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_FEATURE_INCOMPAT_48BIT 0x00000080
34 #define EROFS_ALL_FEATURE_INCOMPAT \
35 ((EROFS_FEATURE_INCOMPAT_48BIT << 1) - 1)
36
37 #define EROFS_SB_EXTSLOT_SIZE 16
38
39 struct erofs_deviceslot {
40 u8 tag[64]; /* digest(sha256), etc. */
41 __le32 blocks_lo; /* total blocks count of this device */
42 __le32 uniaddr_lo; /* unified starting block of this device */
43 __le32 blocks_hi; /* total blocks count MSB */
44 __le16 uniaddr_hi; /* unified starting block MSB */
45 u8 reserved[50];
46 };
47 #define EROFS_DEVT_SLOT_SIZE sizeof(struct erofs_deviceslot)
48
49 /* erofs on-disk super block (currently 128 bytes) */
50 struct erofs_super_block {
51 __le32 magic; /* file system magic number */
52 __le32 checksum; /* crc32c to avoid unexpected on-disk overlap */
53 __le32 feature_compat;
54 __u8 blkszbits; /* filesystem block size in bit shift */
55 __u8 sb_extslots; /* superblock size = 128 + sb_extslots * 16 */
56 union {
57 __le16 rootnid_2b; /* nid of root directory */
58 __le16 blocks_hi; /* (48BIT on) blocks count MSB */
59 } rb;
60 __le64 inos; /* total valid ino # (== f_files - f_favail) */
61 __le64 epoch; /* base seconds used for compact inodes */
62 __le32 fixed_nsec; /* fixed nanoseconds for compact inodes */
63 __le32 blocks_lo; /* blocks count LSB */
64 __le32 meta_blkaddr; /* start block address of metadata area */
65 __le32 xattr_blkaddr; /* start block address of shared xattr area */
66 __u8 uuid[16]; /* 128-bit uuid for volume */
67 __u8 volume_name[16]; /* volume name */
68 __le32 feature_incompat;
69 union {
70 /* bitmap for available compression algorithms */
71 __le16 available_compr_algs;
72 /* customized sliding window size instead of 64k by default */
73 __le16 lz4_max_distance;
74 } __packed u1;
75 __le16 extra_devices; /* # of devices besides the primary device */
76 __le16 devt_slotoff; /* startoff = devt_slotoff * devt_slotsize */
77 __u8 dirblkbits; /* directory block size in bit shift */
78 __u8 xattr_prefix_count; /* # of long xattr name prefixes */
79 __le32 xattr_prefix_start; /* start of long xattr prefixes */
80 __le64 packed_nid; /* nid of the special packed inode */
81 __u8 xattr_filter_reserved; /* reserved for xattr name filter */
82 __u8 reserved[3];
83 __le32 build_time; /* seconds added to epoch for mkfs time */
84 __le64 rootnid_8b; /* (48BIT on) nid of root directory */
85 __u8 reserved2[8];
86 };
87
88 /*
89 * EROFS inode datalayout (i_format in on-disk inode):
90 * 0 - uncompressed flat inode without tail-packing inline data:
91 * 1 - compressed inode with non-compact indexes:
92 * 2 - uncompressed flat inode with tail-packing inline data:
93 * 3 - compressed inode with compact indexes:
94 * 4 - chunk-based inode with (optional) multi-device support:
95 * 5~7 - reserved
96 */
97 enum {
98 EROFS_INODE_FLAT_PLAIN = 0,
99 EROFS_INODE_COMPRESSED_FULL = 1,
100 EROFS_INODE_FLAT_INLINE = 2,
101 EROFS_INODE_COMPRESSED_COMPACT = 3,
102 EROFS_INODE_CHUNK_BASED = 4,
103 EROFS_INODE_DATALAYOUT_MAX
104 };
105
erofs_inode_is_data_compressed(unsigned int datamode)106 static inline bool erofs_inode_is_data_compressed(unsigned int datamode)
107 {
108 return datamode == EROFS_INODE_COMPRESSED_COMPACT ||
109 datamode == EROFS_INODE_COMPRESSED_FULL;
110 }
111
112 /* bit definitions of inode i_format */
