diff --git a/MAINTAINERS b/MAINTAINERS
index 237a022f07..64fb41ece7 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -139,7 +139,7 @@ S:	Maintained
 F:	arch/arm/cpu/armv8/hisilicon
 F:	arch/arm/include/asm/arch-hi6220/
 
-ARM MARVELL KIRKWOOD ARMADA-XP ARMADA-38X
+ARM MARVELL KIRKWOOD ARMADA-XP ARMADA-38X ARMADA-37XX
 M:	Prafulla Wadaskar <prafulla@marvell.com>
 M:	Luka Perkov <luka.perkov@sartura.hr>
 M:	Stefan Roese <sr@denx.de>
@@ -148,6 +148,7 @@ T:	git git://git.denx.de/u-boot-marvell.git
 F:	arch/arm/mach-kirkwood/
 F:	arch/arm/mach-mvebu/
 F:	drivers/ata/ahci_mvebu.c
+F:	drivers/phy/marvell/
 
 ARM MARVELL PXA
 M:	Marek Vasut <marex@denx.de>
diff --git a/arch/arm/dts/armada-3720-turris-mox.dts b/arch/arm/dts/armada-3720-turris-mox.dts
index bef100afce..7babc16679 100644
--- a/arch/arm/dts/armada-3720-turris-mox.dts
+++ b/arch/arm/dts/armada-3720-turris-mox.dts
@@ -94,17 +94,13 @@
 	};
 };
 
-&pinctrl_sb {
-	smi_pins: smi-pins {
-		groups = "smi";
-		function = "smi";
-	};
-};
-
 &spi0 {
 	status = "okay";
 	pinctrl-names = "default";
 	pinctrl-0 = <&spi_cs1_pins>;
+	assigned-clocks = <&nb_periph_clk 7>;
+	assigned-clock-parents = <&tbg 1>;
+	assigned-clock-rates = <20000000>;
 
 	spi-flash@0 {
 		#address-cells = <1>;
@@ -130,3 +126,10 @@
 	vbus-supply = <&reg_usb3_vbus>;
 	status = "okay";
 };
+
+&pcie0 {
+	pinctrl-names = "default";
+	pinctrl-0 = <&pcie_pins>;
+	reset-gpio = <&gpiosb 3 GPIO_ACTIVE_HIGH>;
+	status = "disabled";
+};
diff --git a/arch/arm/dts/armada-8040-mcbin.dts b/arch/arm/dts/armada-8040-mcbin.dts
index 08f1d7df69..f912596c2c 100644
--- a/arch/arm/dts/armada-8040-mcbin.dts
+++ b/arch/arm/dts/armada-8040-mcbin.dts
@@ -154,14 +154,6 @@
 	status = "okay";
 };
 
-/* uSD slot */
-&cpm_sdhci0 {
-	pinctrl-names = "default";
-	pinctrl-0 = <&cpm_sdhci_pins>;
-	bus-width = <4>;
-	status = "okay";
-};
-
 &cpm_comphy {
 	/*
 	 * CP0 Serdes Configuration:
diff --git a/arch/arm/mach-mvebu/armada8k/cpu.c b/arch/arm/mach-mvebu/armada8k/cpu.c
index f8e8e73610..959a7cff76 100644
--- a/arch/arm/mach-mvebu/armada8k/cpu.c
+++ b/arch/arm/mach-mvebu/armada8k/cpu.c
@@ -18,6 +18,10 @@
 #define RFU_GLOBAL_SW_RST		(MVEBU_RFU_BASE + 0x84)
 #define RFU_SW_RESET_OFFSET		0
 
+#define SAR0_REG			(MVEBU_REGISTER(0x2400200))
+#define BOOT_MODE_MASK			0x3f
+#define BOOT_MODE_OFFSET		4
+
 /*
  * The following table includes all memory regions for Armada 7k and
  * 8k SoCs. The Armada 7k is missing the CP110 slave regions here. Lets
@@ -125,3 +129,23 @@ u32 mvebu_get_nand_clock(void)
 	else
 		return 250 * 1000000;
 }
+
+int mmc_get_env_dev(void)
+{
+	u32 reg;
+	unsigned int boot_mode;
+
+	reg = readl(SAR0_REG);
+	boot_mode = (reg >> BOOT_MODE_OFFSET) & BOOT_MODE_MASK;
+
+	switch (boot_mode) {
+	case 0x28:
+	case 0x2a:
+		return 0;
+	case 0x29:
+	case 0x2b:
+		return 1;
+	}
+
+	return CONFIG_SYS_MMC_ENV_DEV;
+}
diff --git a/arch/arm/mach-mvebu/include/mach/soc.h b/arch/arm/mach-mvebu/include/mach/soc.h
index 623ab4eb8d..6e2e14efe0 100644
--- a/arch/arm/mach-mvebu/include/mach/soc.h
+++ b/arch/arm/mach-mvebu/include/mach/soc.h
@@ -110,16 +110,12 @@
 #define COMPHY_REFCLK_ALIGNMENT	(MVEBU_REGISTER(0x182f8))
 
 /* BootROM error register (also includes some status infos) */
-#if defined(CONFIG_ARMADA_38X)
-#define CONFIG_BOOTROM_ERR_REG	(MVEBU_REGISTER(0x182d0))
-#define BOOTROM_ERR_MODE_OFFS	0
-#define BOOTROM_ERR_MODE_MASK	(0xf << BOOTROM_ERR_MODE_OFFS)
-#else
 #define CONFIG_BOOTROM_ERR_REG	(MVEBU_REGISTER(0x182d0))
 #define BOOTROM_ERR_MODE_OFFS	28
 #define BOOTROM_ERR_MODE_MASK	(0xf << BOOTROM_ERR_MODE_OFFS)
 #define BOOTROM_ERR_MODE_UART	0x6
-#endif
+#define BOOTROM_ERR_CODE_OFFS	0
+#define BOOTROM_ERR_CODE_MASK	(0xf << BOOTROM_ERR_CODE_OFFS)
 
 #if defined(CONFIG_ARMADA_375)
 /* SAR values for Armada 375 */
diff --git a/arch/arm/mach-mvebu/spl.c b/arch/arm/mach-mvebu/spl.c
index 50b24f5760..9dd7c84b68 100644
--- a/arch/arm/mach-mvebu/spl.c
+++ b/arch/arm/mach-mvebu/spl.c
@@ -25,18 +25,19 @@ static u32 get_boot_device(void)
 	val = readl(CONFIG_BOOTROM_ERR_REG);
 	boot_device = (val & BOOTROM_ERR_MODE_MASK) >> BOOTROM_ERR_MODE_OFFS;
 	debug("BOOTROM_REG=0x%08x boot_device=0x%x\n", val, boot_device);
-#if defined(CONFIG_ARMADA_38X)
-	/*
-	 * If the bootrom error register contains any else than zeros
-	 * in the first 8 bits it's an error condition. And in that case
-	 * try to boot from UART.
-	 */
-	if (boot_device)
-#else
 	if (boot_device == BOOTROM_ERR_MODE_UART)
-#endif
 		return BOOT_DEVICE_UART;
 
+#ifdef CONFIG_ARMADA_38X
+	/*
+	 * If the bootrom error code contains any other than zeros it's an
+	 * error condition and the bootROM has fallen back to UART boot
+	 */
+	boot_device = (val & BOOTROM_ERR_CODE_MASK) >> BOOTROM_ERR_CODE_OFFS;
+	if (boot_device)
+		return BOOT_DEVICE_UART;
+#endif
+
 	/*
 	 * Now check the SAR register for the strapped boot-device
 	 */
diff --git a/board/CZ.NIC/turris_mox/turris_mox.c b/board/CZ.NIC/turris_mox/turris_mox.c
index 130d4c606d..c4622a49c2 100644
--- a/board/CZ.NIC/turris_mox/turris_mox.c
+++ b/board/CZ.NIC/turris_mox/turris_mox.c
@@ -4,18 +4,100 @@
  */
 
 #include <common.h>
+#include <asm/io.h>
 #include <dm.h>
 #include <clk.h>
 #include <spi.h>
 #include <linux/string.h>
+#include <linux/libfdt.h>
+#include <fdt_support.h>
 
-#ifdef CONFIG_WDT_ARMADA_3720
+#ifdef CONFIG_WDT_ARMADA_37XX
 #include <wdt.h>
 #endif
 
+#define MAX_MOX_MODULES		10
+
+#define MOX_MODULE_SFP		0x1
+#define MOX_MODULE_PCI		0x2
+#define MOX_MODULE_TOPAZ	0x3
+#define MOX_MODULE_PERIDOT	0x4
+#define MOX_MODULE_USB3		0x5
+#define MOX_MODULE_PASSPCI	0x6
+
+#define ARMADA_37XX_NB_GPIO_SEL	0xd0013830
+#define ARMADA_37XX_SPI_CTRL	0xd0010600
+#define ARMADA_37XX_SPI_CFG	0xd0010604
+#define ARMADA_37XX_SPI_DOUT	0xd0010608
+#define ARMADA_37XX_SPI_DIN	0xd001060c
+
+#define PCIE_PATH	"/soc/pcie@d0070000"
+
 DECLARE_GLOBAL_DATA_PTR;
 
