1 // SPDX-License-Identifier: GPL-2.0-only
3 * Copyright (C) 2017-2018 HUAWEI, Inc.
4 * https://www.huawei.com/
5 * Copyright (C) 2021, Alibaba Cloud
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>
18 #define CREATE_TRACE_POINTS
19 #include <trace/events/erofs.h>
21 static struct kmem_cache *erofs_inode_cachep __read_mostly;
23 void _erofs_err(struct super_block *sb, const char *function,
34 pr_err("(device %s): %s: %pV", sb->s_id, function, &vaf);
38 void _erofs_info(struct super_block *sb, const char *function,
49 pr_info("(device %s): %pV", sb->s_id, &vaf);
53 static int erofs_superblock_csum_verify(struct super_block *sb, void *sbdata)
55 size_t len = 1 << EROFS_SB(sb)->blkszbits;
56 struct erofs_super_block *dsb;
57 u32 expected_crc, crc;
59 if (len > EROFS_SUPER_OFFSET)
60 len -= EROFS_SUPER_OFFSET;
62 dsb = kmemdup(sbdata + EROFS_SUPER_OFFSET, len, GFP_KERNEL);
66 expected_crc = le32_to_cpu(dsb->checksum);
68 /* to allow for x86 boot sectors and other oddities. */
69 crc = crc32c(~0, dsb, len);
72 if (crc != expected_crc) {
73 erofs_err(sb, "invalid checksum 0x%08x, 0x%08x expected",
80 static void erofs_inode_init_once(void *ptr)
82 struct erofs_inode *vi = ptr;
84 inode_init_once(&vi->vfs_inode);
87 static struct inode *erofs_alloc_inode(struct super_block *sb)
89 struct erofs_inode *vi =
90 alloc_inode_sb(sb, erofs_inode_cachep, GFP_KERNEL);
95 /* zero out everything except vfs_inode */
96 memset(vi, 0, offsetof(struct erofs_inode, vfs_inode));
97 return &vi->vfs_inode;
100 static void erofs_free_inode(struct inode *inode)
102 struct erofs_inode *vi = EROFS_I(inode);
104 /* be careful of RCU symlink path */
105 if (inode->i_op == &erofs_fast_symlink_iops)
106 kfree(inode->i_link);
107 kfree(vi->xattr_shared_xattrs);
109 kmem_cache_free(erofs_inode_cachep, vi);
112 static bool check_layout_compatibility(struct super_block *sb,
113 struct erofs_super_block *dsb)
115 const unsigned int feature = le32_to_cpu(dsb->feature_incompat);
117 EROFS_SB(sb)->feature_incompat = feature;
119 /* check if current kernel meets all mandatory requirements */
120 if (feature & (~EROFS_ALL_FEATURE_INCOMPAT)) {
122 "unidentified incompatible feature %x, please upgrade kernel version",
123 feature & ~EROFS_ALL_FEATURE_INCOMPAT);
129 /* read variable-sized metadata, offset will be aligned by 4-byte */
130 void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf,
131 erofs_off_t *offset, int *lengthp)
136 *offset = round_up(*offset, 4);
137 ptr = erofs_bread(buf, erofs_blknr(sb, *offset), EROFS_KMAP);
141 len = le16_to_cpu(*(__le16 *)&ptr[erofs_blkoff(sb, *offset)]);
144 buffer = kmalloc(len, GFP_KERNEL);
146 return ERR_PTR(-ENOMEM);
147 *offset += sizeof(__le16);
150 for (i = 0; i < len; i += cnt) {
151 cnt = min_t(int, sb->s_blocksize - erofs_blkoff(sb, *offset),
153 ptr = erofs_bread(buf, erofs_blknr(sb, *offset), EROFS_KMAP);
158 memcpy(buffer + i, ptr + erofs_blkoff(sb, *offset), cnt);
164 #ifdef CONFIG_EROFS_FS_ZIP
165 static int erofs_load_compr_cfgs(struct super_block *sb,
