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 __printf(2, 3) void _erofs_printk(struct super_block *sb, const char *fmt, ...);
24 #define erofs_err(sb, fmt, ...) \
25 _erofs_printk(sb, KERN_ERR fmt "\n", ##__VA_ARGS__)
26 #define erofs_info(sb, fmt, ...) \
27 _erofs_printk(sb, KERN_INFO fmt "\n", ##__VA_ARGS__)
28
29 #ifdef CONFIG_EROFS_FS_DEBUG
30 #define DBG_BUGON BUG_ON
31 #else
32 #define DBG_BUGON(x) ((void)(x))
33 #endif /* !CONFIG_EROFS_FS_DEBUG */
34
35 /* EROFS_SUPER_MAGIC_V1 to represent the whole file system */
36 #define EROFS_SUPER_MAGIC EROFS_SUPER_MAGIC_V1
37
38 typedef u64 erofs_nid_t;
39 typedef u64 erofs_off_t;
40 typedef u64 erofs_blk_t;
41
42 struct erofs_device_info {
43 char *path;
44 struct erofs_fscache *fscache;
45 struct file *file;
46 struct dax_device *dax_dev;
47 u64 dax_part_off;
48
49 erofs_blk_t blocks;
50 erofs_blk_t uniaddr;
51 };
52
53 enum {
54 EROFS_SYNC_DECOMPRESS_AUTO,
55 EROFS_SYNC_DECOMPRESS_FORCE_ON,
56 EROFS_SYNC_DECOMPRESS_FORCE_OFF
57 };
58
59 struct erofs_mount_opts {
60 /* current strategy of how to use managed cache */
61 unsigned char cache_strategy;
62 /* strategy of sync decompression (0 - auto, 1 - force on, 2 - force off) */
63 unsigned int sync_decompress;
64 /* threshold for decompression synchronously */
65 unsigned int max_sync_decompress_pages;
66 unsigned int mount_opt;
67 };
68
69 struct erofs_dev_context {
70 struct idr tree;
71 struct rw_semaphore rwsem;
72
73 unsigned int extra_devices;
74 bool flatdev;
75 };
76
77 /* all filesystem-wide lz4 configurations */
78 struct erofs_sb_lz4_info {
79 /* # of pages needed for EROFS lz4 rolling decompression */
80 u16 max_distance_pages;
81 /* maximum possible blocks for pclusters in the filesystem */
82 u16 max_pclusterblks;
83 };
84
85 struct erofs_domain {
86 refcount_t ref;
87 struct list_head list;
88 struct fscache_volume *volume;
89 char *domain_id;
90 };
91
92 struct erofs_fscache {
93 struct fscache_cookie *cookie;
94 struct inode *inode; /* anonymous inode for the blob */
95
96 /* used for share domain mode */
97 struct erofs_domain *domain;
98 struct list_head node;
99 refcount_t ref;
100 char *name;
101 };
102
103 struct erofs_xattr_prefix_item {
104 struct erofs_xattr_long_prefix *prefix;
105 u8 infix_len;
106 };
107
108 struct erofs_sb_info {
109 struct erofs_device_info dif0;
110 struct erofs_mount_opts opt; /* options */
111 #ifdef CONFIG_EROFS_FS_ZIP
112 /* list for all registered superblocks, mainly for shrinker */
113 struct list_head list;
114 struct mutex umount_mutex;
115
116 /* managed XArray arranged in physical block number */
117 struct xarray managed_pslots;
118
119 unsigned int shrinker_run_no;
120 u16 available_compr_algs;
121
122 /* pseudo inode to manage cached pages */
123 struct inode *managed_cache;
124
125 struct erofs_sb_lz4_info lz4;
126 #endif /* CONFIG_EROFS_FS_ZIP */
127 struct inode *packed_inode;
128 struct erofs_dev_context *devs;
129 u64 total_blocks;
130
131 u32 meta_blkaddr;
132 #ifdef CONFIG_EROFS_FS_XATTR
133 u32 xattr_blkaddr;
134 u32 xattr_prefix_start;
135 u8 xattr_prefix_count;
136 struct erofs_xattr_prefix_item *xattr_prefixes;
137 unsigned int xattr_filter_reserved;
138 #endif
139 u16 device_id_mask; /* valid bits of device id to be used */
140
141 unsigned char islotbits; /* inode slot unit size in bit shift */
