diff --git a/drivers/block/Kconfig b/drivers/block/Kconfig
index 26760895f9..209fb5ebe6 100644
--- a/drivers/block/Kconfig
+++ b/drivers/block/Kconfig
@@ -22,6 +22,15 @@ config SPL_BLK
 	  be partitioned into several areas, called 'partitions' in U-Boot.
 	  A filesystem can be placed in each partition.
 
+config SPL_BLK_READ_PREPARE
+	bool "Support block devices prepare to read data in SPL"
+	depends on SPL_BLK
+	help
+	  Enable support for block devices to prefetch data. MMC and mtd_blk
+	  devices can be attached to block devices. It is applied to prefetch
+	  data in the background and the device run some other process in the
+	  same time.
+
 config BLOCK_CACHE
 	bool "Use block device cache"
 	default n
diff --git a/drivers/block/blk-uclass.c b/drivers/block/blk-uclass.c
index 0d3574ffb7..3d8adc7094 100644
--- a/drivers/block/blk-uclass.c
+++ b/drivers/block/blk-uclass.c
@@ -454,6 +454,20 @@ unsigned long blk_dread(struct blk_desc *block_dev, lbaint_t start,
 	return blks_read;
 }
 
+#ifdef CONFIG_SPL_BLK_READ_PREPARE
+unsigned long blk_dread_prepare(struct blk_desc *block_dev, lbaint_t start,
+				lbaint_t blkcnt, void *buffer)
+{
+	struct udevice *dev = block_dev->bdev;
+	const struct blk_ops *ops = blk_get_ops(dev);
+
+	if (!ops->read)
+		return -ENOSYS;
+
+	return ops->read_prepare(dev, start, blkcnt, buffer);
+}
+#endif
+
 unsigned long blk_dwrite(struct blk_desc *block_dev, lbaint_t start,
 			 lbaint_t blkcnt, const void *buffer)
 {
diff --git a/drivers/mmc/dw_mmc.c b/drivers/mmc/dw_mmc.c
index 6d7cfdfd45..7e8fde320e 100644
--- a/drivers/mmc/dw_mmc.c
+++ b/drivers/mmc/dw_mmc.c
@@ -6,8 +6,8 @@
  * SPDX-License-Identifier:	GPL-2.0+
  */
 
-#include <bouncebuf.h>
 #include <common.h>
+#include <bouncebuf.h>
 #include <errno.h>
 #include <malloc.h>
 #include <memalign.h>
@@ -437,6 +437,135 @@ static int dwmci_send_cmd(struct mmc *mmc, struct mmc_cmd *cmd,
 	return ret;
 }
 
