lwext4/ext4_journal.c

   1 /*
   2  * Copyright (c) 2015 Grzegorz Kostka (kostka.grzegorz@gmail.com)
   3  * Copyright (c) 2015 Kaho Ng (ngkaho1234@gmail.com)
   4  * All rights reserved.
   5  *
   6  * Redistribution and use in source and binary forms, with or without
   7  * modification, are permitted provided that the following conditions
   8  * are met:
   9  *
  10  * - Redistributions of source code must retain the above copyright
  11  *   notice, this list of conditions and the following disclaimer.
  12  * - Redistributions in binary form must reproduce the above copyright
  13  *   notice, this list of conditions and the following disclaimer in the
  14  *   documentation and/or other materials provided with the distribution.
  15  * - The name of the author may not be used to endorse or promote products
  16  *   derived from this software without specific prior written permission.
  17  *
  18  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
  19  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
  20  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
  21  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
  22  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
  23  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  24  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  25  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  26  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
  27  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  28  */
  29
  30 /** @addtogroup lwext4
  31  * @{
  32  */
  33 /**
  34  * @file  ext4_journal.c
  35  * @brief Journal handle functions
  36  */
  37
  38 #include "ext4_config.h"
  39 #include "ext4_types.h"
  40 #include "ext4_fs.h"
  41 #include "ext4_super.h"
  42 #include "ext4_journal.h"
  43 #include "ext4_errno.h"
  44 #include "ext4_blockdev.h"
  45 #include "ext4_crc32c.h"
  46 #include "ext4_debug.h"
  47 #include "tree.h"
  48
  49 #include <string.h>
  50 #include <stdlib.h>
  51
  52 /**@brief  Revoke entry during journal replay.*/
  53 struct revoke_entry {
  54         /**@brief  Block number not to be replayed.*/
  55         ext4_fsblk_t block;
  56
  57         /**@brief  For any transaction id smaller
  58          *         than trans_id, records of @block
  59          *         in those transactions should not
  60          *         be replayed.*/
  61         uint32_t trans_id;
  62
  63         /**@brief  Revoke tree node.*/
  64         RB_ENTRY(revoke_entry) revoke_node;
  65 };
  66
  67 /**@brief  Valid journal replay information.*/
  68 struct recover_info {
  69         /**@brief  Starting transaction id.*/
  70         uint32_t start_trans_id;
  71
  72         /**@brief  Ending transaction id.*/
  73         uint32_t last_trans_id;
  74
  75         /**@brief  Used as internal argument.*/
  76         uint32_t this_trans_id;
  77
  78         /**@brief  RB-Tree storing revoke entries.*/
  79         RB_HEAD(jbd_revoke, revoke_entry) revoke_root;
  80 };
  81
  82 /**@brief  Journal replay internal arguments.*/
  83 struct replay_arg {
  84         /**@brief  Journal replay information.*/
  85         struct recover_info *info;
  86
  87         /**@brief  Current block we are on.*/
  88         uint32_t *this_block;
  89
  90         /**@brief  Current trans_id we are on.*/
  91         uint32_t this_trans_id;
  92 };
  93
  94 static int
  95 jbd_revoke_entry_cmp(struct revoke_entry *a, struct revoke_entry *b)
  96 {
  97         if (a->block > b->block)
  98                 return 1;
  99         else if (a->block < b->block)
 100                 return -1;
 101         return 0;
 102 }
 103
 104 RB_GENERATE_INTERNAL(jbd_revoke, revoke_entry, revoke_node,
 105                      jbd_revoke_entry_cmp, static inline)
 106
 107 #define jbd_alloc_revoke_entry() calloc(1, sizeof(struct revoke_entry))
 108 #define jbd_free_revoke_entry(addr) free(addr)
 109
 110 /**@brief  Write jbd superblock to disk.
 111  * @param  jbd_fs jbd filesystem
 112  * @param  s jbd superblock
 113  * @return standard error code*/
 114 static int jbd_sb_write(struct jbd_fs *jbd_fs, struct jbd_sb *s)
 115 {
 116         int rc;
 117         struct ext4_fs *fs = jbd_fs->inode_ref.fs;
 118         uint64_t offset;
 119         ext4_fsblk_t fblock;
 120         rc = jbd_inode_bmap(jbd_fs, 0, &fblock);
 121         if (rc != EOK)
 122                 return rc;
 123
 124         offset = fblock * ext4_sb_get_block_size(&fs->sb);
 125         return ext4_block_writebytes(fs->bdev, offset, s,
 126                                      EXT4_SUPERBLOCK_SIZE);
 127 }
 128
 129 /**@brief  Read jbd superblock from disk.
 130  * @param  jbd_fs jbd filesystem
 131  * @param  s jbd superblock
 132  * @return standard error code*/
 133 static int jbd_sb_read(struct jbd_fs *jbd_fs, struct jbd_sb *s)
 134 {
 135         int rc;
 136         struct ext4_fs *fs = jbd_fs->inode_ref.fs;
 137         uint64_t offset;
 138         ext4_fsblk_t fblock;
 139         rc = jbd_inode_bmap(jbd_fs, 0, &fblock);
 140         if (rc != EOK)
 141                 return rc;
 142
 143         offset = fblock * ext4_sb_get_block_size(&fs->sb);
 144         return ext4_block_readbytes(fs->bdev, offset, s,
 145                                     EXT4_SUPERBLOCK_SIZE);
 146 }
 147
 148 /**@brief  Verify jbd superblock.
 149  * @param  sb jbd superblock
 150  * @return true if jbd superblock is valid */
 151 static bool jbd_verify_sb(struct jbd_sb *sb)
 152 {
 153         struct jbd_bhdr *header = &sb->header;
 154         if (jbd_get32(header, magic) != JBD_MAGIC_NUMBER)
 155                 return false;
 156
 157         if (jbd_get32(header, blocktype) != JBD_SUPERBLOCK &&
 158             jbd_get32(header, blocktype) != JBD_SUPERBLOCK_V2)
 159                 return false;
 160
 161         return true;
 162 }
 163
 164 /**@brief  Write back dirty jbd superblock to disk.
 165  * @param  jbd_fs jbd filesystem
 166  * @return standard error code*/
 167 static int jbd_write_sb(struct jbd_fs *jbd_fs)
 168 {
 169         int rc = EOK;
 170         if (jbd_fs->dirty) {
 171                 rc = jbd_sb_write(jbd_fs, &jbd_fs->sb);
 172                 if (rc != EOK)
 173                         return rc;
 174
 175                 jbd_fs->dirty = false;
 176         }
 177         return rc;
 178 }
 179
 180 /**@brief  Get reference to jbd filesystem.
 181  * @param  fs Filesystem to load journal of
 182  * @param  jbd_fs jbd filesystem
 183  * @return standard error code*/
 184 int jbd_get_fs(struct ext4_fs *fs,
 185                struct jbd_fs *jbd_fs)
 186 {
 187         int rc;
 188         uint32_t journal_ino;
 189
 190         memset(jbd_fs, 0, sizeof(struct jbd_fs));
 191         /* See if there is journal inode on this filesystem.*/
 192         /* FIXME: detection on existance ofbkejournal bdev is
 193          *        missing.*/
 194         journal_ino = ext4_get32(&fs->sb, journal_inode_number);
 195
 196         rc = ext4_fs_get_inode_ref(fs,
 197                                    journal_ino,
 198                                    &jbd_fs->inode_ref);
 199         if (rc != EOK) {
 200                 memset(jbd_fs, 0, sizeof(struct jbd_fs));
 201                 return rc;
 202         }
 203         rc = jbd_sb_read(jbd_fs, &jbd_fs->sb);
 204         if (rc != EOK) {
 205                 memset(jbd_fs, 0, sizeof(struct jbd_fs));
 206                 ext4_fs_put_inode_ref(&jbd_fs->inode_ref);
 207                 return rc;
 208         }
 209         if (!jbd_verify_sb(&jbd_fs->sb)) {
 210                 memset(jbd_fs, 0, sizeof(struct jbd_fs));
 211                 ext4_fs_put_inode_ref(&jbd_fs->inode_ref);
 212                 rc = EIO;
 213         }
 214
 215         return rc;
 216 }
 217
 218 /**@brief  Put reference of jbd filesystem.
