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 erofs_fscache *fscache;
53 struct block_device *bdev;
54 struct dax_device *dax_dev;
62 EROFS_SYNC_DECOMPRESS_AUTO,
63 EROFS_SYNC_DECOMPRESS_FORCE_ON,
64 EROFS_SYNC_DECOMPRESS_FORCE_OFF
67 struct erofs_mount_opts {
68 #ifdef CONFIG_EROFS_FS_ZIP
69 /* current strategy of how to use managed cache */
70 unsigned char cache_strategy;
71 /* strategy of sync decompression (0 - auto, 1 - force on, 2 - force off) */
72 unsigned int sync_decompress;
74 /* threshold for decompression synchronously */
75 unsigned int max_sync_decompress_pages;
77 unsigned int mount_opt;
81 struct erofs_dev_context {
83 struct rw_semaphore rwsem;
85 unsigned int extra_devices;
88 struct erofs_fs_context {
89 struct erofs_mount_opts opt;
90 struct erofs_dev_context *devs;
93 /* all filesystem-wide lz4 configurations */
94 struct erofs_sb_lz4_info {
95 /* # of pages needed for EROFS lz4 rolling decompression */
96 u16 max_distance_pages;
97 /* maximum possible blocks for pclusters in the filesystem */
101 struct erofs_fscache {
102 struct fscache_cookie *cookie;
106 struct erofs_sb_info {
107 struct erofs_mount_opts opt; /* options */
108 #ifdef CONFIG_EROFS_FS_ZIP
109 /* list for all registered superblocks, mainly for shrinker */
110 struct list_head list;
111 struct mutex umount_mutex;
113 /* managed XArray arranged in physical block number */
114 struct xarray managed_pslots;
116 unsigned int shrinker_run_no;
117 u16 available_compr_algs;
119 /* pseudo inode to manage cached pages */
120 struct inode *managed_cache;
122 struct erofs_sb_lz4_info lz4;
123 #endif /* CONFIG_EROFS_FS_ZIP */
124 struct erofs_dev_context *devs;
125 struct dax_device *dax_dev;
128 u32 primarydevice_blocks;
131 #ifdef CONFIG_EROFS_FS_XATTR
134 u16 device_id_mask; /* valid bits of device id to be used */
136 /* inode slot unit size in bit shift */
137 unsigned char islotbits;
139 u32 sb_size; /* total superblock size */
143 /* what we really care is nid, rather than ino.. */
144 erofs_nid_t root_nid;
145 /* used for statfs, f_files - f_favail */
148 u8 uuid[16]; /* 128-bit uuid for volume */
149 u8 volume_name[16]; /* volume name */
151 u32 feature_incompat;
154 struct kobject s_kobj; /* /sys/fs/erofs/<devname> */
155 struct completion s_kobj_unregister;
157 /* fscache support */
158 struct fscache_volume *volume;
159 struct erofs_fscache *s_fscache;
162 #define EROFS_SB(sb) ((struct erofs_sb_info *)(sb)->s_fs_info)
163 #define EROFS_I_SB(inode) ((struct erofs_sb_info *)(inode)->i_sb->s_fs_info)
165 /* Mount flags set via mount options or defaults */
166 #define EROFS_MOUNT_XATTR_USER 0x00000010
167 #define EROFS_MOUNT_POSIX_ACL 0x00000020
168 #define EROFS_MOUNT_DAX_ALWAYS 0x00000040
169 #define EROFS_MOUNT_DAX_NEVER 0x00000080
171 #define clear_opt(opt, option) ((opt)->mount_opt &= ~EROFS_MOUNT_##option)
172 #define set_opt(opt, option) ((opt)->mount_opt |= EROFS_MOUNT_##option)
173 #define test_opt(opt, option) ((opt)->mount_opt & EROFS_MOUNT_##option)
175 static inline bool erofs_is_fscache_mode(struct super_block *sb)
177 return IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && !sb->s_bdev;
181 EROFS_ZIP_CACHE_DISABLED,
182 EROFS_ZIP_CACHE_READAHEAD,
183 EROFS_ZIP_CACHE_READAROUND
186 #ifdef CONFIG_EROFS_FS_ZIP
187 #define EROFS_LOCKED_MAGIC (INT_MIN | 0xE0F510CCL)
189 /* basic unit of the workstation of a super_block */
190 struct erofs_workgroup {
191 /* the workgroup index in the workstation */
194 /* overall workgroup reference count */
198 #if defined(CONFIG_SMP)
199 static inline bool erofs_workgroup_try_to_freeze(struct erofs_workgroup *grp,
203 if (val != atomic_cmpxchg(&grp->refcount, val, EROFS_LOCKED_MAGIC)) {
210 static inline void erofs_workgroup_unfreeze(struct erofs_workgroup *grp,
214 * other observers should notice all modifications
215 * in the freezing period.
