void jbd_preclean_buffer_check(struct buffer_head *bh)
{
	if (buffer_jbd(bh)) {
	struct journal_head *jh = bh2jh(bh);

	transaction_t *transaction = jh->b_transaction;
	journal_t *journal;

	if (jh->b_jlist == 0 && transaction == NULL)
	return;

	J_ASSERT_JH(jh, transaction != NULL);
	/* The kernel may be unmapping old data. We expect it
	* to be dirty in that case, unless the buffer has
	* already been forgotten by a transaction. */
	if (jh->b_jlist != BJ_Forget) {
#if 1
	if (!buffer_dirty(bh)) {
	printk("%s: clean of clean buffer\n",
	__FUNCTION__);
	print_buffer_trace(bh);
	return;
	}
#endif
	J_ASSERT_BH(bh, buffer_dirty(bh));
	}

	journal = transaction->t_journal;
	J_ASSERT_JH(jh,
	transaction == journal->j_running_transaction ||
	transaction == journal->j_committing_transaction);
	}
}

void print_buffer_fields(struct buffer_head *bh)
{
	printk("b_next:%p, b_blocknr:%lu b_count:%d b_flushtime:%lu\n",
		bh->b_next, bh->b_blocknr, atomic_read(&bh->b_count),
			bh->b_flushtime);
	printk("b_next_free:%p b_prev_free:%p b_this_page:%p b_reqnext:%p\n",
		bh->b_next_free, bh->b_prev_free, bh->b_this_page,
			bh->b_reqnext);
	printk("b_pprev:%p b_data:%p b_page:%p b_inode:%p b_list:%d\n",
		bh->b_pprev, bh->b_data, bh->b_page, bh->b_inode, bh->b_list);
#if defined(CONFIG_JBD) || defined(CONFIG_JBD_MODULE)
	if (buffer_jbd(bh)) {
		struct journal_head *jh = bh2jh(bh);

		printk("b_jlist:%u b_frozen_data:%p b_committed_data:%p\n",
			jh->b_jlist, jh->b_frozen_data, jh->b_committed_data);
		printk(" b_transaction:%p b_next_transaction:%p "
				"b_cp_transaction:%p\n",
			jh->b_transaction, jh->b_next_transaction,
			jh->b_cp_transaction);
		printk("b_cpnext:%p b_cpprev:%p\n",
			jh->b_cpnext, jh->b_cpprev);
	}
#endif
}

int jbd2_journal_try_to_free_buffers(journal_t *journal,
				struct page *page, gfp_t gfp_mask)
{
	struct buffer_head *head;
	struct buffer_head *bh;
	int ret = 0;

	J_ASSERT(PageLocked(page));

	head = page_buffers(page);
	bh = head;
	do {
		struct journal_head *jh;

		jh = jbd2_journal_grab_journal_head(bh);
		if (!jh)
			continue;

		jbd_lock_bh_state(bh);
		__journal_try_to_free_buffer(journal, bh);
		jbd2_journal_put_journal_head(jh);
		jbd_unlock_bh_state(bh);
		if (buffer_jbd(bh))
			goto busy;
	} while ((bh = bh->b_this_page) != head);

	ret = try_to_free_buffers(page);

busy:
	return ret;
}

/*
 * Grab a ref against this buffer_head's journal_head.  If it ended up not
 * having a journal_head, return NULL
 */
struct journal_head *journal_grab_journal_head(struct buffer_head *bh)
{
    struct journal_head *jh = NULL;

    jbd_lock_bh_journal_head(bh);
    if (buffer_jbd(bh)) {
        jh = bh2jh(bh);
        jh->b_jcount++;
    }
    jbd_unlock_bh_journal_head(bh);
    return jh;
}

