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 #define EROFS_SUPER_OFFSET 1024
13
14 #define EROFS_FEATURE_COMPAT_SB_CHKSUM 0x00000001
15 #define EROFS_FEATURE_COMPAT_MTIME 0x00000002
16
17 /*
18 * Any bits that aren't in EROFS_ALL_FEATURE_INCOMPAT should
19 * be incompatible with this kernel version.
20 */
21 #define EROFS_FEATURE_INCOMPAT_ZERO_PADDING 0x00000001
22 #define EROFS_FEATURE_INCOMPAT_COMPR_CFGS 0x00000002
23 #define EROFS_FEATURE_INCOMPAT_BIG_PCLUSTER 0x00000002
24 #define EROFS_FEATURE_INCOMPAT_CHUNKED_FILE 0x00000004
25 #define EROFS_FEATURE_INCOMPAT_DEVICE_TABLE 0x00000008
26 #define EROFS_FEATURE_INCOMPAT_COMPR_HEAD2 0x00000008
27 #define EROFS_FEATURE_INCOMPAT_ZTAILPACKING 0x00000010
28 #define EROFS_FEATURE_INCOMPAT_FRAGMENTS 0x00000020
29 #define EROFS_FEATURE_INCOMPAT_DEDUPE 0x00000020
30 #define EROFS_ALL_FEATURE_INCOMPAT \
31 (EROFS_FEATURE_INCOMPAT_ZERO_PADDING | \
32 EROFS_FEATURE_INCOMPAT_COMPR_CFGS | \
33 EROFS_FEATURE_INCOMPAT_BIG_PCLUSTER | \
34 EROFS_FEATURE_INCOMPAT_CHUNKED_FILE | \
35 EROFS_FEATURE_INCOMPAT_DEVICE_TABLE | \
36 EROFS_FEATURE_INCOMPAT_COMPR_HEAD2 | \
37 EROFS_FEATURE_INCOMPAT_ZTAILPACKING | \
38 EROFS_FEATURE_INCOMPAT_FRAGMENTS | \
39 EROFS_FEATURE_INCOMPAT_DEDUPE)
40
41 #define EROFS_SB_EXTSLOT_SIZE 16
42
43 struct erofs_deviceslot {
44 u8 tag[64]; /* digest(sha256), etc. */
45 __le32 blocks; /* total fs blocks of this device */
46 __le32 mapped_blkaddr; /* map starting at mapped_blkaddr */
47 u8 reserved[56];
48 };
49 #define EROFS_DEVT_SLOT_SIZE sizeof(struct erofs_deviceslot)
50
51 /* erofs on-disk super block (currently 128 bytes) */
52 struct erofs_super_block {
53 __le32 magic; /* file system magic number */
54 __le32 checksum; /* crc32c(super_block) */
55 __le32 feature_compat;
56 __u8 blkszbits; /* filesystem block size in bit shift */
57 __u8 sb_extslots; /* superblock size = 128 + sb_extslots * 16 */
58
59 __le16 root_nid; /* nid of root directory */
60 __le64 inos; /* total valid ino # (== f_files - f_favail) */
61
62 __le64 build_time; /* compact inode time derivation */
63 __le32 build_time_nsec; /* compact inode time derivation in ns scale */
64 __le32 blocks; /* used for statfs */
65 __le32 meta_blkaddr; /* start block address of metadata area */
66 __le32 xattr_blkaddr; /* start block address of shared xattr area */
67 __u8 uuid[16]; /* 128-bit uuid for volume */
68 __u8 volume_name[16]; /* volume name */
69 __le32 feature_incompat;
70 union {
71 /* bitmap for available compression algorithms */
72 __le16 available_compr_algs;
73 /* customized sliding window size instead of 64k by default */
74 __le16 lz4_max_distance;
75 } __packed u1;
76 __le16 extra_devices; /* # of devices besides the primary device */
77 __le16 devt_slotoff; /* startoff = devt_slotoff * devt_slotsize */
78 __u8 dirblkbits; /* directory block size in bit shift */
79 __u8 reserved[5];
80 __le64 packed_nid; /* nid of the special packed inode */
81 __u8 reserved2[24];
82 };
83
84 /*
85 * erofs inode datalayout (i_format in on-disk inode):
86 * 0 - uncompressed flat inode without tail-packing inline data:
87 * inode, [xattrs], ... | ... | no-holed data
88 * 1 - compressed inode with non-compact indexes:
89 * inode, [xattrs], [map_header], extents ... | ...
