Merge tag 'spi-v4.9' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi

Pull spi updates from Mark Brown:
 "The SPI subsystem has also had quite a quiet release, though with a
  fairly large set of per-driver changes and several new drivers. The
  bulk of the changes are:

   - lots and lots of cleanups and improvements for the fsl-espi driver

   - new drivers for Broadcom MSPI/iProc/STB, Cavium ThunderX and
     J-Core"

* tag 'spi-v4.9' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/spi: (80 commits)
  spi: sc18is602: Change gpiod_set_value to gpiod_set_value_cansleep
  spi: pxa2xx: Fix build error because of missing header
  spi: imx: fix error return code in spi_imx_probe()
  spi: pxa2xx: Add support for GPIO descriptor chip selects
  spi: imx: Gracefully handle NULL master->cs_gpios
  spi: iproc-qspi: Add Broadcom iProc SoCs support
  spi: fsl-espi: improve return value handling in fsl_espi_probe
  spi: fsl-espi: simplify of_fsl_espi_probe
  spi: fsl-espi: remove unused variable in fsl_espi_setup
  spi: bcm-qspi: Fix error return code in bcm_qspi_probe()
  spi: bcm-qspi: Fix return value check in bcm_qspi_probe()
  spi: bcm-qspi: fix suspend/resume #ifdef
  spi: bcm-qspi: don't include linux/mtd/cfi.h
  spi: core: Use spi_sync_transfer() in spi_write()/spi_read()
  spi: fsl-espi: improve and extend register bit definitions
  spi: fsl-espi: align register access with other drivers
  spi: fsl-espi: improve and simplify interrupt handler
  spi: fsl-espi: simplify fsl_espi_setup_transfer
  spi: imx: support loopback mode on imx35
  spi: imx: set spi_bus_clk for mx1, mx31 and mx35
  ...

29 files changed, 2718 insertions, 487 deletions

diff --git a/drivers/spi/Kconfig b/drivers/spi/Kconfig
index d6fb8d4b7786..b7995474148c 100644
--- a/drivers/spi/Kconfig
+++ b/drivers/spi/Kconfig
@@ -153,6 +153,16 @@ config SPI_BCM63XX_HSSPI
 	  This enables support for the High Speed SPI controller present on
 	  newer Broadcom BCM63XX SoCs.
 
+config SPI_BCM_QSPI
+	tristate "Broadcom BSPI and MSPI controller support"
+	depends on ARCH_BRCMSTB || ARCH_BCM || ARCH_BCM_IPROC || COMPILE_TEST
+	default ARCH_BCM_IPROC
+	help
+	  Enables support for the Broadcom SPI flash and MSPI controller.
+	  Select this option for any one of BRCMSTB, iProc NSP and NS2 SoCs
+	  based platforms. This driver works for both SPI master for spi-nor
+	  flash device as well as MSPI device.
+
 config SPI_BITBANG
 	tristate "Utilities for Bitbanging SPI masters"
 	help
@@ -285,6 +295,13 @@ config SPI_IMX
 	  This enables using the Freescale i.MX SPI controllers in master
 	  mode.
 
+config SPI_JCORE
+	tristate "J-Core SPI Master"
+	depends on OF && (SUPERH || COMPILE_TEST)
+	help
+	  This enables support for the SPI master controller in the J-Core
+	  synthesizable, open source SoC.
+
 config SPI_LM70_LLP
 	tristate "Parallel port adapter for LM70 eval board (DEVELOPMENT)"
 	depends on PARPORT
@@ -549,7 +566,7 @@ config SPI_SC18IS602
 config SPI_SH_MSIOF
 	tristate "SuperH MSIOF SPI controller"
 	depends on HAVE_CLK && HAS_DMA
-	depends on SUPERH || ARCH_RENESAS || COMPILE_TEST
+	depends on ARCH_SHMOBILE || ARCH_RENESAS || COMPILE_TEST
 	help
 	  SPI driver for SuperH and SH Mobile MSIOF blocks.
 
@@ -631,6 +648,13 @@ config SPI_TEGRA20_SLINK
 	help
 	  SPI driver for Nvidia Tegra20/Tegra30 SLINK Controller interface.
 
+config SPI_THUNDERX
+	tristate "Cavium ThunderX SPI controller"
+	depends on PCI && 64BIT && (ARM64 || COMPILE_TEST)
+	help
+	  SPI host driver for the hardware found on Cavium ThunderX
+	  SOCs.
+
 config SPI_TOPCLIFF_PCH
 	tristate "Intel EG20T PCH/LAPIS Semicon IOH(ML7213/ML7223/ML7831) SPI"
 	depends on PCI && (X86_32 || MIPS || COMPILE_TEST)
diff --git a/drivers/spi/Makefile b/drivers/spi/Makefile
index 185367ef6576..aa939d955521 100644
--- a/drivers/spi/Makefile
+++ b/drivers/spi/Makefile
@@ -21,6 +21,7 @@ obj-$(CONFIG_SPI_BCM2835AUX)		+= spi-bcm2835aux.o
 obj-$(CONFIG_SPI_BCM53XX)		+= spi-bcm53xx.o
 obj-$(CONFIG_SPI_BCM63XX)		+= spi-bcm63xx.o
 obj-$(CONFIG_SPI_BCM63XX_HSSPI)		+= spi-bcm63xx-hsspi.o
+obj-$(CONFIG_SPI_BCM_QSPI)		+= spi-iproc-qspi.o spi-brcmstb-qspi.o spi-bcm-qspi.o
 obj-$(CONFIG_SPI_BFIN5XX)		+= spi-bfin5xx.o
 obj-$(CONFIG_SPI_ADI_V3)                += spi-adi-v3.o
 obj-$(CONFIG_SPI_BFIN_SPORT)		+= spi-bfin-sport.o
@@ -46,6 +47,7 @@ obj-$(CONFIG_SPI_FSL_SPI)		+= spi-fsl-spi.o
 obj-$(CONFIG_SPI_GPIO)			+= spi-gpio.o
 obj-$(CONFIG_SPI_IMG_SPFI)		+= spi-img-spfi.o
 obj-$(CONFIG_SPI_IMX)			+= spi-imx.o
+obj-$(CONFIG_SPI_JCORE)			+= spi-jcore.o
 obj-$(CONFIG_SPI_LM70_LLP)		+= spi-lm70llp.o
 obj-$(CONFIG_SPI_LP8841_RTC)		+= spi-lp8841-rtc.o
 obj-$(CONFIG_SPI_MESON_SPIFC)		+= spi-meson-spifc.o
@@ -91,6 +93,8 @@ obj-$(CONFIG_SPI_TEGRA114)		+= spi-tegra114.o
 obj-$(CONFIG_SPI_TEGRA20_SFLASH)	+= spi-tegra20-sflash.o
 obj-$(CONFIG_SPI_TEGRA20_SLINK)		+= spi-tegra20-slink.o
 obj-$(CONFIG_SPI_TLE62X0)		+= spi-tle62x0.o
+spi-thunderx-objs			:= spi-cavium.o spi-cavium-thunderx.o
+obj-$(CONFIG_SPI_THUNDERX)		+= spi-thunderx.o
 obj-$(CONFIG_SPI_TOPCLIFF_PCH)		+= spi-topcliff-pch.o
 obj-$(CONFIG_SPI_TXX9)			+= spi-txx9.o
 obj-$(CONFIG_SPI_XCOMM)		+= spi-xcomm.o
diff --git a/drivers/spi/spi-bcm-qspi.c b/drivers/spi/spi-bcm-qspi.c
new file mode 100644
index 000000000000..14f9dea3173f
--- /dev/null
+++ b/drivers/spi/spi-bcm-qspi.c
@@ -0,0 +1,1397 @@
+/*
+ * Driver for Broadcom BRCMSTB, NSP,  NS2, Cygnus SPI Controllers
+ *
+ * Copyright 2016 Broadcom
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, version 2, as
+ * published by the Free Software Foundation (the "GPL").
