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;
80 struct erofs_dev_context {
82 struct rw_semaphore rwsem;
84 unsigned int extra_devices;
87 struct erofs_fs_context {
88 struct erofs_mount_opts opt;
89 struct erofs_dev_context *devs;
94 /* all filesystem-wide lz4 configurations */
95 struct erofs_sb_lz4_info {
96 /* # of pages needed for EROFS lz4 rolling decompression */
97 u16 max_distance_pages;
98 /* maximum possible blocks for pclusters in the filesystem */
102 struct erofs_domain {
104 struct list_head list;
105 struct fscache_volume *volume;
109 struct erofs_fscache {
110 struct fscache_cookie *cookie;
112 struct inode *anon_inode;
113 struct erofs_domain *domain;
117 struct erofs_sb_info {
118 struct erofs_mount_opts opt; /* options */
119 #ifdef CONFIG_EROFS_FS_ZIP
120 /* list for all registered superblocks, mainly for shrinker */
121 struct list_head list;
122 struct mutex umount_mutex;
124 /* managed XArray arranged in physical block number */
125 struct xarray managed_pslots;
127 unsigned int shrinker_run_no;
128 u16 available_compr_algs;
130 /* pseudo inode to manage cached pages */
131 struct inode *managed_cache;
133 struct erofs_sb_lz4_info lz4;
134 struct inode *packed_inode;
135 #endif /* CONFIG_EROFS_FS_ZIP */
136 struct erofs_dev_context *devs;
137 struct dax_device *dax_dev;
140 u32 primarydevice_blocks;
143 #ifdef CONFIG_EROFS_FS_XATTR
146 u16 device_id_mask; /* valid bits of device id to be used */
148 /* inode slot unit size in bit shift */
149 unsigned char islotbits;
151 u32 sb_size; /* total superblock size */
155 /* what we really care is nid, rather than ino.. */
156 erofs_nid_t root_nid;
157 /* used for statfs, f_files - f_favail */
160 u8 uuid[16]; /* 128-bit uuid for volume */
161 u8 volume_name[16]; /* volume name */
163 u32 feature_incompat;
166 struct kobject s_kobj; /* /sys/fs/erofs/<devname> */
167 struct completion s_kobj_unregister;
169 /* fscache support */
170 struct fscache_volume *volume;
171 struct erofs_fscache *s_fscache;
172 struct erofs_domain *domain;
177 #define EROFS_SB(sb) ((struct erofs_sb_info *)(sb)->s_fs_info)
178 #define EROFS_I_SB(inode) ((struct erofs_sb_info *)(inode)->i_sb->s_fs_info)
180 /* Mount flags set via mount options or defaults */
181 #define EROFS_MOUNT_XATTR_USER 0x00000010
182 #define EROFS_MOUNT_POSIX_ACL 0x00000020
183 #define EROFS_MOUNT_DAX_ALWAYS 0x00000040
184 #define EROFS_MOUNT_DAX_NEVER 0x00000080
186 #define clear_opt(opt, option) ((opt)->mount_opt &= ~EROFS_MOUNT_##option)
187 #define set_opt(opt, option) ((opt)->mount_opt |= EROFS_MOUNT_##option)
188 #define test_opt(opt, option) ((opt)->mount_opt & EROFS_MOUNT_##option)
190 static inline bool erofs_is_fscache_mode(struct super_block *sb)
192 return IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && !sb->s_bdev;
196 EROFS_ZIP_CACHE_DISABLED,
197 EROFS_ZIP_CACHE_READAHEAD,
198 EROFS_ZIP_CACHE_READAROUND
201 #define EROFS_LOCKED_MAGIC (INT_MIN | 0xE0F510CCL)
203 /* basic unit of the workstation of a super_block */
204 struct erofs_workgroup {
205 /* the workgroup index in the workstation */
208 /* overall workgroup reference count */
212 static inline bool erofs_workgroup_try_to_freeze(struct erofs_workgroup *grp,
216 if (val != atomic_cmpxchg(&grp->refcount, val, EROFS_LOCKED_MAGIC)) {
223 static inline void erofs_workgroup_unfreeze(struct erofs_workgroup *grp,
227 * other observers should notice all modifications
228 * in the freezing period.
