ext4_journal: journal write skeleton code.
[lwext4.git] / lwext4 / ext4_journal.c
index 8fa135245ed3e737704d195860044e951a15f613..306ac75c5c5d95ee8b740840eed32d86e638b8a0 100644 (file)
@@ -1,6 +1,38 @@
+/*
+ * Copyright (c) 2015 Grzegorz Kostka (kostka.grzegorz@gmail.com)
+ * Copyright (c) 2015 Kaho Ng (ngkaho1234@gmail.com)
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions
+ * are met:
+ *
+ * - Redistributions of source code must retain the above copyright
+ *   notice, this list of conditions and the following disclaimer.
+ * - Redistributions in binary form must reproduce the above copyright
+ *   notice, this list of conditions and the following disclaimer in the
+ *   documentation and/or other materials provided with the distribution.
+ * - The name of the author may not be used to endorse or promote products
+ *   derived from this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
+ * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
+ * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
+ * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
+ * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
+ * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
+ * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+/** @addtogroup lwext4
+ * @{
+ */
 /**
  * @file  ext4_journal.c
- * @brief Journalling
+ * @brief Journal handle functions
  */
 
 #include "ext4_config.h"
 #include "tree.h"
 
 #include <string.h>
-#include <malloc.h>
+#include <stdlib.h>
+
+struct revoke_entry {
+       ext4_fsblk_t block;
+       uint32_t trans_id;
+       RB_ENTRY(revoke_entry) revoke_node;
+};
+
+struct recover_info {
+       uint32_t start_trans_id;
+       uint32_t last_trans_id;
+       uint32_t this_trans_id;
+       RB_HEAD(jbd_revoke, revoke_entry) revoke_root;
+};
+
+struct replay_arg {
+       struct recover_info *info;
+       uint32_t *this_block;
+       uint32_t this_trans_id;
+};
+
+static int
+jbd_revoke_entry_cmp(struct revoke_entry *a, struct revoke_entry *b)
+{
+       if (a->block > b->block)
+               return 1;
+       else if (a->block < b->block)
+               return -1;
+       return 0;
+}
+
+RB_GENERATE_INTERNAL(jbd_revoke, revoke_entry, revoke_node,
+                    jbd_revoke_entry_cmp, static inline)
+
+#define jbd_alloc_revoke_entry() calloc(1, sizeof(struct revoke_entry))
+#define jbd_free_revoke_entry(addr) free(addr)
 
 int jbd_inode_bmap(struct jbd_fs *jbd_fs,
                   ext4_lblk_t iblock,
@@ -63,6 +130,19 @@ static bool jbd_verify_sb(struct jbd_sb *sb)
        return true;
 }
 
+static int jbd_write_sb(struct jbd_fs *jbd_fs)
+{
+       int rc = EOK;
+       if (jbd_fs->dirty) {
+               rc = jbd_sb_write(jbd_fs, &jbd_fs->sb);
+               if (rc != EOK)
+                       return rc;
+
+               jbd_fs->dirty = false;
+       }
+       return rc;
+}
+
 int jbd_get_fs(struct ext4_fs *fs,
               struct jbd_fs *jbd_fs)
 {
@@ -83,6 +163,12 @@ int jbd_get_fs(struct ext4_fs *fs,
        if (rc != EOK) {
                memset(jbd_fs, 0, sizeof(struct jbd_fs));
                ext4_fs_put_inode_ref(&jbd_fs->inode_ref);
+               return rc;
+       }
+       if (!jbd_verify_sb(&jbd_fs->sb)) {
+               memset(jbd_fs, 0, sizeof(struct jbd_fs));
+               ext4_fs_put_inode_ref(&jbd_fs->inode_ref);
+               rc = EIO;
        }
 
        return rc;
@@ -90,8 +176,10 @@ int jbd_get_fs(struct ext4_fs *fs,
 
 int jbd_put_fs(struct jbd_fs *jbd_fs)
 {
-       int rc;
-       rc = ext4_fs_put_inode_ref(&jbd_fs->inode_ref);
+       int rc = EOK;
+       rc = jbd_write_sb(jbd_fs);
+
+       ext4_fs_put_inode_ref(&jbd_fs->inode_ref);
        return rc;
 }
 