-#ifdef CONFIG_WDT_ARMADA_3720
+#if defined(CONFIG_OF_BOARD_FIXUP)
+int board_fix_fdt(void *blob)
+{
+	u8 topology[MAX_MOX_MODULES];
+	int i, size, node;
+	bool enable;
+
+	/*
+	 * SPI driver is not loaded in driver model yet, but we have to find out
+	 * if pcie should be enabled in U-Boot's device tree. Therefore we have
+	 * to read SPI by reading/writing SPI registers directly
+	 */
+
+	writel(0x563fa, ARMADA_37XX_NB_GPIO_SEL);
+	writel(0x10df, ARMADA_37XX_SPI_CFG);
+	writel(0x2005b, ARMADA_37XX_SPI_CTRL);
+
+	while (!(readl(ARMADA_37XX_SPI_CTRL) & 0x2))
+		udelay(1);
+
+	for (i = 0; i < MAX_MOX_MODULES; ++i) {
+		writel(0x0, ARMADA_37XX_SPI_DOUT);
+
+		while (!(readl(ARMADA_37XX_SPI_CTRL) & 0x2))
+			udelay(1);
+
+		topology[i] = readl(ARMADA_37XX_SPI_DIN) & 0xff;
+		if (topology[i] == 0xff)
+			break;
+
+		topology[i] &= 0xf;
+	}
+
+	size = i;
+
+	writel(0x5b, ARMADA_37XX_SPI_CTRL);
+
+	if (size > 1 && (topology[1] == MOX_MODULE_PCI ||
+			 topology[1] == MOX_MODULE_USB3 ||
+			 topology[1] == MOX_MODULE_PASSPCI))
+		enable = true;
+	else
+		enable = false;
+
+	node = fdt_path_offset(blob, PCIE_PATH);
+
+	if (node < 0) {
+		printf("Cannot find PCIe node in U-Boot's device tree!\n");
+		return 0;
+	}
+
+	if (fdt_setprop_string(blob, node, "status",
+			       enable ? "okay" : "disabled") < 0) {
+		printf("Cannot %s PCIe in U-Boot's device tree!\n",
+		       enable ? "enable" : "disable");
+		return 0;
+	}
+
+	return 0;
+}
+#endif
+
+#ifdef CONFIG_WDT_ARMADA_37XX
 static struct udevice *watchdog_dev;
 
 void watchdog_reset(void)
@@ -41,7 +123,7 @@ int board_init(void)
 	/* address of boot parameters */
 	gd->bd->bi_boot_params = CONFIG_SYS_SDRAM_BASE + 0x100;
 
-#ifdef CONFIG_WDT_ARMADA_3720
+#ifdef CONFIG_WDT_ARMADA_37XX
 	if (uclass_get_device(UCLASS_WDT, 0, &watchdog_dev)) {
 		printf("Cannot find Armada 3720 watchdog!\n");
 	} else {
diff --git a/cmd/Makefile b/cmd/Makefile
index 3487c80455..a61fab6583 100644
--- a/cmd/Makefile
+++ b/cmd/Makefile
@@ -163,12 +163,13 @@ obj-$(CONFIG_CMD_BLOB) += blob.o
 obj-$(CONFIG_CMD_AVB) += avb.o
 
 obj-$(CONFIG_X86) += x86/
+
+obj-$(CONFIG_ARCH_MVEBU) += mvebu/
 endif # !CONFIG_SPL_BUILD
 
 # core command
 obj-y += nvedit.o
 
-obj-$(CONFIG_ARCH_MVEBU) += mvebu/
 obj-$(CONFIG_TI_COMMON_CMD_OPTIONS) += ti/
 
 filechk_data_gz = (echo "static const char data_gz[] ="; cat $< | scripts/bin2c; echo ";")
diff --git a/configs/turris_mox_defconfig b/configs/turris_mox_defconfig
index 9c37aac867..6463abcd3d 100644
--- a/configs/turris_mox_defconfig
+++ b/configs/turris_mox_defconfig
@@ -13,10 +13,12 @@ CONFIG_SYS_CONSOLE_INFO_QUIET=y
 # CONFIG_DISPLAY_CPUINFO is not set
 # CONFIG_DISPLAY_BOARDINFO is not set
 CONFIG_ARCH_EARLY_INIT_R=y
+CONFIG_OF_BOARD_FIXUP=y
 CONFIG_CMD_CLK=y
 # CONFIG_CMD_FLASH is not set
 CONFIG_CMD_I2C=y
 CONFIG_CMD_MMC=y
+CONFIG_CMD_PCI=y
 CONFIG_CMD_SF=y
 CONFIG_CMD_SPI=y
 CONFIG_CMD_USB=y
@@ -51,6 +53,10 @@ CONFIG_MVEBU_COMPHY_SUPPORT=y
 CONFIG_PINCTRL=y
 CONFIG_PINCTRL_ARMADA_37XX=y
 CONFIG_DM_REGULATOR_FIXED=y
+CONFIG_PCI=y
+CONFIG_DM_PCI=y
+CONFIG_PCI_AARDVARK=y
+# CONFIG_PCI_PNP is not set
 # CONFIG_SPL_SERIAL_PRESENT is not set
 CONFIG_DEBUG_MVEBU_A3700_UART=y
 CONFIG_DEBUG_UART_SHIFT=2
diff --git a/doc/git-mailrc b/doc/git-mailrc
index bf8f2a5b69..5d835d4de5 100644
--- a/doc/git-mailrc
+++ b/doc/git-mailrc
@@ -56,7 +56,7 @@ alias arm            uboot, aaribaud, trini
 alias at91           uboot, abiessmann
 alias davinci        ti
 alias imx            uboot, sbabic
-alias kirkwood       uboot, prafulla, luka
+alias kirkwood       uboot, prafulla, luka, stroese
 alias omap           ti
 alias pxa            uboot, marex
 alias rmobile        uboot, iwamatsu
diff --git a/drivers/clk/mvebu/armada-37xx-periph.c b/drivers/clk/mvebu/armada-37xx-periph.c
index 902a6cc9ef..b1a35968e1 100644
--- a/drivers/clk/mvebu/armada-37xx-periph.c
+++ b/drivers/clk/mvebu/armada-37xx-periph.c
@@ -224,11 +224,21 @@ static const struct clk_periph clks_sb[] = {
 	{ },
 };
 
-static inline int get_mux(struct a37xx_periphclk *priv, int shift)
+static int get_mux(struct a37xx_periphclk *priv, int shift)
 {
 	return (readl(priv->reg + TBG_SEL) >> shift) & 3;
 }
 
+static void set_mux(struct a37xx_periphclk *priv, int shift, int val)
+{
+	u32 reg;
+
+	reg = readl(priv->reg + TBG_SEL);
+	reg &= ~(3 << shift);
+	reg |= (val & 3) << shift;
+	writel(reg, priv->reg + TBG_SEL);
+}
+
 static ulong periph_clk_get_rate(struct a37xx_periphclk *priv, int id);
 
 static ulong get_parent_rate(struct a37xx_periphclk *priv, int id)
@@ -277,6 +287,17 @@ static ulong get_div(struct a37xx_periphclk *priv,
 	return 0;
 }
 
+static void set_div_val(struct a37xx_periphclk *priv,
+			const struct clk_periph *clk, int idx, int val)
+{
+	u32 reg;
+
+	reg = readl(priv->reg + clk->div_reg_off[idx]);
+	reg &= ~(clk->div_mask[idx] << clk->div_shift[idx]);
+	reg |= (val & clk->div_mask[idx]) << clk->div_shift[idx];
+	writel(reg, priv->reg + clk->div_reg_off[idx]);
+}
+
 static ulong periph_clk_get_rate(struct a37xx_periphclk *priv, int id)
 {
 	const struct clk_periph *clk = &priv->clks[id];
@@ -337,6 +358,111 @@ static int armada_37xx_periph_clk_disable(struct clk *clk)
 	return periph_clk_enable(clk, 0);
 }
 