231 atomic_set(&grp->refcount, orig_val);
235 static inline int erofs_wait_on_workgroup_freezed(struct erofs_workgroup *grp)
237 return atomic_cond_read_relaxed(&grp->refcount,
238 VAL != EROFS_LOCKED_MAGIC);
241 /* we strictly follow PAGE_SIZE and no buffer head yet */
242 #define LOG_BLOCK_SIZE PAGE_SHIFT
244 #undef LOG_SECTORS_PER_BLOCK
245 #define LOG_SECTORS_PER_BLOCK (PAGE_SHIFT - 9)
247 #undef SECTORS_PER_BLOCK
248 #define SECTORS_PER_BLOCK (1 << SECTORS_PER_BLOCK)
250 #define EROFS_BLKSIZ (1 << LOG_BLOCK_SIZE)
252 #if (EROFS_BLKSIZ % 4096 || !EROFS_BLKSIZ)
253 #error erofs cannot be used in this platform
256 enum erofs_kmap_type {
257 EROFS_NO_KMAP, /* don't map the buffer */
258 EROFS_KMAP, /* use kmap_local_page() to map the buffer */
264 enum erofs_kmap_type kmap_type;
266 #define __EROFS_BUF_INITIALIZER ((struct erofs_buf){ .page = NULL })
268 #define ROOT_NID(sb) ((sb)->root_nid)
270 #define erofs_blknr(addr) ((addr) / EROFS_BLKSIZ)
271 #define erofs_blkoff(addr) ((addr) % EROFS_BLKSIZ)
272 #define blknr_to_addr(nr) ((erofs_off_t)(nr) * EROFS_BLKSIZ)
274 static inline erofs_off_t iloc(struct erofs_sb_info *sbi, erofs_nid_t nid)
276 return blknr_to_addr(sbi->meta_blkaddr) + (nid << sbi->islotbits);
279 #define EROFS_FEATURE_FUNCS(name, compat, feature) \
280 static inline bool erofs_sb_has_##name(struct erofs_sb_info *sbi) \
282 return sbi->feature_##compat & EROFS_FEATURE_##feature; \
285 EROFS_FEATURE_FUNCS(zero_padding, incompat, INCOMPAT_ZERO_PADDING)
286 EROFS_FEATURE_FUNCS(compr_cfgs, incompat, INCOMPAT_COMPR_CFGS)
287 EROFS_FEATURE_FUNCS(big_pcluster, incompat, INCOMPAT_BIG_PCLUSTER)
288 EROFS_FEATURE_FUNCS(chunked_file, incompat, INCOMPAT_CHUNKED_FILE)
289 EROFS_FEATURE_FUNCS(device_table, incompat, INCOMPAT_DEVICE_TABLE)
290 EROFS_FEATURE_FUNCS(compr_head2, incompat, INCOMPAT_COMPR_HEAD2)
291 EROFS_FEATURE_FUNCS(ztailpacking, incompat, INCOMPAT_ZTAILPACKING)
292 EROFS_FEATURE_FUNCS(fragments, incompat, INCOMPAT_FRAGMENTS)
293 EROFS_FEATURE_FUNCS(dedupe, incompat, INCOMPAT_DEDUPE)
294 EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM)
296 /* atomic flag definitions */
297 #define EROFS_I_EA_INITED_BIT 0
298 #define EROFS_I_Z_INITED_BIT 1
300 /* bitlock definitions (arranged in reverse order) */
301 #define EROFS_I_BL_XATTR_BIT (BITS_PER_LONG - 1)
302 #define EROFS_I_BL_Z_BIT (BITS_PER_LONG - 2)
307 /* atomic flags (including bitlocks) */
310 unsigned char datalayout;
311 unsigned char inode_isize;
312 unsigned short xattr_isize;
314 unsigned int xattr_shared_count;
315 unsigned int *xattr_shared_xattrs;
318 erofs_blk_t raw_blkaddr;
320 unsigned short chunkformat;
321 unsigned char chunkbits;
323 #ifdef CONFIG_EROFS_FS_ZIP
325 unsigned short z_advise;
326 unsigned char z_algorithmtype[2];
327 unsigned char z_logical_clusterbits;
328 unsigned long z_tailextent_headlcn;
331 erofs_off_t z_idataoff;
332 unsigned short z_idata_size;
334 erofs_off_t z_fragmentoff;
337 #endif /* CONFIG_EROFS_FS_ZIP */
339 /* the corresponding vfs inode */
340 struct inode vfs_inode;
343 #define EROFS_I(ptr) \
344 container_of(ptr, struct erofs_inode, vfs_inode)
346 static inline unsigned long erofs_inode_datablocks(struct inode *inode)
348 /* since i_size cannot be changed */
349 return DIV_ROUND_UP(inode->i_size, EROFS_BLKSIZ);
352 static inline unsigned int erofs_bitrange(unsigned int value, unsigned int bit,
356 return (value >> bit) & ((1 << bits) - 1);
360 static inline unsigned int erofs_inode_version(unsigned int value)
362 return erofs_bitrange(value, EROFS_I_VERSION_BIT,
363 EROFS_I_VERSION_BITS);
366 static inline unsigned int erofs_inode_datalayout(unsigned int value)
368 return erofs_bitrange(value, EROFS_I_DATALAYOUT_BIT,
369 EROFS_I_DATALAYOUT_BITS);
373 * Different from grab_cache_page_nowait(), reclaiming is never triggered
374 * when allocating new pages.
