xref: /linux/fs/erofs/erofs_fs.h (revision c2da8b3f914f83fb9089d26a692eb8f22146ddb9)
1 /* SPDX-License-Identifier: GPL-2.0-only OR Apache-2.0 */
2 /*
3  * EROFS (Enhanced ROM File System) on-disk format definition
4  *
5  * Copyright (C) 2017-2018 HUAWEI, Inc.
6  *             https://www.huawei.com/
7  * Copyright (C) 2021, Alibaba Cloud
8  */
9 #ifndef __EROFS_FS_H
10 #define __EROFS_FS_H
11 
12 /* to allow for x86 boot sectors and other oddities. */
13 #define EROFS_SUPER_OFFSET      1024
14 
15 #define EROFS_FEATURE_COMPAT_SB_CHKSUM          0x00000001
16 #define EROFS_FEATURE_COMPAT_MTIME              0x00000002
17 #define EROFS_FEATURE_COMPAT_XATTR_FILTER	0x00000004
18 
19 /*
20  * Any bits that aren't in EROFS_ALL_FEATURE_INCOMPAT should
21  * be incompatible with this kernel version.
22  */
23 #define EROFS_FEATURE_INCOMPAT_ZERO_PADDING	0x00000001
24 #define EROFS_FEATURE_INCOMPAT_COMPR_CFGS	0x00000002
25 #define EROFS_FEATURE_INCOMPAT_BIG_PCLUSTER	0x00000002
26 #define EROFS_FEATURE_INCOMPAT_CHUNKED_FILE	0x00000004
27 #define EROFS_FEATURE_INCOMPAT_DEVICE_TABLE	0x00000008
28 #define EROFS_FEATURE_INCOMPAT_COMPR_HEAD2	0x00000008
29 #define EROFS_FEATURE_INCOMPAT_ZTAILPACKING	0x00000010
30 #define EROFS_FEATURE_INCOMPAT_FRAGMENTS	0x00000020
31 #define EROFS_FEATURE_INCOMPAT_DEDUPE		0x00000020
32 #define EROFS_FEATURE_INCOMPAT_XATTR_PREFIXES	0x00000040
33 #define EROFS_ALL_FEATURE_INCOMPAT		\
34 	(EROFS_FEATURE_INCOMPAT_ZERO_PADDING | \
35 	 EROFS_FEATURE_INCOMPAT_COMPR_CFGS | \
36 	 EROFS_FEATURE_INCOMPAT_BIG_PCLUSTER | \
37 	 EROFS_FEATURE_INCOMPAT_CHUNKED_FILE | \
38 	 EROFS_FEATURE_INCOMPAT_DEVICE_TABLE | \
39 	 EROFS_FEATURE_INCOMPAT_COMPR_HEAD2 | \
40 	 EROFS_FEATURE_INCOMPAT_ZTAILPACKING | \
41 	 EROFS_FEATURE_INCOMPAT_FRAGMENTS | \
42 	 EROFS_FEATURE_INCOMPAT_DEDUPE | \
43 	 EROFS_FEATURE_INCOMPAT_XATTR_PREFIXES)
44 
45 #define EROFS_SB_EXTSLOT_SIZE	16
46 
47 struct erofs_deviceslot {
48 	u8 tag[64];		/* digest(sha256), etc. */
49 	__le32 blocks;		/* total fs blocks of this device */
50 	__le32 mapped_blkaddr;	/* map starting at mapped_blkaddr */
51 	u8 reserved[56];
52 };
53 #define EROFS_DEVT_SLOT_SIZE	sizeof(struct erofs_deviceslot)
54 
55 /* erofs on-disk super block (currently 128 bytes) */
56 struct erofs_super_block {
57 	__le32 magic;           /* file system magic number */
58 	__le32 checksum;        /* crc32c to avoid unexpected on-disk overlap */
59 	__le32 feature_compat;
60 	__u8 blkszbits;         /* filesystem block size in bit shift */
61 	__u8 sb_extslots;	/* superblock size = 128 + sb_extslots * 16 */
62 
63 	__le16 root_nid;	/* nid of root directory */
64 	__le64 inos;            /* total valid ino # (== f_files - f_favail) */
65 
66 	__le64 build_time;      /* compact inode time derivation */