218 atomic_set(&grp->refcount, orig_val);
222 static inline int erofs_wait_on_workgroup_freezed(struct erofs_workgroup *grp)
224 return atomic_cond_read_relaxed(&grp->refcount,
225 VAL != EROFS_LOCKED_MAGIC);
228 static inline bool erofs_workgroup_try_to_freeze(struct erofs_workgroup *grp,
232 /* no need to spin on UP platforms, let's just disable preemption. */
233 if (val != atomic_read(&grp->refcount)) {
240 static inline void erofs_workgroup_unfreeze(struct erofs_workgroup *grp,
246 static inline int erofs_wait_on_workgroup_freezed(struct erofs_workgroup *grp)
248 int v = atomic_read(&grp->refcount);
250 /* workgroup is never freezed on uniprocessor systems */
251 DBG_BUGON(v == EROFS_LOCKED_MAGIC);
254 #endif /* !CONFIG_SMP */
255 #endif /* !CONFIG_EROFS_FS_ZIP */
257 /* we strictly follow PAGE_SIZE and no buffer head yet */
258 #define LOG_BLOCK_SIZE PAGE_SHIFT
260 #undef LOG_SECTORS_PER_BLOCK
261 #define LOG_SECTORS_PER_BLOCK (PAGE_SHIFT - 9)
263 #undef SECTORS_PER_BLOCK
264 #define SECTORS_PER_BLOCK (1 << SECTORS_PER_BLOCK)
266 #define EROFS_BLKSIZ (1 << LOG_BLOCK_SIZE)
268 #if (EROFS_BLKSIZ % 4096 || !EROFS_BLKSIZ)
269 #error erofs cannot be used in this platform
272 enum erofs_kmap_type {
273 EROFS_NO_KMAP, /* don't map the buffer */
274 EROFS_KMAP, /* use kmap() to map the buffer */
275 EROFS_KMAP_ATOMIC, /* use kmap_atomic() to map the buffer */
281 enum erofs_kmap_type kmap_type;
283 #define __EROFS_BUF_INITIALIZER ((struct erofs_buf){ .page = NULL })
285 #define ROOT_NID(sb) ((sb)->root_nid)
287 #define erofs_blknr(addr) ((addr) / EROFS_BLKSIZ)
288 #define erofs_blkoff(addr) ((addr) % EROFS_BLKSIZ)
289 #define blknr_to_addr(nr) ((erofs_off_t)(nr) * EROFS_BLKSIZ)
291 static inline erofs_off_t iloc(struct erofs_sb_info *sbi, erofs_nid_t nid)
293 return blknr_to_addr(sbi->meta_blkaddr) + (nid << sbi->islotbits);
296 #define EROFS_FEATURE_FUNCS(name, compat, feature) \
297 static inline bool erofs_sb_has_##name(struct erofs_sb_info *sbi) \
299 return sbi->feature_##compat & EROFS_FEATURE_##feature; \
302 EROFS_FEATURE_FUNCS(zero_padding, incompat, INCOMPAT_ZERO_PADDING)
303 EROFS_FEATURE_FUNCS(compr_cfgs, incompat, INCOMPAT_COMPR_CFGS)
304 EROFS_FEATURE_FUNCS(big_pcluster, incompat, INCOMPAT_BIG_PCLUSTER)
305 EROFS_FEATURE_FUNCS(chunked_file, incompat, INCOMPAT_CHUNKED_FILE)
306 EROFS_FEATURE_FUNCS(device_table, incompat, INCOMPAT_DEVICE_TABLE)
307 EROFS_FEATURE_FUNCS(compr_head2, incompat, INCOMPAT_COMPR_HEAD2)
308 EROFS_FEATURE_FUNCS(ztailpacking, incompat, INCOMPAT_ZTAILPACKING)
309 EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM)
311 /* atomic flag definitions */
312 #define EROFS_I_EA_INITED_BIT 0
313 #define EROFS_I_Z_INITED_BIT 1
315 /* bitlock definitions (arranged in reverse order) */
