journal.c

嵌入式系统设计与实例开发实验教材二源码 多线程应用程序设计 串行端口程序设计 AD接口实验 CAN总线通信实验 GPS通信实验 Linux内核移植与编译实验 IC卡读写实验 SD驱动使

  }
  c_bh = sb_bread(p_s_sb, reiserfs_get_journal_block(p_s_sb) + ((trans_offset + le32_to_cpu(desc->j_len) + 1) % JOURNAL_BLOCK_COUNT)) ;
  if (!c_bh) {
    brelse(d_bh) ;
    return 1 ;
  }
  commit = (struct reiserfs_journal_commit *)c_bh->b_data ;
  if (journal_compare_desc_commit(p_s_sb, desc, commit)) {
    reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE, "journal_read_transaction, "
                   "commit offset %ld had bad time %d or length %d\n", 
		   c_bh->b_blocknr - reiserfs_get_journal_block(p_s_sb), 
		   le32_to_cpu(commit->j_trans_id), le32_to_cpu(commit->j_len));
    brelse(c_bh) ;
    brelse(d_bh) ;
    return 1; 
  }
  trans_id = le32_to_cpu(desc->j_trans_id) ;
  /* now we know we've got a good transaction, and it was inside the valid time ranges */
  log_blocks = kmalloc(le32_to_cpu(desc->j_len) * sizeof(struct buffer_head *), GFP_NOFS) ;
  real_blocks = kmalloc(le32_to_cpu(desc->j_len) * sizeof(struct buffer_head *), GFP_NOFS) ;
  if (!log_blocks  || !real_blocks) {
    brelse(c_bh) ;
    brelse(d_bh) ;
    kfree(log_blocks) ;
    kfree(real_blocks) ;
    reiserfs_warning("journal-1169: kmalloc failed, unable to mount FS\n") ;
    return -1 ;
  }
  /* get all the buffer heads */
  for(i = 0 ; i < le32_to_cpu(desc->j_len) ; i++) {
    log_blocks[i] = sb_getblk(p_s_sb, reiserfs_get_journal_block(p_s_sb) + (trans_offset + 1 + i) % JOURNAL_BLOCK_COUNT);
    if (i < JOURNAL_TRANS_HALF) {
      real_blocks[i] = sb_getblk(p_s_sb, le32_to_cpu(desc->j_realblock[i])) ;
    } else {
      real_blocks[i] = sb_getblk(p_s_sb, le32_to_cpu(commit->j_realblock[i - JOURNAL_TRANS_HALF])) ;
    }
    if (real_blocks[i]->b_blocknr >= reiserfs_get_journal_block(p_s_sb) &&
        real_blocks[i]->b_blocknr < (reiserfs_get_journal_block(p_s_sb)+JOURNAL_BLOCK_COUNT)) {
      reiserfs_warning("journal-1204: REPLAY FAILURE fsck required! Trying to replay onto a log block\n") ;
      brelse_array(log_blocks, i) ;
      brelse_array(real_blocks, i) ;
      brelse(c_bh) ;
      brelse(d_bh) ;
      kfree(log_blocks) ;
      kfree(real_blocks) ;
      return -1 ;
    }
  }
  /* read in the log blocks, memcpy to the corresponding real block */
  ll_rw_block(READ, le32_to_cpu(desc->j_len), log_blocks) ;
  for (i = 0 ; i < le32_to_cpu(desc->j_len) ; i++) {
    wait_on_buffer(log_blocks[i]) ;
    if (!buffer_uptodate(log_blocks[i])) {
      reiserfs_warning("journal-1212: REPLAY FAILURE fsck required! buffer write failed\n") ;
      brelse_array(log_blocks + i, le32_to_cpu(desc->j_len) - i) ;
      brelse_array(real_blocks, le32_to_cpu(desc->j_len)) ;
      brelse(c_bh) ;
      brelse(d_bh) ;
      kfree(log_blocks) ;
      kfree(real_blocks) ;
      return -1 ;
    }
    memcpy(real_blocks[i]->b_data, log_blocks[i]->b_data, real_blocks[i]->b_size) ;
    mark_buffer_uptodate(real_blocks[i], 1) ;
    brelse(log_blocks[i]) ;
  }
  /* flush out the real blocks */
  for (i = 0 ; i < le32_to_cpu(desc->j_len) ; i++) {
    set_bit(BH_Dirty, &(real_blocks[i]->b_state)) ;
    ll_rw_block(WRITE, 1, real_blocks + i) ;
  }
  for (i = 0 ; i < le32_to_cpu(desc->j_len) ; i++) {
    wait_on_buffer(real_blocks[i]) ; 
    if (!buffer_uptodate(real_blocks[i])) {
      reiserfs_warning("journal-1226: REPLAY FAILURE, fsck required! buffer write failed\n") ;
      brelse_array(real_blocks + i, le32_to_cpu(desc->j_len) - i) ;
      brelse(c_bh) ;
      brelse(d_bh) ;
      kfree(log_blocks) ;
      kfree(real_blocks) ;
      return -1 ;
    }
    brelse(real_blocks[i]) ;
  }
  cur_dblock = reiserfs_get_journal_block(p_s_sb) + ((trans_offset + le32_to_cpu(desc->j_len) + 2) % JOURNAL_BLOCK_COUNT) ;
  reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE, "journal-1095: setting journal "
                 "start to offset %ld\n", 
		 cur_dblock - reiserfs_get_journal_block(p_s_sb)) ;
  
