xref: /openbmc/linux/fs/erofs/super.c (revision 97a532c3)
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/module.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 <linux/exportfs.h>
16  #include "xattr.h"
17  
18  #define CREATE_TRACE_POINTS
19  #include <trace/events/erofs.h>
20  
21  static struct kmem_cache *erofs_inode_cachep __read_mostly;
22  
23  void _erofs_err(struct super_block *sb, const char *func, 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, 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  	pr_info("(device %s): %pV", sb->s_id, &vaf);
48  	va_end(args);
49  }
50  
51  static int erofs_superblock_csum_verify(struct super_block *sb, void *sbdata)
52  {
53  	size_t len = 1 << EROFS_SB(sb)->blkszbits;
54  	struct erofs_super_block *dsb;
55  	u32 expected_crc, crc;
56  
57  	if (len > EROFS_SUPER_OFFSET)
58  		len -= EROFS_SUPER_OFFSET;
59  
60  	dsb = kmemdup(sbdata + EROFS_SUPER_OFFSET, len, GFP_KERNEL);
61  	if (!dsb)
62  		return -ENOMEM;
63  
64  	expected_crc = le32_to_cpu(dsb->checksum);
65  	dsb->checksum = 0;
66  	/* to allow for x86 boot sectors and other oddities. */
67  	crc = crc32c(~0, dsb, len);
68  	kfree(dsb);
69  
70  	if (crc != expected_crc) {
71  		erofs_err(sb, "invalid checksum 0x%08x, 0x%08x expected",
72  			  crc, expected_crc);
73  		return -EBADMSG;
74  	}
75  	return 0;
76  }
77  
78  static void erofs_inode_init_once(void *ptr)
79  {
80  	struct erofs_inode *vi = ptr;
81  
82  	inode_init_once(&vi->vfs_inode);
83  }
84  
85  static struct inode *erofs_alloc_inode(struct super_block *sb)
86  {
87  	struct erofs_inode *vi =
88  		alloc_inode_sb(sb, erofs_inode_cachep, GFP_KERNEL);
89  
90  	if (!vi)
91  		return NULL;
92  
93  	/* zero out everything except vfs_inode */
94  	memset(vi, 0, offsetof(struct erofs_inode, vfs_inode));
95  	return &vi->vfs_inode;
96  }
97  
98  static void erofs_free_inode(struct inode *inode)
99  {
100  	struct erofs_inode *vi = EROFS_I(inode);
101  
102  	if (inode->i_op == &erofs_fast_symlink_iops)
103  		kfree(inode->i_link);
104  	kfree(vi->xattr_shared_xattrs);
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, "unidentified incompatible feature %x, please upgrade kernel",
118  			   feature & ~EROFS_ALL_FEATURE_INCOMPAT);
119  		return false;
120  	}
121  	return true;
122  }
123  
124  /* read variable-sized metadata, offset will be aligned by 4-byte */
125  void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf,
126  			  erofs_off_t *offset, int *lengthp)
127  {
128  	u8 *buffer, *ptr;
129  	int len, i, cnt;
130  
131  	*offset = round_up(*offset, 4);
132  	ptr = erofs_bread(buf, erofs_blknr(sb, *offset), EROFS_KMAP);
133  	if (IS_ERR(ptr))
134  		return ptr;
135  
136  	len = le16_to_cpu(*(__le16 *)&ptr[erofs_blkoff(sb, *offset)]);
137  	if (!len)
138  		len = U16_MAX + 1;
139  	buffer = kmalloc(len, GFP_KERNEL);
140  	if (!buffer)
141  		return ERR_PTR(-ENOMEM);
142  	*offset += sizeof(__le16);
143  	*lengthp = len;
144  
145  	for (i = 0; i < len; i += cnt) {
146  		cnt = min_t(int, sb->s_blocksize - erofs_blkoff(sb, *offset),
147  			    len - i);
148  		ptr = erofs_bread(buf, erofs_blknr(sb, *offset), EROFS_KMAP);
149  		if (IS_ERR(ptr)) {
150  			kfree(buffer);
151  			return ptr;
152  		}
153  		memcpy(buffer + i, ptr + erofs_blkoff(sb, *offset), cnt);
154  		*offset += cnt;
155  	}
156  	return buffer;
157  }
158  
159  #ifndef CONFIG_EROFS_FS_ZIP