/*
 * More or less lifted from ext3. I'll leave their description below:
 *
 * "For ext3 allocations, we must not reuse any blocks which are
 * allocated in the bitmap buffer's "last committed data" copy.  This
 * prevents deletes from freeing up the page for reuse until we have
 * committed the delete transaction.
 *
 * If we didn't do this, then deleting something and reallocating it as
 * data would allow the old block to be overwritten before the
 * transaction committed (because we force data to disk before commit).
 * This would lead to corruption if we crashed between overwriting the
 * data and committing the delete.
 *
 * @@@ We may want to make this allocation behaviour conditional on
 * data-writes at some point, and disable it for metadata allocations or
 * sync-data inodes."
 *
 * Note: OCFS2 already does this differently for metadata vs data
 * allocations, as those bitmaps are seperate and undo access is never
 * called on a metadata group descriptor.
 */
static int ocfs2_test_bg_bit_allocatable(struct buffer_head *bg_bh,
					 int nr)
{
	struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) bg_bh->b_data;

	if (ocfs2_test_bit(nr, (unsigned long *)bg->bg_bitmap))
		return 0;
	if (!buffer_jbd(bg_bh) || !bh2jh(bg_bh)->b_committed_data)
		return 1;

	bg = (struct ocfs2_group_desc *) bh2jh(bg_bh)->b_committed_data;
	return !ocfs2_test_bit(nr, (unsigned long *)bg->bg_bitmap);
}

/*
 * Give a buffer_head a journal_head.
 *
 * Doesn't need the journal lock.
 * May sleep.
 */
struct journal_head *journal_add_journal_head(struct buffer_head *bh)
{
    struct journal_head *jh;
    struct journal_head *new_jh = NULL;

repeat:
    if (!buffer_jbd(bh)) {
        new_jh = journal_alloc_journal_head();
        memset(new_jh, 0, sizeof(*new_jh));
    }

    jbd_lock_bh_journal_head(bh);
    if (buffer_jbd(bh)) {
        jh = bh2jh(bh);
    } else {
        J_ASSERT_BH(bh,
                    (atomic_read(&bh->b_count) > 0) ||
                    (bh->b_page && bh->b_page->mapping));

        if (!new_jh) {
            jbd_unlock_bh_journal_head(bh);
            goto repeat;
        }

        jh = new_jh;
        new_jh = NULL;		/* We consumed it */
        set_buffer_jbd(bh);
        bh->b_private = jh;
        jh->b_bh = bh;
        get_bh(bh);
        BUFFER_TRACE(bh, "added journal_head");
    }
    jh->b_jcount++;
    jbd_unlock_bh_journal_head(bh);
    if (new_jh)
        journal_free_journal_head(new_jh);
    return bh->b_private;
}

int ocfs2_write_block(struct ocfs2_super *osb, struct buffer_head *bh,
		      struct inode *inode)
{
	int ret = 0;

	mlog_entry("(bh->b_blocknr = %llu, inode=%p)\n",
		   (unsigned long long)bh->b_blocknr, inode);

	BUG_ON(bh->b_blocknr < OCFS2_SUPER_BLOCK_BLKNO);
	BUG_ON(buffer_jbd(bh));

	/* No need to check for a soft readonly file system here. non
	 * journalled writes are only ever done on system files which
	 * can get modified during recovery even if read-only. */
	if (ocfs2_is_hard_readonly(osb)) {
		ret = -EROFS;
		goto out;
	}

	mutex_lock(&OCFS2_I(inode)->ip_io_mutex);

	lock_buffer(bh);
	set_buffer_uptodate(bh);

	/* remove from dirty list before I/O. */
	clear_buffer_dirty(bh);

	get_bh(bh); /* for end_buffer_write_sync() */
	bh->b_end_io = end_buffer_write_sync;
	submit_bh(WRITE, bh);

	wait_on_buffer(bh);

	if (buffer_uptodate(bh)) {
		ocfs2_set_buffer_uptodate(inode, bh);
	} else {
		/* We don't need to remove the clustered uptodate
		 * information for this bh as it's not marked locally
		 * uptodate. */
		ret = -EIO;
		put_bh(bh);
	}

	mutex_unlock(&OCFS2_I(inode)->ip_io_mutex);
out:
	mlog_exit(ret);
	return ret;
}

int ocfs2_write_block(struct ocfs2_super *osb, struct buffer_head *bh,
		      struct ocfs2_caching_info *ci)
{
	int ret = 0;

	trace_ocfs2_write_block((unsigned long long)bh->b_blocknr, ci);

