23 files changed, 450 insertions, 117 deletions
diff --git a/arch/arm/dts/dra7-evm.dts b/arch/arm/dts/dra7-evm.dts
index 08ef04e..429b9ed 100644
--- a/arch/arm/dts/dra7-evm.dts
+++ b/arch/arm/dts/dra7-evm.dts
@@ -491,15 +491,13 @@
 	pinctrl-names = "default";
 	pinctrl-0 = <&qspi1_pins>;
 
-	spi-max-frequency = <48000000>;
+	spi-max-frequency = <64000000>;
 	m25p80@0 {
 		compatible = "s25fl256s1","spi-flash";
-		spi-max-frequency = <48000000>;
+		spi-max-frequency = <64000000>;
 		reg = <0>;
 		spi-tx-bus-width = <1>;
 		spi-rx-bus-width = <4>;
-		spi-cpol;
-		spi-cpha;
 		#address-cells = <1>;
 		#size-cells = <1>;
 
diff --git a/arch/arm/dts/dra72-evm.dts b/arch/arm/dts/dra72-evm.dts
index 205103e..ced2f11 100644
--- a/arch/arm/dts/dra72-evm.dts
+++ b/arch/arm/dts/dra72-evm.dts
@@ -603,15 +603,13 @@
 	pinctrl-names = "default";
 	pinctrl-0 = <&qspi1_pins>;
 
-	spi-max-frequency = <48000000>;
+	spi-max-frequency = <64000000>;
 	m25p80@0 {
 		compatible = "s25fl256s1","spi-flash";
-		spi-max-frequency = <48000000>;
+		spi-max-frequency = <64000000>;
 		reg = <0>;
 		spi-tx-bus-width = <1>;
 		spi-rx-bus-width = <4>;
-		spi-cpol;
-		spi-cpha;
 		#address-cells = <1>;
 		#size-cells = <1>;
 
diff --git a/arch/arm/dts/k2e-evm.dts b/arch/arm/dts/k2e-evm.dts
index 50c83c2..e2c3fb4 100644
--- a/arch/arm/dts/k2e-evm.dts
+++ b/arch/arm/dts/k2e-evm.dts
@@ -119,10 +119,11 @@
 };
 
 &spi0 {
+	status = "okay";
 	nor_flash: n25q128a11@0 {
 		#address-cells = <1>;
 		#size-cells = <1>;
-		compatible = "Micron,n25q128a11";
+		compatible = "Micron,n25q128a11", "spi-flash";
 		spi-max-frequency = <54000000>;
 		m25p,fast-read;
 		reg = <0>;
diff --git a/arch/arm/dts/k2g-evm.dts b/arch/arm/dts/k2g-evm.dts
index 0ca36ef..e95efd4 100644
--- a/arch/arm/dts/k2g-evm.dts
+++ b/arch/arm/dts/k2g-evm.dts
@@ -31,3 +31,72 @@
 &gbe0 {
 	phy-handle = <&ethphy0>;
 };
+
+&spi1 {
+	status = "okay";
+
+	spi_nor: flash@0 {
+		#address-cells = <1>;
+		#size-cells = <1>;
+		compatible = "spi-flash";
+		spi-max-frequency = <50000000>;
+		m25p,fast-read;
+		reg = <0>;
+
+		partition@0 {
+			label = "u-boot-spl";
+			reg = <0x0 0x80000>;
+			read-only;
+		};
+
+		partition@1 {
+			label = "misc";
+			reg = <0x80000 0xf80000>;
+		};
+	};
+};
+
+&qspi {
+	status = "okay";
+
+        flash0: m25p80@0 {
+                compatible = "s25fl512s","spi-flash";
+                reg = <0>;
+                spi-tx-bus-width = <1>;
+                spi-rx-bus-width = <4>;
+                spi-max-frequency = <96000000>;
+                #address-cells = <1>;
+                #size-cells = <1>;
+                tshsl-ns = <392>;
+                tsd2d-ns = <392>;
+                tchsh-ns = <100>;
+                tslch-ns = <100>;
+		block-size = <18>;
+
+
+                partition@0 {
+                        label = "QSPI.u-boot-spl-os";
+                        reg = <0x00000000 0x00100000>;
+                };
+                partition@1 {
+                        label = "QSPI.u-boot-env";
+                        reg = <0x00100000 0x00040000>;
+                };
+                partition@2 {
+                        label = "QSPI.skern";
+                        reg = <0x00140000 0x0040000>;
+                };
+                partition@3 {
+                        label = "QSPI.pmmc-firmware";
+                        reg = <0x00180000 0x0040000>;
+                };
+                partition@4 {
+                        label = "QSPI.kernel";
+                        reg = <0x001C0000 0x0800000>;
+                };
+                partition@5 {
+                        label = "QSPI.file-system";
+                        reg = <0x009C0000 0x3640000>;
+                };
+        };
+};
diff --git a/arch/arm/dts/k2g.dtsi b/arch/arm/dts/k2g.dtsi
index a3ed444..00cd492 100644
--- a/arch/arm/dts/k2g.dtsi
+++ b/arch/arm/dts/k2g.dtsi
@@ -19,6 +19,11 @@
 
 	aliases {
 		serial0	= &uart0;
+		spi0 = &spi0;
+		spi1 = &spi1;
+		spi2 = &spi2;
+		spi3 = &spi3;
+		spi4 = &qspi;
 	};
 
 	memory {
@@ -80,6 +85,19 @@
 			bus_freq = <2500000>;
 		};
 
