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/magic.h>
16 #include <linux/slab.h>
17 #include <linux/vmalloc.h>
18 #include <linux/iomap.h>
21 /* redefine pr_fmt "erofs: " */
23 #define pr_fmt(fmt) "erofs: " fmt
25 __printf(3, 4) void _erofs_err(struct super_block *sb,
26 const char *function, const char *fmt, ...);
27 #define erofs_err(sb, fmt, ...) \
28 _erofs_err(sb, __func__, fmt "\n", ##__VA_ARGS__)
29 __printf(3, 4) void _erofs_info(struct super_block *sb,
30 const char *function, const char *fmt, ...);
31 #define erofs_info(sb, fmt, ...) \
32 _erofs_info(sb, __func__, fmt "\n", ##__VA_ARGS__)
33 #ifdef CONFIG_EROFS_FS_DEBUG
34 #define erofs_dbg(x, ...) pr_debug(x "\n", ##__VA_ARGS__)
35 #define DBG_BUGON BUG_ON
37 #define erofs_dbg(x, ...) ((void)0)
38 #define DBG_BUGON(x) ((void)(x))
39 #endif /* !CONFIG_EROFS_FS_DEBUG */
41 /* EROFS_SUPER_MAGIC_V1 to represent the whole file system */
42 #define EROFS_SUPER_MAGIC EROFS_SUPER_MAGIC_V1
44 typedef u64 erofs_nid_t;
45 typedef u64 erofs_off_t;
46 /* data type for filesystem-wide blocks number */
47 typedef u32 erofs_blk_t;
49 struct erofs_device_info {
51 struct erofs_fscache *fscache;
52 struct block_device *bdev;
53 struct dax_device *dax_dev;
61 EROFS_SYNC_DECOMPRESS_AUTO,
62 EROFS_SYNC_DECOMPRESS_FORCE_ON,
63 EROFS_SYNC_DECOMPRESS_FORCE_OFF
66 struct erofs_mount_opts {
67 #ifdef CONFIG_EROFS_FS_ZIP
68 /* current strategy of how to use managed cache */
69 unsigned char cache_strategy;
70 /* strategy of sync decompression (0 - auto, 1 - force on, 2 - force off) */
71 unsigned int sync_decompress;
73 /* threshold for decompression synchronously */
74 unsigned int max_sync_decompress_pages;
76 unsigned int mount_opt;
79 struct erofs_dev_context {
81 struct rw_semaphore rwsem;
83 unsigned int extra_devices;
86 struct erofs_fs_context {
87 struct erofs_mount_opts opt;
88 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_domain {
103 struct list_head list;
104 struct fscache_volume *volume;
108 struct erofs_fscache {
109 struct fscache_cookie *cookie;
111 struct inode *anon_inode;
112 struct erofs_domain *domain;
116 struct erofs_sb_info {
117 struct erofs_mount_opts opt; /* options */
118 #ifdef CONFIG_EROFS_FS_ZIP
119 /* list for all registered superblocks, mainly for shrinker */
120 struct list_head list;
121 struct mutex umount_mutex;
123 /* managed XArray arranged in physical block number */
124 struct xarray managed_pslots;
126 unsigned int shrinker_run_no;
127 u16 available_compr_algs;
129 /* pseudo inode to manage cached pages */
130 struct inode *managed_cache;
132 struct erofs_sb_lz4_info lz4;
133 struct inode *packed_inode;
134 #endif /* CONFIG_EROFS_FS_ZIP */
135 struct erofs_dev_context *devs;
136 struct dax_device *dax_dev;
139 u32 primarydevice_blocks;
142 #ifdef CONFIG_EROFS_FS_XATTR
145 u16 device_id_mask; /* valid bits of device id to be used */
147 /* inode slot unit size in bit shift */
148 unsigned char islotbits;
150 u32 sb_size; /* total superblock size */
154 /* what we really care is nid, rather than ino.. */
155 erofs_nid_t root_nid;
156 /* used for statfs, f_files - f_favail */
159 u8 uuid[16]; /* 128-bit uuid for volume */
160 u8 volume_name[16]; /* volume name */
162 u32 feature_incompat;
165 struct kobject s_kobj; /* /sys/fs/erofs/<devname> */