 219  * @param  jbd_fs jbd filesystem
 220  * @return standard error code*/
 221 int jbd_put_fs(struct jbd_fs *jbd_fs)
 222 {
 223         int rc = EOK;
 224         rc = jbd_write_sb(jbd_fs);
 225
 226         ext4_fs_put_inode_ref(&jbd_fs->inode_ref);
 227         return rc;
 228 }
 229
 230 /**@brief  Data block lookup helper.
 231  * @param  jbd_fs jbd filesystem
 232  * @param  iblock block index
 233  * @param  fblock logical block address
 234  * @return standard error code*/
 235 int jbd_inode_bmap(struct jbd_fs *jbd_fs,
 236                    ext4_lblk_t iblock,
 237                    ext4_fsblk_t *fblock)
 238 {
 239         int rc = ext4_fs_get_inode_dblk_idx(
 240                         &jbd_fs->inode_ref,
 241                         iblock,
 242                         fblock,
 243                         false);
 244         return rc;
 245 }
 246
 247 /**@brief   jbd block get function (through cache).
 248  * @param   jbd_fs jbd filesystem
 249  * @param   block block descriptor
 250  * @param   fblock jbd logical block address
 251  * @return  standard error code*/
 252 static int jbd_block_get(struct jbd_fs *jbd_fs,
 253                   struct ext4_block *block,
 254                   ext4_fsblk_t fblock)
 255 {
 256         /* TODO: journal device. */
 257         int rc;
 258         ext4_lblk_t iblock = (ext4_lblk_t)fblock;
 259
 260         /* Lookup the logical block address of
 261          * fblock.*/
 262         rc = jbd_inode_bmap(jbd_fs, iblock,
 263                             &fblock);
 264         if (rc != EOK)
 265                 return rc;
 266
 267         struct ext4_blockdev *bdev = jbd_fs->inode_ref.fs->bdev;
 268         rc = ext4_block_get(bdev, block, fblock);
 269
 270         /* If succeeded, mark buffer as BC_FLUSH to indicate
 271          * that data should be written to disk immediately.*/
 272         if (rc == EOK)
 273                 ext4_bcache_set_flag(block->buf, BC_FLUSH);
 274
 275         return rc;
 276 }
 277
 278 /**@brief   jbd block get function (through cache, don't read).
 279  * @param   jbd_fs jbd filesystem
 280  * @param   block block descriptor
 281  * @param   fblock jbd logical block address
 282  * @return  standard error code*/
 283 static int jbd_block_get_noread(struct jbd_fs *jbd_fs,
 284                          struct ext4_block *block,
 285                          ext4_fsblk_t fblock)
 286 {
 287         /* TODO: journal device. */
 288         int rc;
 289         ext4_lblk_t iblock = (ext4_lblk_t)fblock;
 290         rc = jbd_inode_bmap(jbd_fs, iblock,
 291                             &fblock);
 292         if (rc != EOK)
 293                 return rc;
 294
 295         struct ext4_blockdev *bdev = jbd_fs->inode_ref.fs->bdev;
 296         rc = ext4_block_get_noread(bdev, block, fblock);
 297         if (rc == EOK)
 298                 ext4_bcache_set_flag(block->buf, BC_FLUSH);
 299
 300         return rc;
 301 }
 302
 303 /**@brief   jbd block set procedure (through cache).
 304  * @param   jbd_fs jbd filesystem
 305  * @param   block block descriptor
 306  * @return  standard error code*/
 307 static int jbd_block_set(struct jbd_fs *jbd_fs,
 308                   struct ext4_block *block)
 309 {
 310         return ext4_block_set(jbd_fs->inode_ref.fs->bdev,
 311                               block);
 312 }
 313
 314 /**@brief  helper functions to calculate
 315  *         block tag size, not including UUID part.
 316  * @param  jbd_fs jbd filesystem
 317  * @return tag size in bytes*/
 318 static int jbd_tag_bytes(struct jbd_fs *jbd_fs)
 319 {
 320         int size;
 321
 322         /* It is very easy to deal with the case which
 323          * JBD_FEATURE_INCOMPAT_CSUM_V3 is enabled.*/
 324         if (JBD_HAS_INCOMPAT_FEATURE(&jbd_fs->sb,
 325                                      JBD_FEATURE_INCOMPAT_CSUM_V3))
 326                 return sizeof(struct jbd_block_tag3);
 327
 328         size = sizeof(struct jbd_block_tag);
 329
 330         /* If JBD_FEATURE_INCOMPAT_CSUM_V2 is enabled,
 331          * add 2 bytes to size.*/
 332         if (JBD_HAS_INCOMPAT_FEATURE(&jbd_fs->sb,
 333                                      JBD_FEATURE_INCOMPAT_CSUM_V2))
 334                 size += sizeof(uint16_t);
 335
 336         if (JBD_HAS_INCOMPAT_FEATURE(&jbd_fs->sb,
 337                                      JBD_FEATURE_INCOMPAT_64BIT))
 338                 return size;
 339
 340         /* If block number is 4 bytes in size,
 341          * minus 4 bytes from size */
 342         return size - sizeof(uint32_t);
 343 }
 344
 345 /**@brief  Tag information. */
 346 struct tag_info {
 347         /**@brief  Tag size in bytes, including UUID part.*/
 348         int tag_bytes;
 349
 350         /**@brief  block number stored in this tag.*/
 351         ext4_fsblk_t block;
 352
 353         /**@brief  whether UUID part exists or not.*/
 354         bool uuid_exist;
 355
 356         /**@brief  UUID content if UUID part exists.*/
 357         uint8_t uuid[UUID_SIZE];
 358
 359         /**@brief  Is this the last tag? */
 360         bool last_tag;
 361 };
 362
 363 /**@brief  Extract information from a block tag.
 364  * @param  __tag pointer to the block tag
 365  * @param  tag_bytes block tag size of this jbd filesystem
 366  * @param  remaining size in buffer containing the block tag
 367  * @param  tag_info information of this tag.