+		qspi: qspi@2940000 {
+			compatible =  "cadence,qspi";
+			#address-cells = <1>;
+			#size-cells = <0>;
+			reg = <0x02940000 0x1000>,
+			      <0x24000000 0x4000000>;
+			interrupts = <GIC_SPI 198 IRQ_TYPE_EDGE_RISING>;
+			num-cs = <4>;
+			fifo-depth = <256>;
+			sram-size = <256>;
+			status = "disabled";
+		};
+
 		#include "k2g-netcp.dtsi"
 
 		pmmc: pmmc@2900000 {
@@ -88,5 +106,48 @@
 			ti,lpsc_module = <1>;
 		};
 
+		spi0: spi@21805400 {
+			compatible = "ti,keystone-spi", "ti,dm6441-spi";
+			reg = <0x21805400 0x200>;
+			num-cs = <4>;
+			ti,davinci-spi-intr-line = <0>;
+			interrupts = <GIC_SPI 64 IRQ_TYPE_EDGE_RISING>;
+			#address-cells = <1>;
+			#size-cells = <0>;
+			status = "disabled";
+		};
+
+		spi1: spi@21805800 {
+			compatible = "ti,keystone-spi", "ti,dm6441-spi";
+			reg = <0x21805800 0x200>;
+			num-cs = <4>;
+			ti,davinci-spi-intr-line = <0>;
+			interrupts = <GIC_SPI 66 IRQ_TYPE_EDGE_RISING>;
+			#address-cells = <1>;
+			#size-cells = <0>;
+			status = "disabled";
+		};
+
+		spi2: spi@21805c00 {
+			compatible = "ti,keystone-spi", "ti,dm6441-spi";
+			reg = <0x21805C00 0x200>;
+			num-cs = <4>;
+			ti,davinci-spi-intr-line = <0>;
+			interrupts = <GIC_SPI 68 IRQ_TYPE_EDGE_RISING>;
+			#address-cells = <1>;
+			#size-cells = <0>;
+			status = "disabled";
+		};
+
+		spi3: spi@21806000 {
+			compatible = "ti,keystone-spi", "ti,dm6441-spi";
+			reg = <0x21806000 0x200>;
+			num-cs = <4>;
+			ti,davinci-spi-intr-line = <0>;
+			interrupts = <GIC_SPI 70 IRQ_TYPE_EDGE_RISING>;
+			#address-cells = <1>;
+			#size-cells = <0>;
+			status = "disabled";
+		};
 	};
 };
diff --git a/arch/arm/dts/k2hk-evm.dts b/arch/arm/dts/k2hk-evm.dts
index 660ebf5..c5cad2c 100644
--- a/arch/arm/dts/k2hk-evm.dts
+++ b/arch/arm/dts/k2hk-evm.dts
@@ -147,10 +147,11 @@
 };
 
 &spi0 {
+	status = "okay";
 	nor_flash: n25q128a11@0 {
 		#address-cells = <1>;
 		#size-cells = <1>;
-		compatible = "Micron,n25q128a11";
+		compatible = "Micron,n25q128a11", "spi-flash";
 		spi-max-frequency = <54000000>;
 		m25p,fast-read;
 		reg = <0>;
diff --git a/arch/arm/dts/k2l-evm.dts b/arch/arm/dts/k2l-evm.dts
index 9a69a6b..da0661b 100644
--- a/arch/arm/dts/k2l-evm.dts
+++ b/arch/arm/dts/k2l-evm.dts
@@ -96,10 +96,11 @@
 };
 
 &spi0 {
+	status ="okay";
 	nor_flash: n25q128a11@0 {
 		#address-cells = <1>;
 		#size-cells = <1>;
-		compatible = "Micron,n25q128a11";
+		compatible = "Micron,n25q128a11", "spi-flash";
 		spi-max-frequency = <54000000>;
 		m25p,fast-read;
 		reg = <0>;
diff --git a/arch/arm/dts/keystone.dtsi b/arch/arm/dts/keystone.dtsi
index f39b969..be97f3f 100644
--- a/arch/arm/dts/keystone.dtsi
+++ b/arch/arm/dts/keystone.dtsi
@@ -19,6 +19,9 @@
 
 	aliases {
 		serial0	= &uart0;
+		spi0 = &spi0;
+		spi1 = &spi1;
+		spi2 = &spi2;
 	};
 
 	chosen {
diff --git a/cmd/sf.c b/cmd/sf.c
index 42862d9..286906c 100644
--- a/cmd/sf.c
+++ b/cmd/sf.c
@@ -88,6 +88,8 @@ static int do_spi_flash_probe(int argc, char * const argv[])
 #ifdef CONFIG_DM_SPI_FLASH
 	struct udevice *new, *bus_dev;
 	int ret;
+	/* In DM mode defaults will be taken from DT */
+	speed = 0, mode = 0;
 #else
 	struct spi_flash *new;
 #endif
diff --git a/common/env_sf.c b/common/env_sf.c
index 273098c..c53200f 100644
--- a/common/env_sf.c
+++ b/common/env_sf.c
@@ -55,9 +55,9 @@ int saveenv(void)
 #ifdef CONFIG_DM_SPI_FLASH
 	struct udevice *new;
 