+#define diff(a, b) abs((long)(a) - (long)(b))
+
+static ulong find_best_div(const struct clk_div_table *t0,
+			   const struct clk_div_table *t1, ulong parent_rate,
+			   ulong req_rate, int *v0, int *v1)
+{
+	const struct clk_div_table *i, *j;
+	ulong rate, best_rate = 0;
+
+	for (i = t0; i && i->div; ++i) {
+		for (j = t1; j && j->div; ++j) {
+			rate = DIV_ROUND_UP(parent_rate, i->div * j->div);
+
+			if (!best_rate ||
+			    diff(rate, req_rate) < diff(best_rate, req_rate)) {
+				best_rate = rate;
+				*v0 = i->val;
+				*v1 = j->val;
+			}
+		}
+	}
+
+	return best_rate;
+}
+
+static ulong armada_37xx_periph_clk_set_rate(struct clk *clk, ulong req_rate)
+{
+	struct a37xx_periphclk *priv = dev_get_priv(clk->dev);
+	const struct clk_periph *periph_clk = &priv->clks[clk->id];
+	ulong rate, old_rate, parent_rate;
+	int div_val0 = 0, div_val1 = 0;
+	const struct clk_div_table *t1;
+	static const struct clk_div_table empty_table[2] = {
+		{ 1, 0 },
+		{ 0, 0 }
+	};
+
+	if (clk->id > priv->count)
+		return -EINVAL;
+
+	old_rate = periph_clk_get_rate(priv, clk->id);
+	if (old_rate == -EINVAL)
+		return -EINVAL;
+
+	if (old_rate == req_rate)
+		return old_rate;
+
+	if (!periph_clk->can_gate || !periph_clk->dividers)
+		return -ENOTSUPP;
+
+	parent_rate = get_parent_rate(priv, clk->id);
+	if (parent_rate == -EINVAL)
+		return -EINVAL;
+
+	t1 = empty_table;
+	if (periph_clk->dividers > 1)
+		t1 = periph_clk->div_table[1];
+
+	rate = find_best_div(periph_clk->div_table[0], t1, parent_rate,
+			     req_rate, &div_val0, &div_val1);
+
+	periph_clk_enable(clk, 0);
+
+	set_div_val(priv, periph_clk, 0, div_val0);
+	if (periph_clk->dividers > 1)
+		set_div_val(priv, periph_clk, 1, div_val1);
+
+	periph_clk_enable(clk, 1);
+
+	return rate;
+}
+
+static int armada_37xx_periph_clk_set_parent(struct clk *clk,
+					     struct clk *parent)
+{
+	struct a37xx_periphclk *priv = dev_get_priv(clk->dev);
+	const struct clk_periph *periph_clk = &priv->clks[clk->id];
+	struct clk check_parent;
+	int ret;
+
+	/* We also check if parent is our TBG clock */
+	if (clk->id > priv->count || parent->id >= MAX_TBG_PARENTS)
+		return -EINVAL;
+
+	if (!periph_clk->can_mux || !periph_clk->can_gate)
+		return -ENOTSUPP;
+
+	ret = clk_get_by_index(clk->dev, 0, &check_parent);
+	if (ret < 0)
+		return ret;
+
+	if (parent->dev != check_parent.dev)
+		ret = -EINVAL;
+
+	clk_free(&check_parent);
+	if (ret < 0)
+		return ret;
+
+	periph_clk_enable(clk, 0);
+	set_mux(priv, periph_clk->mux_shift, parent->id);
+	periph_clk_enable(clk, 1);
+
+	return 0;
+}
+
 #if defined(CONFIG_CMD_CLK) && defined(CONFIG_CLK_ARMADA_3720)
 static int armada_37xx_periph_clk_dump(struct udevice *dev)
 {
@@ -473,6 +599,8 @@ static int armada_37xx_periph_clk_probe(struct udevice *dev)
 
 static const struct clk_ops armada_37xx_periph_clk_ops = {
 	.get_rate = armada_37xx_periph_clk_get_rate,
+	.set_rate = armada_37xx_periph_clk_set_rate,
+	.set_parent = armada_37xx_periph_clk_set_parent,
 	.enable = armada_37xx_periph_clk_enable,
 	.disable = armada_37xx_periph_clk_disable,
 };
diff --git a/drivers/mtd/nand/pxa3xx_nand.c b/drivers/mtd/nand/pxa3xx_nand.c
index 9d02fd8eb1..2a02a9d58e 100644
--- a/drivers/mtd/nand/pxa3xx_nand.c
+++ b/drivers/mtd/nand/pxa3xx_nand.c
@@ -24,14 +24,16 @@ DECLARE_GLOBAL_DATA_PTR;
 #define TIMEOUT_DRAIN_FIFO	5	/* in ms */
 #define	CHIP_DELAY_TIMEOUT	200
 #define NAND_STOP_DELAY		40
-#define PAGE_CHUNK_SIZE		(2048)
 
 /*
  * Define a buffer size for the initial command that detects the flash device:
- * STATUS, READID and PARAM. The largest of these is the PARAM command,
- * needing 256 bytes.
+ * STATUS, READID and PARAM.
+ * ONFI param page is 256 bytes, and there are three redundant copies
+ * to be read. JEDEC param page is 512 bytes, and there are also three
+ * redundant copies to be read.
+ * Hence this buffer should be at least 512 x 3. Let's pick 2048.
  */
-#define INIT_BUFFER_SIZE	256
+#define INIT_BUFFER_SIZE	2048
 
 /* registers and bit definitions */
 #define NDCR		(0x00) /* Control register */
@@ -58,7 +60,7 @@ DECLARE_GLOBAL_DATA_PTR;
 #define NDCR_ND_MODE		(0x3 << 21)
 #define NDCR_NAND_MODE		(0x0)
 #define NDCR_CLR_PG_CNT		(0x1 << 20)
-#define NDCR_STOP_ON_UNCOR	(0x1 << 19)
+#define NFCV1_NDCR_ARB_CNTL	(0x1 << 19)
 #define NDCR_RD_ID_CNT_MASK	(0x7 << 16)
 #define NDCR_RD_ID_CNT(x)	(((x) << 16) & NDCR_RD_ID_CNT_MASK)
 
@@ -109,6 +111,13 @@ DECLARE_GLOBAL_DATA_PTR;
 #define EXT_CMD_TYPE_LAST_RW	1 /* Last naked read/write */
 #define EXT_CMD_TYPE_MONO	0 /* Monolithic read/write */
 
+/*
+ * This should be large enough to read 'ONFI' and 'JEDEC'.
+ * Let's use 7 bytes, which is the maximum ID count supported
+ * by the controller (see NDCR_RD_ID_CNT_MASK).
+ */
+#define READ_ID_BYTES		7
+
 /* macros for registers read/write */
 #define nand_writel(info, off, val)	\
 	writel((val), (info)->mmio_base + (off))
@@ -146,7 +155,6 @@ enum pxa3xx_nand_variant {
 
 struct pxa3xx_nand_host {
 	struct nand_chip	chip;
-	struct mtd_info         *mtd;
 	void			*info_data;
 
 	/* page size of attached chip */
@@ -156,8 +164,6 @@ struct pxa3xx_nand_host {
 	/* calculated from pxa3xx_nand_flash data */
 	unsigned int		col_addr_cycles;
 	unsigned int		row_addr_cycles;
-	size_t			read_id_bytes;
-
 };
 
 struct pxa3xx_nand_info {
@@ -193,15 +199,44 @@ struct pxa3xx_nand_info {
 	int			use_spare;	/* use spare ? */
 	int			need_wait;
 
-	unsigned int		data_size;	/* data to be read from FIFO */
-	unsigned int		chunk_size;	/* split commands chunk size */
-	unsigned int		oob_size;
+	/* Amount of real data per full chunk */
+	unsigned int		chunk_size;
+
+	/* Amount of spare data per full chunk */
 	unsigned int		spare_size;
+
+	/* Number of full chunks (i.e chunk_size + spare_size) */
+	unsigned int            nfullchunks;
+
+	/*
+	 * Total number of chunks. If equal to nfullchunks, then there
+	 * are only full chunks. Otherwise, there is one last chunk of
+	 * size (last_chunk_size + last_spare_size)
+	 */
+	unsigned int            ntotalchunks;
+
+	/* Amount of real data in the last chunk */
+	unsigned int		last_chunk_size;
+
+	/* Amount of spare data in the last chunk */
+	unsigned int		last_spare_size;
+
 	unsigned int		ecc_size;
 	unsigned int		ecc_err_cnt;
 	unsigned int		max_bitflips;
 	int			retcode;
 
+	/*
+	 * Variables only valid during command
+	 * execution. step_chunk_size and step_spare_size is the
+	 * amount of real data and spare data in the current
+	 * chunk. cur_chunk is the current chunk being
+	 * read/programmed.
+	 */
+	unsigned int		step_chunk_size;
+	unsigned int		step_spare_size;
+	unsigned int            cur_chunk;
+
 	/* cached register value */
 	uint32_t		reg_ndcr;
 	uint32_t		ndtr0cs0;
@@ -215,13 +250,33 @@ struct pxa3xx_nand_info {
 };
 
 static struct pxa3xx_nand_timing timing[] = {
+	/*
+	 * tCH	Enable signal hold time
+	 * tCS	Enable signal setup time
+	 * tWH	ND_nWE high duration
+	 * tWP	ND_nWE pulse time
+	 * tRH	ND_nRE high duration
+	 * tRP	ND_nRE pulse width
+	 * tR	ND_nWE high to ND_nRE low for read
+	 * tWHR	ND_nWE high to ND_nRE low for status read
+	 * tAR	ND_ALE low to ND_nRE low delay
+	 */
+	/*ch  cs  wh  wp   rh  rp   r      whr  ar */
 	{ 40, 80, 60, 100, 80, 100, 90000, 400, 40, },
 	{ 10,  0, 20,  40, 30,  40, 11123, 110, 10, },
 	{ 10, 25, 15,  25, 15,  30, 25000,  60, 10, },
 	{ 10, 35, 15,  25, 15,  25, 25000,  60, 10, },
+	{  5, 20, 10,  12, 10,  12, 25000,  60, 10, },
 };
 
 static struct pxa3xx_nand_flash builtin_flash_types[] = {
+	/*
+	 * chip_id
+	 * flash_width	Width of Flash memory (DWIDTH_M)
+	 * dfc_width	Width of flash controller(DWIDTH_C)
+	 * *timing
+	 * http://www.linux-mtd.infradead.org/nand-data/nanddata.html
+	 */
 	{ 0x46ec, 16, 16, &timing[1] },
 	{ 0xdaec,  8,  8, &timing[1] },
 	{ 0xd7ec,  8,  8, &timing[1] },
@@ -230,6 +285,7 @@ static struct pxa3xx_nand_flash builtin_flash_types[] = {
 	{ 0xdc2c,  8,  8, &timing[2] },
 	{ 0xcc2c, 16, 16, &timing[2] },
 	{ 0xba20, 16, 16, &timing[3] },
+	{ 0xda98,  8,  8, &timing[4] },
 };
 