160  static int z_erofs_parse_cfgs(struct super_block *sb,
161  			      struct erofs_super_block *dsb)
162  {
163  	if (!dsb->u1.available_compr_algs)
164  		return 0;
165  
166  	erofs_err(sb, "compression disabled, unable to mount compressed EROFS");
167  	return -EOPNOTSUPP;
168  }
169  #endif
170  
171  static int erofs_init_device(struct erofs_buf *buf, struct super_block *sb,
172  			     struct erofs_device_info *dif, erofs_off_t *pos)
173  {
174  	struct erofs_sb_info *sbi = EROFS_SB(sb);
175  	struct erofs_fscache *fscache;
176  	struct erofs_deviceslot *dis;
177  	struct bdev_handle *bdev_handle;
178  	void *ptr;
179  
180  	ptr = erofs_read_metabuf(buf, sb, erofs_blknr(sb, *pos), EROFS_KMAP);
181  	if (IS_ERR(ptr))
182  		return PTR_ERR(ptr);
183  	dis = ptr + erofs_blkoff(sb, *pos);
184  
185  	if (!sbi->devs->flatdev && !dif->path) {
186  		if (!dis->tag[0]) {
187  			erofs_err(sb, "empty device tag @ pos %llu", *pos);
188  			return -EINVAL;
189  		}
190  		dif->path = kmemdup_nul(dis->tag, sizeof(dis->tag), GFP_KERNEL);
191  		if (!dif->path)
192  			return -ENOMEM;
193  	}
194  
195  	if (erofs_is_fscache_mode(sb)) {
196  		fscache = erofs_fscache_register_cookie(sb, dif->path, 0);
197  		if (IS_ERR(fscache))
198  			return PTR_ERR(fscache);
199  		dif->fscache = fscache;
200  	} else if (!sbi->devs->flatdev) {
201  		bdev_handle = bdev_open_by_path(dif->path, BLK_OPEN_READ,
202  						sb->s_type, NULL);
203  		if (IS_ERR(bdev_handle))
204  			return PTR_ERR(bdev_handle);
205  		dif->bdev_handle = bdev_handle;
206  		dif->dax_dev = fs_dax_get_by_bdev(bdev_handle->bdev,
207  				&dif->dax_part_off, NULL, NULL);
208  	}
209  
210  	dif->blocks = le32_to_cpu(dis->blocks);
211  	dif->mapped_blkaddr = le32_to_cpu(dis->mapped_blkaddr);
212  	sbi->total_blocks += dif->blocks;
213  	*pos += EROFS_DEVT_SLOT_SIZE;
214  	return 0;
215  }
216  
217  static int erofs_scan_devices(struct super_block *sb,
218  			      struct erofs_super_block *dsb)
219  {
220  	struct erofs_sb_info *sbi = EROFS_SB(sb);
221  	unsigned int ondisk_extradevs;
222  	erofs_off_t pos;
223  	struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
224  	struct erofs_device_info *dif;
225  	int id, err = 0;
226  
227  	sbi->total_blocks = sbi->primarydevice_blocks;
228  	if (!erofs_sb_has_device_table(sbi))
229  		ondisk_extradevs = 0;
230  	else
231  		ondisk_extradevs = le16_to_cpu(dsb->extra_devices);
232  
233  	if (sbi->devs->extra_devices &&
234  	    ondisk_extradevs != sbi->devs->extra_devices) {
235  		erofs_err(sb, "extra devices don't match (ondisk %u, given %u)",
236  			  ondisk_extradevs, sbi->devs->extra_devices);
237  		return -EINVAL;
238  	}
239  	if (!ondisk_extradevs)
240  		return 0;
241  
242  	if (!sbi->devs->extra_devices && !erofs_is_fscache_mode(sb))
243  		sbi->devs->flatdev = true;
244  
245  	sbi->device_id_mask = roundup_pow_of_two(ondisk_extradevs + 1) - 1;
246  	pos = le16_to_cpu(dsb->devt_slotoff) * EROFS_DEVT_SLOT_SIZE;
247  	down_read(&sbi->devs->rwsem);
248  	if (sbi->devs->extra_devices) {
249  		idr_for_each_entry(&sbi->devs->tree, dif, id) {
250  			err = erofs_init_device(&buf, sb, dif, &pos);
251  			if (err)
252  				break;
253  		}
254  	} else {
255  		for (id = 0; id < ondisk_extradevs; id++) {
256  			dif = kzalloc(sizeof(*dif), GFP_KERNEL);
257  			if (!dif) {
258  				err = -ENOMEM;
259  				break;
260  			}
261  
262  			err = idr_alloc(&sbi->devs->tree, dif, 0, 0, GFP_KERNEL);