+	/* speed and mode will be read from DT */
 	ret = spi_flash_probe_bus_cs(CONFIG_ENV_SPI_BUS, CONFIG_ENV_SPI_CS,
-				     CONFIG_ENV_SPI_MAX_HZ,
-				     CONFIG_ENV_SPI_MODE, &new);
+				     0, 0, &new);
 	if (ret) {
 		set_default_env("!spi_flash_probe_bus_cs() failed");
 		return 1;
@@ -245,9 +245,9 @@ int saveenv(void)
 #ifdef CONFIG_DM_SPI_FLASH
 	struct udevice *new;
 
+	/* speed and mode will be read from DT */
 	ret = spi_flash_probe_bus_cs(CONFIG_ENV_SPI_BUS, CONFIG_ENV_SPI_CS,
-				     CONFIG_ENV_SPI_MAX_HZ,
-				     CONFIG_ENV_SPI_MODE, &new);
+				     0, 0, &new);
 	if (ret) {
 		set_default_env("!spi_flash_probe_bus_cs() failed");
 		return 1;
diff --git a/configs/k2e_evm_defconfig b/configs/k2e_evm_defconfig
index 9fcdfe9..65561b1 100644
--- a/configs/k2e_evm_defconfig
+++ b/configs/k2e_evm_defconfig
@@ -28,6 +28,8 @@ CONFIG_CMD_FS_GENERIC=y
 CONFIG_OF_CONTROL=y
 CONFIG_DM=y
 CONFIG_TI_AEMIF=y
+CONFIG_DM_SPI=y
+CONFIG_DM_SPI_FLASH=y
 CONFIG_SPI_FLASH=y
 CONFIG_SPI_FLASH_STMICRO=y
 CONFIG_DM_ETH=y
diff --git a/configs/k2g_evm_defconfig b/configs/k2g_evm_defconfig
index 8efa58c..5d44e8d 100644
--- a/configs/k2g_evm_defconfig
+++ b/configs/k2g_evm_defconfig
@@ -27,6 +27,8 @@ CONFIG_CMD_FAT=y
 CONFIG_CMD_FS_GENERIC=y
 CONFIG_OF_CONTROL=y
 CONFIG_DM=y
+CONFIG_DM_SPI=y
+CONFIG_DM_SPI_FLASH=y
 CONFIG_SPI_FLASH=y
 CONFIG_SPI_FLASH_STMICRO=y
 CONFIG_DM_ETH=y
@@ -35,3 +37,5 @@ CONFIG_SYS_NS16550=y
 CONFIG_USB=y
 CONFIG_USB_XHCI_HCD=y
 CONFIG_USB_XHCI_DWC3=y
+CONFIG_SPI_FLASH_SPANSION=y
+CONFIG_SPI_FLASH_BAR=y
diff --git a/configs/k2hk_evm_defconfig b/configs/k2hk_evm_defconfig
index 278eaf3..8623e1c 100644
--- a/configs/k2hk_evm_defconfig
+++ b/configs/k2hk_evm_defconfig
@@ -28,6 +28,8 @@ CONFIG_CMD_FS_GENERIC=y
 CONFIG_OF_CONTROL=y
 CONFIG_DM=y
 CONFIG_TI_AEMIF=y
+CONFIG_DM_SPI=y
+CONFIG_DM_SPI_FLASH=y
 CONFIG_SPI_FLASH=y
 CONFIG_SPI_FLASH_STMICRO=y
 CONFIG_DM_ETH=y
diff --git a/configs/k2l_evm_defconfig b/configs/k2l_evm_defconfig
index 8417e0a..9aa429c 100644
--- a/configs/k2l_evm_defconfig
+++ b/configs/k2l_evm_defconfig
@@ -28,6 +28,8 @@ CONFIG_CMD_FS_GENERIC=y
 CONFIG_OF_CONTROL=y
 CONFIG_DM=y
 CONFIG_TI_AEMIF=y
+CONFIG_DM_SPI=y
+CONFIG_DM_SPI_FLASH=y
 CONFIG_SPI_FLASH=y
 CONFIG_SPI_FLASH_STMICRO=y
 CONFIG_DM_ETH=y
diff --git a/drivers/core/device.c b/drivers/core/device.c
index eb75b17..f7fb0cc 100644
--- a/drivers/core/device.c
+++ b/drivers/core/device.c
@@ -10,6 +10,7 @@
  */
 
 #include <common.h>
+#include <asm/io.h>
 #include <fdtdec.h>
 #include <fdt_support.h>
 #include <malloc.h>
@@ -697,6 +698,16 @@ void *dev_get_addr_ptr(struct udevice *dev)
 	return (void *)(uintptr_t)dev_get_addr_index(dev, 0);
 }
 