	BUG_ON(bh->b_blocknr < OCFS2_SUPER_BLOCK_BLKNO);
	BUG_ON(buffer_jbd(bh));

	/* No need to check for a soft readonly file system here. non
	 * journalled writes are only ever done on system files which
	 * can get modified during recovery even if read-only. */
	if (ocfs2_is_hard_readonly(osb)) {
		ret = -EROFS;
		mlog_errno(ret);
		goto out;
	}

	ocfs2_metadata_cache_io_lock(ci);

	lock_buffer(bh);
	set_buffer_uptodate(bh);

	/* remove from dirty list before I/O. */
	clear_buffer_dirty(bh);

	get_bh(bh); /* for end_buffer_write_sync() */
	bh->b_end_io = end_buffer_write_sync;
	submit_bh(REQ_OP_WRITE, 0, bh);

	wait_on_buffer(bh);

	if (buffer_uptodate(bh)) {
		ocfs2_set_buffer_uptodate(ci, bh);
	} else {
		/* We don't need to remove the clustered uptodate
		 * information for this bh as it's not marked locally
		 * uptodate. */
		ret = -EIO;
		mlog_errno(ret);
	}

	ocfs2_metadata_cache_io_unlock(ci);
out:
	return ret;
}

/* Warning: even if it returns true, this does *not* guarantee that
 * the block is stored in our inode metadata cache. 
 * 
 * This can be called under lock_buffer()
 */
int ocfs2_buffer_uptodate(struct inode *inode,
			  struct buffer_head *bh)
{
	/* Doesn't matter if the bh is in our cache or not -- if it's
	 * not marked uptodate then we know it can't have correct
	 * data. */
	if (!buffer_uptodate(bh))
		return 0;

	/* OCFS2 does not allow multiple nodes to be changing the same
	 * block at the same time. */
	if (buffer_jbd(bh))
		return 1;

	/* Ok, locally the buffer is marked as up to date, now search
	 * our cache to see if we can trust that. */
	return ocfs2_buffer_cached(OCFS2_I(inode), bh);
}

void print_buffer_fields(struct buffer_head *bh)
{
	printk("b_blocknr:%llu b_count:%d\n",
	(unsigned long long)bh->b_blocknr, atomic_read(&bh->b_count));
	printk("b_this_page:%p b_data:%p b_page:%p\n",
	bh->b_this_page, bh->b_data, bh->b_page);
#if defined(CONFIG_JBD2) || defined(CONFIG_JBD2_MODULE)
	if (buffer_jbd(bh)) {
	struct journal_head *jh = bh2jh(bh);

	printk("b_jlist:%u b_frozen_data:%p b_committed_data:%p\n",
	jh->b_jlist, jh->b_frozen_data, jh->b_committed_data);
	printk(" b_transaction:%p b_next_transaction:%p "
	"b_cp_transaction:%p\n",
	jh->b_transaction, jh->b_next_transaction,
	jh->b_cp_transaction);
	printk("b_cpnext:%p b_cpprev:%p\n",
	jh->b_cpnext, jh->b_cpprev);
	}
#endif
}

/*
 * More or less lifted from ext3. I'll leave their description below:
 *
 * "For ext3 allocations, we must not reuse any blocks which are
 * allocated in the bitmap buffer's "last committed data" copy.  This
 * prevents deletes from freeing up the page for reuse until we have
 * committed the delete transaction.
 *
 * If we didn't do this, then deleting something and reallocating it as
 * data would allow the old block to be overwritten before the
 * transaction committed (because we force data to disk before commit).
 * This would lead to corruption if we crashed between overwriting the
 * data and committing the delete.
 *
 * @@@ We may want to make this allocation behaviour conditional on
 * data-writes at some point, and disable it for metadata allocations or
 * sync-data inodes."
 *
 * Note: OCFS2 already does this differently for metadata vs data
 * allocations, as those bitmaps are separate and undo access is never
 * called on a metadata group descriptor.
 */
static int ocfs2_test_bg_bit_allocatable(struct buffer_head *bg_bh,
					 int nr)
{
	struct ocfs2_group_desc *bg = (struct ocfs2_group_desc *) bg_bh->b_data;
	int ret;

	if (ocfs2_test_bit(nr, (unsigned long *)bg->bg_bitmap))
		return 0;

	if (!buffer_jbd(bg_bh))
		return 1;

	jbd_lock_bh_state(bg_bh);
	bg = (struct ocfs2_group_desc *) bh2jh(bg_bh)->b_committed_data;
	if (bg)
		ret = !ocfs2_test_bit(nr, (unsigned long *)bg->bg_bitmap);
	else
		ret = 1;
	jbd_unlock_bh_state(bg_bh);

	return ret;
}

static void __journal_remove_journal_head(struct buffer_head *bh)
{
    struct journal_head *jh = bh2jh(bh);

