xref: /openbmc/linux/fs/erofs/internal.h (revision fe4549b1)
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/dcache.h>
12 #include <linux/mm.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>
19 #include "erofs_fs.h"
20 
21 /* redefine pr_fmt "erofs: " */
22 #undef pr_fmt
23 #define pr_fmt(fmt) "erofs: " fmt
24 
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 DBG_BUGON               BUG_ON
35 #else
36 #define DBG_BUGON(x)            ((void)(x))
37 #endif	/* !CONFIG_EROFS_FS_DEBUG */
38 
39 /* EROFS_SUPER_MAGIC_V1 to represent the whole file system */
40 #define EROFS_SUPER_MAGIC   EROFS_SUPER_MAGIC_V1
41 
42 typedef u64 erofs_nid_t;
43 typedef u64 erofs_off_t;
44 /* data type for filesystem-wide blocks number */
45 typedef u32 erofs_blk_t;
46 
47 struct erofs_device_info {
48 	char *path;
49 	struct erofs_fscache *fscache;
50 	struct bdev_handle *bdev_handle;
51 	struct dax_device *dax_dev;
52 	u64 dax_part_off;
53 
54 	u32 blocks;
55 	u32 mapped_blkaddr;
56 };
57 
58 enum {
59 	EROFS_SYNC_DECOMPRESS_AUTO,
60 	EROFS_SYNC_DECOMPRESS_FORCE_ON,
61 	EROFS_SYNC_DECOMPRESS_FORCE_OFF
62 };
63 
64 struct erofs_mount_opts {
65 #ifdef CONFIG_EROFS_FS_ZIP
66 	/* current strategy of how to use managed cache */
67 	unsigned char cache_strategy;
68 	/* strategy of sync decompression (0 - auto, 1 - force on, 2 - force off) */
69 	unsigned int sync_decompress;
70 
71 	/* threshold for decompression synchronously */
72 	unsigned int max_sync_decompress_pages;
73 #endif
74 	unsigned int mount_opt;
75 };
76 
77 struct erofs_dev_context {
78 	struct idr tree;
79 	struct rw_semaphore rwsem;
80 
81 	unsigned int extra_devices;
82 	bool flatdev;
83 };
84 
85 /* all filesystem-wide lz4 configurations */
86 struct erofs_sb_lz4_info {
87 	/* # of pages needed for EROFS lz4 rolling decompression */
88 	u16 max_distance_pages;
89 	/* maximum possible blocks for pclusters in the filesystem */
90 	u16 max_pclusterblks;
91 };
92 
93 struct erofs_domain {
94 	refcount_t ref;
95 	struct list_head list;
96 	struct fscache_volume *volume;
97 	char *domain_id;
98 };
99 
100 struct erofs_fscache {
101 	struct fscache_cookie *cookie;
102 	struct inode *inode;	/* anonymous inode for the blob */
103 
104 	/* used for share domain mode */
105 	struct erofs_domain *domain;
106 	struct list_head node;
107 	refcount_t ref;
108 	char *name;
109 };
110 
111 struct erofs_xattr_prefix_item {
112 	struct erofs_xattr_long_prefix *prefix;
113 	u8 infix_len;
114 };
115 
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;
122 
123 	/* managed XArray arranged in physical block number */
124 	struct xarray managed_pslots;
125 
126 	unsigned int shrinker_run_no;
127 	u16 available_compr_algs;
128 
129 	/* pseudo inode to manage cached pages */
130 	struct inode *managed_cache;
131 
132 	struct erofs_sb_lz4_info lz4;
133 #endif	/* CONFIG_EROFS_FS_ZIP */
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_fscache_mode(struct super_block *sb)
195 {
196 	return IS_ENABLED(CONFIG_EROFS_FS_ONDEMAND) && !sb->s_bdev;
197 }
198 
199 enum {
200 	EROFS_ZIP_CACHE_DISABLED,
201 	EROFS_ZIP_CACHE_READAHEAD,
202 	EROFS_ZIP_CACHE_READAROUND
203 };
204 
205 /* basic unit of the workstation of a super_block */
206 struct erofs_workgroup {
207 	pgoff_t index;
208 	struct lockref lockref;
209 };
210 
211 enum erofs_kmap_type {
212 	EROFS_NO_KMAP,		/* don't map the buffer */
213 	EROFS_KMAP,		/* use kmap_local_page() to map the buffer */
214 };
215 
216 struct erofs_buf {
217 	struct inode *inode;
218 	struct page *page;