166 struct erofs_super_block *dsb)
168 struct erofs_sb_info *sbi = EROFS_SB(sb);
169 struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
170 unsigned int algs, alg;
174 sbi->available_compr_algs = le16_to_cpu(dsb->u1.available_compr_algs);
175 if (sbi->available_compr_algs & ~Z_EROFS_ALL_COMPR_ALGS) {
176 erofs_err(sb, "try to load compressed fs with unsupported algorithms %x",
177 sbi->available_compr_algs & ~Z_EROFS_ALL_COMPR_ALGS);
181 erofs_init_metabuf(&buf, sb);
182 offset = EROFS_SUPER_OFFSET + sbi->sb_size;
184 for (algs = sbi->available_compr_algs; algs; algs >>= 1, ++alg) {
190 data = erofs_read_metadata(sb, &buf, &offset, &size);
197 case Z_EROFS_COMPRESSION_LZ4:
198 ret = z_erofs_load_lz4_config(sb, dsb, data, size);
200 case Z_EROFS_COMPRESSION_LZMA:
201 ret = z_erofs_load_lzma_config(sb, dsb, data, size);
211 erofs_put_metabuf(&buf);
215 static int erofs_load_compr_cfgs(struct super_block *sb,
216 struct erofs_super_block *dsb)
218 if (dsb->u1.available_compr_algs) {
219 erofs_err(sb, "try to load compressed fs when compression is disabled");
226 static int erofs_init_device(struct erofs_buf *buf, struct super_block *sb,
227 struct erofs_device_info *dif, erofs_off_t *pos)
229 struct erofs_sb_info *sbi = EROFS_SB(sb);
230 struct erofs_fscache *fscache;
231 struct erofs_deviceslot *dis;
232 struct block_device *bdev;
235 ptr = erofs_read_metabuf(buf, sb, erofs_blknr(sb, *pos), EROFS_KMAP);
238 dis = ptr + erofs_blkoff(sb, *pos);
242 erofs_err(sb, "empty device tag @ pos %llu", *pos);
245 dif->path = kmemdup_nul(dis->tag, sizeof(dis->tag), GFP_KERNEL);
250 if (erofs_is_fscache_mode(sb)) {
251 fscache = erofs_fscache_register_cookie(sb, dif->path, 0);
253 return PTR_ERR(fscache);
254 dif->fscache = fscache;
255 } else if (!sbi->devs->flatdev) {
256 bdev = blkdev_get_by_path(dif->path, FMODE_READ | FMODE_EXCL,
259 return PTR_ERR(bdev);
261 dif->dax_dev = fs_dax_get_by_bdev(bdev, &dif->dax_part_off,
265 dif->blocks = le32_to_cpu(dis->blocks);
266 dif->mapped_blkaddr = le32_to_cpu(dis->mapped_blkaddr);
267 sbi->total_blocks += dif->blocks;
268 *pos += EROFS_DEVT_SLOT_SIZE;
272 static int erofs_scan_devices(struct super_block *sb,
273 struct erofs_super_block *dsb)
275 struct erofs_sb_info *sbi = EROFS_SB(sb);
276 unsigned int ondisk_extradevs;
278 struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
279 struct erofs_device_info *dif;
282 sbi->total_blocks = sbi->primarydevice_blocks;
283 if (!erofs_sb_has_device_table(sbi))
284 ondisk_extradevs = 0;
286 ondisk_extradevs = le16_to_cpu(dsb->extra_devices);
288 if (sbi->devs->extra_devices &&
289 ondisk_extradevs != sbi->devs->extra_devices) {
290 erofs_err(sb, "extra devices don't match (ondisk %u, given %u)",
291 ondisk_extradevs, sbi->devs->extra_devices);
294 if (!ondisk_extradevs)
297 if (!sbi->devs->extra_devices && !erofs_is_fscache_mode(sb))
298 sbi->devs->flatdev = true;
300 sbi->device_id_mask = roundup_pow_of_two(ondisk_extradevs + 1) - 1;
301 pos = le16_to_cpu(dsb->devt_slotoff) * EROFS_DEVT_SLOT_SIZE;