    J_ASSERT_JH(jh, jh->b_jcount >= 0);

    get_bh(bh);
    if (jh->b_jcount == 0) {
        if (jh->b_transaction == NULL &&
                jh->b_next_transaction == NULL &&
                jh->b_cp_transaction == NULL) {
            J_ASSERT_JH(jh, jh->b_jlist == BJ_None);
            J_ASSERT_BH(bh, buffer_jbd(bh));
            J_ASSERT_BH(bh, jh2bh(jh) == bh);
            BUFFER_TRACE(bh, "remove journal_head");
            if (jh->b_frozen_data) {
                printk(KERN_WARNING "%s: freeing "
                       "b_frozen_data\n",
                       __FUNCTION__);
                jbd_free(jh->b_frozen_data, bh->b_size);
            }
            if (jh->b_committed_data) {
                printk(KERN_WARNING "%s: freeing "
                       "b_committed_data\n",
                       __FUNCTION__);
                jbd_free(jh->b_committed_data, bh->b_size);
            }
            bh->b_private = NULL;
            jh->b_bh = NULL;	/* debug, really */
            clear_buffer_jbd(bh);
            __brelse(bh);
            journal_free_journal_head(jh);
        } else {
            BUFFER_TRACE(bh, "journal_head was locked");
        }
    }
}

/*
 * Write super block and backups doesn't need to collaborate with journal,
 * so we don't need to lock ip_io_mutex and inode doesn't need to bea passed
 * into this function.
 */
int ocfs2_write_super_or_backup(struct ocfs2_super *osb,
				struct buffer_head *bh)
{
	int ret = 0;

	mlog_entry_void();

	BUG_ON(buffer_jbd(bh));
	ocfs2_check_super_or_backup(osb->sb, bh->b_blocknr);

	if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb)) {
		ret = -EROFS;
		goto out;
	}

	lock_buffer(bh);
	set_buffer_uptodate(bh);

	/* remove from dirty list before I/O. */
	clear_buffer_dirty(bh);

	get_bh(bh); /* for end_buffer_write_sync() */
	bh->b_end_io = end_buffer_write_sync;
	submit_bh(WRITE, bh);

	wait_on_buffer(bh);

	if (!buffer_uptodate(bh)) {
		ret = -EIO;
		put_bh(bh);
	}

out:
	mlog_exit(ret);
	return ret;
}

int journal_try_to_free_buffers(journal_t *journal,
				struct page *page, gfp_t gfp_mask)
{
	struct buffer_head *head;
	struct buffer_head *bh;
	int ret = 0;

	J_ASSERT(PageLocked(page));

	head = page_buffers(page);
	bh = head;
	do {
		struct journal_head *jh;

		/*
		 * We take our own ref against the journal_head here to avoid
		 * having to add tons of locking around each instance of
		 * journal_remove_journal_head() and journal_put_journal_head().
		 */
		jh = journal_grab_journal_head(bh);
		if (!jh)
			continue;

		jbd_lock_bh_state(bh);
		__journal_try_to_free_buffer(journal, bh);
		journal_put_journal_head(jh);
		jbd_unlock_bh_state(bh);
		if (buffer_jbd(bh))
			goto busy;
	} while ((bh = bh->b_this_page) != head);

	ret = try_to_free_buffers(page);

busy:
	return ret;
}

/*
 * Write super block and backups doesn't need to collaborate with journal,
 * so we don't need to lock ip_io_mutex and ci doesn't need to bea passed
 * into this function.
 */
int ocfs2_write_super_or_backup(struct ocfs2_super *osb,
				struct buffer_head *bh)
{
	int ret = 0;
	struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;