 368  * @return  EOK when succeed, otherwise return EINVAL.*/
 369 static int
 370 jbd_extract_block_tag(struct jbd_fs *jbd_fs,
 371                       void *__tag,
 372                       int tag_bytes,
 373                       int32_t remain_buf_size,
 374                       struct tag_info *tag_info)
 375 {
 376         char *uuid_start;
 377         tag_info->tag_bytes = tag_bytes;
 378         tag_info->uuid_exist = false;
 379         tag_info->last_tag = false;
 380
 381         /* See whether it is possible to hold a valid block tag.*/
 382         if (remain_buf_size - tag_bytes < 0)
 383                 return EINVAL;
 384
 385         if (JBD_HAS_INCOMPAT_FEATURE(&jbd_fs->sb,
 386                                      JBD_FEATURE_INCOMPAT_CSUM_V3)) {
 387                 struct jbd_block_tag3 *tag = __tag;
 388                 tag_info->block = jbd_get32(tag, blocknr);
 389                 if (JBD_HAS_INCOMPAT_FEATURE(&jbd_fs->sb,
 390                                              JBD_FEATURE_INCOMPAT_64BIT))
 391                          tag_info->block |=
 392                                  (uint64_t)jbd_get32(tag, blocknr_high) << 32;
 393
 394                 if (jbd_get32(tag, flags) & JBD_FLAG_ESCAPE)
 395                         tag_info->block = 0;
 396
 397                 if (!(jbd_get32(tag, flags) & JBD_FLAG_SAME_UUID)) {
 398                         /* See whether it is possible to hold UUID part.*/
 399                         if (remain_buf_size - tag_bytes < UUID_SIZE)
 400                                 return EINVAL;
 401
 402                         uuid_start = (char *)tag + tag_bytes;
 403                         tag_info->uuid_exist = true;
 404                         tag_info->tag_bytes += UUID_SIZE;
 405                         memcpy(tag_info->uuid, uuid_start, UUID_SIZE);
 406                 }
 407
 408                 if (jbd_get32(tag, flags) & JBD_FLAG_LAST_TAG)
 409                         tag_info->last_tag = true;
 410
 411         } else {
 412                 struct jbd_block_tag *tag = __tag;
 413                 tag_info->block = jbd_get32(tag, blocknr);
 414                 if (JBD_HAS_INCOMPAT_FEATURE(&jbd_fs->sb,
 415                                              JBD_FEATURE_INCOMPAT_64BIT))
 416                          tag_info->block |=
 417                                  (uint64_t)jbd_get32(tag, blocknr_high) << 32;
 418
 419                 if (jbd_get16(tag, flags) & JBD_FLAG_ESCAPE)
 420                         tag_info->block = 0;
 421
 422                 if (!(jbd_get16(tag, flags) & JBD_FLAG_SAME_UUID)) {
 423                         /* See whether it is possible to hold UUID part.*/
 424                         if (remain_buf_size - tag_bytes < UUID_SIZE)
 425                                 return EINVAL;
 426
 427                         uuid_start = (char *)tag + tag_bytes;
 428                         tag_info->uuid_exist = true;
 429                         tag_info->tag_bytes += UUID_SIZE;
 430                         memcpy(tag_info->uuid, uuid_start, UUID_SIZE);
 431                 }
 432
 433                 if (jbd_get16(tag, flags) & JBD_FLAG_LAST_TAG)
 434                         tag_info->last_tag = true;
 435
 436         }
 437         return EOK;
 438 }
 439
 440 /**@brief  Write information to a block tag.
 441  * @param  __tag pointer to the block tag
 442  * @param  remaining size in buffer containing the block tag
 443  * @param  tag_info information of this tag.
 444  * @return  EOK when succeed, otherwise return EINVAL.*/
 445 static int
 446 jbd_write_block_tag(struct jbd_fs *jbd_fs,
 447                     void *__tag,
 448                     int32_t remain_buf_size,
 449                     struct tag_info *tag_info)
 450 {
 451         char *uuid_start;
 452         int tag_bytes = jbd_tag_bytes(jbd_fs);
 453
 454         tag_info->tag_bytes = tag_bytes;
 455
 456         /* See whether it is possible to hold a valid block tag.*/
 457         if (remain_buf_size - tag_bytes < 0)
 458                 return EINVAL;
 459
 460         if (JBD_HAS_INCOMPAT_FEATURE(&jbd_fs->sb,
 461                                      JBD_FEATURE_INCOMPAT_CSUM_V3)) {
 462                 struct jbd_block_tag3 *tag = __tag;
 463                 jbd_set32(tag, blocknr, tag_info->block);
 464                 if (JBD_HAS_INCOMPAT_FEATURE(&jbd_fs->sb,
 465                                              JBD_FEATURE_INCOMPAT_64BIT))
 466                         jbd_set32(tag, blocknr_high, tag_info->block >> 32);
 467
 468                 if (tag_info->uuid_exist) {
 469                         /* See whether it is possible to hold UUID part.*/
 470                         if (remain_buf_size - tag_bytes < UUID_SIZE)
 471                                 return EINVAL;
 472
 473                         uuid_start = (char *)tag + tag_bytes;
 474                         tag_info->tag_bytes += UUID_SIZE;
 475                         memcpy(uuid_start, tag_info->uuid, UUID_SIZE);
 476                 } else
 477                         jbd_set32(tag, flags,
 478                                   jbd_get32(tag, flags) | JBD_FLAG_SAME_UUID);
 479
 480                 if (tag_info->last_tag)
 481                         jbd_set32(tag, flags,
 482                                   jbd_get32(tag, flags) | JBD_FLAG_LAST_TAG);
 483
 484         } else {
 485                 struct jbd_block_tag *tag = __tag;
 486                 jbd_set32(tag, blocknr, tag_info->block);
 487                 if (JBD_HAS_INCOMPAT_FEATURE(&jbd_fs->sb,
 488                                              JBD_FEATURE_INCOMPAT_64BIT))
 489                         jbd_set32(tag, blocknr_high, tag_info->block >> 32);
 490
 491                 if (tag_info->uuid_exist) {
 492                         /* See whether it is possible to hold UUID part.*/
 493                         if (remain_buf_size - tag_bytes < UUID_SIZE)
 494                                 return EINVAL;
 495
 496                         uuid_start = (char *)tag + tag_bytes;
 497                         tag_info->tag_bytes += UUID_SIZE;
 498                         memcpy(uuid_start, tag_info->uuid, UUID_SIZE);
 499                 } else
 500                         jbd_set16(tag, flags,
 501                                   jbd_get16(tag, flags) | JBD_FLAG_SAME_UUID);
 502
 503                 if (tag_info->last_tag)
 504                         jbd_set16(tag, flags,
 505                                   jbd_get16(tag, flags) | JBD_FLAG_LAST_TAG);
 506
 507         }
 508         return EOK;
 509 }
 510
 511 /**@brief  Iterate all block tags in a block.
 512  * @param  jbd_fs jbd filesystem
 513  * @param  __tag_start pointer to the block
 514  * @param  tag_tbl_size size of the block
 515  * @param  func callback routine to indicate that
 516  *         a block tag is found
 517  * @param  arg additional argument to be passed to func */
 518 static void
 519 jbd_iterate_block_table(struct jbd_fs *jbd_fs,
 520                         void *__tag_start,
 521                         int32_t tag_tbl_size,
 522                         void (*func)(struct jbd_fs * jbd_fs,
 523                                         ext4_fsblk_t block,
 524                                         uint8_t *uuid,
 525                                         void *arg),
 526                         void *arg)
 527 {
 528         char *tag_start, *tag_ptr;
 529         int tag_bytes = jbd_tag_bytes(jbd_fs);
 530         tag_start = __tag_start;
 531         tag_ptr = tag_start;
 532
 533         /* Cut off the size of block tail storing checksum. */
 534         if (JBD_HAS_INCOMPAT_FEATURE(&jbd_fs->sb,
 535                                      JBD_FEATURE_INCOMPAT_CSUM_V2) ||
 536             JBD_HAS_INCOMPAT_FEATURE(&jbd_fs->sb,
 537                                      JBD_FEATURE_INCOMPAT_CSUM_V3))
 538                 tag_tbl_size -= sizeof(struct jbd_block_tail);
 539
 540         while (tag_tbl_size) {
 541                 struct tag_info tag_info;
 542                 int rc = jbd_extract_block_tag(jbd_fs,
 543                                       tag_ptr,
 544                                       tag_bytes,
 545                                       tag_tbl_size,
 546                                       &tag_info);
 547                 if (rc != EOK)
 548                         break;
 549
 550                 if (func)
 551                         func(jbd_fs, tag_info.block, tag_info.uuid, arg);
 552
 553                 /* Stop the iteration when we reach the last tag. */
 554                 if (tag_info.last_tag)
 555                         break;
 556
 557                 tag_ptr += tag_info.tag_bytes;
 558                 tag_tbl_size -= tag_info.tag_bytes;
 559         }
 560 }
 561
 562 static void jbd_display_block_tags(struct jbd_fs *jbd_fs,
 563                                    ext4_fsblk_t block,
 564                                    uint8_t *uuid,
 565                                    void *arg)
 566 {
 567         uint32_t *iblock = arg;
 568         ext4_dbg(DEBUG_JBD, "Block in block_tag: %" PRIu64 "\n", block);
 569         (*iblock)++;
 570         (void)jbd_fs;
 571         (void)uuid;
 572         return;
 573 }
 574
 575 static struct revoke_entry *
 576 jbd_revoke_entry_lookup(struct recover_info *info, ext4_fsblk_t block)
 577 {
 578         struct revoke_entry tmp = {
 579                 .block = block
 580         };
 581
 582         return RB_FIND(jbd_revoke, &info->revoke_root, &tmp);
 583 }
 584
 585 /**@brief  Replay a block in a transaction.