+void *dev_map_physmem(struct udevice *dev, unsigned long size)
+{
+	fdt_addr_t addr = dev_get_addr(dev);
+
+	if (addr == FDT_ADDR_T_NONE)
+		return NULL;
+
+	return map_physmem(addr, size, MAP_NOCACHE);
+}
+
 bool device_has_children(struct udevice *dev)
 {
 	return !list_empty(&dev->child_head);
diff --git a/drivers/spi/cadence_qspi.c b/drivers/spi/cadence_qspi.c
index 4f7fd52..a5244ff 100644
--- a/drivers/spi/cadence_qspi.c
+++ b/drivers/spi/cadence_qspi.c
@@ -191,6 +191,7 @@ static int cadence_spi_xfer(struct udevice *dev, unsigned int bitlen,
 	struct udevice *bus = dev->parent;
 	struct cadence_spi_platdata *plat = bus->platdata;
 	struct cadence_spi_priv *priv = dev_get_priv(bus);
+	struct dm_spi_slave_platdata *dm_plat = dev_get_parent_platdata(dev);
 	void *base = priv->regbase;
 	u8 *cmd_buf = priv->cmd_buf;
 	size_t data_bytes;
@@ -250,7 +251,7 @@ static int cadence_spi_xfer(struct udevice *dev, unsigned int bitlen,
 		break;
 		case CQSPI_INDIRECT_READ:
 			err = cadence_qspi_apb_indirect_read_setup(plat,
-				priv->cmd_len, cmd_buf);
+				priv->cmd_len, dm_plat->mode_rx, cmd_buf);
 			if (!err) {
 				err = cadence_qspi_apb_indirect_read_execute
 				(plat, data_bytes, din);
diff --git a/drivers/spi/cadence_qspi.h b/drivers/spi/cadence_qspi.h
index 2912e36..a849f7b 100644
--- a/drivers/spi/cadence_qspi.h
+++ b/drivers/spi/cadence_qspi.h
@@ -53,7 +53,7 @@ int cadence_qspi_apb_command_write(void *reg_base_addr,
 	unsigned int txlen,  const u8 *txbuf);
 
 int cadence_qspi_apb_indirect_read_setup(struct cadence_spi_platdata *plat,
-	unsigned int cmdlen, const u8 *cmdbuf);
+	unsigned int cmdlen, unsigned int rx_width, const u8 *cmdbuf);
 int cadence_qspi_apb_indirect_read_execute(struct cadence_spi_platdata *plat,
 	unsigned int rxlen, u8 *rxbuf);
 int cadence_qspi_apb_indirect_write_setup(struct cadence_spi_platdata *plat,
diff --git a/drivers/spi/cadence_qspi_apb.c b/drivers/spi/cadence_qspi_apb.c
index a71531d..1a35d55 100644
--- a/drivers/spi/cadence_qspi_apb.c
+++ b/drivers/spi/cadence_qspi_apb.c
@@ -29,6 +29,7 @@
 #include <asm/io.h>
 #include <asm/errno.h>
 #include <wait_bit.h>
+#include <spi.h>
 #include "cadence_qspi.h"
 
 #define CQSPI_REG_POLL_US			(1) /* 1us */
@@ -45,7 +46,6 @@
 #define CQSPI_INST_TYPE_QUAD			(2)
 
 #define CQSPI_STIG_DATA_LEN_MAX			(8)
-#define CQSPI_INDIRECTTRIGGER_ADDR_MASK		(0xFFFFF)
 
 #define CQSPI_DUMMY_CLKS_PER_BYTE		(8)
 #define CQSPI_DUMMY_BYTES_MAX			(4)
@@ -549,7 +549,7 @@ int cadence_qspi_apb_command_write(void *reg_base, unsigned int cmdlen,
 
 /* Opcode + Address (3/4 bytes) + dummy bytes (0-4 bytes) */
 int cadence_qspi_apb_indirect_read_setup(struct cadence_spi_platdata *plat,
-	unsigned int cmdlen, const u8 *cmdbuf)
+	unsigned int cmdlen, unsigned int rx_width, const u8 *cmdbuf)
 {
 	unsigned int reg;
 	unsigned int rd_reg;
@@ -573,16 +573,15 @@ int cadence_qspi_apb_indirect_read_setup(struct cadence_spi_platdata *plat,
 		addr_bytes = cmdlen - 1;
 
 	/* Setup the indirect trigger address */
-	writel(((u32)plat->ahbbase & CQSPI_INDIRECTTRIGGER_ADDR_MASK),
+	writel((u32)plat->ahbbase,
 	       plat->regbase + CQSPI_REG_INDIRECTTRIGGER);
 
 	/* Configure the opcode */
 	rd_reg = cmdbuf[0] << CQSPI_REG_RD_INSTR_OPCODE_LSB;
 
-#if (CONFIG_SPI_FLASH_QUAD == 1)
-	/* Instruction and address at DQ0, data at DQ0-3. */
-	rd_reg |= CQSPI_INST_TYPE_QUAD << CQSPI_REG_RD_INSTR_TYPE_DATA_LSB;
-#endif
+	if (rx_width & SPI_RX_QUAD)
+		/* Instruction and address at DQ0, data at DQ0-3. */
+		rd_reg |= CQSPI_INST_TYPE_QUAD << CQSPI_REG_RD_INSTR_TYPE_DATA_LSB;
 
 	/* Get address */
 	addr_value = cadence_qspi_apb_cmd2addr(&cmdbuf[1], addr_bytes);
@@ -714,7 +713,7 @@ int cadence_qspi_apb_indirect_write_setup(struct cadence_spi_platdata *plat,
 		return -EINVAL;
 	}
 	/* Setup the indirect trigger address */
-	writel(((u32)plat->ahbbase & CQSPI_INDIRECTTRIGGER_ADDR_MASK),
+	writel((u32)plat->ahbbase,
 	       plat->regbase + CQSPI_REG_INDIRECTTRIGGER);
 
 	/* Configure the opcode */
diff --git a/drivers/spi/davinci_spi.c b/drivers/spi/davinci_spi.c
index 0bd4f88..20aa99a 100644
--- a/drivers/spi/davinci_spi.c
+++ b/drivers/spi/davinci_spi.c
@@ -14,6 +14,7 @@
 #include <malloc.h>
 #include <asm/io.h>
 #include <asm/arch/hardware.h>
+#include <dm.h>
 