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License version 2 (GPLv2) for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * version 2 (GPLv2) along with this source code.
+ */
+
+#include <linux/clk.h>
+#include <linux/delay.h>
+#include <linux/device.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/ioport.h>
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/mtd/spi-nor.h>
+#include <linux/of.h>
+#include <linux/of_irq.h>
+#include <linux/platform_device.h>
+#include <linux/slab.h>
+#include <linux/spi/spi.h>
+#include <linux/sysfs.h>
+#include <linux/types.h>
+#include "spi-bcm-qspi.h"
+
+#define DRIVER_NAME "bcm_qspi"
+
+
+/* BSPI register offsets */
+#define BSPI_REVISION_ID			0x000
+#define BSPI_SCRATCH				0x004
+#define BSPI_MAST_N_BOOT_CTRL			0x008
+#define BSPI_BUSY_STATUS			0x00c
+#define BSPI_INTR_STATUS			0x010
+#define BSPI_B0_STATUS				0x014
+#define BSPI_B0_CTRL				0x018
+#define BSPI_B1_STATUS				0x01c
+#define BSPI_B1_CTRL				0x020
+#define BSPI_STRAP_OVERRIDE_CTRL		0x024
+#define BSPI_FLEX_MODE_ENABLE			0x028
+#define BSPI_BITS_PER_CYCLE			0x02c
+#define BSPI_BITS_PER_PHASE			0x030
+#define BSPI_CMD_AND_MODE_BYTE			0x034
+#define BSPI_BSPI_FLASH_UPPER_ADDR_BYTE	0x038
+#define BSPI_BSPI_XOR_VALUE			0x03c
+#define BSPI_BSPI_XOR_ENABLE			0x040
+#define BSPI_BSPI_PIO_MODE_ENABLE		0x044
+#define BSPI_BSPI_PIO_IODIR			0x048
+#define BSPI_BSPI_PIO_DATA			0x04c
+
+/* RAF register offsets */
+#define BSPI_RAF_START_ADDR			0x100
+#define BSPI_RAF_NUM_WORDS			0x104
+#define BSPI_RAF_CTRL				0x108
+#define BSPI_RAF_FULLNESS			0x10c
+#define BSPI_RAF_WATERMARK			0x110
+#define BSPI_RAF_STATUS			0x114
+#define BSPI_RAF_READ_DATA			0x118
+#define BSPI_RAF_WORD_CNT			0x11c
+#define BSPI_RAF_CURR_ADDR			0x120
+
+/* Override mode masks */
+#define BSPI_STRAP_OVERRIDE_CTRL_OVERRIDE	BIT(0)
+#define BSPI_STRAP_OVERRIDE_CTRL_DATA_DUAL	BIT(1)
+#define BSPI_STRAP_OVERRIDE_CTRL_ADDR_4BYTE	BIT(2)
+#define BSPI_STRAP_OVERRIDE_CTRL_DATA_QUAD	BIT(3)
+#define BSPI_STRAP_OVERRIDE_CTRL_ENDAIN_MODE	BIT(4)
+
+#define BSPI_ADDRLEN_3BYTES			3
+#define BSPI_ADDRLEN_4BYTES			4
+
+#define BSPI_RAF_STATUS_FIFO_EMPTY_MASK	BIT(1)
+
+#define BSPI_RAF_CTRL_START_MASK		BIT(0)
+#define BSPI_RAF_CTRL_CLEAR_MASK		BIT(1)
+
+#define BSPI_BPP_MODE_SELECT_MASK		BIT(8)
+#define BSPI_BPP_ADDR_SELECT_MASK		BIT(16)
+
+#define BSPI_READ_LENGTH			256
+
+/* MSPI register offsets */
+#define MSPI_SPCR0_LSB				0x000
+#define MSPI_SPCR0_MSB				0x004
+#define MSPI_SPCR1_LSB				0x008
+#define MSPI_SPCR1_MSB				0x00c
+#define MSPI_NEWQP				0x010
+#define MSPI_ENDQP				0x014
+#define MSPI_SPCR2				0x018
+#define MSPI_MSPI_STATUS			0x020
+#define MSPI_CPTQP				0x024
+#define MSPI_SPCR3				0x028
+#define MSPI_TXRAM				0x040
+#define MSPI_RXRAM				0x0c0
+#define MSPI_CDRAM				0x140
+#define MSPI_WRITE_LOCK			0x180
+
+#define MSPI_MASTER_BIT			BIT(7)
+
+#define MSPI_NUM_CDRAM				16
+#define MSPI_CDRAM_CONT_BIT			BIT(7)
+#define MSPI_CDRAM_BITSE_BIT			BIT(6)
+#define MSPI_CDRAM_PCS				0xf
+
+#define MSPI_SPCR2_SPE				BIT(6)
+#define MSPI_SPCR2_CONT_AFTER_CMD		BIT(7)
+
+#define MSPI_MSPI_STATUS_SPIF			BIT(0)
+
+#define INTR_BASE_BIT_SHIFT			0x02
+#define INTR_COUNT				0x07
+
+#define NUM_CHIPSELECT				4
+#define QSPI_SPBR_MIN				8U
+#define QSPI_SPBR_MAX				255U
+
+#define OPCODE_DIOR				0xBB
+#define OPCODE_QIOR				0xEB
+#define OPCODE_DIOR_4B				0xBC
+#define OPCODE_QIOR_4B				0xEC
+
+#define MAX_CMD_SIZE				6
+
+#define ADDR_4MB_MASK				GENMASK(22, 0)
+
+/* stop at end of transfer, no other reason */
+#define TRANS_STATUS_BREAK_NONE		0
+/* stop at end of spi_message */
+#define TRANS_STATUS_BREAK_EOM			1
+/* stop at end of spi_transfer if delay */
+#define TRANS_STATUS_BREAK_DELAY		2
+/* stop at end of spi_transfer if cs_change */
+#define TRANS_STATUS_BREAK_CS_CHANGE		4
+/* stop if we run out of bytes */
+#define TRANS_STATUS_BREAK_NO_BYTES		8
+
+/* events that make us stop filling TX slots */
+#define TRANS_STATUS_BREAK_TX (TRANS_STATUS_BREAK_EOM |		\
+			       TRANS_STATUS_BREAK_DELAY |		\
+			       TRANS_STATUS_BREAK_CS_CHANGE)
+
+/* events that make us deassert CS */
+#define TRANS_STATUS_BREAK_DESELECT (TRANS_STATUS_BREAK_EOM |		\
+				     TRANS_STATUS_BREAK_CS_CHANGE)
+
+struct bcm_qspi_parms {
+	u32 speed_hz;
+	u8 mode;
+	u8 bits_per_word;
+};
+
+struct bcm_xfer_mode {
+	bool flex_mode;
+	unsigned int width;
+	unsigned int addrlen;
+	unsigned int hp;
+};
+
+enum base_type {
+	MSPI,
+	BSPI,
+	CHIP_SELECT,
+	BASEMAX,
+};
+
+enum irq_source {
