1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /*
3 * Copyright (C) 2017-2018 HUAWEI, Inc.
4 * https://www.huawei.com/
5 * Copyright (C) 2021, Alibaba Cloud
6 */
7 #ifndef __EROFS_INTERNAL_H
8 #define __EROFS_INTERNAL_H
9
10 #include <linux/fs.h>
11 #include <linux/dax.h>
12 #include <linux/dcache.h>
13 #include <linux/mm.h>
14 #include <linux/module.h>
15 #include <linux/pagemap.h>
16 #include <linux/bio.h>
17 #include <linux/magic.h>
18 #include <linux/slab.h>
19 #include <linux/vmalloc.h>
20 #include <linux/iomap.h>
21 #include "erofs_fs.h"
22
23 /* redefine pr_fmt "erofs: " */
24 #undef pr_fmt
25 #define pr_fmt(fmt) "erofs: " fmt
26
27 __printf(3, 4) void _erofs_err(struct super_block *sb,
28 const char *function, const char *fmt, ...);
29 #define erofs_err(sb, fmt, ...) \
30 _erofs_err(sb, __func__, fmt "\n", ##__VA_ARGS__)
31 __printf(3, 4) void _erofs_info(struct super_block *sb,
32 const char *function, const char *fmt, ...);
33 #define erofs_info(sb, fmt, ...) \
34 _erofs_info(sb, __func__, fmt "\n", ##__VA_ARGS__)
35 #ifdef CONFIG_EROFS_FS_DEBUG
36 #define DBG_BUGON BUG_ON
37 #else
38 #define DBG_BUGON(x) ((void)(x))
39 #endif /* !CONFIG_EROFS_FS_DEBUG */
40
41 /* EROFS_SUPER_MAGIC_V1 to represent the whole file system */
42 #define EROFS_SUPER_MAGIC EROFS_SUPER_MAGIC_V1
43
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;
48
49 struct erofs_device_info {
50 char *path;
51 struct erofs_fscache *fscache;
52 struct file *bdev_file;
53 struct dax_device *dax_dev;
54 u64 dax_part_off;
55
56 u32 blocks;
57 u32 mapped_blkaddr;
58 };
59
60 enum {
61 EROFS_SYNC_DECOMPRESS_AUTO,
62 EROFS_SYNC_DECOMPRESS_FORCE_ON,
63 EROFS_SYNC_DECOMPRESS_FORCE_OFF
64 };
65
66 struct erofs_mount_opts {
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;
71 /* threshold for decompression synchronously */
72 unsigned int max_sync_decompress_pages;
73 unsigned int mount_opt;
74 };
75
76 struct erofs_dev_context {
77 struct idr tree;
78 struct rw_semaphore rwsem;
79
80 unsigned int extra_devices;
81 bool flatdev;
82 };
83
84 /* all filesystem-wide lz4 configurations */
85 struct erofs_sb_lz4_info {
86 /* # of pages needed for EROFS lz4 rolling decompression */
87 u16 max_distance_pages;
88 /* maximum possible blocks for pclusters in the filesystem */
89 u16 max_pclusterblks;
90 };
91
92 struct erofs_domain {
93 refcount_t ref;
94 struct list_head list;
95 struct fscache_volume *volume;
96 char *domain_id;
97 };
98
99 struct erofs_fscache {
100 struct fscache_cookie *cookie;
101 struct inode *inode; /* anonymous inode for the blob */
102
103 /* used for share domain mode */
104 struct erofs_domain *domain;
105 struct list_head node;
106 refcount_t ref;
107 char *name;
108 };
109
110 struct erofs_xattr_prefix_item {
111 struct erofs_xattr_long_prefix *prefix;
112 u8 infix_len;
113 };
114
115 struct erofs_sb_info {
116 struct erofs_mount_opts opt; /* options */
117 #ifdef CONFIG_EROFS_FS_ZIP
118 /* list for all registered superblocks, mainly for shrinker */
119 struct list_head list;
120 struct mutex umount_mutex;
121
122 /* managed XArray arranged in physical block number */
123 struct xarray managed_pslots;
124
125 unsigned int shrinker_run_no;
126 u16 available_compr_algs;
127
128 /* pseudo inode to manage cached pages */
129 struct inode *managed_cache;
130
131 struct erofs_sb_lz4_info lz4;
132 #endif /* CONFIG_EROFS_FS_ZIP */