90 * 2 - uncompressed flat inode with tail-packing inline data:
91 * inode, [xattrs], tailpacking data, ... | ... | no-holed data
92 * 3 - compressed inode with compact indexes:
93 * inode, [xattrs], map_header, extents ... | ...
94 * 4 - chunk-based inode with (optional) multi-device support:
95 * inode, [xattrs], chunk indexes ... | ...
96 * 5~7 - reserved
97 */
98 enum {
99 EROFS_INODE_FLAT_PLAIN = 0,
100 EROFS_INODE_FLAT_COMPRESSION_LEGACY = 1,
101 EROFS_INODE_FLAT_INLINE = 2,
102 EROFS_INODE_FLAT_COMPRESSION = 3,
103 EROFS_INODE_CHUNK_BASED = 4,
104 EROFS_INODE_DATALAYOUT_MAX
105 };
106
erofs_inode_is_data_compressed(unsigned int datamode)107 static inline bool erofs_inode_is_data_compressed(unsigned int datamode)
108 {
109 return datamode == EROFS_INODE_FLAT_COMPRESSION ||
110 datamode == EROFS_INODE_FLAT_COMPRESSION_LEGACY;
111 }
112
113 /* bit definitions of inode i_format */
114 #define EROFS_I_VERSION_BITS 1
115 #define EROFS_I_DATALAYOUT_BITS 3
116
117 #define EROFS_I_VERSION_BIT 0
118 #define EROFS_I_DATALAYOUT_BIT 1
119
120 #define EROFS_I_ALL \
121 ((1 << (EROFS_I_DATALAYOUT_BIT + EROFS_I_DATALAYOUT_BITS)) - 1)
122
123 /* indicate chunk blkbits, thus 'chunksize = blocksize << chunk blkbits' */
124 #define EROFS_CHUNK_FORMAT_BLKBITS_MASK 0x001F
125 /* with chunk indexes or just a 4-byte blkaddr array */
126 #define EROFS_CHUNK_FORMAT_INDEXES 0x0020
127
128 #define EROFS_CHUNK_FORMAT_ALL \
129 (EROFS_CHUNK_FORMAT_BLKBITS_MASK | EROFS_CHUNK_FORMAT_INDEXES)
130
131 struct erofs_inode_chunk_info {
132 __le16 format; /* chunk blkbits, etc. */
133 __le16 reserved;
134 };
135
136 /* 32-byte reduced form of an ondisk inode */
137 struct erofs_inode_compact {
138 __le16 i_format; /* inode format hints */
139
140 /* 1 header + n-1 * 4 bytes inline xattr to keep continuity */
141 __le16 i_xattr_icount;
142 __le16 i_mode;
143 __le16 i_nlink;
144 __le32 i_size;
145 __le32 i_reserved;
146 union {
147 /* total compressed blocks for compressed inodes */
148 __le32 compressed_blocks;
149 /* block address for uncompressed flat inodes */
150 __le32 raw_blkaddr;
151
152 /* for device files, used to indicate old/new device # */
153 __le32 rdev;
154
155 /* for chunk-based files, it contains the summary info */
156 struct erofs_inode_chunk_info c;
157 } i_u;
158 __le32 i_ino; /* only used for 32-bit stat compatibility */
159 __le16 i_uid;
160 __le16 i_gid;
161 __le32 i_reserved2;
162 };
163
164 /* 32-byte on-disk inode */
165 #define EROFS_INODE_LAYOUT_COMPACT 0
166 /* 64-byte on-disk inode */
167 #define EROFS_INODE_LAYOUT_EXTENDED 1
168
169 /* 64-byte complete form of an ondisk inode */
170 struct erofs_inode_extended {
171 __le16 i_format; /* inode format hints */
172
173 /* 1 header + n-1 * 4 bytes inline xattr to keep continuity */
174 __le16 i_xattr_icount;
175 __le16 i_mode;
176 __le16 i_reserved;
177 __le64 i_size;
178 union {
179 /* total compressed blocks for compressed inodes */
180 __le32 compressed_blocks;
181 /* block address for uncompressed flat inodes */
182 __le32 raw_blkaddr;
183
184 /* for device files, used to indicate old/new device # */
185 __le32 rdev;
186
187 /* for chunk-based files, it contains the summary info */
188 struct erofs_inode_chunk_info c;
189 } i_u;
190
191 /* only used for 32-bit stat compatibility */
192 __le32 i_ino;
193
194 __le32 i_uid;
195 __le32 i_gid;
196 __le64 i_mtime;
197 __le32 i_mtime_nsec;