67 	__le32 build_time_nsec;	/* compact inode time derivation in ns scale */
68 	__le32 blocks;          /* used for statfs */
69 	__le32 meta_blkaddr;	/* start block address of metadata area */
70 	__le32 xattr_blkaddr;	/* start block address of shared xattr area */
71 	__u8 uuid[16];          /* 128-bit uuid for volume */
72 	__u8 volume_name[16];   /* volume name */
73 	__le32 feature_incompat;
74 	union {
75 		/* bitmap for available compression algorithms */
76 		__le16 available_compr_algs;
77 		/* customized sliding window size instead of 64k by default */
78 		__le16 lz4_max_distance;
79 	} __packed u1;
80 	__le16 extra_devices;	/* # of devices besides the primary device */
81 	__le16 devt_slotoff;	/* startoff = devt_slotoff * devt_slotsize */
82 	__u8 dirblkbits;	/* directory block size in bit shift */
83 	__u8 xattr_prefix_count;	/* # of long xattr name prefixes */
84 	__le32 xattr_prefix_start;	/* start of long xattr prefixes */
85 	__le64 packed_nid;	/* nid of the special packed inode */
86 	__u8 xattr_filter_reserved; /* reserved for xattr name filter */
87 	__u8 reserved2[23];
88 };
89 
90 /*
91  * EROFS inode datalayout (i_format in on-disk inode):
92  * 0 - uncompressed flat inode without tail-packing inline data:
93  * 1 - compressed inode with non-compact indexes:
94  * 2 - uncompressed flat inode with tail-packing inline data:
95  * 3 - compressed inode with compact indexes:
96  * 4 - chunk-based inode with (optional) multi-device support:
97  * 5~7 - reserved
98  */
99 enum {
100 	EROFS_INODE_FLAT_PLAIN			= 0,
101 	EROFS_INODE_COMPRESSED_FULL		= 1,
102 	EROFS_INODE_FLAT_INLINE			= 2,
103 	EROFS_INODE_COMPRESSED_COMPACT		= 3,
104 	EROFS_INODE_CHUNK_BASED			= 4,
105 	EROFS_INODE_DATALAYOUT_MAX
106 };
107 
erofs_inode_is_data_compressed(unsigned int datamode)108 static inline bool erofs_inode_is_data_compressed(unsigned int datamode)
109 {
110 	return datamode == EROFS_INODE_COMPRESSED_COMPACT ||
111 		datamode == EROFS_INODE_COMPRESSED_FULL;
112 }
113 
114 /* bit definitions of inode i_format */
115 #define EROFS_I_VERSION_MASK            0x01
116 #define EROFS_I_DATALAYOUT_MASK         0x07
117 
118 #define EROFS_I_VERSION_BIT             0
119 #define EROFS_I_DATALAYOUT_BIT          1
120 #define EROFS_I_ALL_BIT			4
121 
122 #define EROFS_I_ALL	((1 << EROFS_I_ALL_BIT) - 1)
123 
124 /* indicate chunk blkbits, thus 'chunksize = blocksize << chunk blkbits' */
125 #define EROFS_CHUNK_FORMAT_BLKBITS_MASK		0x001F
126 /* with chunk indexes or just a 4-byte blkaddr array */
127 #define EROFS_CHUNK_FORMAT_INDEXES		0x0020
128 
129 #define EROFS_CHUNK_FORMAT_ALL	\
130 	(EROFS_CHUNK_FORMAT_BLKBITS_MASK | EROFS_CHUNK_FORMAT_INDEXES)
131 
132 /* 32-byte on-disk inode */
133 #define EROFS_INODE_LAYOUT_COMPACT	0
134 /* 64-byte on-disk inode */
135 #define EROFS_INODE_LAYOUT_EXTENDED	1
136 