+	SINGLE_L2,
+	MUXED_L1,
+};
+
+struct bcm_qspi_irq {
+	const char *irq_name;
+	const irq_handler_t irq_handler;
+	int irq_source;
+	u32 mask;
+};
+
+struct bcm_qspi_dev_id {
+	const struct bcm_qspi_irq *irqp;
+	void *dev;
+};
+
+struct qspi_trans {
+	struct spi_transfer *trans;
+	int byte;
+};
+
+struct bcm_qspi {
+	struct platform_device *pdev;
+	struct spi_master *master;
+	struct clk *clk;
+	u32 base_clk;
+	u32 max_speed_hz;
+	void __iomem *base[BASEMAX];
+
+	/* Some SoCs provide custom interrupt status register(s) */
+	struct bcm_qspi_soc_intc	*soc_intc;
+
+	struct bcm_qspi_parms last_parms;
+	struct qspi_trans  trans_pos;
+	int curr_cs;
+	int bspi_maj_rev;
+	int bspi_min_rev;
+	int bspi_enabled;
+	struct spi_flash_read_message *bspi_rf_msg;
+	u32 bspi_rf_msg_idx;
+	u32 bspi_rf_msg_len;
+	u32 bspi_rf_msg_status;
+	struct bcm_xfer_mode xfer_mode;
+	u32 s3_strap_override_ctrl;
+	bool bspi_mode;
+	bool big_endian;
+	int num_irqs;
+	struct bcm_qspi_dev_id *dev_ids;
+	struct completion mspi_done;
+	struct completion bspi_done;
+};
+
+static inline bool has_bspi(struct bcm_qspi *qspi)
+{
+	return qspi->bspi_mode;
+}
+
+/* Read qspi controller register*/
+static inline u32 bcm_qspi_read(struct bcm_qspi *qspi, enum base_type type,
+				unsigned int offset)
+{
+	return bcm_qspi_readl(qspi->big_endian, qspi->base[type] + offset);
+}
+
+/* Write qspi controller register*/
+static inline void bcm_qspi_write(struct bcm_qspi *qspi, enum base_type type,
+				  unsigned int offset, unsigned int data)
+{
+	bcm_qspi_writel(qspi->big_endian, data, qspi->base[type] + offset);
+}
+
+/* BSPI helpers */
+static int bcm_qspi_bspi_busy_poll(struct bcm_qspi *qspi)
+{
+	int i;
+
+	/* this should normally finish within 10us */
+	for (i = 0; i < 1000; i++) {
+		if (!(bcm_qspi_read(qspi, BSPI, BSPI_BUSY_STATUS) & 1))
+			return 0;
+		udelay(1);
+	}
+	dev_warn(&qspi->pdev->dev, "timeout waiting for !busy_status\n");
+	return -EIO;
+}
+
+static inline bool bcm_qspi_bspi_ver_three(struct bcm_qspi *qspi)
+{
+	if (qspi->bspi_maj_rev < 4)
+		return true;
+	return false;
+}
+
+static void bcm_qspi_bspi_flush_prefetch_buffers(struct bcm_qspi *qspi)
+{
+	bcm_qspi_bspi_busy_poll(qspi);
+	/* Force rising edge for the b0/b1 'flush' field */
+	bcm_qspi_write(qspi, BSPI, BSPI_B0_CTRL, 1);
+	bcm_qspi_write(qspi, BSPI, BSPI_B1_CTRL, 1);
+	bcm_qspi_write(qspi, BSPI, BSPI_B0_CTRL, 0);
+	bcm_qspi_write(qspi, BSPI, BSPI_B1_CTRL, 0);
+}
+
+static int bcm_qspi_bspi_lr_is_fifo_empty(struct bcm_qspi *qspi)
+{
+	return (bcm_qspi_read(qspi, BSPI, BSPI_RAF_STATUS) &
+				BSPI_RAF_STATUS_FIFO_EMPTY_MASK);
+}
+
+static inline u32 bcm_qspi_bspi_lr_read_fifo(struct bcm_qspi *qspi)
+{
+	u32 data = bcm_qspi_read(qspi, BSPI, BSPI_RAF_READ_DATA);
+
+	/* BSPI v3 LR is LE only, convert data to host endianness */
+	if (bcm_qspi_bspi_ver_three(qspi))
+		data = le32_to_cpu(data);
+
+	return data;
+}
+
+static inline void bcm_qspi_bspi_lr_start(struct bcm_qspi *qspi)
+{
+	bcm_qspi_bspi_busy_poll(qspi);
+	bcm_qspi_write(qspi, BSPI, BSPI_RAF_CTRL,
+		       BSPI_RAF_CTRL_START_MASK);
+}
+
+static inline void bcm_qspi_bspi_lr_clear(struct bcm_qspi *qspi)
+{
+	bcm_qspi_write(qspi, BSPI, BSPI_RAF_CTRL,
+		       BSPI_RAF_CTRL_CLEAR_MASK);
+	bcm_qspi_bspi_flush_prefetch_buffers(qspi);
+}
+
+static void bcm_qspi_bspi_lr_data_read(struct bcm_qspi *qspi)
+{
+	u32 *buf = (u32 *)qspi->bspi_rf_msg->buf;
+	u32 data = 0;
+
+	dev_dbg(&qspi->pdev->dev, "xfer %p rx %p rxlen %d\n", qspi->bspi_rf_msg,
+		qspi->bspi_rf_msg->buf, qspi->bspi_rf_msg_len);
+	while (!bcm_qspi_bspi_lr_is_fifo_empty(qspi)) {
+		data = bcm_qspi_bspi_lr_read_fifo(qspi);
+		if (likely(qspi->bspi_rf_msg_len >= 4) &&
+		    IS_ALIGNED((uintptr_t)buf, 4)) {
+			buf[qspi->bspi_rf_msg_idx++] = data;
+			qspi->bspi_rf_msg_len -= 4;
+		} else {
+			/* Read out remaining bytes, make sure*/
+			u8 *cbuf = (u8 *)&buf[qspi->bspi_rf_msg_idx];
+
+			data = cpu_to_le32(data);
+			while (qspi->bspi_rf_msg_len) {
+				*cbuf++ = (u8)data;
+				data >>= 8;
+				qspi->bspi_rf_msg_len--;
+			}
+		}
+	}
+}
+
+static void bcm_qspi_bspi_set_xfer_params(struct bcm_qspi *qspi, u8 cmd_byte,
+					  int bpp, int bpc, int flex_mode)
+{
+	bcm_qspi_write(qspi, BSPI, BSPI_FLEX_MODE_ENABLE, 0);
+	bcm_qspi_write(qspi, BSPI, BSPI_BITS_PER_CYCLE, bpc);
+	bcm_qspi_write(qspi, BSPI, BSPI_BITS_PER_PHASE, bpp);
+	bcm_qspi_write(qspi, BSPI, BSPI_CMD_AND_MODE_BYTE, cmd_byte);
+	bcm_qspi_write(qspi, BSPI, BSPI_FLEX_MODE_ENABLE, flex_mode);
+}
+