 /* SPIGCR0 */
 #define SPIGCR0_SPIENA_MASK	0x1
@@ -51,6 +52,7 @@
 /* SPIDEF */
 #define SPIDEF_CSDEF0_MASK	BIT(0)
 
+#ifndef CONFIG_DM_SPI
 #define SPI0_BUS		0
 #define SPI0_BASE		CONFIG_SYS_SPI_BASE
 /*
@@ -83,6 +85,9 @@
 #define SPI2_NUM_CS		CONFIG_SYS_SPI2_NUM_CS
 #define SPI2_BASE		CONFIG_SYS_SPI2_BASE
 #endif
+#endif
+
+DECLARE_GLOBAL_DATA_PTR;
 
 /* davinci spi register set */
 struct davinci_spi_regs {
@@ -114,16 +119,17 @@ struct davinci_spi_regs {
 
 /* davinci spi slave */
 struct davinci_spi_slave {
+#ifndef CONFIG_DM_SPI
 	struct spi_slave slave;
+#endif
 	struct davinci_spi_regs *regs;
-	unsigned int freq;
+	unsigned int freq; /* current SPI bus frequency */
+	unsigned int mode; /* current SPI mode used */
+	u8 num_cs;	   /* total no. of CS available */
+	u8 cur_cs;	   /* CS of current slave */
+	bool half_duplex;  /* true, if master is half-duplex only */
 };
 
-static inline struct davinci_spi_slave *to_davinci_spi(struct spi_slave *slave)
-{
-	return container_of(slave, struct davinci_spi_slave, slave);
-}
-
 /*
  * This functions needs to act like a macro to avoid pipeline reloads in the
  * loops below. Use always_inline. This gains us about 160KiB/s and the bloat
@@ -144,15 +150,14 @@ static inline u32 davinci_spi_xfer_data(struct davinci_spi_slave *ds, u32 data)
 	return buf_reg_val;
 }
 
-static int davinci_spi_read(struct spi_slave *slave, unsigned int len,
+static int davinci_spi_read(struct davinci_spi_slave *ds, unsigned int len,
 			    u8 *rxp, unsigned long flags)
 {
-	struct davinci_spi_slave *ds = to_davinci_spi(slave);
 	unsigned int data1_reg_val;
 
 	/* enable CS hold, CS[n] and clear the data bits */
 	data1_reg_val = ((1 << SPIDAT1_CSHOLD_SHIFT) |
-			 (slave->cs << SPIDAT1_CSNR_SHIFT));
+			 (ds->cur_cs << SPIDAT1_CSNR_SHIFT));
 
 	/* wait till TXFULL is deasserted */
 	while (readl(&ds->regs->buf) & SPIBUF_TXFULL_MASK)
@@ -175,15 +180,14 @@ static int davinci_spi_read(struct spi_slave *slave, unsigned int len,
 	return 0;
 }
 
-static int davinci_spi_write(struct spi_slave *slave, unsigned int len,
+static int davinci_spi_write(struct davinci_spi_slave *ds, unsigned int len,
 			     const u8 *txp, unsigned long flags)
 {
-	struct davinci_spi_slave *ds = to_davinci_spi(slave);
 	unsigned int data1_reg_val;
 
 	/* enable CS hold and clear the data bits */
 	data1_reg_val = ((1 << SPIDAT1_CSHOLD_SHIFT) |
-			 (slave->cs << SPIDAT1_CSNR_SHIFT));
+			 (ds->cur_cs << SPIDAT1_CSNR_SHIFT));
 
 	/* wait till TXFULL is deasserted */
 	while (readl(&ds->regs->buf) & SPIBUF_TXFULL_MASK)
@@ -209,16 +213,15 @@ static int davinci_spi_write(struct spi_slave *slave, unsigned int len,
 	return 0;
 }
 
-#ifndef CONFIG_SPI_HALF_DUPLEX
-static int davinci_spi_read_write(struct spi_slave *slave, unsigned int len,
-				  u8 *rxp, const u8 *txp, unsigned long flags)
+static int davinci_spi_read_write(struct davinci_spi_slave *ds, unsigned
+				  int len, u8 *rxp, const u8 *txp,
+				  unsigned long flags)
 {
-	struct davinci_spi_slave *ds = to_davinci_spi(slave);
 	unsigned int data1_reg_val;
 
 	/* enable CS hold and clear the data bits */
 	data1_reg_val = ((1 << SPIDAT1_CSHOLD_SHIFT) |
-			 (slave->cs << SPIDAT1_CSNR_SHIFT));
+			 (ds->cur_cs << SPIDAT1_CSNR_SHIFT));
 
 	/* wait till TXFULL is deasserted */
 	while (readl(&ds->regs->buf) & SPIBUF_TXFULL_MASK)
@@ -237,7 +240,115 @@ static int davinci_spi_read_write(struct spi_slave *slave, unsigned int len,
 
