src/ext4_fs.c

   1 /*
   2  * Copyright (c) 2013 Grzegorz Kostka (kostka.grzegorz@gmail.com)
   3  *
   4  *
   5  * HelenOS:
   6  * Copyright (c) 2012 Martin Sucha
   7  * Copyright (c) 2012 Frantisek Princ
   8  * All rights reserved.
   9  *
  10  * Redistribution and use in source and binary forms, with or without
  11  * modification, are permitted provided that the following conditions
  12  * are met:
  13  *
  14  * - Redistributions of source code must retain the above copyright
  15  *   notice, this list of conditions and the following disclaimer.
  16  * - Redistributions in binary form must reproduce the above copyright
  17  *   notice, this list of conditions and the following disclaimer in the
  18  *   documentation and/or other materials provided with the distribution.
  19  * - The name of the author may not be used to endorse or promote products
  20  *   derived from this software without specific prior written permission.
  21  *
  22  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
  23  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
  24  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
  25  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
  26  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
  27  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  28  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  29  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  30  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
  31  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  32  */
  33 /** @addtogroup lwext4
  34  * @{
  35  */
  36 /**
  37  * @file  ext4_fs.c
  38  * @brief More complex filesystem functions.
  39  */
  40
  41 #include "ext4_config.h"
  42 #include "ext4_types.h"
  43 #include "ext4_fs.h"
  44 #include "ext4_errno.h"
  45 #include "ext4_blockdev.h"
  46 #include "ext4_super.h"
  47 #include "ext4_crc32.h"
  48 #include "ext4_debug.h"
  49 #include "ext4_block_group.h"
  50 #include "ext4_balloc.h"
  51 #include "ext4_bitmap.h"
  52 #include "ext4_inode.h"
  53 #include "ext4_ialloc.h"
  54 #include "ext4_extent.h"
  55
  56 #include <string.h>
  57
  58 int ext4_fs_init(struct ext4_fs *fs, struct ext4_blockdev *bdev,
  59                  bool read_only)
  60 {
  61         int r, i;
  62         uint16_t tmp;
  63         uint32_t bsize;
  64
  65         ext4_assert(fs && bdev);
  66
  67         fs->bdev = bdev;
  68
  69         r = ext4_sb_read(fs->bdev, &fs->sb);
  70         if (r != EOK)
  71                 return r;
  72
  73         if (!ext4_sb_check(&fs->sb))
  74                 return ENOTSUP;
  75
  76         bsize = ext4_sb_get_block_size(&fs->sb);
  77         if (bsize > EXT4_MAX_BLOCK_SIZE)
  78                 return ENXIO;
  79
  80         r = ext4_fs_check_features(fs, &read_only);
  81         if (r != EOK)
  82                 return r;
  83
  84         if (read_only)
  85                 fs->read_only = true;
  86
  87         /* Compute limits for indirect block levels */
  88         uint32_t blocks_id = bsize / sizeof(uint32_t);
  89
  90         fs->inode_block_limits[0] = EXT4_INODE_DIRECT_BLOCK_COUNT;
  91         fs->inode_blocks_per_level[0] = 1;
  92
  93         for (i = 1; i < 4; i++) {
  94                 fs->inode_blocks_per_level[i] =
  95                     fs->inode_blocks_per_level[i - 1] * blocks_id;
  96                 fs->inode_block_limits[i] = fs->inode_block_limits[i - 1] +
  97                                             fs->inode_blocks_per_level[i];
  98         }
  99
 100         /*Validate FS*/
 101         tmp = ext4_get16(&fs->sb, state);
 102         if (tmp & EXT4_SUPERBLOCK_STATE_ERROR_FS)
 103                 ext4_dbg(DEBUG_FS, DBG_WARN
 104                                 "last umount error: superblock fs_error flag\n");
 105
 106
 107         if (!fs->read_only) {
 108                 /* Mark system as mounted */
 109                 ext4_set16(&fs->sb, state, EXT4_SUPERBLOCK_STATE_ERROR_FS);
 110                 r = ext4_sb_write(fs->bdev, &fs->sb);
 111                 if (r != EOK)
 112                         return r;
 113
 114                 /*Update mount count*/
 115                 ext4_set16(&fs->sb, mount_count, ext4_get16(&fs->sb, mount_count) + 1);
 116         }
 117
 118         return r;
 119 }
 120
 121 int ext4_fs_fini(struct ext4_fs *fs)
 122 {
 123         ext4_assert(fs);
 124
 125         /*Set superblock state*/
 126         ext4_set16(&fs->sb, state, EXT4_SUPERBLOCK_STATE_VALID_FS);
 127
 128         if (!fs->read_only)
 129                 return ext4_sb_write(fs->bdev, &fs->sb);
 130
 131         return EOK;
 132 }
 133
 134 static void ext4_fs_debug_features_inc(uint32_t features_incompatible)
 135 {
 136         if (features_incompatible & EXT4_FINCOM_COMPRESSION)
 137                 ext4_dbg(DEBUG_FS, DBG_NONE "compression\n");
 138         if (features_incompatible & EXT4_FINCOM_FILETYPE)
 139                 ext4_dbg(DEBUG_FS, DBG_NONE "filetype\n");
 140         if (features_incompatible & EXT4_FINCOM_RECOVER)
 141                 ext4_dbg(DEBUG_FS, DBG_NONE "recover\n");
 142         if (features_incompatible & EXT4_FINCOM_JOURNAL_DEV)
 143                 ext4_dbg(DEBUG_FS, DBG_NONE "journal_dev\n");
 144         if (features_incompatible & EXT4_FINCOM_META_BG)
 145                 ext4_dbg(DEBUG_FS, DBG_NONE "meta_bg\n");
 146         if (features_incompatible & EXT4_FINCOM_EXTENTS)
 147                 ext4_dbg(DEBUG_FS, DBG_NONE "extents\n");
 148         if (features_incompatible & EXT4_FINCOM_64BIT)
 149                 ext4_dbg(DEBUG_FS, DBG_NONE "64bit\n");
 150         if (features_incompatible & EXT4_FINCOM_MMP)
 151                 ext4_dbg(DEBUG_FS, DBG_NONE "mnp\n");
 152         if (features_incompatible & EXT4_FINCOM_FLEX_BG)
 153                 ext4_dbg(DEBUG_FS, DBG_NONE "flex_bg\n");
 154         if (features_incompatible & EXT4_FINCOM_EA_INODE)
 155                 ext4_dbg(DEBUG_FS, DBG_NONE "ea_inode\n");
 156         if (features_incompatible & EXT4_FINCOM_DIRDATA)
 157                 ext4_dbg(DEBUG_FS, DBG_NONE "dirdata\n");
 158         if (features_incompatible & EXT4_FINCOM_BG_USE_META_CSUM)
 159                 ext4_dbg(DEBUG_FS, DBG_NONE "meta_csum\n");
 160         if (features_incompatible & EXT4_FINCOM_LARGEDIR)
 161                 ext4_dbg(DEBUG_FS, DBG_NONE "largedir\n");
 162         if (features_incompatible & EXT4_FINCOM_INLINE_DATA)
 163                 ext4_dbg(DEBUG_FS, DBG_NONE "inline_data\n");
 164 }
 165 static void ext4_fs_debug_features_comp(uint32_t features_compatible)
 166 {
 167         if (features_compatible & EXT4_FCOM_DIR_PREALLOC)
 168                 ext4_dbg(DEBUG_FS, DBG_NONE "dir_prealloc\n");
 169         if (features_compatible & EXT4_FCOM_IMAGIC_INODES)
 170                 ext4_dbg(DEBUG_FS, DBG_NONE "imagic_inodes\n");
 171         if (features_compatible & EXT4_FCOM_HAS_JOURNAL)
 172                 ext4_dbg(DEBUG_FS, DBG_NONE "has_journal\n");
 173         if (features_compatible & EXT4_FCOM_EXT_ATTR)
 174                 ext4_dbg(DEBUG_FS, DBG_NONE "ext_attr\n");
 175         if (features_compatible & EXT4_FCOM_RESIZE_INODE)
 176                 ext4_dbg(DEBUG_FS, DBG_NONE "resize_inode\n");
 177         if (features_compatible & EXT4_FCOM_DIR_INDEX)
 178                 ext4_dbg(DEBUG_FS, DBG_NONE "dir_index\n");
 179 }
 180
 181 static void ext4_fs_debug_features_ro(uint32_t features_ro)
 182 {
 183         if (features_ro & EXT4_FRO_COM_SPARSE_SUPER)
 184                 ext4_dbg(DEBUG_FS, DBG_NONE "sparse_super\n");
 185         if (features_ro & EXT4_FRO_COM_LARGE_FILE)
 186                 ext4_dbg(DEBUG_FS, DBG_NONE "large_file\n");
 187         if (features_ro & EXT4_FRO_COM_BTREE_DIR)
 188                 ext4_dbg(DEBUG_FS, DBG_NONE "btree_dir\n");
 189         if (features_ro & EXT4_FRO_COM_HUGE_FILE)
 190                 ext4_dbg(DEBUG_FS, DBG_NONE "huge_file\n");
 191         if (features_ro & EXT4_FRO_COM_GDT_CSUM)
 192                 ext4_dbg(DEBUG_FS, DBG_NONE "gtd_csum\n");