+#ifdef CONFIG_SPL_BLK_READ_PREPARE
+#ifdef CONFIG_DM_MMC
+static int dwmci_send_cmd_prepare(struct udevice *dev, struct mmc_cmd *cmd,
+				  struct mmc_data *data)
+{
+	struct mmc *mmc = mmc_get_mmc_dev(dev);
+#else
+static int dwmci_send_cmd_prepare(struct mmc *mmc, struct mmc_cmd *cmd,
+				  struct mmc_data *data)
+{
+#endif
+	struct dwmci_host *host = mmc->priv;
+	struct dwmci_idmac *cur_idmac;
+	int ret = 0, flags = 0, i;
+	unsigned int timeout = 500;
+	u32 retry = 100000;
+	u32 mask;
+	ulong start = get_timer(0);
+	struct bounce_buffer bbstate;
+
+	cur_idmac = malloc(ROUND(DIV_ROUND_UP(data->blocks, 8) *
+			   sizeof(struct dwmci_idmac),
+			   ARCH_DMA_MINALIGN) + ARCH_DMA_MINALIGN - 1);
+	if (!cur_idmac)
+		return -ENODATA;
+
+	while (dwmci_readl(host, DWMCI_STATUS) & DWMCI_BUSY) {
+		if (get_timer(start) > timeout) {
+			debug("%s: Timeout on data busy\n", __func__);
+			return -ETIMEDOUT;
+		}
+	}
+
+	dwmci_writel(host, DWMCI_RINTSTS, DWMCI_INTMSK_ALL);
+
+	if (data) {
+		if (host->fifo_mode) {
+			dwmci_writel(host, DWMCI_BLKSIZ, data->blocksize);
+			dwmci_writel(host, DWMCI_BYTCNT,
+				     data->blocksize * data->blocks);
+			dwmci_wait_reset(host, DWMCI_CTRL_FIFO_RESET);
+		} else {
+			if (data->flags == MMC_DATA_READ) {
+				bounce_buffer_start(&bbstate, (void *)data->dest,
+						    data->blocksize *
+						    data->blocks, GEN_BB_WRITE);
+			} else {
+				bounce_buffer_start(&bbstate, (void *)data->src,
+						    data->blocksize *
+						    data->blocks, GEN_BB_READ);
+			}
+			dwmci_prepare_data(host, data, cur_idmac,
+					   bbstate.bounce_buffer);
+		}
+	}
+
+	dwmci_writel(host, DWMCI_CMDARG, cmd->cmdarg);
+
+	if (data)
+		flags = dwmci_set_transfer_mode(host, data);
+
+	if ((cmd->resp_type & MMC_RSP_136) && (cmd->resp_type & MMC_RSP_BUSY))
+		return -1;
+
+	if (cmd->cmdidx == MMC_CMD_STOP_TRANSMISSION)
+		flags |= DWMCI_CMD_ABORT_STOP;
+	else
+		flags |= DWMCI_CMD_PRV_DAT_WAIT;
+
+	if (cmd->resp_type & MMC_RSP_PRESENT) {
+		flags |= DWMCI_CMD_RESP_EXP;
+		if (cmd->resp_type & MMC_RSP_136)
+			flags |= DWMCI_CMD_RESP_LENGTH;
+	}
+
+	if (cmd->resp_type & MMC_RSP_CRC)
+		flags |= DWMCI_CMD_CHECK_CRC;
+
+	flags |= (cmd->cmdidx | DWMCI_CMD_START | DWMCI_CMD_USE_HOLD_REG);
+
+	debug("Sending CMD%d\n", cmd->cmdidx);
+
+	dwmci_writel(host, DWMCI_CMD, flags);
+
+	for (i = 0; i < retry; i++) {
+		mask = dwmci_readl(host, DWMCI_RINTSTS);
+		if (mask & DWMCI_INTMSK_CDONE) {
+			if (!data)
+				dwmci_writel(host, DWMCI_RINTSTS, mask);
+			break;
+		}
+	}
+
+	if (i == retry) {
+		debug("%s: Timeout.\n", __func__);
+		return -ETIMEDOUT;
+	}
+
+	if (mask & DWMCI_INTMSK_RTO) {
+		/*
+		 * Timeout here is not necessarily fatal. (e)MMC cards
+		 * will splat here when they receive CMD55 as they do
+		 * not support this command and that is exactly the way
+		 * to tell them apart from SD cards. Thus, this output
+		 * below shall be debug(). eMMC cards also do not favor
+		 * CMD8, please keep that in mind.
+		 */
+		debug("%s: Response Timeout.\n", __func__);
+		return -ETIMEDOUT;
+	} else if (mask & DWMCI_INTMSK_RE) {
+		debug("%s: Response Error.\n", __func__);
+		return -EIO;
+	}
+
+	if (cmd->resp_type & MMC_RSP_PRESENT) {
+		if (cmd->resp_type & MMC_RSP_136) {
+			cmd->response[0] = dwmci_readl(host, DWMCI_RESP3);
+			cmd->response[1] = dwmci_readl(host, DWMCI_RESP2);
+			cmd->response[2] = dwmci_readl(host, DWMCI_RESP1);
+			cmd->response[3] = dwmci_readl(host, DWMCI_RESP0);
+		} else {
+			cmd->response[0] = dwmci_readl(host, DWMCI_RESP0);
+		}
+	}
+
+	return ret;
+}
+#endif
+
 static int dwmci_setup_bus(struct dwmci_host *host, u32 freq)
 {
 	u32 div, status;
@@ -669,6 +798,9 @@ int dwmci_probe(struct udevice *dev)
 const struct dm_mmc_ops dm_dwmci_ops = {
 	.card_busy	= dwmci_card_busy,
 	.send_cmd	= dwmci_send_cmd,
+#ifdef CONFIG_SPL_BLK_READ_PREPARE
+	.send_cmd_prepare = dwmci_send_cmd_prepare,
+#endif
 	.set_ios	= dwmci_set_ios,
 	.get_cd         = dwmci_get_cd,
 	.execute_tuning	= dwmci_execute_tuning,
diff --git a/drivers/mmc/mmc-uclass.c b/drivers/mmc/mmc-uclass.c
index 7feb225186..06c3b8fc14 100644
--- a/drivers/mmc/mmc-uclass.c
+++ b/drivers/mmc/mmc-uclass.c
@@ -32,11 +32,37 @@ int dm_mmc_send_cmd(struct udevice *dev, struct mmc_cmd *cmd,
 	return ret;
 }
 