	BUG_ON(buffer_jbd(bh));
	ocfs2_check_super_or_backup(osb->sb, bh->b_blocknr);

	if (ocfs2_is_hard_readonly(osb) || ocfs2_is_soft_readonly(osb)) {
		ret = -EROFS;
		mlog_errno(ret);
		goto out;
	}

	lock_buffer(bh);
	set_buffer_uptodate(bh);

	/* remove from dirty list before I/O. */
	clear_buffer_dirty(bh);

	get_bh(bh); /* for end_buffer_write_sync() */
	bh->b_end_io = end_buffer_write_sync;
	ocfs2_compute_meta_ecc(osb->sb, bh->b_data, &di->i_check);
	submit_bh(REQ_OP_WRITE, 0, bh);

	wait_on_buffer(bh);

	if (!buffer_uptodate(bh)) {
		ret = -EIO;
		mlog_errno(ret);
	}

out:
	return ret;
}

int ocfs2_read_blocks(struct inode *inode, u64 block, int nr,
		      struct buffer_head *bhs[], int flags,
		      int (*validate)(struct super_block *sb,
				      struct buffer_head *bh))
{
	int status = 0;
	int i, ignore_cache = 0;
	struct buffer_head *bh;

	mlog_entry("(inode=%p, block=(%llu), nr=(%d), flags=%d)\n",
		   inode, (unsigned long long)block, nr, flags);

	BUG_ON(!inode);
	BUG_ON((flags & OCFS2_BH_READAHEAD) &&
	       (flags & OCFS2_BH_IGNORE_CACHE));

	if (bhs == NULL) {
		status = -EINVAL;
		mlog_errno(status);
		goto bail;
	}

	if (nr < 0) {
		mlog(ML_ERROR, "asked to read %d blocks!\n", nr);
		status = -EINVAL;
		mlog_errno(status);
		goto bail;
	}

	if (nr == 0) {
		mlog(ML_BH_IO, "No buffers will be read!\n");
		status = 0;
		goto bail;
	}

	mutex_lock(&OCFS2_I(inode)->ip_io_mutex);
	for (i = 0 ; i < nr ; i++) {
		if (bhs[i] == NULL) {
			bhs[i] = sb_getblk(inode->i_sb, block++);
			if (bhs[i] == NULL) {
				mutex_unlock(&OCFS2_I(inode)->ip_io_mutex);
				status = -EIO;
				mlog_errno(status);
				goto bail;
			}
		}
		bh = bhs[i];
		ignore_cache = (flags & OCFS2_BH_IGNORE_CACHE);

		/* There are three read-ahead cases here which we need to
		 * be concerned with. All three assume a buffer has
		 * previously been submitted with OCFS2_BH_READAHEAD
		 * and it hasn't yet completed I/O.
		 *
		 * 1) The current request is sync to disk. This rarely
		 *    happens these days, and never when performance
		 *    matters - the code can just wait on the buffer
		 *    lock and re-submit.
		 *
		 * 2) The current request is cached, but not
		 *    readahead. ocfs2_buffer_uptodate() will return
		 *    false anyway, so we'll wind up waiting on the
		 *    buffer lock to do I/O. We re-check the request
		 *    with after getting the lock to avoid a re-submit.
		 *
		 * 3) The current request is readahead (and so must
		 *    also be a caching one). We short circuit if the
		 *    buffer is locked (under I/O) and if it's in the
		 *    uptodate cache. The re-check from #2 catches the
		 *    case that the previous read-ahead completes just
		 *    before our is-it-in-flight check.
		 */

		if (!ignore_cache && !ocfs2_buffer_uptodate(inode, bh)) {
			mlog(ML_UPTODATE,
			     "bh (%llu), inode %llu not uptodate\n",
			     (unsigned long long)bh->b_blocknr,
			     (unsigned long long)OCFS2_I(inode)->ip_blkno);
			/* We're using ignore_cache here to say
			 * "go to disk" */
			ignore_cache = 1;
		}

		if (buffer_jbd(bh)) {
			if (ignore_cache)
				mlog(ML_BH_IO, "trying to sync read a jbd "
					       "managed bh (blocknr = %llu)\n",
				     (unsigned long long)bh->b_blocknr);
			continue;
		}

		if (ignore_cache) {
			if (buffer_dirty(bh)) {
				/* This should probably be a BUG, or
				 * at least return an error. */
				mlog(ML_BH_IO, "asking me to sync read a dirty "
					       "buffer! (blocknr = %llu)\n",
				     (unsigned long long)bh->b_blocknr);
				continue;
			}
//.........这里部分代码省略.........