166 struct completion s_kobj_unregister;
168 /* fscache support */
169 struct fscache_volume *volume;
170 struct erofs_fscache *s_fscache;
171 struct erofs_domain *domain;
176 #define EROFS_SB(sb) ((struct erofs_sb_info *)(sb)->s_fs_info)
177 #define EROFS_I_SB(inode) ((struct erofs_sb_info *)(inode)->i_sb->s_fs_info)
179 /* Mount flags set via mount options or defaults */
180 #define EROFS_MOUNT_XATTR_USER 0x00000010
181 #define EROFS_MOUNT_POSIX_ACL 0x00000020
182 #define EROFS_MOUNT_DAX_ALWAYS 0x00000040
183 #define EROFS_MOUNT_DAX_NEVER 0x00000080
185 #define clear_opt(opt, option) ((opt)->mount_opt &= ~EROFS_MOUNT_##option)
186 #define set_opt(opt, option) ((opt)->mount_opt |= EROFS_MOUNT_##option)
187 #define test_opt(opt, option) ((opt)->mount_opt & EROFS_MOUNT_##option)
189 static inline bool erofs_is_fscache_mode(struct super_block *sb)
191 return IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && !sb->s_bdev;
195 EROFS_ZIP_CACHE_DISABLED,
196 EROFS_ZIP_CACHE_READAHEAD,
197 EROFS_ZIP_CACHE_READAROUND
200 #define EROFS_LOCKED_MAGIC (INT_MIN | 0xE0F510CCL)
202 /* basic unit of the workstation of a super_block */
203 struct erofs_workgroup {
204 /* the workgroup index in the workstation */
207 /* overall workgroup reference count */
211 static inline bool erofs_workgroup_try_to_freeze(struct erofs_workgroup *grp,
215 if (val != atomic_cmpxchg(&grp->refcount, val, EROFS_LOCKED_MAGIC)) {
222 static inline void erofs_workgroup_unfreeze(struct erofs_workgroup *grp,
226 * other observers should notice all modifications
227 * in the freezing period.
230 atomic_set(&grp->refcount, orig_val);
234 static inline int erofs_wait_on_workgroup_freezed(struct erofs_workgroup *grp)
236 return atomic_cond_read_relaxed(&grp->refcount,
237 VAL != EROFS_LOCKED_MAGIC);
240 /* we strictly follow PAGE_SIZE and no buffer head yet */
241 #define LOG_BLOCK_SIZE PAGE_SHIFT
243 #undef LOG_SECTORS_PER_BLOCK
244 #define LOG_SECTORS_PER_BLOCK (PAGE_SHIFT - 9)
246 #undef SECTORS_PER_BLOCK
247 #define SECTORS_PER_BLOCK (1 << SECTORS_PER_BLOCK)
249 #define EROFS_BLKSIZ (1 << LOG_BLOCK_SIZE)
251 #if (EROFS_BLKSIZ % 4096 || !EROFS_BLKSIZ)
252 #error erofs cannot be used in this platform
255 enum erofs_kmap_type {
256 EROFS_NO_KMAP, /* don't map the buffer */
257 EROFS_KMAP, /* use kmap_local_page() 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 #define EROFS_FEATURE_FUNCS(name, compat, feature) \
274 static inline bool erofs_sb_has_##name(struct erofs_sb_info *sbi) \
276 return sbi->feature_##compat & EROFS_FEATURE_##feature; \
279 EROFS_FEATURE_FUNCS(zero_padding, incompat, INCOMPAT_ZERO_PADDING)
280 EROFS_FEATURE_FUNCS(compr_cfgs, incompat, INCOMPAT_COMPR_CFGS)
281 EROFS_FEATURE_FUNCS(big_pcluster, incompat, INCOMPAT_BIG_PCLUSTER)
282 EROFS_FEATURE_FUNCS(chunked_file, incompat, INCOMPAT_CHUNKED_FILE)
283 EROFS_FEATURE_FUNCS(device_table, incompat, INCOMPAT_DEVICE_TABLE)
284 EROFS_FEATURE_FUNCS(compr_head2, incompat, INCOMPAT_COMPR_HEAD2)
285 EROFS_FEATURE_FUNCS(ztailpacking, incompat, INCOMPAT_ZTAILPACKING)
286 EROFS_FEATURE_FUNCS(fragments, incompat, INCOMPAT_FRAGMENTS)
287 EROFS_FEATURE_FUNCS(dedupe, incompat, INCOMPAT_DEDUPE)
288 EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM)
290 /* atomic flag definitions */