  /* init starting values for the first transaction, in case this is the last transaction to be replayed. */
  SB_JOURNAL(p_s_sb)->j_start = cur_dblock - reiserfs_get_journal_block(p_s_sb) ;
  SB_JOURNAL(p_s_sb)->j_last_flush_trans_id = trans_id ;
  SB_JOURNAL(p_s_sb)->j_trans_id = trans_id + 1;
  brelse(c_bh) ;
  brelse(d_bh) ;
  kfree(log_blocks) ;
  kfree(real_blocks) ;
  return 0 ;
}

/*
** read and replay the log
** on a clean unmount, the journal header's next unflushed pointer will be to an invalid
** transaction.  This tests that before finding all the transactions in the log, whic makes normal mount times fast.
**
** After a crash, this starts with the next unflushed transaction, and replays until it finds one too old, or invalid.
**
** On exit, it sets things up so the first transaction will work correctly.
*/
static int journal_read(struct super_block *p_s_sb) {
  struct reiserfs_journal_desc *desc ;
  unsigned long last_flush_trans_id = 0 ;
  unsigned long oldest_trans_id = 0;
  unsigned long oldest_invalid_trans_id = 0 ;
  time_t start ;
  unsigned long last_flush_start = 0;
  unsigned long oldest_start = 0;
  unsigned long cur_dblock = 0 ;
  unsigned long newest_mount_id = 9 ;
  struct buffer_head *d_bh ;
  struct reiserfs_journal_header *jh ;
  int valid_journal_header = 0 ;
  int replay_count = 0 ;
  int continue_replay = 1 ;
  int ret ;

  cur_dblock = reiserfs_get_journal_block(p_s_sb) ;
  printk("reiserfs: checking transaction log (device %s) ...\n",
          kdevname(p_s_sb->s_dev)) ;
  start = CURRENT_TIME ;

  /* step 1, read in the journal header block.  Check the transaction it says 
  ** is the first unflushed, and if that transaction is not valid, 
  ** replay is done
  */
  SB_JOURNAL(p_s_sb)->j_header_bh = sb_bread(p_s_sb, 
                                          reiserfs_get_journal_block(p_s_sb) + 
					  JOURNAL_BLOCK_COUNT) ;
  if (!SB_JOURNAL(p_s_sb)->j_header_bh) {
    return 1 ;
  }
  jh = (struct reiserfs_journal_header *)(SB_JOURNAL(p_s_sb)->j_header_bh->b_data) ;
  if (le32_to_cpu(jh->j_first_unflushed_offset) >= 0 && 
      le32_to_cpu(jh->j_first_unflushed_offset) < JOURNAL_BLOCK_COUNT &&
      le32_to_cpu(jh->j_last_flush_trans_id) > 0) {
    last_flush_start = reiserfs_get_journal_block(p_s_sb) + 
                       le32_to_cpu(jh->j_first_unflushed_offset) ;
    last_flush_trans_id = le32_to_cpu(jh->j_last_flush_trans_id) ;
    reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE, "journal-1153: found in "
                   "header: first_unflushed_offset %d, last_flushed_trans_id "
		   "%lu\n", le32_to_cpu(jh->j_first_unflushed_offset), 
		   last_flush_trans_id) ;
    valid_journal_header = 1 ;

    /* now, we try to read the first unflushed offset.  If it is not valid, 
    ** there is nothing more we can do, and it makes no sense to read 
    ** through the whole log.
    */
    d_bh = sb_bread(p_s_sb, reiserfs_get_journal_block(p_s_sb) + le32_to_cpu(jh->j_first_unflushed_offset)) ;
    ret = journal_transaction_is_valid(p_s_sb, d_bh, NULL, NULL) ;
    if (!ret) {
      continue_replay = 0 ;
    }
    brelse(d_bh) ;
  }

  if (continue_replay && is_read_only(p_s_sb->s_dev)) {
    printk("clm-2076: device is readonly, unable to replay log\n") ;
    return -1 ;
  }
  if (continue_replay && (p_s_sb->s_flags & MS_RDONLY)) {
    printk("Warning, log replay starting on readonly filesystem\n") ;    
  }