+#ifdef CONFIG_SPL_BLK_READ_PREPARE
+int dm_mmc_send_cmd_prepare(struct udevice *dev, struct mmc_cmd *cmd,
+			    struct mmc_data *data)
+{
+	struct mmc *mmc = mmc_get_mmc_dev(dev);
+	struct dm_mmc_ops *ops = mmc_get_ops(dev);
+	int ret;
+
+	mmmc_trace_before_send(mmc, cmd);
+	if (ops->send_cmd_prepare)
+		ret = ops->send_cmd_prepare(dev, cmd, data);
+	else
+		ret = -ENOSYS;
+	mmmc_trace_after_send(mmc, cmd, ret);
+
+	return ret;
+}
+#endif
+
 int mmc_send_cmd(struct mmc *mmc, struct mmc_cmd *cmd, struct mmc_data *data)
 {
 	return dm_mmc_send_cmd(mmc->dev, cmd, data);
 }
 
+#ifdef CONFIG_SPL_BLK_READ_PREPARE
+int mmc_send_cmd_prepare(struct mmc *mmc, struct mmc_cmd *cmd, struct mmc_data *data)
+{
+	return dm_mmc_send_cmd_prepare(mmc->dev, cmd, data);
+}
+#endif
+
 bool mmc_card_busy(struct mmc *mmc)
 {
 	struct dm_mmc_ops *ops = mmc_get_ops(mmc->dev);
@@ -289,6 +315,9 @@ static int mmc_blk_probe(struct udevice *dev)
 
 static const struct blk_ops mmc_blk_ops = {
 	.read	= mmc_bread,
+#ifdef CONFIG_SPL_BLK_READ_PREPARE
+	.read_prepare	= mmc_bread_prepare,
+#endif
 #ifndef CONFIG_SPL_BUILD
 	.write	= mmc_bwrite,
 	.erase	= mmc_berase,
diff --git a/drivers/mmc/mmc.c b/drivers/mmc/mmc.c
index 14141601a3..941005a19e 100644
--- a/drivers/mmc/mmc.c
+++ b/drivers/mmc/mmc.c
@@ -263,6 +263,37 @@ static int mmc_read_blocks(struct mmc *mmc, void *dst, lbaint_t start,
 	return blkcnt;
 }
 
