_fs_info;
751
752	if (!sbi)
753		return;
754
755	erofs_free_dev_context(sbi->devs);
756	kfree(sbi->fsid);
757	kfree(sbi->domain_id);
758	kfree(sbi);
759}
760
761static const struct fs_context_operations erofs_context_ops = {
762	.parse_param	= erofs_fc_parse_param,
763	.get_tree       = erofs_fc_get_tree,
764	.reconfigure    = erofs_fc_reconfigure,
765	.free		= erofs_fc_free,
766};
767
768static int erofs_init_fs_context(struct fs_context *fc)
769{
770	struct erofs_sb_info *sbi;
771
772	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
773	if (!sbi)
774		return -ENOMEM;
775
776	sbi->devs = kzalloc(sizeof(struct erofs_dev_context), GFP_KERNEL);
777	if (!sbi->devs) {
778		kfree(sbi);
779		return -ENOMEM;
780	}
781	fc->s_fs_info = sbi;
782
783	idr_init(&sbi->devs->tree);
784	init_rwsem(&sbi->devs->rwsem);
785	erofs_default_options(sbi);
786	fc->ops = &erofs_context_ops;
787	return 0;
788}
789
790static void erofs_kill_sb(struct super_block *sb)
791{
792	struct erofs_sb_info *sbi = EROFS_SB(sb);
793
794	if (IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid)
795		kill_anon_super(sb);
796	else
797		kill_block_super(sb);
798
799	erofs_free_dev_context(sbi->devs);
800	fs_put_dax(sbi->dax_dev, NULL);
801	erofs_fscache_unregister_fs(sb);
802	kfree(sbi->fsid);
803	kfree(sbi->domain_id);
804	kfree(sbi);
805	sb->s_fs_info = NULL;
806}
807
808static void erofs_put_super(struct super_block *sb)
809{
810	struct erofs_sb_info *const sbi = EROFS_SB(sb);
811
812	DBG_BUGON(!sbi);
813
814	erofs_unregister_sysfs(sb);
815	erofs_shrinker_unregister(sb);
816	erofs_xattr_prefixes_cleanup(sb);
817#ifdef CONFIG_EROFS_FS_ZIP
818	iput(sbi->managed_cache);
819	sbi->managed_cache = NULL;
820#endif
821	iput(sbi->packed_inode);
822	sbi->packed_inode = NULL;
823	erofs_free_dev_context(sbi->devs);
824	sbi->devs = NULL;
825	erofs_fscache_unregister_fs(sb);
826}
827
828static struct file_system_type erofs_fs_type = {
829	.owner          = THIS_MODULE,
830	.name           = "erofs",
831	.init_fs_context = erofs_init_fs_context,
832	.kill_sb        = erofs_kill_sb,
833	.fs_flags       = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
834};
835MODULE_ALIAS_FS("erofs");
836
837static int __init erofs_module_init(void)
838{
839	int err;
840
841	erofs_check_ondisk_layout_definitions();
842
843	erofs_inode_cachep = kmem_cache_create("erofs_inode",
844			sizeof(struct erofs_inode), 0,
845			SLAB_RECLAIM_ACCOUNT | SLAB_ACCOUNT,
846			erofs_inode_init_once);
847	if (!erofs_inode_cachep)
848		return -ENOMEM;
849
850	err = erofs_init_shrinker();
851	if (err)
852		goto shrinker_err;
853
854	err = z_erofs_lzma_init();
855	if (err)
856		goto lzma_err;
857
858	err = z_erofs_deflate_init();
859	if (err)
860		goto deflate_err;
861
862	erofs_pcpubuf_init();
863	err = z_erofs_init_zip_subsystem();
864	if (err)
865		goto zip_err;
866
867	err = erofs_init_sysfs();
868	if (err)
869		goto sysfs_err;
870
871	err = register_filesystem(&erofs_fs_type);