142 unsigned char blkszbits; /* filesystem block size in bit shift */
143
144 u32 sb_size; /* total superblock size */
145 u32 fixed_nsec;
146 s64 epoch;
147
148 /* what we really care is nid, rather than ino.. */
149 erofs_nid_t root_nid;
150 erofs_nid_t packed_nid;
151 /* used for statfs, f_files - f_favail */
152 u64 inos;
153
154 u32 feature_compat;
155 u32 feature_incompat;
156
157 /* sysfs support */
158 struct kobject s_kobj; /* /sys/fs/erofs/<devname> */
159 struct completion s_kobj_unregister;
160
161 /* fscache support */
162 struct fscache_volume *volume;
163 struct erofs_domain *domain;
164 char *fsid;
165 char *domain_id;
166 };
167
168 #define EROFS_SB(sb) ((struct erofs_sb_info *)(sb)->s_fs_info)
169 #define EROFS_I_SB(inode) ((struct erofs_sb_info *)(inode)->i_sb->s_fs_info)
170
171 /* Mount flags set via mount options or defaults */
172 #define EROFS_MOUNT_XATTR_USER 0x00000010
173 #define EROFS_MOUNT_POSIX_ACL 0x00000020
174 #define EROFS_MOUNT_DAX_ALWAYS 0x00000040
175 #define EROFS_MOUNT_DAX_NEVER 0x00000080
176 #define EROFS_MOUNT_DIRECT_IO 0x00000100
177
178 #define clear_opt(opt, option) ((opt)->mount_opt &= ~EROFS_MOUNT_##option)
179 #define set_opt(opt, option) ((opt)->mount_opt |= EROFS_MOUNT_##option)
180 #define test_opt(opt, option) ((opt)->mount_opt & EROFS_MOUNT_##option)
181
erofs_is_fileio_mode(struct erofs_sb_info * sbi)182 static inline bool erofs_is_fileio_mode(struct erofs_sb_info *sbi)
183 {
184 return IS_ENABLED(CONFIG_EROFS_FS_BACKED_BY_FILE) && sbi->dif0.file;
185 }
186
erofs_is_fscache_mode(struct super_block * sb)187 static inline bool erofs_is_fscache_mode(struct super_block *sb)
188 {
189 return IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) &&
190 !erofs_is_fileio_mode(EROFS_SB(sb)) && !sb->s_bdev;
191 }
192
193 enum {
194 EROFS_ZIP_CACHE_DISABLED,
195 EROFS_ZIP_CACHE_READAHEAD,
196 EROFS_ZIP_CACHE_READAROUND
197 };
198
199 struct erofs_buf {
200 struct address_space *mapping;
201 struct file *file;
202 struct page *page;
203 void *base;
204 };
205 #define __EROFS_BUF_INITIALIZER ((struct erofs_buf){ .page = NULL })
206
207 #define erofs_blknr(sb, pos) ((erofs_blk_t)((pos) >> (sb)->s_blocksize_bits))
208 #define erofs_blkoff(sb, pos) ((pos) & ((sb)->s_blocksize - 1))
209 #define erofs_pos(sb, blk) ((erofs_off_t)(blk) << (sb)->s_blocksize_bits)
210 #define erofs_iblks(i) (round_up((i)->i_size, i_blocksize(i)) >> (i)->i_blkbits)
211
212 #define EROFS_FEATURE_FUNCS(name, compat, feature) \
213 static inline bool erofs_sb_has_##name(struct erofs_sb_info *sbi) \
214 { \
215 return sbi->feature_##compat & EROFS_FEATURE_##feature; \
216 }
217
218 EROFS_FEATURE_FUNCS(zero_padding, incompat, INCOMPAT_ZERO_PADDING)
219 EROFS_FEATURE_FUNCS(compr_cfgs, incompat, INCOMPAT_COMPR_CFGS)
220 EROFS_FEATURE_FUNCS(big_pcluster, incompat, INCOMPAT_BIG_PCLUSTER)
221 EROFS_FEATURE_FUNCS(chunked_file, incompat, INCOMPAT_CHUNKED_FILE)
222 EROFS_FEATURE_FUNCS(device_table, incompat, INCOMPAT_DEVICE_TABLE)
223 EROFS_FEATURE_FUNCS(compr_head2, incompat, INCOMPAT_COMPR_HEAD2)
224 EROFS_FEATURE_FUNCS(ztailpacking, incompat, INCOMPAT_ZTAILPACKING)
225 EROFS_FEATURE_FUNCS(fragments, incompat, INCOMPAT_FRAGMENTS)
226 EROFS_FEATURE_FUNCS(dedupe, incompat, INCOMPAT_DEDUPE)
227 EROFS_FEATURE_FUNCS(xattr_prefixes, incompat, INCOMPAT_XATTR_PREFIXES)