 586  * @param  jbd_fs jbd filesystem
 587  * @param  block  block address to be replayed.*/
 588 static void jbd_replay_block_tags(struct jbd_fs *jbd_fs,
 589                                   ext4_fsblk_t block,
 590                                   uint8_t *uuid __unused,
 591                                   void *__arg)
 592 {
 593         int r;
 594         struct replay_arg *arg = __arg;
 595         struct recover_info *info = arg->info;
 596         uint32_t *this_block = arg->this_block;
 597         struct revoke_entry *revoke_entry;
 598         struct ext4_block journal_block, ext4_block;
 599         struct ext4_fs *fs = jbd_fs->inode_ref.fs;
 600
 601         (*this_block)++;
 602
 603         /* We replay this block only if the current transaction id
 604          * is equal or greater than that in revoke entry.*/
 605         revoke_entry = jbd_revoke_entry_lookup(info, block);
 606         if (revoke_entry &&
 607             arg->this_trans_id < revoke_entry->trans_id)
 608                 return;
 609
 610         ext4_dbg(DEBUG_JBD,
 611                  "Replaying block in block_tag: %" PRIu64 "\n",
 612                  block);
 613
 614         r = jbd_block_get(jbd_fs, &journal_block, *this_block);
 615         if (r != EOK)
 616                 return;
 617
 618         /* We need special treatment for ext4 superblock. */
 619         if (block) {
 620                 r = ext4_block_get_noread(fs->bdev, &ext4_block, block);
 621                 if (r != EOK) {
 622                         jbd_block_set(jbd_fs, &journal_block);
 623                         return;
 624                 }
 625
 626                 memcpy(ext4_block.data,
 627                         journal_block.data,
 628                         jbd_get32(&jbd_fs->sb, blocksize));
 629
 630                 ext4_bcache_set_dirty(ext4_block.buf);
 631                 ext4_block_set(fs->bdev, &ext4_block);
 632         } else {
 633                 uint16_t mount_count, state;
 634                 mount_count = ext4_get16(&fs->sb, mount_count);
 635                 state = ext4_get16(&fs->sb, state);
 636
 637                 memcpy(&fs->sb,
 638                         journal_block.data + EXT4_SUPERBLOCK_OFFSET,
 639                         EXT4_SUPERBLOCK_SIZE);
 640
 641                 /* Mark system as mounted */
 642                 ext4_set16(&fs->sb, state, state);
 643                 r = ext4_sb_write(fs->bdev, &fs->sb);
 644                 if (r != EOK)
 645                         return;
 646
 647                 /*Update mount count*/
 648                 ext4_set16(&fs->sb, mount_count, mount_count);
 649         }
 650
 651         jbd_block_set(jbd_fs, &journal_block);
 652
 653         return;
 654 }
 655
 656 /**@brief  Add block address to revoke tree, along with
 657  *         its transaction id.
 658  * @param  info  journal replay info
 659  * @param  block  block address to be replayed.*/
 660 static void jbd_add_revoke_block_tags(struct recover_info *info,
 661                                       ext4_fsblk_t block)
 662 {
 663         struct revoke_entry *revoke_entry;
 664
 665         ext4_dbg(DEBUG_JBD, "Add block %" PRIu64 " to revoke tree\n", block);
 666         /* If the revoke entry with respect to the block address
 667          * exists already, update its transaction id.*/
 668         revoke_entry = jbd_revoke_entry_lookup(info, block);
 669         if (revoke_entry) {
 670                 revoke_entry->trans_id = info->this_trans_id;
 671                 return;
 672         }
 673
 674         revoke_entry = jbd_alloc_revoke_entry();
 675         ext4_assert(revoke_entry);
 676         revoke_entry->block = block;
 677         revoke_entry->trans_id = info->this_trans_id;
 678         RB_INSERT(jbd_revoke, &info->revoke_root, revoke_entry);
 679
 680         return;
 681 }
 682
 683 static void jbd_destroy_revoke_tree(struct recover_info *info)
 684 {
 685         while (!RB_EMPTY(&info->revoke_root)) {
 686                 struct revoke_entry *revoke_entry =
 687                         RB_MIN(jbd_revoke, &info->revoke_root);
 688                 ext4_assert(revoke_entry);
 689                 RB_REMOVE(jbd_revoke, &info->revoke_root, revoke_entry);
 690                 jbd_free_revoke_entry(revoke_entry);
 691         }
 692 }
 693
 694 /* Make sure we wrap around the log correctly! */
 695 #define wrap(sb, var)                                           \
 696 do {                                                                    \
 697         if (var >= jbd_get32((sb), maxlen))                                     \
 698                 var -= (jbd_get32((sb), maxlen) - jbd_get32((sb), first));      \
 699 } while (0)
 700
 701 #define ACTION_SCAN 0
 702 #define ACTION_REVOKE 1
 703 #define ACTION_RECOVER 2
 704
 705 /**@brief  Add entries in a revoke block to revoke tree.