 #ifdef CONFIG_SYS_NAND_USE_FLASH_BBT
@@ -267,6 +323,20 @@ static struct nand_ecclayout ecc_layout_2KB_bch4bit = {
 	.oobfree = { {2, 30} }
 };
 
+static struct nand_ecclayout ecc_layout_2KB_bch8bit = {
+	.eccbytes = 64,
+	.eccpos = {
+		64,  65,  66,  67,  68,  69,  70,  71,
+		72,  73,  74,  75,  76,  77,  78,  79,
+		80,  81,  82,  83,  84,  85,  86,  87,
+		88,  89,  90,  91,  92,  93,  94,  95,
+		96,  97,  98,  99,  100, 101, 102, 103,
+		104, 105, 106, 107, 108, 109, 110, 111,
+		112, 113, 114, 115, 116, 117, 118, 119,
+		120, 121, 122, 123, 124, 125, 126, 127},
+	.oobfree = { {1, 4}, {6, 26} }
+};
+
 static struct nand_ecclayout ecc_layout_4KB_bch4bit = {
 	.eccbytes = 64,
 	.eccpos = {
@@ -282,6 +352,33 @@ static struct nand_ecclayout ecc_layout_4KB_bch4bit = {
 	.oobfree = { {6, 26}, { 64, 32} }
 };
 
+static struct nand_ecclayout ecc_layout_8KB_bch4bit = {
+	.eccbytes = 128,
+	.eccpos = {
+		32,  33,  34,  35,  36,  37,  38,  39,
+		40,  41,  42,  43,  44,  45,  46,  47,
+		48,  49,  50,  51,  52,  53,  54,  55,
+		56,  57,  58,  59,  60,  61,  62,  63,
+
+		96,  97,  98,  99,  100, 101, 102, 103,
+		104, 105, 106, 107, 108, 109, 110, 111,
+		112, 113, 114, 115, 116, 117, 118, 119,
+		120, 121, 122, 123, 124, 125, 126, 127,
+
+		160, 161, 162, 163, 164, 165, 166, 167,
+		168, 169, 170, 171, 172, 173, 174, 175,
+		176, 177, 178, 179, 180, 181, 182, 183,
+		184, 185, 186, 187, 188, 189, 190, 191,
+
+		224, 225, 226, 227, 228, 229, 230, 231,
+		232, 233, 234, 235, 236, 237, 238, 239,
+		240, 241, 242, 243, 244, 245, 246, 247,
+		248, 249, 250, 251, 252, 253, 254, 255},
+
+	/* Bootrom looks in bytes 0 & 5 for bad blocks */
+	.oobfree = { {1, 4}, {6, 26}, { 64, 32}, {128, 32}, {192, 32} }
+};
+
 static struct nand_ecclayout ecc_layout_4KB_bch8bit = {
 	.eccbytes = 128,
 	.eccpos = {
@@ -292,6 +389,13 @@ static struct nand_ecclayout ecc_layout_4KB_bch8bit = {
 	.oobfree = { }
 };
 
+static struct nand_ecclayout ecc_layout_8KB_bch8bit = {
+	.eccbytes = 256,
+	.eccpos = {},
+	/* HW ECC handles all ECC data and all spare area is free for OOB */
+	.oobfree = {{0, 160} }
+};
+
 #define NDTR0_tCH(c)	(min((c), 7) << 19)
 #define NDTR0_tCS(c)	(min((c), 7) << 16)
 #define NDTR0_tWH(c)	(min((c), 7) << 11)
@@ -347,9 +451,9 @@ static void pxa3xx_nand_set_sdr_timing(struct pxa3xx_nand_host *host,
 	u32 tCH_min = DIV_ROUND_UP(t->tCH_min, 1000);
 	u32 tCS_min = DIV_ROUND_UP(t->tCS_min, 1000);
 	u32 tWH_min = DIV_ROUND_UP(t->tWH_min, 1000);
-	u32 tWP_min = DIV_ROUND_UP(t->tWC_min - tWH_min, 1000);
+	u32 tWP_min = DIV_ROUND_UP(t->tWC_min - t->tWH_min, 1000);
 	u32 tREH_min = DIV_ROUND_UP(t->tREH_min, 1000);
-	u32 tRP_min = DIV_ROUND_UP(t->tRC_min - tREH_min, 1000);
+	u32 tRP_min = DIV_ROUND_UP(t->tRC_min - t->tREH_min, 1000);
 	u32 tR = chip->chip_delay * 1000;
 	u32 tWHR_min = DIV_ROUND_UP(t->tWHR_min, 1000);
 	u32 tAR_min = DIV_ROUND_UP(t->tAR_min, 1000);
@@ -381,16 +485,17 @@ static int pxa3xx_nand_init_timings(struct pxa3xx_nand_host *host)
 	struct nand_chip *chip = &host->chip;
 	struct pxa3xx_nand_info *info = host->info_data;
 	const struct pxa3xx_nand_flash *f = NULL;
+	struct mtd_info *mtd = nand_to_mtd(&host->chip);
 	int mode, id, ntypes, i;
 
 	mode = onfi_get_async_timing_mode(chip);
 	if (mode == ONFI_TIMING_MODE_UNKNOWN) {
 		ntypes = ARRAY_SIZE(builtin_flash_types);
 
-		chip->cmdfunc(host->mtd, NAND_CMD_READID, 0x00, -1);
+		chip->cmdfunc(mtd, NAND_CMD_READID, 0x00, -1);
 
-		id = chip->read_byte(host->mtd);
-		id |= chip->read_byte(host->mtd) << 0x8;
+		id = chip->read_byte(mtd);
+		id |= chip->read_byte(mtd) << 0x8;
 
 		for (i = 0; i < ntypes; i++) {
 			f = &builtin_flash_types[i];
@@ -427,25 +532,6 @@ static int pxa3xx_nand_init_timings(struct pxa3xx_nand_host *host)
 	return 0;
 }
 
-/*
- * Set the data and OOB size, depending on the selected
- * spare and ECC configuration.
- * Only applicable to READ0, READOOB and PAGEPROG commands.
- */
-static void pxa3xx_set_datasize(struct pxa3xx_nand_info *info,
-				struct mtd_info *mtd)
-{
-	int oob_enable = info->reg_ndcr & NDCR_SPARE_EN;
-
-	info->data_size = mtd->writesize;
-	if (!oob_enable)
-		return;
-
-	info->oob_size = info->spare_size;
-	if (!info->use_ecc)
-		info->oob_size += info->ecc_size;
-}
-
 /**
  * NOTE: it is a must to set ND_RUN first, then write
  * command buffer, otherwise, it does not work.
@@ -478,8 +564,8 @@ static void pxa3xx_nand_start(struct pxa3xx_nand_info *info)
 	ndcr |= NDCR_ND_RUN;
 
 	/* clear status bits and run */
-	nand_writel(info, NDCR, 0);
 	nand_writel(info, NDSR, NDSR_MASK);
+	nand_writel(info, NDCR, 0);
 	nand_writel(info, NDCR, ndcr);
 }
 
@@ -526,39 +612,38 @@ static void drain_fifo(struct pxa3xx_nand_info *info, void *data, int len)
 
 static void handle_data_pio(struct pxa3xx_nand_info *info)
 {
-	unsigned int do_bytes = min(info->data_size, info->chunk_size);
-
 	switch (info->state) {
 	case STATE_PIO_WRITING:
-		writesl(info->mmio_base + NDDB,
-			info->data_buff + info->data_buff_pos,
-			DIV_ROUND_UP(do_bytes, 4));
+		if (info->step_chunk_size)
+			writesl(info->mmio_base + NDDB,
+				info->data_buff + info->data_buff_pos,
+				DIV_ROUND_UP(info->step_chunk_size, 4));
 
-		if (info->oob_size > 0)
+		if (info->step_spare_size)
 			writesl(info->mmio_base + NDDB,
 				info->oob_buff + info->oob_buff_pos,
-				DIV_ROUND_UP(info->oob_size, 4));
+				DIV_ROUND_UP(info->step_spare_size, 4));
 		break;
 	case STATE_PIO_READING:
-		drain_fifo(info,
-			   info->data_buff + info->data_buff_pos,
-			   DIV_ROUND_UP(do_bytes, 4));
+		if (info->step_chunk_size)
+			drain_fifo(info,
+				   info->data_buff + info->data_buff_pos,
+				   DIV_ROUND_UP(info->step_chunk_size, 4));
 
-		if (info->oob_size > 0)
+		if (info->step_spare_size)
 			drain_fifo(info,
 				   info->oob_buff + info->oob_buff_pos,
-				   DIV_ROUND_UP(info->oob_size, 4));
+				   DIV_ROUND_UP(info->step_spare_size, 4));
 		break;
 	default:
 		dev_err(&info->pdev->dev, "%s: invalid state %d\n", __func__,
-			info->state);
+				info->state);
 		BUG();
 	}
 
 	/* Update buffer pointers for multi-page read/write */
-	info->data_buff_pos += do_bytes;
-	info->oob_buff_pos += info->oob_size;
-	info->data_size -= do_bytes;
+	info->data_buff_pos += info->step_chunk_size;
+	info->oob_buff_pos += info->step_spare_size;
 }
 
 static void pxa3xx_nand_irq_thread(struct pxa3xx_nand_info *info)
@@ -583,6 +668,9 @@ static irqreturn_t pxa3xx_nand_irq(struct pxa3xx_nand_info *info)
 		cmd_done        = NDSR_CS1_CMDD;
 	}
 