@@ -99,7 +187,7 @@ int jbd_inode_bmap(struct jbd_fs *jbd_fs,
                   ext4_lblk_t iblock,
                   ext4_fsblk_t *fblock)
 {
-       int rc = ext4_fs_get_inode_data_block_index(
+       int rc = ext4_fs_get_inode_dblk_idx(
                        &jbd_fs->inode_ref,
                        iblock,
                        fblock,
@@ -172,64 +260,150 @@ int jbd_tag_bytes(struct jbd_fs *jbd_fs)
        return size - sizeof(uint32_t);
 }
 
-static void
+/**@brief: tag information. */
+struct tag_info {
+       int tag_bytes;
+       ext4_fsblk_t block;
+       bool uuid_exist;
+       uint8_t uuid[UUID_SIZE];
+       bool last_tag;
+};
+
+static int
 jbd_extract_block_tag(struct jbd_fs *jbd_fs,
-                     uint32_t tag_bytes,
                      void *__tag,
-                     ext4_fsblk_t *block,
-                     bool *uuid_exist,
-                     uint8_t *uuid,
-                     bool *last_tag)
+                     int tag_bytes,
+                     int32_t remain_buf_size,
+                     struct tag_info *tag_info)
 {
        char *uuid_start;
-       *uuid_exist = false;
-       *last_tag = false;
+       tag_info->tag_bytes = tag_bytes;
+       tag_info->uuid_exist = false;
+       tag_info->last_tag = false;
+
+       if (remain_buf_size - tag_bytes < 0)
+               return EINVAL;
+
        if (JBD_HAS_INCOMPAT_FEATURE(&jbd_fs->sb,
                                     JBD_FEATURE_INCOMPAT_CSUM_V3)) {
                struct jbd_block_tag3 *tag = __tag;
-               *block = jbd_get32(tag, blocknr);
+               tag_info->block = jbd_get32(tag, blocknr);
                if (JBD_HAS_INCOMPAT_FEATURE(&jbd_fs->sb,
                                             JBD_FEATURE_INCOMPAT_64BIT))
-                        *block |= (uint64_t)jbd_get32(tag, blocknr_high) << 32;
+                        tag_info->block |=
+                                (uint64_t)jbd_get32(tag, blocknr_high) << 32;
 
                if (jbd_get32(tag, flags) & JBD_FLAG_ESCAPE)
-                       *block = 0;
+                       tag_info->block = 0;
 
                if (!(jbd_get32(tag, flags) & JBD_FLAG_SAME_UUID)) {
+                       if (remain_buf_size - tag_bytes < UUID_SIZE)
+                               return EINVAL;
+
                        uuid_start = (char *)tag + tag_bytes;
-                       *uuid_exist = true;
-                       memcpy(uuid, uuid_start, UUID_SIZE);
+                       tag_info->uuid_exist = true;
+                       tag_info->tag_bytes += UUID_SIZE;
+                       memcpy(tag_info->uuid, uuid_start, UUID_SIZE);
                }
 
                if (jbd_get32(tag, flags) & JBD_FLAG_LAST_TAG)
-                       *last_tag = true;
+                       tag_info->last_tag = true;
 
        } else {
                struct jbd_block_tag *tag = __tag;
-               *block = jbd_get32(tag, blocknr);
+               tag_info->block = jbd_get32(tag, blocknr);
                if (JBD_HAS_INCOMPAT_FEATURE(&jbd_fs->sb,
                                             JBD_FEATURE_INCOMPAT_64BIT))
-                        *block |= (uint64_t)jbd_get32(tag, blocknr_high) << 32;
+                        tag_info->block |=
+                                (uint64_t)jbd_get32(tag, blocknr_high) << 32;
 