113 #define EROFS_I_VERSION_MASK 0x01
114 #define EROFS_I_DATALAYOUT_MASK 0x07
115
116 #define EROFS_I_VERSION_BIT 0
117 #define EROFS_I_DATALAYOUT_BIT 1
118 #define EROFS_I_NLINK_1_BIT 4 /* non-directory compact inodes only */
119 #define EROFS_I_DOT_OMITTED_BIT 4 /* (directories) omit the `.` dirent */
120 #define EROFS_I_ALL ((1 << (EROFS_I_NLINK_1_BIT + 1)) - 1)
121
122 /* indicate chunk blkbits, thus 'chunksize = blocksize << chunk blkbits' */
123 #define EROFS_CHUNK_FORMAT_BLKBITS_MASK 0x001F
124 /* with chunk indexes or just a 4-byte block array */
125 #define EROFS_CHUNK_FORMAT_INDEXES 0x0020
126 #define EROFS_CHUNK_FORMAT_48BIT 0x0040
127
128 #define EROFS_CHUNK_FORMAT_ALL ((EROFS_CHUNK_FORMAT_48BIT << 1) - 1)
129
130 /* 32-byte on-disk inode */
131 #define EROFS_INODE_LAYOUT_COMPACT 0
132 /* 64-byte on-disk inode */
133 #define EROFS_INODE_LAYOUT_EXTENDED 1
134
135 struct erofs_inode_chunk_info {
136 __le16 format; /* chunk blkbits, etc. */
137 __le16 reserved;
138 };
139
140 union erofs_inode_i_u {
141 __le32 blocks_lo; /* total blocks count (if compressed inodes) */
142 __le32 startblk_lo; /* starting block number (if flat inodes) */
143 __le32 rdev; /* device ID (if special inodes) */
144 struct erofs_inode_chunk_info c;
145 };
146
147 union erofs_inode_i_nb {
148 __le16 nlink; /* if EROFS_I_NLINK_1_BIT is unset */
149 __le16 blocks_hi; /* total blocks count MSB */
150 __le16 startblk_hi; /* starting block number MSB */
151 };
152
153 /* 32-byte reduced form of an ondisk inode */
154 struct erofs_inode_compact {
155 __le16 i_format; /* inode format hints */
156 __le16 i_xattr_icount;
157 __le16 i_mode;
158 union erofs_inode_i_nb i_nb;
159 __le32 i_size;
160 __le32 i_mtime;
161 union erofs_inode_i_u i_u;
162
163 __le32 i_ino; /* only used for 32-bit stat compatibility */
164 __le16 i_uid;
165 __le16 i_gid;
166 __le32 i_reserved;
167 };
168
169 /* 64-byte complete form of an ondisk inode */
170 struct erofs_inode_extended {
171 __le16 i_format; /* inode format hints */
172 __le16 i_xattr_icount;
173 __le16 i_mode;
174 union erofs_inode_i_nb i_nb;
175 __le64 i_size;
176 union erofs_inode_i_u i_u;
177
178 __le32 i_ino; /* only used for 32-bit stat compatibility */
179 __le32 i_uid;
180 __le32 i_gid;
181 __le64 i_mtime;
182 __le32 i_mtime_nsec;
183 __le32 i_nlink;
184 __u8 i_reserved2[16];
185 };
186
187 /*
188 * inline xattrs (n == i_xattr_icount):
189 * erofs_xattr_ibody_header(1) + (n - 1) * 4 bytes
190 * 12 bytes / \
191 * / \
192 * /-----------------------\
193 * | erofs_xattr_entries+ |
194 * +-----------------------+
195 * inline xattrs must starts in erofs_xattr_ibody_header,
196 * for read-only fs, no need to introduce h_refcount
197 */
198 struct erofs_xattr_ibody_header {
199 __le32 h_name_filter; /* bit value 1 indicates not-present */
200 __u8 h_shared_count;
201 __u8 h_reserved2[7];
202 __le32 h_shared_xattrs[]; /* shared xattr id array */
203 };
204
205 /* Name indexes */
206 #define EROFS_XATTR_INDEX_USER 1
207 #define EROFS_XATTR_INDEX_POSIX_ACL_ACCESS 2
208 #define EROFS_XATTR_INDEX_POSIX_ACL_DEFAULT 3
209 #define EROFS_XATTR_INDEX_TRUSTED 4
210 #define EROFS_XATTR_INDEX_LUSTRE 5
211 #define EROFS_XATTR_INDEX_SECURITY 6
212
213 /*
214 * bit 7 of e_name_index is set when it refers to a long xattr name prefix,
215 * while the remained lower bits represent the index of the prefix.