 706  * @param  jbd_fs jbd filesystem
 707  * @param  header revoke block header
 708  * @param  recover_info  journal replay info*/
 709 static void jbd_build_revoke_tree(struct jbd_fs *jbd_fs,
 710                                   struct jbd_bhdr *header,
 711                                   struct recover_info *info)
 712 {
 713         char *blocks_entry;
 714         struct jbd_revoke_header *revoke_hdr =
 715                 (struct jbd_revoke_header *)header;
 716         uint32_t i, nr_entries, record_len = 4;
 717
 718         /* If we are working on a 64bit jbd filesystem, */
 719         if (JBD_HAS_INCOMPAT_FEATURE(&jbd_fs->sb,
 720                                      JBD_FEATURE_INCOMPAT_64BIT))
 721                 record_len = 8;
 722
 723         nr_entries = (jbd_get32(revoke_hdr, count) -
 724                         sizeof(struct jbd_revoke_header)) /
 725                         record_len;
 726
 727         blocks_entry = (char *)(revoke_hdr + 1);
 728
 729         for (i = 0;i < nr_entries;i++) {
 730                 if (record_len == 8) {
 731                         uint64_t *blocks =
 732                                 (uint64_t *)blocks_entry;
 733                         jbd_add_revoke_block_tags(info, to_be64(*blocks));
 734                 } else {
 735                         uint32_t *blocks =
 736                                 (uint32_t *)blocks_entry;
 737                         jbd_add_revoke_block_tags(info, to_be32(*blocks));
 738                 }
 739                 blocks_entry += record_len;
 740         }
 741 }
 742
 743 static void jbd_debug_descriptor_block(struct jbd_fs *jbd_fs,
 744                                        struct jbd_bhdr *header,
 745                                        uint32_t *iblock)
 746 {
 747         jbd_iterate_block_table(jbd_fs,
 748                                 header + 1,
 749                                 jbd_get32(&jbd_fs->sb, blocksize) -
 750                                         sizeof(struct jbd_bhdr),
 751                                 jbd_display_block_tags,
 752                                 iblock);
 753 }
 754
 755 static void jbd_replay_descriptor_block(struct jbd_fs *jbd_fs,
 756                                         struct jbd_bhdr *header,
 757                                         struct replay_arg *arg)
 758 {
 759         jbd_iterate_block_table(jbd_fs,
 760                                 header + 1,
 761                                 jbd_get32(&jbd_fs->sb, blocksize) -
 762                                         sizeof(struct jbd_bhdr),
 763                                 jbd_replay_block_tags,
 764                                 arg);
 765 }
 766
 767 /**@brief  The core routine of journal replay.
 768  * @param  jbd_fs jbd filesystem
 769  * @param  recover_info  journal replay info
 770  * @param  action action needed to be taken
 771  * @return standard error code*/
 772 static int jbd_iterate_log(struct jbd_fs *jbd_fs,
 773                            struct recover_info *info,
 774                            int action)
 775 {
 776         int r = EOK;
 777         bool log_end = false;
 778         struct jbd_sb *sb = &jbd_fs->sb;
 779         uint32_t start_trans_id, this_trans_id;
 780         uint32_t start_block, this_block;
 781
 782         /* We start iterating valid blocks in the whole journal.*/
 783         start_trans_id = this_trans_id = jbd_get32(sb, sequence);
 784         start_block = this_block = jbd_get32(sb, start);
 785
 786         ext4_dbg(DEBUG_JBD, "Start of journal at trans id: %" PRIu32 "\n",
 787                             start_trans_id);
 788
 789         while (!log_end) {
 790                 struct ext4_block block;
 791                 struct jbd_bhdr *header;
 792                 /* If we are not scanning for the last
 793                  * valid transaction in the journal,
 794                  * we will stop when we reach the end of
 795                  * the journal.*/
 796                 if (action != ACTION_SCAN)
 797                         if (this_trans_id > info->last_trans_id) {
 798                                 log_end = true;
 799                                 continue;
 800                         }
 801
 802                 r = jbd_block_get(jbd_fs, &block, this_block);
 803                 if (r != EOK)
 804                         break;
 805
 806                 header = (struct jbd_bhdr *)block.data;
 807                 /* This block does not have a valid magic number,
 808                  * so we have reached the end of the journal.*/
 809                 if (jbd_get32(header, magic) != JBD_MAGIC_NUMBER) {
 810                         jbd_block_set(jbd_fs, &block);
 811                         log_end = true;
 812                         continue;
 813                 }
 814
 815                 /* If the transaction id we found is not expected,
 816                  * we may have reached the end of the journal.
 817                  *
 818                  * If we are not scanning the journal, something
 819                  * bad might have taken place. :-( */
 820                 if (jbd_get32(header, sequence) != this_trans_id) {
 821                         if (action != ACTION_SCAN)
 822                                 r = EIO;
 823
 824                         jbd_block_set(jbd_fs, &block);
 825                         log_end = true;
 826                         continue;
 827                 }
 828
 829                 switch (jbd_get32(header, blocktype)) {
 830                 case JBD_DESCRIPTOR_BLOCK:
 831                         ext4_dbg(DEBUG_JBD, "Descriptor block: %" PRIu32", "
 832                                             "trans_id: %" PRIu32"\n",
 833                                             this_block, this_trans_id);
 834                         if (action == ACTION_RECOVER) {
 835                                 struct replay_arg replay_arg;
 836                                 replay_arg.info = info;
 837                                 replay_arg.this_block = &this_block;
 838                                 replay_arg.this_trans_id = this_trans_id;
 839
 840                                 jbd_replay_descriptor_block(jbd_fs,
 841                                                 header, &replay_arg);
 842                         } else
 843                                 jbd_debug_descriptor_block(jbd_fs,
 844                                                 header, &this_block);
 845
 846                         break;
 847                 case JBD_COMMIT_BLOCK:
 848                         ext4_dbg(DEBUG_JBD, "Commit block: %" PRIu32", "
 849                                             "trans_id: %" PRIu32"\n",
 850                                             this_block, this_trans_id);
 851                         /* This is the end of a transaction,
 852                          * we may now proceed to the next transaction.
 853                          */
 854                         this_trans_id++;
 855                         break;
 856                 case JBD_REVOKE_BLOCK:
 857                         ext4_dbg(DEBUG_JBD, "Revoke block: %" PRIu32", "
 858                                             "trans_id: %" PRIu32"\n",
 859                                             this_block, this_trans_id);
 860                         if (action == ACTION_REVOKE) {
 861                                 info->this_trans_id = this_trans_id;
 862                                 jbd_build_revoke_tree(jbd_fs,
 863                                                 header, info);
 864                         }
 865                         break;
 866                 default:
 867                         log_end = true;
 868                         break;
 869                 }
 870                 jbd_block_set(jbd_fs, &block);
 871                 this_block++;
 872                 wrap(sb, this_block);
 873                 if (this_block == start_block)
 874                         log_end = true;
 875
 876         }
 877         ext4_dbg(DEBUG_JBD, "End of journal.\n");
 878         if (r == EOK && action == ACTION_SCAN) {
 879                 /* We have finished scanning the journal. */
 880                 info->start_trans_id = start_trans_id;
 881                 if (this_trans_id > start_trans_id)
 882                         info->last_trans_id = this_trans_id - 1;
 883                 else
 884                         info->last_trans_id = this_trans_id;
 885         }
 886
 887         return r;
 888 }
 889
 890 /**@brief  Replay journal.
 891  * @param  jbd_fs jbd filesystem
 892  * @return standard error code*/
 893 int jbd_recover(struct jbd_fs *jbd_fs)
 894 {
 895         int r;
 896         struct recover_info info;
 897         struct jbd_sb *sb = &jbd_fs->sb;
 898         if (!sb->start)
 899                 return EOK;
 900
 901         RB_INIT(&info.revoke_root);
 902
 903         r = jbd_iterate_log(jbd_fs, &info, ACTION_SCAN);
 904         if (r != EOK)
 905                 return r;
 906
 907         r = jbd_iterate_log(jbd_fs, &info, ACTION_REVOKE);
 908         if (r != EOK)
 909                 return r;
 910
 911         r = jbd_iterate_log(jbd_fs, &info, ACTION_RECOVER);
 912         if (r == EOK) {
 913                 /* If we successfully replay the journal,
 914                  * clear EXT4_FINCOM_RECOVER flag on the
 915                  * ext4 superblock, and set the start of
 916                  * journal to 0.*/
 917                 uint32_t features_incompatible =
 918                         ext4_get32(&jbd_fs->inode_ref.fs->sb,
 919                                    features_incompatible);
 920                 jbd_set32(&jbd_fs->sb, start, 0);
 921                 features_incompatible &= ~EXT4_FINCOM_RECOVER;
 922                 ext4_set32(&jbd_fs->inode_ref.fs->sb,
 923                            features_incompatible,
 924                            features_incompatible);
 925                 jbd_fs->dirty = true;
 926                 r = ext4_sb_write(jbd_fs->inode_ref.fs->bdev,
 927                                   &jbd_fs->inode_ref.fs->sb);
 928         }
 929         jbd_destroy_revoke_tree(&info);
 930         return r;
 931 }
 932
 933 static void jbd_journal_write_sb(struct jbd_journal *journal)
 934 {
 935         struct jbd_fs *jbd_fs = journal->jbd_fs;
 936         jbd_set32(&jbd_fs->sb, start, journal->start);
 937         jbd_set32(&jbd_fs->sb, sequence, journal->trans_id);
 938         jbd_fs->dirty = true;
 939 }
 940
 941 /**@brief  Start accessing the journal.