 193         if (features_ro & EXT4_FRO_COM_DIR_NLINK)
 194                 ext4_dbg(DEBUG_FS, DBG_NONE "dir_nlink\n");
 195         if (features_ro & EXT4_FRO_COM_EXTRA_ISIZE)
 196                 ext4_dbg(DEBUG_FS, DBG_NONE "extra_isize\n");
 197         if (features_ro & EXT4_FRO_COM_QUOTA)
 198                 ext4_dbg(DEBUG_FS, DBG_NONE "quota\n");
 199         if (features_ro & EXT4_FRO_COM_BIGALLOC)
 200                 ext4_dbg(DEBUG_FS, DBG_NONE "bigalloc\n");
 201         if (features_ro & EXT4_FRO_COM_METADATA_CSUM)
 202                 ext4_dbg(DEBUG_FS, DBG_NONE "metadata_csum\n");
 203 }
 204
 205 int ext4_fs_check_features(struct ext4_fs *fs, bool *read_only)
 206 {
 207         ext4_assert(fs && read_only);
 208         uint32_t v;
 209         if (ext4_get32(&fs->sb, rev_level) == 0) {
 210                 *read_only = false;
 211                 return EOK;
 212         }
 213
 214         ext4_dbg(DEBUG_FS, DBG_INFO "sblock features_incompatible:\n");
 215         ext4_fs_debug_features_inc(ext4_get32(&fs->sb, features_incompatible));
 216
 217         ext4_dbg(DEBUG_FS, DBG_INFO "sblock features_compatible:\n");
 218         ext4_fs_debug_features_comp(ext4_get32(&fs->sb, features_compatible));
 219
 220         ext4_dbg(DEBUG_FS, DBG_INFO "sblock features_read_only:\n");
 221         ext4_fs_debug_features_ro(ext4_get32(&fs->sb, features_read_only));
 222
 223         /*Check features_incompatible*/
 224         v = (ext4_get32(&fs->sb, features_incompatible) &
 225              (~CONFIG_SUPPORTED_FINCOM));
 226         if (v) {
 227                 ext4_dbg(DEBUG_FS, DBG_ERROR
 228                                 "sblock has unsupported features incompatible:\n");
 229                 ext4_fs_debug_features_inc(v);
 230                 return ENOTSUP;
 231         }
 232
 233         /*Check features_read_only*/
 234         v = ext4_get32(&fs->sb, features_read_only);
 235         v &= ~CONFIG_SUPPORTED_FRO_COM;
 236         if (v) {
 237                 ext4_dbg(DEBUG_FS, DBG_WARN
 238                         "sblock has unsupported features read only:\n");
 239                 ext4_fs_debug_features_ro(v);
 240                 *read_only = true;
 241                 return EOK;
 242         }
 243         *read_only = false;
 244
 245         return EOK;
 246 }
 247
 248 /**@brief Determine whether the block is inside the group.
 249  * @param baddr   block address
 250  * @param bgid    block group id
 251  * @return Error code
 252  */
 253 static bool ext4_block_in_group(struct ext4_sblock *s, ext4_fsblk_t baddr,
 254                                 uint32_t bgid)
 255 {
 256         uint32_t actual_bgid;
 257         actual_bgid = ext4_balloc_get_bgid_of_block(s, baddr);
 258         if (actual_bgid == bgid)
 259                 return true;
 260         return false;
 261 }
 262
 263 /**@brief   To avoid calling the atomic setbit hundreds or thousands of times, we only
 264  *          need to use it within a single byte (to ensure we get endianness right).
 265  *          We can use memset for the rest of the bitmap as there are no other users.
 266  */
 267 static void ext4_fs_mark_bitmap_end(int start_bit, int end_bit, void *bitmap)
 268 {
 269         int i;
 270
 271         if (start_bit >= end_bit)
 272                 return;
 273
 274         for (i = start_bit; (unsigned)i < ((start_bit + 7) & ~7UL); i++)
 275                 ext4_bmap_bit_set(bitmap, i);
 276
 277         if (i < end_bit)
 278                 memset((char *)bitmap + (i >> 3), 0xff, (end_bit - i) >> 3);
 279 }
 280
 281 /**@brief Initialize block bitmap in block group.
 282  * @param bg_ref Reference to block group
 283  * @return Error code
 284  */
 285 static int ext4_fs_init_block_bitmap(struct ext4_block_group_ref *bg_ref)
 286 {
 287         struct ext4_sblock *sb = &bg_ref->fs->sb;
 288         struct ext4_bgroup *bg = bg_ref->block_group;
 289         int rc;
 290
 291         uint32_t i, bit, bit_max;
 292         uint32_t group_blocks;
 293         uint16_t inode_size = ext4_get16(sb, inode_size);
 294         uint32_t block_size = ext4_sb_get_block_size(sb);
 295         uint32_t inodes_per_group = ext4_get32(sb, inodes_per_group);
 296
 297         ext4_fsblk_t bmp_blk = ext4_bg_get_block_bitmap(bg, sb);
 298         ext4_fsblk_t bmp_inode = ext4_bg_get_inode_bitmap(bg, sb);
 299         ext4_fsblk_t inode_table = ext4_bg_get_inode_table_first_block(bg, sb);
 300         ext4_fsblk_t first_bg = ext4_balloc_get_block_of_bgid(sb, bg_ref->index);
 301
 302         uint32_t dsc_per_block =  block_size / ext4_sb_get_desc_size(sb);
 303
 304         bool flex_bg = ext4_sb_feature_incom(sb, EXT4_FINCOM_FLEX_BG);
 305         bool meta_bg = ext4_sb_feature_incom(sb, EXT4_FINCOM_META_BG);
 306
 307         uint32_t inode_table_bcnt = inodes_per_group * inode_size / block_size;
 308
 309         struct ext4_block block_bitmap;
 310         rc = ext4_trans_block_get_noread(bg_ref->fs->bdev, &block_bitmap, bmp_blk);
 311         if (rc != EOK)
 312                 return rc;
 313
 314         memset(block_bitmap.data, 0, block_size);
 315         bit_max = ext4_sb_is_super_in_bg(sb, bg_ref->index);
 316
 317         uint32_t count = ext4_sb_first_meta_bg(sb) * dsc_per_block;
 318         if (!meta_bg || bg_ref->index < count) {
 319                 if (bit_max) {
 320                         bit_max += ext4_bg_num_gdb(sb, bg_ref->index);
 321                         bit_max += ext4_get16(sb, s_reserved_gdt_blocks);
 322                 }
 323         } else { /* For META_BG_BLOCK_GROUPS */
 324                 bit_max += ext4_bg_num_gdb(sb, bg_ref->index);
 325         }
 326         for (bit = 0; bit < bit_max; bit++)
 327                 ext4_bmap_bit_set(block_bitmap.data, bit);
 328
 329         if (bg_ref->index == ext4_block_group_cnt(sb) - 1) {
 330                 /*
 331                  * Even though mke2fs always initialize first and last group
 332                  * if some other tool enabled the EXT4_BG_BLOCK_UNINIT we need
 333                  * to make sure we calculate the right free blocks
 334                  */
 335
 336                 group_blocks = (uint32_t)(ext4_sb_get_blocks_cnt(sb) -
 337                                           ext4_get32(sb, first_data_block) -
 338                                           ext4_get32(sb, blocks_per_group) *
 339                                           (ext4_block_group_cnt(sb) - 1));
 340         } else {
 341                 group_blocks = ext4_get32(sb, blocks_per_group);
 342         }
 343
 344         bool in_bg;
 345         in_bg = ext4_block_in_group(sb, bmp_blk, bg_ref->index);
 346         if (!flex_bg || in_bg)
 347                 ext4_bmap_bit_set(block_bitmap.data,
 348                                   (uint32_t)(bmp_blk - first_bg));
 349
 350         in_bg = ext4_block_in_group(sb, bmp_inode, bg_ref->index);
 351         if (!flex_bg || in_bg)
 352                 ext4_bmap_bit_set(block_bitmap.data,
 353                                   (uint32_t)(bmp_inode - first_bg));
 354
 355         for (i = inode_table; i < inode_table + inode_table_bcnt; i++) {
 356                 in_bg = ext4_block_in_group(sb, i, bg_ref->index);
 357                 if (!flex_bg || in_bg)
 358                         ext4_bmap_bit_set(block_bitmap.data,
 359                                           (uint32_t)(i - first_bg));
 360         }
 361         /*
 362          * Also if the number of blocks within the group is
 363          * less than the blocksize * 8 ( which is the size
 364          * of bitmap ), set rest of the block bitmap to 1
 365          */
 366         ext4_fs_mark_bitmap_end(group_blocks, block_size * 8, block_bitmap.data);
 367         ext4_trans_set_block_dirty(block_bitmap.buf);
 368
 369         ext4_balloc_set_bitmap_csum(sb, bg_ref->block_group, block_bitmap.data);
 370         bg_ref->dirty = true;
 371
 372         /* Save bitmap */
 373         return ext4_block_set(bg_ref->fs->bdev, &block_bitmap);
 374 }
 375
 376 /**@brief Initialize i-node bitmap in block group.