198 __le32 i_nlink;
199 __u8 i_reserved2[16];
200 };
201
202 #define EROFS_MAX_SHARED_XATTRS (128)
203 /* h_shared_count between 129 ... 255 are special # */
204 #define EROFS_SHARED_XATTR_EXTENT (255)
205
206 /*
207 * inline xattrs (n == i_xattr_icount):
208 * erofs_xattr_ibody_header(1) + (n - 1) * 4 bytes
209 * 12 bytes / \
210 * / \
211 * /-----------------------\
212 * | erofs_xattr_entries+ |
213 * +-----------------------+
214 * inline xattrs must starts in erofs_xattr_ibody_header,
215 * for read-only fs, no need to introduce h_refcount
216 */
217 struct erofs_xattr_ibody_header {
218 __le32 h_reserved;
219 __u8 h_shared_count;
220 __u8 h_reserved2[7];
221 __le32 h_shared_xattrs[]; /* shared xattr id array */
222 };
223
224 /* Name indexes */
225 #define EROFS_XATTR_INDEX_USER 1
226 #define EROFS_XATTR_INDEX_POSIX_ACL_ACCESS 2
227 #define EROFS_XATTR_INDEX_POSIX_ACL_DEFAULT 3
228 #define EROFS_XATTR_INDEX_TRUSTED 4
229 #define EROFS_XATTR_INDEX_LUSTRE 5
230 #define EROFS_XATTR_INDEX_SECURITY 6
231
232 /* xattr entry (for both inline & shared xattrs) */
233 struct erofs_xattr_entry {
234 __u8 e_name_len; /* length of name */
235 __u8 e_name_index; /* attribute name index */
236 __le16 e_value_size; /* size of attribute value */
237 /* followed by e_name and e_value */
238 char e_name[]; /* attribute name */
239 };
240
erofs_xattr_ibody_size(__le16 i_xattr_icount)241 static inline unsigned int erofs_xattr_ibody_size(__le16 i_xattr_icount)
242 {
243 if (!i_xattr_icount)
244 return 0;
245
246 return sizeof(struct erofs_xattr_ibody_header) +
247 sizeof(__u32) * (le16_to_cpu(i_xattr_icount) - 1);
248 }
249
250 #define EROFS_XATTR_ALIGN(size) round_up(size, sizeof(struct erofs_xattr_entry))
251
erofs_xattr_entry_size(struct erofs_xattr_entry * e)252 static inline unsigned int erofs_xattr_entry_size(struct erofs_xattr_entry *e)
253 {
254 return EROFS_XATTR_ALIGN(sizeof(struct erofs_xattr_entry) +
255 e->e_name_len + le16_to_cpu(e->e_value_size));
256 }
257
258 /* represent a zeroed chunk (hole) */
259 #define EROFS_NULL_ADDR -1
260
261 /* 4-byte block address array */
262 #define EROFS_BLOCK_MAP_ENTRY_SIZE sizeof(__le32)
263
264 /* 8-byte inode chunk indexes */
265 struct erofs_inode_chunk_index {
266 __le16 advise; /* always 0, don't care for now */
267 __le16 device_id; /* back-end storage id (with bits masked) */
268 __le32 blkaddr; /* start block address of this inode chunk */
269 };
270
271 /* maximum supported size of a physical compression cluster */
272 #define Z_EROFS_PCLUSTER_MAX_SIZE (1024 * 1024)
273
274 /* available compression algorithm types (for h_algorithmtype) */
275 enum {
276 Z_EROFS_COMPRESSION_LZ4 = 0,
277 Z_EROFS_COMPRESSION_LZMA = 1,
278 Z_EROFS_COMPRESSION_MAX
279 };
280 #define Z_EROFS_ALL_COMPR_ALGS ((1 << Z_EROFS_COMPRESSION_MAX) - 1)
281
282 /* 14 bytes (+ length field = 16 bytes) */
283 struct z_erofs_lz4_cfgs {
284 __le16 max_distance;
285 __le16 max_pclusterblks;
286 u8 reserved[10];
287 } __packed;
288
289 /* 14 bytes (+ length field = 16 bytes) */
290 struct z_erofs_lzma_cfgs {
291 __le32 dict_size;
292 __le16 format;
293 u8 reserved[8];
294 } __packed;
295
296 #define Z_EROFS_LZMA_MAX_DICT_SIZE (8 * Z_EROFS_PCLUSTER_MAX_SIZE)
297
298 /*
299 * bit 0 : COMPACTED_2B indexes (0 - off; 1 - on)
300 * e.g. for 4k logical cluster size, 4B if compacted 2B is off;
301 * (4B) + 2B + (4B) if compacted 2B is on.