133 struct inode *packed_inode;
134 struct erofs_dev_context *devs;
135 struct dax_device *dax_dev;
136 u64 dax_part_off;
137 u64 total_blocks;
138 u32 primarydevice_blocks;
139
140 u32 meta_blkaddr;
141 #ifdef CONFIG_EROFS_FS_XATTR
142 u32 xattr_blkaddr;
143 u32 xattr_prefix_start;
144 u8 xattr_prefix_count;
145 struct erofs_xattr_prefix_item *xattr_prefixes;
146 unsigned int xattr_filter_reserved;
147 #endif
148 u16 device_id_mask; /* valid bits of device id to be used */
149
150 unsigned char islotbits; /* inode slot unit size in bit shift */
151 unsigned char blkszbits; /* filesystem block size in bit shift */
152
153 u32 sb_size; /* total superblock size */
154 u32 build_time_nsec;
155 u64 build_time;
156
157 /* what we really care is nid, rather than ino.. */
158 erofs_nid_t root_nid;
159 erofs_nid_t packed_nid;
160 /* used for statfs, f_files - f_favail */
161 u64 inos;
162
163 u8 uuid[16]; /* 128-bit uuid for volume */
164 u8 volume_name[16]; /* volume name */
165 u32 feature_compat;
166 u32 feature_incompat;
167
168 /* sysfs support */
169 struct kobject s_kobj; /* /sys/fs/erofs/<devname> */
170 struct completion s_kobj_unregister;
171
172 /* fscache support */
173 struct fscache_volume *volume;
174 struct erofs_fscache *s_fscache;
175 struct erofs_domain *domain;
176 char *fsid;
177 char *domain_id;
178 };
179
180 #define EROFS_SB(sb) ((struct erofs_sb_info *)(sb)->s_fs_info)
181 #define EROFS_I_SB(inode) ((struct erofs_sb_info *)(inode)->i_sb->s_fs_info)
182
183 /* Mount flags set via mount options or defaults */
184 #define EROFS_MOUNT_XATTR_USER 0x00000010
185 #define EROFS_MOUNT_POSIX_ACL 0x00000020
186 #define EROFS_MOUNT_DAX_ALWAYS 0x00000040
187 #define EROFS_MOUNT_DAX_NEVER 0x00000080
188
189 #define clear_opt(opt, option) ((opt)->mount_opt &= ~EROFS_MOUNT_##option)
190 #define set_opt(opt, option) ((opt)->mount_opt |= EROFS_MOUNT_##option)
191 #define test_opt(opt, option) ((opt)->mount_opt & EROFS_MOUNT_##option)
192
193 static inline bool erofs_is_fscache_mode(struct super_block *sb)
194 {
195 return IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && !sb->s_bdev;
196 }
197
198 enum {
199 EROFS_ZIP_CACHE_DISABLED,
200 EROFS_ZIP_CACHE_READAHEAD,
201 EROFS_ZIP_CACHE_READAROUND
202 };
203
204 /* basic unit of the workstation of a super_block */
205 struct erofs_workgroup {
206 pgoff_t index;
207 struct lockref lockref;
208 };
209
210 enum erofs_kmap_type {
211 EROFS_NO_KMAP, /* don't map the buffer */
212 EROFS_KMAP, /* use kmap_local_page() to map the buffer */
213 };
214
215 struct erofs_buf {
216 struct address_space *mapping;
217 struct page *page;
218 void *base;
219 enum erofs_kmap_type kmap_type;
220 };
221 #define __EROFS_BUF_INITIALIZER ((struct erofs_buf){ .page = NULL })
222
223 #define erofs_blknr(sb, addr) ((erofs_blk_t)((addr) >> (sb)->s_blocksize_bits))
224 #define erofs_blkoff(sb, addr) ((addr) & ((sb)->s_blocksize - 1))
225 #define erofs_pos(sb, blk) ((erofs_off_t)(blk) << (sb)->s_blocksize_bits)
226 #define erofs_iblks(i) (round_up((i)->i_size, i_blocksize(i)) >> (i)->i_blkbits)
227
228 #define EROFS_FEATURE_FUNCS(name, compat, feature) \
229 static inline bool erofs_sb_has_##name(struct erofs_sb_info *sbi) \
230 { \
231 return sbi->feature_##compat & EROFS_FEATURE_##feature; \
232 }
233
234 EROFS_FEATURE_FUNCS(zero_padding, incompat, INCOMPAT_ZERO_PADDING)
235 EROFS_FEATURE_FUNCS(compr_cfgs, incompat, INCOMPAT_COMPR_CFGS)