-               if (jbd_get32(tag, flags) & JBD_FLAG_ESCAPE)
-                       *block = 0;
+               if (jbd_get16(tag, flags) & JBD_FLAG_ESCAPE)
+                       tag_info->block = 0;
+
+               if (!(jbd_get16(tag, flags) & JBD_FLAG_SAME_UUID)) {
+                       if (remain_buf_size - tag_bytes < UUID_SIZE)
+                               return EINVAL;
 
-               if (!(jbd_get32(tag, flags) & JBD_FLAG_SAME_UUID)) {
                        uuid_start = (char *)tag + tag_bytes;
-                       *uuid_exist = true;
-                       memcpy(uuid, uuid_start, UUID_SIZE);
+                       tag_info->uuid_exist = true;
+                       tag_info->tag_bytes += UUID_SIZE;
+                       memcpy(tag_info->uuid, uuid_start, UUID_SIZE);
                }
 
-               if (jbd_get32(tag, flags) & JBD_FLAG_LAST_TAG)
-                       *last_tag = true;
+               if (jbd_get16(tag, flags) & JBD_FLAG_LAST_TAG)
+                       tag_info->last_tag = true;
+
+       }
+       return EOK;
+}
+
+static int
+jbd_write_block_tag(struct jbd_fs *jbd_fs,
+                   void *__tag,
+                   int32_t remain_buf_size,
+                   struct tag_info *tag_info)
+{
+       char *uuid_start;
+       int tag_bytes = jbd_tag_bytes(jbd_fs);
+
+       tag_info->tag_bytes = tag_bytes;
+
+       if (remain_buf_size - tag_bytes < 0)
+               return EINVAL;
+
+       if (JBD_HAS_INCOMPAT_FEATURE(&jbd_fs->sb,
+                                    JBD_FEATURE_INCOMPAT_CSUM_V3)) {
+               struct jbd_block_tag3 *tag = __tag;
+               jbd_set32(tag, blocknr, tag_info->block);
+               if (JBD_HAS_INCOMPAT_FEATURE(&jbd_fs->sb,
+                                            JBD_FEATURE_INCOMPAT_64BIT))
+                       jbd_set32(tag, blocknr_high, tag_info->block >> 32);
+
+               if (!tag_info->uuid_exist) {
+                       if (remain_buf_size - tag_bytes < UUID_SIZE)
+                               return EINVAL;
+
+                       uuid_start = (char *)tag + tag_bytes;
+                       tag_info->tag_bytes += UUID_SIZE;
+                       memcpy(uuid_start, tag_info->uuid, UUID_SIZE);
+                       jbd_set32(tag, flags,
+                                 jbd_get32(tag, flags) | JBD_FLAG_SAME_UUID);
+               }
+
+               if (tag_info->last_tag)
+                       jbd_set32(tag, flags,
+                                 jbd_get32(tag, flags) | JBD_FLAG_LAST_TAG);
+
+       } else {
+               struct jbd_block_tag *tag = __tag;
+               jbd_set32(tag, blocknr, tag_info->block);
+               if (JBD_HAS_INCOMPAT_FEATURE(&jbd_fs->sb,
+                                            JBD_FEATURE_INCOMPAT_64BIT))
+                       jbd_set32(tag, blocknr_high, tag_info->block >> 32);
+
+               if (!tag_info->uuid_exist) {
+                       if (remain_buf_size - tag_bytes < UUID_SIZE)
+                               return EINVAL;
+
+                       uuid_start = (char *)tag + tag_bytes;
+                       tag_info->tag_bytes += UUID_SIZE;
+                       memcpy(uuid_start, tag_info->uuid, UUID_SIZE);
+                       jbd_set16(tag, flags,
+                                 jbd_get16(tag, flags) | JBD_FLAG_SAME_UUID);
+               }
+
+               if (tag_info->last_tag)
+                       jbd_set16(tag, flags,
+                                 jbd_get16(tag, flags) | JBD_FLAG_LAST_TAG);
 
