LXR linux/fs/erofs/erofs

   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
  16/*
  17 * Any bits that aren't in EROFS_ALL_FEATURE_INCOMPAT should
  18 * be incompatible with this kernel version.
  19 */
  20#define EROFS_FEATURE_INCOMPAT_LZ4_0PADDING     0x00000001
  21#define EROFS_FEATURE_INCOMPAT_COMPR_CFGS       0x00000002
  22#define EROFS_FEATURE_INCOMPAT_BIG_PCLUSTER     0x00000002
  23#define EROFS_FEATURE_INCOMPAT_CHUNKED_FILE     0x00000004
  24#define EROFS_ALL_FEATURE_INCOMPAT              \
  25        (EROFS_FEATURE_INCOMPAT_LZ4_0PADDING | \
  26         EROFS_FEATURE_INCOMPAT_COMPR_CFGS | \
  27         EROFS_FEATURE_INCOMPAT_BIG_PCLUSTER | \
  28         EROFS_FEATURE_INCOMPAT_CHUNKED_FILE)
  29
  30#define EROFS_SB_EXTSLOT_SIZE   16
  31
  32/* erofs on-disk super block (currently 128 bytes) */
  33struct erofs_super_block {
  34        __le32 magic;           /* file system magic number */
  35        __le32 checksum;        /* crc32c(super_block) */
  36        __le32 feature_compat;
  37        __u8 blkszbits;         /* support block_size == PAGE_SIZE only */
  38        __u8 sb_extslots;       /* superblock size = 128 + sb_extslots * 16 */
  39
  40        __le16 root_nid;        /* nid of root directory */
  41        __le64 inos;            /* total valid ino # (== f_files - f_favail) */
  42
  43        __le64 build_time;      /* inode v1 time derivation */
  44        __le32 build_time_nsec; /* inode v1 time derivation in nano scale */
  45        __le32 blocks;          /* used for statfs */
  46        __le32 meta_blkaddr;    /* start block address of metadata area */
  47        __le32 xattr_blkaddr;   /* start block address of shared xattr area */
  48        __u8 uuid[16];          /* 128-bit uuid for volume */
  49        __u8 volume_name[16];   /* volume name */
  50        __le32 feature_incompat;
  51        union {
  52                /* bitmap for available compression algorithms */
  53                __le16 available_compr_algs;
  54                /* customized sliding window size instead of 64k by default */
  55                __le16 lz4_max_distance;
  56        } __packed u1;
  57        __u8 reserved2[42];
  58};
  59
  60/*
  61 * erofs inode datalayout (i_format in on-disk inode):
  62 * 0 - inode plain without inline data A:
  63 * inode, [xattrs], ... | ... | no-holed data
  64 * 1 - inode VLE compression B (legacy):
  65 * inode, [xattrs], extents ... | ...
  66 * 2 - inode plain with inline data C:
  67 * inode, [xattrs], last_inline_data, ... | ... | no-holed data
  68 * 3 - inode compression D:
  69 * inode, [xattrs], map_header, extents ... | ...
  70 * 4 - inode chunk-based E:
  71 * inode, [xattrs], chunk indexes ... | ...
  72 * 5~7 - reserved
  73 */
  74enum {
  75        EROFS_INODE_FLAT_PLAIN                  = 0,
  76        EROFS_INODE_FLAT_COMPRESSION_LEGACY     = 1,
  77        EROFS_INODE_FLAT_INLINE                 = 2,
  78        EROFS_INODE_FLAT_COMPRESSION            = 3,
  79        EROFS_INODE_CHUNK_BASED                 = 4,
  80        EROFS_INODE_DATALAYOUT_MAX
  81};
  82
  83static inline bool erofs_inode_is_data_compressed(unsigned int datamode)
  84{
  85        return datamode == EROFS_INODE_FLAT_COMPRESSION ||
  86                datamode == EROFS_INODE_FLAT_COMPRESSION_LEGACY;
  87}
  88
  89/* bit definitions of inode i_advise */
  90#define EROFS_I_VERSION_BITS            1
  91#define EROFS_I_DATALAYOUT_BITS         3
  92
  93#define EROFS_I_VERSION_BIT             0
  94#define EROFS_I_DATALAYOUT_BIT          1
  95