int ocfs2_read_blocks_sync(struct ocfs2_super *osb, u64 block,
			   unsigned int nr, struct buffer_head *bhs[])
{
	int status = 0;
	unsigned int i;
	struct buffer_head *bh;

	if (!nr) {
		mlog(ML_BH_IO, "No buffers will be read!\n");
		goto bail;
	}

	for (i = 0 ; i < nr ; i++) {
		if (bhs[i] == NULL) {
			bhs[i] = sb_getblk(osb->sb, block++);
			if (bhs[i] == NULL) {
				status = -EIO;
				mlog_errno(status);
				goto bail;
			}
		}
		bh = bhs[i];

		if (buffer_jbd(bh)) {
			mlog(ML_BH_IO,
			     "trying to sync read a jbd "
			     "managed bh (blocknr = %llu), skipping\n",
			     (unsigned long long)bh->b_blocknr);
			continue;
		}

		if (buffer_dirty(bh)) {
			/* This should probably be a BUG, or
			 * at least return an error. */
			mlog(ML_ERROR,
			     "trying to sync read a dirty "
			     "buffer! (blocknr = %llu), skipping\n",
			     (unsigned long long)bh->b_blocknr);
			continue;
		}

		lock_buffer(bh);
		if (buffer_jbd(bh)) {
			mlog(ML_ERROR,
			     "block %llu had the JBD bit set "
			     "while I was in lock_buffer!",
			     (unsigned long long)bh->b_blocknr);
			BUG();
		}

		clear_buffer_uptodate(bh);
		get_bh(bh); /* for end_buffer_read_sync() */
		bh->b_end_io = end_buffer_read_sync;
		submit_bh(READ, bh);
	}

	for (i = nr; i > 0; i--) {
		bh = bhs[i - 1];

		/* No need to wait on the buffer if it's managed by JBD. */
		if (!buffer_jbd(bh))
			wait_on_buffer(bh);

		if (!buffer_uptodate(bh)) {
			/* Status won't be cleared from here on out,
			 * so we can safely record this and loop back
			 * to cleanup the other buffers. */
			status = -EIO;
			put_bh(bh);
			bhs[i - 1] = NULL;
		}
	}

bail:
	return status;
}

//.........这里部分代码省略.........
	/*
	 * Wait for all previously submitted IO to complete.
	 */
	spin_lock(&journal->j_list_lock);
	while (commit_transaction->t_locked_list) {
		struct buffer_head *bh;

		jh = commit_transaction->t_locked_list->b_tprev;
		bh = jh2bh(jh);
		get_bh(bh);
		if (buffer_locked(bh)) {
			spin_unlock(&journal->j_list_lock);
			wait_on_buffer(bh);
			spin_lock(&journal->j_list_lock);
		}
		if (unlikely(!buffer_uptodate(bh))) {
			if (!trylock_page(bh->b_page)) {
				spin_unlock(&journal->j_list_lock);
				lock_page(bh->b_page);
				spin_lock(&journal->j_list_lock);
			}
			if (bh->b_page->mapping)
				set_bit(AS_EIO, &bh->b_page->mapping->flags);

			unlock_page(bh->b_page);
			SetPageError(bh->b_page);
			err = -EIO;
		}
		if (!inverted_lock(journal, bh)) {
			put_bh(bh);
			spin_lock(&journal->j_list_lock);
			continue;
		}
		if (buffer_jbd(bh) && bh2jh(bh) == jh &&
		    jh->b_transaction == commit_transaction &&
		    jh->b_jlist == BJ_Locked)
			__journal_unfile_buffer(jh);
		jbd_unlock_bh_state(bh);
		release_data_buffer(bh);
		cond_resched_lock(&journal->j_list_lock);
	}
	spin_unlock(&journal->j_list_lock);

	if (err) {
		char b[BDEVNAME_SIZE];

		printk(KERN_WARNING
			"JBD: Detected IO errors while flushing file data "
			"on %s\n", bdevname(journal->j_fs_dev, b));
		if (journal->j_flags & JFS_ABORT_ON_SYNCDATA_ERR)
			journal_abort(journal, err);
		err = 0;
	}

	blk_start_plug(&plug);

	journal_write_revoke_records(journal, commit_transaction, WRITE_SYNC);

