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 #ifndef __EROFS_INTERNAL_H
8 #define __EROFS_INTERNAL_H
11 #include <linux/dcache.h>
13 #include <linux/pagemap.h>
14 #include <linux/bio.h>
15 #include <linux/buffer_head.h>
16 #include <linux/magic.h>
17 #include <linux/slab.h>
18 #include <linux/vmalloc.h>
19 #include <linux/iomap.h>
22 /* redefine pr_fmt "erofs: " */
24 #define pr_fmt(fmt) "erofs: " fmt
26 __printf(3, 4) void _erofs_err(struct super_block *sb,
27 const char *function, const char *fmt, ...);
28 #define erofs_err(sb, fmt, ...) \
29 _erofs_err(sb, __func__, fmt "\n", ##__VA_ARGS__)
30 __printf(3, 4) void _erofs_info(struct super_block *sb,
31 const char *function, const char *fmt, ...);
32 #define erofs_info(sb, fmt, ...) \
33 _erofs_info(sb, __func__, fmt "\n", ##__VA_ARGS__)
34 #ifdef CONFIG_EROFS_FS_DEBUG
35 #define erofs_dbg(x, ...) pr_debug(x "\n", ##__VA_ARGS__)
36 #define DBG_BUGON BUG_ON
38 #define erofs_dbg(x, ...) ((void)0)
39 #define DBG_BUGON(x) ((void)(x))
40 #endif /* !CONFIG_EROFS_FS_DEBUG */
42 /* EROFS_SUPER_MAGIC_V1 to represent the whole file system */
43 #define EROFS_SUPER_MAGIC EROFS_SUPER_MAGIC_V1
45 typedef u64 erofs_nid_t;
46 typedef u64 erofs_off_t;
47 /* data type for filesystem-wide blocks number */
48 typedef u32 erofs_blk_t;
50 struct erofs_device_info {
52 struct block_device *bdev;
53 struct dax_device *dax_dev;
60 EROFS_SYNC_DECOMPRESS_AUTO,
61 EROFS_SYNC_DECOMPRESS_FORCE_ON,
62 EROFS_SYNC_DECOMPRESS_FORCE_OFF
65 struct erofs_mount_opts {
66 #ifdef CONFIG_EROFS_FS_ZIP
67 /* current strategy of how to use managed cache */
68 unsigned char cache_strategy;
69 /* strategy of sync decompression (0 - auto, 1 - force on, 2 - force off) */
70 unsigned int sync_decompress;
72 /* threshold for decompression synchronously */
73 unsigned int max_sync_decompress_pages;
75 unsigned int mount_opt;
78 struct erofs_dev_context {
80 struct rw_semaphore rwsem;
82 unsigned int extra_devices;
85 struct erofs_fs_context {
86 struct erofs_mount_opts opt;
87 struct erofs_dev_context *devs;
90 /* all filesystem-wide lz4 configurations */
91 struct erofs_sb_lz4_info {
92 /* # of pages needed for EROFS lz4 rolling decompression */
93 u16 max_distance_pages;
94 /* maximum possible blocks for pclusters in the filesystem */
98 struct erofs_sb_info {
99 struct erofs_mount_opts opt; /* options */
100 #ifdef CONFIG_EROFS_FS_ZIP
101 /* list for all registered superblocks, mainly for shrinker */
102 struct list_head list;
103 struct mutex umount_mutex;
105 /* managed XArray arranged in physical block number */
106 struct xarray managed_pslots;
108 unsigned int shrinker_run_no;
109 u16 available_compr_algs;
111 /* pseudo inode to manage cached pages */
112 struct inode *managed_cache;
114 struct erofs_sb_lz4_info lz4;
115 #endif /* CONFIG_EROFS_FS_ZIP */
116 struct erofs_dev_context *devs;
117 struct dax_device *dax_dev;
119 u32 primarydevice_blocks;
122 #ifdef CONFIG_EROFS_FS_XATTR
125 u16 device_id_mask; /* valid bits of device id to be used */