216 */
217 #define EROFS_XATTR_LONG_PREFIX 0x80
218 #define EROFS_XATTR_LONG_PREFIX_MASK 0x7f
219
220 #define EROFS_XATTR_FILTER_BITS 32
221 #define EROFS_XATTR_FILTER_DEFAULT UINT32_MAX
222 #define EROFS_XATTR_FILTER_SEED 0x25BBE08F
223
224 /* xattr entry (for both inline & shared xattrs) */
225 struct erofs_xattr_entry {
226 __u8 e_name_len; /* length of name */
227 __u8 e_name_index; /* attribute name index */
228 __le16 e_value_size; /* size of attribute value */
229 /* followed by e_name and e_value */
230 char e_name[]; /* attribute name */
231 };
232
233 /* long xattr name prefix */
234 struct erofs_xattr_long_prefix {
235 __u8 base_index; /* short xattr name prefix index */
236 char infix[]; /* infix apart from short prefix */
237 };
238
erofs_xattr_ibody_size(__le16 i_xattr_icount)239 static inline unsigned int erofs_xattr_ibody_size(__le16 i_xattr_icount)
240 {
241 if (!i_xattr_icount)
242 return 0;
243
244 /* 1 header + n-1 * 4 bytes inline xattr to keep continuity */
245 return sizeof(struct erofs_xattr_ibody_header) +
246 sizeof(__u32) * (le16_to_cpu(i_xattr_icount) - 1);
247 }
248
249 #define EROFS_XATTR_ALIGN(size) round_up(size, sizeof(struct erofs_xattr_entry))
250
erofs_xattr_entry_size(struct erofs_xattr_entry * e)251 static inline unsigned int erofs_xattr_entry_size(struct erofs_xattr_entry *e)
252 {
253 return EROFS_XATTR_ALIGN(sizeof(struct erofs_xattr_entry) +
254 e->e_name_len + le16_to_cpu(e->e_value_size));
255 }
256
257 /* represent a zeroed chunk (hole) */
258 #define EROFS_NULL_ADDR -1
259
260 /* 4-byte block address array */
261 #define EROFS_BLOCK_MAP_ENTRY_SIZE sizeof(__le32)
262
263 /* 8-byte inode chunk index */
264 struct erofs_inode_chunk_index {
265 __le16 startblk_hi; /* starting block number MSB */
266 __le16 device_id; /* back-end storage id (with bits masked) */
267 __le32 startblk_lo; /* starting block number of this chunk */
268 };
269
270 /* dirent sorts in alphabet order, thus we can do binary search */
271 struct erofs_dirent {
272 __le64 nid; /* node number */
273 __le16 nameoff; /* start offset of file name */
274 __u8 file_type; /* file type */
275 __u8 reserved; /* reserved */
276 } __packed;
277
278 /*
279 * EROFS file types should match generic FT_* types and
280 * it seems no need to add BUILD_BUG_ONs since potential
281 * unmatchness will break other fses as well...