302 * bit 1 : HEAD1 big pcluster (0 - off; 1 - on)
303 * bit 2 : HEAD2 big pcluster (0 - off; 1 - on)
304 * bit 3 : tailpacking inline pcluster (0 - off; 1 - on)
305 * bit 4 : interlaced plain pcluster (0 - off; 1 - on)
306 * bit 5 : fragment pcluster (0 - off; 1 - on)
307 */
308 #define Z_EROFS_ADVISE_COMPACTED_2B 0x0001
309 #define Z_EROFS_ADVISE_BIG_PCLUSTER_1 0x0002
310 #define Z_EROFS_ADVISE_BIG_PCLUSTER_2 0x0004
311 #define Z_EROFS_ADVISE_INLINE_PCLUSTER 0x0008
312 #define Z_EROFS_ADVISE_INTERLACED_PCLUSTER 0x0010
313 #define Z_EROFS_ADVISE_FRAGMENT_PCLUSTER 0x0020
314
315 #define Z_EROFS_FRAGMENT_INODE_BIT 7
316 struct z_erofs_map_header {
317 union {
318 /* fragment data offset in the packed inode */
319 __le32 h_fragmentoff;
320 struct {
321 __le16 h_reserved1;
322 /* indicates the encoded size of tailpacking data */
323 __le16 h_idata_size;
324 };
325 };
326 __le16 h_advise;
327 /*
328 * bit 0-3 : algorithm type of head 1 (logical cluster type 01);
329 * bit 4-7 : algorithm type of head 2 (logical cluster type 11).
330 */
331 __u8 h_algorithmtype;
332 /*
333 * bit 0-2 : logical cluster bits - 12, e.g. 0 for 4096;
334 * bit 3-6 : reserved;
335 * bit 7 : move the whole file into packed inode or not.
336 */
337 __u8 h_clusterbits;
338 };
339
340 #define Z_EROFS_VLE_LEGACY_HEADER_PADDING 8
341
342 /*
343 * Fixed-sized output compression on-disk logical cluster type:
344 * 0 - literal (uncompressed) lcluster
345 * 1,3 - compressed lcluster (for HEAD lclusters)
346 * 2 - compressed lcluster (for NONHEAD lclusters)
347 *
348 * In detail,
349 * 0 - literal (uncompressed) lcluster,
350 * di_advise = 0
351 * di_clusterofs = the literal data offset of the lcluster
352 * di_blkaddr = the blkaddr of the literal pcluster
353 *
354 * 1,3 - compressed lcluster (for HEAD lclusters)
355 * di_advise = 1 or 3
356 * di_clusterofs = the decompressed data offset of the lcluster
357 * di_blkaddr = the blkaddr of the compressed pcluster
358 *
359 * 2 - compressed lcluster (for NONHEAD lclusters)
360 * di_advise = 2
361 * di_clusterofs =
362 * the decompressed data offset in its own HEAD lcluster
363 * di_u.delta[0] = distance to this HEAD lcluster
364 * di_u.delta[1] = distance to the next HEAD lcluster
365 */
366 enum {
367 Z_EROFS_VLE_CLUSTER_TYPE_PLAIN = 0,
368 Z_EROFS_VLE_CLUSTER_TYPE_HEAD1 = 1,
369 Z_EROFS_VLE_CLUSTER_TYPE_NONHEAD = 2,
370 Z_EROFS_VLE_CLUSTER_TYPE_HEAD2 = 3,
371 Z_EROFS_VLE_CLUSTER_TYPE_MAX
372 };
373
374 #define Z_EROFS_VLE_DI_CLUSTER_TYPE_BITS 2
375 #define Z_EROFS_VLE_DI_CLUSTER_TYPE_BIT 0
376
377 /* (noncompact only, HEAD) This pcluster refers to partial decompressed data */
378 #define Z_EROFS_VLE_DI_PARTIAL_REF (1 << 15)
379
380 /*
381 * D0_CBLKCNT will be marked _only_ at the 1st non-head lcluster to store the
382 * compressed block count of a compressed extent (in logical clusters, aka.