316 #define EROFS_I_BL_XATTR_BIT (BITS_PER_LONG - 1)
317 #define EROFS_I_BL_Z_BIT (BITS_PER_LONG - 2)
322 /* atomic flags (including bitlocks) */
325 unsigned char datalayout;
326 unsigned char inode_isize;
327 unsigned short xattr_isize;
329 unsigned int xattr_shared_count;
330 unsigned int *xattr_shared_xattrs;
333 erofs_blk_t raw_blkaddr;
335 unsigned short chunkformat;
336 unsigned char chunkbits;
338 #ifdef CONFIG_EROFS_FS_ZIP
340 unsigned short z_advise;
341 unsigned char z_algorithmtype[2];
342 unsigned char z_logical_clusterbits;
343 unsigned long z_tailextent_headlcn;
344 erofs_off_t z_idataoff;
345 unsigned short z_idata_size;
347 #endif /* CONFIG_EROFS_FS_ZIP */
349 /* the corresponding vfs inode */
350 struct inode vfs_inode;
353 #define EROFS_I(ptr) \
354 container_of(ptr, struct erofs_inode, vfs_inode)
356 static inline unsigned long erofs_inode_datablocks(struct inode *inode)
358 /* since i_size cannot be changed */
359 return DIV_ROUND_UP(inode->i_size, EROFS_BLKSIZ);
362 static inline unsigned int erofs_bitrange(unsigned int value, unsigned int bit,
366 return (value >> bit) & ((1 << bits) - 1);
370 static inline unsigned int erofs_inode_version(unsigned int value)
372 return erofs_bitrange(value, EROFS_I_VERSION_BIT,
373 EROFS_I_VERSION_BITS);
376 static inline unsigned int erofs_inode_datalayout(unsigned int value)
378 return erofs_bitrange(value, EROFS_I_DATALAYOUT_BIT,
379 EROFS_I_DATALAYOUT_BITS);
383 * Different from grab_cache_page_nowait(), reclaiming is never triggered
384 * when allocating new pages.
387 struct page *erofs_grab_cache_page_nowait(struct address_space *mapping,
390 return pagecache_get_page(mapping, index,
391 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
392 readahead_gfp_mask(mapping) & ~__GFP_RECLAIM);
395 extern const struct super_operations erofs_sops;
397 extern const struct address_space_operations erofs_raw_access_aops;
398 extern const struct address_space_operations z_erofs_aops;
401 BH_Encoded = BH_PrivateStart,
405 /* Has a disk mapping */
406 #define EROFS_MAP_MAPPED (1 << BH_Mapped)
407 /* Located in metadata (could be copied from bd_inode) */
408 #define EROFS_MAP_META (1 << BH_Meta)
409 /* The extent is encoded */
410 #define EROFS_MAP_ENCODED (1 << BH_Encoded)
411 /* The length of extent is full */
412 #define EROFS_MAP_FULL_MAPPED (1 << BH_FullMapped)
414 struct erofs_map_blocks {
415 struct erofs_buf buf;
417 erofs_off_t m_pa, m_la;
420 unsigned short m_deviceid;
421 char m_algorithmformat;
422 unsigned int m_flags;
425 /* Flags used by erofs_map_blocks_flatmode() */
426 #define EROFS_GET_BLOCKS_RAW 0x0001
428 * Used to get the exact decompressed length, e.g. fiemap (consider lookback
429 * approach instead if possible since it's more metadata lightweight.)