291 #define EROFS_I_EA_INITED_BIT 0
292 #define EROFS_I_Z_INITED_BIT 1
294 /* bitlock definitions (arranged in reverse order) */
295 #define EROFS_I_BL_XATTR_BIT (BITS_PER_LONG - 1)
296 #define EROFS_I_BL_Z_BIT (BITS_PER_LONG - 2)
301 /* atomic flags (including bitlocks) */
304 unsigned char datalayout;
305 unsigned char inode_isize;
306 unsigned short xattr_isize;
308 unsigned int xattr_shared_count;
309 unsigned int *xattr_shared_xattrs;
312 erofs_blk_t raw_blkaddr;
314 unsigned short chunkformat;
315 unsigned char chunkbits;
317 #ifdef CONFIG_EROFS_FS_ZIP
319 unsigned short z_advise;
320 unsigned char z_algorithmtype[2];
321 unsigned char z_logical_clusterbits;
322 unsigned long z_tailextent_headlcn;
325 erofs_off_t z_idataoff;
326 unsigned short z_idata_size;
328 erofs_off_t z_fragmentoff;
331 #endif /* CONFIG_EROFS_FS_ZIP */
333 /* the corresponding vfs inode */
334 struct inode vfs_inode;
337 #define EROFS_I(ptr) container_of(ptr, struct erofs_inode, vfs_inode)
339 static inline erofs_off_t erofs_iloc(struct inode *inode)
341 struct erofs_sb_info *sbi = EROFS_I_SB(inode);
343 return blknr_to_addr(sbi->meta_blkaddr) +
344 (EROFS_I(inode)->nid << sbi->islotbits);
347 static inline unsigned long erofs_inode_datablocks(struct inode *inode)
349 /* since i_size cannot be changed */
350 return DIV_ROUND_UP(inode->i_size, EROFS_BLKSIZ);
353 static inline unsigned int erofs_bitrange(unsigned int value, unsigned int bit,
357 return (value >> bit) & ((1 << bits) - 1);
361 static inline unsigned int erofs_inode_version(unsigned int value)
363 return erofs_bitrange(value, EROFS_I_VERSION_BIT,
364 EROFS_I_VERSION_BITS);
367 static inline unsigned int erofs_inode_datalayout(unsigned int value)
369 return erofs_bitrange(value, EROFS_I_DATALAYOUT_BIT,
370 EROFS_I_DATALAYOUT_BITS);
374 * Different from grab_cache_page_nowait(), reclaiming is never triggered
375 * when allocating new pages.
378 struct page *erofs_grab_cache_page_nowait(struct address_space *mapping,
381 return pagecache_get_page(mapping, index,
382 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
383 readahead_gfp_mask(mapping) & ~__GFP_RECLAIM);
386 extern const struct super_operations erofs_sops;
387 extern struct file_system_type erofs_fs_type;
389 extern const struct address_space_operations erofs_raw_access_aops;
390 extern const struct address_space_operations z_erofs_aops;
392 /* Has a disk mapping */
393 #define EROFS_MAP_MAPPED 0x0001
394 /* Located in metadata (could be copied from bd_inode) */
395 #define EROFS_MAP_META 0x0002
396 /* The extent is encoded */
397 #define EROFS_MAP_ENCODED 0x0004
398 /* The length of extent is full */
399 #define EROFS_MAP_FULL_MAPPED 0x0008
400 /* Located in the special packed inode */
401 #define EROFS_MAP_FRAGMENT 0x0010
402 /* The extent refers to partial decompressed data */
403 #define EROFS_MAP_PARTIAL_REF 0x0020
405 struct erofs_map_blocks {
406 struct erofs_buf buf;
408 erofs_off_t m_pa, m_la;
411 unsigned short m_deviceid;
412 char m_algorithmformat;
413 unsigned int m_flags;
416 /* Flags used by erofs_map_blocks_flatmode() */
417 #define EROFS_GET_BLOCKS_RAW 0x0001
419 * Used to get the exact decompressed length, e.g. fiemap (consider lookback
420 * approach instead if possible since it's more metadata lightweight.)