228 EROFS_FEATURE_FUNCS(48bit, incompat, INCOMPAT_48BIT)
229 EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM)
230 EROFS_FEATURE_FUNCS(xattr_filter, compat, COMPAT_XATTR_FILTER)
231
232 /* atomic flag definitions */
233 #define EROFS_I_EA_INITED_BIT 0
234 #define EROFS_I_Z_INITED_BIT 1
235
236 /* bitlock definitions (arranged in reverse order) */
237 #define EROFS_I_BL_XATTR_BIT (BITS_PER_LONG - 1)
238 #define EROFS_I_BL_Z_BIT (BITS_PER_LONG - 2)
239
240 struct erofs_inode {
241 erofs_nid_t nid;
242
243 /* atomic flags (including bitlocks) */
244 unsigned long flags;
245
246 unsigned char datalayout;
247 unsigned char inode_isize;
248 bool dot_omitted;
249 unsigned int xattr_isize;
250
251 unsigned int xattr_name_filter;
252 unsigned int xattr_shared_count;
253 unsigned int *xattr_shared_xattrs;
254
255 union {
256 erofs_blk_t startblk;
257 struct {
258 unsigned short chunkformat;
259 unsigned char chunkbits;
260 };
261 #ifdef CONFIG_EROFS_FS_ZIP
262 struct {
263 unsigned short z_advise;
264 unsigned char z_algorithmtype[2];
265 unsigned char z_lclusterbits;
266 union {
267 u64 z_tailextent_headlcn;
268 u64 z_extents;
269 };
270 erofs_off_t z_fragmentoff;
271 unsigned short z_idata_size;
272 };
273 #endif /* CONFIG_EROFS_FS_ZIP */
274 };
275 /* the corresponding vfs inode */
276 struct inode vfs_inode;
277 };
278
279 #define EROFS_I(ptr) container_of(ptr, struct erofs_inode, vfs_inode)
280
erofs_iloc(struct inode * inode)281 static inline erofs_off_t erofs_iloc(struct inode *inode)
282 {
283 struct erofs_sb_info *sbi = EROFS_I_SB(inode);
284
285 return erofs_pos(inode->i_sb, sbi->meta_blkaddr) +
286 (EROFS_I(inode)->nid << sbi->islotbits);
287 }
288
erofs_inode_version(unsigned int ifmt)289 static inline unsigned int erofs_inode_version(unsigned int ifmt)
290 {
291 return (ifmt >> EROFS_I_VERSION_BIT) & EROFS_I_VERSION_MASK;
292 }
293
erofs_inode_datalayout(unsigned int ifmt)294 static inline unsigned int erofs_inode_datalayout(unsigned int ifmt)
295 {
296 return (ifmt >> EROFS_I_DATALAYOUT_BIT) & EROFS_I_DATALAYOUT_MASK;
297 }
298
299 /* reclaiming is never triggered when allocating new folios. */
erofs_grab_folio_nowait(struct address_space * as,pgoff_t index)300 static inline struct folio *erofs_grab_folio_nowait(struct address_space *as,
301 pgoff_t index)
302 {
303 return __filemap_get_folio(as, index,
304 FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
305 readahead_gfp_mask(as) & ~__GFP_RECLAIM);
306 }
307
308 /* Has a disk mapping */
309 #define EROFS_MAP_MAPPED 0x0001
310 /* Located in metadata (could be copied from bd_inode) */
311 #define EROFS_MAP_META 0x0002
312 /* The extent is encoded */
313 #define EROFS_MAP_ENCODED 0x0004
314 /* The length of extent is full */
315 #define EROFS_MAP_FULL_MAPPED 0x0008
316 /* Located in the special packed inode */
317 #define EROFS_MAP_FRAGMENT 0x0010
318 /* The extent refers to partial decompressed data */
319 #define EROFS_MAP_PARTIAL_REF 0x0020
320
321 struct erofs_map_blocks {
322 struct erofs_buf buf;
323
324 erofs_off_t m_pa, m_la;
325 u64 m_plen, m_llen;
326
327 unsigned short m_deviceid;
328 char m_algorithmformat;
329 unsigned int m_flags;
330 };
331
332 /*
333 * Used to get the exact decompressed length, e.g. fiemap (consider lookback
334 * approach instead if possible since it's more metadata lightweight.)