 377  * @param bg_ref Reference to block group
 378  * @return Error code
 379  */
 380 static int ext4_fs_init_inode_bitmap(struct ext4_block_group_ref *bg_ref)
 381 {
 382         int rc;
 383         struct ext4_sblock *sb = &bg_ref->fs->sb;
 384         struct ext4_bgroup *bg = bg_ref->block_group;
 385
 386         /* Load bitmap */
 387         ext4_fsblk_t bitmap_block_addr = ext4_bg_get_inode_bitmap(bg, sb);
 388
 389         struct ext4_block b;
 390         rc = ext4_trans_block_get_noread(bg_ref->fs->bdev, &b, bitmap_block_addr);
 391         if (rc != EOK)
 392                 return rc;
 393
 394         /* Initialize all bitmap bits to zero */
 395         uint32_t block_size = ext4_sb_get_block_size(sb);
 396         uint32_t inodes_per_group = ext4_get32(sb, inodes_per_group);
 397
 398         memset(b.data, 0, (inodes_per_group + 7) / 8);
 399
 400         uint32_t start_bit = inodes_per_group;
 401         uint32_t end_bit = block_size * 8;
 402
 403         uint32_t i;
 404         for (i = start_bit; i < ((start_bit + 7) & ~7UL); i++)
 405                 ext4_bmap_bit_set(b.data, i);
 406
 407         if (i < end_bit)
 408                 memset(b.data + (i >> 3), 0xff, (end_bit - i) >> 3);
 409
 410         ext4_trans_set_block_dirty(b.buf);
 411
 412         ext4_ialloc_set_bitmap_csum(sb, bg, b.data);
 413         bg_ref->dirty = true;
 414
 415         /* Save bitmap */
 416         return ext4_block_set(bg_ref->fs->bdev, &b);
 417 }
 418
 419 /**@brief Initialize i-node table in block group.
 420  * @param bg_ref Reference to block group
 421  * @return Error code
 422  */
 423 static int ext4_fs_init_inode_table(struct ext4_block_group_ref *bg_ref)
 424 {
 425         struct ext4_sblock *sb = &bg_ref->fs->sb;
 426         struct ext4_bgroup *bg = bg_ref->block_group;
 427
 428         uint32_t inode_size = ext4_get32(sb, inode_size);
 429         uint32_t block_size = ext4_sb_get_block_size(sb);
 430         uint32_t inodes_per_block = block_size / inode_size;
 431         uint32_t inodes_in_group = ext4_inodes_in_group_cnt(sb, bg_ref->index);
 432         uint32_t table_blocks = inodes_in_group / inodes_per_block;
 433         ext4_fsblk_t fblock;
 434
 435         if (inodes_in_group % inodes_per_block)
 436                 table_blocks++;
 437
 438         /* Compute initialization bounds */
 439         ext4_fsblk_t first_block = ext4_bg_get_inode_table_first_block(bg, sb);
 440
 441         ext4_fsblk_t last_block = first_block + table_blocks - 1;
 442
 443         /* Initialization of all itable blocks */
 444         for (fblock = first_block; fblock <= last_block; ++fblock) {
 445                 struct ext4_block b;
 446                 int rc = ext4_trans_block_get_noread(bg_ref->fs->bdev, &b, fblock);
 447                 if (rc != EOK)
 448                         return rc;
 449
 450                 memset(b.data, 0, block_size);
 451                 ext4_trans_set_block_dirty(b.buf);
 452
 453                 ext4_block_set(bg_ref->fs->bdev, &b);
 454                 if (rc != EOK)
 455                         return rc;
 456         }
 457
 458         return EOK;
 459 }
 460
 461 static ext4_fsblk_t ext4_fs_get_descriptor_block(struct ext4_sblock *s,
 462                                              uint32_t bgid,
 463                                              uint32_t dsc_per_block)
 464 {
 465         uint32_t first_meta_bg, dsc_id;
 466         int has_super = 0;
 467         dsc_id = bgid / dsc_per_block;
 468         first_meta_bg = ext4_sb_first_meta_bg(s);
 469
 470         bool meta_bg = ext4_sb_feature_incom(s, EXT4_FINCOM_META_BG);
 471
 472         if (!meta_bg || dsc_id < first_meta_bg)
 473                 return ext4_get32(s, first_data_block) + dsc_id + 1;
 474
 475         if (ext4_sb_is_super_in_bg(s, bgid))
 476                 has_super = 1;
 477
 478         return (has_super + ext4_fs_first_bg_block_no(s, bgid));
 479 }
 480
 481 /**@brief  Compute checksum of block group descriptor.
 482  * @param sb   Superblock
 483  * @param bgid Index of block group in the filesystem
 484  * @param bg   Block group to compute checksum for
 485  * @return Checksum value
 486  */
 487 static uint16_t ext4_fs_bg_checksum(struct ext4_sblock *sb, uint32_t bgid,
 488                                     struct ext4_bgroup *bg)
 489 {
 490         /* If checksum not supported, 0 will be returned */
 491         uint16_t crc = 0;
 492 #if CONFIG_META_CSUM_ENABLE
 493         /* Compute the checksum only if the filesystem supports it */
 494         if (ext4_sb_feature_ro_com(sb, EXT4_FRO_COM_METADATA_CSUM)) {
 495                 /* Use metadata_csum algorithm instead */
 496                 uint32_t le32_bgid = to_le32(bgid);
 497                 uint32_t orig_checksum, checksum;
 498
 499                 /* Preparation: temporarily set bg checksum to 0 */
 500                 orig_checksum = bg->checksum;
 501                 bg->checksum = 0;
 502
 503                 /* First calculate crc32 checksum against fs uuid */
 504                 checksum = ext4_crc32c(EXT4_CRC32_INIT, sb->uuid,
 505                                 sizeof(sb->uuid));
 506                 /* Then calculate crc32 checksum against bgid */
 507                 checksum = ext4_crc32c(checksum, &le32_bgid, sizeof(bgid));
 508                 /* Finally calculate crc32 checksum against block_group_desc */
 509                 checksum = ext4_crc32c(checksum, bg, ext4_sb_get_desc_size(sb));
 510                 bg->checksum = orig_checksum;
 511
 512                 crc = checksum & 0xFFFF;
 513                 return crc;
 514         }
 515 #endif
 516         if (ext4_sb_feature_ro_com(sb, EXT4_FRO_COM_GDT_CSUM)) {
 517                 uint8_t *base = (uint8_t *)bg;
 518                 uint8_t *checksum = (uint8_t *)&bg->checksum;
 519
 520                 uint32_t offset = (uint32_t)(checksum - base);
 521
 522                 /* Convert block group index to little endian */
 523                 uint32_t group = to_le32(bgid);
 524
 525                 /* Initialization */
 526                 crc = ext4_bg_crc16(~0, sb->uuid, sizeof(sb->uuid));
 527
 528                 /* Include index of block group */
 529                 crc = ext4_bg_crc16(crc, (uint8_t *)&group, sizeof(group));
 530
 531                 /* Compute crc from the first part (stop before checksum field)
 532                  */
 533                 crc = ext4_bg_crc16(crc, (uint8_t *)bg, offset);
 534
 535                 /* Skip checksum */
 536                 offset += sizeof(bg->checksum);
 537
 538                 /* Checksum of the rest of block group descriptor */
 539                 if ((ext4_sb_feature_incom(sb, EXT4_FINCOM_64BIT)) &&
 540                     (offset < ext4_sb_get_desc_size(sb))) {
 541
 542                         const uint8_t *start = ((uint8_t *)bg) + offset;
 543                         size_t len = ext4_sb_get_desc_size(sb) - offset;
 544                         crc = ext4_bg_crc16(crc, start, len);
 545                 }
 546         }
 547         return crc;
 548 }
 549
 550 #if CONFIG_META_CSUM_ENABLE
 551 static bool ext4_fs_verify_bg_csum(struct ext4_sblock *sb,
 552                                    uint32_t bgid,
 553                                    struct ext4_bgroup *bg)
 554 {
 555         if (!ext4_sb_feature_ro_com(sb, EXT4_FRO_COM_METADATA_CSUM))
 556                 return true;
 557
 558         return ext4_fs_bg_checksum(sb, bgid, bg) == to_le16(bg->checksum);
 559 }
 560 #else
 561 #define ext4_fs_verify_bg_csum(...) true
 562 #endif
 563
 564 int ext4_fs_get_block_group_ref(struct ext4_fs *fs, uint32_t bgid,
 565                                 struct ext4_block_group_ref *ref)
 566 {
 567         /* Compute number of descriptors, that fits in one data block */
 568         uint32_t block_size = ext4_sb_get_block_size(&fs->sb);
 569         uint32_t dsc_cnt = block_size / ext4_sb_get_desc_size(&fs->sb);
 570
 571         /* Block group descriptor table starts at the next block after
 572          * superblock */
 573         uint64_t block_id = ext4_fs_get_descriptor_block(&fs->sb, bgid, dsc_cnt);
 574
 575         uint32_t offset = (bgid % dsc_cnt) * ext4_sb_get_desc_size(&fs->sb);
 576
 577         int rc = ext4_trans_block_get(fs->bdev, &ref->block, block_id);
 578         if (rc != EOK)
 579                 return rc;
 580
 581         ref->block_group = (void *)(ref->block.data + offset);
 582         ref->fs = fs;
 583         ref->index = bgid;
 584         ref->dirty = false;
 585         struct ext4_bgroup *bg = ref->block_group;