263  			if (err < 0) {
264  				kfree(dif);
265  				break;
266  			}
267  			++sbi->devs->extra_devices;
268  
269  			err = erofs_init_device(&buf, sb, dif, &pos);
270  			if (err)
271  				break;
272  		}
273  	}
274  	up_read(&sbi->devs->rwsem);
275  	erofs_put_metabuf(&buf);
276  	return err;
277  }
278  
279  static int erofs_read_superblock(struct super_block *sb)
280  {
281  	struct erofs_sb_info *sbi;
282  	struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
283  	struct erofs_super_block *dsb;
284  	void *data;
285  	int ret;
286  
287  	data = erofs_read_metabuf(&buf, sb, 0, EROFS_KMAP);
288  	if (IS_ERR(data)) {
289  		erofs_err(sb, "cannot read erofs superblock");
290  		return PTR_ERR(data);
291  	}
292  
293  	sbi = EROFS_SB(sb);
294  	dsb = (struct erofs_super_block *)(data + EROFS_SUPER_OFFSET);
295  
296  	ret = -EINVAL;
297  	if (le32_to_cpu(dsb->magic) != EROFS_SUPER_MAGIC_V1) {
298  		erofs_err(sb, "cannot find valid erofs superblock");
299  		goto out;
300  	}
301  
302  	sbi->blkszbits  = dsb->blkszbits;
303  	if (sbi->blkszbits < 9 || sbi->blkszbits > PAGE_SHIFT) {
304  		erofs_err(sb, "blkszbits %u isn't supported", sbi->blkszbits);
305  		goto out;
306  	}
307  	if (dsb->dirblkbits) {
308  		erofs_err(sb, "dirblkbits %u isn't supported", dsb->dirblkbits);
309  		goto out;
310  	}
311  
312  	sbi->feature_compat = le32_to_cpu(dsb->feature_compat);
313  	if (erofs_sb_has_sb_chksum(sbi)) {
314  		ret = erofs_superblock_csum_verify(sb, data);
315  		if (ret)
316  			goto out;
317  	}
318  
319  	ret = -EINVAL;
320  	if (!check_layout_compatibility(sb, dsb))
321  		goto out;
322  
323  	sbi->sb_size = 128 + dsb->sb_extslots * EROFS_SB_EXTSLOT_SIZE;
324  	if (sbi->sb_size > PAGE_SIZE - EROFS_SUPER_OFFSET) {
325  		erofs_err(sb, "invalid sb_extslots %u (more than a fs block)",
326  			  sbi->sb_size);
327  		goto out;
328  	}
329  	sbi->primarydevice_blocks = le32_to_cpu(dsb->blocks);
330  	sbi->meta_blkaddr = le32_to_cpu(dsb->meta_blkaddr);
331  #ifdef CONFIG_EROFS_FS_XATTR
332  	sbi->xattr_blkaddr = le32_to_cpu(dsb->xattr_blkaddr);
333  	sbi->xattr_prefix_start = le32_to_cpu(dsb->xattr_prefix_start);
334  	sbi->xattr_prefix_count = dsb->xattr_prefix_count;
335  	sbi->xattr_filter_reserved = dsb->xattr_filter_reserved;
336  #endif
337  	sbi->islotbits = ilog2(sizeof(struct erofs_inode_compact));
338  	sbi->root_nid = le16_to_cpu(dsb->root_nid);
339  	sbi->packed_nid = le64_to_cpu(dsb->packed_nid);
340  	sbi->inos = le64_to_cpu(dsb->inos);
341  
342  	sbi->build_time = le64_to_cpu(dsb->build_time);
343  	sbi->build_time_nsec = le32_to_cpu(dsb->build_time_nsec);
344  
345  	memcpy(&sb->s_uuid, dsb->uuid, sizeof(dsb->uuid));
346  
347  	ret = strscpy(sbi->volume_name, dsb->volume_name,
348  		      sizeof(dsb->volume_name));
349  	if (ret < 0) {	/* -E2BIG */
350  		erofs_err(sb, "bad volume name without NIL terminator");
351  		ret = -EFSCORRUPTED;
352  		goto out;
353  	}
354  
355  	/* parse on-disk compression configurations */
356  	ret = z_erofs_parse_cfgs(sb, dsb);
357  	if (ret < 0)
358  		goto out;
359  
360  	/* handle multiple devices */
361  	ret = erofs_scan_devices(sb, dsb);
362  
363  	if (erofs_is_fscache_mode(sb))
364  		erofs_info(sb, "EXPERIMENTAL fscache-based on-demand read feature in use. Use at your own risk!");
365  out:
366  	erofs_put_metabuf(&buf);
367  	return ret;
368  }
369  
370  static void erofs_default_options(struct erofs_sb_info *sbi)