+static int bcm_qspi_bspi_set_flex_mode(struct bcm_qspi *qspi, int width,
+				       int addrlen, int hp)
+{
+	int bpc = 0, bpp = 0;
+	u8 command = SPINOR_OP_READ_FAST;
+	int flex_mode = 1, rv = 0;
+	bool spans_4byte = false;
+
+	dev_dbg(&qspi->pdev->dev, "set flex mode w %x addrlen %x hp %d\n",
+		width, addrlen, hp);
+
+	if (addrlen == BSPI_ADDRLEN_4BYTES) {
+		bpp = BSPI_BPP_ADDR_SELECT_MASK;
+		spans_4byte = true;
+	}
+
+	bpp |= 8;
+
+	switch (width) {
+	case SPI_NBITS_SINGLE:
+		if (addrlen == BSPI_ADDRLEN_3BYTES)
+			/* default mode, does not need flex_cmd */
+			flex_mode = 0;
+		else
+			command = SPINOR_OP_READ4_FAST;
+		break;
+	case SPI_NBITS_DUAL:
+		bpc = 0x00000001;
+		if (hp) {
+			bpc |= 0x00010100; /* address and mode are 2-bit */
+			bpp = BSPI_BPP_MODE_SELECT_MASK;
+			command = OPCODE_DIOR;
+			if (spans_4byte)
+				command = OPCODE_DIOR_4B;
+		} else {
+			command = SPINOR_OP_READ_1_1_2;
+			if (spans_4byte)
+				command = SPINOR_OP_READ4_1_1_2;
+		}
+		break;
+	case SPI_NBITS_QUAD:
+		bpc = 0x00000002;
+		if (hp) {
+			bpc |= 0x00020200; /* address and mode are 4-bit */
+			bpp = 4; /* dummy cycles */
+			bpp |= BSPI_BPP_ADDR_SELECT_MASK;
+			command = OPCODE_QIOR;
+			if (spans_4byte)
+				command = OPCODE_QIOR_4B;
+		} else {
+			command = SPINOR_OP_READ_1_1_4;
+			if (spans_4byte)
+				command = SPINOR_OP_READ4_1_1_4;
+		}
+		break;
+	default:
+		rv = -EINVAL;
+		break;
+	}
+
+	if (rv == 0)
+		bcm_qspi_bspi_set_xfer_params(qspi, command, bpp, bpc,
+					      flex_mode);
+
+	return rv;
+}
+
+static int bcm_qspi_bspi_set_override(struct bcm_qspi *qspi, int width,
+				      int addrlen, int hp)
+{
+	u32 data = bcm_qspi_read(qspi, BSPI, BSPI_STRAP_OVERRIDE_CTRL);
+
+	dev_dbg(&qspi->pdev->dev, "set override mode w %x addrlen %x hp %d\n",
+		width, addrlen, hp);
+
+	switch (width) {
+	case SPI_NBITS_SINGLE:
+		/* clear quad/dual mode */
+		data &= ~(BSPI_STRAP_OVERRIDE_CTRL_DATA_QUAD |
+			  BSPI_STRAP_OVERRIDE_CTRL_DATA_DUAL);
+		break;
+
+	case SPI_NBITS_QUAD:
+		/* clear dual mode and set quad mode */
+		data &= ~BSPI_STRAP_OVERRIDE_CTRL_DATA_DUAL;
+		data |= BSPI_STRAP_OVERRIDE_CTRL_DATA_QUAD;
+		break;
+	case SPI_NBITS_DUAL:
+		/* clear quad mode set dual mode */
+		data &= ~BSPI_STRAP_OVERRIDE_CTRL_DATA_QUAD;
+		data |= BSPI_STRAP_OVERRIDE_CTRL_DATA_DUAL;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	if (addrlen == BSPI_ADDRLEN_4BYTES)
+		/* set 4byte mode*/
+		data |= BSPI_STRAP_OVERRIDE_CTRL_ADDR_4BYTE;
+	else
+		/* clear 4 byte mode */
+		data &= ~BSPI_STRAP_OVERRIDE_CTRL_ADDR_4BYTE;
+
+	/* set the override mode */
+	data |=	BSPI_STRAP_OVERRIDE_CTRL_OVERRIDE;
+	bcm_qspi_write(qspi, BSPI, BSPI_STRAP_OVERRIDE_CTRL, data);
+	bcm_qspi_bspi_set_xfer_params(qspi, SPINOR_OP_READ_FAST, 0, 0, 0);
+
+	return 0;
+}
+
+static int bcm_qspi_bspi_set_mode(struct bcm_qspi *qspi,
+				  int width, int addrlen, int hp)
+{
+	int error = 0;
+
+	/* default mode */
+	qspi->xfer_mode.flex_mode = true;
+
+	if (!bcm_qspi_bspi_ver_three(qspi)) {
+		u32 val, mask;
+
+		val = bcm_qspi_read(qspi, BSPI, BSPI_STRAP_OVERRIDE_CTRL);
+		mask = BSPI_STRAP_OVERRIDE_CTRL_OVERRIDE;
+		if (val & mask || qspi->s3_strap_override_ctrl & mask) {
+			qspi->xfer_mode.flex_mode = false;
+			bcm_qspi_write(qspi, BSPI, BSPI_FLEX_MODE_ENABLE,
+				       0);
+
+			if ((val | qspi->s3_strap_override_ctrl) &
+			    BSPI_STRAP_OVERRIDE_CTRL_DATA_DUAL)
+				width = SPI_NBITS_DUAL;
+			else if ((val |  qspi->s3_strap_override_ctrl) &
+				 BSPI_STRAP_OVERRIDE_CTRL_DATA_QUAD)
+				width = SPI_NBITS_QUAD;
+
+			error = bcm_qspi_bspi_set_override(qspi, width, addrlen,
+							   hp);
+		}
+	}
+
+	if (qspi->xfer_mode.flex_mode)
+		error = bcm_qspi_bspi_set_flex_mode(qspi, width, addrlen, hp);
+
+	if (error) {
+		dev_warn(&qspi->pdev->dev,
+			 "INVALID COMBINATION: width=%d addrlen=%d hp=%d\n",
+			 width, addrlen, hp);
+	} else if (qspi->xfer_mode.width != width ||
+		   qspi->xfer_mode.addrlen != addrlen ||
+		   qspi->xfer_mode.hp != hp) {
+		qspi->xfer_mode.width = width;
+		qspi->xfer_mode.addrlen = addrlen;
+		qspi->xfer_mode.hp = hp;
+		dev_dbg(&qspi->pdev->dev,
+			"cs:%d %d-lane output, %d-byte address%s\n",
+			qspi->curr_cs,
+			qspi->xfer_mode.width,
+			qspi->xfer_mode.addrlen,
+			qspi->xfer_mode.hp != -1 ? ", hp mode" : "");
+	}
+
+	return error;
+}
+
+static void bcm_qspi_enable_bspi(struct bcm_qspi *qspi)
+{
+	if (!has_bspi(qspi) || (qspi->bspi_enabled))
+		return;
+
+	qspi->bspi_enabled = 1;
+	if ((bcm_qspi_read(qspi, BSPI, BSPI_MAST_N_BOOT_CTRL) & 1) == 0)
