xref: /linux/fs/erofs/internal.h (revision c532de5a67a70f8533d495f8f2aaa9a0491c3ad0)
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 *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 file *fdev;
134 	struct inode *packed_inode;
135 	struct erofs_dev_context *devs;
136 	struct dax_device *dax_dev;
137 	u64 dax_part_off;
138 	u64 total_blocks;
139 	u32 primarydevice_blocks;
140 
141 	u32 meta_blkaddr;
142 #ifdef CONFIG_EROFS_FS_XATTR
143 	u32 xattr_blkaddr;
144 	u32 xattr_prefix_start;
145 	u8 xattr_prefix_count;
146 	struct erofs_xattr_prefix_item *xattr_prefixes;
147 	unsigned int xattr_filter_reserved;
148 #endif
149 	u16 device_id_mask;	/* valid bits of device id to be used */
150 
151 	unsigned char islotbits;	/* inode slot unit size in bit shift */
152 	unsigned char blkszbits;	/* filesystem block size in bit shift */
153 
154 	u32 sb_size;			/* total superblock size */
155 	u32 build_time_nsec;
156 	u64 build_time;
157 
158 	/* what we really care is nid, rather than ino.. */
159 	erofs_nid_t root_nid;
160 	erofs_nid_t packed_nid;
161 	/* used for statfs, f_files - f_favail */
162 	u64 inos;
163 
164 	u8 uuid[16];                    /* 128-bit uuid for volume */
165 	u8 volume_name[16];             /* volume name */
166 	u32 feature_compat;
167 	u32 feature_incompat;
168 
169 	/* sysfs support */
170 	struct kobject s_kobj;		/* /sys/fs/erofs/<devname> */
171 	struct completion s_kobj_unregister;
172 
173 	/* fscache support */
174 	struct fscache_volume *volume;
175 	struct erofs_fscache *s_fscache;
176 	struct erofs_domain *domain;
177 	char *fsid;
178 	char *domain_id;
179 };
180 
181 #define EROFS_SB(sb) ((struct erofs_sb_info *)(sb)->s_fs_info)
182 #define EROFS_I_SB(inode) ((struct erofs_sb_info *)(inode)->i_sb->s_fs_info)
183 
184 /* Mount flags set via mount options or defaults */
185 #define EROFS_MOUNT_XATTR_USER		0x00000010
186 #define EROFS_MOUNT_POSIX_ACL		0x00000020
187 #define EROFS_MOUNT_DAX_ALWAYS		0x00000040
188 #define EROFS_MOUNT_DAX_NEVER		0x00000080
189 
190 #define clear_opt(opt, option)	((opt)->mount_opt &= ~EROFS_MOUNT_##option)
191 #define set_opt(opt, option)	((opt)->mount_opt |= EROFS_MOUNT_##option)
192 #define test_opt(opt, option)	((opt)->mount_opt & EROFS_MOUNT_##option)
193 
194 static inline bool erofs_is_fileio_mode(struct erofs_sb_info *sbi)
195 {
196 	return IS_ENABLED(CONFIG_EROFS_FS_BACKED_BY_FILE) && sbi->fdev;
197 }
198 
199 static inline bool erofs_is_fscache_mode(struct super_block *sb)
200 {
201 	return IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) &&
202 			!erofs_is_fileio_mode(EROFS_SB(sb)) && !sb->s_bdev;
203 }
204 
205 enum {
206 	EROFS_ZIP_CACHE_DISABLED,
207 	EROFS_ZIP_CACHE_READAHEAD,
208 	EROFS_ZIP_CACHE_READAROUND
209 };
210 
211 /* basic unit of the workstation of a super_block */
212 struct erofs_workgroup {
213 	pgoff_t index;
214 	struct lockref lockref;
215 };
216 
217 enum erofs_kmap_type {
218 	EROFS_NO_KMAP,		/* don't map the buffer */
219 	EROFS_KMAP,		/* use kmap_local_page() to map the buffer */
220 };
221 
222 struct erofs_buf {
223 	struct address_space *mapping;
224 	struct page *page;
225 	void *base;
226 	enum erofs_kmap_type kmap_type;
227 };
228 #define __EROFS_BUF_INITIALIZER	((struct erofs_buf){ .page = NULL })
229 
230 #define erofs_blknr(sb, addr)	((erofs_blk_t)((addr) >> (sb)->s_blocksize_bits))
231 #define erofs_blkoff(sb, addr)	((addr) & ((sb)->s_blocksize - 1))
232 #define erofs_pos(sb, blk)	((erofs_off_t)(blk) << (sb)->s_blocksize_bits)
233 #define erofs_iblks(i)	(round_up((i)->i_size, i_blocksize(i)) >> (i)->i_blkbits)
234 
235 #define EROFS_FEATURE_FUNCS(name, compat, feature) \
236 static inline bool erofs_sb_has_##name(struct erofs_sb_info *sbi) \
237 { \
238 	return sbi->feature_##compat & EROFS_FEATURE_##feature; \
239 }
240 
241 EROFS_FEATURE_FUNCS(zero_padding, incompat, INCOMPAT_ZERO_PADDING)