        }
+       return EOK;
 }
 
 static void
 jbd_iterate_block_table(struct jbd_fs *jbd_fs,
                        void *__tag_start,
-                       uint32_t tag_tbl_size,
+                       int32_t tag_tbl_size,
                        void (*func)(struct jbd_fs * jbd_fs,
                                        ext4_fsblk_t block,
                                        uint8_t *uuid,
@@ -237,9 +411,8 @@ jbd_iterate_block_table(struct jbd_fs *jbd_fs,
                        void *arg)
 {
        ext4_fsblk_t block = 0;
-       uint8_t uuid[UUID_SIZE];
        char *tag_start, *tag_ptr;
-       uint32_t tag_bytes = jbd_tag_bytes(jbd_fs);
+       int tag_bytes = jbd_tag_bytes(jbd_fs);
        tag_start = __tag_start;
        tag_ptr = tag_start;
 
@@ -249,26 +422,24 @@ jbd_iterate_block_table(struct jbd_fs *jbd_fs,
                                     JBD_FEATURE_INCOMPAT_CSUM_V3))
                tag_tbl_size -= sizeof(struct jbd_block_tail);
 
-       while (tag_ptr - tag_start + tag_bytes <= tag_tbl_size) {
-               bool uuid_exist;
-               bool last_tag;
-               jbd_extract_block_tag(jbd_fs,
-                                     tag_bytes,
+       while (tag_tbl_size) {
+               struct tag_info tag_info;
+               int rc = jbd_extract_block_tag(jbd_fs,
                                      tag_ptr,
-                                     &block,
-                                     &uuid_exist,
-                                     uuid,
-                                     &last_tag);
+                                     tag_bytes,
+                                     tag_tbl_size,
+                                     &tag_info);
+               if (rc != EOK)
+                       break;
+
                if (func)
-                       func(jbd_fs, block, uuid, arg);
+                       func(jbd_fs, block, tag_info.uuid, arg);
 
-               if (last_tag)
+               if (tag_info.last_tag)
                        break;
 
-               tag_ptr += tag_bytes;
-               if (uuid_exist)
-                       tag_ptr += UUID_SIZE;
-
+               tag_ptr += tag_info.tag_bytes;
+               tag_tbl_size -= tag_info.tag_bytes;
        }
 }
 
@@ -285,11 +456,112 @@ static void jbd_display_block_tags(struct jbd_fs *jbd_fs,
        return;
 }
 