431 #define EROFS_GET_BLOCKS_FIEMAP 0x0002
432 /* Used to map the whole extent if non-negligible data is requested for LZMA */
433 #define EROFS_GET_BLOCKS_READMORE 0x0004
434 /* Used to map tail extent for tailpacking inline pcluster */
435 #define EROFS_GET_BLOCKS_FINDTAIL 0x0008
438 Z_EROFS_COMPRESSION_SHIFTED = Z_EROFS_COMPRESSION_MAX,
439 Z_EROFS_COMPRESSION_RUNTIME_MAX
443 extern const struct iomap_ops z_erofs_iomap_report_ops;
445 #ifdef CONFIG_EROFS_FS_ZIP
446 int z_erofs_fill_inode(struct inode *inode);
447 int z_erofs_map_blocks_iter(struct inode *inode,
448 struct erofs_map_blocks *map,
451 static inline int z_erofs_fill_inode(struct inode *inode) { return -EOPNOTSUPP; }
452 static inline int z_erofs_map_blocks_iter(struct inode *inode,
453 struct erofs_map_blocks *map,
458 #endif /* !CONFIG_EROFS_FS_ZIP */
460 struct erofs_map_dev {
461 struct erofs_fscache *m_fscache;
462 struct block_device *m_bdev;
463 struct dax_device *m_daxdev;
467 unsigned int m_deviceid;
471 extern const struct file_operations erofs_file_fops;
472 void erofs_unmap_metabuf(struct erofs_buf *buf);
473 void erofs_put_metabuf(struct erofs_buf *buf);
474 void *erofs_bread(struct erofs_buf *buf, struct inode *inode,
475 erofs_blk_t blkaddr, enum erofs_kmap_type type);
476 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
477 erofs_blk_t blkaddr, enum erofs_kmap_type type);
478 int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev);
479 int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
481 int erofs_map_blocks(struct inode *inode,
482 struct erofs_map_blocks *map, int flags);
485 static inline unsigned long erofs_inode_hash(erofs_nid_t nid)
487 #if BITS_PER_LONG == 32
488 return (nid >> 32) ^ (nid & 0xffffffff);
494 extern const struct inode_operations erofs_generic_iops;
495 extern const struct inode_operations erofs_symlink_iops;
496 extern const struct inode_operations erofs_fast_symlink_iops;
498 struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid, bool dir);
499 int erofs_getattr(struct user_namespace *mnt_userns, const struct path *path,
500 struct kstat *stat, u32 request_mask,
501 unsigned int query_flags);
504 extern const struct inode_operations erofs_dir_iops;
506 int erofs_namei(struct inode *dir, const struct qstr *name,
507 erofs_nid_t *nid, unsigned int *d_type);
510 extern const struct file_operations erofs_dir_fops;
512 static inline void *erofs_vm_map_ram(struct page **pages, unsigned int count)
517 void *p = vm_map_ram(pages, count, -1);
519 /* retry two more times (totally 3 times) */
520 if (p || ++retried >= 3)
528 void *erofs_get_pcpubuf(unsigned int requiredpages);
529 void erofs_put_pcpubuf(void *ptr);
530 int erofs_pcpubuf_growsize(unsigned int nrpages);
531 void erofs_pcpubuf_init(void);
532 void erofs_pcpubuf_exit(void);
535 int erofs_register_sysfs(struct super_block *sb);
536 void erofs_unregister_sysfs(struct super_block *sb);
537 int __init erofs_init_sysfs(void);
538 void erofs_exit_sysfs(void);
540 /* utils.c / zdata.c */
541 struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp);
542 static inline void erofs_pagepool_add(struct page **pagepool,
545 set_page_private(page, (unsigned long)*pagepool);