 586
 587         if (!ext4_fs_verify_bg_csum(&fs->sb, bgid, bg)) {
 588                 ext4_dbg(DEBUG_FS,
 589                          DBG_WARN "Block group descriptor checksum failed."
 590                          "Block group index: %" PRIu32"\n",
 591                          bgid);
 592         }
 593
 594         if (ext4_bg_has_flag(bg, EXT4_BLOCK_GROUP_BLOCK_UNINIT)) {
 595                 rc = ext4_fs_init_block_bitmap(ref);
 596                 if (rc != EOK) {
 597                         ext4_block_set(fs->bdev, &ref->block);
 598                         return rc;
 599                 }
 600                 ext4_bg_clear_flag(bg, EXT4_BLOCK_GROUP_BLOCK_UNINIT);
 601                 ref->dirty = true;
 602         }
 603
 604         if (ext4_bg_has_flag(bg, EXT4_BLOCK_GROUP_INODE_UNINIT)) {
 605                 rc = ext4_fs_init_inode_bitmap(ref);
 606                 if (rc != EOK) {
 607                         ext4_block_set(ref->fs->bdev, &ref->block);
 608                         return rc;
 609                 }
 610
 611                 ext4_bg_clear_flag(bg, EXT4_BLOCK_GROUP_INODE_UNINIT);
 612
 613                 if (!ext4_bg_has_flag(bg, EXT4_BLOCK_GROUP_ITABLE_ZEROED)) {
 614                         rc = ext4_fs_init_inode_table(ref);
 615                         if (rc != EOK) {
 616                                 ext4_block_set(fs->bdev, &ref->block);
 617                                 return rc;
 618                         }
 619
 620                         ext4_bg_set_flag(bg, EXT4_BLOCK_GROUP_ITABLE_ZEROED);
 621                 }
 622
 623                 ref->dirty = true;
 624         }
 625
 626         return EOK;
 627 }
 628
 629 int ext4_fs_put_block_group_ref(struct ext4_block_group_ref *ref)
 630 {
 631         /* Check if reference modified */
 632         if (ref->dirty) {
 633                 /* Compute new checksum of block group */
 634                 uint16_t cs;
 635                 cs = ext4_fs_bg_checksum(&ref->fs->sb, ref->index,
 636                                          ref->block_group);
 637                 ref->block_group->checksum = to_le16(cs);
 638
 639                 /* Mark block dirty for writing changes to physical device */
 640                 ext4_trans_set_block_dirty(ref->block.buf);
 641         }
 642
 643         /* Put back block, that contains block group descriptor */
 644         return ext4_block_set(ref->fs->bdev, &ref->block);
 645 }
 646
 647 #if CONFIG_META_CSUM_ENABLE
 648 static uint32_t ext4_fs_inode_checksum(struct ext4_inode_ref *inode_ref)
 649 {
 650         uint32_t checksum = 0;
 651         struct ext4_sblock *sb = &inode_ref->fs->sb;
 652         uint16_t inode_size = ext4_get16(sb, inode_size);
 653
 654         if (ext4_sb_feature_ro_com(sb, EXT4_FRO_COM_METADATA_CSUM)) {
 655                 uint32_t orig_checksum;
 656
 657                 uint32_t ino_index = to_le32(inode_ref->index);
 658                 uint32_t ino_gen =
 659                         to_le32(ext4_inode_get_generation(inode_ref->inode));
 660
 661                 /* Preparation: temporarily set bg checksum to 0 */
 662                 orig_checksum = ext4_inode_get_csum(sb, inode_ref->inode);
 663                 ext4_inode_set_csum(sb, inode_ref->inode, 0);
 664
 665                 /* First calculate crc32 checksum against fs uuid */
 666                 checksum = ext4_crc32c(EXT4_CRC32_INIT, sb->uuid,
 667                                        sizeof(sb->uuid));
 668                 /* Then calculate crc32 checksum against inode number
 669                  * and inode generation */
 670                 checksum = ext4_crc32c(checksum, &ino_index, sizeof(ino_index));
 671                 checksum = ext4_crc32c(checksum, &ino_gen, sizeof(ino_gen));
 672                 /* Finally calculate crc32 checksum against
 673                  * the entire inode */
 674                 checksum = ext4_crc32c(checksum, inode_ref->inode, inode_size);
 675                 ext4_inode_set_csum(sb, inode_ref->inode, orig_checksum);
 676
 677                 /* If inode size is not large enough to hold the
 678                  * upper 16bit of the checksum */
 679                 if (inode_size == EXT4_GOOD_OLD_INODE_SIZE)
 680                         checksum &= 0xFFFF;
 681
 682         }
 683         return checksum;
 684 }
 685 #else
 686 #define ext4_fs_inode_checksum(...) 0
 687 #endif
 688
 689 static void ext4_fs_set_inode_checksum(struct ext4_inode_ref *inode_ref)
 690 {
 691         struct ext4_sblock *sb = &inode_ref->fs->sb;
 692         if (!ext4_sb_feature_ro_com(sb, EXT4_FRO_COM_METADATA_CSUM))
 693                 return;
 694
 695         uint32_t csum = ext4_fs_inode_checksum(inode_ref);
 696         ext4_inode_set_csum(sb, inode_ref->inode, csum);
 697 }
 698
 699 #if CONFIG_META_CSUM_ENABLE
 700 static bool ext4_fs_verify_inode_csum(struct ext4_inode_ref *inode_ref)
 701 {
 702         struct ext4_sblock *sb = &inode_ref->fs->sb;
 703         if (!ext4_sb_feature_ro_com(sb, EXT4_FRO_COM_METADATA_CSUM))
 704                 return true;
 705
 706         return ext4_inode_get_csum(sb, inode_ref->inode) ==
 707                 ext4_fs_inode_checksum(inode_ref);
 708 }
 709 #else
 710 #define ext4_fs_verify_inode_csum(...) true
 711 #endif
 712
 713 static int
 714 __ext4_fs_get_inode_ref(struct ext4_fs *fs, uint32_t index,
 715                         struct ext4_inode_ref *ref,
 716                         bool initialized)
 717 {
 718         /* Compute number of i-nodes, that fits in one data block */
 719         uint32_t inodes_per_group = ext4_get32(&fs->sb, inodes_per_group);
 720
 721         /*
 722          * Inode numbers are 1-based, but it is simpler to work with 0-based
 723          * when computing indices
 724          */
 725         index -= 1;
 726         uint32_t block_group = index / inodes_per_group;
 727         uint32_t offset_in_group = index % inodes_per_group;
 728
 729         /* Load block group, where i-node is located */
 730         struct ext4_block_group_ref bg_ref;
 731
 732         int rc = ext4_fs_get_block_group_ref(fs, block_group, &bg_ref);
 733         if (rc != EOK) {
 734                 return rc;
 735         }
 736
 737         /* Load block address, where i-node table is located */
 738         ext4_fsblk_t inode_table_start =
 739             ext4_bg_get_inode_table_first_block(bg_ref.block_group, &fs->sb);
 740
 741         /* Put back block group reference (not needed more) */
 742         rc = ext4_fs_put_block_group_ref(&bg_ref);
 743         if (rc != EOK) {
 744                 return rc;
 745         }
 746
 747         /* Compute position of i-node in the block group */
 748         uint16_t inode_size = ext4_get16(&fs->sb, inode_size);
 749         uint32_t block_size = ext4_sb_get_block_size(&fs->sb);
 750         uint32_t byte_offset_in_group = offset_in_group * inode_size;
 751
 752         /* Compute block address */
 753         ext4_fsblk_t block_id =
 754             inode_table_start + (byte_offset_in_group / block_size);
 755
 756         rc = ext4_trans_block_get(fs->bdev, &ref->block, block_id);
 757         if (rc != EOK) {
 758                 return rc;
 759         }
 760
 761         /* Compute position of i-node in the data block */
 762         uint32_t offset_in_block = byte_offset_in_group % block_size;
 763         ref->inode = (struct ext4_inode *)(ref->block.data + offset_in_block);
 764
 765         /* We need to store the original value of index in the reference */
 766         ref->index = index + 1;
 767         ref->fs = fs;
 768         ref->dirty = false;