377 struct page *erofs_grab_cache_page_nowait(struct address_space *mapping,
380 return pagecache_get_page(mapping, index,
381 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
382 readahead_gfp_mask(mapping) & ~__GFP_RECLAIM);
385 extern const struct super_operations erofs_sops;
386 extern struct file_system_type erofs_fs_type;
388 extern const struct address_space_operations erofs_raw_access_aops;
389 extern const struct address_space_operations z_erofs_aops;
392 BH_Encoded = BH_PrivateStart,
398 /* Has a disk mapping */
399 #define EROFS_MAP_MAPPED (1 << BH_Mapped)
400 /* Located in metadata (could be copied from bd_inode) */
401 #define EROFS_MAP_META (1 << BH_Meta)
402 /* The extent is encoded */
403 #define EROFS_MAP_ENCODED (1 << BH_Encoded)
404 /* The length of extent is full */
405 #define EROFS_MAP_FULL_MAPPED (1 << BH_FullMapped)
406 /* Located in the special packed inode */
407 #define EROFS_MAP_FRAGMENT (1 << BH_Fragment)
408 /* The extent refers to partial decompressed data */
409 #define EROFS_MAP_PARTIAL_REF (1 << BH_Partialref)
411 struct erofs_map_blocks {
412 struct erofs_buf buf;
414 erofs_off_t m_pa, m_la;
417 unsigned short m_deviceid;
418 char m_algorithmformat;
419 unsigned int m_flags;
422 /* Flags used by erofs_map_blocks_flatmode() */
423 #define EROFS_GET_BLOCKS_RAW 0x0001
425 * Used to get the exact decompressed length, e.g. fiemap (consider lookback
426 * approach instead if possible since it's more metadata lightweight.)
428 #define EROFS_GET_BLOCKS_FIEMAP 0x0002
429 /* Used to map the whole extent if non-negligible data is requested for LZMA */
430 #define EROFS_GET_BLOCKS_READMORE 0x0004
431 /* Used to map tail extent for tailpacking inline or fragment pcluster */
432 #define EROFS_GET_BLOCKS_FINDTAIL 0x0008
435 Z_EROFS_COMPRESSION_SHIFTED = Z_EROFS_COMPRESSION_MAX,
436 Z_EROFS_COMPRESSION_INTERLACED,
437 Z_EROFS_COMPRESSION_RUNTIME_MAX
441 extern const struct iomap_ops z_erofs_iomap_report_ops;
443 #ifdef CONFIG_EROFS_FS_ZIP
444 int z_erofs_fill_inode(struct inode *inode);
445 int z_erofs_map_blocks_iter(struct inode *inode,
446 struct erofs_map_blocks *map,
449 static inline int z_erofs_fill_inode(struct inode *inode) { return -EOPNOTSUPP; }
450 static inline int z_erofs_map_blocks_iter(struct inode *inode,
451 struct erofs_map_blocks *map,
456 #endif /* !CONFIG_EROFS_FS_ZIP */
458 struct erofs_map_dev {
459 struct erofs_fscache *m_fscache;
460 struct block_device *m_bdev;
461 struct dax_device *m_daxdev;
465 unsigned int m_deviceid;
469 extern const struct file_operations erofs_file_fops;
470 void erofs_unmap_metabuf(struct erofs_buf *buf);
471 void erofs_put_metabuf(struct erofs_buf *buf);
472 void *erofs_bread(struct erofs_buf *buf, struct inode *inode,
473 erofs_blk_t blkaddr, enum erofs_kmap_type type);
474 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
475 erofs_blk_t blkaddr, enum erofs_kmap_type type);
476 int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev);
477 int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
479 int erofs_map_blocks(struct inode *inode,
480 struct erofs_map_blocks *map, int flags);
483 static inline unsigned long erofs_inode_hash(erofs_nid_t nid)
485 #if BITS_PER_LONG == 32
486 return (nid >> 32) ^ (nid & 0xffffffff);
492 extern const struct inode_operations erofs_generic_iops;
493 extern const struct inode_operations erofs_symlink_iops;
494 extern const struct inode_operations erofs_fast_symlink_iops;
496 struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid);
497 int erofs_getattr(struct user_namespace *mnt_userns, const struct path *path,
498 struct kstat *stat, u32 request_mask,
499 unsigned int query_flags);
502 extern const struct inode_operations erofs_dir_iops;
504 int erofs_namei(struct inode *dir, const struct qstr *name,
505 erofs_nid_t *nid, unsigned int *d_type);
508 extern const struct file_operations erofs_dir_fops;
510 static inline void *erofs_vm_map_ram(struct page **pages, unsigned int count)