548 void erofs_release_pages(struct page **pagepool);
550 #ifdef CONFIG_EROFS_FS_ZIP
551 int erofs_workgroup_put(struct erofs_workgroup *grp);
552 struct erofs_workgroup *erofs_find_workgroup(struct super_block *sb,
554 struct erofs_workgroup *erofs_insert_workgroup(struct super_block *sb,
555 struct erofs_workgroup *grp);
556 void erofs_workgroup_free_rcu(struct erofs_workgroup *grp);
557 void erofs_shrinker_register(struct super_block *sb);
558 void erofs_shrinker_unregister(struct super_block *sb);
559 int __init erofs_init_shrinker(void);
560 void erofs_exit_shrinker(void);
561 int __init z_erofs_init_zip_subsystem(void);
562 void z_erofs_exit_zip_subsystem(void);
563 int erofs_try_to_free_all_cached_pages(struct erofs_sb_info *sbi,
564 struct erofs_workgroup *egrp);
565 int erofs_try_to_free_cached_page(struct page *page);
566 int z_erofs_load_lz4_config(struct super_block *sb,
567 struct erofs_super_block *dsb,
568 struct z_erofs_lz4_cfgs *lz4, int len);
570 static inline void erofs_shrinker_register(struct super_block *sb) {}
571 static inline void erofs_shrinker_unregister(struct super_block *sb) {}
572 static inline int erofs_init_shrinker(void) { return 0; }
573 static inline void erofs_exit_shrinker(void) {}
574 static inline int z_erofs_init_zip_subsystem(void) { return 0; }
575 static inline void z_erofs_exit_zip_subsystem(void) {}
576 static inline int z_erofs_load_lz4_config(struct super_block *sb,
577 struct erofs_super_block *dsb,
578 struct z_erofs_lz4_cfgs *lz4, int len)
580 if (lz4 || dsb->u1.lz4_max_distance) {
581 erofs_err(sb, "lz4 algorithm isn't enabled");
586 #endif /* !CONFIG_EROFS_FS_ZIP */
588 #ifdef CONFIG_EROFS_FS_ZIP_LZMA
589 int z_erofs_lzma_init(void);
590 void z_erofs_lzma_exit(void);
591 int z_erofs_load_lzma_config(struct super_block *sb,
592 struct erofs_super_block *dsb,
593 struct z_erofs_lzma_cfgs *lzma, int size);
595 static inline int z_erofs_lzma_init(void) { return 0; }
596 static inline int z_erofs_lzma_exit(void) { return 0; }
597 static inline int z_erofs_load_lzma_config(struct super_block *sb,
598 struct erofs_super_block *dsb,
599 struct z_erofs_lzma_cfgs *lzma, int size) {
601 erofs_err(sb, "lzma algorithm isn't enabled");
606 #endif /* !CONFIG_EROFS_FS_ZIP */
609 #ifdef CONFIG_EROFS_FS_ONDEMAND
610 int erofs_fscache_register_fs(struct super_block *sb);
611 void erofs_fscache_unregister_fs(struct super_block *sb);
613 int erofs_fscache_register_cookie(struct super_block *sb,
614 struct erofs_fscache **fscache,
615 char *name, bool need_inode);
616 void erofs_fscache_unregister_cookie(struct erofs_fscache **fscache);
618 extern const struct address_space_operations erofs_fscache_access_aops;
620 static inline int erofs_fscache_register_fs(struct super_block *sb)
624 static inline void erofs_fscache_unregister_fs(struct super_block *sb) {}
626 static inline int erofs_fscache_register_cookie(struct super_block *sb,
627 struct erofs_fscache **fscache,
628 char *name, bool need_inode)
633 static inline void erofs_fscache_unregister_cookie(struct erofs_fscache **fscache)
638 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */
640 #endif /* __EROFS_INTERNAL_H */