+#ifdef CONFIG_SPL_BLK_READ_PREPARE
+static int mmc_read_blocks_prepare(struct mmc *mmc, void *dst, lbaint_t start,
+				   lbaint_t blkcnt)
+{
+	struct mmc_cmd cmd;
+	struct mmc_data data;
+
+	if (blkcnt > 1)
+		cmd.cmdidx = MMC_CMD_READ_MULTIPLE_BLOCK;
+	else
+		cmd.cmdidx = MMC_CMD_READ_SINGLE_BLOCK;
+
+	if (mmc->high_capacity)
+		cmd.cmdarg = start;
+	else
+		cmd.cmdarg = start * mmc->read_bl_len;
+
+	cmd.resp_type = MMC_RSP_R1;
+
+	data.dest = dst;
+	data.blocks = blkcnt;
+	data.blocksize = mmc->read_bl_len;
+	data.flags = MMC_DATA_READ;
+
+	if (mmc_send_cmd_prepare(mmc, &cmd, &data))
+		return 0;
+
+	return blkcnt;
+}
+#endif
+
 #if CONFIG_IS_ENABLED(BLK)
 ulong mmc_bread(struct udevice *dev, lbaint_t start, lbaint_t blkcnt, void *dst)
 #else
@@ -339,6 +370,77 @@ re_init_retry:
 	return blkcnt;
 }
 
+#ifdef CONFIG_SPL_BLK_READ_PREPARE
+#if CONFIG_IS_ENABLED(BLK)
+ulong mmc_bread_prepare(struct udevice *dev, lbaint_t start, lbaint_t blkcnt, void *dst)
+#else
+ulong mmc_bread_prepare(struct blk_desc *block_dev, lbaint_t start, lbaint_t blkcnt,
+			void *dst)
+#endif
+{
+#if CONFIG_IS_ENABLED(BLK)
+	struct blk_desc *block_dev = dev_get_uclass_platdata(dev);
+#endif
+	int dev_num = block_dev->devnum;
+	int timeout = 0;
+	int err;
+
+	if (blkcnt == 0)
+		return 0;
+
+	struct mmc *mmc = find_mmc_device(dev_num);
+
+	if (!mmc)
+		return 0;
+
+	if (CONFIG_IS_ENABLED(MMC_TINY))
+		err = mmc_switch_part(mmc, block_dev->hwpart);
+	else
+		err = blk_dselect_hwpart(block_dev, block_dev->hwpart);
+
+	if (err < 0)
+		return 0;
+
+	if ((start + blkcnt) > block_dev->lba) {
+#if !defined(CONFIG_SPL_BUILD) || defined(CONFIG_SPL_LIBCOMMON_SUPPORT)
+		printf("MMC: block number 0x" LBAF " exceeds max(0x" LBAF ")\n",
+		       start + blkcnt, block_dev->lba);
+#endif
+		return 0;
+	}
+
+	if (mmc_set_blocklen(mmc, mmc->read_bl_len)) {
+		debug("%s: Failed to set blocklen\n", __func__);
+		return 0;
+	}
+
+	if (mmc_read_blocks_prepare(mmc, dst, start, blkcnt) != blkcnt) {
+		debug("%s: Failed to read blocks\n", __func__);
+re_init_retry:
+		timeout++;
+		/*
+		 * Try re-init seven times.
+		 */
+		if (timeout > 7) {
+			printf("Re-init retry timeout\n");
+			return 0;
+		}
+
+		mmc->has_init = 0;
+		if (mmc_init(mmc))
+			return 0;
+
+		if (mmc_read_blocks_prepare(mmc, dst, start, blkcnt) != blkcnt) {
+			printf("%s: Re-init mmc_read_blocks_prepare error\n",
+			       __func__);
+			goto re_init_retry;
+		}
+	}
+
+	return blkcnt;
+}
+#endif
+
 void mmc_set_clock(struct mmc *mmc, uint clock)
 {
 	if (clock > mmc->cfg->f_max)
diff --git a/drivers/mmc/mmc_private.h b/drivers/mmc/mmc_private.h
index 1290eed590..f76e6f8490 100644
--- a/drivers/mmc/mmc_private.h
+++ b/drivers/mmc/mmc_private.h
@@ -14,6 +14,10 @@
 