282 */
283
284 #define EROFS_NAME_LEN 255
285
286 /* maximum supported encoded size of a physical compressed cluster */
287 #define Z_EROFS_PCLUSTER_MAX_SIZE (1024 * 1024)
288
289 /* maximum supported decoded size of a physical compressed cluster */
290 #define Z_EROFS_PCLUSTER_MAX_DSIZE (12 * 1024 * 1024)
291
292 /* available compression algorithm types (for h_algorithmtype) */
293 enum {
294 Z_EROFS_COMPRESSION_LZ4 = 0,
295 Z_EROFS_COMPRESSION_LZMA = 1,
296 Z_EROFS_COMPRESSION_DEFLATE = 2,
297 Z_EROFS_COMPRESSION_ZSTD = 3,
298 Z_EROFS_COMPRESSION_MAX
299 };
300 #define Z_EROFS_ALL_COMPR_ALGS ((1 << Z_EROFS_COMPRESSION_MAX) - 1)
301
302 /* 14 bytes (+ length field = 16 bytes) */
303 struct z_erofs_lz4_cfgs {
304 __le16 max_distance;
305 __le16 max_pclusterblks;
306 u8 reserved[10];
307 } __packed;
308
309 /* 14 bytes (+ length field = 16 bytes) */
310 struct z_erofs_lzma_cfgs {
311 __le32 dict_size;
312 __le16 format;
313 u8 reserved[8];
314 } __packed;
315
316 #define Z_EROFS_LZMA_MAX_DICT_SIZE (8 * Z_EROFS_PCLUSTER_MAX_SIZE)
317
318 /* 6 bytes (+ length field = 8 bytes) */
319 struct z_erofs_deflate_cfgs {
320 u8 windowbits; /* 8..15 for DEFLATE */
321 u8 reserved[5];
322 } __packed;
323
324 /* 6 bytes (+ length field = 8 bytes) */
325 struct z_erofs_zstd_cfgs {
326 u8 format;
327 u8 windowlog; /* windowLog - ZSTD_WINDOWLOG_ABSOLUTEMIN(10) */
328 u8 reserved[4];
329 } __packed;
330
331 #define Z_EROFS_ZSTD_MAX_DICT_SIZE Z_EROFS_PCLUSTER_MAX_SIZE
332
333 /*
334 * Enable COMPACTED_2B for EROFS_INODE_COMPRESSED_COMPACT inodes:
335 * 4B (disabled) vs 4B+2B+4B (enabled)
336 */
337 #define Z_EROFS_ADVISE_COMPACTED_2B 0x0001
338 /* Enable extent metadata for EROFS_INODE_COMPRESSED_FULL inodes */
339 #define Z_EROFS_ADVISE_EXTENTS 0x0001
340 #define Z_EROFS_ADVISE_BIG_PCLUSTER_1 0x0002
341 #define Z_EROFS_ADVISE_BIG_PCLUSTER_2 0x0004
342 #define Z_EROFS_ADVISE_INLINE_PCLUSTER 0x0008
343 #define Z_EROFS_ADVISE_INTERLACED_PCLUSTER 0x0010
344 #define Z_EROFS_ADVISE_FRAGMENT_PCLUSTER 0x0020
345 /* Indicate the record size for each extent if extent metadata is used */
346 #define Z_EROFS_ADVISE_EXTRECSZ_BIT 1
347 #define Z_EROFS_ADVISE_EXTRECSZ_MASK 0x3
348
349 #define Z_EROFS_FRAGMENT_INODE_BIT 7
350 struct z_erofs_map_header {
351 union {
352 /* fragment data offset in the packed inode */
353 __le32 h_fragmentoff;
354 struct {
355 __le16 h_reserved1;
356 /* indicates the encoded size of tailpacking data */
357 __le16 h_idata_size;
358 };
359 __le32 h_extents_lo; /* extent count LSB */
360 };
361 __le16 h_advise;
362 union {
363 struct {
364 /* algorithm type (bit 0-3: HEAD1; bit 4-7: HEAD2) */
365 __u8 h_algorithmtype;
366 /*
367 * bit 0-3 : logical cluster bits - blkszbits
368 * bit 4-6 : reserved
369 * bit 7 : pack the whole file into packed inode
370 */
371 __u8 h_clusterbits;
372 };
373 __le16 h_extents_hi; /* extent count MSB */
374 };
375 };
376
377 enum {
378 Z_EROFS_LCLUSTER_TYPE_PLAIN = 0,
379 Z_EROFS_LCLUSTER_TYPE_HEAD1 = 1,
380 Z_EROFS_LCLUSTER_TYPE_NONHEAD = 2,