127 /* inode slot unit size in bit shift */
128 unsigned char islotbits;
130 u32 sb_size; /* total superblock size */
134 /* what we really care is nid, rather than ino.. */
135 erofs_nid_t root_nid;
136 /* used for statfs, f_files - f_favail */
139 u8 uuid[16]; /* 128-bit uuid for volume */
140 u8 volume_name[16]; /* volume name */
142 u32 feature_incompat;
145 struct kobject s_kobj; /* /sys/fs/erofs/<devname> */
146 struct completion s_kobj_unregister;
149 #define EROFS_SB(sb) ((struct erofs_sb_info *)(sb)->s_fs_info)
150 #define EROFS_I_SB(inode) ((struct erofs_sb_info *)(inode)->i_sb->s_fs_info)
152 /* Mount flags set via mount options or defaults */
153 #define EROFS_MOUNT_XATTR_USER 0x00000010
154 #define EROFS_MOUNT_POSIX_ACL 0x00000020
155 #define EROFS_MOUNT_DAX_ALWAYS 0x00000040
156 #define EROFS_MOUNT_DAX_NEVER 0x00000080
158 #define clear_opt(opt, option) ((opt)->mount_opt &= ~EROFS_MOUNT_##option)
159 #define set_opt(opt, option) ((opt)->mount_opt |= EROFS_MOUNT_##option)
160 #define test_opt(opt, option) ((opt)->mount_opt & EROFS_MOUNT_##option)
163 EROFS_ZIP_CACHE_DISABLED,
164 EROFS_ZIP_CACHE_READAHEAD,
165 EROFS_ZIP_CACHE_READAROUND
168 #ifdef CONFIG_EROFS_FS_ZIP
169 #define EROFS_LOCKED_MAGIC (INT_MIN | 0xE0F510CCL)
171 /* basic unit of the workstation of a super_block */
172 struct erofs_workgroup {
173 /* the workgroup index in the workstation */
176 /* overall workgroup reference count */
180 #if defined(CONFIG_SMP)
181 static inline bool erofs_workgroup_try_to_freeze(struct erofs_workgroup *grp,
185 if (val != atomic_cmpxchg(&grp->refcount, val, EROFS_LOCKED_MAGIC)) {
192 static inline void erofs_workgroup_unfreeze(struct erofs_workgroup *grp,
196 * other observers should notice all modifications
197 * in the freezing period.
200 atomic_set(&grp->refcount, orig_val);
204 static inline int erofs_wait_on_workgroup_freezed(struct erofs_workgroup *grp)
206 return atomic_cond_read_relaxed(&grp->refcount,
207 VAL != EROFS_LOCKED_MAGIC);
210 static inline bool erofs_workgroup_try_to_freeze(struct erofs_workgroup *grp,
214 /* no need to spin on UP platforms, let's just disable preemption. */
215 if (val != atomic_read(&grp->refcount)) {
222 static inline void erofs_workgroup_unfreeze(struct erofs_workgroup *grp,
228 static inline int erofs_wait_on_workgroup_freezed(struct erofs_workgroup *grp)
230 int v = atomic_read(&grp->refcount);
232 /* workgroup is never freezed on uniprocessor systems */
233 DBG_BUGON(v == EROFS_LOCKED_MAGIC);
236 #endif /* !CONFIG_SMP */
237 #endif /* !CONFIG_EROFS_FS_ZIP */
239 /* we strictly follow PAGE_SIZE and no buffer head yet */
240 #define LOG_BLOCK_SIZE PAGE_SHIFT
242 #undef LOG_SECTORS_PER_BLOCK
243 #define LOG_SECTORS_PER_BLOCK (PAGE_SHIFT - 9)
245 #undef SECTORS_PER_BLOCK
246 #define SECTORS_PER_BLOCK (1 << SECTORS_PER_BLOCK)
248 #define EROFS_BLKSIZ (1 << LOG_BLOCK_SIZE)
250 #if (EROFS_BLKSIZ % 4096 || !EROFS_BLKSIZ)
251 #error erofs cannot be used in this platform
254 enum erofs_kmap_type {
255 EROFS_NO_KMAP, /* don't map the buffer */