302 down_read(&sbi->devs->rwsem);
303 if (sbi->devs->extra_devices) {
304 idr_for_each_entry(&sbi->devs->tree, dif, id) {
305 err = erofs_init_device(&buf, sb, dif, &pos);
310 for (id = 0; id < ondisk_extradevs; id++) {
311 dif = kzalloc(sizeof(*dif), GFP_KERNEL);
317 err = idr_alloc(&sbi->devs->tree, dif, 0, 0, GFP_KERNEL);
322 ++sbi->devs->extra_devices;
324 err = erofs_init_device(&buf, sb, dif, &pos);
329 up_read(&sbi->devs->rwsem);
330 erofs_put_metabuf(&buf);
334 static int erofs_read_superblock(struct super_block *sb)
336 struct erofs_sb_info *sbi;
337 struct erofs_buf buf = __EROFS_BUF_INITIALIZER;
338 struct erofs_super_block *dsb;
342 data = erofs_read_metabuf(&buf, sb, 0, EROFS_KMAP);
344 erofs_err(sb, "cannot read erofs superblock");
345 return PTR_ERR(data);
349 dsb = (struct erofs_super_block *)(data + EROFS_SUPER_OFFSET);
352 if (le32_to_cpu(dsb->magic) != EROFS_SUPER_MAGIC_V1) {
353 erofs_err(sb, "cannot find valid erofs superblock");
357 sbi->blkszbits = dsb->blkszbits;
358 if (sbi->blkszbits < 9 || sbi->blkszbits > PAGE_SHIFT) {
359 erofs_err(sb, "blkszbits %u isn't supported", sbi->blkszbits);
362 if (dsb->dirblkbits) {
363 erofs_err(sb, "dirblkbits %u isn't supported", dsb->dirblkbits);
367 sbi->feature_compat = le32_to_cpu(dsb->feature_compat);
368 if (erofs_sb_has_sb_chksum(sbi)) {
369 ret = erofs_superblock_csum_verify(sb, data);
375 if (!check_layout_compatibility(sb, dsb))
378 sbi->sb_size = 128 + dsb->sb_extslots * EROFS_SB_EXTSLOT_SIZE;
379 if (sbi->sb_size > PAGE_SIZE - EROFS_SUPER_OFFSET) {
380 erofs_err(sb, "invalid sb_extslots %u (more than a fs block)",
384 sbi->primarydevice_blocks = le32_to_cpu(dsb->blocks);
385 sbi->meta_blkaddr = le32_to_cpu(dsb->meta_blkaddr);
386 #ifdef CONFIG_EROFS_FS_XATTR
387 sbi->xattr_blkaddr = le32_to_cpu(dsb->xattr_blkaddr);
388 sbi->xattr_prefix_start = le32_to_cpu(dsb->xattr_prefix_start);
389 sbi->xattr_prefix_count = dsb->xattr_prefix_count;
391 sbi->islotbits = ilog2(sizeof(struct erofs_inode_compact));
392 sbi->root_nid = le16_to_cpu(dsb->root_nid);
393 sbi->packed_nid = le64_to_cpu(dsb->packed_nid);
394 sbi->inos = le64_to_cpu(dsb->inos);
396 sbi->build_time = le64_to_cpu(dsb->build_time);
397 sbi->build_time_nsec = le32_to_cpu(dsb->build_time_nsec);
399 memcpy(&sb->s_uuid, dsb->uuid, sizeof(dsb->uuid));
401 ret = strscpy(sbi->volume_name, dsb->volume_name,
402 sizeof(dsb->volume_name));
403 if (ret < 0) { /* -E2BIG */
404 erofs_err(sb, "bad volume name without NIL terminator");
409 /* parse on-disk compression configurations */
410 if (erofs_sb_has_compr_cfgs(sbi))
411 ret = erofs_load_compr_cfgs(sb, dsb);
413 ret = z_erofs_load_lz4_config(sb, dsb, NULL, 0);
417 /* handle multiple devices */
418 ret = erofs_scan_devices(sb, dsb);
420 if (erofs_is_fscache_mode(sb))
421 erofs_info(sb, "EXPERIMENTAL fscache-based on-demand read feature in use. Use at your own risk!");
422 if (erofs_sb_has_fragments(sbi))
423 erofs_info(sb, "EXPERIMENTAL compressed fragments feature in use. Use at your own risk!");