242 EROFS_FEATURE_FUNCS(compr_cfgs, incompat, INCOMPAT_COMPR_CFGS)
243 EROFS_FEATURE_FUNCS(big_pcluster, incompat, INCOMPAT_BIG_PCLUSTER)
244 EROFS_FEATURE_FUNCS(chunked_file, incompat, INCOMPAT_CHUNKED_FILE)
245 EROFS_FEATURE_FUNCS(device_table, incompat, INCOMPAT_DEVICE_TABLE)
246 EROFS_FEATURE_FUNCS(compr_head2, incompat, INCOMPAT_COMPR_HEAD2)
247 EROFS_FEATURE_FUNCS(ztailpacking, incompat, INCOMPAT_ZTAILPACKING)
248 EROFS_FEATURE_FUNCS(fragments, incompat, INCOMPAT_FRAGMENTS)
249 EROFS_FEATURE_FUNCS(dedupe, incompat, INCOMPAT_DEDUPE)
250 EROFS_FEATURE_FUNCS(xattr_prefixes, incompat, INCOMPAT_XATTR_PREFIXES)
251 EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM)
252 EROFS_FEATURE_FUNCS(xattr_filter, compat, COMPAT_XATTR_FILTER)
253 
254 /* atomic flag definitions */
255 #define EROFS_I_EA_INITED_BIT	0
256 #define EROFS_I_Z_INITED_BIT	1
257 
258 /* bitlock definitions (arranged in reverse order) */
259 #define EROFS_I_BL_XATTR_BIT	(BITS_PER_LONG - 1)
260 #define EROFS_I_BL_Z_BIT	(BITS_PER_LONG - 2)
261 
262 struct erofs_inode {
263 	erofs_nid_t nid;
264 
265 	/* atomic flags (including bitlocks) */
266 	unsigned long flags;
267 
268 	unsigned char datalayout;
269 	unsigned char inode_isize;
270 	unsigned int xattr_isize;
271 
272 	unsigned int xattr_name_filter;
273 	unsigned int xattr_shared_count;
274 	unsigned int *xattr_shared_xattrs;
275 
276 	union {
277 		erofs_blk_t raw_blkaddr;
278 		struct {
279 			unsigned short	chunkformat;
280 			unsigned char	chunkbits;
281 		};
282 #ifdef CONFIG_EROFS_FS_ZIP
283 		struct {
284 			unsigned short z_advise;
285 			unsigned char  z_algorithmtype[2];
286 			unsigned char  z_logical_clusterbits;
287 			unsigned long  z_tailextent_headlcn;
288 			union {
289 				struct {
290 					erofs_off_t    z_idataoff;
291 					unsigned short z_idata_size;
292 				};
293 				erofs_off_t z_fragmentoff;
294 			};
295 		};
296 #endif	/* CONFIG_EROFS_FS_ZIP */
297 	};
298 	/* the corresponding vfs inode */
299 	struct inode vfs_inode;
300 };
301 
302 #define EROFS_I(ptr)	container_of(ptr, struct erofs_inode, vfs_inode)
303 
304 static inline erofs_off_t erofs_iloc(struct inode *inode)
305 {
306 	struct erofs_sb_info *sbi = EROFS_I_SB(inode);
307 
308 	return erofs_pos(inode->i_sb, sbi->meta_blkaddr) +
309 		(EROFS_I(inode)->nid << sbi->islotbits);
310 }
311 
312 static inline unsigned int erofs_inode_version(unsigned int ifmt)
313 {
314 	return (ifmt >> EROFS_I_VERSION_BIT) & EROFS_I_VERSION_MASK;
315 }
316 
317 static inline unsigned int erofs_inode_datalayout(unsigned int ifmt)
318 {
319 	return (ifmt >> EROFS_I_DATALAYOUT_BIT) & EROFS_I_DATALAYOUT_MASK;
320 }
321 
322 /* reclaiming is never triggered when allocating new folios. */
323 static inline struct folio *erofs_grab_folio_nowait(struct address_space *as,
324 						    pgoff_t index)
325 {
326 	return __filemap_get_folio(as, index,
327 			FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
328 			readahead_gfp_mask(as) & ~__GFP_RECLAIM);
329 }
330 
331 /* Has a disk mapping */
332 #define EROFS_MAP_MAPPED	0x0001
333 /* Located in metadata (could be copied from bd_inode) */
334 #define EROFS_MAP_META		0x0002
335 /* The extent is encoded */
336 #define EROFS_MAP_ENCODED	0x0004
337 /* The length of extent is full */
338 #define EROFS_MAP_FULL_MAPPED	0x0008
339 /* Located in the special packed inode */
340 #define EROFS_MAP_FRAGMENT	0x0010
341 /* The extent refers to partial decompressed data */
342 #define EROFS_MAP_PARTIAL_REF	0x0020
343 
344 struct erofs_map_blocks {
345 	struct erofs_buf buf;
346 
347 	erofs_off_t m_pa, m_la;
348 	u64 m_plen, m_llen;
349 
350 	unsigned short m_deviceid;
351 	char m_algorithmformat;
352 	unsigned int m_flags;
353 };
354 
355 /*
356  * Used to get the exact decompressed length, e.g. fiemap (consider lookback
357  * approach instead if possible since it's more metadata lightweight.)