 769
 770         if (initialized && !ext4_fs_verify_inode_csum(ref)) {
 771                 ext4_dbg(DEBUG_FS,
 772                         DBG_WARN "Inode checksum failed."
 773                         "Inode: %" PRIu32"\n",
 774                         ref->index);
 775         }
 776
 777         return EOK;
 778 }
 779
 780 int ext4_fs_get_inode_ref(struct ext4_fs *fs, uint32_t index,
 781                           struct ext4_inode_ref *ref)
 782 {
 783         return __ext4_fs_get_inode_ref(fs, index, ref, true);
 784 }
 785
 786 int ext4_fs_put_inode_ref(struct ext4_inode_ref *ref)
 787 {
 788         /* Check if reference modified */
 789         if (ref->dirty) {
 790                 /* Mark block dirty for writing changes to physical device */
 791                 ext4_fs_set_inode_checksum(ref);
 792                 ext4_trans_set_block_dirty(ref->block.buf);
 793         }
 794
 795         /* Put back block, that contains i-node */
 796         return ext4_block_set(ref->fs->bdev, &ref->block);
 797 }
 798
 799 void ext4_fs_inode_blocks_init(struct ext4_fs *fs,
 800                                struct ext4_inode_ref *inode_ref)
 801 {
 802         int i;
 803         struct ext4_inode *inode = inode_ref->inode;
 804
 805         for (i = 0; i < EXT4_INODE_BLOCKS; i++)
 806                 inode->blocks[i] = 0;
 807
 808         (void)fs;
 809 #if CONFIG_EXTENT_ENABLE
 810         /* Initialize extents if needed */
 811         if (ext4_sb_feature_incom(&fs->sb, EXT4_FINCOM_EXTENTS)) {
 812                 ext4_inode_set_flag(inode, EXT4_INODE_FLAG_EXTENTS);
 813
 814                 /* Initialize extent root header */
 815                 ext4_extent_tree_init(inode_ref);
 816         }
 817 #endif
 818 }
 819
 820 uint32_t ext4_fs_correspond_inode_mode(int filetype)
 821 {
 822         switch (filetype) {
 823         case EXT4_DE_DIR:
 824                 return EXT4_INODE_MODE_DIRECTORY;
 825         case EXT4_DE_REG_FILE:
 826                 return EXT4_INODE_MODE_FILE;
 827         case EXT4_DE_SYMLINK:
 828                 return EXT4_INODE_MODE_SOFTLINK;
 829         default:
 830                 /* FIXME: right now we only support 3 file type. */
 831                 ext4_assert(0);
 832         }
 833         return 0;
 834 }
 835
 836 int ext4_fs_alloc_inode(struct ext4_fs *fs, struct ext4_inode_ref *inode_ref,
 837                         int filetype)
 838 {
 839         /* Check if newly allocated i-node will be a directory */
 840         bool is_dir;
 841         uint16_t inode_size = ext4_get16(&fs->sb, inode_size);
 842
 843         is_dir = (filetype == EXT4_DE_DIR);
 844
 845         /* Allocate inode by allocation algorithm */
 846         uint32_t index;
 847         int rc = ext4_ialloc_alloc_inode(fs, &index, is_dir);
 848         if (rc != EOK)
 849                 return rc;
 850
 851         /* Load i-node from on-disk i-node table */
 852         rc = __ext4_fs_get_inode_ref(fs, index, inode_ref, false);
 853         if (rc != EOK) {
 854                 ext4_ialloc_free_inode(fs, index, is_dir);
 855                 return rc;
 856         }
 857
 858         /* Initialize i-node */
 859         struct ext4_inode *inode = inode_ref->inode;
 860
 861         uint32_t mode;
 862         if (is_dir) {
 863                 /*
 864                  * Default directory permissions to be compatible with other
 865                  * systems
 866                  * 0777 (octal) == rwxrwxrwx
 867                  */
 868
 869                 mode = 0777;
 870                 mode |= EXT4_INODE_MODE_DIRECTORY;
 871         } else {
 872                 /*
 873                  * Default file permissions to be compatible with other systems
 874                  * 0666 (octal) == rw-rw-rw-
 875                  */
 876
 877                 mode = 0666;
 878                 mode |= ext4_fs_correspond_inode_mode(filetype);
 879         }
 880         ext4_inode_set_mode(&fs->sb, inode, mode);
 881
 882         ext4_inode_set_links_cnt(inode, 0);
 883         ext4_inode_set_uid(inode, 0);
 884         ext4_inode_set_gid(inode, 0);
 885         ext4_inode_set_size(inode, 0);
 886         ext4_inode_set_access_time(inode, 0);
 887         ext4_inode_set_change_inode_time(inode, 0);
 888         ext4_inode_set_modif_time(inode, 0);
 889         ext4_inode_set_del_time(inode, 0);
 890         ext4_inode_set_blocks_count(&fs->sb, inode, 0);
 891         ext4_inode_set_flags(inode, 0);
 892         ext4_inode_set_generation(inode, 0);
 893         if (inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
 894                 uint16_t off = offsetof(struct ext4_inode, extra_isize);
 895                 uint16_t size = sizeof(struct ext4_inode) - off;
 896                 ext4_inode_set_extra_isize(inode, size);
 897         }
 898
 899         /* Reset blocks array. For symbolic link inode, just
 900          * fill in blocks with 0 */
 901         if (ext4_inode_is_type(&fs->sb, inode, EXT4_INODE_MODE_SOFTLINK)) {
 902                 for (int i = 0; i < EXT4_INODE_BLOCKS; i++)
 903                         inode->blocks[i] = 0;
 904
 905         } else {
 906                 ext4_fs_inode_blocks_init(fs, inode_ref);
 907         }
 908
 909         inode_ref->dirty = true;
 910
 911         return EOK;
 912 }
 913
 914 int ext4_fs_free_inode(struct ext4_inode_ref *inode_ref)
 915 {
 916         struct ext4_fs *fs = inode_ref->fs;
 917         uint32_t offset;
 918         uint32_t suboff;
 919         int rc;
 920 #if CONFIG_EXTENT_ENABLE
 921         /* For extents must be data block destroyed by other way */
 922         if ((ext4_sb_feature_incom(&fs->sb, EXT4_FINCOM_EXTENTS)) &&
 923             (ext4_inode_has_flag(inode_ref->inode, EXT4_INODE_FLAG_EXTENTS))) {
 924                 /* Data structures are released during truncate operation... */
 925                 goto finish;
 926         }
 927 #endif
 928         /* Release all indirect (no data) blocks */
 929
 930         /* 1) Single indirect */
 931         ext4_fsblk_t fblock = ext4_inode_get_indirect_block(inode_ref->inode, 0);
 932         if (fblock != 0) {
 933                 int rc = ext4_balloc_free_block(inode_ref, fblock);
 934                 if (rc != EOK)
 935                         return rc;
 936
 937                 ext4_inode_set_indirect_block(inode_ref->inode, 0, 0);
 938         }
 939
 940         uint32_t block_size = ext4_sb_get_block_size(&fs->sb);
 941         uint32_t count = block_size / sizeof(uint32_t);
 942
 943         struct ext4_block block;
 944
 945         /* 2) Double indirect */
 946         fblock = ext4_inode_get_indirect_block(inode_ref->inode, 1);
 947         if (fblock != 0) {
 948                 int rc = ext4_trans_block_get(fs->bdev, &block, fblock);
 949                 if (rc != EOK)
 950                         return rc;
 951
 952                 ext4_fsblk_t ind_block;
 953                 for (offset = 0; offset < count; ++offset) {
 954                         ind_block = to_le32(((uint32_t *)block.data)[offset]);
 955
 956                         if (ind_block == 0)
 957                                 continue;
 958                         rc = ext4_balloc_free_block(inode_ref, ind_block);
 959                         if (rc != EOK) {
 960                                 ext4_block_set(fs->bdev, &block);
 961                                 return rc;
 962                         }
 963
 964                 }
 965
 966                 ext4_block_set(fs->bdev, &block);
 967                 rc = ext4_balloc_free_block(inode_ref, fblock);
 968                 if (rc != EOK)
 969                         return rc;
 970
 971                 ext4_inode_set_indirect_block(inode_ref->inode, 1, 0);
 972         }
 973
 974         /* 3) Tripple indirect */
 975         struct ext4_block subblock;
 976         fblock = ext4_inode_get_indirect_block(inode_ref->inode, 2);
 977         if (fblock == 0)
 978                 goto finish;
 979         rc = ext4_trans_block_get(fs->bdev, &block, fblock);
 980         if (rc != EOK)
 981                 return rc;
 982
 983         ext4_fsblk_t ind_block;
 984         for (offset = 0; offset < count; ++offset) {
 985                 ind_block = to_le32(((uint32_t *)block.data)[offset]);
 986
 987                 if (ind_block == 0)
 988                         continue;
 989                 rc = ext4_trans_block_get(fs->bdev, &subblock,
 990                                 ind_block);
 991                 if (rc != EOK) {
 992                         ext4_block_set(fs->bdev, &block);
 993                         return rc;
 994                 }
 995
 996                 ext4_fsblk_t ind_subblk;
 997                 for (suboff = 0; suboff < count; ++suboff) {
 998                         ind_subblk = to_le32(((uint32_t *)subblock.data)[suboff]);
 999
1000                         if (ind_subblk == 0)
1001                                 continue;
1002                         rc = ext4_balloc_free_block(inode_ref, ind_subblk);
1003                         if (rc != EOK) {
1004                                 ext4_block_set(fs->bdev, &subblock);
1005                                 ext4_block_set(fs->bdev, &block);