424 if (erofs_sb_has_dedupe(sbi))
425 erofs_info(sb, "EXPERIMENTAL global deduplication feature in use. Use at your own risk!");
427 erofs_put_metabuf(&buf);
431 /* set up default EROFS parameters */
432 static void erofs_default_options(struct erofs_fs_context *ctx)
434 #ifdef CONFIG_EROFS_FS_ZIP
435 ctx->opt.cache_strategy = EROFS_ZIP_CACHE_READAROUND;
436 ctx->opt.max_sync_decompress_pages = 3;
437 ctx->opt.sync_decompress = EROFS_SYNC_DECOMPRESS_AUTO;
439 #ifdef CONFIG_EROFS_FS_XATTR
440 set_opt(&ctx->opt, XATTR_USER);
442 #ifdef CONFIG_EROFS_FS_POSIX_ACL
443 set_opt(&ctx->opt, POSIX_ACL);
459 static const struct constant_table erofs_param_cache_strategy[] = {
460 {"disabled", EROFS_ZIP_CACHE_DISABLED},
461 {"readahead", EROFS_ZIP_CACHE_READAHEAD},
462 {"readaround", EROFS_ZIP_CACHE_READAROUND},
466 static const struct constant_table erofs_dax_param_enums[] = {
467 {"always", EROFS_MOUNT_DAX_ALWAYS},
468 {"never", EROFS_MOUNT_DAX_NEVER},
472 static const struct fs_parameter_spec erofs_fs_parameters[] = {
473 fsparam_flag_no("user_xattr", Opt_user_xattr),
474 fsparam_flag_no("acl", Opt_acl),
475 fsparam_enum("cache_strategy", Opt_cache_strategy,
476 erofs_param_cache_strategy),
477 fsparam_flag("dax", Opt_dax),
478 fsparam_enum("dax", Opt_dax_enum, erofs_dax_param_enums),
479 fsparam_string("device", Opt_device),
480 fsparam_string("fsid", Opt_fsid),
481 fsparam_string("domain_id", Opt_domain_id),
485 static bool erofs_fc_set_dax_mode(struct fs_context *fc, unsigned int mode)
488 struct erofs_fs_context *ctx = fc->fs_private;
491 case EROFS_MOUNT_DAX_ALWAYS:
492 warnfc(fc, "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
493 set_opt(&ctx->opt, DAX_ALWAYS);
494 clear_opt(&ctx->opt, DAX_NEVER);
496 case EROFS_MOUNT_DAX_NEVER:
497 set_opt(&ctx->opt, DAX_NEVER);
498 clear_opt(&ctx->opt, DAX_ALWAYS);
505 errorfc(fc, "dax options not supported");
510 static int erofs_fc_parse_param(struct fs_context *fc,
511 struct fs_parameter *param)
513 struct erofs_fs_context *ctx = fc->fs_private;
514 struct fs_parse_result result;
515 struct erofs_device_info *dif;
518 opt = fs_parse(fc, erofs_fs_parameters, param, &result);
524 #ifdef CONFIG_EROFS_FS_XATTR
526 set_opt(&ctx->opt, XATTR_USER);
528 clear_opt(&ctx->opt, XATTR_USER);
530 errorfc(fc, "{,no}user_xattr options not supported");
534 #ifdef CONFIG_EROFS_FS_POSIX_ACL
536 set_opt(&ctx->opt, POSIX_ACL);
538 clear_opt(&ctx->opt, POSIX_ACL);
540 errorfc(fc, "{,no}acl options not supported");
543 case Opt_cache_strategy:
544 #ifdef CONFIG_EROFS_FS_ZIP
545 ctx->opt.cache_strategy = result.uint_32;
547 errorfc(fc, "compression not supported, cache_strategy ignored");
551 if (!erofs_fc_set_dax_mode(fc, EROFS_MOUNT_DAX_ALWAYS))
555 if (!erofs_fc_set_dax_mode(fc, result.uint_32))
559 dif = kzalloc(sizeof(*dif), GFP_KERNEL);
562 dif->path = kstrdup(param->string, GFP_KERNEL);
567 down_write(&ctx->devs->rwsem);
568 ret = idr_alloc(&ctx->devs->tree, dif, 0, 0, GFP_KERNEL);
569 up_write(&ctx->devs->rwsem);
575 ++ctx->devs->extra_devices;
577 #ifdef CONFIG_EROFS_FS_ONDEMAND