 942  * @param  jbd_fs jbd filesystem
 943  * @param  journal current journal session
 944  * @return standard error code*/
 945 int jbd_journal_start(struct jbd_fs *jbd_fs,
 946                       struct jbd_journal *journal)
 947 {
 948         int r;
 949         uint32_t features_incompatible =
 950                         ext4_get32(&jbd_fs->inode_ref.fs->sb,
 951                                    features_incompatible);
 952         features_incompatible |= EXT4_FINCOM_RECOVER;
 953         ext4_set32(&jbd_fs->inode_ref.fs->sb,
 954                         features_incompatible,
 955                         features_incompatible);
 956         r = ext4_sb_write(jbd_fs->inode_ref.fs->bdev,
 957                         &jbd_fs->inode_ref.fs->sb);
 958         if (r != EOK)
 959                 return r;
 960
 961         journal->first = jbd_get32(&jbd_fs->sb, first);
 962         journal->start = journal->first;
 963         journal->last = journal->first;
 964         journal->trans_id = 1;
 965         journal->alloc_trans_id = 1;
 966
 967         journal->block_size = jbd_get32(&jbd_fs->sb, blocksize);
 968
 969         TAILQ_INIT(&journal->trans_queue);
 970         TAILQ_INIT(&journal->cp_queue);
 971         journal->jbd_fs = jbd_fs;
 972         jbd_journal_write_sb(journal);
 973         return jbd_write_sb(jbd_fs);
 974 }
 975
 976 /**@brief  Stop accessing the journal.
 977  * @param  journal current journal session
 978  * @return standard error code*/
 979 int jbd_journal_stop(struct jbd_journal *journal)
 980 {
 981         int r;
 982         struct jbd_fs *jbd_fs = journal->jbd_fs;
 983         uint32_t features_incompatible;
 984
 985         /* Commit all the transactions to the journal.*/
 986         jbd_journal_commit_all(journal);
 987         /* Make sure that journalled content have reached
 988          * the disk.*/
 989         ext4_block_cache_flush(jbd_fs->inode_ref.fs->bdev);
 990
 991         features_incompatible =
 992                 ext4_get32(&jbd_fs->inode_ref.fs->sb,
 993                            features_incompatible);
 994         features_incompatible &= ~EXT4_FINCOM_RECOVER;
 995         ext4_set32(&jbd_fs->inode_ref.fs->sb,
 996                         features_incompatible,
 997                         features_incompatible);
 998         r = ext4_sb_write(jbd_fs->inode_ref.fs->bdev,
 999                         &jbd_fs->inode_ref.fs->sb);
1000         if (r != EOK)
1001                 return r;
1002
1003         journal->start = 0;
1004         journal->trans_id = 0;
1005         jbd_journal_write_sb(journal);
1006         return jbd_write_sb(journal->jbd_fs);
1007 }
1008
1009 /**@brief  Allocate a block in the journal.
1010  * @param  journal current journal session
1011  * @param  trans transaction
1012  * @return allocated block address*/
1013 static uint32_t jbd_journal_alloc_block(struct jbd_journal *journal,
1014                                         struct jbd_trans *trans)
1015 {
1016         uint32_t start_block;
1017
1018         start_block = journal->last++;
1019         trans->alloc_blocks++;
1020         wrap(&journal->jbd_fs->sb, journal->last);
1021
1022         /* If there is no space left, flush all journalled
1023          * blocks to disk first.*/
1024         if (journal->last == journal->start)
1025                 ext4_block_cache_flush(journal->jbd_fs->inode_ref.fs->bdev);
1026
1027         return start_block;
1028 }
1029
1030 /**@brief  Allocate a new transaction
1031  * @param  journal current journal session
1032  * @return transaction allocated*/
1033 struct jbd_trans *
1034 jbd_journal_new_trans(struct jbd_journal *journal)
1035 {
1036         struct jbd_trans *trans = calloc(1, sizeof(struct jbd_trans));
1037         if (!trans)
1038                 return NULL;
1039
1040         /* We will assign a trans_id to this transaction,
1041          * once it has been committed.*/
1042         trans->journal = journal;
1043         trans->error = EOK;
1044         return trans;
1045 }
1046
1047 static void jbd_trans_end_write(struct ext4_bcache *bc __unused,
1048                           struct ext4_buf *buf __unused,
1049                           int res,
1050                           void *arg);
1051
1052 /**@brief  Add block to a transaction
1053  * @param  trans transaction
1054  * @param  block block descriptor
1055  * @return standard error code*/
1056 int jbd_trans_add_block(struct jbd_trans *trans,
1057                         struct ext4_block *block)
1058 {
1059         struct jbd_buf *buf;
1060         /* We do not need to add those unmodified buffer to
1061          * a transaction. */
1062         if (!ext4_bcache_test_flag(block->buf, BC_DIRTY))
1063                 return EOK;
1064
1065         buf = calloc(1, sizeof(struct jbd_buf));
1066         if (!buf)
1067                 return ENOMEM;
1068
1069         buf->trans = trans;
1070         buf->block = *block;
1071         ext4_bcache_inc_ref(block->buf);
1072
1073         /* If the content reach the disk, notify us
1074          * so that we may do a checkpoint. */
1075         block->buf->end_write = jbd_trans_end_write;
1076         block->buf->end_write_arg = trans;
1077
1078         trans->data_cnt++;
1079         LIST_INSERT_HEAD(&trans->buf_list, buf, buf_node);
1080         return EOK;
1081 }
1082
1083 /**@brief  Add block to be revoked to a transaction
1084  * @param  trans transaction
1085  * @param  lba logical block address
1086  * @return standard error code*/
1087 int jbd_trans_revoke_block(struct jbd_trans *trans,
1088                            ext4_fsblk_t lba)
1089 {
1090         struct jbd_revoke_rec *rec =
1091                 calloc(1, sizeof(struct jbd_revoke_rec));
1092         if (!rec)
1093                 return ENOMEM;
1094
1095         rec->lba = lba;
1096         LIST_INSERT_HEAD(&trans->revoke_list, rec, revoke_node);
1097         return EOK;
1098 }
1099
1100 /**@brief  Free a transaction
1101  * @param  journal current journal session
1102  * @param  trans transaction
1103  * @param  abort discard all the modifications on the block?