  /* ok, there are transactions that need to be replayed.  start with the first log block, find
  ** all the valid transactions, and pick out the oldest.
  */
  while(continue_replay && cur_dblock < (reiserfs_get_journal_block(p_s_sb) + JOURNAL_BLOCK_COUNT)) {
    d_bh = sb_bread(p_s_sb, cur_dblock) ;
    ret = journal_transaction_is_valid(p_s_sb, d_bh, &oldest_invalid_trans_id, &newest_mount_id) ;
    if (ret == 1) {
      desc = (struct reiserfs_journal_desc *)d_bh->b_data ;
      if (oldest_start == 0) { /* init all oldest_ values */
        oldest_trans_id = le32_to_cpu(desc->j_trans_id) ;
	oldest_start = d_bh->b_blocknr ;
	newest_mount_id = le32_to_cpu(desc->j_mount_id) ;
	reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE, "journal-1179: Setting "
	               "oldest_start to offset %lu, trans_id %lu\n", 
		       oldest_start - reiserfs_get_journal_block(p_s_sb), 
		       oldest_trans_id) ;
      } else if (oldest_trans_id > le32_to_cpu(desc->j_trans_id)) { 
        /* one we just read was older */
        oldest_trans_id = le32_to_cpu(desc->j_trans_id) ;
	oldest_start = d_bh->b_blocknr ;
	reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE, "journal-1180: Resetting "
	               "oldest_start to offset %lu, trans_id %lu\n", 
			oldest_start - reiserfs_get_journal_block(p_s_sb), 
			oldest_trans_id) ;
      }
      if (newest_mount_id < le32_to_cpu(desc->j_mount_id)) {
        newest_mount_id = le32_to_cpu(desc->j_mount_id) ;
	reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE, "journal-1299: Setting "
	              "newest_mount_id to %d\n", le32_to_cpu(desc->j_mount_id));
      }
      cur_dblock += le32_to_cpu(desc->j_len) + 2 ;
    } 
    else {
      cur_dblock++ ;
    }
    brelse(d_bh) ;
  }
  /* step three, starting at the oldest transaction, replay */
  if (last_flush_start > 0) {
    oldest_start = last_flush_start ;
    oldest_trans_id = last_flush_trans_id + 1 ;
  } 
  cur_dblock = oldest_start ;
  if (oldest_trans_id)  {
    reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE, "journal-1206: Starting replay "
                   "from offset %lu, trans_id %lu\n", 
		   cur_dblock - reiserfs_get_journal_block(p_s_sb), 
		   oldest_trans_id) ;

  }
  replay_count = 0 ;
  while(continue_replay && oldest_trans_id > 0) {
    ret = journal_read_transaction(p_s_sb, cur_dblock, oldest_start, oldest_trans_id, newest_mount_id) ;
    if (ret < 0) {
      return ret ;
    } else if (ret != 0) {
      break ;
    }
    cur_dblock = reiserfs_get_journal_block(p_s_sb) + SB_JOURNAL(p_s_sb)->j_start ;
    replay_count++ ;
   if (cur_dblock == oldest_start)
        break;
  }

  if (oldest_trans_id == 0) {
    reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE, "journal-1225: No valid "
                   "transactions found\n") ;
  }
  /* j_start does not get set correctly if we don't replay any transactions.
  ** if we had a valid journal_header, set j_start to the first unflushed transaction value,
  ** copy the trans_id from the header
  */
  if (valid_journal_header && replay_count == 0) { 
    SB_JOURNAL(p_s_sb)->j_start = le32_to_cpu(jh->j_first_unflushed_offset) ;
    SB_JOURNAL(p_s_sb)->j_trans_id = le32_to_cpu(jh->j_last_flush_trans_id) + 1;
    SB_JOURNAL(p_s_sb)->j_last_flush_trans_id = le32_to_cpu(jh->j_last_flush_trans_id) ;
    SB_JOURNAL(p_s_sb)->j_mount_id = le32_to_cpu(jh->j_mount_id) + 1;
  } else {
    SB_JOURNAL(p_s_sb)->j_mount_id = newest_mount_id + 1 ;
  }
  reiserfs_debug(p_s_sb, REISERFS_DEBUG_CODE, "journal-1299: Setting "
                 "newest_mount_id to %lu\n", SB_JOURNAL(p_s_sb)->j_mount_id) ;
  SB_JOURNAL(p_s_sb)->j_first_unflushed_offset = SB_JOURNAL(p_s_sb)->j_start ; 
  if (replay_count > 0) {
    printk("reiserfs: replayed %d transactions in %lu seconds\n", replay_count, 
	    CURRENT_TIME - start) ;
  }
  if (!is_read_only(p_s_sb->s_dev) && 
       _update_journal_header_block(p_s_sb, SB_JOURNAL(p_s_sb)->j_start, 
                                   SB_JOURNAL(p_s_sb)->j_last_flush_trans_id))
  {
      /* replay failed, caller must call free_journal_ram and abort
      ** the mount
      */
      return -1 ;
  }
  return 0 ;
}