+	/* TODO - find out why we need the delay during write operation. */
+	ndelay(1);
+
 	status = nand_readl(info, NDSR);
 
 	if (status & NDSR_UNCORERR)
@@ -620,8 +708,14 @@ static irqreturn_t pxa3xx_nand_irq(struct pxa3xx_nand_info *info)
 		is_ready = 1;
 	}
 
+	/*
+	 * Clear all status bit before issuing the next command, which
+	 * can and will alter the status bits and will deserve a new
+	 * interrupt on its own. This lets the controller exit the IRQ
+	 */
+	nand_writel(info, NDSR, status);
+
 	if (status & NDSR_WRCMDREQ) {
-		nand_writel(info, NDSR, NDSR_WRCMDREQ);
 		status &= ~NDSR_WRCMDREQ;
 		info->state = STATE_CMD_HANDLE;
 
@@ -642,8 +736,6 @@ static irqreturn_t pxa3xx_nand_irq(struct pxa3xx_nand_info *info)
 			nand_writel(info, NDCB0, info->ndcb3);
 	}
 
-	/* clear NDSR to let the controller exit the IRQ */
-	nand_writel(info, NDSR, status);
 	if (is_completed)
 		info->cmd_complete = 1;
 	if (is_ready)
@@ -664,7 +756,7 @@ static void set_command_address(struct pxa3xx_nand_info *info,
 		unsigned int page_size, uint16_t column, int page_addr)
 {
 	/* small page addr setting */
-	if (page_size < PAGE_CHUNK_SIZE) {
+	if (page_size < info->chunk_size) {
 		info->ndcb1 = ((page_addr & 0xFFFFFF) << 8)
 				| (column & 0xFF);
 
@@ -683,14 +775,16 @@ static void set_command_address(struct pxa3xx_nand_info *info,
 static void prepare_start_command(struct pxa3xx_nand_info *info, int command)
 {
 	struct pxa3xx_nand_host *host = info->host[info->cs];
-	struct mtd_info *mtd = host->mtd;
+	struct mtd_info *mtd = nand_to_mtd(&host->chip);
 
 	/* reset data and oob column point to handle data */
 	info->buf_start		= 0;
 	info->buf_count		= 0;
-	info->oob_size		= 0;
 	info->data_buff_pos	= 0;
 	info->oob_buff_pos	= 0;
+	info->step_chunk_size   = 0;
+	info->step_spare_size   = 0;
+	info->cur_chunk         = 0;
 	info->use_ecc		= 0;
 	info->use_spare		= 1;
 	info->retcode		= ERR_NONE;
@@ -700,10 +794,9 @@ static void prepare_start_command(struct pxa3xx_nand_info *info, int command)
 
 	switch (command) {
 	case NAND_CMD_READ0:
+	case NAND_CMD_READOOB:
 	case NAND_CMD_PAGEPROG:
 		info->use_ecc = 1;
-	case NAND_CMD_READOOB:
-		pxa3xx_set_datasize(info, mtd);
 		break;
 	case NAND_CMD_PARAM:
 		info->use_spare = 0;
@@ -734,7 +827,7 @@ static int prepare_set_command(struct pxa3xx_nand_info *info, int command,
 	struct mtd_info *mtd;
 
 	host = info->host[info->cs];
-	mtd = host->mtd;
+	mtd = nand_to_mtd(&host->chip);
 	addr_cycle = 0;
 	exec_cmd = 1;
 
@@ -760,19 +853,27 @@ static int prepare_set_command(struct pxa3xx_nand_info *info, int command,
 		if (command == NAND_CMD_READOOB)
 			info->buf_start += mtd->writesize;
 
+		if (info->cur_chunk < info->nfullchunks) {
+			info->step_chunk_size = info->chunk_size;
+			info->step_spare_size = info->spare_size;
+		} else {
+			info->step_chunk_size = info->last_chunk_size;
+			info->step_spare_size = info->last_spare_size;
+		}
+
 		/*
 		 * Multiple page read needs an 'extended command type' field,
 		 * which is either naked-read or last-read according to the
 		 * state.
 		 */
-		if (mtd->writesize == PAGE_CHUNK_SIZE) {
+		if (mtd->writesize == info->chunk_size) {
 			info->ndcb0 |= NDCB0_DBC | (NAND_CMD_READSTART << 8);
-		} else if (mtd->writesize > PAGE_CHUNK_SIZE) {
+		} else if (mtd->writesize > info->chunk_size) {
 			info->ndcb0 |= NDCB0_DBC | (NAND_CMD_READSTART << 8)
 					| NDCB0_LEN_OVRD
 					| NDCB0_EXT_CMD_TYPE(ext_cmd_type);
-			info->ndcb3 = info->chunk_size +
-				      info->oob_size;
+			info->ndcb3 = info->step_chunk_size +
+				info->step_spare_size;
 		}
 
 		set_command_address(info, mtd->writesize, column, page_addr);
@@ -787,13 +888,11 @@ static int prepare_set_command(struct pxa3xx_nand_info *info, int command,
 		 * Multiple page programming needs to execute the initial
 		 * SEQIN command that sets the page address.
 		 */
-		if (mtd->writesize > PAGE_CHUNK_SIZE) {
+		if (mtd->writesize > info->chunk_size) {
 			info->ndcb0 |= NDCB0_CMD_TYPE(0x1)
 				| NDCB0_EXT_CMD_TYPE(ext_cmd_type)
 				| addr_cycle
 				| command;
-			/* No data transfer in this case */
-			info->data_size = 0;
 			exec_cmd = 1;
 		}
 		break;
@@ -805,8 +904,16 @@ static int prepare_set_command(struct pxa3xx_nand_info *info, int command,
 			break;
 		}
 
+		if (info->cur_chunk < info->nfullchunks) {
+			info->step_chunk_size = info->chunk_size;
+			info->step_spare_size = info->spare_size;
+		} else {
+			info->step_chunk_size = info->last_chunk_size;
+			info->step_spare_size = info->last_spare_size;
+		}
+
 		/* Second command setting for large pages */
-		if (mtd->writesize > PAGE_CHUNK_SIZE) {
+		if (mtd->writesize > info->chunk_size) {
 			/*
 			 * Multiple page write uses the 'extended command'
 			 * field. This can be used to issue a command dispatch
@@ -815,14 +922,14 @@ static int prepare_set_command(struct pxa3xx_nand_info *info, int command,
 			info->ndcb0 |= NDCB0_CMD_TYPE(0x1)
 					| NDCB0_LEN_OVRD
 					| NDCB0_EXT_CMD_TYPE(ext_cmd_type);
-			info->ndcb3 = info->chunk_size +
-				      info->oob_size;
+			info->ndcb3 = info->step_chunk_size +
+				      info->step_spare_size;
 
 			/*
 			 * This is the command dispatch that completes a chunked
 			 * page program operation.
 			 */
-			if (info->data_size == 0) {
+			if (info->cur_chunk == info->ntotalchunks) {
 				info->ndcb0 = NDCB0_CMD_TYPE(0x1)
 					| NDCB0_EXT_CMD_TYPE(ext_cmd_type)
 					| command;
@@ -842,24 +949,24 @@ static int prepare_set_command(struct pxa3xx_nand_info *info, int command,
 		break;
 
 	case NAND_CMD_PARAM:
-		info->buf_count = 256;
+		info->buf_count = INIT_BUFFER_SIZE;
 		info->ndcb0 |= NDCB0_CMD_TYPE(0)
 				| NDCB0_ADDR_CYC(1)
 				| NDCB0_LEN_OVRD
 				| command;
 		info->ndcb1 = (column & 0xFF);
-		info->ndcb3 = 256;
-		info->data_size = 256;
+		info->ndcb3 = INIT_BUFFER_SIZE;
+		info->step_chunk_size = INIT_BUFFER_SIZE;
 		break;
 
 	case NAND_CMD_READID:
-		info->buf_count = host->read_id_bytes;
+		info->buf_count = READ_ID_BYTES;
 		info->ndcb0 |= NDCB0_CMD_TYPE(3)
 				| NDCB0_ADDR_CYC(1)
 				| command;
 		info->ndcb1 = (column & 0xFF);
 
-		info->data_size = 8;
+		info->step_chunk_size = 8;
 		break;
 	case NAND_CMD_STATUS:
 		info->buf_count = 1;
@@ -867,7 +974,7 @@ static int prepare_set_command(struct pxa3xx_nand_info *info, int command,
 				| NDCB0_ADDR_CYC(1)
 				| command;
 
-		info->data_size = 8;
+		info->step_chunk_size = 8;
 		break;
 
 	case NAND_CMD_ERASE1:
@@ -1051,22 +1158,31 @@ static void nand_cmdfunc_extended(struct mtd_info *mtd,
 			}
 		}
 
+		/* Only a few commands need several steps */
+		if (command != NAND_CMD_PAGEPROG &&
+		    command != NAND_CMD_READ0    &&
+		    command != NAND_CMD_READOOB)
+			break;
+
+		info->cur_chunk++;
+
 		/* Check if the sequence is complete */
-		if (info->data_size == 0 && command != NAND_CMD_PAGEPROG)
+		if (info->cur_chunk == info->ntotalchunks &&
+		    command != NAND_CMD_PAGEPROG)
 			break;
 