1104  * @return standard error code*/
1105 void jbd_journal_free_trans(struct jbd_journal *journal,
1106                             struct jbd_trans *trans,
1107                             bool abort)
1108 {
1109         struct jbd_buf *jbd_buf, *tmp;
1110         struct jbd_revoke_rec *rec, *tmp2;
1111         struct ext4_fs *fs = journal->jbd_fs->inode_ref.fs;
1112         LIST_FOREACH_SAFE(jbd_buf, &trans->buf_list, buf_node,
1113                           tmp) {
1114                 if (abort) {
1115                         ext4_bcache_clear_dirty(jbd_buf->block.buf);
1116                         ext4_block_set(fs->bdev, &jbd_buf->block);
1117                 }
1118
1119                 LIST_REMOVE(jbd_buf, buf_node);
1120                 free(jbd_buf);
1121         }
1122         LIST_FOREACH_SAFE(rec, &trans->revoke_list, revoke_node,
1123                           tmp2) {
1124                 LIST_REMOVE(rec, revoke_node);
1125                 free(rec);
1126         }
1127
1128         free(trans);
1129 }
1130
1131 /**@brief  Write commit block for a transaction
1132  * @param  trans transaction
1133  * @return standard error code*/
1134 static int jbd_trans_write_commit_block(struct jbd_trans *trans)
1135 {
1136         int rc;
1137         struct jbd_commit_header *header;
1138         uint32_t commit_iblock = 0;
1139         struct ext4_block commit_block;
1140         struct jbd_journal *journal = trans->journal;
1141
1142         commit_iblock = jbd_journal_alloc_block(journal, trans);
1143         rc = jbd_block_get_noread(journal->jbd_fs,
1144                         &commit_block, commit_iblock);
1145         if (rc != EOK)
1146                 return rc;
1147
1148         header = (struct jbd_commit_header *)commit_block.data;
1149         jbd_set32(&header->header, magic, JBD_MAGIC_NUMBER);
1150         jbd_set32(&header->header, blocktype, JBD_COMMIT_BLOCK);
1151         jbd_set32(&header->header, sequence, trans->trans_id);
1152
1153         ext4_bcache_set_dirty(commit_block.buf);
1154         rc = jbd_block_set(journal->jbd_fs, &commit_block);
1155         if (rc != EOK)
1156                 return rc;
1157
1158         return EOK;
1159 }
1160
1161 /**@brief  Write descriptor block for a transaction
1162  * @param  journal current journal session
1163  * @param  trans transaction
1164  * @return standard error code*/
1165 static int jbd_journal_prepare(struct jbd_journal *journal,
1166                                struct jbd_trans *trans)
1167 {
1168         int rc = EOK, i = 0;
1169         int32_t tag_tbl_size;
1170         uint32_t desc_iblock = 0;
1171         uint32_t data_iblock = 0;
1172         char *tag_start = NULL, *tag_ptr = NULL;
1173         struct jbd_buf *jbd_buf;
1174         struct ext4_block desc_block, data_block;
1175
1176         LIST_FOREACH(jbd_buf, &trans->buf_list, buf_node) {
1177                 struct tag_info tag_info;
1178                 bool uuid_exist = false;
1179 again:
1180                 if (!desc_iblock) {
1181                         struct jbd_bhdr *bhdr;
1182                         desc_iblock = jbd_journal_alloc_block(journal, trans);
1183                         rc = jbd_block_get_noread(journal->jbd_fs,
1184                                            &desc_block, desc_iblock);
1185                         if (rc != EOK)
1186                                 break;
1187
1188                         ext4_bcache_set_dirty(desc_block.buf);
1189
1190                         bhdr = (struct jbd_bhdr *)desc_block.data;
1191                         jbd_set32(bhdr, magic, JBD_MAGIC_NUMBER);
1192                         jbd_set32(bhdr, blocktype, JBD_DESCRIPTOR_BLOCK);
1193                         jbd_set32(bhdr, sequence, trans->trans_id);
1194
1195                         tag_start = (char *)(bhdr + 1);
1196                         tag_ptr = tag_start;
1197                         uuid_exist = true;
1198                         tag_tbl_size = journal->block_size -
1199                                 sizeof(struct jbd_bhdr);
1200
1201                         if (!trans->start_iblock)
1202                                 trans->start_iblock = desc_iblock;
1203
1204                 }
1205                 tag_info.block = jbd_buf->block.lb_id;
1206                 tag_info.uuid_exist = uuid_exist;
1207                 if (i == trans->data_cnt - 1)
1208                         tag_info.last_tag = true;
1209
1210                 if (uuid_exist)
1211                         memcpy(tag_info.uuid, journal->jbd_fs->sb.uuid,
1212                                         UUID_SIZE);
1213
1214                 rc = jbd_write_block_tag(journal->jbd_fs,
1215                                 tag_ptr,
1216                                 tag_tbl_size,
1217                                 &tag_info);
1218                 if (rc != EOK) {
1219                         jbd_block_set(journal->jbd_fs, &desc_block);
1220                         desc_iblock = 0;
1221                         goto again;
1222                 }
1223
1224                 data_iblock = jbd_journal_alloc_block(journal, trans);
1225                 rc = jbd_block_get_noread(journal->jbd_fs,
1226                                 &data_block, data_iblock);
1227                 if (rc != EOK)
1228                         break;
1229
1230                 ext4_bcache_set_dirty(data_block.buf);
1231
1232                 memcpy(data_block.data, jbd_buf->block.data,
1233                         journal->block_size);
1234
1235                 rc = jbd_block_set(journal->jbd_fs, &data_block);
1236                 if (rc != EOK)
1237                         break;
1238
1239                 tag_ptr += tag_info.tag_bytes;
1240                 tag_tbl_size -= tag_info.tag_bytes;
1241
1242                 i++;
1243         }
1244         if (rc == EOK && desc_iblock)
1245                 jbd_block_set(journal->jbd_fs, &desc_block);
1246
1247         return rc;
1248 }
1249
1250 /**@brief  Write revoke block for a transaction
1251  * @param  journal current journal session
1252  * @param  trans transaction
1253  * @return standard error code*/
1254 static int
1255 jbd_journal_prepare_revoke(struct jbd_journal *journal,
1256                            struct jbd_trans *trans)
1257 {
1258         int rc = EOK, i = 0;
1259         int32_t tag_tbl_size;
1260         uint32_t desc_iblock = 0;
1261         char *blocks_entry = NULL;
1262         struct jbd_revoke_rec *rec, *tmp;
1263         struct ext4_block desc_block;
1264         struct jbd_revoke_header *header = NULL;
1265         int32_t record_len = 4;
1266
1267         if (JBD_HAS_INCOMPAT_FEATURE(&journal->jbd_fs->sb,
1268                                      JBD_FEATURE_INCOMPAT_64BIT))
1269                 record_len = 8;
1270
1271         LIST_FOREACH_SAFE(rec, &trans->revoke_list, revoke_node,
1272                           tmp) {
1273 again:
1274                 if (!desc_iblock) {
1275                         struct jbd_bhdr *bhdr;
1276                         desc_iblock = jbd_journal_alloc_block(journal, trans);
1277                         rc = jbd_block_get_noread(journal->jbd_fs,
1278                                            &desc_block, desc_iblock);
1279                         if (rc != EOK) {
1280                                 break;
1281                         }
1282
1283                         ext4_bcache_set_dirty(desc_block.buf);
1284
1285                         bhdr = (struct jbd_bhdr *)desc_block.data;
1286                         jbd_set32(bhdr, magic, JBD_MAGIC_NUMBER);
1287                         jbd_set32(bhdr, blocktype, JBD_REVOKE_BLOCK);
1288                         jbd_set32(bhdr, sequence, trans->trans_id);
1289
1290                         header = (struct jbd_revoke_header *)bhdr;
1291                         blocks_entry = (char *)(header + 1);
1292                         tag_tbl_size = journal->block_size -
1293                                 sizeof(struct jbd_revoke_header);
1294
1295                         if (!trans->start_iblock)
1296                                 trans->start_iblock = desc_iblock;
1297
1298                 }
1299
1300                 if (tag_tbl_size < record_len) {
1301                         jbd_set32(header, count,
1302                                   journal->block_size - tag_tbl_size);
1303                         jbd_block_set(journal->jbd_fs, &desc_block);
1304                         desc_iblock = 0;
1305                         header = NULL;
1306                         goto again;
1307                 }
1308                 if (record_len == 8) {
1309                         uint64_t *blocks =
1310                                 (uint64_t *)blocks_entry;
1311                         *blocks = to_be64(rec->lba);
1312                 } else {
1313                         uint32_t *blocks =
1314                                 (uint32_t *)blocks_entry;
1315                         *blocks = to_be32(rec->lba);
1316                 }
1317                 blocks_entry += record_len;
1318                 tag_tbl_size -= record_len;
1319
1320                 i++;
1321         }
1322         if (rc == EOK && desc_iblock) {
1323                 if (header != NULL)
1324                         jbd_set32(header, count,
1325                                   journal->block_size - tag_tbl_size);
1326
1327                 jbd_block_set(journal->jbd_fs, &desc_block);
1328         }
1329
1330         return rc;
1331 }
1332
1333 /**@brief  Submit the transaction to transaction queue.