struct reiserfs_journal_commit_task {
  struct super_block *p_s_sb ;
  int jindex ;
  int wake_on_finish ; /* if this is one, we wake the task_done queue, if it
                       ** is zero, we free the whole struct on finish
		       */
  struct reiserfs_journal_commit_task *self ;
  struct wait_queue *task_done ;
  struct tq_struct task ;
} ;

static void reiserfs_journal_commit_task_func(struct reiserfs_journal_commit_task *ct) {

  struct reiserfs_journal_list *jl ;
  jl = SB_JOURNAL_LIST(ct->p_s_sb) + ct->jindex ;

  flush_commit_list(ct->p_s_sb, SB_JOURNAL_LIST(ct->p_s_sb) + ct->jindex, 1) ; 
  if (jl->j_len > 0 && atomic_read(&(jl->j_nonzerolen)) > 0 && 
      atomic_read(&(jl->j_commit_left)) == 0) {
    kupdate_one_transaction(ct->p_s_sb, jl) ;
  }
  kfree(ct->self) ;
}

static void setup_commit_task_arg(struct reiserfs_journal_commit_task *ct,
                                  struct super_block *p_s_sb, 
				  int jindex) {
  if (!ct) {
    reiserfs_panic(NULL, "journal-1360: setup_commit_task_arg called with NULL struct\n") ;
  }
  ct->p_s_sb = p_s_sb ;
  ct->jindex = jindex ;
  ct->task_done = NULL ;
  INIT_LIST_HEAD(&ct->task.list) ;
  ct->task.sync = 0 ;
  ct->task.routine = (void *)(void *)reiserfs_journal_commit_task_func ; 
  ct->self = ct ;
  ct->task.data = (void *)ct ;
}

static void commit_flush_async(struct super_block *p_s_sb, int jindex) {
  struct reiserfs_journal_commit_task *ct ;
  /* using GFP_NOFS, GFP_KERNEL could try to flush inodes, which will try
  ** to start/join a transaction, which will deadlock
  */
  ct = kmalloc(sizeof(struct reiserfs_journal_commit_task), GFP_NOFS) ;
  if (ct) {
    setup_commit_task_arg(ct, p_s_sb, jindex) ;
    queue_task(&(ct->task), &reiserfs_commit_thread_tq);
    wake_up(&reiserfs_commit_thread_wait) ;
  } else {
#ifdef CONFIG_REISERFS_CHECK
    reiserfs_warning("journal-1540: kmalloc failed, doing sync commit\n") ;
#endif
    flush_commit_list(p_s_sb, SB_JOURNAL_LIST(p_s_sb) + jindex, 1) ;
  }
}

/*
** this is the commit thread.  It is started with kernel_thread on
** FS mount, and journal_release() waits for it to exit.
**
** It could do a periodic commit, but there is a lot code for that
** elsewhere right now, and I only wanted to implement this little
** piece for starters.
**
** All we do here is sleep on the j_commit_thread_wait wait queue, and
** then run the per filesystem commit task queue when we wakeup.
*/
static int reiserfs_journal_commit_thread(void *nullp) {

  daemonize() ;

  spin_lock_irq(&current->sigmask_lock);
  sigfillset(&current->blocked);
  recalc_sigpending(current);
  spin_unlock_irq(&current->sigmask_lock);

  sprintf(current->comm, "kreiserfsd") ;
  lock_kernel() ;
  while(1) {

    while(TQ_ACTIVE(reiserfs_commit_thread_tq)) {
      run_task_queue(&reiserfs_commit_thread_tq) ;
    }

    /* if there aren't any more filesystems left, break */
    if (reiserfs_mounted_fs_count <= 0) {
      run_task_queue(&reiserfs_commit_thread_tq) ;
      break ;
    }
    wake_up(&reiserfs_commit_thread_done) ;
    interruptible_sleep_on_timeout(&reiserfs_commit_thread_wait, 5 * HZ) ;
  }
  unlock_kernel() ;
  wake_up(&reiserfs_c