+		return;
+
+	bcm_qspi_bspi_flush_prefetch_buffers(qspi);
+	udelay(1);
+	bcm_qspi_write(qspi, BSPI, BSPI_MAST_N_BOOT_CTRL, 0);
+	udelay(1);
+}
+
+static void bcm_qspi_disable_bspi(struct bcm_qspi *qspi)
+{
+	if (!has_bspi(qspi) || (!qspi->bspi_enabled))
+		return;
+
+	qspi->bspi_enabled = 0;
+	if ((bcm_qspi_read(qspi, BSPI, BSPI_MAST_N_BOOT_CTRL) & 1))
+		return;
+
+	bcm_qspi_bspi_busy_poll(qspi);
+	bcm_qspi_write(qspi, BSPI, BSPI_MAST_N_BOOT_CTRL, 1);
+	udelay(1);
+}
+
+static void bcm_qspi_chip_select(struct bcm_qspi *qspi, int cs)
+{
+	u32 data = 0;
+
+	if (qspi->curr_cs == cs)
+		return;
+	if (qspi->base[CHIP_SELECT]) {
+		data = bcm_qspi_read(qspi, CHIP_SELECT, 0);
+		data = (data & ~0xff) | (1 << cs);
+		bcm_qspi_write(qspi, CHIP_SELECT, 0, data);
+		usleep_range(10, 20);
+	}
+	qspi->curr_cs = cs;
+}
+
+/* MSPI helpers */
+static void bcm_qspi_hw_set_parms(struct bcm_qspi *qspi,
+				  const struct bcm_qspi_parms *xp)
+{
+	u32 spcr, spbr = 0;
+
+	if (xp->speed_hz)
+		spbr = qspi->base_clk / (2 * xp->speed_hz);
+
+	spcr = clamp_val(spbr, QSPI_SPBR_MIN, QSPI_SPBR_MAX);
+	bcm_qspi_write(qspi, MSPI, MSPI_SPCR0_LSB, spcr);
+
+	spcr = MSPI_MASTER_BIT;
+	/* for 16 bit the data should be zero */
+	if (xp->bits_per_word != 16)
+		spcr |= xp->bits_per_word << 2;
+	spcr |= xp->mode & 3;
+	bcm_qspi_write(qspi, MSPI, MSPI_SPCR0_MSB, spcr);
+
+	qspi->last_parms = *xp;
+}
+
+static void bcm_qspi_update_parms(struct bcm_qspi *qspi,
+				  struct spi_device *spi,
+				  struct spi_transfer *trans)
+{
+	struct bcm_qspi_parms xp;
+
+	xp.speed_hz = trans->speed_hz;
+	xp.bits_per_word = trans->bits_per_word;
+	xp.mode = spi->mode;
+
+	bcm_qspi_hw_set_parms(qspi, &xp);
+}
+
+static int bcm_qspi_setup(struct spi_device *spi)
+{
+	struct bcm_qspi_parms *xp;
+
+	if (spi->bits_per_word > 16)
+		return -EINVAL;
+
+	xp = spi_get_ctldata(spi);
+	if (!xp) {
+		xp = kzalloc(sizeof(*xp), GFP_KERNEL);
+		if (!xp)
+			return -ENOMEM;
+		spi_set_ctldata(spi, xp);
+	}
+	xp->speed_hz = spi->max_speed_hz;
+	xp->mode = spi->mode;
+
+	if (spi->bits_per_word)
+		xp->bits_per_word = spi->bits_per_word;
+	else
+		xp->bits_per_word = 8;
+
+	return 0;
+}
+
+static int update_qspi_trans_byte_count(struct bcm_qspi *qspi,
+					struct qspi_trans *qt, int flags)
+{
+	int ret = TRANS_STATUS_BREAK_NONE;
+
+	/* count the last transferred bytes */
+	if (qt->trans->bits_per_word <= 8)
+		qt->byte++;
+	else
+		qt->byte += 2;
+
+	if (qt->byte >= qt->trans->len) {
+		/* we're at the end of the spi_transfer */
+
+		/* in TX mode, need to pause for a delay or CS change */
+		if (qt->trans->delay_usecs &&
+		    (flags & TRANS_STATUS_BREAK_DELAY))
+			ret |= TRANS_STATUS_BREAK_DELAY;
+		if (qt->trans->cs_change &&
+		    (flags & TRANS_STATUS_BREAK_CS_CHANGE))
+			ret |= TRANS_STATUS_BREAK_CS_CHANGE;
+		if (ret)
+			goto done;
+
+		dev_dbg(&qspi->pdev->dev, "advance msg exit\n");
+		if (spi_transfer_is_last(qspi->master, qt->trans))
+			ret = TRANS_STATUS_BREAK_EOM;
+		else
+			ret = TRANS_STATUS_BREAK_NO_BYTES;
+
+		qt->trans = NULL;
+	}
+
+done:
+	dev_dbg(&qspi->pdev->dev, "trans %p len %d byte %d ret %x\n",
+		qt->trans, qt->trans ? qt->trans->len : 0, qt->byte, ret);
+	return ret;
+}
+
+static inline u8 read_rxram_slot_u8(struct bcm_qspi *qspi, int slot)
+{
+	u32 slot_offset = MSPI_RXRAM + (slot << 3) + 0x4;
+
+	/* mask out reserved bits */
+	return bcm_qspi_read(qspi, MSPI, slot_offset) & 0xff;
+}
+
+static inline u16 read_rxram_slot_u16(struct bcm_qspi *qspi, int slot)
+{
+	u32 reg_offset = MSPI_RXRAM;
+	u32 lsb_offset = reg_offset + (slot << 3) + 0x4;
+	u32 msb_offset = reg_offset + (slot << 3);
+
+	return (bcm_qspi_read(qspi, MSPI, lsb_offset) & 0xff) |
+		((bcm_qspi_read(qspi, MSPI, msb_offset) & 0xff) << 8);
+}
+
+static void read_from_hw(struct bcm_qspi *qspi, int slots)
+{
+	struct qspi_trans tp;
+	int slot;
+
+	bcm_qspi_disable_bspi(qspi);
+
+	if (slots > MSPI_NUM_CDRAM) {
+		/* should never happen */
+		dev_err(&qspi->pdev->dev, "%s: too many slots!\n", __func__);
+		return;
+	}
+
+	tp = qspi->trans_pos;
+
+	for (slot = 0; slot < slots; slot++) {
+		if (tp.trans->bits_per_word <= 8) {
+			u8 *buf = tp.trans->rx_buf;
+
+			if (buf)
+				buf[tp.byte] = read_rxram_slot_u8(qspi, slot);
+			dev_dbg(&qspi->pdev->dev, "RD %02x\n",
+				buf ? buf[tp.byte] : 0xff);
+		} else {
+			u16 *buf = tp.trans->rx_buf;
+
+			if (buf)
+				buf[tp.byte / 2] = read_rxram_slot_u16(qspi,
+								      slot);
+			dev_dbg(&qspi->pdev->dev, "RD %04x\n",