371  {
372  #ifdef CONFIG_EROFS_FS_ZIP
373  	sbi->opt.cache_strategy = EROFS_ZIP_CACHE_READAROUND;
374  	sbi->opt.max_sync_decompress_pages = 3;
375  	sbi->opt.sync_decompress = EROFS_SYNC_DECOMPRESS_AUTO;
376  #endif
377  #ifdef CONFIG_EROFS_FS_XATTR
378  	set_opt(&sbi->opt, XATTR_USER);
379  #endif
380  #ifdef CONFIG_EROFS_FS_POSIX_ACL
381  	set_opt(&sbi->opt, POSIX_ACL);
382  #endif
383  }
384  
385  enum {
386  	Opt_user_xattr,
387  	Opt_acl,
388  	Opt_cache_strategy,
389  	Opt_dax,
390  	Opt_dax_enum,
391  	Opt_device,
392  	Opt_fsid,
393  	Opt_domain_id,
394  	Opt_err
395  };
396  
397  static const struct constant_table erofs_param_cache_strategy[] = {
398  	{"disabled",	EROFS_ZIP_CACHE_DISABLED},
399  	{"readahead",	EROFS_ZIP_CACHE_READAHEAD},
400  	{"readaround",	EROFS_ZIP_CACHE_READAROUND},
401  	{}
402  };
403  
404  static const struct constant_table erofs_dax_param_enums[] = {
405  	{"always",	EROFS_MOUNT_DAX_ALWAYS},
406  	{"never",	EROFS_MOUNT_DAX_NEVER},
407  	{}
408  };
409  
410  static const struct fs_parameter_spec erofs_fs_parameters[] = {
411  	fsparam_flag_no("user_xattr",	Opt_user_xattr),
412  	fsparam_flag_no("acl",		Opt_acl),
413  	fsparam_enum("cache_strategy",	Opt_cache_strategy,
414  		     erofs_param_cache_strategy),
415  	fsparam_flag("dax",             Opt_dax),
416  	fsparam_enum("dax",		Opt_dax_enum, erofs_dax_param_enums),
417  	fsparam_string("device",	Opt_device),
418  	fsparam_string("fsid",		Opt_fsid),
419  	fsparam_string("domain_id",	Opt_domain_id),
420  	{}
421  };
422  
423  static bool erofs_fc_set_dax_mode(struct fs_context *fc, unsigned int mode)
424  {
425  #ifdef CONFIG_FS_DAX
426  	struct erofs_sb_info *sbi = fc->s_fs_info;
427  
428  	switch (mode) {
429  	case EROFS_MOUNT_DAX_ALWAYS:
430  		warnfc(fc, "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
431  		set_opt(&sbi->opt, DAX_ALWAYS);
432  		clear_opt(&sbi->opt, DAX_NEVER);
433  		return true;
434  	case EROFS_MOUNT_DAX_NEVER:
435  		set_opt(&sbi->opt, DAX_NEVER);
436  		clear_opt(&sbi->opt, DAX_ALWAYS);
437  		return true;
438  	default:
439  		DBG_BUGON(1);
440  		return false;
441  	}
442  #else
443  	errorfc(fc, "dax options not supported");
444  	return false;
445  #endif
446  }
447  
448  static int erofs_fc_parse_param(struct fs_context *fc,
449  				struct fs_parameter *param)
450  {
451  	struct erofs_sb_info *sbi = fc->s_fs_info;
452  	struct fs_parse_result result;
453  	struct erofs_device_info *dif;
454  	int opt, ret;
455  
456  	opt = fs_parse(fc, erofs_fs_parameters, param, &result);
457  	if (opt < 0)
458  		return opt;
459  
460  	switch (opt) {
461  	case Opt_user_xattr:
462  #ifdef CONFIG_EROFS_FS_XATTR
463  		if (result.boolean)
464  			set_opt(&sbi->opt, XATTR_USER);
465  		else
466  			clear_opt(&sbi->opt, XATTR_USER);
467  #else
468  		errorfc(fc, "{,no}user_xattr options not supported");
469  #endif
470  		break;
471  	case Opt_acl:
472  #ifdef CONFIG_EROFS_FS_POSIX_ACL
473  		if (result.boolean)
474  			set_opt(&sbi->opt, POSIX_ACL);
475  		else
476  			clear_opt(&sbi->opt, POSIX_ACL);
477  #else
478  		errorfc(fc, "{,no}acl options not supported");
479  #endif
480  		break;
481  	case Opt_cache_strategy:
482  #ifdef CONFIG_EROFS_FS_ZIP
483  		sbi->opt.cache_strategy = result.uint_32;
484  #else
485  		errorfc(fc, "compression not supported, cache_strategy ignored");
486  #endif
487  		break;
488  	case Opt_dax:
489  		if (!erofs_fc_set_dax_mode(fc, EROFS_MOUNT_DAX_ALWAYS))
490  			return -EINVAL;
491  		break;
492  	case Opt_dax_enum:
493  		if (!erofs_fc_set_dax_mode(fc, result.uint_32))
494  			return -EINVAL;
495  		break;
496  	case Opt_device:
497  		dif = kzalloc(sizeof(*dif), GFP_KERNEL);
498  		if (!dif)
499  			return -ENOMEM;
500  		dif->path = kstrdup(param->string, GFP_KERNEL);
501  		if (!dif->path) {
502  			kfree(dif);
503  			return -ENOMEM;
504  		}
505  		down_write(&sbi->devs->rwsem);
506  		ret = idr_alloc(&sbi->devs->tree, dif, 0, 0, GFP_KERNEL);
507  		up_write(&sbi->devs->rwsem);
508  		if (ret < 0) {
509  			kfree(dif->path);
510  			kfree(dif);
511  			return ret;
512  		}
513  		++sbi->devs->extra_devices;
514  		break;
515  #ifdef CONFIG_EROFS_FS_ONDEMAND
516  	case Opt_fsid:
517  		kfree(sbi->fsid);
518  		sbi->fsid = kstrdup(param->string, GFP_KERNEL);
519  		if (!sbi->fsid)
520  			return -ENOMEM;
521  		break;
522  	case Opt_domain_id:
523  		kfree(sbi->domain_id);
524  		sbi->domain_id = kstrdup(param->string, GFP_KERNEL);
525  		if (!sbi->domain_id)
526  			return -ENOMEM;
527  		break;
528  #else
529  	case Opt_fsid:
530  	case Opt_domain_id:
531  		errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
532  		break;
533  #endif
534  	default:
535  		return -ENOPARAM;
536  	}
537  	return 0;
538  }
539  
540  static struct inode *erofs_nfs_get_inode(struct super_block *sb,
541  					 u64 ino, u32 generation)
542  {
543  	return erofs_iget(sb, ino);
544  }
545  
546  static struct dentry *erofs_fh_to_dentry(struct super_block *sb,
547  		struct fid *fid, int fh_len, int fh_type)
548  {
549  	return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
550  				    erofs_nfs_get_inode);
551  }
552  
553  static struct dentry *erofs_fh_to_parent(struct super_block *sb,
554  		struct fid *fid, int fh_len, int fh_type)
555  {
556  	return generic_fh_to_parent(sb, fid, fh_len, fh_type,
557  				    erofs_nfs_get_inode);
558  }
559  
560  static struct dentry *erofs_get_parent(struct dentry *child)
561  {
562  	erofs_nid_t nid;
563  	unsigned int d_type;
564  	int err;
565  
566  	err = erofs_namei(d_inode(child), &dotdot_name, &nid, &d_type);
567  	if (err)
568  		return ERR_PTR(err);
569  	return d_obtain_alias(erofs_iget(child->d_sb, nid));
570  }
571  
572  static const struct export_operations erofs_export_ops = {
573  	.fh_to_dentry = erofs_fh_to_dentry,
574  	.fh_to_parent = erofs_fh_to_parent,
575  	.get_parent = erofs_get_parent,
576  };
577  
578  static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
579  {
580  	struct inode *inode;
581  	struct erofs_sb_info *sbi = EROFS_SB(sb);
582  	int err;
583  
584  	sb->s_magic = EROFS_SUPER_MAGIC;
585  	sb->s_flags |= SB_RDONLY | SB_NOATIME;
586  	sb->s_maxbytes = MAX_LFS_FILESIZE;
587  	sb->s_op = &erofs_sops;
588  
589  	sbi->blkszbits = PAGE_SHIFT;
590  	if (erofs_is_fscache_mode(sb)) {
591  		sb->s_blocksize = PAGE_SIZE;
592  		sb->s_blocksize_bits = PAGE_SHIFT;
593  
594  		err = erofs_fscache_register_fs(sb);
595  		if (err)
596  			return err;
597  
598  		err = super_setup_bdi(sb);
599  		if (err)
600  			return err;
601  	} else {
602  		if (!sb_set_blocksize(sb, PAGE_SIZE)) {
603  			errorfc(fc, "failed to set initial blksize");
604  			return -EINVAL;
605  		}
606  
607  		sbi->dax_dev = fs_dax_get_by_bdev(sb->s_bdev,
608  						  &sbi->dax_part_off,
609  						  NULL, NULL);
610  	}
611  