219 	void *base;
220 	enum erofs_kmap_type kmap_type;
221 };
222 #define __EROFS_BUF_INITIALIZER	((struct erofs_buf){ .page = NULL })
223 
224 #define ROOT_NID(sb)		((sb)->root_nid)
225 
226 #define erofs_blknr(sb, addr)	((addr) >> (sb)->s_blocksize_bits)
227 #define erofs_blkoff(sb, addr)	((addr) & ((sb)->s_blocksize - 1))
228 #define erofs_pos(sb, blk)	((erofs_off_t)(blk) << (sb)->s_blocksize_bits)
229 #define erofs_iblks(i)	(round_up((i)->i_size, i_blocksize(i)) >> (i)->i_blkbits)
230 
231 #define EROFS_FEATURE_FUNCS(name, compat, feature) \
232 static inline bool erofs_sb_has_##name(struct erofs_sb_info *sbi) \
233 { \
234 	return sbi->feature_##compat & EROFS_FEATURE_##feature; \
235 }
236 
237 EROFS_FEATURE_FUNCS(zero_padding, incompat, INCOMPAT_ZERO_PADDING)
238 EROFS_FEATURE_FUNCS(compr_cfgs, incompat, INCOMPAT_COMPR_CFGS)
239 EROFS_FEATURE_FUNCS(big_pcluster, incompat, INCOMPAT_BIG_PCLUSTER)
240 EROFS_FEATURE_FUNCS(chunked_file, incompat, INCOMPAT_CHUNKED_FILE)
241 EROFS_FEATURE_FUNCS(device_table, incompat, INCOMPAT_DEVICE_TABLE)
242 EROFS_FEATURE_FUNCS(compr_head2, incompat, INCOMPAT_COMPR_HEAD2)
243 EROFS_FEATURE_FUNCS(ztailpacking, incompat, INCOMPAT_ZTAILPACKING)
244 EROFS_FEATURE_FUNCS(fragments, incompat, INCOMPAT_FRAGMENTS)
245 EROFS_FEATURE_FUNCS(dedupe, incompat, INCOMPAT_DEDUPE)
246 EROFS_FEATURE_FUNCS(xattr_prefixes, incompat, INCOMPAT_XATTR_PREFIXES)
247 EROFS_FEATURE_FUNCS(sb_chksum, compat, COMPAT_SB_CHKSUM)
248 EROFS_FEATURE_FUNCS(xattr_filter, compat, COMPAT_XATTR_FILTER)
249 
250 /* atomic flag definitions */
251 #define EROFS_I_EA_INITED_BIT	0
252 #define EROFS_I_Z_INITED_BIT	1
253 
254 /* bitlock definitions (arranged in reverse order) */
255 #define EROFS_I_BL_XATTR_BIT	(BITS_PER_LONG - 1)
256 #define EROFS_I_BL_Z_BIT	(BITS_PER_LONG - 2)
257 
258 struct erofs_inode {
259 	erofs_nid_t nid;
260 
261 	/* atomic flags (including bitlocks) */
262 	unsigned long flags;
263 
264 	unsigned char datalayout;
265 	unsigned char inode_isize;
266 	unsigned int xattr_isize;
267 
268 	unsigned int xattr_name_filter;
269 	unsigned int xattr_shared_count;
270 	unsigned int *xattr_shared_xattrs;
271 
272 	union {
273 		erofs_blk_t raw_blkaddr;
274 		struct {
275 			unsigned short	chunkformat;
276 			unsigned char	chunkbits;
277 		};
278 #ifdef CONFIG_EROFS_FS_ZIP
279 		struct {
280 			unsigned short z_advise;
281 			unsigned char  z_algorithmtype[2];
282 			unsigned char  z_logical_clusterbits;
283 			unsigned long  z_tailextent_headlcn;
284 			union {
285 				struct {
286 					erofs_off_t    z_idataoff;
287 					unsigned short z_idata_size;
288 				};
289 				erofs_off_t z_fragmentoff;
290 			};
291 		};
292 #endif	/* CONFIG_EROFS_FS_ZIP */
293 	};
294 	/* the corresponding vfs inode */
295 	struct inode vfs_inode;
296 };
297 
298 #define EROFS_I(ptr)	container_of(ptr, struct erofs_inode, vfs_inode)
299 
300 static inline erofs_off_t erofs_iloc(struct inode *inode)
301 {
302 	struct erofs_sb_info *sbi = EROFS_I_SB(inode);
303 
304 	return erofs_pos(inode->i_sb, sbi->meta_blkaddr) +
305 		(EROFS_I(inode)->nid << sbi->islotbits);
306 }
307 
308 static inline unsigned int erofs_inode_version(unsigned int ifmt)
309 {
310 	return (ifmt >> EROFS_I_VERSION_BIT) & EROFS_I_VERSION_MASK;
311 }
312 
313 static inline unsigned int erofs_inode_datalayout(unsigned int ifmt)
314 {
315 	return (ifmt >> EROFS_I_DATALAYOUT_BIT) & EROFS_I_DATALAYOUT_MASK;
316 }
317 
318 /*
319  * Different from grab_cache_page_nowait(), reclaiming is never triggered
320  * when allocating new pages.