515 void *p = vm_map_ram(pages, count, -1);
517 /* retry two more times (totally 3 times) */
518 if (p || ++retried >= 3)
526 void *erofs_get_pcpubuf(unsigned int requiredpages);
527 void erofs_put_pcpubuf(void *ptr);
528 int erofs_pcpubuf_growsize(unsigned int nrpages);
529 void erofs_pcpubuf_init(void);
530 void erofs_pcpubuf_exit(void);
533 int erofs_register_sysfs(struct super_block *sb);
534 void erofs_unregister_sysfs(struct super_block *sb);
535 int __init erofs_init_sysfs(void);
536 void erofs_exit_sysfs(void);
538 /* utils.c / zdata.c */
539 struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp);
540 static inline void erofs_pagepool_add(struct page **pagepool,
543 set_page_private(page, (unsigned long)*pagepool);
546 void erofs_release_pages(struct page **pagepool);
548 #ifdef CONFIG_EROFS_FS_ZIP
549 int erofs_workgroup_put(struct erofs_workgroup *grp);
550 struct erofs_workgroup *erofs_find_workgroup(struct super_block *sb,
552 struct erofs_workgroup *erofs_insert_workgroup(struct super_block *sb,
553 struct erofs_workgroup *grp);
554 void erofs_workgroup_free_rcu(struct erofs_workgroup *grp);
555 void erofs_shrinker_register(struct super_block *sb);
556 void erofs_shrinker_unregister(struct super_block *sb);
557 int __init erofs_init_shrinker(void);
558 void erofs_exit_shrinker(void);
559 int __init z_erofs_init_zip_subsystem(void);
560 void z_erofs_exit_zip_subsystem(void);
561 int erofs_try_to_free_all_cached_pages(struct erofs_sb_info *sbi,
562 struct erofs_workgroup *egrp);
563 int erofs_try_to_free_cached_page(struct page *page);
564 int z_erofs_load_lz4_config(struct super_block *sb,
565 struct erofs_super_block *dsb,
566 struct z_erofs_lz4_cfgs *lz4, int len);
568 static inline void erofs_shrinker_register(struct super_block *sb) {}
569 static inline void erofs_shrinker_unregister(struct super_block *sb) {}
570 static inline int erofs_init_shrinker(void) { return 0; }
571 static inline void erofs_exit_shrinker(void) {}
572 static inline int z_erofs_init_zip_subsystem(void) { return 0; }
573 static inline void z_erofs_exit_zip_subsystem(void) {}
574 static inline int z_erofs_load_lz4_config(struct super_block *sb,
575 struct erofs_super_block *dsb,
576 struct z_erofs_lz4_cfgs *lz4, int len)
578 if (lz4 || dsb->u1.lz4_max_distance) {
579 erofs_err(sb, "lz4 algorithm isn't enabled");
584 #endif /* !CONFIG_EROFS_FS_ZIP */
586 #ifdef CONFIG_EROFS_FS_ZIP_LZMA
587 int z_erofs_lzma_init(void);
588 void z_erofs_lzma_exit(void);
589 int z_erofs_load_lzma_config(struct super_block *sb,
590 struct erofs_super_block *dsb,
591 struct z_erofs_lzma_cfgs *lzma, int size);
593 static inline int z_erofs_lzma_init(void) { return 0; }
594 static inline int z_erofs_lzma_exit(void) { return 0; }
595 static inline int z_erofs_load_lzma_config(struct super_block *sb,
596 struct erofs_super_block *dsb,
597 struct z_erofs_lzma_cfgs *lzma, int size) {
599 erofs_err(sb, "lzma algorithm isn't enabled");
604 #endif /* !CONFIG_EROFS_FS_ZIP */
606 /* flags for erofs_fscache_register_cookie() */
607 #define EROFS_REG_COOKIE_NEED_INODE 1
608 #define EROFS_REG_COOKIE_NEED_NOEXIST 2
611 #ifdef CONFIG_EROFS_FS_ONDEMAND
612 int erofs_fscache_register_fs(struct super_block *sb);
613 void erofs_fscache_unregister_fs(struct super_block *sb);
615 struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
618 void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache);
620 extern const struct address_space_operations erofs_fscache_access_aops;
622 static inline int erofs_fscache_register_fs(struct super_block *sb)
626 static inline void erofs_fscache_unregister_fs(struct super_block *sb) {}
629 struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
633 return ERR_PTR(-EOPNOTSUPP);
636 static inline void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache)
641 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */
643 #endif /* __EROFS_INTERNAL_H */