-struct revoke_entry {
-       ext4_fsblk_t block;
-       uint32_t trans_id;
-       RB_ENTRY(revoke_entry) revoke_node;
-};
+static struct revoke_entry *
+jbd_revoke_entry_lookup(struct recover_info *info, ext4_fsblk_t block)
+{
+       struct revoke_entry tmp = {
+               .block = block
+       };
+
+       return RB_FIND(jbd_revoke, &info->revoke_root, &tmp);
+}
+
+static void jbd_replay_block_tags(struct jbd_fs *jbd_fs,
+                                 ext4_fsblk_t block,
+                                 uint8_t *uuid __unused,
+                                 void *__arg)
+{
+       int r;
+       struct replay_arg *arg = __arg;
+       struct recover_info *info = arg->info;
+       uint32_t *this_block = arg->this_block;
+       struct revoke_entry *revoke_entry;
+       struct ext4_block journal_block, ext4_block;
+       struct ext4_fs *fs = jbd_fs->inode_ref.fs;
+
+       (*this_block)++;
+
+       revoke_entry = jbd_revoke_entry_lookup(info, block);
+       if (revoke_entry &&
+           arg->this_trans_id < revoke_entry->trans_id)
+               return;
+
+       ext4_dbg(DEBUG_JBD,
+                "Replaying block in block_tag: %" PRIu64 "\n",
+                block);
+
+       r = jbd_block_get(jbd_fs, &journal_block, *this_block);
+       if (r != EOK)
+               return;
+
+       if (block) {
+               r = ext4_block_get_noread(fs->bdev, &ext4_block, block);
+               if (r != EOK) {
+                       jbd_block_set(jbd_fs, &journal_block);
+                       return;
+               }
+
+               memcpy(ext4_block.data,
+                       journal_block.data,
+                       jbd_get32(&jbd_fs->sb, blocksize));
+
+               ext4_bcache_set_dirty(ext4_block.buf);
+               ext4_block_set(fs->bdev, &ext4_block);
+       } else {
+               uint16_t mount_count, state;
+               mount_count = ext4_get16(&fs->sb, mount_count);
+               state = ext4_get16(&fs->sb, state);
+
+               memcpy(&fs->sb,
+                       journal_block.data + EXT4_SUPERBLOCK_OFFSET,
+                       EXT4_SUPERBLOCK_SIZE);
+
+               /* Mark system as mounted */
+               ext4_set16(&fs->sb, state, state);
+               r = ext4_sb_write(fs->bdev, &fs->sb);
+               if (r != EOK)
+                       return;
+
+               /*Update mount count*/
+               ext4_set16(&fs->sb, mount_count, mount_count);
+       }
+
+       jbd_block_set(jbd_fs, &journal_block);
+       
+       return;
+}
+
+static void jbd_add_revoke_block_tags(struct recover_info *info,
+                                     ext4_fsblk_t block)
+{
+       struct revoke_entry *revoke_entry;
+
+       ext4_dbg(DEBUG_JBD, "Add block %" PRIu64 " to revoke tree\n", block);
+       revoke_entry = jbd_revoke_entry_lookup(info, block);
+       if (revoke_entry) {
+               revoke_entry->trans_id = info->this_trans_id;
+               return;
+       }
+
+       revoke_entry = jbd_alloc_revoke_entry();
+       ext4_assert(revoke_entry);
+       revoke_entry->block = block;
+       revoke_entry->trans_id = info->this_trans_id;
+       RB_INSERT(jbd_revoke, &info->revoke_root, revoke_entry);
+
+       return;
+}
+
+static void jbd_destroy_revoke_tree(struct recover_info *info)
+{
+       while (!RB_EMPTY(&info->revoke_root)) {
+               struct revoke_entry *revoke_entry =
+                       RB_MIN(jbd_revoke, &info->revoke_root);
+               ext4_assert(revoke_entry);
+               RB_REMOVE(jbd_revoke, &info->revoke_root, revoke_entry);
+               jbd_free_revoke_entry(revoke_entry);
+       }
+}
 
 /* Make sure we wrap around the log correctly! */
 #define wrap(sb, var)                                          \
@@ -302,27 +574,37 @@ do {                                                                      \
 #define ACTION_REVOKE 1
 #define ACTION_RECOVER 2
 
-struct recover_info {
-       uint32_t start_trans_id;
-       uint32_t last_trans_id;
-       RB_HEAD(jbd_revoke, revoke_entry) revoke_root;
-};
 
-static void jbd_build_revoke_root(struct jbd_fs *jbd_fs,
+static void jbd_build_revoke_tree(struct jbd_fs *jbd_fs,
                                  struct jbd_bhdr *header,
                                  struct recover_info *info)
 {
+       char *blocks_entry;
        struct jbd_revoke_header *revoke_hdr =
                (struct jbd_revoke_header *)header;
-
-       jbd_iterate_block_table(jbd_fs,
-                               revoke_hdr + 1,
-                               jbd_get32(&jbd_fs->sb, blocksize) -
-                                       sizeof(struct jbd_revoke_header),
-                               jbd_display_block_tags,
-                               NULL);
-
-       (void)info;
+       uint32_t i, nr_entries, record_len = 4;
+       if (JBD_HAS_INCOMPAT_FEATURE(&jbd_fs->sb,
+                                    JBD_FEATURE_INCOMPAT_64BIT))
+               record_len = 8;
+
+       nr_entries = (jbd_get32(revoke_hdr, count) -
+                       sizeof(struct jbd_revoke_header)) /
+                       record_len;
+
+       blocks_entry = (char *)(revoke_hdr + 1);
+
+       for (i = 0;i < nr_entries;i++) {
+               if (record_len == 8) {
+                       uint64_t *blocks =
+                               (uint64_t *)blocks_entry;
+                       jbd_add_revoke_block_tags(info, to_be64(*blocks));
+               } else {
+                       uint32_t *blocks =
+                               (uint32_t *)blocks_entry;
+                       jbd_add_revoke_block_tags(info, to_be32(*blocks));
+               }
+               blocks_entry += record_len;
+       }
 }
 