335 */
336 #define EROFS_GET_BLOCKS_FIEMAP 0x0001
337 /* Used to map the whole extent if non-negligible data is requested for LZMA */
338 #define EROFS_GET_BLOCKS_READMORE 0x0002
339 /* Used to map tail extent for tailpacking inline or fragment pcluster */
340 #define EROFS_GET_BLOCKS_FINDTAIL 0x0004
341
342 enum {
343 Z_EROFS_COMPRESSION_SHIFTED = Z_EROFS_COMPRESSION_MAX,
344 Z_EROFS_COMPRESSION_INTERLACED,
345 Z_EROFS_COMPRESSION_RUNTIME_MAX
346 };
347
348 struct erofs_map_dev {
349 struct super_block *m_sb;
350 struct erofs_device_info *m_dif;
351 struct block_device *m_bdev;
352
353 erofs_off_t m_pa;
354 unsigned int m_deviceid;
355 };
356
357 extern const struct super_operations erofs_sops;
358
359 extern const struct address_space_operations erofs_aops;
360 extern const struct address_space_operations erofs_fileio_aops;
361 extern const struct address_space_operations z_erofs_aops;
362 extern const struct address_space_operations erofs_fscache_access_aops;
363
364 extern const struct inode_operations erofs_generic_iops;
365 extern const struct inode_operations erofs_symlink_iops;
366 extern const struct inode_operations erofs_fast_symlink_iops;
367 extern const struct inode_operations erofs_dir_iops;
368
369 extern const struct file_operations erofs_file_fops;
370 extern const struct file_operations erofs_dir_fops;
371
372 extern const struct iomap_ops z_erofs_iomap_report_ops;
373
374 /* flags for erofs_fscache_register_cookie() */
375 #define EROFS_REG_COOKIE_SHARE 0x0001
376 #define EROFS_REG_COOKIE_NEED_NOEXIST 0x0002
377
378 void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf,
379 erofs_off_t *offset, int *lengthp);
380 void erofs_unmap_metabuf(struct erofs_buf *buf);
381 void erofs_put_metabuf(struct erofs_buf *buf);
382 void *erofs_bread(struct erofs_buf *buf, erofs_off_t offset, bool need_kmap);
383 void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb);
384 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
385 erofs_off_t offset, bool need_kmap);
386 int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev);
387 int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
388 u64 start, u64 len);
389 int erofs_map_blocks(struct inode *inode, struct erofs_map_blocks *map);
390 void erofs_onlinefolio_init(struct folio *folio);
391 void erofs_onlinefolio_split(struct folio *folio);
392 void erofs_onlinefolio_end(struct folio *folio, int err);
393 struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid);
394 int erofs_getattr(struct mnt_idmap *idmap, const struct path *path,
395 struct kstat *stat, u32 request_mask,
396 unsigned int query_flags);
397 int erofs_namei(struct inode *dir, const struct qstr *name,
398 erofs_nid_t *nid, unsigned int *d_type);
399
erofs_vm_map_ram(struct page ** pages,unsigned int count)400 static inline void *erofs_vm_map_ram(struct page **pages, unsigned int count)
401 {
402 int retried = 0;
403
404 while (1) {
405 void *p = vm_map_ram(pages, count, -1);
406
407 /* retry two more times (totally 3 times) */
408 if (p || ++retried >= 3)
409 return p;
410 vm_unmap_aliases();
411 }
412 return NULL;
413 }
414
415 int erofs_register_sysfs(struct super_block *sb);
416 void erofs_unregister_sysfs(struct super_block *sb);
417 int __init erofs_init_sysfs(void);
418 void erofs_exit_sysfs(void);
419
420 struct page *__erofs_allocpage(struct page **pagepool, gfp_t gfp, bool tryrsv);
erofs_allocpage(struct page ** pagepool,gfp_t gfp)421 static inline struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp)
422 {
423 return __erofs_allocpage(pagepool, gfp, false);
424 }
erofs_pagepool_add(struct page ** pagepool,struct page * page)425 static inline void erofs_pagepool_add(struct page **pagepool, struct page *page)
426 {
427 set_page_private(page, (unsigned long)*pagepool);
428 *pagepool = page;