  96#define EROFS_I_ALL     \
  97        ((1 << (EROFS_I_DATALAYOUT_BIT + EROFS_I_DATALAYOUT_BITS)) - 1)
  98
  99/* indicate chunk blkbits, thus 'chunksize = blocksize << chunk blkbits' */
 100#define EROFS_CHUNK_FORMAT_BLKBITS_MASK         0x001F
 101/* with chunk indexes or just a 4-byte blkaddr array */
 102#define EROFS_CHUNK_FORMAT_INDEXES              0x0020
 103
 104#define EROFS_CHUNK_FORMAT_ALL  \
 105        (EROFS_CHUNK_FORMAT_BLKBITS_MASK | EROFS_CHUNK_FORMAT_INDEXES)
 106
 107struct erofs_inode_chunk_info {
 108        __le16 format;          /* chunk blkbits, etc. */
 109        __le16 reserved;
 110};
 111
 112/* 32-byte reduced form of an ondisk inode */
 113struct erofs_inode_compact {
 114        __le16 i_format;        /* inode format hints */
 115
 116/* 1 header + n-1 * 4 bytes inline xattr to keep continuity */
 117        __le16 i_xattr_icount;
 118        __le16 i_mode;
 119        __le16 i_nlink;
 120        __le32 i_size;
 121        __le32 i_reserved;
 122        union {
 123                /* file total compressed blocks for data mapping 1 */
 124                __le32 compressed_blocks;
 125                __le32 raw_blkaddr;
 126
 127                /* for device files, used to indicate old/new device # */
 128                __le32 rdev;
 129
 130                /* for chunk-based files, it contains the summary info */
 131                struct erofs_inode_chunk_info c;
 132        } i_u;
 133        __le32 i_ino;           /* only used for 32-bit stat compatibility */
 134        __le16 i_uid;
 135        __le16 i_gid;
 136        __le32 i_reserved2;
 137};
 138
 139/* 32 bytes on-disk inode */
 140#define EROFS_INODE_LAYOUT_COMPACT      0
 141/* 64 bytes on-disk inode */
 142#define EROFS_INODE_LAYOUT_EXTENDED     1
 143
 144/* 64-byte complete form of an ondisk inode */
 145struct erofs_inode_extended {
 146        __le16 i_format;        /* inode format hints */
 147
 148/* 1 header + n-1 * 4 bytes inline xattr to keep continuity */
 149        __le16 i_xattr_icount;
 150        __le16 i_mode;
 151        __le16 i_reserved;
 152        __le64 i_size;
 153        union {
 154                /* file total compressed blocks for data mapping 1 */
 155                __le32 compressed_blocks;
 156                __le32 raw_blkaddr;
 157
 158                /* for device files, used to indicate old/new device # */
 159                __le32 rdev;
 160
 161                /* for chunk-based files, it contains the summary info */
 162                struct erofs_inode_chunk_info c;
 163        } i_u;
 164
 165        /* only used for 32-bit stat compatibility */
 166        __le32 i_ino;
 167
 168        __le32 i_uid;
 169        __le32 i_gid;
 170        __le64 i_ctime;
 171        __le32 i_ctime_nsec;
 172        __le32 i_nlink;
 173        __u8   i_reserved2[16];
 174};
 175
 176#define EROFS_MAX_SHARED_XATTRS         (128)
 177/* h_shared_count between 129 ... 255 are special # */
 178#define EROFS_SHARED_XATTR_EXTENT       (255)
 179
 180/*
 181 * inline xattrs (n == i_xattr_icount):
 182 * erofs_xattr_ibody_header(1) + (n - 1) * 4 bytes
 183 *          12 bytes           /                   \
 184 *                            /                     \
 185 *                           /-----------------------\
 186 *                           |  erofs_xattr_entries+ |
 187 *                           +-----------------------+
 188 * inline xattrs must starts in erofs_xattr_ibody_header,
 189 * for read-only fs, no need to introduce h_refcount
 190 */
 191struct erofs_xattr_ibody_header {