 		/*
 		 * After a splitted program command sequence has issued
 		 * the command dispatch, the command sequence is complete.
 		 */
-		if (info->data_size == 0 &&
+		if (info->cur_chunk == (info->ntotalchunks + 1) &&
 		    command == NAND_CMD_PAGEPROG &&
 		    ext_cmd_type == EXT_CMD_TYPE_DISPATCH)
 			break;
 
 		if (command == NAND_CMD_READ0 || command == NAND_CMD_READOOB) {
 			/* Last read: issue a 'last naked read' */
-			if (info->data_size == info->chunk_size)
+			if (info->cur_chunk == info->ntotalchunks - 1)
 				ext_cmd_type = EXT_CMD_TYPE_LAST_RW;
 			else
 				ext_cmd_type = EXT_CMD_TYPE_NAKED_RW;
@@ -1076,7 +1192,7 @@ static void nand_cmdfunc_extended(struct mtd_info *mtd,
 		 * the command dispatch must be issued to complete.
 		 */
 		} else if (command == NAND_CMD_PAGEPROG &&
-			   info->data_size == 0) {
+			   info->cur_chunk == info->ntotalchunks) {
 				ext_cmd_type = EXT_CMD_TYPE_DISPATCH;
 		}
 	} while (1);
@@ -1218,42 +1334,42 @@ static int pxa3xx_nand_waitfunc(struct mtd_info *mtd, struct nand_chip *this)
 	return NAND_STATUS_READY;
 }
 
-static int pxa3xx_nand_config_flash(struct pxa3xx_nand_info *info)
+static int pxa3xx_nand_config_ident(struct pxa3xx_nand_info *info)
+{
+	struct pxa3xx_nand_platform_data *pdata = info->pdata;
+
+	/* Configure default flash values */
+	info->reg_ndcr = 0x0; /* enable all interrupts */
+	info->reg_ndcr |= (pdata->enable_arbiter) ? NDCR_ND_ARB_EN : 0;
+	info->reg_ndcr |= NDCR_RD_ID_CNT(READ_ID_BYTES);
+	info->reg_ndcr |= NDCR_SPARE_EN;
+
+	return 0;
+}
+
+static void pxa3xx_nand_config_tail(struct pxa3xx_nand_info *info)
 {
 	struct pxa3xx_nand_host *host = info->host[info->cs];
-	struct mtd_info *mtd = host->mtd;
+	struct mtd_info *mtd = nand_to_mtd(&info->host[info->cs]->chip);
 	struct nand_chip *chip = mtd_to_nand(mtd);
 
 	info->reg_ndcr |= (host->col_addr_cycles == 2) ? NDCR_RA_START : 0;
 	info->reg_ndcr |= (chip->page_shift == 6) ? NDCR_PG_PER_BLK : 0;
 	info->reg_ndcr |= (mtd->writesize == 2048) ? NDCR_PAGE_SZ : 0;
-
-	return 0;
 }
 
-static int pxa3xx_nand_detect_config(struct pxa3xx_nand_info *info)
+static void pxa3xx_nand_detect_config(struct pxa3xx_nand_info *info)
 {
-	/*
-	 * We set 0 by hard coding here, for we don't support keep_config
-	 * when there is more than one chip attached to the controller
-	 */
-	struct pxa3xx_nand_host *host = info->host[0];
+	struct pxa3xx_nand_platform_data *pdata = info->pdata;
 	uint32_t ndcr = nand_readl(info, NDCR);
 
-	if (ndcr & NDCR_PAGE_SZ) {
-		/* Controller's FIFO size */
-		info->chunk_size = 2048;
-		host->read_id_bytes = 4;
-	} else {
-		info->chunk_size = 512;
-		host->read_id_bytes = 2;
-	}
-
 	/* Set an initial chunk size */
-	info->reg_ndcr = ndcr & ~NDCR_INT_MASK;
+	info->chunk_size = ndcr & NDCR_PAGE_SZ ? 2048 : 512;
+	info->reg_ndcr = ndcr &
+		~(NDCR_INT_MASK | NDCR_ND_ARB_EN | NFCV1_NDCR_ARB_CNTL);
+	info->reg_ndcr |= (pdata->enable_arbiter) ? NDCR_ND_ARB_EN : 0;
 	info->ndtr0cs0 = nand_readl(info, NDTR0CS0);
 	info->ndtr1cs0 = nand_readl(info, NDTR1CS0);
-	return 0;
 }
 
 static int pxa3xx_nand_init_buff(struct pxa3xx_nand_info *info)
@@ -1273,13 +1389,13 @@ static int pxa3xx_nand_sensing(struct pxa3xx_nand_host *host)
 	const struct nand_sdr_timings *timings;
 	int ret;
 
-	mtd = info->host[info->cs]->mtd;
+	mtd = nand_to_mtd(&info->host[info->cs]->chip);
 	chip = mtd_to_nand(mtd);
 
 	/* configure default flash values */
 	info->reg_ndcr = 0x0; /* enable all interrupts */
 	info->reg_ndcr |= (pdata->enable_arbiter) ? NDCR_ND_ARB_EN : 0;
-	info->reg_ndcr |= NDCR_RD_ID_CNT(host->read_id_bytes);
+	info->reg_ndcr |= NDCR_RD_ID_CNT(READ_ID_BYTES);
 	info->reg_ndcr |= NDCR_SPARE_EN; /* enable spare by default */
 
 	/* use the common timing to make a try */
@@ -1302,6 +1418,8 @@ static int pxa_ecc_init(struct pxa3xx_nand_info *info,
 			int strength, int ecc_stepsize, int page_size)
 {
 	if (strength == 1 && ecc_stepsize == 512 && page_size == 2048) {
+		info->nfullchunks = 1;
+		info->ntotalchunks = 1;
 		info->chunk_size = 2048;
 		info->spare_size = 40;
 		info->ecc_size = 24;
@@ -1310,6 +1428,8 @@ static int pxa_ecc_init(struct pxa3xx_nand_info *info,
 		ecc->strength = 1;
 
 	} else if (strength == 1 && ecc_stepsize == 512 && page_size == 512) {
+		info->nfullchunks = 1;
+		info->ntotalchunks = 1;
 		info->chunk_size = 512;
 		info->spare_size = 8;
 		info->ecc_size = 8;
@@ -1323,6 +1443,8 @@ static int pxa_ecc_init(struct pxa3xx_nand_info *info,
 	 */
 	} else if (strength == 4 && ecc_stepsize == 512 && page_size == 2048) {
 		info->ecc_bch = 1;
+		info->nfullchunks = 1;
+		info->ntotalchunks = 1;
 		info->chunk_size = 2048;
 		info->spare_size = 32;
 		info->ecc_size = 32;
@@ -1333,6 +1455,8 @@ static int pxa_ecc_init(struct pxa3xx_nand_info *info,
 
 	} else if (strength == 4 && ecc_stepsize == 512 && page_size == 4096) {
 		info->ecc_bch = 1;
+		info->nfullchunks = 2;
+		info->ntotalchunks = 2;
 		info->chunk_size = 2048;
 		info->spare_size = 32;
 		info->ecc_size = 32;
@@ -1341,19 +1465,64 @@ static int pxa_ecc_init(struct pxa3xx_nand_info *info,
 		ecc->layout = &ecc_layout_4KB_bch4bit;
 		ecc->strength = 16;
 
+	} else if (strength == 4 && ecc_stepsize == 512 && page_size == 8192) {
+		info->ecc_bch = 1;
+		info->nfullchunks = 4;
+		info->ntotalchunks = 4;
+		info->chunk_size = 2048;
+		info->spare_size = 32;
+		info->ecc_size = 32;
+		ecc->mode = NAND_ECC_HW;
+		ecc->size = info->chunk_size;
+		ecc->layout = &ecc_layout_8KB_bch4bit;
+		ecc->strength = 16;
+
 	/*
 	 * Required ECC: 8-bit correction per 512 bytes
 	 * Select: 16-bit correction per 1024 bytes
 	 */
-	} else if (strength == 8 && ecc_stepsize == 512 && page_size == 4096) {
+	} else if (strength == 8 && ecc_stepsize == 512 && page_size == 2048) {
 		info->ecc_bch = 1;
+		info->nfullchunks = 1;
+		info->ntotalchunks = 2;
 		info->chunk_size = 1024;
 		info->spare_size = 0;
+		info->last_chunk_size = 1024;
+		info->last_spare_size = 64;
+		info->ecc_size = 32;
+		ecc->mode = NAND_ECC_HW;
+		ecc->size = info->chunk_size;
+		ecc->layout = &ecc_layout_2KB_bch8bit;
+		ecc->strength = 16;
+
+	} else if (strength == 8 && ecc_stepsize == 512 && page_size == 4096) {
+		info->ecc_bch = 1;
+		info->nfullchunks = 4;
+		info->ntotalchunks = 5;
+		info->chunk_size = 1024;
+		info->spare_size = 0;
+		info->last_chunk_size = 0;
+		info->last_spare_size = 64;
 		info->ecc_size = 32;
 		ecc->mode = NAND_ECC_HW;
 		ecc->size = info->chunk_size;
 		ecc->layout = &ecc_layout_4KB_bch8bit;
 		ecc->strength = 16;
+
+	} else if (strength == 8 && ecc_stepsize == 512 && page_size == 8192) {
+		info->ecc_bch = 1;
+		info->nfullchunks = 8;
+		info->ntotalchunks = 9;
+		info->chunk_size = 1024;
+		info->spare_size = 0;
+		info->last_chunk_size = 0;
+		info->last_spare_size = 160;
+		info->ecc_size = 32;
+		ecc->mode = NAND_ECC_HW;
+		ecc->size = info->chunk_size;
+		ecc->layout = &ecc_layout_8KB_bch8bit;
+		ecc->strength = 16;
+
 	} else {
 		dev_err(&info->pdev->dev,
 			"ECC strength %d at page size %d is not supported\n",
@@ -1373,21 +1542,21 @@ static int pxa3xx_nand_scan(struct mtd_info *mtd)
 	int ret;
 	uint16_t ecc_strength, ecc_step;
 