+				buf ? buf[tp.byte] : 0xffff);
+		}
+
+		update_qspi_trans_byte_count(qspi, &tp,
+					     TRANS_STATUS_BREAK_NONE);
+	}
+
+	qspi->trans_pos = tp;
+}
+
+static inline void write_txram_slot_u8(struct bcm_qspi *qspi, int slot,
+				       u8 val)
+{
+	u32 reg_offset = MSPI_TXRAM + (slot << 3);
+
+	/* mask out reserved bits */
+	bcm_qspi_write(qspi, MSPI, reg_offset, val);
+}
+
+static inline void write_txram_slot_u16(struct bcm_qspi *qspi, int slot,
+					u16 val)
+{
+	u32 reg_offset = MSPI_TXRAM;
+	u32 msb_offset = reg_offset + (slot << 3);
+	u32 lsb_offset = reg_offset + (slot << 3) + 0x4;
+
+	bcm_qspi_write(qspi, MSPI, msb_offset, (val >> 8));
+	bcm_qspi_write(qspi, MSPI, lsb_offset, (val & 0xff));
+}
+
+static inline u32 read_cdram_slot(struct bcm_qspi *qspi, int slot)
+{
+	return bcm_qspi_read(qspi, MSPI, MSPI_CDRAM + (slot << 2));
+}
+
+static inline void write_cdram_slot(struct bcm_qspi *qspi, int slot, u32 val)
+{
+	bcm_qspi_write(qspi, MSPI, (MSPI_CDRAM + (slot << 2)), val);
+}
+
+/* Return number of slots written */
+static int write_to_hw(struct bcm_qspi *qspi, struct spi_device *spi)
+{
+	struct qspi_trans tp;
+	int slot = 0, tstatus = 0;
+	u32 mspi_cdram = 0;
+
+	bcm_qspi_disable_bspi(qspi);
+	tp = qspi->trans_pos;
+	bcm_qspi_update_parms(qspi, spi, tp.trans);
+
+	/* Run until end of transfer or reached the max data */
+	while (!tstatus && slot < MSPI_NUM_CDRAM) {
+		if (tp.trans->bits_per_word <= 8) {
+			const u8 *buf = tp.trans->tx_buf;
+			u8 val = buf ? buf[tp.byte] : 0xff;
+
+			write_txram_slot_u8(qspi, slot, val);
+			dev_dbg(&qspi->pdev->dev, "WR %02x\n", val);
+		} else {
+			const u16 *buf = tp.trans->tx_buf;
+			u16 val = buf ? buf[tp.byte / 2] : 0xffff;
+
+			write_txram_slot_u16(qspi, slot, val);
+			dev_dbg(&qspi->pdev->dev, "WR %04x\n", val);
+		}
+		mspi_cdram = MSPI_CDRAM_CONT_BIT;
+		mspi_cdram |= (~(1 << spi->chip_select) &
+			       MSPI_CDRAM_PCS);
+		mspi_cdram |= ((tp.trans->bits_per_word <= 8) ? 0 :
+				MSPI_CDRAM_BITSE_BIT);
+
+		write_cdram_slot(qspi, slot, mspi_cdram);
+
+		tstatus = update_qspi_trans_byte_count(qspi, &tp,
+						       TRANS_STATUS_BREAK_TX);
+		slot++;
+	}
+
+	if (!slot) {
+		dev_err(&qspi->pdev->dev, "%s: no data to send?", __func__);
+		goto done;
+	}
+
+	dev_dbg(&qspi->pdev->dev, "submitting %d slots\n", slot);
+	bcm_qspi_write(qspi, MSPI, MSPI_NEWQP, 0);
+	bcm_qspi_write(qspi, MSPI, MSPI_ENDQP, slot - 1);
+
+	if (tstatus & TRANS_STATUS_BREAK_DESELECT) {
+		mspi_cdram = read_cdram_slot(qspi, slot - 1) &
+			~MSPI_CDRAM_CONT_BIT;
+		write_cdram_slot(qspi, slot - 1, mspi_cdram);
+	}
+
+	if (has_bspi(qspi))
+		bcm_qspi_write(qspi, MSPI, MSPI_WRITE_LOCK, 1);
+
+	/* Must flush previous writes before starting MSPI operation */
+	mb();
+	/* Set cont | spe | spifie */
+	bcm_qspi_write(qspi, MSPI, MSPI_SPCR2, 0xe0);
+
+done:
+	return slot;
+}
+
+static int bcm_qspi_bspi_flash_read(struct spi_device *spi,
+				    struct spi_flash_read_message *msg)
+{
+	struct bcm_qspi *qspi = spi_master_get_devdata(spi->master);
+	u32 addr = 0, len, len_words;
+	int ret = 0;
+	unsigned long timeo = msecs_to_jiffies(100);
+	struct bcm_qspi_soc_intc *soc_intc = qspi->soc_intc;
+
+	if (bcm_qspi_bspi_ver_three(qspi))
+		if (msg->addr_width == BSPI_ADDRLEN_4BYTES)
+			return -EIO;
+
+	bcm_qspi_chip_select(qspi, spi->chip_select);
+	bcm_qspi_write(qspi, MSPI, MSPI_WRITE_LOCK, 0);
+
+	/*
+	 * when using flex mode mode we need to send
+	 * the upper address byte to bspi
+	 */
+	if (bcm_qspi_bspi_ver_three(qspi) == false) {
+		addr = msg->from & 0xff000000;
+		bcm_qspi_write(qspi, BSPI,
+			       BSPI_BSPI_FLASH_UPPER_ADDR_BYTE, addr);
+	}
+
+	if (!qspi->xfer_mode.flex_mode)
+		addr = msg->from;
+	else
+		addr = msg->from & 0x00ffffff;
+
+	/* set BSPI RAF buffer max read length */
+	len = msg->len;
+	if (len > BSPI_READ_LENGTH)
+		len = BSPI_READ_LENGTH;
+
+	if (bcm_qspi_bspi_ver_three(qspi) == true)
+		addr = (addr + 0xc00000) & 0xffffff;
+
+	reinit_completion(&qspi->bspi_done);
+	bcm_qspi_enable_bspi(qspi);
+	len_words = (len + 3) >> 2;
+	qspi->bspi_rf_msg = msg;
+	qspi->bspi_rf_msg_status = 0;
+	qspi->bspi_rf_msg_idx = 0;
+	qspi->bspi_rf_msg_len = len;
+	dev_dbg(&qspi->pdev->dev, "bspi xfr addr 0x%x len 0x%x", addr, len);
+
+	bcm_qspi_write(qspi, BSPI, BSPI_RAF_START_ADDR, addr);
+	bcm_qspi_write(qspi, BSPI, BSPI_RAF_NUM_WORDS, len_words);
+	bcm_qspi_write(qspi, BSPI, BSPI_RAF_WATERMARK, 0);
+
+	if (qspi->soc_intc) {
+		/*
+		 * clear soc MSPI and BSPI interrupts and enable
+		 * BSPI interrupts.