321  */
322 static inline
323 struct page *erofs_grab_cache_page_nowait(struct address_space *mapping,
324 					  pgoff_t index)
325 {
326 	return pagecache_get_page(mapping, index,
327 			FGP_LOCK|FGP_CREAT|FGP_NOFS|FGP_NOWAIT,
328 			readahead_gfp_mask(mapping) & ~__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 	u64 m_dax_part_off;
376 
377 	erofs_off_t m_pa;
378 	unsigned int m_deviceid;
379 };
380 
381 extern const struct super_operations erofs_sops;
382 
383 extern const struct address_space_operations erofs_raw_access_aops;
384 extern const struct address_space_operations z_erofs_aops;
385 extern const struct address_space_operations erofs_fscache_access_aops;
386 
387 extern const struct inode_operations erofs_generic_iops;
388 extern const struct inode_operations erofs_symlink_iops;
389 extern const struct inode_operations erofs_fast_symlink_iops;
390 extern const struct inode_operations erofs_dir_iops;
391 
392 extern const struct file_operations erofs_file_fops;
393 extern const struct file_operations erofs_dir_fops;
394 
395 extern const struct iomap_ops z_erofs_iomap_report_ops;
396 
397 /* flags for erofs_fscache_register_cookie() */
398 #define EROFS_REG_COOKIE_SHARE		0x0001
399 #define EROFS_REG_COOKIE_NEED_NOEXIST	0x0002
400 
401 void *erofs_read_metadata(struct super_block *sb, struct erofs_buf *buf,
402 			  erofs_off_t *offset, int *lengthp);
403 void erofs_unmap_metabuf(struct erofs_buf *buf);
404 void erofs_put_metabuf(struct erofs_buf *buf);
405 void *erofs_bread(struct erofs_buf *buf, erofs_blk_t blkaddr,
406 		  enum erofs_kmap_type type);
407 void erofs_init_metabuf(struct erofs_buf *buf, struct super_block *sb);
408 void *erofs_read_metabuf(struct erofs_buf *buf, struct super_block *sb,
409 			 erofs_blk_t blkaddr, enum erofs_kmap_type type);
410 int erofs_map_dev(struct super_block *sb, struct erofs_map_dev *dev);
411 int erofs_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
412 		 u64 start, u64 len);
413 int erofs_map_blocks(struct inode *inode, struct erofs_map_blocks *map);
414 struct inode *erofs_iget(struct super_block *sb, erofs_nid_t nid);
415 int erofs_getattr(struct mnt_idmap *idmap, const struct path *path,
416 		  struct kstat *stat, u32 request_mask,
417 		  unsigned int query_flags);
418 int erofs_namei(struct inode *dir, const struct qstr *name,
419 		erofs_nid_t *nid, unsigned int *d_type);
420 
421 static inline void *erofs_vm_map_ram(struct page **pages, unsigned int count)
422 {
423 	int retried = 0;
424 
425 	while (1) {
426 		void *p = vm_map_ram(pages, count, -1);
427 
428 		/* retry two more times (totally 3 times) */
429 		if (p || ++retried >= 3)
430 			return p;
431 		vm_unmap_aliases();
432 	}
433 	return NULL;
434 }
435 
436 int erofs_register_sysfs(struct super_block *sb);
437 void erofs_unregister_sysfs(struct super_block *sb);
438 int __init erofs_init_sysfs(void);
439 void erofs_exit_sysfs(void);
440 
441 struct page *erofs_allocpage(struct page **pagepool, gfp_t gfp);
442 static inline void erofs_pagepool_add(struct page **pagepool, struct page *page)
443 {
444 	set_page_private(page, (unsigned long)*pagepool);
445 	*pagepool = page;
446 }
447 void erofs_release_pages(struct page **pagepool);
448 
449 #ifdef CONFIG_EROFS_FS_ZIP
450 void erofs_workgroup_put(struct erofs_workgroup *grp);