 static void jbd_debug_descriptor_block(struct jbd_fs *jbd_fs,
@@ -330,13 +612,25 @@ static void jbd_debug_descriptor_block(struct jbd_fs *jbd_fs,
                                       uint32_t *iblock)
 {
        jbd_iterate_block_table(jbd_fs,
-                               header,
+                               header + 1,
                                jbd_get32(&jbd_fs->sb, blocksize) -
                                        sizeof(struct jbd_bhdr),
                                jbd_display_block_tags,
                                iblock);
 }
 
+static void jbd_replay_descriptor_block(struct jbd_fs *jbd_fs,
+                                       struct jbd_bhdr *header,
+                                       struct replay_arg *arg)
+{
+       jbd_iterate_block_table(jbd_fs,
+                               header + 1,
+                               jbd_get32(&jbd_fs->sb, blocksize) -
+                                       sizeof(struct jbd_bhdr),
+                               jbd_replay_block_tags,
+                               arg);
+}
+
 int jbd_iterate_log(struct jbd_fs *jbd_fs,
                    struct recover_info *info,
                    int action)
@@ -384,22 +678,37 @@ int jbd_iterate_log(struct jbd_fs *jbd_fs,
 
                switch (jbd_get32(header, blocktype)) {
                case JBD_DESCRIPTOR_BLOCK:
-                       ext4_dbg(DEBUG_JBD, "Descriptor block: %u, "
-                                           "trans_id: %u\n",
+                       ext4_dbg(DEBUG_JBD, "Descriptor block: %" PRIu32", "
+                                           "trans_id: %" PRIu32"\n",
                                            this_block, this_trans_id);
-                       jbd_debug_descriptor_block(jbd_fs, header, &this_block);
+                       if (action == ACTION_RECOVER) {
+                               struct replay_arg replay_arg;
+                               replay_arg.info = info;
+                               replay_arg.this_block = &this_block;
+                               replay_arg.this_trans_id = this_trans_id;
+
+                               jbd_replay_descriptor_block(jbd_fs,
+                                               header, &replay_arg);
+                       } else
+                               jbd_debug_descriptor_block(jbd_fs,
+                                               header, &this_block);
+
                        break;
                case JBD_COMMIT_BLOCK:
-                       ext4_dbg(DEBUG_JBD, "Commit block: %u, "
-                                           "trans_id: %u\n",
+                       ext4_dbg(DEBUG_JBD, "Commit block: %" PRIu32", "
+                                           "trans_id: %" PRIu32"\n",
                                            this_block, this_trans_id);
                        this_trans_id++;
                        break;
                case JBD_REVOKE_BLOCK:
-                       ext4_dbg(DEBUG_JBD, "Revoke block: %u, "
-                                           "trans_id: %u\n",
+                       ext4_dbg(DEBUG_JBD, "Revoke block: %" PRIu32", "
+                                           "trans_id: %" PRIu32"\n",
                                            this_block, this_trans_id);
-                       jbd_build_revoke_root(jbd_fs, header, info);
+                       if (action == ACTION_REVOKE) {
+                               info->this_trans_id = this_trans_id;
+                               jbd_build_revoke_tree(jbd_fs,
+                                               header, info);
+                       }
                        break;
                default:
                        log_end = true;
@@ -432,6 +741,234 @@ int jbd_recover(struct jbd_fs *jbd_fs)
        if (!sb->start)
                return EOK;
 