+		 */
+		soc_intc->bcm_qspi_int_ack(soc_intc, MSPI_BSPI_DONE);
+		soc_intc->bcm_qspi_int_set(soc_intc, BSPI_DONE, true);
+	}
+
+	/* Must flush previous writes before starting BSPI operation */
+	mb();
+
+	bcm_qspi_bspi_lr_start(qspi);
+	if (!wait_for_completion_timeout(&qspi->bspi_done, timeo)) {
+		dev_err(&qspi->pdev->dev, "timeout waiting for BSPI\n");
+		ret = -ETIMEDOUT;
+	} else {
+		/* set the return length for the caller */
+		msg->retlen = len;
+	}
+
+	return ret;
+}
+
+static int bcm_qspi_flash_read(struct spi_device *spi,
+			       struct spi_flash_read_message *msg)
+{
+	struct bcm_qspi *qspi = spi_master_get_devdata(spi->master);
+	int ret = 0;
+	bool mspi_read = false;
+	u32 io_width, addrlen, addr, len;
+	u_char *buf;
+
+	buf = msg->buf;
+	addr = msg->from;
+	len = msg->len;
+
+	if (bcm_qspi_bspi_ver_three(qspi) == true) {
+		/*
+		 * The address coming into this function is a raw flash offset.
+		 * But for BSPI <= V3, we need to convert it to a remapped BSPI
+		 * address. If it crosses a 4MB boundary, just revert back to
+		 * using MSPI.
+		 */
+		addr = (addr + 0xc00000) & 0xffffff;
+
+		if ((~ADDR_4MB_MASK & addr) ^
+		    (~ADDR_4MB_MASK & (addr + len - 1)))
+			mspi_read = true;
+	}
+
+	/* non-aligned and very short transfers are handled by MSPI */
+	if (!IS_ALIGNED((uintptr_t)addr, 4) || !IS_ALIGNED((uintptr_t)buf, 4) ||
+	    len < 4)
+		mspi_read = true;
+
+	if (mspi_read)
+		/* this will make the m25p80 read to fallback to mspi read */
+		return -EAGAIN;
+
+	io_width = msg->data_nbits ? msg->data_nbits : SPI_NBITS_SINGLE;
+	addrlen = msg->addr_width;
+	ret = bcm_qspi_bspi_set_mode(qspi, io_width, addrlen, -1);
+
+	if (!ret)
+		ret = bcm_qspi_bspi_flash_read(spi, msg);
+
+	return ret;
+}
+
+static int bcm_qspi_transfer_one(struct spi_master *master,
+				 struct spi_device *spi,
+				 struct spi_transfer *trans)
+{
+	struct bcm_qspi *qspi = spi_master_get_devdata(master);
+	int slots;
+	unsigned long timeo = msecs_to_jiffies(100);
+
+	bcm_qspi_chip_select(qspi, spi->chip_select);
+	qspi->trans_pos.trans = trans;
+	qspi->trans_pos.byte = 0;
+
+	while (qspi->trans_pos.byte < trans->len) {
+		reinit_completion(&qspi->mspi_done);
+
+		slots = write_to_hw(qspi, spi);
+		if (!wait_for_completion_timeout(&qspi->mspi_done, timeo)) {
+			dev_err(&qspi->pdev->dev, "timeout waiting for MSPI\n");
+			return -ETIMEDOUT;
+		}
+
+		read_from_hw(qspi, slots);
+	}
+
+	return 0;
+}
+
+static void bcm_qspi_cleanup(struct spi_device *spi)
+{
+	struct bcm_qspi_parms *xp = spi_get_ctldata(spi);
+
+	kfree(xp);
+}
+
+static irqreturn_t bcm_qspi_mspi_l2_isr(int irq, void *dev_id)
+{
+	struct bcm_qspi_dev_id *qspi_dev_id = dev_id;
+	struct bcm_qspi *qspi = qspi_dev_id->dev;
+	u32 status = bcm_qspi_read(qspi, MSPI, MSPI_MSPI_STATUS);
+
+	if (status & MSPI_MSPI_STATUS_SPIF) {
+		struct bcm_qspi_soc_intc *soc_intc = qspi->soc_intc;
+		/* clear interrupt */
+		status &= ~MSPI_MSPI_STATUS_SPIF;
+		bcm_qspi_write(qspi, MSPI, MSPI_MSPI_STATUS, status);
+		if (qspi->soc_intc)
+			soc_intc->bcm_qspi_int_ack(soc_intc, MSPI_DONE);
+		complete(&qspi->mspi_done);
+		return IRQ_HANDLED;
+	}
+
+	return IRQ_NONE;
+}
+
+static irqreturn_t bcm_qspi_bspi_lr_l2_isr(int irq, void *dev_id)
+{
+	struct bcm_qspi_dev_id *qspi_dev_id = dev_id;
+	struct bcm_qspi *qspi = qspi_dev_id->dev;
+	struct bcm_qspi_soc_intc *soc_intc = qspi->soc_intc;
+	u32 status = qspi_dev_id->irqp->mask;
+
+	if (qspi->bspi_enabled && qspi->bspi_rf_msg) {
+		bcm_qspi_bspi_lr_data_read(qspi);
+		if (qspi->bspi_rf_msg_len == 0) {
+			qspi->bspi_rf_msg = NULL;
+			if (qspi->soc_intc) {
+				/* disable soc BSPI interrupt */
+				soc_intc->bcm_qspi_int_set(soc_intc, BSPI_DONE,
+							   false);
+				/* indicate done */