429 }
430 void erofs_release_pages(struct page **pagepool);
431
432 #ifdef CONFIG_EROFS_FS_ZIP
433 #define MNGD_MAPPING(sbi) ((sbi)->managed_cache->i_mapping)
434
435 extern atomic_long_t erofs_global_shrink_cnt;
436 void erofs_shrinker_register(struct super_block *sb);
437 void erofs_shrinker_unregister(struct super_block *sb);
438 int __init erofs_init_shrinker(void);
439 void erofs_exit_shrinker(void);
440 int __init z_erofs_init_subsystem(void);
441 void z_erofs_exit_subsystem(void);
442 int z_erofs_init_super(struct super_block *sb);
443 unsigned long z_erofs_shrink_scan(struct erofs_sb_info *sbi,
444 unsigned long nr_shrink);
445 int z_erofs_map_blocks_iter(struct inode *inode, struct erofs_map_blocks *map,
446 int flags);
447 void *z_erofs_get_gbuf(unsigned int requiredpages);
448 void z_erofs_put_gbuf(void *ptr);
449 int z_erofs_gbuf_growsize(unsigned int nrpages);
450 int __init z_erofs_gbuf_init(void);
451 void z_erofs_gbuf_exit(void);
452 int z_erofs_parse_cfgs(struct super_block *sb, struct erofs_super_block *dsb);
453 #else
erofs_shrinker_register(struct super_block * sb)454 static inline void erofs_shrinker_register(struct super_block *sb) {}
erofs_shrinker_unregister(struct super_block * sb)455 static inline void erofs_shrinker_unregister(struct super_block *sb) {}
erofs_init_shrinker(void)456 static inline int erofs_init_shrinker(void) { return 0; }
erofs_exit_shrinker(void)457 static inline void erofs_exit_shrinker(void) {}
z_erofs_init_subsystem(void)458 static inline int z_erofs_init_subsystem(void) { return 0; }
z_erofs_exit_subsystem(void)459 static inline void z_erofs_exit_subsystem(void) {}
z_erofs_init_super(struct super_block * sb)460 static inline int z_erofs_init_super(struct super_block *sb) { return 0; }
461 #endif /* !CONFIG_EROFS_FS_ZIP */
462
463 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
464 struct bio *erofs_fileio_bio_alloc(struct erofs_map_dev *mdev);
465 void erofs_fileio_submit_bio(struct bio *bio);
466 #else
erofs_fileio_bio_alloc(struct erofs_map_dev * mdev)467 static inline struct bio *erofs_fileio_bio_alloc(struct erofs_map_dev *mdev) { return NULL; }
erofs_fileio_submit_bio(struct bio * bio)468 static inline void erofs_fileio_submit_bio(struct bio *bio) {}
469 #endif
470
471 #ifdef CONFIG_EROFS_FS_ONDEMAND
472 int erofs_fscache_register_fs(struct super_block *sb);
473 void erofs_fscache_unregister_fs(struct super_block *sb);
474
475 struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
476 char *name, unsigned int flags);
477 void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache);
478 struct bio *erofs_fscache_bio_alloc(struct erofs_map_dev *mdev);
479 void erofs_fscache_submit_bio(struct bio *bio);
480 #else
erofs_fscache_register_fs(struct super_block * sb)481 static inline int erofs_fscache_register_fs(struct super_block *sb)
482 {
483 return -EOPNOTSUPP;
484 }
erofs_fscache_unregister_fs(struct super_block * sb)485 static inline void erofs_fscache_unregister_fs(struct super_block *sb) {}
486
487 static inline
erofs_fscache_register_cookie(struct super_block * sb,char * name,unsigned int flags)488 struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
489 char *name, unsigned int flags)
490 {
491 return ERR_PTR(-EOPNOTSUPP);
492 }
493
erofs_fscache_unregister_cookie(struct erofs_fscache * fscache)494 static inline void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache)
495 {
496 }
erofs_fscache_bio_alloc(struct erofs_map_dev * mdev)497 static inline struct bio *erofs_fscache_bio_alloc(struct erofs_map_dev *mdev) { return NULL; }
erofs_fscache_submit_bio(struct bio * bio)498 static inline void erofs_fscache_submit_bio(struct bio *bio) {}
499 #endif
500
501 #define EFSCORRUPTED EUCLEAN /* Filesystem is corrupted */
502
503 #endif /* __EROFS_INTERNAL_H */
504