 	return 0;
 }
-#endif
+
+
+static int __davinci_spi_claim_bus(struct davinci_spi_slave *ds, int cs)
+{
+	unsigned int mode = 0, scalar;
+
+	/* Enable the SPI hardware */
+	writel(SPIGCR0_SPIRST_MASK, &ds->regs->gcr0);
+	udelay(1000);
+	writel(SPIGCR0_SPIENA_MASK, &ds->regs->gcr0);
+
+	/* Set master mode, powered up and not activated */
+	writel(SPIGCR1_MASTER_MASK | SPIGCR1_CLKMOD_MASK, &ds->regs->gcr1);
+
+	/* CS, CLK, SIMO and SOMI are functional pins */
+	writel(((1 << cs) | SPIPC0_CLKFUN_MASK |
+		SPIPC0_DOFUN_MASK | SPIPC0_DIFUN_MASK), &ds->regs->pc0);
+
+	/* setup format */
+	scalar = ((CONFIG_SYS_SPI_CLK / ds->freq) - 1) & 0xFF;
+
+	/*
+	 * Use following format:
+	 *   character length = 8,
+	 *   MSB shifted out first
+	 */
+	if (ds->mode & SPI_CPOL)
+		mode |= SPI_CPOL;
+	if (!(ds->mode & SPI_CPHA))
+		mode |= SPI_CPHA;
+	writel(8 | (scalar << SPIFMT_PRESCALE_SHIFT) |
+		(mode << SPIFMT_PHASE_SHIFT), &ds->regs->fmt0);
+
+	/*
+	 * Including a minor delay. No science here. Should be good even with
+	 * no delay
+	 */
+	writel((50 << SPI_C2TDELAY_SHIFT) |
+		(50 << SPI_T2CDELAY_SHIFT), &ds->regs->delay);
+
+	/* default chip select register */
+	writel(SPIDEF_CSDEF0_MASK, &ds->regs->def);
+
+	/* no interrupts */
+	writel(0, &ds->regs->int0);
+	writel(0, &ds->regs->lvl);
+
+	/* enable SPI */
+	writel((readl(&ds->regs->gcr1) | SPIGCR1_SPIENA_MASK), &ds->regs->gcr1);
+
+	return 0;
+}
+
+static int __davinci_spi_release_bus(struct davinci_spi_slave *ds)
+{
+	/* Disable the SPI hardware */
+	writel(SPIGCR0_SPIRST_MASK, &ds->regs->gcr0);
+
+	return 0;
+}
+
+static int __davinci_spi_xfer(struct davinci_spi_slave *ds,
+		unsigned int bitlen,  const void *dout, void *din,
+		unsigned long flags)
+{
+	unsigned int len;
+
+	if (bitlen == 0)
+		/* Finish any previously submitted transfers */
+		goto out;
+
+	/*
+	 * It's not clear how non-8-bit-aligned transfers are supposed to be
+	 * represented as a stream of bytes...this is a limitation of
+	 * the current SPI interface - here we terminate on receiving such a
+	 * transfer request.
+	 */
+	if (bitlen % 8) {
+		/* Errors always terminate an ongoing transfer */
+		flags |= SPI_XFER_END;
+		goto out;
+	}
+
+	len = bitlen / 8;
+
+	if (!dout)
+		return davinci_spi_read(ds, len, din, flags);
+	if (!din)
+		return davinci_spi_write(ds, len, dout, flags);
+	if (!ds->half_duplex)
+		return davinci_spi_read_write(ds, len, din, dout, flags);
+
+	printf("SPI full duplex not supported\n");
+	flags |= SPI_XFER_END;
+
+out:
+	if (flags & SPI_XFER_END) {
+		u8 dummy = 0;
+		davinci_spi_write(ds, 1, &dummy, flags);
+	}
+	return 0;
+}
+
+#ifndef CONFIG_DM_SPI
+
+static inline struct davinci_spi_slave *to_davinci_spi(struct spi_slave *slave)
+{
+	return container_of(slave, struct davinci_spi_slave, slave);
+}
 
 int spi_cs_is_valid(unsigned int bus, unsigned int cs)
 {
@@ -313,6 +424,7 @@ struct spi_slave *spi_setup_slave(unsigned int bus, unsigned int cs,
 	}
 
 	ds->freq = max_hz;
+	ds->mode = mode;
 
 	return &ds->slave;
 }
@@ -324,104 +436,143 @@ void spi_free_slave(struct spi_slave *slave)
 	free(ds);
 }
 
+int spi_xfer(struct spi_slave *slave, unsigned int bitlen,
+	     const void *dout, void *din, unsigned long flags)
+{
+	struct davinci_spi_slave *ds = to_davinci_spi(slave);
+
+	ds->cur_cs = slave->cs;
+
+	return __davinci_spi_xfer(ds, bitlen, dout, din, flags);
+}
+
 int spi_claim_bus(struct spi_slave *slave)
 {
 	struct davinci_spi_slave *ds = to_davinci_spi(slave);
-	unsigned int scalar;
 
-	/* Enable the SPI hardware */
-	writel(SPIGCR0_SPIRST_MASK, &ds->regs->gcr0);
-	udelay(1000);
-	writel(SPIGCR0_SPIENA_MASK, &ds->regs->gcr0);
+#ifdef CONFIG_SPI_HALF_DUPLEX
+	ds->half_duplex = true;
+#else
+	ds->half_duplex = false;
+#endif
+	return __davinci_spi_claim_bus(ds, ds->slave.cs);
+}
 
-	/* Set master mode, powered up and not activated */
-	writel(SPIGCR1_MASTER_MASK | SPIGCR1_CLKMOD_MASK, &ds->regs->gcr1);
+void spi_release_bus(struct spi_slave *slave)
+{
+	struct davinci_spi_slave *ds = to_davinci_spi(slave);
 
-	/* CS, CLK, SIMO and SOMI are functional pins */
-	writel(((1 << slave->cs) | SPIPC0_CLKFUN_MASK |
-		SPIPC0_DOFUN_MASK | SPIPC0_DIFUN_MASK), &ds->regs->pc0);
+	__davinci_spi_release_bus(ds);
+}
 