256 EROFS_KMAP, /* use kmap() to map the buffer */
257 EROFS_KMAP_ATOMIC, /* use kmap_atomic() to map the buffer */
263 enum erofs_kmap_type kmap_type;
265 #define __EROFS_BUF_INITIALIZER ((struct erofs_buf){ .page = NULL })
267 #define ROOT_NID(sb) ((sb)->root_nid)
269 #define erofs_blknr(addr) ((addr) / EROFS_BLKSIZ)
270 #define erofs_blkoff(addr) ((addr) % EROFS_BLKSIZ)
271 #define blknr_to_addr(nr) ((erofs_off_t)(nr) * EROFS_BLKSIZ)
273 static inline erofs_off_t iloc(struct erofs_sb_info *sbi, erofs_nid_t nid)
275 return blknr_to_addr(sbi->meta_blkaddr) + (nid << sbi->islotbits);
278 #define EROFS_FEATURE_FUNCS(name, compat, feature) \
279 static inline bool erofs_sb_has_##name(struct erofs_sb_info *sbi) \
281 return sbi->feature_##compat & EROFS_FEATURE_##feature; \
284 EROFS_FEATURE_FUNCS(zero_padding, incompat, INCOMPAT_ZERO_PADDING)
285 EROFS_FEATURE_FUNCS(compr_cfgs, incompat, INCOMPAT_COMPR_CFGS)
286 EROFS_FEATURE_FUNCS(big_pcluster, incompat, INCOMPAT_BIG_PCLUSTER)
287 EROFS_FEATURE_FUNCS(chunked_file, incompat, INCOMPAT_CHUNKED_FILE)
288 EROFS_FEATURE_FUNCS(device_table, incompat, INCOMPAT_DEVICE_TABLE)
289 EROFS_FEATURE_FUNCS(compr_head2, incompat, INCOMPAT_COMPR_HEAD2)
290 EROFS_FEATURE_FUNCS(ztailpacking, incompat, INCOMPAT_ZTAILPACKING)
291 EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM)
293 /* atomic flag definitions */
294 #define EROFS_I_EA_INITED_BIT 0
295 #define EROFS_I_Z_INITED_BIT 1
297 /* bitlock definitions (arranged in reverse order) */
298 #define EROFS_I_BL_XATTR_BIT (BITS_PER_LONG - 1)
299 #define EROFS_I_BL_Z_BIT (BITS_PER_LONG - 2)
304 /* atomic flags (including bitlocks) */
307 unsigned char datalayout;
308 unsigned char inode_isize;
309 unsigned short xattr_isize;
311 unsigned int xattr_shared_count;
312 unsigned int *xattr_shared_xattrs;
315 erofs_blk_t raw_blkaddr;
317 unsigned short chunkformat;
318 unsigned char chunkbits;
320 #ifdef CONFIG_EROFS_FS_ZIP
322 unsigned short z_advise;
323 unsigned char z_algorithmtype[2];
324 unsigned char z_logical_clusterbits;
325 unsigned long z_tailextent_headlcn;
326 unsigned int z_idataoff;
327 unsigned short z_idata_size;
329 #endif /* CONFIG_EROFS_FS_ZIP */
331 /* the corresponding vfs inode */
332 struct inode vfs_inode;
335 #define EROFS_I(ptr) \
336 container_of(ptr, struct erofs_inode, vfs_inode)
338 static inline unsigned long erofs_inode_datablocks(struct inode *inode)
340 /* since i_size cannot be changed */
341 return DIV_ROUND_UP(inode->i_size, EROFS_BLKSIZ);
344 static inline unsigned int erofs_bitrange(unsigned int value, unsigned int bit,
348 return (value >> bit) & ((1 << bits) - 1);
352 static inline unsigned int erofs_inode_version(unsigned int value)
354 return erofs_bitrange(value, EROFS_I_VERSION_BIT,
355 EROFS_I_VERSION_BITS);
358 static inline unsigned int erofs_inode_datalayout(unsigned int value)
360 return erofs_bitrange(value, EROFS_I_DATALAYOUT_BIT,
361 EROFS_I_DATALAYOUT_BITS);
365 * Different from grab_cache_page_nowait(), reclaiming is never triggered
366 * when allocating new pages.