612  	err = erofs_read_superblock(sb);
613  	if (err)
614  		return err;
615  
616  	if (sb->s_blocksize_bits != sbi->blkszbits) {
617  		if (erofs_is_fscache_mode(sb)) {
618  			errorfc(fc, "unsupported blksize for fscache mode");
619  			return -EINVAL;
620  		}
621  		if (!sb_set_blocksize(sb, 1 << sbi->blkszbits)) {
622  			errorfc(fc, "failed to set erofs blksize");
623  			return -EINVAL;
624  		}
625  	}
626  
627  	if (test_opt(&sbi->opt, DAX_ALWAYS)) {
628  		if (!sbi->dax_dev) {
629  			errorfc(fc, "DAX unsupported by block device. Turning off DAX.");
630  			clear_opt(&sbi->opt, DAX_ALWAYS);
631  		} else if (sbi->blkszbits != PAGE_SHIFT) {
632  			errorfc(fc, "unsupported blocksize for DAX");
633  			clear_opt(&sbi->opt, DAX_ALWAYS);
634  		}
635  	}
636  
637  	sb->s_time_gran = 1;
638  	sb->s_xattr = erofs_xattr_handlers;
639  	sb->s_export_op = &erofs_export_ops;
640  
641  	if (test_opt(&sbi->opt, POSIX_ACL))
642  		sb->s_flags |= SB_POSIXACL;
643  	else
644  		sb->s_flags &= ~SB_POSIXACL;
645  
646  #ifdef CONFIG_EROFS_FS_ZIP
647  	xa_init(&sbi->managed_pslots);
648  #endif
649  
650  	inode = erofs_iget(sb, ROOT_NID(sbi));
651  	if (IS_ERR(inode))
652  		return PTR_ERR(inode);
653  
654  	if (!S_ISDIR(inode->i_mode)) {
655  		erofs_err(sb, "rootino(nid %llu) is not a directory(i_mode %o)",
656  			  ROOT_NID(sbi), inode->i_mode);
657  		iput(inode);
658  		return -EINVAL;
659  	}
660  
661  	sb->s_root = d_make_root(inode);
662  	if (!sb->s_root)
663  		return -ENOMEM;
664  
665  	erofs_shrinker_register(sb);
666  	if (erofs_sb_has_fragments(sbi) && sbi->packed_nid) {
667  		sbi->packed_inode = erofs_iget(sb, sbi->packed_nid);
668  		if (IS_ERR(sbi->packed_inode)) {
669  			err = PTR_ERR(sbi->packed_inode);
670  			sbi->packed_inode = NULL;
671  			return err;
672  		}
673  	}
674  	err = erofs_init_managed_cache(sb);
675  	if (err)
676  		return err;
677  
678  	err = erofs_xattr_prefixes_init(sb);
679  	if (err)
680  		return err;
681  
682  	err = erofs_register_sysfs(sb);
683  	if (err)
684  		return err;
685  
686  	erofs_info(sb, "mounted with root inode @ nid %llu.", ROOT_NID(sbi));
687  	return 0;
688  }
689  
690  static int erofs_fc_get_tree(struct fs_context *fc)
691  {
692  	struct erofs_sb_info *sbi = fc->s_fs_info;
693  
694  	if (IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid)
695  		return get_tree_nodev(fc, erofs_fc_fill_super);
696  
697  	return get_tree_bdev(fc, erofs_fc_fill_super);
698  }
699  
700  static int erofs_fc_reconfigure(struct fs_context *fc)
701  {
702  	struct super_block *sb = fc->root->d_sb;
703  	struct erofs_sb_info *sbi = EROFS_SB(sb);
704  	struct erofs_sb_info *new_sbi = fc->s_fs_info;
705  
706  	DBG_BUGON(!sb_rdonly(sb));
707  
708  	if (new_sbi->fsid || new_sbi->domain_id)
709  		erofs_info(sb, "ignoring reconfiguration for fsid|domain_id.");
710  
711  	if (test_opt(&new_sbi->opt, POSIX_ACL))
712  		fc->sb_flags |= SB_POSIXACL;
713  	else
714  		fc->sb_flags &= ~SB_POSIXACL;
715  
716  	sbi->opt = new_sbi->opt;
717  
718  	fc->sb_flags |= SB_RDONLY;
719  	return 0;
720  }
721  
722  static int erofs_release_device_info(int id, void *ptr, void *data)
723  {
724  	struct erofs_device_info *dif = ptr;
725  
726  	fs_put_dax(dif->dax_dev, NULL);
727  	if (dif->bdev_handle)
728  		bdev_release(dif->bdev_handle);
729  	erofs_fscache_unregister_cookie(dif->fscache);
730  	dif->fscache = NULL;
731  	kfree(dif->path);