-	if (pdata->keep_config && !pxa3xx_nand_detect_config(info))
-		goto KEEP_CONFIG;
-
-	/* Set a default chunk size */
-	info->chunk_size = 512;
-
-	ret = pxa3xx_nand_sensing(host);
-	if (ret) {
-		dev_info(&info->pdev->dev, "There is no chip on cs %d!\n",
-			 info->cs);
-
-		return ret;
+	if (pdata->keep_config) {
+		pxa3xx_nand_detect_config(info);
+	} else {
+		ret = pxa3xx_nand_config_ident(info);
+		if (ret)
+			return ret;
+		ret = pxa3xx_nand_sensing(host);
+		if (ret) {
+			dev_info(&info->pdev->dev,
+				 "There is no chip on cs %d!\n",
+				 info->cs);
+			return ret;
+		}
 	}
 
-KEEP_CONFIG:
 	/* Device detection must be done with ECC disabled */
 	if (info->variant == PXA3XX_NAND_VARIANT_ARMADA370)
 		nand_writel(info, NDECCCTRL, 0x0);
@@ -1404,10 +1573,6 @@ KEEP_CONFIG:
 		}
 	}
 
-	ret = pxa3xx_nand_config_flash(info);
-	if (ret)
-		return ret;
-
 #ifdef CONFIG_SYS_NAND_USE_FLASH_BBT
 	/*
 	 * We'll use a bad block table stored in-flash and don't
@@ -1418,21 +1583,6 @@ KEEP_CONFIG:
 	chip->bbt_md = &bbt_mirror_descr;
 #endif
 
-	/*
-	 * If the page size is bigger than the FIFO size, let's check
-	 * we are given the right variant and then switch to the extended
-	 * (aka splitted) command handling,
-	 */
-	if (mtd->writesize > PAGE_CHUNK_SIZE) {
-		if (info->variant == PXA3XX_NAND_VARIANT_ARMADA370) {
-			chip->cmdfunc = nand_cmdfunc_extended;
-		} else {
-			dev_err(&info->pdev->dev,
-				"unsupported page size on this variant\n");
-			return -ENODEV;
-		}
-	}
-
 	if (pdata->ecc_strength && pdata->ecc_step_size) {
 		ecc_strength = pdata->ecc_strength;
 		ecc_step = pdata->ecc_step_size;
@@ -1452,6 +1602,21 @@ KEEP_CONFIG:
 	if (ret)
 		return ret;
 
+	/*
+	 * If the page size is bigger than the FIFO size, let's check
+	 * we are given the right variant and then switch to the extended
+	 * (aka split) command handling,
+	 */
+	if (mtd->writesize > info->chunk_size) {
+		if (info->variant == PXA3XX_NAND_VARIANT_ARMADA370) {
+			chip->cmdfunc = nand_cmdfunc_extended;
+		} else {
+			dev_err(&info->pdev->dev,
+				"unsupported page size on this variant\n");
+			return -ENODEV;
+		}
+	}
+
 	/* calculate addressing information */
 	if (mtd->writesize >= 2048)
 		host->col_addr_cycles = 2;
@@ -1472,6 +1637,10 @@ KEEP_CONFIG:
 		host->row_addr_cycles = 3;
 	else
 		host->row_addr_cycles = 2;
+
+	if (!pdata->keep_config)
+		pxa3xx_nand_config_tail(info);
+
 	return nand_scan_tail(mtd);
 }
 
@@ -1494,10 +1663,8 @@ static int alloc_nand_resource(struct pxa3xx_nand_info *info)
 		mtd = nand_to_mtd(chip);
 		host = (struct pxa3xx_nand_host *)chip;
 		info->host[cs] = host;
-		host->mtd = mtd;
 		host->cs = cs;
 		host->info_data = info;
-		host->read_id_bytes = 4;
 		mtd->owner = THIS_MODULE;
 
 		nand_set_controller_data(chip, host);
@@ -1612,7 +1779,7 @@ static int pxa3xx_nand_probe(struct pxa3xx_nand_info *info)
 
 	probe_success = 0;
 	for (cs = 0; cs < pdata->num_cs; cs++) {
-		struct mtd_info *mtd = info->host[cs]->mtd;
+		struct mtd_info *mtd = nand_to_mtd(&info->host[cs]->chip);
 
 		/*
 		 * The mtd name matches the one used in 'mtdparts' kernel
diff --git a/drivers/phy/marvell/comphy_a3700.h b/drivers/phy/marvell/comphy_a3700.h
index a14767d809..b0941ffb37 100644
--- a/drivers/phy/marvell/comphy_a3700.h
+++ b/drivers/phy/marvell/comphy_a3700.h
@@ -6,7 +6,7 @@
 #ifndef _COMPHY_A3700_H_
 #define _COMPHY_A3700_H_
 
-#include "comphy.h"
+#include "comphy_core.h"
 #include "comphy_hpipe.h"
 
 #define MVEBU_REG(offs)			\
diff --git a/drivers/phy/marvell/comphy_core.c b/drivers/phy/marvell/comphy_core.c
index c6e2cc8897..9c24692629 100644
--- a/drivers/phy/marvell/comphy_core.c
+++ b/drivers/phy/marvell/comphy_core.c
@@ -11,7 +11,7 @@
 #include <linux/errno.h>
 #include <asm/io.h>
 
-#include "comphy.h"
+#include "comphy_core.h"
 
 #define COMPHY_MAX_CHIP 4
 
@@ -66,6 +66,11 @@ void comphy_print(struct chip_serdes_phy_config *chip_cfg,
 	}
 }
 
+__weak int comphy_update_map(struct comphy_map *serdes_map, int count)
+{
+	return 0;
+}
+
 static int comphy_probe(struct udevice *dev)
 {
 	const void *blob = gd->fdt_blob;
@@ -76,6 +81,7 @@ static int comphy_probe(struct udevice *dev)
 	int lane;
 	int last_idx = 0;
 	static int current_idx;
+	int res;
 
 	/* Save base addresses for later use */
 	chip_cfg->comphy_base_addr = (void *)devfdt_get_addr_index(dev, 0);
@@ -143,6 +149,10 @@ static int comphy_probe(struct udevice *dev)
 		lane++;
 	}
 
+	res = comphy_update_map(comphy_map_data, chip_cfg->comphy_lanes_count);
+	if (res < 0)
+		return res;
+
 	/* Save CP index for MultiCP devices (A8K) */
 	chip_cfg->cp_index = current_idx++;
 	/* PHY power UP sequence */
diff --git a/drivers/phy/marvell/comphy.h b/drivers/phy/marvell/comphy_core.h
similarity index 96%
rename from drivers/phy/marvell/comphy.h
rename to drivers/phy/marvell/comphy_core.h
index b588ae41f0..12ab921d24 100644
--- a/drivers/phy/marvell/comphy.h
+++ b/drivers/phy/marvell/comphy_core.h
@@ -3,11 +3,11 @@
  * Copyright (C) 2015-2016 Marvell International Ltd.
  */
 
-#ifndef _COMPHY_H_
-#define _COMPHY_H_
+#ifndef _COMPHY_CORE_H_
+#define _COMPHY_CORE_H_
 
-#include <dt-bindings/comphy/comphy_data.h>
 #include <fdtdec.h>
+#include <mvebu/comphy.h>
 
 #if defined(DEBUG)
 #define debug_enter()	printf("----> Enter %s\n", __func__);
@@ -80,14 +80,6 @@ struct comphy_mux_data {
 	struct comphy_mux_options mux_values[MAX_LANE_OPTIONS];
 };
 
-struct comphy_map {
-	u32 type;
-	u32 speed;
-	u32 invert;
-	bool clk_src;
-	bool end_point;
-};
-
 struct chip_serdes_phy_config {
 	struct comphy_mux_data *mux_data;
 	int (*ptr_comphy_chip_init)(struct chip_serdes_phy_config *,
@@ -183,5 +175,5 @@ void comphy_pcie_config_detect(u32 comphy_max_count,
 			       struct comphy_map *serdes_map);
 void comphy_pcie_unit_general_config(u32 pex_index);
 
-#endif /* _COMPHY_H_ */
+#endif /* _COMPHY_CORE_H_ */
 
diff --git a/drivers/phy/marvell/comphy_cp110.c b/drivers/phy/marvell/comphy_cp110.c
index b0d5d5ca26..9998c07a47 100644
--- a/drivers/phy/marvell/comphy_cp110.c
+++ b/drivers/phy/marvell/comphy_cp110.c
@@ -9,7 +9,7 @@
 #include <asm/arch/cpu.h>
 #include <asm/arch/soc.h>
 