369 struct page *erofs_grab_cache_page_nowait(struct address_space *mapping,
372 return pagecache_get_page(mapping, index,
373 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
374 readahead_gfp_mask(mapping) & ~__GFP_RECLAIM);
377 extern const struct super_operations erofs_sops;
379 extern const struct address_space_operations erofs_raw_access_aops;
380 extern const struct address_space_operations z_erofs_aops;
383 * Logical to physical block mapping
385 * Different with other file systems, it is used for 2 access modes:
387 * 1) RAW access mode:
389 * Users pass a valid (m_lblk, m_lofs -- usually 0) pair,
390 * and get the valid m_pblk, m_pofs and the longest m_len(in bytes).
392 * Note that m_lblk in the RAW access mode refers to the number of
393 * the compressed ondisk block rather than the uncompressed
394 * in-memory block for the compressed file.
396 * m_pofs equals to m_lofs except for the inline data page.
398 * 2) Normal access mode:
400 * If the inode is not compressed, it has no difference with
401 * the RAW access mode. However, if the inode is compressed,
402 * users should pass a valid (m_lblk, m_lofs) pair, and get
403 * the needed m_pblk, m_pofs, m_len to get the compressed data
404 * and the updated m_lblk, m_lofs which indicates the start
405 * of the corresponding uncompressed data in the file.
408 BH_Encoded = BH_PrivateStart,
412 /* Has a disk mapping */
413 #define EROFS_MAP_MAPPED (1 << BH_Mapped)
414 /* Located in metadata (could be copied from bd_inode) */
415 #define EROFS_MAP_META (1 << BH_Meta)
416 /* The extent is encoded */
417 #define EROFS_MAP_ENCODED (1 << BH_Encoded)
418 /* The length of extent is full */
419 #define EROFS_MAP_FULL_MAPPED (1 << BH_FullMapped)
421 struct erofs_map_blocks {
422 struct erofs_buf buf;
424 erofs_off_t m_pa, m_la;
427 unsigned short m_deviceid;
428 char m_algorithmformat;
429 unsigned int m_flags;
432 /* Flags used by erofs_map_blocks_flatmode() */
433 #define EROFS_GET_BLOCKS_RAW 0x0001
435 * Used to get the exact decompressed length, e.g. fiemap (consider lookback
436 * approach instead if possible since it's more metadata lightweight.)
438 #define EROFS_GET_BLOCKS_FIEMAP 0x0002
439 /* Used to map the whole extent if non-negligible data is requested for LZMA */
440 #define EROFS_GET_BLOCKS_READMORE 0x0004
441 /* Used to map tail extent for tailpacking inline pcluster */
442 #define EROFS_GET_BLOCKS_FINDTAIL 0x0008
445 Z_EROFS_COMPRESSION_SHIFTED = Z_EROFS_COMPRESSION_MAX,
446 Z_EROFS_COMPRESSION_RUNTIME_MAX
450 extern const struct iomap_ops z_erofs_iomap_report_ops;
452 #ifdef CONFIG_EROFS_FS_ZIP
453 int z_erofs_fill_inode(struct inode *inode);
454 int z_erofs_map_blocks_iter(struct inode *inode,
455 struct erofs_map_blocks *map,
458 static inline int z_erofs_fill_inode(struct inode *inode) { return -EOPNOTSUPP; }
459 static inline int z_erofs_map_blocks_iter(struct inode *inode,
460 struct erofs_map_blocks *map,
465 #endif /* !CONFIG_EROFS_FS_ZIP */
467 struct erofs_map_dev {
468 struct block_device *m_bdev;
469 struct dax_device *m_daxdev;
472 unsigned int m_deviceid;
476 extern const struct file_operations erofs_file_fops;
477 void erofs_unmap_metabuf(struct erofs_buf *buf);
478 void erofs_put_metabuf(struct erofs_buf *buf);
479 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
480 erofs_blk_t blkaddr, enum erofs_kmap_type type);
481 int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev);
482 int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
486 static inline unsigned long erofs_inode_hash(erofs_nid_t nid)
488 #if BITS_PER_LONG == 32
489 return (nid >> 32) ^ (nid & 0xffffffff);
495 extern const struct inode_operations erofs_generic_iops;