383 * block count of a pcluster).
384 */
385 #define Z_EROFS_VLE_DI_D0_CBLKCNT (1 << 11)
386
387 struct z_erofs_vle_decompressed_index {
388 __le16 di_advise;
389 /* where to decompress in the head lcluster */
390 __le16 di_clusterofs;
391
392 union {
393 /* for the HEAD lclusters */
394 __le32 blkaddr;
395 /*
396 * for the NONHEAD lclusters
397 * [0] - distance to its HEAD lcluster
398 * [1] - distance to the next HEAD lcluster
399 */
400 __le16 delta[2];
401 } di_u;
402 };
403
404 #define Z_EROFS_VLE_LEGACY_INDEX_ALIGN(size) \
405 (round_up(size, sizeof(struct z_erofs_vle_decompressed_index)) + \
406 sizeof(struct z_erofs_map_header) + Z_EROFS_VLE_LEGACY_HEADER_PADDING)
407
408 /* dirent sorts in alphabet order, thus we can do binary search */
409 struct erofs_dirent {
410 __le64 nid; /* node number */
411 __le16 nameoff; /* start offset of file name */
412 __u8 file_type; /* file type */
413 __u8 reserved; /* reserved */
414 } __packed;
415
416 /*
417 * EROFS file types should match generic FT_* types and
418 * it seems no need to add BUILD_BUG_ONs since potential
419 * unmatchness will break other fses as well...
420 */
421
422 #define EROFS_NAME_LEN 255
423
424 /* check the EROFS on-disk layout strictly at compile time */
erofs_check_ondisk_layout_definitions(void)425 static inline void erofs_check_ondisk_layout_definitions(void)
426 {
427 const __le64 fmh = *(__le64 *)&(struct z_erofs_map_header) {
428 .h_clusterbits = 1 << Z_EROFS_FRAGMENT_INODE_BIT
429 };
430
431 BUILD_BUG_ON(sizeof(struct erofs_super_block) != 128);
432 BUILD_BUG_ON(sizeof(struct erofs_inode_compact) != 32);
433 BUILD_BUG_ON(sizeof(struct erofs_inode_extended) != 64);
434 BUILD_BUG_ON(sizeof(struct erofs_xattr_ibody_header) != 12);
435 BUILD_BUG_ON(sizeof(struct erofs_xattr_entry) != 4);
436 BUILD_BUG_ON(sizeof(struct erofs_inode_chunk_info) != 4);
437 BUILD_BUG_ON(sizeof(struct erofs_inode_chunk_index) != 8);
438 BUILD_BUG_ON(sizeof(struct z_erofs_map_header) != 8);
439 BUILD_BUG_ON(sizeof(struct z_erofs_vle_decompressed_index) != 8);
440 BUILD_BUG_ON(sizeof(struct erofs_dirent) != 12);
441 /* keep in sync between 2 index structures for better extendibility */
442 BUILD_BUG_ON(sizeof(struct erofs_inode_chunk_index) !=
443 sizeof(struct z_erofs_vle_decompressed_index));
444 BUILD_BUG_ON(sizeof(struct erofs_deviceslot) != 128);
445
446 BUILD_BUG_ON(BIT(Z_EROFS_VLE_DI_CLUSTER_TYPE_BITS) <
447 Z_EROFS_VLE_CLUSTER_TYPE_MAX - 1);
448 /* exclude old compiler versions like gcc 7.5.0 */
449 BUILD_BUG_ON(__builtin_constant_p(fmh) ?
450 fmh != cpu_to_le64(1ULL << 63) : 0);
451 }
452
453 #endif
454