-#include "comphy.h"
+#include "comphy_core.h"
 #include "comphy_hpipe.h"
 #include "sata.h"
 #include "utmi_phy.h"
@@ -641,7 +641,8 @@ static int comphy_usb3_power_up(u32 lane, void __iomem *hpipe_base,
 }
 
 static int comphy_sata_power_up(u32 lane, void __iomem *hpipe_base,
-				void __iomem *comphy_base, int cp_index)
+				void __iomem *comphy_base, int cp_index,
+				u32 invert)
 {
 	u32 mask, data, i, ret = 1;
 	void __iomem *hpipe_addr = HPIPE_ADDR(hpipe_base, lane);
@@ -927,6 +928,19 @@ static int comphy_sata_power_up(u32 lane, void __iomem *hpipe_base,
 	reg_set(hpipe_addr + HPIPE_PWR_CTR_REG,
 		0x0 << HPIPE_PWR_CTR_RST_DFE_OFFSET,
 		HPIPE_PWR_CTR_RST_DFE_MASK);
+
+	/* Set RX / TX swaps */
+	data = mask = 0;
+	if (invert & PHY_POLARITY_TXD_INVERT) {
+		data |= (1 << HPIPE_SYNC_PATTERN_TXD_SWAP_OFFSET);
+		mask |= HPIPE_SYNC_PATTERN_TXD_SWAP_MASK;
+	}
+	if (invert & PHY_POLARITY_RXD_INVERT) {
+		data |= (1 << HPIPE_SYNC_PATTERN_RXD_SWAP_OFFSET);
+		mask |= HPIPE_SYNC_PATTERN_RXD_SWAP_MASK;
+	}
+	reg_set(hpipe_addr + HPIPE_SYNC_PATTERN_REG, data, mask);
+
 	/* SW reset for interupt logic */
 	reg_set(hpipe_addr + HPIPE_PWR_CTR_REG,
 		0x1 << HPIPE_PWR_CTR_SFT_RST_OFFSET,
@@ -2006,7 +2020,8 @@ int comphy_cp110_init(struct chip_serdes_phy_config *ptr_chip_cfg,
 		case PHY_TYPE_SATA3:
 			ret = comphy_sata_power_up(
 				lane, hpipe_base_addr, comphy_base_addr,
-				ptr_chip_cfg->cp_index);
+				ptr_chip_cfg->cp_index,
+				serdes_map[lane].invert);
 			break;
 		case PHY_TYPE_USB3_HOST0:
 		case PHY_TYPE_USB3_HOST1:
diff --git a/drivers/phy/marvell/comphy_hpipe.h b/drivers/phy/marvell/comphy_hpipe.h
index d99da7b9ff..a692035c94 100644
--- a/drivers/phy/marvell/comphy_hpipe.h
+++ b/drivers/phy/marvell/comphy_hpipe.h
@@ -221,6 +221,12 @@
 	(0x7 << HPIPE_LOOPBACK_SEL_OFFSET)
 
 #define HPIPE_SYNC_PATTERN_REG                  0x090
+#define HPIPE_SYNC_PATTERN_TXD_SWAP_OFFSET	10
+#define HPIPE_SYNC_PATTERN_TXD_SWAP_MASK	\
+	(0x1 << HPIPE_SYNC_PATTERN_TXD_SWAP_OFFSET)
+#define HPIPE_SYNC_PATTERN_RXD_SWAP_OFFSET	11
+#define HPIPE_SYNC_PATTERN_RXD_SWAP_MASK	\
+	(0x1 << HPIPE_SYNC_PATTERN_RXD_SWAP_OFFSET)
 
 #define HPIPE_INTERFACE_REG			0x94
 #define HPIPE_INTERFACE_GEN_MAX_OFFSET		10
diff --git a/drivers/phy/marvell/comphy_mux.c b/drivers/phy/marvell/comphy_mux.c
index 1f757d8e04..c67ba99762 100644
--- a/drivers/phy/marvell/comphy_mux.c
+++ b/drivers/phy/marvell/comphy_mux.c
@@ -6,7 +6,7 @@
 #include <common.h>
 #include <asm/io.h>
 
-#include "comphy.h"
+#include "comphy_core.h"
 #include "comphy_hpipe.h"
 
 /*
diff --git a/env/sf.c b/env/sf.c
index 494510533a..2e3c600483 100644
--- a/env/sf.c
+++ b/env/sf.c
@@ -58,7 +58,8 @@ static int setup_flash_device(void)
 
 	/* speed and mode will be read from DT */
 	ret = spi_flash_probe_bus_cs(CONFIG_ENV_SPI_BUS, CONFIG_ENV_SPI_CS,
-				     0, 0, &new);
+				     CONFIG_ENV_SPI_MAX_HZ, CONFIG_ENV_SPI_MODE,
+				     &new);
 	if (ret) {
 		set_default_env("spi_flash_probe_bus_cs() failed", 0);
 		return ret;
diff --git a/include/configs/clearfog.h b/include/configs/clearfog.h
index 7d56dfd86e..77ab6caf52 100644
--- a/include/configs/clearfog.h
+++ b/include/configs/clearfog.h
@@ -29,6 +29,10 @@
  */
 #define CONFIG_SYS_MMC_BASE		MVEBU_SDIO_BASE
 
+#ifdef CONFIG_CMD_MMC
+#define CONFIG_SUPPORT_EMMC_BOOT
+#endif
+
 /* USB/EHCI configuration */
 #define CONFIG_EHCI_IS_TDI
 
diff --git a/include/configs/db-88f6820-gp.h b/include/configs/db-88f6820-gp.h
index ac810b0cbc..f2aa21a43e 100644
--- a/include/configs/db-88f6820-gp.h
+++ b/include/configs/db-88f6820-gp.h
@@ -28,6 +28,12 @@
 #define CONFIG_SYS_I2C_SLAVE		0x0
 #define CONFIG_SYS_I2C_SPEED		100000
 
+/*
+ * SPI Flash configuration for the environemnt access
+ */
+#define CONFIG_ENV_SPI_BUS		0
+#define CONFIG_ENV_SPI_CS		0
+
 /* SPI NOR flash default params, used by sf commands */
 #define CONFIG_SF_DEFAULT_SPEED		1000000
 #define CONFIG_SF_DEFAULT_MODE		SPI_MODE_3
diff --git a/include/configs/mvebu_armada-8k.h b/include/configs/mvebu_armada-8k.h
index 93c50488a2..a6636e19e8 100644
--- a/include/configs/mvebu_armada-8k.h
+++ b/include/configs/mvebu_armada-8k.h
@@ -64,6 +64,9 @@
 #define CONFIG_ENV_SIZE			(64 << 10) /* 64KiB */
 #define CONFIG_ENV_SECT_SIZE		(64 << 10) /* 64KiB sectors */
 
+/* When runtime detection fails this is the default */
+#define CONFIG_SYS_MMC_ENV_DEV		0
+
 #define CONFIG_SYS_MAX_NAND_DEVICE	1
 #define CONFIG_SYS_NAND_MAX_CHIPS	1
 #define CONFIG_SYS_NAND_ONFI_DETECTION
diff --git a/include/configs/nas220.h b/include/configs/nas220.h
index ca5cb2a838..bdfa42fd30 100644
--- a/include/configs/nas220.h
+++ b/include/configs/nas220.h
@@ -69,7 +69,6 @@
 	"0x500000@0xc0000(uimage),"\
 	"0x1a40000@0x5c0000(rootfs)\0" \
 	"mtdids=nand0=orion_nand\0"\
-	"bootdelay=-1\0"\
 	"autostart=no\0"\
 	"autoload=no\0"
 
diff --git a/include/mvebu/comphy.h b/include/mvebu/comphy.h
new file mode 100644
index 0000000000..cde7a022af
--- /dev/null
+++ b/include/mvebu/comphy.h
@@ -0,0 +1,22 @@
+/* SPDX-License-Identifier: GPL-2.0+ */
+/*
+ * Copyright (C) 2015-2016 Marvell International Ltd.
+ */
+
+#ifndef _MVEBU_COMPHY_H_
+#define _MVEBU_COMPHY_H_
+
+#include <dt-bindings/comphy/comphy_data.h>
+
+struct comphy_map {
+	u32 type;
+	u32 speed;
+	u32 invert;
+	bool clk_src;
+	bool end_point;
+};
+
+int comphy_update_map(struct comphy_map *serdes_map, int count);
+
+#endif /* _MVEBU_COMPHY_H_ */
+
diff --git a/tools/kwboot.c b/tools/kwboot.c
index 50ae2b4b77..4be094c9c8 100644
--- a/tools/kwboot.c
+++ b/tools/kwboot.c
@@ -286,6 +286,7 @@ kwboot_bootmsg(int tty, void *msg)
 {
 	int rc;
 	char c;
+	int count;
 
 	if (msg == NULL)
 		kwboot_printv("Please reboot the target into UART boot mode...");
@@ -297,10 +298,12 @@ kwboot_bootmsg(int tty, void *msg)
 		if (rc)
 			break;
 
-		rc = kwboot_tty_send(tty, msg, 8);
-		if (rc) {
-			usleep(msg_req_delay * 1000);
-			continue;
+		for (count = 0; count < 128; count++) {
+			rc = kwboot_tty_send(tty, msg, 8);
+			if (rc) {
+				usleep(msg_req_delay * 1000);
+				continue;
+			}
 		}
 
 		rc = kwboot_tty_recv(tty, &c, 1, msg_rsp_timeo);
@@ -426,6 +429,9 @@ kwboot_xmodem(int tty, const void *_data, size_t size)
 
 	kwboot_printv("Sending boot image...\n");
 
+	sleep(2); /* flush isn't effective without it */
+	tcflush(tty, TCIOFLUSH);
+
 	do {
 		struct kwboot_block block;
 		int n;