381 Z_EROFS_LCLUSTER_TYPE_HEAD2 = 3,
382 Z_EROFS_LCLUSTER_TYPE_MAX
383 };
384
385 #define Z_EROFS_LI_LCLUSTER_TYPE_MASK (Z_EROFS_LCLUSTER_TYPE_MAX - 1)
386
387 /* (noncompact only, HEAD) This pcluster refers to partial decompressed data */
388 #define Z_EROFS_LI_PARTIAL_REF (1 << 15)
389
390 /* Set on 1st non-head lcluster to store compressed block counti (in blocks) */
391 #define Z_EROFS_LI_D0_CBLKCNT (1 << 11)
392
393 struct z_erofs_lcluster_index {
394 __le16 di_advise;
395 /* where to decompress in the head lcluster */
396 __le16 di_clusterofs;
397
398 union {
399 __le32 blkaddr; /* for the HEAD lclusters */
400 /*
401 * [0] - distance to its HEAD lcluster
402 * [1] - distance to the next HEAD lcluster
403 */
404 __le16 delta[2]; /* for the NONHEAD lclusters */
405 } di_u;
406 };
407
408 #define Z_EROFS_MAP_HEADER_END(end) \
409 (ALIGN(end, 8) + sizeof(struct z_erofs_map_header))
410 #define Z_EROFS_FULL_INDEX_START(end) (Z_EROFS_MAP_HEADER_END(end) + 8)
411
412 #define Z_EROFS_EXTENT_PLEN_PARTIAL BIT(27)
413 #define Z_EROFS_EXTENT_PLEN_FMT_BIT 28
414 #define Z_EROFS_EXTENT_PLEN_MASK ((Z_EROFS_PCLUSTER_MAX_SIZE << 1) - 1)
415 struct z_erofs_extent {
416 __le32 plen; /* encoded length */
417 __le32 pstart_lo; /* physical offset */
418 __le32 pstart_hi; /* physical offset MSB */
419 __le32 lstart_lo; /* logical offset */
420 __le32 lstart_hi; /* logical offset MSB (>= 4GiB inodes) */
421 __u8 reserved[12]; /* for future use */
422 };
423
z_erofs_extent_recsize(unsigned int advise)424 static inline int z_erofs_extent_recsize(unsigned int advise)
425 {
426 return 4 << ((advise >> Z_EROFS_ADVISE_EXTRECSZ_BIT) &
427 Z_EROFS_ADVISE_EXTRECSZ_MASK);
428 }
429
430 /* check the EROFS on-disk layout strictly at compile time */
erofs_check_ondisk_layout_definitions(void)431 static inline void erofs_check_ondisk_layout_definitions(void)
432 {
433 const __le64 fmh = *(__le64 *)&(struct z_erofs_map_header) {
434 .h_clusterbits = 1 << Z_EROFS_FRAGMENT_INODE_BIT
435 };
436
437 BUILD_BUG_ON(sizeof(struct erofs_super_block) != 128);
438 BUILD_BUG_ON(sizeof(struct erofs_inode_compact) != 32);
439 BUILD_BUG_ON(sizeof(struct erofs_inode_extended) != 64);
440 BUILD_BUG_ON(sizeof(struct erofs_xattr_ibody_header) != 12);
441 BUILD_BUG_ON(sizeof(struct erofs_xattr_entry) != 4);
442 BUILD_BUG_ON(sizeof(struct erofs_inode_chunk_info) != 4);
443 BUILD_BUG_ON(sizeof(struct erofs_inode_chunk_index) != 8);
444 BUILD_BUG_ON(sizeof(struct z_erofs_map_header) != 8);
445 BUILD_BUG_ON(sizeof(struct z_erofs_lcluster_index) != 8);
446 BUILD_BUG_ON(sizeof(struct erofs_dirent) != 12);
447 /* keep in sync between 2 index structures for better extendibility */
448 BUILD_BUG_ON(sizeof(struct erofs_inode_chunk_index) !=
449 sizeof(struct z_erofs_lcluster_index));
450 BUILD_BUG_ON(sizeof(struct erofs_deviceslot) != 128);
451
452 /* exclude old compiler versions like gcc 7.5.0 */
453 BUILD_BUG_ON(__builtin_constant_p(fmh) ?
454 fmh != cpu_to_le64(1ULL << 63) : 0);
455 }
456
457 #endif
458