-	/* setup format */
-	scalar = ((CONFIG_SYS_SPI_CLK / ds->freq) - 1) & 0xFF;
+#else
+static int davinci_spi_set_speed(struct udevice *bus, uint max_hz)
+{
+	struct davinci_spi_slave *ds = dev_get_priv(bus);
 
-	/*
-	 * Use following format:
-	 *   character length = 8,
-	 *   clock signal delayed by half clk cycle,
-	 *   clock low in idle state - Mode 0,
-	 *   MSB shifted out first
-	 */
-	writel(8 | (scalar << SPIFMT_PRESCALE_SHIFT) |
-		(1 << SPIFMT_PHASE_SHIFT), &ds->regs->fmt0);
+	debug("%s speed %u\n", __func__, max_hz);
+	if (max_hz > CONFIG_SYS_SPI_CLK / 2)
+		return -EINVAL;
 
-	/*
-	 * Including a minor delay. No science here. Should be good even with
-	 * no delay
-	 */
-	writel((50 << SPI_C2TDELAY_SHIFT) |
-		(50 << SPI_T2CDELAY_SHIFT), &ds->regs->delay);
+	ds->freq = max_hz;
 
-	/* default chip select register */
-	writel(SPIDEF_CSDEF0_MASK, &ds->regs->def);
+	return 0;
+}
 
-	/* no interrupts */
-	writel(0, &ds->regs->int0);
-	writel(0, &ds->regs->lvl);
+static int davinci_spi_set_mode(struct udevice *bus, uint mode)
+{
+	struct davinci_spi_slave *ds = dev_get_priv(bus);
 
-	/* enable SPI */
-	writel((readl(&ds->regs->gcr1) | SPIGCR1_SPIENA_MASK), &ds->regs->gcr1);
+	debug("%s mode %u\n", __func__, mode);
+	ds->mode = mode;
 
 	return 0;
 }
 
-void spi_release_bus(struct spi_slave *slave)
+static int davinci_spi_claim_bus(struct udevice *dev)
 {
-	struct davinci_spi_slave *ds = to_davinci_spi(slave);
+	struct dm_spi_slave_platdata *slave_plat =
+		dev_get_parent_platdata(dev);
+	struct udevice *bus = dev->parent;
+	struct davinci_spi_slave *ds = dev_get_priv(bus);
+
+	if (slave_plat->cs >= ds->num_cs) {
+		printf("Invalid SPI chipselect\n");
+		return -EINVAL;
+	}
+	ds->half_duplex = slave_plat->mode & SPI_PREAMBLE;
 
-	/* Disable the SPI hardware */
-	writel(SPIGCR0_SPIRST_MASK, &ds->regs->gcr0);
+	return __davinci_spi_claim_bus(ds, slave_plat->cs);
 }
 
-int spi_xfer(struct spi_slave *slave, unsigned int bitlen,
-	     const void *dout, void *din, unsigned long flags)
+static int davinci_spi_release_bus(struct udevice *dev)
 {
-	unsigned int len;
+	struct davinci_spi_slave *ds = dev_get_priv(dev->parent);
 
-	if (bitlen == 0)
-		/* Finish any previously submitted transfers */
-		goto out;
+	return __davinci_spi_release_bus(ds);
+}
 
-	/*
-	 * It's not clear how non-8-bit-aligned transfers are supposed to be
-	 * represented as a stream of bytes...this is a limitation of
-	 * the current SPI interface - here we terminate on receiving such a
-	 * transfer request.
-	 */
-	if (bitlen % 8) {
-		/* Errors always terminate an ongoing transfer */
-		flags |= SPI_XFER_END;
-		goto out;
+static int davinci_spi_xfer(struct udevice *dev, unsigned int bitlen,
+			    const void *dout, void *din,
+			    unsigned long flags)
+{
+	struct dm_spi_slave_platdata *slave =
+		dev_get_parent_platdata(dev);
+	struct udevice *bus = dev->parent;
+	struct davinci_spi_slave *ds = dev_get_priv(bus);
+
+	if (slave->cs >= ds->num_cs) {
+		printf("Invalid SPI chipselect\n");
+		return -EINVAL;
 	}
+	ds->cur_cs = slave->cs;
 
-	len = bitlen / 8;
+	return __davinci_spi_xfer(ds, bitlen, dout, din, flags);
+}
 
-	if (!dout)
-		return davinci_spi_read(slave, len, din, flags);
-	else if (!din)
-		return davinci_spi_write(slave, len, dout, flags);
-#ifndef CONFIG_SPI_HALF_DUPLEX
-	else
-		return davinci_spi_read_write(slave, len, din, dout, flags);
-#else
-	printf("SPI full duplex transaction requested with "
-	       "CONFIG_SPI_HALF_DUPLEX defined.\n");
-	flags |= SPI_XFER_END;
-#endif
+static int davinci_spi_probe(struct udevice *bus)
+{
+	/* Nothing to do */
+	return 0;
+}
 