236 EROFS_FEATURE_FUNCS(big_pcluster, incompat, INCOMPAT_BIG_PCLUSTER)
237 EROFS_FEATURE_FUNCS(chunked_file, incompat, INCOMPAT_CHUNKED_FILE)
238 EROFS_FEATURE_FUNCS(device_table, incompat, INCOMPAT_DEVICE_TABLE)
239 EROFS_FEATURE_FUNCS(compr_head2, incompat, INCOMPAT_COMPR_HEAD2)
240 EROFS_FEATURE_FUNCS(ztailpacking, incompat, INCOMPAT_ZTAILPACKING)
241 EROFS_FEATURE_FUNCS(fragments, incompat, INCOMPAT_FRAGMENTS)
242 EROFS_FEATURE_FUNCS(dedupe, incompat, INCOMPAT_DEDUPE)
243 EROFS_FEATURE_FUNCS(xattr_prefixes, incompat, INCOMPAT_XATTR_PREFIXES)
244 EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM)
245 EROFS_FEATURE_FUNCS(xattr_filter, compat, COMPAT_XATTR_FILTER)
246
247 /* atomic flag definitions */
248 #define EROFS_I_EA_INITED_BIT 0
249 #define EROFS_I_Z_INITED_BIT 1
250
251 /* bitlock definitions (arranged in reverse order) */
252 #define EROFS_I_BL_XATTR_BIT (BITS_PER_LONG - 1)
253 #define EROFS_I_BL_Z_BIT (BITS_PER_LONG - 2)
254
255 struct erofs_inode {
256 erofs_nid_t nid;
257
258 /* atomic flags (including bitlocks) */
259 unsigned long flags;
260
261 unsigned char datalayout;
262 unsigned char inode_isize;
263 unsigned int xattr_isize;
264
265 unsigned int xattr_name_filter;
266 unsigned int xattr_shared_count;
267 unsigned int *xattr_shared_xattrs;
268
269 union {
270 erofs_blk_t raw_blkaddr;
271 struct {
272 unsigned short chunkformat;
273 unsigned char chunkbits;
274 };
275 #ifdef CONFIG_EROFS_FS_ZIP
276 struct {
277 unsigned short z_advise;
278 unsigned char z_algorithmtype[2];
279 unsigned char z_logical_clusterbits;
280 unsigned long z_tailextent_headlcn;
281 union {
282 struct {
283 erofs_off_t z_idataoff;
284 unsigned short z_idata_size;
285 };
286 erofs_off_t z_fragmentoff;
287 };
288 };
289 #endif /* CONFIG_EROFS_FS_ZIP */
290 };
291 /* the corresponding vfs inode */
292 struct inode vfs_inode;
293 };
294
295 #define EROFS_I(ptr) container_of(ptr, struct erofs_inode, vfs_inode)
296
297 static inline erofs_off_t erofs_iloc(struct inode *inode)
298 {
299 struct erofs_sb_info *sbi = EROFS_I_SB(inode);
300
301 return erofs_pos(inode->i_sb, sbi->meta_blkaddr) +
302 (EROFS_I(inode)->nid << sbi->islotbits);
303 }
304
305 static inline unsigned int erofs_inode_version(unsigned int ifmt)
306 {
307 return (ifmt >> EROFS_I_VERSION_BIT) & EROFS_I_VERSION_MASK;
308 }
309
310 static inline unsigned int erofs_inode_datalayout(unsigned int ifmt)
311 {
312 return (ifmt >> EROFS_I_DATALAYOUT_BIT) & EROFS_I_DATALAYOUT_MASK;
313 }
314
315 /* reclaiming is never triggered when allocating new folios. */
316 static inline struct folio *erofs_grab_folio_nowait(struct address_space *as,
317 pgoff_t index)
318 {
319 return __filemap_get_folio(as, index,
320 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
321 readahead_gfp_mask(as) & ~__GFP_RECLAIM);
322 }
323
324 /* Has a disk mapping */
325 #define EROFS_MAP_MAPPED 0x0001
326 /* Located in metadata (could be copied from bd_inode) */
327 #define EROFS_MAP_META 0x0002
328 /* The extent is encoded */
329 #define EROFS_MAP_ENCODED 0x0004
330 /* The length of extent is full */
331 #define EROFS_MAP_FULL_MAPPED 0x0008
332 /* Located in the special packed inode */
333 #define EROFS_MAP_FRAGMENT 0x0010
334 /* The extent refers to partial decompressed data */
335 #define EROFS_MAP_PARTIAL_REF 0x0020
336
337 struct erofs_map_blocks {
338 struct erofs_buf buf;
339
340 erofs_off_t m_pa, m_la;