137 struct erofs_inode_chunk_info {
138 	__le16 format;		/* chunk blkbits, etc. */
139 	__le16 reserved;
140 };
141 
142 union erofs_inode_i_u {
143 	/* total compressed blocks for compressed inodes */
144 	__le32 compressed_blocks;
145 
146 	/* block address for uncompressed flat inodes */
147 	__le32 raw_blkaddr;
148 
149 	/* for device files, used to indicate old/new device # */
150 	__le32 rdev;
151 
152 	/* for chunk-based files, it contains the summary info */
153 	struct erofs_inode_chunk_info c;
154 };
155 
156 /* 32-byte reduced form of an ondisk inode */
157 struct erofs_inode_compact {
158 	__le16 i_format;	/* inode format hints */
159 
160 /* 1 header + n-1 * 4 bytes inline xattr to keep continuity */
161 	__le16 i_xattr_icount;
162 	__le16 i_mode;
163 	__le16 i_nlink;
164 	__le32 i_size;
165 	__le32 i_reserved;
166 	union erofs_inode_i_u i_u;
167 
168 	__le32 i_ino;		/* only used for 32-bit stat compatibility */
169 	__le16 i_uid;
170 	__le16 i_gid;
171 	__le32 i_reserved2;
172 };
173 
174 /* 64-byte complete form of an ondisk inode */
175 struct erofs_inode_extended {
176 	__le16 i_format;	/* inode format hints */
177 
178 /* 1 header + n-1 * 4 bytes inline xattr to keep continuity */
179 	__le16 i_xattr_icount;
180 	__le16 i_mode;
181 	__le16 i_reserved;
182 	__le64 i_size;
183 	union erofs_inode_i_u i_u;
184 
185 	__le32 i_ino;		/* only used for 32-bit stat compatibility */
186 	__le32 i_uid;
187 	__le32 i_gid;
188 	__le64 i_mtime;
189 	__le32 i_mtime_nsec;
190 	__le32 i_nlink;
191 	__u8   i_reserved2[16];
192 };
193 
194 /*
195  * inline xattrs (n == i_xattr_icount):
196  * erofs_xattr_ibody_header(1) + (n - 1) * 4 bytes
197  *          12 bytes           /                   \
198  *                            /                     \
199  *                           /-----------------------\
200  *                           |  erofs_xattr_entries+ |
201  *                           +-----------------------+
202  * inline xattrs must starts in erofs_xattr_ibody_header,
203  * for read-only fs, no need to introduce h_refcount
204  */
205 struct erofs_xattr_ibody_header {
206 	__le32 h_name_filter;		/* bit value 1 indicates not-present */
207 	__u8   h_shared_count;
208 	__u8   h_reserved2[7];
209 	__le32 h_shared_xattrs[];       /* shared xattr id array */
210 };
211 
212 /* Name indexes */
213 #define EROFS_XATTR_INDEX_USER              1
214 #define EROFS_XATTR_INDEX_POSIX_ACL_ACCESS  2
215 #define EROFS_XATTR_INDEX_POSIX_ACL_DEFAULT 3
216 #define EROFS_XATTR_INDEX_TRUSTED           4
217 #define EROFS_XATTR_INDEX_LUSTRE            5
218 #define EROFS_XATTR_INDEX_SECURITY          6
219 
220 /*
221  * bit 7 of e_name_index is set when it refers to a long xattr name prefix,
222  * while the remained lower bits represent the index of the prefix.