+       RB_INIT(&info.revoke_root);
+
        r = jbd_iterate_log(jbd_fs, &info, ACTION_SCAN);
+       if (r != EOK)
+               return r;
+
+       r = jbd_iterate_log(jbd_fs, &info, ACTION_REVOKE);
+       if (r != EOK)
+               return r;
+
+       r = jbd_iterate_log(jbd_fs, &info, ACTION_RECOVER);
+       if (r == EOK) {
+               jbd_set32(&jbd_fs->sb, start, 0);
+               jbd_fs->dirty = true;
+       }
+       jbd_destroy_revoke_tree(&info);
        return r;
 }
+
+void jbd_journal_write_sb(struct jbd_journal *journal)
+{
+       struct jbd_fs *jbd_fs = journal->jbd_fs;
+       jbd_set32(&jbd_fs->sb, start, journal->first);
+       jbd_set32(&jbd_fs->sb, sequence, journal->trans_id);
+       jbd_fs->dirty = true;
+}
+
+int jbd_journal_start(struct jbd_fs *jbd_fs,
+                     struct jbd_journal *journal)
+{
+       journal->first = jbd_get32(&jbd_fs->sb, first);
+       journal->start = journal->first;
+       journal->last = journal->first;
+       journal->trans_id = 1;
+
+       journal->block_size = jbd_get32(&jbd_fs->sb, blocksize);
+
+       TAILQ_INIT(&journal->trans_queue);
+       journal->jbd_fs = jbd_fs;
+       jbd_journal_write_sb(journal);
+       return jbd_write_sb(jbd_fs);
+}
+
+int jbd_journal_stop(struct jbd_journal *journal)
+{
+       journal->start = 0;
+       journal->trans_id = 0;
+       jbd_journal_write_sb(journal);
+       return jbd_write_sb(journal->jbd_fs);
+}
+
+static inline bool jbd_has_enough_space(struct jbd_journal *journal,
+                                       uint32_t blk_cnt)
+{
+       uint32_t new_last = journal->last + blk_cnt;
+       wrap(&journal->jbd_fs->sb, new_last);
+       if (new_last >= journal->start)
+               return false;
+
+       return true;
+}
+
+static uint32_t jbd_journal_alloc_block(struct jbd_journal *journal)
+{
+       uint32_t start_block = journal->last++;
+       wrap(&journal->jbd_fs->sb, journal->last);
+       return start_block;
+}
+
+struct jbd_trans *
+jbd_journal_new_trans()
+{
+       struct jbd_trans *trans = calloc(1, sizeof(struct jbd_trans));
+       if (!trans)
+               return NULL;
+
+       /* We will assign a trans_id to this transaction,
+        * once it has been committed.*/
+       return trans;
+}
+
+int jbd_trans_add_block(struct jbd_trans *trans,
+                       struct ext4_block *block)
+{
+       struct jbd_buf *buf = calloc(1, sizeof(struct jbd_buf));
+       if (!buf)
+               return ENOMEM;
+
+       buf->trans = trans;
+       buf->block = *block;
+       ext4_bcache_inc_ref(block->buf);
+       trans->buf_cnt++;
+       LIST_INSERT_HEAD(&trans->buf_list, buf, buf_node);
+       return EOK;
+}
+
+int jbd_trans_revoke_block(struct jbd_trans *trans,
+                          ext4_fsblk_t lba)
+{
+       struct jbd_revoke_rec *rec =
+               calloc(1, sizeof(struct jbd_revoke_rec));
+       if (!rec)
+               return ENOMEM;
+
+       rec->lba = lba;
+       trans->revoke_cnt++;
+       LIST_INSERT_HEAD(&trans->revoke_list, rec, revoke_node);
+       return EOK;
+}
+
+void jbd_journal_abort_trans(struct jbd_journal *journal,
+                            struct jbd_trans *trans)
+{
+       struct jbd_buf *var, *tmp;
+       struct jbd_revoke_rec *rec, *tmp2;
+       LIST_FOREACH_SAFE(var, &trans->buf_list, buf_node,
+                         tmp) {
+               ext4_block_set(journal->jbd_fs->bdev, &var->block);
+               LIST_REMOVE(var, buf_node);
+               free(var);
+       }
+       LIST_FOREACH_SAFE(rec, &trans->revoke_list, revoke_node,