341 u64 m_plen, m_llen;
342
343 unsigned short m_deviceid;
344 char m_algorithmformat;
345 unsigned int m_flags;
346 };
347
348 /*
349 * Used to get the exact decompressed length, e.g. fiemap (consider lookback
350 * approach instead if possible since it's more metadata lightweight.)
351 */
352 #define EROFS_GET_BLOCKS_FIEMAP 0x0001
353 /* Used to map the whole extent if non-negligible data is requested for LZMA */
354 #define EROFS_GET_BLOCKS_READMORE 0x0002
355 /* Used to map tail extent for tailpacking inline or fragment pcluster */
356 #define EROFS_GET_BLOCKS_FINDTAIL 0x0004
357
358 enum {
359 Z_EROFS_COMPRESSION_SHIFTED = Z_EROFS_COMPRESSION_MAX,
360 Z_EROFS_COMPRESSION_INTERLACED,
361 Z_EROFS_COMPRESSION_RUNTIME_MAX
362 };
363
364 struct erofs_map_dev {
365 struct erofs_fscache *m_fscache;
366 struct block_device *m_bdev;
367 struct dax_device *m_daxdev;
368 u64 m_dax_part_off;
369
370 erofs_off_t m_pa;
371 unsigned int m_deviceid;
372 };
373
374 extern const struct super_operations erofs_sops;
375
376 extern const struct address_space_operations erofs_raw_access_aops;
377 extern const struct address_space_operations z_erofs_aops;
378 extern const struct address_space_operations erofs_fscache_access_aops;
379
380 extern const struct inode_operations erofs_generic_iops;
381 extern const struct inode_operations erofs_symlink_iops;
382 extern const struct inode_operations erofs_fast_symlink_iops;
383 extern const struct inode_operations erofs_dir_iops;
384
385 extern const struct file_operations erofs_file_fops;
386 extern const struct file_operations erofs_dir_fops;
387
388 extern const struct iomap_ops z_erofs_iomap_report_ops;
389
390 /* flags for erofs_fscache_register_cookie() */
391 #define EROFS_REG_COOKIE_SHARE 0x0001
392 #define EROFS_REG_COOKIE_NEED_NOEXIST 0x0002
393
394 void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf,
395 erofs_off_t *offset, int *lengthp);
396 void erofs_unmap_metabuf(struct erofs_buf *buf);
397 void erofs_put_metabuf(struct erofs_buf *buf);
398 void *erofs_bread(struct erofs_buf *buf, erofs_off_t offset,
399 enum erofs_kmap_type type);
400 void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb);
401 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
402 erofs_off_t offset, enum erofs_kmap_type type);
403 int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev);
404 int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
405 u64 start, u64 len);
406 int erofs_map_blocks(struct inode *inode, struct erofs_map_blocks *map);
407 struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid);
408 int erofs_getattr(struct mnt_idmap *idmap, const struct path *path,
409 struct kstat *stat, u32 request_mask,
410 unsigned int query_flags);
411 int erofs_namei(struct inode *dir, const struct qstr *name,
412 erofs_nid_t *nid, unsigned int *d_type);
413
414 static inline void *erofs_vm_map_ram(struct page **pages, unsigned int count)
415 {
416 int retried = 0;
417
418 while (1) {
419 void *p = vm_map_ram(pages, count, -1);
420
421 /* retry two more times (totally 3 times) */
422 if (p || ++retried >= 3)
423 return p;
424 vm_unmap_aliases();
425 }
426 return NULL;
427 }
428
429 int erofs_register_sysfs(struct super_block *sb);
430 void erofs_unregister_sysfs(struct super_block *sb);
431 int __init erofs_init_sysfs(void);
432 void erofs_exit_sysfs(void);
433