223  */
224 #define EROFS_XATTR_LONG_PREFIX		0x80
225 #define EROFS_XATTR_LONG_PREFIX_MASK	0x7f
226 
227 #define EROFS_XATTR_FILTER_BITS		32
228 #define EROFS_XATTR_FILTER_DEFAULT	UINT32_MAX
229 #define EROFS_XATTR_FILTER_SEED		0x25BBE08F
230 
231 /* xattr entry (for both inline & shared xattrs) */
232 struct erofs_xattr_entry {
233 	__u8   e_name_len;      /* length of name */
234 	__u8   e_name_index;    /* attribute name index */
235 	__le16 e_value_size;    /* size of attribute value */
236 	/* followed by e_name and e_value */
237 	char   e_name[];        /* attribute name */
238 };
239 
240 /* long xattr name prefix */
241 struct erofs_xattr_long_prefix {
242 	__u8   base_index;	/* short xattr name prefix index */
243 	char   infix[];		/* infix apart from short prefix */
244 };
245 
erofs_xattr_ibody_size(__le16 i_xattr_icount)246 static inline unsigned int erofs_xattr_ibody_size(__le16 i_xattr_icount)
247 {
248 	if (!i_xattr_icount)
249 		return 0;
250 
251 	return sizeof(struct erofs_xattr_ibody_header) +
252 		sizeof(__u32) * (le16_to_cpu(i_xattr_icount) - 1);
253 }
254 
255 #define EROFS_XATTR_ALIGN(size) round_up(size, sizeof(struct erofs_xattr_entry))
256 
erofs_xattr_entry_size(struct erofs_xattr_entry * e)257 static inline unsigned int erofs_xattr_entry_size(struct erofs_xattr_entry *e)
258 {
259 	return EROFS_XATTR_ALIGN(sizeof(struct erofs_xattr_entry) +
260 				 e->e_name_len + le16_to_cpu(e->e_value_size));
261 }
262 
263 /* represent a zeroed chunk (hole) */
264 #define EROFS_NULL_ADDR			-1
265 
266 /* 4-byte block address array */
267 #define EROFS_BLOCK_MAP_ENTRY_SIZE	sizeof(__le32)
268 
269 /* 8-byte inode chunk indexes */
270 struct erofs_inode_chunk_index {
271 	__le16 advise;		/* always 0, don't care for now */
272 	__le16 device_id;	/* back-end storage id (with bits masked) */
273 	__le32 blkaddr;		/* start block address of this inode chunk */
274 };
275 
276 /* dirent sorts in alphabet order, thus we can do binary search */
277 struct erofs_dirent {
278 	__le64 nid;     /* node number */
279 	__le16 nameoff; /* start offset of file name */
280 	__u8 file_type; /* file type */
281 	__u8 reserved;  /* reserved */
282 } __packed;
283 
284 /*
285  * EROFS file types should match generic FT_* types and
286  * it seems no need to add BUILD_BUG_ONs since potential
287  * unmatchness will break other fses as well...
288  */
289 
290 #define EROFS_NAME_LEN      255
291 
292 /* maximum supported encoded size of a physical compressed cluster */
293 #define Z_EROFS_PCLUSTER_MAX_SIZE	(1024 * 1024)
294 
295 /* maximum supported decoded size of a physical compressed cluster */
296 #define Z_EROFS_PCLUSTER_MAX_DSIZE	(12 * 1024 * 1024)
297 
298 /* available compression algorithm types (for h_algorithmtype) */
299 enum {
300 	Z_EROFS_COMPRESSION_LZ4		= 0,
301 	Z_EROFS_COMPRESSION_LZMA	= 1,
302 	Z_EROFS_COMPRESSION_DEFLATE	= 2,
303 	Z_EROFS_COMPRESSION_ZSTD	= 3,
304 	Z_EROFS_COMPRESSION_MAX
305 };
306 #define Z_EROFS_ALL_COMPR_ALGS		((1 << Z_EROFS_COMPRESSION_MAX) - 1)
307 
308 /* 14 bytes (+ length field = 16 bytes) */
309 struct z_erofs_lz4_cfgs {
310 	__le16 max_distance;
311 	__le16 max_pclusterblks;
312 	u8 reserved[10];
313 } __packed;
314 
315 /* 14 bytes (+ length field = 16 bytes) */
316 struct z_erofs_lzma_cfgs {
317 	__le32 dict_size;
318 	__le16 format;
319 	u8 reserved[8];
320 } __packed;
321 
322 #define Z_EROFS_LZMA_MAX_DICT_SIZE	(8 * Z_EROFS_PCLUSTER_MAX_SIZE)
323 
324 /* 6 bytes (+ length field = 8 bytes) */
325 struct z_erofs_deflate_cfgs {
326 	u8 windowbits;			/* 8..15 for DEFLATE */
327 	u8 reserved[5];
328 } __packed;
329 
330 /* 6 bytes (+ length field = 8 bytes) */
331 struct z_erofs_zstd_cfgs {
332 	u8 format;
333 	u8 windowlog;           /* windowLog - ZSTD_WINDOWLOG_ABSOLUTEMIN(10) */
334 	u8 reserved[4];
335 } __packed;
336 
337 #define Z_EROFS_ZSTD_MAX_DICT_SIZE      Z_EROFS_PCLUSTER_MAX_SIZE
338 
339 /*
340  * bit 0 : COMPACTED_2B indexes (0 - off; 1 - on)
341  *  e.g. for 4k logical cluster size,      4B        if compacted 2B is off;
342  *                                  (4B) + 2B + (4B) if compacted 2B is on.