872	if (err)
873		goto fs_err;
874
875	return 0;
876
877fs_err:
878	erofs_exit_sysfs();
879sysfs_err:
880	z_erofs_exit_zip_subsystem();
881zip_err:
882	z_erofs_deflate_exit();
883deflate_err:
884	z_erofs_lzma_exit();
885lzma_err:
886	erofs_exit_shrinker();
887shrinker_err:
888	kmem_cache_destroy(erofs_inode_cachep);
889	return err;
890}
891
892static void __exit erofs_module_exit(void)
893{
894	unregister_filesystem(&erofs_fs_type);
895
896	/* Ensure all RCU free inodes / pclusters are safe to be destroyed. */
897	rcu_barrier();
898
899	erofs_exit_sysfs();
900	z_erofs_exit_zip_subsystem();
901	z_erofs_deflate_exit();
902	z_erofs_lzma_exit();
903	erofs_exit_shrinker();
904	kmem_cache_destroy(erofs_inode_cachep);
905	erofs_pcpubuf_exit();
906}
907
908static int erofs_statfs(struct dentry *dentry, struct kstatfs *buf)
909{
910	struct super_block *sb = dentry->d_sb;
911	struct erofs_sb_info *sbi = EROFS_SB(sb);
912	u64 id = 0;
913
914	if (!erofs_is_fscache_mode(sb))
915		id = huge_encode_dev(sb->s_bdev->bd_dev);
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
922	buf->f_files = ULLONG_MAX;
923	buf->f_ffree = ULLONG_MAX - sbi->inos;
924
925	buf->f_namelen = EROFS_NAME_LEN;
926
927	buf->f_fsid    = u64_to_fsid(id);
928	return 0;
929}
930
931static int erofs_show_options(struct seq_file *seq, struct dentry *root)
932{
933	struct erofs_sb_info *sbi = EROFS_SB(root->d_sb);
934	struct erofs_mount_opts *opt = &sbi->opt;
935
936#ifdef CONFIG_EROFS_FS_XATTR
937	if (test_opt(opt, XATTR_USER))
938		seq_puts(seq, ",user_xattr");
939	else
940		seq_puts(seq, ",nouser_xattr");
941#endif
942#ifdef CONFIG_EROFS_FS_POSIX_ACL
943	if (test_opt(opt, POSIX_ACL))
944		seq_puts(seq, ",acl");
945	else
946		seq_puts(seq, ",noacl");
947#endif
948#ifdef CONFIG_EROFS_FS_ZIP
949	if (opt->cache_strategy == EROFS_ZIP_CACHE_DISABLED)
950		seq_puts(seq, ",cache_strategy=disabled");
951	else if (opt->cache_strategy == EROFS_ZIP_CACHE_READAHEAD)
952		seq_puts(seq, ",cache_strategy=readahead");
953	else if (opt->cache_strategy == EROFS_ZIP_CACHE_READAROUND)
954		seq_puts(seq, ",cache_strategy=readaround");
955#endif
956	if (test_opt(opt, DAX_ALWAYS))
957		seq_puts(seq, ",dax=always");
958	if (test_opt(opt, DAX_NEVER))
959		seq_puts(seq, ",dax=never");
960#ifdef CONFIG_EROFS_FS_ONDEMAND
961	if (sbi->fsid)
962		seq_printf(seq, ",fsid=%s", sbi->fsid);
963	if (sbi->domain_id)
964		seq_printf(seq, ",domain_id=%s", sbi->domain_id);
965#endif
966	return 0;
967}
968
969const struct super_operations erofs_sops = {
970	.put_super = erofs_put_super,
971	.alloc_inode = erofs_alloc_inode,
972	.free_inode = erofs_free_inode,
973	.statfs = erofs_statfs,
974	.show_options = erofs_show_options,
975};
976
977module_init(erofs_module_init);
978module_exit(erofs_module_exit);
979
980MODULE_DESCRIPTION("Enhanced ROM File System");
981MODULE_AUTHOR("Gao Xiang, Chao Yu, Miao Xie, CONSUMER BG, HUAWEI Inc.");
982MODULE_LICENSE("GPL");
983