358  */
359 #define EROFS_GET_BLOCKS_FIEMAP		0x0001
360 /* Used to map the whole extent if non-negligible data is requested for LZMA */
361 #define EROFS_GET_BLOCKS_READMORE	0x0002
362 /* Used to map tail extent for tailpacking inline or fragment pcluster */
363 #define EROFS_GET_BLOCKS_FINDTAIL	0x0004
364 
365 enum {
366 	Z_EROFS_COMPRESSION_SHIFTED = Z_EROFS_COMPRESSION_MAX,
367 	Z_EROFS_COMPRESSION_INTERLACED,
368 	Z_EROFS_COMPRESSION_RUNTIME_MAX
369 };
370 
371 struct erofs_map_dev {
372 	struct erofs_fscache *m_fscache;
373 	struct block_device *m_bdev;
374 	struct dax_device *m_daxdev;
375 	struct file *m_fp;
376 	u64 m_dax_part_off;
377 
378 	erofs_off_t m_pa;
379 	unsigned int m_deviceid;
380 };
381 
382 extern const struct super_operations erofs_sops;
383 
384 extern const struct address_space_operations erofs_aops;
385 extern const struct address_space_operations erofs_fileio_aops;
386 extern const struct address_space_operations z_erofs_aops;
387 extern const struct address_space_operations erofs_fscache_access_aops;
388 
389 extern const struct inode_operations erofs_generic_iops;
390 extern const struct inode_operations erofs_symlink_iops;
391 extern const struct inode_operations erofs_fast_symlink_iops;
392 extern const struct inode_operations erofs_dir_iops;
393 
394 extern const struct file_operations erofs_file_fops;
395 extern const struct file_operations erofs_dir_fops;
396 
397 extern const struct iomap_ops z_erofs_iomap_report_ops;
398 
399 /* flags for erofs_fscache_register_cookie() */
400 #define EROFS_REG_COOKIE_SHARE		0x0001
401 #define EROFS_REG_COOKIE_NEED_NOEXIST	0x0002
402 
403 void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf,
404 			  erofs_off_t *offset, int *lengthp);
405 void erofs_unmap_metabuf(struct erofs_buf *buf);
406 void erofs_put_metabuf(struct erofs_buf *buf);
407 void *erofs_bread(struct erofs_buf *buf, erofs_off_t offset,
408 		  enum erofs_kmap_type type);
409 void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb);
410 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
411 			 erofs_off_t offset, enum erofs_kmap_type type);
412 int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev);
413 int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
414 		 u64 start, u64 len);
415 int erofs_map_blocks(struct inode *inode, struct erofs_map_blocks *map);
416 void erofs_onlinefolio_init(struct folio *folio);
417 void erofs_onlinefolio_split(struct folio *folio);
418 void erofs_onlinefolio_end(struct folio *folio, int err);
419 struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid);
420 int erofs_getattr(struct mnt_idmap *idmap, const struct path *path,
421 		  struct kstat *stat, u32 request_mask,
422 		  unsigned int query_flags);
423 int erofs_namei(struct inode *dir, const struct qstr *name,
424 		erofs_nid_t *nid, unsigned int *d_type);
425 
426 static inline void *erofs_vm_map_ram(struct page **pages, unsigned int count)
427 {
428 	int retried = 0;
429 
430 	while (1) {
431 		void *p = vm_map_ram(pages, count, -1);
432 
433 		/* retry two more times (totally 3 times) */
434 		if (p || ++retried >= 3)
435 			return p;
436 		vm_unmap_aliases();
437 	}
438 	return NULL;
439 }
440 
441 int erofs_register_sysfs(struct super_block *sb);
442 void erofs_unregister_sysfs(struct super_block *sb);
443 int __init erofs_init_sysfs(void);
444 void erofs_exit_sysfs(void);
445 
446 struct page *__erofs_allocpage(struct page **pagepool, gfp_t gfp, bool tryrsv);
447 static inline struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp)
448 {
449 	return __erofs_allocpage(pagepool, gfp, false);