434 struct page *__erofs_allocpage(struct page **pagepool, gfp_t gfp, bool tryrsv);
435 static inline struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp)
436 {
437 return __erofs_allocpage(pagepool, gfp, false);
438 }
439 static inline void erofs_pagepool_add(struct page **pagepool, struct page *page)
440 {
441 set_page_private(page, (unsigned long)*pagepool);
442 *pagepool = page;
443 }
444 void erofs_release_pages(struct page **pagepool);
445
446 #ifdef CONFIG_EROFS_FS_ZIP
447 void erofs_workgroup_put(struct erofs_workgroup *grp);
448 struct erofs_workgroup *erofs_find_workgroup(struct super_block *sb,
449 pgoff_t index);
450 struct erofs_workgroup *erofs_insert_workgroup(struct super_block *sb,
451 struct erofs_workgroup *grp);
452 void erofs_workgroup_free_rcu(struct erofs_workgroup *grp);
453 void erofs_shrinker_register(struct super_block *sb);
454 void erofs_shrinker_unregister(struct super_block *sb);
455 int __init erofs_init_shrinker(void);
456 void erofs_exit_shrinker(void);
457 int __init z_erofs_init_subsystem(void);
458 void z_erofs_exit_subsystem(void);
459 int erofs_try_to_free_all_cached_folios(struct erofs_sb_info *sbi,
460 struct erofs_workgroup *egrp);
461 int z_erofs_map_blocks_iter(struct inode *inode, struct erofs_map_blocks *map,
462 int flags);
463 void *z_erofs_get_gbuf(unsigned int requiredpages);
464 void z_erofs_put_gbuf(void *ptr);
465 int z_erofs_gbuf_growsize(unsigned int nrpages);
466 int __init z_erofs_gbuf_init(void);
467 void z_erofs_gbuf_exit(void);
468 int erofs_init_managed_cache(struct super_block *sb);
469 int z_erofs_parse_cfgs(struct super_block *sb, struct erofs_super_block *dsb);
470 #else
471 static inline void erofs_shrinker_register(struct super_block *sb) {}
472 static inline void erofs_shrinker_unregister(struct super_block *sb) {}
473 static inline int erofs_init_shrinker(void) { return 0; }
474 static inline void erofs_exit_shrinker(void) {}
475 static inline int z_erofs_init_subsystem(void) { return 0; }
476 static inline void z_erofs_exit_subsystem(void) {}
477 static inline int erofs_init_managed_cache(struct super_block *sb) { return 0; }
478 #endif /* !CONFIG_EROFS_FS_ZIP */
479
480 #ifdef CONFIG_EROFS_FS_ONDEMAND
481 int erofs_fscache_register_fs(struct super_block *sb);
482 void erofs_fscache_unregister_fs(struct super_block *sb);
483
484 struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
485 char *name, unsigned int flags);
486 void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache);
487 struct bio *erofs_fscache_bio_alloc(struct erofs_map_dev *mdev);
488 void erofs_fscache_submit_bio(struct bio *bio);
489 #else
490 static inline int erofs_fscache_register_fs(struct super_block *sb)
491 {
492 return -EOPNOTSUPP;
493 }
494 static inline void erofs_fscache_unregister_fs(struct super_block *sb) {}
495
496 static inline
497 struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
498 char *name, unsigned int flags)
499 {
500 return ERR_PTR(-EOPNOTSUPP);
501 }
502
503 static inline void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache)
504 {
505 }
506 static inline struct bio *erofs_fscache_bio_alloc(struct erofs_map_dev *mdev) { return NULL; }
507 static inline void erofs_fscache_submit_bio(struct bio *bio) {}
508 #endif
509
510 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */
511
512 #endif /* __EROFS_INTERNAL_H */
513
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.