422 #define EROFS_GET_BLOCKS_FIEMAP 0x0002
423 /* Used to map the whole extent if non-negligible data is requested for LZMA */
424 #define EROFS_GET_BLOCKS_READMORE 0x0004
425 /* Used to map tail extent for tailpacking inline or fragment pcluster */
426 #define EROFS_GET_BLOCKS_FINDTAIL 0x0008
429 Z_EROFS_COMPRESSION_SHIFTED = Z_EROFS_COMPRESSION_MAX,
430 Z_EROFS_COMPRESSION_INTERLACED,
431 Z_EROFS_COMPRESSION_RUNTIME_MAX
435 extern const struct iomap_ops z_erofs_iomap_report_ops;
437 #ifdef CONFIG_EROFS_FS_ZIP
438 int z_erofs_fill_inode(struct inode *inode);
439 int z_erofs_map_blocks_iter(struct inode *inode,
440 struct erofs_map_blocks *map,
443 static inline int z_erofs_fill_inode(struct inode *inode) { return -EOPNOTSUPP; }
444 static inline int z_erofs_map_blocks_iter(struct inode *inode,
445 struct erofs_map_blocks *map,
450 #endif /* !CONFIG_EROFS_FS_ZIP */
452 struct erofs_map_dev {
453 struct erofs_fscache *m_fscache;
454 struct block_device *m_bdev;
455 struct dax_device *m_daxdev;
459 unsigned int m_deviceid;
463 extern const struct file_operations erofs_file_fops;
464 void erofs_unmap_metabuf(struct erofs_buf *buf);
465 void erofs_put_metabuf(struct erofs_buf *buf);
466 void *erofs_bread(struct erofs_buf *buf, struct inode *inode,
467 erofs_blk_t blkaddr, enum erofs_kmap_type type);
468 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
469 erofs_blk_t blkaddr, enum erofs_kmap_type type);
470 int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev);
471 int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
473 int erofs_map_blocks(struct inode *inode,
474 struct erofs_map_blocks *map, int flags);
477 extern const struct inode_operations erofs_generic_iops;
478 extern const struct inode_operations erofs_symlink_iops;
479 extern const struct inode_operations erofs_fast_symlink_iops;
481 struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid);
482 int erofs_getattr(struct user_namespace *mnt_userns, const struct path *path,
483 struct kstat *stat, u32 request_mask,
484 unsigned int query_flags);
487 extern const struct inode_operations erofs_dir_iops;
489 int erofs_namei(struct inode *dir, const struct qstr *name,
490 erofs_nid_t *nid, unsigned int *d_type);
493 extern const struct file_operations erofs_dir_fops;
495 static inline void *erofs_vm_map_ram(struct page **pages, unsigned int count)
500 void *p = vm_map_ram(pages, count, -1);
502 /* retry two more times (totally 3 times) */
503 if (p || ++retried >= 3)
511 void *erofs_get_pcpubuf(unsigned int requiredpages);
512 void erofs_put_pcpubuf(void *ptr);
513 int erofs_pcpubuf_growsize(unsigned int nrpages);
514 void erofs_pcpubuf_init(void);
515 void erofs_pcpubuf_exit(void);
518 int erofs_register_sysfs(struct super_block *sb);
519 void erofs_unregister_sysfs(struct super_block *sb);
520 int __init erofs_init_sysfs(void);
521 void erofs_exit_sysfs(void);
523 /* utils.c / zdata.c */
524 struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp);
525 static inline void erofs_pagepool_add(struct page **pagepool,
528 set_page_private(page, (unsigned long)*pagepool);
531 void erofs_release_pages(struct page **pagepool);