1006                                 return rc;
1007                         }
1008
1009                 }
1010
1011                 ext4_block_set(fs->bdev, &subblock);
1012
1013                 rc = ext4_balloc_free_block(inode_ref,
1014                                 ind_block);
1015                 if (rc != EOK) {
1016                         ext4_block_set(fs->bdev, &block);
1017                         return rc;
1018                 }
1019
1020         }
1021
1022         ext4_block_set(fs->bdev, &block);
1023         rc = ext4_balloc_free_block(inode_ref, fblock);
1024         if (rc != EOK)
1025                 return rc;
1026
1027         ext4_inode_set_indirect_block(inode_ref->inode, 2, 0);
1028 finish:
1029         /* Mark inode dirty for writing to the physical device */
1030         inode_ref->dirty = true;
1031
1032         /* Free block with extended attributes if present */
1033         ext4_fsblk_t xattr_block =
1034             ext4_inode_get_file_acl(inode_ref->inode, &fs->sb);
1035         if (xattr_block) {
1036                 int rc = ext4_balloc_free_block(inode_ref, xattr_block);
1037                 if (rc != EOK)
1038                         return rc;
1039
1040                 ext4_inode_set_file_acl(inode_ref->inode, &fs->sb, 0);
1041         }
1042
1043         /* Free inode by allocator */
1044         if (ext4_inode_is_type(&fs->sb, inode_ref->inode,
1045                                EXT4_INODE_MODE_DIRECTORY))
1046                 rc = ext4_ialloc_free_inode(fs, inode_ref->index, true);
1047         else
1048                 rc = ext4_ialloc_free_inode(fs, inode_ref->index, false);
1049
1050         return rc;
1051 }
1052
1053
1054 /**@brief Release data block from i-node
1055  * @param inode_ref I-node to release block from
1056  * @param iblock    Logical block to be released
1057  * @return Error code
1058  */
1059 static int ext4_fs_release_inode_block(struct ext4_inode_ref *inode_ref,
1060                                 uint32_t iblock)
1061 {
1062         ext4_fsblk_t fblock;
1063
1064         struct ext4_fs *fs = inode_ref->fs;
1065
1066         /* Extents are handled otherwise = there is not support in this function
1067          */
1068         ext4_assert(!(
1069             ext4_sb_feature_incom(&fs->sb, EXT4_FINCOM_EXTENTS) &&
1070             (ext4_inode_has_flag(inode_ref->inode, EXT4_INODE_FLAG_EXTENTS))));
1071
1072         struct ext4_inode *inode = inode_ref->inode;
1073
1074         /* Handle simple case when we are dealing with direct reference */
1075         if (iblock < EXT4_INODE_DIRECT_BLOCK_COUNT) {
1076                 fblock = ext4_inode_get_direct_block(inode, iblock);
1077
1078                 /* Sparse file */
1079                 if (fblock == 0)
1080                         return EOK;
1081
1082                 ext4_inode_set_direct_block(inode, iblock, 0);
1083                 return ext4_balloc_free_block(inode_ref, fblock);
1084         }
1085
1086         /* Determine the indirection level needed to get the desired block */
1087         unsigned int level = 0;
1088         unsigned int i;
1089         for (i = 1; i < 4; i++) {
1090                 if (iblock < fs->inode_block_limits[i]) {
1091                         level = i;
1092                         break;
1093                 }
1094         }
1095
1096         if (level == 0)
1097                 return EIO;
1098
1099         /* Compute offsets for the topmost level */
1100         uint32_t block_offset_in_level =
1101             iblock - fs->inode_block_limits[level - 1];
1102         ext4_fsblk_t current_block =
1103             ext4_inode_get_indirect_block(inode, level - 1);
1104         uint32_t offset_in_block =
1105             block_offset_in_level / fs->inode_blocks_per_level[level - 1];
1106
1107         /*
1108          * Navigate through other levels, until we find the block number
1109          * or find null reference meaning we are dealing with sparse file
1110          */
1111         struct ext4_block block;
1112
1113         while (level > 0) {
1114
1115                 /* Sparse check */
1116                 if (current_block == 0)
1117                         return EOK;
1118
1119                 int rc = ext4_trans_block_get(fs->bdev, &block, current_block);
1120                 if (rc != EOK)
1121                         return rc;
1122
1123                 current_block =
1124                     to_le32(((uint32_t *)block.data)[offset_in_block]);
1125
1126                 /* Set zero if physical data block address found */
1127                 if (level == 1) {
1128                         ((uint32_t *)block.data)[offset_in_block] = to_le32(0);
1129                         ext4_trans_set_block_dirty(block.buf);
1130                 }
1131
1132                 rc = ext4_block_set(fs->bdev, &block);
1133                 if (rc != EOK)
1134                         return rc;
1135
1136                 level--;
1137
1138                 /*
1139                  * If we are on the last level, break here as
1140                  * there is no next level to visit
1141                  */
1142                 if (level == 0)
1143                         break;
1144
1145                 /* Visit the next level */
1146                 block_offset_in_level %= fs->inode_blocks_per_level[level];
1147                 offset_in_block = block_offset_in_level /
1148                                   fs->inode_blocks_per_level[level - 1];
1149         }
1150
1151         fblock = current_block;
1152         if (fblock == 0)
1153                 return EOK;
1154
1155         /* Physical block is not referenced, it can be released */
1156         return ext4_balloc_free_block(inode_ref, fblock);
1157 }
1158
1159 int ext4_fs_truncate_inode(struct ext4_inode_ref *inode_ref, uint64_t new_size)
1160 {
1161         struct ext4_sblock *sb = &inode_ref->fs->sb;
1162         uint32_t i;
1163         int r;
1164
1165         /* Check flags, if i-node can be truncated */
1166         if (!ext4_inode_can_truncate(sb, inode_ref->inode))
1167                 return EINVAL;
1168
1169         /* If sizes are equal, nothing has to be done. */
1170         uint64_t old_size = ext4_inode_get_size(sb, inode_ref->inode);
1171         if (old_size == new_size)
1172                 return EOK;
1173
1174         /* It's not supported to make the larger file by truncate operation */
1175         if (old_size < new_size)
1176                 return EINVAL;
1177
1178         bool v;
1179         v = ext4_inode_is_type(sb, inode_ref->inode, EXT4_INODE_MODE_SOFTLINK);
1180         if (v && old_size < sizeof(inode_ref->inode->blocks) &&
1181             !ext4_inode_get_blocks_count(sb, inode_ref->inode)) {
1182                 char *content = (char *)inode_ref->inode->blocks + new_size;
1183                 memset(content, 0,
1184                        sizeof(inode_ref->inode->blocks) - (uint32_t)new_size);
1185                 ext4_inode_set_size(inode_ref->inode, new_size);
1186                 inode_ref->dirty = true;
1187
1188                 return EOK;
1189         }
1190
1191         /* Compute how many blocks will be released */
1192         uint32_t block_size = ext4_sb_get_block_size(sb);
1193         uint32_t new_blocks_cnt = (uint32_t)((new_size + block_size - 1) / block_size);
1194         uint32_t old_blocks_cnt = (uint32_t)((old_size + block_size - 1) / block_size);
1195         uint32_t diff_blocks_cnt = old_blocks_cnt - new_blocks_cnt;
1196 #if CONFIG_EXTENT_ENABLE
1197         if ((ext4_sb_feature_incom(sb, EXT4_FINCOM_EXTENTS)) &&
1198             (ext4_inode_has_flag(inode_ref->inode, EXT4_INODE_FLAG_EXTENTS))) {
1199
1200                 /* Extents require special operation */
1201                 if (diff_blocks_cnt) {
1202                         r = ext4_extent_remove_space(inode_ref, new_blocks_cnt,
1203                                                      EXT_MAX_BLOCKS);
1204                         if (r != EOK)
1205                                 return r;
1206
1207                 }
1208         } else
1209 #endif
1210         {
1211                 /* Release data blocks from the end of file */
1212
1213                 /* Starting from 1 because of logical blocks are numbered from 0
1214                  */
1215                 for (i = 0; i < diff_blocks_cnt; ++i) {
1216                         r = ext4_fs_release_inode_block(inode_ref,
1217                                                         new_blocks_cnt + i);
1218                         if (r != EOK)
1219                                 return r;
1220                 }
1221         }
1222
1223         /* Update i-node */
1224         ext4_inode_set_size(inode_ref->inode, new_size);
1225         inode_ref->dirty = true;
1226
1227         return EOK;
1228 }
1229
1230 /**@brief Compute 'goal' for inode index
1231  * @param inode_ref Reference to inode, to allocate block for
1232  * @return goal
1233  */
1234 ext4_fsblk_t ext4_fs_inode_to_goal_block(struct ext4_inode_ref *inode_ref)
1235 {
1236         uint32_t grp_inodes = ext4_get32(&inode_ref->fs->sb, inodes_per_group);
1237         return (inode_ref->index - 1) / grp_inodes;
1238 }
1239
1240 /**@brief Compute 'goal' for allocation algorithm (For blockmap).
