diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 81ce6bf2dacfecb5dd2b44af66d4f98af34c66e1..0e16c3118ea14c71750db48024b2777a70fa5246 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -3819,25 +3819,31 @@ static ssize_t ext4_direct_IO_write(struct kiocb *iocb, struct iov_iter *iter)
 
 static ssize_t ext4_direct_IO_read(struct kiocb *iocb, struct iov_iter *iter)
 {
-	struct address_space *mapping = iocb->ki_filp->f_mapping;
-	struct inode *inode = mapping->host;
-	size_t count = iov_iter_count(iter);
+	int unlocked = 0;
+	struct inode *inode = iocb->ki_filp->f_mapping->host;
 	ssize_t ret;
 
-	/*
-	 * Shared inode_lock is enough for us - it protects against concurrent
-	 * writes & truncates and since we take care of writing back page cache,
-	 * we are protected against page writeback as well.
-	 */
-	inode_lock_shared(inode);
-	ret = filemap_write_and_wait_range(mapping, iocb->ki_pos,
-					   iocb->ki_pos + count - 1);
-	if (ret)
-		goto out_unlock;
+	if (ext4_should_dioread_nolock(inode)) {
+		/*
+		 * Nolock dioread optimization may be dynamically disabled
+		 * via ext4_inode_block_unlocked_dio(). Check inode's state
+		 * while holding extra i_dio_count ref.
+		 */
+		inode_dio_begin(inode);
+		smp_mb();
+		if (unlikely(ext4_test_inode_state(inode,
+						    EXT4_STATE_DIOREAD_LOCK)))
+			inode_dio_end(inode);
+		else
+			unlocked = 1;
+	}
 	ret = __blockdev_direct_IO(iocb, inode, inode->i_sb->s_bdev,
-				   iter, ext4_dio_get_block, NULL, NULL, 0);
-out_unlock:
-	inode_unlock_shared(inode);
+				   iter, ext4_dio_get_block,
+				   NULL, NULL,
+				   unlocked ? 0 : DIO_LOCKING);
+
+	if (unlocked)
+		inode_dio_end(inode);
 	return ret;
 }