533 #ifdef CONFIG_EROFS_FS_ZIP
534 int erofs_workgroup_put(struct erofs_workgroup *grp);
535 struct erofs_workgroup *erofs_find_workgroup(struct super_block *sb,
537 struct erofs_workgroup *erofs_insert_workgroup(struct super_block *sb,
538 struct erofs_workgroup *grp);
539 void erofs_workgroup_free_rcu(struct erofs_workgroup *grp);
540 void erofs_shrinker_register(struct super_block *sb);
541 void erofs_shrinker_unregister(struct super_block *sb);
542 int __init erofs_init_shrinker(void);
543 void erofs_exit_shrinker(void);
544 int __init z_erofs_init_zip_subsystem(void);
545 void z_erofs_exit_zip_subsystem(void);
546 int erofs_try_to_free_all_cached_pages(struct erofs_sb_info *sbi,
547 struct erofs_workgroup *egrp);
548 int erofs_try_to_free_cached_page(struct page *page);
549 int z_erofs_load_lz4_config(struct super_block *sb,
550 struct erofs_super_block *dsb,
551 struct z_erofs_lz4_cfgs *lz4, int len);
553 static inline void erofs_shrinker_register(struct super_block *sb) {}
554 static inline void erofs_shrinker_unregister(struct super_block *sb) {}
555 static inline int erofs_init_shrinker(void) { return 0; }
556 static inline void erofs_exit_shrinker(void) {}
557 static inline int z_erofs_init_zip_subsystem(void) { return 0; }
558 static inline void z_erofs_exit_zip_subsystem(void) {}
559 static inline int z_erofs_load_lz4_config(struct super_block *sb,
560 struct erofs_super_block *dsb,
561 struct z_erofs_lz4_cfgs *lz4, int len)
563 if (lz4 || dsb->u1.lz4_max_distance) {
564 erofs_err(sb, "lz4 algorithm isn't enabled");
569 #endif /* !CONFIG_EROFS_FS_ZIP */
571 #ifdef CONFIG_EROFS_FS_ZIP_LZMA
572 int z_erofs_lzma_init(void);
573 void z_erofs_lzma_exit(void);
574 int z_erofs_load_lzma_config(struct super_block *sb,
575 struct erofs_super_block *dsb,
576 struct z_erofs_lzma_cfgs *lzma, int size);
578 static inline int z_erofs_lzma_init(void) { return 0; }
579 static inline int z_erofs_lzma_exit(void) { return 0; }
580 static inline int z_erofs_load_lzma_config(struct super_block *sb,
581 struct erofs_super_block *dsb,
582 struct z_erofs_lzma_cfgs *lzma, int size) {
584 erofs_err(sb, "lzma algorithm isn't enabled");
589 #endif /* !CONFIG_EROFS_FS_ZIP */
591 /* flags for erofs_fscache_register_cookie() */
592 #define EROFS_REG_COOKIE_NEED_INODE 1
593 #define EROFS_REG_COOKIE_NEED_NOEXIST 2
596 #ifdef CONFIG_EROFS_FS_ONDEMAND
597 int erofs_fscache_register_fs(struct super_block *sb);
598 void erofs_fscache_unregister_fs(struct super_block *sb);
600 struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
603 void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache);
605 extern const struct address_space_operations erofs_fscache_access_aops;
607 static inline int erofs_fscache_register_fs(struct super_block *sb)
611 static inline void erofs_fscache_unregister_fs(struct super_block *sb) {}
614 struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
618 return ERR_PTR(-EOPNOTSUPP);
621 static inline void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache)
626 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */
628 #endif /* __EROFS_INTERNAL_H */