	/*
	 * If we found any dirty or locked buffers, then we should have
	 * looped back up to the write_out_data label.  If there weren't
	 * any then journal_clean_data_list should have wiped the list
	 * clean by now, so check that it is in fact empty.
	 */
	J_ASSERT (commit_transaction->t_sync_datalist == NULL);

/*
 *  Submit all the data buffers to disk
 */
static int journal_submit_data_buffers(journal_t *journal,
				       transaction_t *commit_transaction,
				       int write_op)
{
	struct journal_head *jh;
	struct buffer_head *bh;
	int locked;
	int bufs = 0;
	struct buffer_head **wbuf = journal->j_wbuf;
	int err = 0;

	/*
	 * Whenever we unlock the journal and sleep, things can get added
	 * onto ->t_sync_datalist, so we have to keep looping back to
	 * write_out_data until we *know* that the list is empty.
	 *
	 * Cleanup any flushed data buffers from the data list.  Even in
	 * abort mode, we want to flush this out as soon as possible.
	 */
write_out_data:
	cond_resched();
	spin_lock(&journal->j_list_lock);

	while (commit_transaction->t_sync_datalist) {
		jh = commit_transaction->t_sync_datalist;
		bh = jh2bh(jh);
		locked = 0;

		/* Get reference just to make sure buffer does not disappear
		 * when we are forced to drop various locks */
		get_bh(bh);
		/* If the buffer is dirty, we need to submit IO and hence
		 * we need the buffer lock. We try to lock the buffer without
		 * blocking. If we fail, we need to drop j_list_lock and do
		 * blocking lock_buffer().
		 */
		if (buffer_dirty(bh)) {
			if (!trylock_buffer(bh)) {
				BUFFER_TRACE(bh, "needs blocking lock");
				spin_unlock(&journal->j_list_lock);
				trace_jbd_do_submit_data(journal,
						     commit_transaction);
				/* Write out all data to prevent deadlocks */
				journal_do_submit_data(wbuf, bufs, write_op);
				bufs = 0;
				lock_buffer(bh);
				spin_lock(&journal->j_list_lock);
			}
			locked = 1;
		}
		/* We have to get bh_state lock. Again out of order, sigh. */
		if (!inverted_lock(journal, bh)) {
			jbd_lock_bh_state(bh);
			spin_lock(&journal->j_list_lock);
		}
		/* Someone already cleaned up the buffer? */
		if (!buffer_jbd(bh) || bh2jh(bh) != jh
			|| jh->b_transaction != commit_transaction
			|| jh->b_jlist != BJ_SyncData) {
			jbd_unlock_bh_state(bh);
			if (locked)
				unlock_buffer(bh);
			BUFFER_TRACE(bh, "already cleaned up");
			release_data_buffer(bh);
			continue;
		}
		if (locked && test_clear_buffer_dirty(bh)) {
			BUFFER_TRACE(bh, "needs writeout, adding to array");
			wbuf[bufs++] = bh;
			__journal_file_buffer(jh, commit_transaction,
						BJ_Locked);
			jbd_unlock_bh_state(bh);
			if (bufs == journal->j_wbufsize) {
				spin_unlock(&journal->j_list_lock);
				trace_jbd_do_submit_data(journal,
						     commit_transaction);
				journal_do_submit_data(wbuf, bufs, write_op);
				bufs = 0;
				goto write_out_data;
			}
		} else if (!locked && buffer_locked(bh)) {
			__journal_file_buffer(jh, commit_transaction,
						BJ_Locked);
			jbd_unlock_bh_state(bh);
			put_bh(bh);
		} else {
			BUFFER_TRACE(bh, "writeout complete: unfile");
			if (unlikely(!buffer_uptodate(bh)))
				err = -EIO;
			__journal_unfile_buffer(jh);
			jbd_unlock_bh_state(bh);
			if (locked)
				unlock_buffer(bh);
			release_data_buffer(bh);
		}

		if (need_resched() || spin_needbreak(&journal->j_list_lock)) {
//.........这里部分代码省略.........