-out:
-	if (flags & SPI_XFER_END) {
-		u8 dummy = 0;
-		davinci_spi_write(slave, 1, &dummy, flags);
+static int davinci_ofdata_to_platadata(struct udevice *bus)
+{
+	struct davinci_spi_slave *ds = dev_get_priv(bus);
+	const void *blob = gd->fdt_blob;
+	int node = bus->of_offset;
+
+	ds->regs = dev_map_physmem(bus, sizeof(struct davinci_spi_regs));
+	if (!ds->regs) {
+		printf("%s: could not map device address\n", __func__);
+		return -EINVAL;
 	}
+	ds->num_cs = fdtdec_get_int(blob, node, "num-cs", 4);
+
 	return 0;
 }
+
+static const struct dm_spi_ops davinci_spi_ops = {
+	.claim_bus	= davinci_spi_claim_bus,
+	.release_bus	= davinci_spi_release_bus,
+	.xfer		= davinci_spi_xfer,
+	.set_speed	= davinci_spi_set_speed,
+	.set_mode	= davinci_spi_set_mode,
+};
+
+static const struct udevice_id davinci_spi_ids[] = {
+	{ .compatible = "ti,keystone-spi" },
+	{ .compatible = "ti,dm6441-spi" },
+	{ }
+};
+
+U_BOOT_DRIVER(davinci_spi) = {
+	.name = "davinci_spi",
+	.id = UCLASS_SPI,
+	.of_match = davinci_spi_ids,
+	.ops = &davinci_spi_ops,
+	.ofdata_to_platdata = davinci_ofdata_to_platadata,
+	.priv_auto_alloc_size = sizeof(struct davinci_spi_slave),
+	.probe = davinci_spi_probe,
+};
+#endif
diff --git a/drivers/spi/spi-uclass.c b/drivers/spi/spi-uclass.c
index 84b6786..8003f9b 100644
--- a/drivers/spi/spi-uclass.c
+++ b/drivers/spi/spi-uclass.c
@@ -278,6 +278,7 @@ int spi_get_bus_and_cs(int busnum, int cs, int speed, int mode,
 		       struct udevice **busp, struct spi_slave **devp)
 {
 	struct udevice *bus, *dev;
+	struct dm_spi_slave_platdata *plat;
 	bool created = false;
 	int ret;
 
@@ -294,8 +295,6 @@ int spi_get_bus_and_cs(int busnum, int cs, int speed, int mode,
 	 * SPI flash chip - we will bind to the correct driver.
 	 */
 	if (ret == -ENODEV && drv_name) {
-		struct dm_spi_slave_platdata *plat;
-
 		debug("%s: Binding new device '%s', busnum=%d, cs=%d, driver=%s\n",
 		      __func__, dev_name, busnum, cs, drv_name);
 		ret = device_bind_driver(bus, drv_name, dev_name, &dev);
@@ -322,6 +321,11 @@ int spi_get_bus_and_cs(int busnum, int cs, int speed, int mode,
 		slave->dev = dev;
 	}
 
+	plat = dev_get_parent_platdata(dev);
+	if (!speed) {
+		speed = plat->max_hz;
+		mode = plat->mode;
+	}
 	ret = spi_set_speed_mode(bus, speed, mode);
 	if (ret)
 		goto err;
@@ -333,7 +337,7 @@ int spi_get_bus_and_cs(int busnum, int cs, int speed, int mode,
 	return 0;
 
 err:
-	debug("%s: Error path, credted=%d, device '%s'\n", __func__,
+	debug("%s: Error path, created=%d, device '%s'\n", __func__,
 	      created, dev->name);
 	if (created) {
 		device_remove(dev);
diff --git a/include/configs/k2g_evm.h b/include/configs/k2g_evm.h
index f8bba67..71b0037 100644
--- a/include/configs/k2g_evm.h
+++ b/include/configs/k2g_evm.h
@@ -73,4 +73,10 @@
 #define CONFIG_SF_DEFAULT_BUS		1
 #define CONFIG_SF_DEFAULT_CS		0
 
+#ifndef CONFIG_SPL_BUILD
+#define CONFIG_CADENCE_QSPI
+#define CONFIG_CQSPI_REF_CLK 384000000
+#define CONFIG_CQSPI_DECODER 0x0
+#endif
+
 #endif /* __CONFIG_K2G_EVM_H */
diff --git a/include/configs/ti_armv7_keystone2.h b/include/configs/ti_armv7_keystone2.h
index 2ee26c4..4aa262e 100644
--- a/include/configs/ti_armv7_keystone2.h
+++ b/include/configs/ti_armv7_keystone2.h
@@ -89,6 +89,10 @@
 #define CONFIG_SYS_SPI2
 #define CONFIG_SYS_SPI2_BASE		KS2_SPI2_BASE
 #define CONFIG_SYS_SPI2_NUM_CS		4
+#ifdef CONFIG_SPL_BUILD
+#undef CONFIG_DM_SPI
+#undef CONFIG_DM_SPI_FLASH
+#endif
 
 /* Network Configuration */
 #define CONFIG_PHYLIB
diff --git a/include/dm/device.h b/include/dm/device.h
index f03bcd3..1bfcf3b 100644
--- a/include/dm/device.h
+++ b/include/dm/device.h
@@ -467,6 +467,19 @@ fdt_addr_t dev_get_addr(struct udevice *dev);
 void *dev_get_addr_ptr(struct udevice *dev);
 
 /**
+ * dev_map_physmem() - Read device address from reg property of the
+ *                     device node and map the address into CPU address
+ *                     space.
+ *
+ * @dev: Pointer to device
+ * @size: size of the memory to map
+ *
+ * @return  mapped address, or NULL if the device does not have reg
+ *          property.
+ */
+void *dev_map_physmem(struct udevice *dev, unsigned long size);
+
+/**
  * dev_get_addr_index() - Get the indexed reg property of a device
  *
  * @dev: Pointer to a device