 192        __le32 h_reserved;
 193        __u8   h_shared_count;
 194        __u8   h_reserved2[7];
 195        __le32 h_shared_xattrs[0];      /* shared xattr id array */
 196};
 197
 198/* Name indexes */
 199#define EROFS_XATTR_INDEX_USER              1
 200#define EROFS_XATTR_INDEX_POSIX_ACL_ACCESS  2
 201#define EROFS_XATTR_INDEX_POSIX_ACL_DEFAULT 3
 202#define EROFS_XATTR_INDEX_TRUSTED           4
 203#define EROFS_XATTR_INDEX_LUSTRE            5
 204#define EROFS_XATTR_INDEX_SECURITY          6
 205
 206/* xattr entry (for both inline & shared xattrs) */
 207struct erofs_xattr_entry {
 208        __u8   e_name_len;      /* length of name */
 209        __u8   e_name_index;    /* attribute name index */
 210        __le16 e_value_size;    /* size of attribute value */
 211        /* followed by e_name and e_value */
 212        char   e_name[0];       /* attribute name */
 213};
 214
 215static inline unsigned int erofs_xattr_ibody_size(__le16 i_xattr_icount)
 216{
 217        if (!i_xattr_icount)
 218                return 0;
 219
 220        return sizeof(struct erofs_xattr_ibody_header) +
 221                sizeof(__u32) * (le16_to_cpu(i_xattr_icount) - 1);
 222}
 223
 224#define EROFS_XATTR_ALIGN(size) round_up(size, sizeof(struct erofs_xattr_entry))
 225
 226static inline unsigned int erofs_xattr_entry_size(struct erofs_xattr_entry *e)
 227{
 228        return EROFS_XATTR_ALIGN(sizeof(struct erofs_xattr_entry) +
 229                                 e->e_name_len + le16_to_cpu(e->e_value_size));
 230}
 231
 232/* represent a zeroed chunk (hole) */
 233#define EROFS_NULL_ADDR                 -1
 234
 235/* 4-byte block address array */
 236#define EROFS_BLOCK_MAP_ENTRY_SIZE      sizeof(__le32)
 237
 238/* 8-byte inode chunk indexes */
 239struct erofs_inode_chunk_index {
 240        __le16 advise;          /* always 0, don't care for now */
 241        __le16 device_id;       /* back-end storage id, always 0 for now */
 242        __le32 blkaddr;         /* start block address of this inode chunk */
 243};
 244
 245/* maximum supported size of a physical compression cluster */
 246#define Z_EROFS_PCLUSTER_MAX_SIZE       (1024 * 1024)
 247
 248/* available compression algorithm types (for h_algorithmtype) */
 249enum {
 250        Z_EROFS_COMPRESSION_LZ4 = 0,
 251        Z_EROFS_COMPRESSION_MAX
 252};
 253#define Z_EROFS_ALL_COMPR_ALGS          (1 << (Z_EROFS_COMPRESSION_MAX - 1))
 254
 255/* 14 bytes (+ length field = 16 bytes) */
 256struct z_erofs_lz4_cfgs {
 257        __le16 max_distance;
 258        __le16 max_pclusterblks;
 259        u8 reserved[10];
 260} __packed;
 261
 262/*
 263 * bit 0 : COMPACTED_2B indexes (0 - off; 1 - on)
 264 *  e.g. for 4k logical cluster size,      4B        if compacted 2B is off;
 265 *                                  (4B) + 2B + (4B) if compacted 2B is on.
 266 * bit 1 : HEAD1 big pcluster (0 - off; 1 - on)
 267 * bit 2 : HEAD2 big pcluster (0 - off; 1 - on)
 268 */
 269#define Z_EROFS_ADVISE_COMPACTED_2B             0x0001
 270#define Z_EROFS_ADVISE_BIG_PCLUSTER_1           0x0002
 271#define Z_EROFS_ADVISE_BIG_PCLUSTER_2           0x0004
 272
 273struct z_erofs_map_header {
 274        __le32  h_reserved1;
 275        __le16  h_advise;
 276        /*
 277         * bit 0-3 : algorithm type of head 1 (logical cluster type 01);
 278         * bit 4-7 : algorithm type of head 2 (logical cluster type 11).
 279         */
 280        __u8    h_algorithmtype;
 281        /*
 282         * bit 0-2 : logical cluster bits - 12, e.g. 0 for 4096;
 283         * bit 3-7 : reserved.