580 ctx->fsid = kstrdup(param->string, GFP_KERNEL);
585 kfree(ctx->domain_id);
586 ctx->domain_id = kstrdup(param->string, GFP_KERNEL);
593 errorfc(fc, "%s option not supported", erofs_fs_parameters[opt].name);
602 #ifdef CONFIG_EROFS_FS_ZIP
603 static const struct address_space_operations managed_cache_aops;
605 static bool erofs_managed_cache_release_folio(struct folio *folio, gfp_t gfp)
608 struct address_space *const mapping = folio->mapping;
610 DBG_BUGON(!folio_test_locked(folio));
611 DBG_BUGON(mapping->a_ops != &managed_cache_aops);
613 if (folio_test_private(folio))
614 ret = erofs_try_to_free_cached_page(&folio->page);
620 * It will be called only on inode eviction. In case that there are still some
621 * decompression requests in progress, wait with rescheduling for a bit here.
622 * We could introduce an extra locking instead but it seems unnecessary.
624 static void erofs_managed_cache_invalidate_folio(struct folio *folio,
625 size_t offset, size_t length)
627 const size_t stop = length + offset;
629 DBG_BUGON(!folio_test_locked(folio));
631 /* Check for potential overflow in debug mode */
632 DBG_BUGON(stop > folio_size(folio) || stop < length);
634 if (offset == 0 && stop == folio_size(folio))
635 while (!erofs_managed_cache_release_folio(folio, GFP_NOFS))
639 static const struct address_space_operations managed_cache_aops = {
640 .release_folio = erofs_managed_cache_release_folio,
641 .invalidate_folio = erofs_managed_cache_invalidate_folio,
644 static int erofs_init_managed_cache(struct super_block *sb)
646 struct erofs_sb_info *const sbi = EROFS_SB(sb);
647 struct inode *const inode = new_inode(sb);
653 inode->i_size = OFFSET_MAX;
655 inode->i_mapping->a_ops = &managed_cache_aops;
656 mapping_set_gfp_mask(inode->i_mapping, GFP_NOFS);
657 sbi->managed_cache = inode;
661 static int erofs_init_managed_cache(struct super_block *sb) { return 0; }
664 static struct inode *erofs_nfs_get_inode(struct super_block *sb,
665 u64 ino, u32 generation)
667 return erofs_iget(sb, ino);
670 static struct dentry *erofs_fh_to_dentry(struct super_block *sb,
671 struct fid *fid, int fh_len, int fh_type)
673 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
674 erofs_nfs_get_inode);
677 static struct dentry *erofs_fh_to_parent(struct super_block *sb,
678 struct fid *fid, int fh_len, int fh_type)
680 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
681 erofs_nfs_get_inode);
684 static struct dentry *erofs_get_parent(struct dentry *child)
690 err = erofs_namei(d_inode(child), &dotdot_name, &nid, &d_type);
693 return d_obtain_alias(erofs_iget(child->d_sb, nid));
696 static const struct export_operations erofs_export_ops = {
697 .fh_to_dentry = erofs_fh_to_dentry,
698 .fh_to_parent = erofs_fh_to_parent,
699 .get_parent = erofs_get_parent,
702 static int erofs_fc_fill_pseudo_super(struct super_block *sb, struct fs_context *fc)
704 static const struct tree_descr empty_descr = {""};
706 return simple_fill_super(sb, EROFS_SUPER_MAGIC, &empty_descr);