 extern int mmc_send_cmd(struct mmc *mmc, struct mmc_cmd *cmd,
 			struct mmc_data *data);
+#ifdef CONFIG_SPL_BLK_READ_PREPARE
+int mmc_send_cmd_prepare(struct mmc *mmc, struct mmc_cmd *cmd,
+			 struct mmc_data *data);
+#endif
 extern int mmc_send_status(struct mmc *mmc, int timeout);
 extern int mmc_set_blocklen(struct mmc *mmc, int len);
 #ifdef CONFIG_FSL_ESDHC_ADAPTER_IDENT
@@ -23,9 +27,17 @@ void mmc_adapter_card_type_ident(void);
 #if CONFIG_IS_ENABLED(BLK)
 ulong mmc_bread(struct udevice *dev, lbaint_t start, lbaint_t blkcnt,
 		void *dst);
+#ifdef CONFIG_SPL_BLK_READ_PREPARE
+ulong mmc_bread_prepare(struct udevice *dev, lbaint_t start, lbaint_t blkcnt,
+			void *dst);
+#endif
 #else
 ulong mmc_bread(struct blk_desc *block_dev, lbaint_t start, lbaint_t blkcnt,
 		void *dst);
+#ifdef CONFIG_SPL_BLK_READ_PREPARE
+ulong mmc_bread_prepare(struct blk_desc *block_dev, lbaint_t start, lbaint_t blkcnt,
+			void *dst);
+#endif
 #endif
 
 #if !(defined(CONFIG_SPL_BUILD) && !defined(CONFIG_SPL_SAVEENV))
diff --git a/include/blk.h b/include/blk.h
index 215929c945..d4a259335c 100644
--- a/include/blk.h
+++ b/include/blk.h
@@ -224,6 +224,20 @@ struct blk_ops {
 	unsigned long (*read)(struct udevice *dev, lbaint_t start,
 			      lbaint_t blkcnt, void *buffer);
 
+	/**
+	 * read_prepare() - read from a block device
+	 *
+	 * @dev:	Device to read from
+	 * @start:	Start block number to read (0=first)
+	 * @blkcnt:	Number of blocks to read
+	 * @buffer:	Destination buffer for data read
+	 * @return number of blocks read, or -ve error number (see the
+	 * IS_ERR_VALUE() macro
+	 */
+#ifdef CONFIG_SPL_BLK_READ_PREPARE
+	unsigned long (*read_prepare)(struct udevice *dev, lbaint_t start,
+				      lbaint_t blkcnt, void *buffer);
+#endif
 	/**
 	 * write() - write to a block device
 	 *
@@ -279,6 +293,10 @@ struct blk_ops {
  */
 unsigned long blk_dread(struct blk_desc *block_dev, lbaint_t start,
 			lbaint_t blkcnt, void *buffer);
+#ifdef CONFIG_SPL_BLK_READ_PREPARE
+unsigned long blk_dread_prepare(struct blk_desc *block_dev, lbaint_t start,
+				lbaint_t blkcnt, void *buffer);
+#endif
 unsigned long blk_dwrite(struct blk_desc *block_dev, lbaint_t start,
 			 lbaint_t blkcnt, const void *buffer);
 unsigned long blk_derase(struct blk_desc *block_dev, lbaint_t start,
diff --git a/include/mmc.h b/include/mmc.h
index 8bf4761457..0d23e40dd5 100644
--- a/include/mmc.h
+++ b/include/mmc.h
@@ -416,6 +416,18 @@ struct dm_mmc_ops {
 	int (*send_cmd)(struct udevice *dev, struct mmc_cmd *cmd,
 			struct mmc_data *data);
 
+	/**
+	 * send_cmd_prepare() - Send a command to the MMC device
+	 *
+	 * @dev:	Device to receive the command
+	 * @cmd:	Command to send
+	 * @data:	Additional data to send/receive
+	 * @return 0 if OK, -ve on error
+	 */
+#ifdef CONFIG_SPL_BLK_READ_PREPARE
+	int (*send_cmd_prepare)(struct udevice *dev, struct mmc_cmd *cmd,
+				struct mmc_data *data);
+#endif
 	/**
 	 * card_busy() - Query the card device status
 	 *