732  	kfree(dif);
733  	return 0;
734  }
735  
736  static void erofs_free_dev_context(struct erofs_dev_context *devs)
737  {
738  	if (!devs)
739  		return;
740  	idr_for_each(&devs->tree, &erofs_release_device_info, NULL);
741  	idr_destroy(&devs->tree);
742  	kfree(devs);
743  }
744  
745  static void erofs_fc_free(struct fs_context *fc)
746  {
747  	struct erofs_sb_info *sbi = fc->s_fs_info;
748  
749  	if (!sbi)
750  		return;
751  
752  	erofs_free_dev_context(sbi->devs);
753  	kfree(sbi->fsid);
754  	kfree(sbi->domain_id);
755  	kfree(sbi);
756  }
757  
758  static const struct fs_context_operations erofs_context_ops = {
759  	.parse_param	= erofs_fc_parse_param,
760  	.get_tree       = erofs_fc_get_tree,
761  	.reconfigure    = erofs_fc_reconfigure,
762  	.free		= erofs_fc_free,
763  };
764  
765  static int erofs_init_fs_context(struct fs_context *fc)
766  {
767  	struct erofs_sb_info *sbi;
768  
769  	sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
770  	if (!sbi)
771  		return -ENOMEM;
772  
773  	sbi->devs = kzalloc(sizeof(struct erofs_dev_context), GFP_KERNEL);
774  	if (!sbi->devs) {
775  		kfree(sbi);
776  		return -ENOMEM;
777  	}
778  	fc->s_fs_info = sbi;
779  
780  	idr_init(&sbi->devs->tree);
781  	init_rwsem(&sbi->devs->rwsem);
782  	erofs_default_options(sbi);
783  	fc->ops = &erofs_context_ops;
784  	return 0;
785  }
786  
787  static void erofs_kill_sb(struct super_block *sb)
788  {
789  	struct erofs_sb_info *sbi = EROFS_SB(sb);
790  
791  	if (IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && sbi->fsid)
792  		kill_anon_super(sb);
793  	else
794  		kill_block_super(sb);
795  
796  	erofs_free_dev_context(sbi->devs);
797  	fs_put_dax(sbi->dax_dev, NULL);
798  	erofs_fscache_unregister_fs(sb);
799  	kfree(sbi->fsid);
800  	kfree(sbi->domain_id);
801  	kfree(sbi);
802  	sb->s_fs_info = NULL;
803  }
804  
805  static void erofs_put_super(struct super_block *sb)
806  {
807  	struct erofs_sb_info *const sbi = EROFS_SB(sb);
808  
809  	DBG_BUGON(!sbi);
810  
811  	erofs_unregister_sysfs(sb);
812  	erofs_shrinker_unregister(sb);
813  	erofs_xattr_prefixes_cleanup(sb);
814  #ifdef CONFIG_EROFS_FS_ZIP
815  	iput(sbi->managed_cache);
816  	sbi->managed_cache = NULL;
817  #endif
818  	iput(sbi->packed_inode);
819  	sbi->packed_inode = NULL;
820  	erofs_free_dev_context(sbi->devs);
821  	sbi->devs = NULL;
822  	erofs_fscache_unregister_fs(sb);
823  }
824  
825  static struct file_system_type erofs_fs_type = {
826  	.owner          = THIS_MODULE,
827  	.name           = "erofs",
828  	.init_fs_context = erofs_init_fs_context,
829  	.kill_sb        = erofs_kill_sb,
830  	.fs_flags       = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
831  };
832  MODULE_ALIAS_FS("erofs");
833  
834  static int __init erofs_module_init(void)
835  {
836  	int err;
837  
838  	erofs_check_ondisk_layout_definitions();
839  
840  	erofs_inode_cachep = kmem_cache_create("erofs_inode",
841  			sizeof(struct erofs_inode), 0,
842  			SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD | SLAB_ACCOUNT,
843  			erofs_inode_init_once);
844  	if (!erofs_inode_cachep)
845  		return -ENOMEM;
846  
847  	err = erofs_init_shrinker();
848  	if (err)
849  		goto shrinker_err;
850  
851  	err = z_erofs_lzma_init();
852  	if (err)
853  		goto lzma_err;
854  
855  	err = z_erofs_deflate_init();
856  	if (err)
857  		goto deflate_err;
858  
859  	erofs_pcpubuf_init();