709 static int erofs_fc_fill_super(struct super_block *sb, struct fs_context *fc)
712 struct erofs_sb_info *sbi;
713 struct erofs_fs_context *ctx = fc->fs_private;
716 sb->s_magic = EROFS_SUPER_MAGIC;
717 sb->s_flags |= SB_RDONLY | SB_NOATIME;
718 sb->s_maxbytes = MAX_LFS_FILESIZE;
719 sb->s_op = &erofs_sops;
721 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
727 sbi->devs = ctx->devs;
729 sbi->fsid = ctx->fsid;
731 sbi->domain_id = ctx->domain_id;
732 ctx->domain_id = NULL;
734 sbi->blkszbits = PAGE_SHIFT;
735 if (erofs_is_fscache_mode(sb)) {
736 sb->s_blocksize = PAGE_SIZE;
737 sb->s_blocksize_bits = PAGE_SHIFT;
739 err = erofs_fscache_register_fs(sb);
743 err = super_setup_bdi(sb);
747 if (!sb_set_blocksize(sb, PAGE_SIZE)) {
748 errorfc(fc, "failed to set initial blksize");
752 sbi->dax_dev = fs_dax_get_by_bdev(sb->s_bdev,
757 err = erofs_read_superblock(sb);
761 if (sb->s_blocksize_bits != sbi->blkszbits) {
762 if (erofs_is_fscache_mode(sb)) {
763 errorfc(fc, "unsupported blksize for fscache mode");
766 if (!sb_set_blocksize(sb, 1 << sbi->blkszbits)) {
767 errorfc(fc, "failed to set erofs blksize");
772 if (test_opt(&sbi->opt, DAX_ALWAYS)) {
774 errorfc(fc, "DAX unsupported by block device. Turning off DAX.");
775 clear_opt(&sbi->opt, DAX_ALWAYS);
776 } else if (sbi->blkszbits != PAGE_SHIFT) {
777 errorfc(fc, "unsupported blocksize for DAX");
778 clear_opt(&sbi->opt, DAX_ALWAYS);
783 sb->s_xattr = erofs_xattr_handlers;
784 sb->s_export_op = &erofs_export_ops;
786 if (test_opt(&sbi->opt, POSIX_ACL))
787 sb->s_flags |= SB_POSIXACL;
789 sb->s_flags &= ~SB_POSIXACL;
791 #ifdef CONFIG_EROFS_FS_ZIP
792 xa_init(&sbi->managed_pslots);
795 /* get the root inode */
796 inode = erofs_iget(sb, ROOT_NID(sbi));
798 return PTR_ERR(inode);
800 if (!S_ISDIR(inode->i_mode)) {
801 erofs_err(sb, "rootino(nid %llu) is not a directory(i_mode %o)",
802 ROOT_NID(sbi), inode->i_mode);
807 sb->s_root = d_make_root(inode);
811 erofs_shrinker_register(sb);
812 /* sb->s_umount is already locked, SB_ACTIVE and SB_BORN are not set */
813 if (erofs_sb_has_fragments(sbi) && sbi->packed_nid) {
814 sbi->packed_inode = erofs_iget(sb, sbi->packed_nid);
815 if (IS_ERR(sbi->packed_inode)) {
816 err = PTR_ERR(sbi->packed_inode);
817 sbi->packed_inode = NULL;
821 err = erofs_init_managed_cache(sb);
825 err = erofs_xattr_prefixes_init(sb);
829 err = erofs_register_sysfs(sb);
833 erofs_info(sb, "mounted with root inode @ nid %llu.", ROOT_NID(sbi));
837 static int erofs_fc_anon_get_tree(struct fs_context *fc)
839 return get_tree_nodev(fc, erofs_fc_fill_pseudo_super);
842 static int erofs_fc_get_tree(struct fs_context *fc)
844 struct erofs_fs_context *ctx = fc->fs_private;
846 if (IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && ctx->fsid)
847 return get_tree_nodev(fc, erofs_fc_fill_super);
849 return get_tree_bdev(fc, erofs_fc_fill_super);
852 static int erofs_fc_reconfigure(struct fs_context *fc)
854 struct super_block *sb = fc->root->d_sb;
855 struct erofs_sb_info *sbi = EROFS_SB(sb);
856 struct erofs_fs_context *ctx = fc->fs_private;