1241  * @param inode_ref Reference to inode, to allocate block for
1242  * @param goal
1243  * @return error code
1244  */
1245 int ext4_fs_indirect_find_goal(struct ext4_inode_ref *inode_ref,
1246                                ext4_fsblk_t *goal)
1247 {
1248         int r;
1249         struct ext4_sblock *sb = &inode_ref->fs->sb;
1250         *goal = 0;
1251
1252         uint64_t inode_size = ext4_inode_get_size(sb, inode_ref->inode);
1253         uint32_t block_size = ext4_sb_get_block_size(sb);
1254         uint32_t iblock_cnt = (uint32_t)(inode_size / block_size);
1255
1256         if (inode_size % block_size != 0)
1257                 iblock_cnt++;
1258
1259         /* If inode has some blocks, get last block address + 1 */
1260         if (iblock_cnt > 0) {
1261                 r = ext4_fs_get_inode_dblk_idx(inode_ref, iblock_cnt - 1,
1262                                                goal, false);
1263                 if (r != EOK)
1264                         return r;
1265
1266                 if (*goal != 0) {
1267                         (*goal)++;
1268                         return r;
1269                 }
1270
1271                 /* If goal == 0, sparse file -> continue */
1272         }
1273
1274         /* Identify block group of inode */
1275
1276         uint32_t inodes_per_bg = ext4_get32(sb, inodes_per_group);
1277         uint32_t block_group = (inode_ref->index - 1) / inodes_per_bg;
1278         block_size = ext4_sb_get_block_size(sb);
1279
1280         /* Load block group reference */
1281         struct ext4_block_group_ref bg_ref;
1282         r = ext4_fs_get_block_group_ref(inode_ref->fs, block_group, &bg_ref);
1283         if (r != EOK)
1284                 return r;
1285
1286         struct ext4_bgroup *bg = bg_ref.block_group;
1287
1288         /* Compute indexes */
1289         uint32_t bg_count = ext4_block_group_cnt(sb);
1290         ext4_fsblk_t itab_first_block = ext4_bg_get_inode_table_first_block(bg, sb);
1291         uint16_t itab_item_size = ext4_get16(sb, inode_size);
1292         uint32_t itab_bytes;
1293
1294         /* Check for last block group */
1295         if (block_group < bg_count - 1) {
1296                 itab_bytes = inodes_per_bg * itab_item_size;
1297         } else {
1298                 /* Last block group could be smaller */
1299                 uint32_t inodes_cnt = ext4_get32(sb, inodes_count);
1300
1301                 itab_bytes = (inodes_cnt - ((bg_count - 1) * inodes_per_bg));
1302                 itab_bytes *= itab_item_size;
1303         }
1304
1305         ext4_fsblk_t inode_table_blocks = itab_bytes / block_size;
1306
1307         if (itab_bytes % block_size)
1308                 inode_table_blocks++;
1309
1310         *goal = itab_first_block + inode_table_blocks;
1311
1312         return ext4_fs_put_block_group_ref(&bg_ref);
1313 }
1314
1315 static int ext4_fs_get_inode_dblk_idx_internal(struct ext4_inode_ref *inode_ref,
1316                                        uint64_t iblock, ext4_fsblk_t *fblock,
1317                                        bool extent_create,
1318                                        bool support_unwritten __unused)
1319 {
1320         struct ext4_fs *fs = inode_ref->fs;
1321
1322         /* For empty file is situation simple */
1323         if (ext4_inode_get_size(&fs->sb, inode_ref->inode) == 0) {
1324                 *fblock = 0;
1325                 return EOK;
1326         }
1327
1328         ext4_fsblk_t current_block;
1329
1330         (void)extent_create;
1331 #if CONFIG_EXTENT_ENABLE
1332         /* Handle i-node using extents */
1333         if ((ext4_sb_feature_incom(&fs->sb, EXT4_FINCOM_EXTENTS)) &&
1334             (ext4_inode_has_flag(inode_ref->inode, EXT4_INODE_FLAG_EXTENTS))) {
1335
1336                 ext4_fsblk_t current_fsblk;
1337                 int rc = ext4_extent_get_blocks(inode_ref, iblock, 1,
1338                                 &current_fsblk, extent_create, NULL);
1339                 if (rc != EOK)
1340                         return rc;
1341
1342                 current_block = current_fsblk;
1343                 *fblock = current_block;
1344
1345                 ext4_assert(*fblock || support_unwritten);
1346                 return EOK;
1347         }
1348 #endif
1349
1350         struct ext4_inode *inode = inode_ref->inode;
1351
1352         /* Direct block are read directly from array in i-node structure */
1353         if (iblock < EXT4_INODE_DIRECT_BLOCK_COUNT) {
1354                 current_block =
1355                     ext4_inode_get_direct_block(inode, (uint32_t)iblock);
1356                 *fblock = current_block;
1357                 return EOK;
1358         }
1359
1360         /* Determine indirection level of the target block */
1361         unsigned int l = 0;
1362         unsigned int i;
1363         for (i = 1; i < 4; i++) {
1364                 if (iblock < fs->inode_block_limits[i]) {
1365                         l = i;
1366                         break;
1367                 }
1368         }
1369
1370         if (l == 0)
1371                 return EIO;
1372
1373         /* Compute offsets for the topmost level */
1374         uint32_t blk_off_in_lvl =  iblock - fs->inode_block_limits[l - 1];
1375         current_block = ext4_inode_get_indirect_block(inode, l - 1);
1376         uint32_t off_in_blk = blk_off_in_lvl / fs->inode_blocks_per_level[l - 1];
1377
1378         /* Sparse file */
1379         if (current_block == 0) {
1380                 *fblock = 0;
1381                 return EOK;
1382         }
1383
1384         struct ext4_block block;
1385
1386         /*
1387          * Navigate through other levels, until we find the block number
1388          * or find null reference meaning we are dealing with sparse file
1389          */
1390         while (l > 0) {
1391                 /* Load indirect block */
1392                 int rc = ext4_trans_block_get(fs->bdev, &block, current_block);
1393                 if (rc != EOK)
1394                         return rc;
1395
1396                 /* Read block address from indirect block */
1397                 current_block =
1398                     to_le32(((uint32_t *)block.data)[off_in_blk]);
1399
1400                 /* Put back indirect block untouched */
1401                 rc = ext4_block_set(fs->bdev, &block);
1402                 if (rc != EOK)
1403                         return rc;
1404
1405                 /* Check for sparse file */
1406                 if (current_block == 0) {
1407                         *fblock = 0;
1408                         return EOK;
1409                 }
1410
1411                 /* Jump to the next level */
1412                 l--;
1413
1414                 /* Termination condition - we have address of data block loaded
1415                  */
1416                 if (l == 0)
1417                         break;
1418
1419                 /* Visit the next level */
1420                 blk_off_in_lvl %= fs->inode_blocks_per_level[l];
1421                 off_in_blk = blk_off_in_lvl / fs->inode_blocks_per_level[l - 1];
1422         }
1423
1424         *fblock = current_block;
1425
1426         return EOK;
1427 }
1428
1429
1430 int ext4_fs_get_inode_dblk_idx(struct ext4_inode_ref *inode_ref,
1431                                uint64_t iblock, ext4_fsblk_t *fblock,
1432                                bool support_unwritten)
1433 {
1434         return ext4_fs_get_inode_dblk_idx_internal(inode_ref, iblock, fblock,
1435                                                    false, support_unwritten);
1436 }
1437
1438 int ext4_fs_init_inode_dblk_idx(struct ext4_inode_ref *inode_ref,
1439                                 uint64_t iblock, ext4_fsblk_t *fblock)
1440 {
1441         return ext4_fs_get_inode_dblk_idx_internal(inode_ref, iblock, fblock,
1442                                                    true, true);
1443 }
1444
1445 static int ext4_fs_set_inode_data_block_index(struct ext4_inode_ref *inode_ref,
1446                                        uint32_t iblock, ext4_fsblk_t fblock)
1447 {
1448         struct ext4_fs *fs = inode_ref->fs;
1449
1450 #if CONFIG_EXTENT_ENABLE
1451         /* Handle inode using extents */
1452         if ((ext4_sb_feature_incom(&fs->sb, EXT4_FINCOM_EXTENTS)) &&
1453             (ext4_inode_has_flag(inode_ref->inode, EXT4_INODE_FLAG_EXTENTS))) {
1454                 /* Not reachable */
1455                 return ENOTSUP;
1456         }
1457 #endif
1458
1459         /* Handle simple case when we are dealing with direct reference */
1460         if (iblock < EXT4_INODE_DIRECT_BLOCK_COUNT) {
1461                 ext4_inode_set_direct_block(inode_ref->inode, (uint32_t)iblock,
1462                                             (uint32_t)fblock);
1463                 inode_ref->dirty = true;
1464
1465                 return EOK;
1466         }
1467
1468         /* Determine the indirection level needed to get the desired block */
1469         unsigned int l = 0;
1470         unsigned int i;
1471         for (i = 1; i < 4; i++) {
1472                 if (iblock < fs->inode_block_limits[i]) {
1473                         l = i;
1474                         break;
1475                 }
1476         }
1477
1478         if (l == 0)
1479                 return EIO;
1480
1481         uint32_t block_size = ext4_sb_get_block_size(&fs->sb);
1482
1483         /* Compute offsets for the topmost level */
1484         uint32_t blk_off_in_lvl =  iblock - fs->inode_block_limits[l - 1];
1485         ext4_fsblk_t current_block =
1486                         ext4_inode_get_indirect_block(inode_ref->inode, l - 1);