+                         tmp2) {
+               LIST_REMOVE(rec, revoke_node);
+               free(rec);
+       }
+
+       free(trans);
+}
+
+static void jbd_journal_end_write()
+{
+
+}
+
+static int jbd_journal_prepare(struct jbd_journal *journal,
+                              struct jbd_trans *trans)
+{
+       int rc = EOK, i = 0;
+       int32_t tag_tbl_size;
+       uint32_t desc_iblock = 0;
+       uint32_t data_iblock = 0;
+       char *tag_start = NULL, *tag_ptr = NULL;
+       struct jbd_buf *jbd_buf;
+       struct ext4_block desc_block, data_block;
+
+       tag_tbl_size = journal->block_size - sizeof(struct jbd_bhdr);
+
+       LIST_FOREACH(jbd_buf, &trans->buf_list, buf_node) {
+               struct tag_info tag_info;
+               bool uuid_exist = false;
+again:
+               if (!desc_iblock) {
+                       struct jbd_bhdr *bhdr;
+                       desc_iblock = jbd_journal_alloc_block(journal);
+                       rc = jbd_block_get_noread(journal->jbd_fs,
+                                          &desc_block, desc_iblock);
+                       if (!rc)
+                               break;
+
+                       ext4_bcache_set_dirty(desc_block.buf);
+
+                       bhdr = (struct jbd_bhdr *)desc_block.data;
+                       bhdr->magic = JBD_MAGIC_NUMBER;
+                       bhdr->blocktype = JBD_DESCRIPTOR_BLOCK;
+                       bhdr->sequence = trans->trans_id;
+                       tag_start = (char *)(bhdr + 1);
+                       tag_ptr = tag_start;
+                       uuid_exist = true;
+               }
+               tag_info.block = jbd_buf->block.lb_id;
+               tag_info.uuid_exist = uuid_exist;
+               if (i == trans->buf_cnt - 1)
+                       tag_info.last_tag = true;
+
+               if (uuid_exist)
+                       memcpy(tag_info.uuid, journal->jbd_fs->sb.uuid,
+                                       UUID_SIZE);
+
+               rc = jbd_write_block_tag(journal->jbd_fs,
+                               tag_ptr,
+                               tag_tbl_size,
+                               &tag_info);
+               if (rc != EOK) {
+                       jbd_block_set(journal->jbd_fs, &desc_block);
+                       desc_iblock = 0;
+                       goto again;
+               }
+
+               data_iblock = jbd_journal_alloc_block(journal);
+               rc = jbd_block_get_noread(journal->jbd_fs,
+                               &data_block, data_iblock);
+               if (rc != EOK)
+                       break;
+
+               ext4_bcache_set_dirty(data_block.buf);
+
+               memcpy(data_block.data, jbd_buf->block.data,
+                       journal->block_size);
+
+               rc = jbd_block_set(journal->jbd_fs, &data_block);
+               if (rc != EOK)
+                       break;
+
+               tag_ptr += tag_info.tag_bytes;
+               tag_tbl_size -= tag_info.tag_bytes;
+
+               i++;
+       }
+       if (rc != EOK)
+               jbd_journal_abort_trans(journal, trans);
+
+       if (desc_iblock)
+               jbd_block_set(journal->jbd_fs, &desc_block);
+
+       return rc;
+}
+
+void
+jbd_journal_commit_trans(struct jbd_journal *journal,
+                        struct jbd_trans *trans)
+{
+       trans->trans_id = journal->trans_id++;
+       TAILQ_INSERT_TAIL(&journal->trans_queue,
+                         trans,
+                         trans_node);
+}
+
+/**
+ * @}
+ */