858 DBG_BUGON(!sb_rdonly(sb));
860 if (ctx->fsid || ctx->domain_id)
861 erofs_info(sb, "ignoring reconfiguration for fsid|domain_id.");
863 if (test_opt(&ctx->opt, POSIX_ACL))
864 fc->sb_flags |= SB_POSIXACL;
866 fc->sb_flags &= ~SB_POSIXACL;
870 fc->sb_flags |= SB_RDONLY;
874 static int erofs_release_device_info(int id, void *ptr, void *data)
876 struct erofs_device_info *dif = ptr;
878 fs_put_dax(dif->dax_dev, NULL);
880 blkdev_put(dif->bdev, FMODE_READ | FMODE_EXCL);
881 erofs_fscache_unregister_cookie(dif->fscache);
888 static void erofs_free_dev_context(struct erofs_dev_context *devs)
892 idr_for_each(&devs->tree, &erofs_release_device_info, NULL);
893 idr_destroy(&devs->tree);
897 static void erofs_fc_free(struct fs_context *fc)
899 struct erofs_fs_context *ctx = fc->fs_private;
901 erofs_free_dev_context(ctx->devs);
903 kfree(ctx->domain_id);
907 static const struct fs_context_operations erofs_context_ops = {
908 .parse_param = erofs_fc_parse_param,
909 .get_tree = erofs_fc_get_tree,
910 .reconfigure = erofs_fc_reconfigure,
911 .free = erofs_fc_free,
914 static const struct fs_context_operations erofs_anon_context_ops = {
915 .get_tree = erofs_fc_anon_get_tree,
918 static int erofs_init_fs_context(struct fs_context *fc)
920 struct erofs_fs_context *ctx;
922 /* pseudo mount for anon inodes */
923 if (fc->sb_flags & SB_KERNMOUNT) {
924 fc->ops = &erofs_anon_context_ops;
928 ctx = kzalloc(sizeof(*ctx), GFP_KERNEL);
931 ctx->devs = kzalloc(sizeof(struct erofs_dev_context), GFP_KERNEL);
936 fc->fs_private = ctx;
938 idr_init(&ctx->devs->tree);
939 init_rwsem(&ctx->devs->rwsem);
940 erofs_default_options(ctx);
941 fc->ops = &erofs_context_ops;
946 * could be triggered after deactivate_locked_super()
947 * is called, thus including umount and failed to initialize.
949 static void erofs_kill_sb(struct super_block *sb)
951 struct erofs_sb_info *sbi;
953 WARN_ON(sb->s_magic != EROFS_SUPER_MAGIC);
955 /* pseudo mount for anon inodes */
956 if (sb->s_flags & SB_KERNMOUNT) {
961 if (erofs_is_fscache_mode(sb))
964 kill_block_super(sb);
970 erofs_free_dev_context(sbi->devs);
971 fs_put_dax(sbi->dax_dev, NULL);
972 erofs_fscache_unregister_fs(sb);
974 kfree(sbi->domain_id);
976 sb->s_fs_info = NULL;
979 /* called when ->s_root is non-NULL */
980 static void erofs_put_super(struct super_block *sb)
982 struct erofs_sb_info *const sbi = EROFS_SB(sb);
986 erofs_unregister_sysfs(sb);
987 erofs_shrinker_unregister(sb);
988 erofs_xattr_prefixes_cleanup(sb);
989 #ifdef CONFIG_EROFS_FS_ZIP
990 iput(sbi->managed_cache);
991 sbi->managed_cache = NULL;
993 iput(sbi->packed_inode);
994 sbi->packed_inode = NULL;
995 erofs_free_dev_context(sbi->devs);
997 erofs_fscache_unregister_fs(sb);
1000 struct file_system_type erofs_fs_type = {
1001 .owner = THIS_MODULE,
1003 .init_fs_context = erofs_init_fs_context,
1004 .kill_sb = erofs_kill_sb,
1005 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
1007 MODULE_ALIAS_FS("erofs");
1009 static int __init erofs_module_init(void)