450 }
451 static inline void erofs_pagepool_add(struct page **pagepool, struct page *page)
452 {
453 	set_page_private(page, (unsigned long)*pagepool);
454 	*pagepool = page;
455 }
456 void erofs_release_pages(struct page **pagepool);
457 
458 #ifdef CONFIG_EROFS_FS_ZIP
459 void erofs_workgroup_put(struct erofs_workgroup *grp);
460 struct erofs_workgroup *erofs_find_workgroup(struct super_block *sb,
461 					     pgoff_t index);
462 struct erofs_workgroup *erofs_insert_workgroup(struct super_block *sb,
463 					       struct erofs_workgroup *grp);
464 void erofs_workgroup_free_rcu(struct erofs_workgroup *grp);
465 void erofs_shrinker_register(struct super_block *sb);
466 void erofs_shrinker_unregister(struct super_block *sb);
467 int __init erofs_init_shrinker(void);
468 void erofs_exit_shrinker(void);
469 int __init z_erofs_init_subsystem(void);
470 void z_erofs_exit_subsystem(void);
471 int erofs_try_to_free_all_cached_folios(struct erofs_sb_info *sbi,
472 					struct erofs_workgroup *egrp);
473 int z_erofs_map_blocks_iter(struct inode *inode, struct erofs_map_blocks *map,
474 			    int flags);
475 void *z_erofs_get_gbuf(unsigned int requiredpages);
476 void z_erofs_put_gbuf(void *ptr);
477 int z_erofs_gbuf_growsize(unsigned int nrpages);
478 int __init z_erofs_gbuf_init(void);
479 void z_erofs_gbuf_exit(void);
480 int erofs_init_managed_cache(struct super_block *sb);
481 int z_erofs_parse_cfgs(struct super_block *sb, struct erofs_super_block *dsb);
482 #else
483 static inline void erofs_shrinker_register(struct super_block *sb) {}
484 static inline void erofs_shrinker_unregister(struct super_block *sb) {}
485 static inline int erofs_init_shrinker(void) { return 0; }
486 static inline void erofs_exit_shrinker(void) {}
487 static inline int z_erofs_init_subsystem(void) { return 0; }
488 static inline void z_erofs_exit_subsystem(void) {}
489 static inline int erofs_init_managed_cache(struct super_block *sb) { return 0; }
490 #endif	/* !CONFIG_EROFS_FS_ZIP */
491 
492 #ifdef CONFIG_EROFS_FS_BACKED_BY_FILE
493 struct bio *erofs_fileio_bio_alloc(struct erofs_map_dev *mdev);
494 void erofs_fileio_submit_bio(struct bio *bio);
495 #else
496 static inline struct bio *erofs_fileio_bio_alloc(struct erofs_map_dev *mdev) { return NULL; }
497 static inline void erofs_fileio_submit_bio(struct bio *bio) {}
498 #endif
499 
500 #ifdef CONFIG_EROFS_FS_ONDEMAND
501 int erofs_fscache_register_fs(struct super_block *sb);
502 void erofs_fscache_unregister_fs(struct super_block *sb);
503 
504 struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
505 					char *name, unsigned int flags);
506 void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache);
507 struct bio *erofs_fscache_bio_alloc(struct erofs_map_dev *mdev);
508 void erofs_fscache_submit_bio(struct bio *bio);
509 #else
510 static inline int erofs_fscache_register_fs(struct super_block *sb)
511 {
512 	return -EOPNOTSUPP;
513 }
514 static inline void erofs_fscache_unregister_fs(struct super_block *sb) {}
515 
516 static inline
517 struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
518 					char *name, unsigned int flags)
519 {
520 	return ERR_PTR(-EOPNOTSUPP);
521 }
522 
523 static inline void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache)
524 {
525 }
526 static inline struct bio *erofs_fscache_bio_alloc(struct erofs_map_dev *mdev) { return NULL; }
527 static inline void erofs_fscache_submit_bio(struct bio *bio) {}
528 #endif
529 
530 #define EFSCORRUPTED    EUCLEAN         /* Filesystem is corrupted */
531 
532 #endif	/* __EROFS_INTERNAL_H */
533