 284         */
 285        __u8    h_clusterbits;
 286};
 287
 288#define Z_EROFS_VLE_LEGACY_HEADER_PADDING       8
 289
 290/*
 291 * Fixed-sized output compression ondisk Logical Extent cluster type:
 292 *    0 - literal (uncompressed) cluster
 293 *    1 - compressed cluster (for the head logical cluster)
 294 *    2 - compressed cluster (for the other logical clusters)
 295 *
 296 * In detail,
 297 *    0 - literal (uncompressed) cluster,
 298 *        di_advise = 0
 299 *        di_clusterofs = the literal data offset of the cluster
 300 *        di_blkaddr = the blkaddr of the literal cluster
 301 *
 302 *    1 - compressed cluster (for the head logical cluster)
 303 *        di_advise = 1
 304 *        di_clusterofs = the decompressed data offset of the cluster
 305 *        di_blkaddr = the blkaddr of the compressed cluster
 306 *
 307 *    2 - compressed cluster (for the other logical clusters)
 308 *        di_advise = 2
 309 *        di_clusterofs =
 310 *           the decompressed data offset in its own head cluster
 311 *        di_u.delta[0] = distance to its corresponding head cluster
 312 *        di_u.delta[1] = distance to its corresponding tail cluster
 313 *                (di_advise could be 0, 1 or 2)
 314 */
 315enum {
 316        Z_EROFS_VLE_CLUSTER_TYPE_PLAIN          = 0,
 317        Z_EROFS_VLE_CLUSTER_TYPE_HEAD           = 1,
 318        Z_EROFS_VLE_CLUSTER_TYPE_NONHEAD        = 2,
 319        Z_EROFS_VLE_CLUSTER_TYPE_RESERVED       = 3,
 320        Z_EROFS_VLE_CLUSTER_TYPE_MAX
 321};
 322
 323#define Z_EROFS_VLE_DI_CLUSTER_TYPE_BITS        2
 324#define Z_EROFS_VLE_DI_CLUSTER_TYPE_BIT         0
 325
 326/*
 327 * D0_CBLKCNT will be marked _only_ at the 1st non-head lcluster to store the
 328 * compressed block count of a compressed extent (in logical clusters, aka.
 329 * block count of a pcluster).
 330 */
 331#define Z_EROFS_VLE_DI_D0_CBLKCNT               (1 << 11)
 332
 333struct z_erofs_vle_decompressed_index {
 334        __le16 di_advise;
 335        /* where to decompress in the head cluster */
 336        __le16 di_clusterofs;
 337
 338        union {
 339                /* for the head cluster */
 340                __le32 blkaddr;
 341                /*
 342                 * for the rest clusters
 343                 * eg. for 4k page-sized cluster, maximum 4K*64k = 256M)
 344                 * [0] - pointing to the head cluster
 345                 * [1] - pointing to the tail cluster
 346                 */
 347                __le16 delta[2];
 348        } di_u;
 349};
 350
 351#define Z_EROFS_VLE_LEGACY_INDEX_ALIGN(size) \
 352        (round_up(size, sizeof(struct z_erofs_vle_decompressed_index)) + \
 353         sizeof(struct z_erofs_map_header) + Z_EROFS_VLE_LEGACY_HEADER_PADDING)
 354
 355/* dirent sorts in alphabet order, thus we can do binary search */
 356struct erofs_dirent {
 357        __le64 nid;     /* node number */
 358        __le16 nameoff; /* start offset of file name */
 359        __u8 file_type; /* file type */
 360        __u8 reserved;  /* reserved */
 361} __packed;
 362
 363/*
 364 * EROFS file types should match generic FT_* types and
 365 * it seems no need to add BUILD_BUG_ONs since potential
 366 * unmatchness will break other fses as well...
 367 */
 368
 369#define EROFS_NAME_LEN      255
 370
 371/* check the EROFS on-disk layout strictly at compile time */
 372static inline void erofs_check_ondisk_layout_definitions(void)
 373{
 374        BUILD_BUG_ON(sizeof(struct erofs_super_block) != 128);
 375        BUILD_BUG_ON(sizeof(struct erofs_inode_compact) != 32);
 376        BUILD_BUG_ON(sizeof(struct erofs_inode_extended) != 64);
 377        BUILD_BUG_ON(sizeof(struct erofs_xattr_ibody_header) != 12);
 378        BUILD_BUG_ON(sizeof(struct erofs_xattr_entry) != 4);
 379        BUILD_BUG_ON(sizeof(struct erofs_inode_chunk_info) != 4);
 380        BUILD_BUG_ON(sizeof(struct erofs_inode_chunk_index) != 8);
 381        BUILD_BUG_ON(sizeof(struct z_erofs_map_header) != 8);
 382        BUILD_BUG_ON(sizeof(struct z_erofs_vle_decompressed_index) != 8);
 383        BUILD_BUG_ON(sizeof(struct erofs_dirent) != 12);
 384        /* keep in sync between 2 index structures for better extendibility */
 385        BUILD_BUG_ON(sizeof(struct erofs_inode_chunk_index) !=
 386                     sizeof(struct z_erofs_vle_decompressed_index));
 387
 388        BUILD_BUG_ON(BIT(Z_EROFS_VLE_DI_CLUSTER_TYPE_BITS) <
 389                     Z_EROFS_VLE_CLUSTER_TYPE_MAX - 1);
 390}
 391
 392#endif
 393