1013 erofs_check_ondisk_layout_definitions();
1015 erofs_inode_cachep = kmem_cache_create("erofs_inode",
1016 sizeof(struct erofs_inode), 0,
1017 SLAB_RECLAIM_ACCOUNT,
1018 erofs_inode_init_once);
1019 if (!erofs_inode_cachep) {
1024 err = erofs_init_shrinker();
1028 err = z_erofs_lzma_init();
1032 erofs_pcpubuf_init();
1033 err = z_erofs_init_zip_subsystem();
1037 err = erofs_init_sysfs();
1041 err = register_filesystem(&erofs_fs_type);
1050 z_erofs_exit_zip_subsystem();
1052 z_erofs_lzma_exit();
1054 erofs_exit_shrinker();
1056 kmem_cache_destroy(erofs_inode_cachep);
1061 static void __exit erofs_module_exit(void)
1063 unregister_filesystem(&erofs_fs_type);
1065 /* Ensure all RCU free inodes / pclusters are safe to be destroyed. */
1069 z_erofs_exit_zip_subsystem();
1070 z_erofs_lzma_exit();
1071 erofs_exit_shrinker();
1072 kmem_cache_destroy(erofs_inode_cachep);
1073 erofs_pcpubuf_exit();
1076 /* get filesystem statistics */
1077 static int erofs_statfs(struct dentry *dentry, struct kstatfs *buf)
1079 struct super_block *sb = dentry->d_sb;
1080 struct erofs_sb_info *sbi = EROFS_SB(sb);
1083 if (!erofs_is_fscache_mode(sb))
1084 id = huge_encode_dev(sb->s_bdev->bd_dev);
1086 buf->f_type = sb->s_magic;
1087 buf->f_bsize = sb->s_blocksize;
1088 buf->f_blocks = sbi->total_blocks;
1089 buf->f_bfree = buf->f_bavail = 0;
1091 buf->f_files = ULLONG_MAX;
1092 buf->f_ffree = ULLONG_MAX - sbi->inos;
1094 buf->f_namelen = EROFS_NAME_LEN;
1096 buf->f_fsid = u64_to_fsid(id);
1100 static int erofs_show_options(struct seq_file *seq, struct dentry *root)
1102 struct erofs_sb_info *sbi = EROFS_SB(root->d_sb);
1103 struct erofs_mount_opts *opt = &sbi->opt;
1105 #ifdef CONFIG_EROFS_FS_XATTR
1106 if (test_opt(opt, XATTR_USER))
1107 seq_puts(seq, ",user_xattr");
1109 seq_puts(seq, ",nouser_xattr");
1111 #ifdef CONFIG_EROFS_FS_POSIX_ACL
1112 if (test_opt(opt, POSIX_ACL))
1113 seq_puts(seq, ",acl");
1115 seq_puts(seq, ",noacl");
1117 #ifdef CONFIG_EROFS_FS_ZIP
1118 if (opt->cache_strategy == EROFS_ZIP_CACHE_DISABLED)
1119 seq_puts(seq, ",cache_strategy=disabled");
1120 else if (opt->cache_strategy == EROFS_ZIP_CACHE_READAHEAD)
1121 seq_puts(seq, ",cache_strategy=readahead");
1122 else if (opt->cache_strategy == EROFS_ZIP_CACHE_READAROUND)
1123 seq_puts(seq, ",cache_strategy=readaround");
1125 if (test_opt(opt, DAX_ALWAYS))
1126 seq_puts(seq, ",dax=always");
1127 if (test_opt(opt, DAX_NEVER))
1128 seq_puts(seq, ",dax=never");
1129 #ifdef CONFIG_EROFS_FS_ONDEMAND
1131 seq_printf(seq, ",fsid=%s", sbi->fsid);
1133 seq_printf(seq, ",domain_id=%s", sbi->domain_id);
1138 const struct super_operations erofs_sops = {
1139 .put_super = erofs_put_super,
1140 .alloc_inode = erofs_alloc_inode,
1141 .free_inode = erofs_free_inode,
1142 .statfs = erofs_statfs,
1143 .show_options = erofs_show_options,
1146 module_init(erofs_module_init);
1147 module_exit(erofs_module_exit);
1149 MODULE_DESCRIPTION("Enhanced ROM File System");
1150 MODULE_AUTHOR("Gao Xiang, Chao Yu, Miao Xie, CONSUMER BG, HUAWEI Inc.");
1151 MODULE_LICENSE("GPL");