451 struct erofs_workgroup *erofs_find_workgroup(struct super_block *sb,
452 					     pgoff_t index);
453 struct erofs_workgroup *erofs_insert_workgroup(struct super_block *sb,
454 					       struct erofs_workgroup *grp);
455 void erofs_workgroup_free_rcu(struct erofs_workgroup *grp);
456 void erofs_shrinker_register(struct super_block *sb);
457 void erofs_shrinker_unregister(struct super_block *sb);
458 int __init erofs_init_shrinker(void);
459 void erofs_exit_shrinker(void);
460 int __init z_erofs_init_zip_subsystem(void);
461 void z_erofs_exit_zip_subsystem(void);
462 int erofs_try_to_free_all_cached_pages(struct erofs_sb_info *sbi,
463 				       struct erofs_workgroup *egrp);
464 int z_erofs_map_blocks_iter(struct inode *inode, struct erofs_map_blocks *map,
465 			    int flags);
466 void *erofs_get_pcpubuf(unsigned int requiredpages);
467 void erofs_put_pcpubuf(void *ptr);
468 int erofs_pcpubuf_growsize(unsigned int nrpages);
469 void __init erofs_pcpubuf_init(void);
470 void erofs_pcpubuf_exit(void);
471 int erofs_init_managed_cache(struct super_block *sb);
472 int z_erofs_parse_cfgs(struct super_block *sb, struct erofs_super_block *dsb);
473 #else
474 static inline void erofs_shrinker_register(struct super_block *sb) {}
475 static inline void erofs_shrinker_unregister(struct super_block *sb) {}
476 static inline int erofs_init_shrinker(void) { return 0; }
477 static inline void erofs_exit_shrinker(void) {}
478 static inline int z_erofs_init_zip_subsystem(void) { return 0; }
479 static inline void z_erofs_exit_zip_subsystem(void) {}
480 static inline void erofs_pcpubuf_init(void) {}
481 static inline void erofs_pcpubuf_exit(void) {}
482 static inline int erofs_init_managed_cache(struct super_block *sb) { return 0; }
483 #endif	/* !CONFIG_EROFS_FS_ZIP */
484 
485 #ifdef CONFIG_EROFS_FS_ZIP_LZMA
486 int __init z_erofs_lzma_init(void);
487 void z_erofs_lzma_exit(void);
488 #else
489 static inline int z_erofs_lzma_init(void) { return 0; }
490 static inline int z_erofs_lzma_exit(void) { return 0; }
491 #endif	/* !CONFIG_EROFS_FS_ZIP_LZMA */
492 
493 #ifdef CONFIG_EROFS_FS_ZIP_DEFLATE
494 int __init z_erofs_deflate_init(void);
495 void z_erofs_deflate_exit(void);
496 #else
497 static inline int z_erofs_deflate_init(void) { return 0; }
498 static inline int z_erofs_deflate_exit(void) { return 0; }
499 #endif	/* !CONFIG_EROFS_FS_ZIP_DEFLATE */
500 
501 #ifdef CONFIG_EROFS_FS_ONDEMAND
502 int erofs_fscache_register_fs(struct super_block *sb);
503 void erofs_fscache_unregister_fs(struct super_block *sb);
504 
505 struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
506 					char *name, unsigned int flags);
507 void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache);
508 #else
509 static inline int erofs_fscache_register_fs(struct super_block *sb)
510 {
511 	return -EOPNOTSUPP;
512 }
513 static inline void erofs_fscache_unregister_fs(struct super_block *sb) {}
514 
515 static inline
516 struct erofs_fscache *erofs_fscache_register_cookie(struct super_block *sb,
517 					char *name, unsigned int flags)
518 {
519 	return ERR_PTR(-EOPNOTSUPP);
520 }
521 
522 static inline void erofs_fscache_unregister_cookie(struct erofs_fscache *fscache)
523 {
524 }
525 #endif
526 
527 #define EFSCORRUPTED    EUCLEAN         /* Filesystem is corrupted */
528 
529 #endif	/* __EROFS_INTERNAL_H */
530