1487         uint32_t off_in_blk = blk_off_in_lvl / fs->inode_blocks_per_level[l - 1];
1488
1489         ext4_fsblk_t new_blk;
1490
1491         struct ext4_block block;
1492         struct ext4_block new_block;
1493
1494         /* Is needed to allocate indirect block on the i-node level */
1495         if (current_block == 0) {
1496                 /* Allocate new indirect block */
1497                 ext4_fsblk_t goal;
1498                 int rc = ext4_fs_indirect_find_goal(inode_ref, &goal);
1499                 if (rc != EOK)
1500                         return rc;
1501
1502                 rc = ext4_balloc_alloc_block(inode_ref, goal, &new_blk);
1503                 if (rc != EOK)
1504                         return rc;
1505
1506                 /* Update i-node */
1507                 ext4_inode_set_indirect_block(inode_ref->inode, l - 1,
1508                                 (uint32_t)new_blk);
1509                 inode_ref->dirty = true;
1510
1511                 /* Load newly allocated block */
1512                 rc = ext4_trans_block_get_noread(fs->bdev, &new_block, new_blk);
1513                 if (rc != EOK) {
1514                         ext4_balloc_free_block(inode_ref, new_blk);
1515                         return rc;
1516                 }
1517
1518                 /* Initialize new block */
1519                 memset(new_block.data, 0, block_size);
1520                 ext4_trans_set_block_dirty(new_block.buf);
1521
1522                 /* Put back the allocated block */
1523                 rc = ext4_block_set(fs->bdev, &new_block);
1524                 if (rc != EOK)
1525                         return rc;
1526
1527                 current_block = new_blk;
1528         }
1529
1530         /*
1531          * Navigate through other levels, until we find the block number
1532          * or find null reference meaning we are dealing with sparse file
1533          */
1534         while (l > 0) {
1535                 int rc = ext4_trans_block_get(fs->bdev, &block, current_block);
1536                 if (rc != EOK)
1537                         return rc;
1538
1539                 current_block = to_le32(((uint32_t *)block.data)[off_in_blk]);
1540                 if ((l > 1) && (current_block == 0)) {
1541                         ext4_fsblk_t goal;
1542                         rc = ext4_fs_indirect_find_goal(inode_ref, &goal);
1543                         if (rc != EOK) {
1544                                 ext4_block_set(fs->bdev, &block);
1545                                 return rc;
1546                         }
1547
1548                         /* Allocate new block */
1549                         rc =
1550                             ext4_balloc_alloc_block(inode_ref, goal, &new_blk);
1551                         if (rc != EOK) {
1552                                 ext4_block_set(fs->bdev, &block);
1553                                 return rc;
1554                         }
1555
1556                         /* Load newly allocated block */
1557                         rc = ext4_trans_block_get_noread(fs->bdev, &new_block,
1558                                             new_blk);
1559
1560                         if (rc != EOK) {
1561                                 ext4_block_set(fs->bdev, &block);
1562                                 return rc;
1563                         }
1564
1565                         /* Initialize allocated block */
1566                         memset(new_block.data, 0, block_size);
1567                         ext4_trans_set_block_dirty(new_block.buf);
1568
1569                         rc = ext4_block_set(fs->bdev, &new_block);
1570                         if (rc != EOK) {
1571                                 ext4_block_set(fs->bdev, &block);
1572                                 return rc;
1573                         }
1574
1575                         /* Write block address to the parent */
1576                         uint32_t * p = (uint32_t * )block.data;
1577                         p[off_in_blk] = to_le32((uint32_t)new_blk);
1578                         ext4_trans_set_block_dirty(block.buf);
1579                         current_block = new_blk;
1580                 }
1581
1582                 /* Will be finished, write the fblock address */
1583                 if (l == 1) {
1584                         uint32_t * p = (uint32_t * )block.data;
1585                         p[off_in_blk] = to_le32((uint32_t)fblock);
1586                         ext4_trans_set_block_dirty(block.buf);
1587                 }
1588
1589                 rc = ext4_block_set(fs->bdev, &block);
1590                 if (rc != EOK)
1591                         return rc;
1592
1593                 l--;
1594
1595                 /*
1596                  * If we are on the last level, break here as
1597                  * there is no next level to visit
1598                  */
1599                 if (l == 0)
1600                         break;
1601
1602                 /* Visit the next level */
1603                 blk_off_in_lvl %= fs->inode_blocks_per_level[l];
1604                 off_in_blk = blk_off_in_lvl / fs->inode_blocks_per_level[l - 1];
1605         }
1606
1607         return EOK;
1608 }
1609
1610
1611 int ext4_fs_append_inode_dblk(struct ext4_inode_ref *inode_ref,
1612                               ext4_fsblk_t *fblock, uint32_t *iblock)
1613 {
1614 #if CONFIG_EXTENT_ENABLE
1615         /* Handle extents separately */
1616         if ((ext4_sb_feature_incom(&inode_ref->fs->sb, EXT4_FINCOM_EXTENTS)) &&
1617             (ext4_inode_has_flag(inode_ref->inode, EXT4_INODE_FLAG_EXTENTS))) {
1618                 int rc;
1619                 ext4_fsblk_t current_fsblk;
1620                 struct ext4_sblock *sb = &inode_ref->fs->sb;
1621                 uint64_t inode_size = ext4_inode_get_size(sb, inode_ref->inode);
1622                 uint32_t block_size = ext4_sb_get_block_size(sb);
1623                 *iblock = (uint32_t)((inode_size + block_size - 1) / block_size);
1624
1625                 rc = ext4_extent_get_blocks(inode_ref, *iblock, 1,
1626                                                 &current_fsblk, true, NULL);
1627
1628                 *fblock = current_fsblk;
1629                 ext4_assert(*fblock);
1630
1631                 ext4_inode_set_size(inode_ref->inode, inode_size + block_size);
1632                 inode_ref->dirty = true;
1633
1634
1635                 return rc;
1636         }
1637 #endif
1638         struct ext4_sblock *sb = &inode_ref->fs->sb;
1639
1640         /* Compute next block index and allocate data block */
1641         uint64_t inode_size = ext4_inode_get_size(sb, inode_ref->inode);
1642         uint32_t block_size = ext4_sb_get_block_size(sb);
1643
1644         /* Align size i-node size */
1645         if ((inode_size % block_size) != 0)
1646                 inode_size += block_size - (inode_size % block_size);
1647
1648         /* Logical blocks are numbered from 0 */
1649         uint32_t new_block_idx = (uint32_t)(inode_size / block_size);
1650
1651         /* Allocate new physical block */
1652         ext4_fsblk_t goal, phys_block;
1653         int rc = ext4_fs_indirect_find_goal(inode_ref, &goal);
1654         if (rc != EOK)
1655                 return rc;
1656
1657         rc = ext4_balloc_alloc_block(inode_ref, goal, &phys_block);
1658         if (rc != EOK)
1659                 return rc;
1660
1661         /* Add physical block address to the i-node */
1662         rc = ext4_fs_set_inode_data_block_index(inode_ref, new_block_idx,
1663                                                 phys_block);
1664         if (rc != EOK) {
1665                 ext4_balloc_free_block(inode_ref, phys_block);
1666                 return rc;
1667         }
1668
1669         /* Update i-node */
1670         ext4_inode_set_size(inode_ref->inode, inode_size + block_size);
1671         inode_ref->dirty = true;
1672
1673         *fblock = phys_block;
1674         *iblock = new_block_idx;
1675
1676         return EOK;
1677 }
1678
1679 void ext4_fs_inode_links_count_inc(struct ext4_inode_ref *inode_ref)
1680 {
1681         uint16_t link;
1682         bool is_dx;
1683         link = ext4_inode_get_links_cnt(inode_ref->inode);
1684         link++;
1685         ext4_inode_set_links_cnt(inode_ref->inode, link);
1686
1687         is_dx = ext4_sb_feature_com(&inode_ref->fs->sb, EXT4_FCOM_DIR_INDEX) &&
1688                 ext4_inode_has_flag(inode_ref->inode, EXT4_INODE_FLAG_INDEX);
1689
1690         if (is_dx && link > 1) {
1691                 if (link >= EXT4_LINK_MAX || link == 2) {
1692                         ext4_inode_set_links_cnt(inode_ref->inode, 1);
1693
1694                         uint32_t v;
1695                         v = ext4_get32(&inode_ref->fs->sb, features_read_only);
1696                         v |= EXT4_FRO_COM_DIR_NLINK;
1697                         ext4_set32(&inode_ref->fs->sb, features_read_only, v);
1698                 }
1699         }
1700 }
1701
1702 void ext4_fs_inode_links_count_dec(struct ext4_inode_ref *inode_ref)
1703 {
1704         uint16_t links = ext4_inode_get_links_cnt(inode_ref->inode);
1705         if (!ext4_inode_is_type(&inode_ref->fs->sb, inode_ref->inode,
1706                                 EXT4_INODE_MODE_DIRECTORY)) {
1707                 if (links > 0)
1708                         ext4_inode_set_links_cnt(inode_ref->inode, links - 1);
1709                 return;
1710         }
1711
1712         if (links > 2)
1713                 ext4_inode_set_links_cnt(inode_ref->inode, links - 1);
1714 }
1715
1716 /**
1717  * @}
1718  */