860  	err = z_erofs_init_zip_subsystem();
861  	if (err)
862  		goto zip_err;
863  
864  	err = erofs_init_sysfs();
865  	if (err)
866  		goto sysfs_err;
867  
868  	err = register_filesystem(&erofs_fs_type);
869  	if (err)
870  		goto fs_err;
871  
872  	return 0;
873  
874  fs_err:
875  	erofs_exit_sysfs();
876  sysfs_err:
877  	z_erofs_exit_zip_subsystem();
878  zip_err:
879  	z_erofs_deflate_exit();
880  deflate_err:
881  	z_erofs_lzma_exit();
882  lzma_err:
883  	erofs_exit_shrinker();
884  shrinker_err:
885  	kmem_cache_destroy(erofs_inode_cachep);
886  	return err;
887  }
888  
889  static void __exit erofs_module_exit(void)
890  {
891  	unregister_filesystem(&erofs_fs_type);
892  
893  	/* Ensure all RCU free inodes / pclusters are safe to be destroyed. */
894  	rcu_barrier();
895  
896  	erofs_exit_sysfs();
897  	z_erofs_exit_zip_subsystem();
898  	z_erofs_deflate_exit();
899  	z_erofs_lzma_exit();
900  	erofs_exit_shrinker();
901  	kmem_cache_destroy(erofs_inode_cachep);
902  	erofs_pcpubuf_exit();
903  }
904  
905  static int erofs_statfs(struct dentry *dentry, struct kstatfs *buf)
906  {
907  	struct super_block *sb = dentry->d_sb;
908  	struct erofs_sb_info *sbi = EROFS_SB(sb);
909  	u64 id = 0;
910  
911  	if (!erofs_is_fscache_mode(sb))
912  		id = huge_encode_dev(sb->s_bdev->bd_dev);
913  
914  	buf->f_type = sb->s_magic;
915  	buf->f_bsize = sb->s_blocksize;
916  	buf->f_blocks = sbi->total_blocks;
917  	buf->f_bfree = buf->f_bavail = 0;
918  
919  	buf->f_files = ULLONG_MAX;
920  	buf->f_ffree = ULLONG_MAX - sbi->inos;
921  
922  	buf->f_namelen = EROFS_NAME_LEN;
923  
924  	buf->f_fsid    = u64_to_fsid(id);
925  	return 0;
926  }
927  
928  static int erofs_show_options(struct seq_file *seq, struct dentry *root)
929  {
930  	struct erofs_sb_info *sbi = EROFS_SB(root->d_sb);
931  	struct erofs_mount_opts *opt = &sbi->opt;
932  
933  #ifdef CONFIG_EROFS_FS_XATTR
934  	if (test_opt(opt, XATTR_USER))
935  		seq_puts(seq, ",user_xattr");
936  	else
937  		seq_puts(seq, ",nouser_xattr");
938  #endif
939  #ifdef CONFIG_EROFS_FS_POSIX_ACL
940  	if (test_opt(opt, POSIX_ACL))
941  		seq_puts(seq, ",acl");
942  	else
943  		seq_puts(seq, ",noacl");
944  #endif
945  #ifdef CONFIG_EROFS_FS_ZIP
946  	if (opt->cache_strategy == EROFS_ZIP_CACHE_DISABLED)
947  		seq_puts(seq, ",cache_strategy=disabled");
948  	else if (opt->cache_strategy == EROFS_ZIP_CACHE_READAHEAD)
949  		seq_puts(seq, ",cache_strategy=readahead");
950  	else if (opt->cache_strategy == EROFS_ZIP_CACHE_READAROUND)
951  		seq_puts(seq, ",cache_strategy=readaround");
952  #endif
953  	if (test_opt(opt, DAX_ALWAYS))
954  		seq_puts(seq, ",dax=always");
955  	if (test_opt(opt, DAX_NEVER))
956  		seq_puts(seq, ",dax=never");
957  #ifdef CONFIG_EROFS_FS_ONDEMAND
958  	if (sbi->fsid)
959  		seq_printf(seq, ",fsid=%s", sbi->fsid);
960  	if (sbi->domain_id)
961  		seq_printf(seq, ",domain_id=%s", sbi->domain_id);
962  #endif
963  	return 0;
964  }
965  
966  const struct super_operations erofs_sops = {
967  	.put_super = erofs_put_super,
968  	.alloc_inode = erofs_alloc_inode,
969  	.free_inode = erofs_free_inode,
970  	.statfs = erofs_statfs,
971  	.show_options = erofs_show_options,
972  };
973  
974  module_init(erofs_module_init);
975  module_exit(erofs_module_exit);
976  
977  MODULE_DESCRIPTION("Enhanced ROM File System");
978  MODULE_AUTHOR("Gao Xiang, Chao Yu, Miao Xie, CONSUMER BG, HUAWEI Inc.");
979  MODULE_LICENSE("GPL");
980