343  * bit 1 : HEAD1 big pcluster (0 - off; 1 - on)
344  * bit 2 : HEAD2 big pcluster (0 - off; 1 - on)
345  * bit 3 : tailpacking inline pcluster (0 - off; 1 - on)
346  * bit 4 : interlaced plain pcluster (0 - off; 1 - on)
347  * bit 5 : fragment pcluster (0 - off; 1 - on)
348  */
349 #define Z_EROFS_ADVISE_COMPACTED_2B		0x0001
350 #define Z_EROFS_ADVISE_BIG_PCLUSTER_1		0x0002
351 #define Z_EROFS_ADVISE_BIG_PCLUSTER_2		0x0004
352 #define Z_EROFS_ADVISE_INLINE_PCLUSTER		0x0008
353 #define Z_EROFS_ADVISE_INTERLACED_PCLUSTER	0x0010
354 #define Z_EROFS_ADVISE_FRAGMENT_PCLUSTER	0x0020
355 
356 #define Z_EROFS_FRAGMENT_INODE_BIT              7
357 struct z_erofs_map_header {
358 	union {
359 		/* fragment data offset in the packed inode */
360 		__le32  h_fragmentoff;
361 		struct {
362 			__le16  h_reserved1;
363 			/* indicates the encoded size of tailpacking data */
364 			__le16  h_idata_size;
365 		};
366 	};
367 	__le16	h_advise;
368 	/*
369 	 * bit 0-3 : algorithm type of head 1 (logical cluster type 01);
370 	 * bit 4-7 : algorithm type of head 2 (logical cluster type 11).
371 	 */
372 	__u8	h_algorithmtype;
373 	/*
374 	 * bit 0-2 : logical cluster bits - 12, e.g. 0 for 4096;
375 	 * bit 3-6 : reserved;
376 	 * bit 7   : move the whole file into packed inode or not.