static int journal_unmap_buffer(journal_t *journal, struct buffer_head *bh)
{
	transaction_t *transaction;
	struct journal_head *jh;
	int may_free = 1;
	int ret;

	BUFFER_TRACE(bh, "entry");

	/*
	 * It is safe to proceed here without the j_list_lock because the
	 * buffers cannot be stolen by try_to_free_buffers as long as we are
	 * holding the page lock. --sct
	 */

	if (!buffer_jbd(bh))
		goto zap_buffer_unlocked;

	spin_lock(&journal->j_state_lock);
	jbd_lock_bh_state(bh);
	spin_lock(&journal->j_list_lock);

	jh = journal_grab_journal_head(bh);
	if (!jh)
		goto zap_buffer_no_jh;

	/*
	 * We cannot remove the buffer from checkpoint lists until the
	 * transaction adding inode to orphan list (let's call it T)
	 * is committed.  Otherwise if the transaction changing the
	 * buffer would be cleaned from the journal before T is
	 * committed, a crash will cause that the correct contents of
	 * the buffer will be lost.  On the other hand we have to
	 * clear the buffer dirty bit at latest at the moment when the
	 * transaction marking the buffer as freed in the filesystem
	 * structures is committed because from that moment on the
	 * buffer can be reallocated and used by a different page.
	 * Since the block hasn't been freed yet but the inode has
	 * already been added to orphan list, it is safe for us to add
	 * the buffer to BJ_Forget list of the newest transaction.
	 */
	transaction = jh->b_transaction;
	if (transaction == NULL) {
		/* First case: not on any transaction.  If it
		 * has no checkpoint link, then we can zap it:
		 * it's a writeback-mode buffer so we don't care
		 * if it hits disk safely. */
		if (!jh->b_cp_transaction) {
			JBUFFER_TRACE(jh, "not on any transaction: zap");
			goto zap_buffer;
		}

		if (!buffer_dirty(bh)) {
			/* bdflush has written it.  We can drop it now */
			goto zap_buffer;
		}

		/* OK, it must be in the journal but still not
		 * written fully to disk: it's metadata or
		 * journaled data... */

		if (journal->j_running_transaction) {
			/* ... and once the current transaction has
			 * committed, the buffer won't be needed any
			 * longer. */
			JBUFFER_TRACE(jh, "checkpointed: add to BJ_Forget");
			ret = __dispose_buffer(jh,
					journal->j_running_transaction);
			journal_put_journal_head(jh);
			spin_unlock(&journal->j_list_lock);
			jbd_unlock_bh_state(bh);
			spin_unlock(&journal->j_state_lock);
			return ret;
		} else {
			/* There is no currently-running transaction. So the
			 * orphan record which we wrote for this file must have
			 * passed into commit.  We must attach this buffer to
			 * the committing transaction, if it exists. */
			if (journal->j_committing_transaction) {
				JBUFFER_TRACE(jh, "give to committing trans");
				ret = __dispose_buffer(jh,
					journal->j_committing_transaction);
				journal_put_journal_head(jh);
				spin_unlock(&journal->j_list_lock);
				jbd_unlock_bh_state(bh);
				spin_unlock(&journal->j_state_lock);
				return ret;
			} else {
				/* The orphan record's transaction has
				 * committed.  We can cleanse this buffer */
				clear_buffer_jbddirty(bh);
				goto zap_buffer;
			}
		}
	} else if (transaction == journal->j_committing_transaction) {
		JBUFFER_TRACE(jh, "on committing transaction");
		if (jh->b_jlist == BJ_Locked) {
			/*
			 * The buffer is on the committing transaction's locked
			 * list.  We have the buffer locked, so I/O has
//.........这里部分代码省略.........