496 extern const struct inode_operations erofs_symlink_iops;
497 extern const struct inode_operations erofs_fast_symlink_iops;
499 struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid, bool dir);
500 int erofs_getattr(struct user_namespace *mnt_userns, const struct path *path,
501 struct kstat *stat, u32 request_mask,
502 unsigned int query_flags);
505 extern const struct inode_operations erofs_dir_iops;
507 int erofs_namei(struct inode *dir, struct qstr *name,
508 erofs_nid_t *nid, unsigned int *d_type);
511 extern const struct file_operations erofs_dir_fops;
513 static inline void *erofs_vm_map_ram(struct page **pages, unsigned int count)
518 void *p = vm_map_ram(pages, count, -1);
520 /* retry two more times (totally 3 times) */
521 if (p || ++retried >= 3)
529 void *erofs_get_pcpubuf(unsigned int requiredpages);
530 void erofs_put_pcpubuf(void *ptr);
531 int erofs_pcpubuf_growsize(unsigned int nrpages);
532 void erofs_pcpubuf_init(void);
533 void erofs_pcpubuf_exit(void);
536 int erofs_register_sysfs(struct super_block *sb);
537 void erofs_unregister_sysfs(struct super_block *sb);
538 int __init erofs_init_sysfs(void);
539 void erofs_exit_sysfs(void);
541 /* utils.c / zdata.c */
542 struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp);
543 static inline void erofs_pagepool_add(struct page **pagepool,
546 set_page_private(page, (unsigned long)*pagepool);
549 void erofs_release_pages(struct page **pagepool);
551 #ifdef CONFIG_EROFS_FS_ZIP
552 int erofs_workgroup_put(struct erofs_workgroup *grp);
553 struct erofs_workgroup *erofs_find_workgroup(struct super_block *sb,
555 struct erofs_workgroup *erofs_insert_workgroup(struct super_block *sb,
556 struct erofs_workgroup *grp);
557 void erofs_workgroup_free_rcu(struct erofs_workgroup *grp);
558 void erofs_shrinker_register(struct super_block *sb);
559 void erofs_shrinker_unregister(struct super_block *sb);
560 int __init erofs_init_shrinker(void);
561 void erofs_exit_shrinker(void);
562 int __init z_erofs_init_zip_subsystem(void);
563 void z_erofs_exit_zip_subsystem(void);
564 int erofs_try_to_free_all_cached_pages(struct erofs_sb_info *sbi,
565 struct erofs_workgroup *egrp);
566 int erofs_try_to_free_cached_page(struct page *page);
567 int z_erofs_load_lz4_config(struct super_block *sb,
568 struct erofs_super_block *dsb,
569 struct z_erofs_lz4_cfgs *lz4, int len);
571 static inline void erofs_shrinker_register(struct super_block *sb) {}
572 static inline void erofs_shrinker_unregister(struct super_block *sb) {}
573 static inline int erofs_init_shrinker(void) { return 0; }
574 static inline void erofs_exit_shrinker(void) {}
575 static inline int z_erofs_init_zip_subsystem(void) { return 0; }
576 static inline void z_erofs_exit_zip_subsystem(void) {}
577 static inline int z_erofs_load_lz4_config(struct super_block *sb,
578 struct erofs_super_block *dsb,
579 struct z_erofs_lz4_cfgs *lz4, int len)
581 if (lz4 || dsb->u1.lz4_max_distance) {
582 erofs_err(sb, "lz4 algorithm isn't enabled");
587 #endif /* !CONFIG_EROFS_FS_ZIP */
589 #ifdef CONFIG_EROFS_FS_ZIP_LZMA
590 int z_erofs_lzma_init(void);
591 void z_erofs_lzma_exit(void);
592 int z_erofs_load_lzma_config(struct super_block *sb,
593 struct erofs_super_block *dsb,
594 struct z_erofs_lzma_cfgs *lzma, int size);
596 static inline int z_erofs_lzma_init(void) { return 0; }
597 static inline int z_erofs_lzma_exit(void) { return 0; }
598 static inline int z_erofs_load_lzma_config(struct super_block *sb,
599 struct erofs_super_block *dsb,
600 struct z_erofs_lzma_cfgs *lzma, int size) {
602 erofs_err(sb, "lzma algorithm isn't enabled");
607 #endif /* !CONFIG_EROFS_FS_ZIP */
609 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */
611 #endif /* __EROFS_INTERNAL_H */