377 	 */
378 	__u8	h_clusterbits;
379 };
380 
381 /*
382  * On-disk logical cluster type:
383  *    0   - literal (uncompressed) lcluster
384  *    1,3 - compressed lcluster (for HEAD lclusters)
385  *    2   - compressed lcluster (for NONHEAD lclusters)
386  *
387  * In detail,
388  *    0 - literal (uncompressed) lcluster,
389  *        di_advise = 0
390  *        di_clusterofs = the literal data offset of the lcluster
391  *        di_blkaddr = the blkaddr of the literal pcluster
392  *
393  *    1,3 - compressed lcluster (for HEAD lclusters)
394  *        di_advise = 1 or 3
395  *        di_clusterofs = the decompressed data offset of the lcluster
396  *        di_blkaddr = the blkaddr of the compressed pcluster
397  *
398  *    2 - compressed lcluster (for NONHEAD lclusters)
399  *        di_advise = 2
400  *        di_clusterofs =
401  *           the decompressed data offset in its own HEAD lcluster
402  *        di_u.delta[0] = distance to this HEAD lcluster
403  *        di_u.delta[1] = distance to the next HEAD lcluster
404  */
405 enum {
406 	Z_EROFS_LCLUSTER_TYPE_PLAIN	= 0,
407 	Z_EROFS_LCLUSTER_TYPE_HEAD1	= 1,
408 	Z_EROFS_LCLUSTER_TYPE_NONHEAD	= 2,
409 	Z_EROFS_LCLUSTER_TYPE_HEAD2	= 3,
410 	Z_EROFS_LCLUSTER_TYPE_MAX
411 };
412 
413 #define Z_EROFS_LI_LCLUSTER_TYPE_MASK	(Z_EROFS_LCLUSTER_TYPE_MAX - 1)
414 
415 /* (noncompact only, HEAD) This pcluster refers to partial decompressed data */
416 #define Z_EROFS_LI_PARTIAL_REF		(1 << 15)
417 
418 /*
419  * D0_CBLKCNT will be marked _only_ at the 1st non-head lcluster to store the
420  * compressed block count of a compressed extent (in logical clusters, aka.
421  * block count of a pcluster).
422  */
423 #define Z_EROFS_LI_D0_CBLKCNT		(1 << 11)
424 
425 struct z_erofs_lcluster_index {
426 	__le16 di_advise;
427 	/* where to decompress in the head lcluster */
428 	__le16 di_clusterofs;
429 
430 	union {
431 		/* for the HEAD lclusters */
432 		__le32 blkaddr;
433 		/*
434 		 * for the NONHEAD lclusters
435 		 * [0] - distance to its HEAD lcluster
436 		 * [1] - distance to the next HEAD lcluster
437 		 */
438 		__le16 delta[2];
439 	} di_u;
440 };
441 
442 #define Z_EROFS_FULL_INDEX_ALIGN(end)	\
443 	(ALIGN(end, 8) + sizeof(struct z_erofs_map_header) + 8)
444 
445 /* check the EROFS on-disk layout strictly at compile time */
erofs_check_ondisk_layout_definitions(void)446 static inline void erofs_check_ondisk_layout_definitions(void)
447 {
448 	const __le64 fmh = *(__le64 *)&(struct z_erofs_map_header) {
449 		.h_clusterbits = 1 << Z_EROFS_FRAGMENT_INODE_BIT
450 	};
451 
452 	BUILD_BUG_ON(sizeof(struct erofs_super_block) != 128);
453 	BUILD_BUG_ON(sizeof(struct erofs_inode_compact) != 32);
454 	BUILD_BUG_ON(sizeof(struct erofs_inode_extended) != 64);
455 	BUILD_BUG_ON(sizeof(struct erofs_xattr_ibody_header) != 12);
456 	BUILD_BUG_ON(sizeof(struct erofs_xattr_entry) != 4);
457 	BUILD_BUG_ON(sizeof(struct erofs_inode_chunk_info) != 4);
458 	BUILD_BUG_ON(sizeof(struct erofs_inode_chunk_index) != 8);
459 	BUILD_BUG_ON(sizeof(struct z_erofs_map_header) != 8);
460 	BUILD_BUG_ON(sizeof(struct z_erofs_lcluster_index) != 8);
461 	BUILD_BUG_ON(sizeof(struct erofs_dirent) != 12);
462 	/* keep in sync between 2 index structures for better extendibility */
463 	BUILD_BUG_ON(sizeof(struct erofs_inode_chunk_index) !=
464 		     sizeof(struct z_erofs_lcluster_index));
465 	BUILD_BUG_ON(sizeof(struct erofs_deviceslot) != 128);
466 
467 	/* exclude old compiler versions like gcc 7.5.0 */
468 	BUILD_BUG_ON(__builtin_constant_p(fmh) ?
469 		     fmh != cpu_to_le64(1ULL << 63) : 0);
470 }
471 
472 #endif
473