1334  * @param  journal current journal session
1335  * @param  trans transaction*/
1336 void
1337 jbd_journal_submit_trans(struct jbd_journal *journal,
1338                          struct jbd_trans *trans)
1339 {
1340         TAILQ_INSERT_TAIL(&journal->trans_queue,
1341                           trans,
1342                           trans_node);
1343 }
1344
1345 /**@brief  Put references of block descriptors in a transaction.
1346  * @param  journal current journal session
1347  * @param  trans transaction*/
1348 void jbd_journal_cp_trans(struct jbd_journal *journal, struct jbd_trans *trans)
1349 {
1350         struct jbd_buf *jbd_buf, *tmp;
1351         struct ext4_fs *fs = journal->jbd_fs->inode_ref.fs;
1352         LIST_FOREACH_SAFE(jbd_buf, &trans->buf_list, buf_node,
1353                         tmp) {
1354                 struct ext4_block block = jbd_buf->block;
1355                 ext4_block_set(fs->bdev, &block);
1356         }
1357 }
1358
1359 /**@brief  Update the start block of the journal when
1360  *         all the contents in a transaction reach the disk.*/
1361 static void jbd_trans_end_write(struct ext4_bcache *bc __unused,
1362                           struct ext4_buf *buf __unused,
1363                           int res,
1364                           void *arg)
1365 {
1366         struct jbd_trans *trans = arg;
1367         struct jbd_journal *journal = trans->journal;
1368         bool first_in_queue =
1369                 trans == TAILQ_FIRST(&journal->cp_queue);
1370         if (res != EOK)
1371                 trans->error = res;
1372
1373         trans->written_cnt++;
1374         if (trans->written_cnt == trans->data_cnt) {
1375                 TAILQ_REMOVE(&journal->cp_queue, trans, trans_node);
1376
1377                 if (first_in_queue) {
1378                         journal->start = trans->start_iblock +
1379                                 trans->alloc_blocks;
1380                         wrap(&journal->jbd_fs->sb, journal->start);
1381                         journal->trans_id = trans->trans_id + 1;
1382                 }
1383                 jbd_journal_free_trans(journal, trans, false);
1384
1385                 if (first_in_queue) {
1386                         while ((trans = TAILQ_FIRST(&journal->cp_queue))) {
1387                                 if (!trans->data_cnt) {
1388                                         TAILQ_REMOVE(&journal->cp_queue,
1389                                                      trans,
1390                                                      trans_node);
1391                                         journal->start = trans->start_iblock +
1392                                                 trans->alloc_blocks;
1393                                         wrap(&journal->jbd_fs->sb, journal->start);
1394                                         journal->trans_id = trans->trans_id + 1;
1395                                         jbd_journal_free_trans(journal,
1396                                                                trans, false);
1397                                 } else {
1398                                         journal->start = trans->start_iblock;
1399                                         wrap(&journal->jbd_fs->sb, journal->start);
1400                                         journal->trans_id = trans->trans_id;
1401                                         break;
1402                                 }
1403                         }
1404                         jbd_journal_write_sb(journal);
1405                         jbd_write_sb(journal->jbd_fs);
1406                 }
1407         }
1408 }
1409
1410 /**@brief  Commit a transaction to the journal immediately.
1411  * @param  journal current journal session
1412  * @param  trans transaction
1413  * @return standard error code*/
1414 int jbd_journal_commit_trans(struct jbd_journal *journal,
1415                              struct jbd_trans *trans)
1416 {
1417         int rc = EOK;
1418         uint32_t last = journal->last;
1419
1420         trans->trans_id = journal->alloc_trans_id;
1421         rc = jbd_journal_prepare(journal, trans);
1422         if (rc != EOK)
1423                 goto Finish;
1424
1425         rc = jbd_journal_prepare_revoke(journal, trans);
1426         if (rc != EOK)
1427                 goto Finish;
1428
1429         rc = jbd_trans_write_commit_block(trans);
1430         if (rc != EOK)
1431                 goto Finish;
1432
1433         journal->alloc_trans_id++;
1434         if (TAILQ_EMPTY(&journal->cp_queue)) {
1435                 if (trans->data_cnt) {
1436                         journal->start = trans->start_iblock;
1437                         wrap(&journal->jbd_fs->sb, journal->start);
1438                         journal->trans_id = trans->trans_id;
1439                         jbd_journal_write_sb(journal);
1440                         jbd_write_sb(journal->jbd_fs);
1441                         TAILQ_INSERT_TAIL(&journal->cp_queue, trans,
1442                                         trans_node);
1443                         jbd_journal_cp_trans(journal, trans);
1444                 } else {
1445                         journal->start = trans->start_iblock +
1446                                 trans->alloc_blocks;
1447                         wrap(&journal->jbd_fs->sb, journal->start);
1448                         journal->trans_id = trans->trans_id + 1;
1449                         jbd_journal_write_sb(journal);
1450                         jbd_journal_free_trans(journal, trans, false);
1451                 }
1452         } else {
1453                 TAILQ_INSERT_TAIL(&journal->cp_queue, trans,
1454                                 trans_node);
1455                 if (trans->data_cnt)
1456                         jbd_journal_cp_trans(journal, trans);
1457
1458         }
1459 Finish:
1460         if (rc != EOK) {
1461                 journal->last = last;
1462                 jbd_journal_free_trans(journal, trans, true);
1463         }
1464         return rc;
1465 }
1466
1467 /**@brief  Commit one transaction on transaction queue
1468  *         to the journal.
1469  * @param  journal current journal session.*/
1470 void jbd_journal_commit_one(struct jbd_journal *journal)
1471 {
1472         struct jbd_trans *trans;
1473
1474         if ((trans = TAILQ_FIRST(&journal->trans_queue))) {
1475                 TAILQ_REMOVE(&journal->trans_queue, trans, trans_node);
1476                 jbd_journal_commit_trans(journal, trans);
1477         }
1478 }
1479
1480 /**@brief  Commit all the transactions on transaction queue
1481  *         to the journal.
1482  * @param  journal current journal session.*/
1483 void jbd_journal_commit_all(struct jbd_journal *journal)
1484 {
1485         while (!TAILQ_EMPTY(&journal->trans_queue)) {
1486                 jbd_journal_commit_one(journal);
1487         }
1488 }
1489
1490 /**
1491  * @}
1492  */