+				status = INTR_BSPI_LR_SESSION_DONE_MASK;
+			}
+
+			if (qspi->bspi_rf_msg_status)
+				bcm_qspi_bspi_lr_clear(qspi);
+			else
+				bcm_qspi_bspi_flush_prefetch_buffers(qspi);
+		}
+
+		if (qspi->soc_intc)
+			/* clear soc BSPI interrupt */
+			soc_intc->bcm_qspi_int_ack(soc_intc, BSPI_DONE);
+	}
+
+	status &= INTR_BSPI_LR_SESSION_DONE_MASK;
+	if (qspi->bspi_enabled && status && qspi->bspi_rf_msg_len == 0)
+		complete(&qspi->bspi_done);
+
+	return IRQ_HANDLED;
+}
+
+static irqreturn_t bcm_qspi_bspi_lr_err_l2_isr(int irq, void *dev_id)
+{
+	struct bcm_qspi_dev_id *qspi_dev_id = dev_id;
+	struct bcm_qspi *qspi = qspi_dev_id->dev;
+	struct bcm_qspi_soc_intc *soc_intc = qspi->soc_intc;
+
+	dev_err(&qspi->pdev->dev, "BSPI INT error\n");
+	qspi->bspi_rf_msg_status = -EIO;
+	if (qspi->soc_intc)
+		/* clear soc interrupt */
+		soc_intc->bcm_qspi_int_ack(soc_intc, BSPI_ERR);
+
+	complete(&qspi->bspi_done);
+	return IRQ_HANDLED;
+}
+
+static irqreturn_t bcm_qspi_l1_isr(int irq, void *dev_id)
+{
+	struct bcm_qspi_dev_id *qspi_dev_id = dev_id;
+	struct bcm_qspi *qspi = qspi_dev_id->dev;
+	struct bcm_qspi_soc_intc *soc_intc = qspi->soc_intc;
+	irqreturn_t ret = IRQ_NONE;
+
+	if (soc_intc) {
+		u32 status = soc_intc->bcm_qspi_get_int_status(soc_intc);
+
+		if (status & MSPI_DONE)
+			ret = bcm_qspi_mspi_l2_isr(irq, dev_id);
+		else if (status & BSPI_DONE)
+			ret = bcm_qspi_bspi_lr_l2_isr(irq, dev_id);
+		else if (status & BSPI_ERR)
+			ret = bcm_qspi_bspi_lr_err_l2_isr(irq, dev_id);
+	}
+
+	return ret;
+}
+
+static const struct bcm_qspi_irq qspi_irq_tab[] = {
+	{
+		.irq_name = "spi_lr_fullness_reached",
+		.irq_handler = bcm_qspi_bspi_lr_l2_isr,
+		.mask = INTR_BSPI_LR_FULLNESS_REACHED_MASK,
+	},
+	{
+		.irq_name = "spi_lr_session_aborted",
+		.irq_handler = bcm_qspi_bspi_lr_err_l2_isr,
+		.mask = INTR_BSPI_LR_SESSION_ABORTED_MASK,
+	},
+	{
+		.irq_name = "spi_lr_impatient",
+		.irq_handler = bcm_qspi_bspi_lr_err_l2_isr,
+		.mask = INTR_BSPI_LR_IMPATIENT_MASK,
+	},
+	{
+		.irq_name = "spi_lr_session_done",
+		.irq_handler = bcm_qspi_bspi_lr_l2_isr,
+		.mask = INTR_BSPI_LR_SESSION_DONE_MASK,
+	},
+#ifdef QSPI_INT_DEBUG
+	/* this interrupt is for debug purposes only, dont request irq */
+	{
+		.irq_name = "spi_lr_overread",
+		.irq_handler = bcm_qspi_bspi_lr_err_l2_isr,
+		.mask = INTR_BSPI_LR_OVERREAD_MASK,
+	},
+#endif
+	{
+		.irq_name = "mspi_done",
+		.irq_handler = bcm_qspi_mspi_l2_isr,
+		.mask = INTR_MSPI_DONE_MASK,
+	},
+	{
+		.irq_name = "mspi_halted",
+		.irq_handler = bcm_qspi_mspi_l2_isr,
+		.mask = INTR_MSPI_HALTED_MASK,
+	},
+	{
+		/* single muxed L1 interrupt source */
+		.irq_name = "spi_l1_intr",
+		.irq_handler = bcm_qspi_l1_isr,
+		.irq_source = MUXED_L1,
+		.mask = QSPI_INTERRUPTS_ALL,
+	},
+};
+
+static void bcm_qspi_bspi_init(struct bcm_qspi *qspi)
+{
+	u32 val = 0;
+
+	val = bcm_qspi_read(qspi, BSPI, BSPI_REVISION_ID);
+	qspi->bspi_maj_rev = (val >> 8) & 0xff;
+	qspi->bspi_min_rev = val & 0xff;
+	if (!(bcm_qspi_bspi_ver_three(qspi))) {
+		/* Force mapping of BSPI address -> flash offset */
+		bcm_qspi_write(qspi, BSPI, BSPI_BSPI_XOR_VALUE, 0);
+		bcm_qspi_write(qspi, BSPI, BSPI_BSPI_XOR_ENABLE, 1);
+	}
+	qspi->bspi_enabled = 1;
+	bcm_qspi_disable_bspi(qspi);
+	bcm_qspi_write(qspi, BSPI, BSPI_B0_CTRL, 0);
+	bcm_qspi_write(qspi, BSPI, BSPI_B1_CTRL, 0);
+}
+
+static void bcm_qspi_hw_init(struct bcm_qspi *qspi)
+{
+	struct bcm_qspi_parms parms;
+
+	bcm_qspi_write(qspi, MSPI, MSPI_SPCR1_LSB, 0);
+	bcm_qspi_write(qspi, MSPI, MSPI_SPCR1_MSB, 0);
+	bcm_qspi_write(qspi, MSPI, MSPI_NEWQP, 0);
+	bcm_qspi_write(qspi, MSPI, MSPI_ENDQP, 0);
+	bcm_qspi_write(qspi, MSPI, MSPI_SPCR2, 0x20);
+
+	parms.mode = SPI_MODE_3;
+	parms.bits_per_word = 8;
+	parms.speed_hz = qspi->max_speed_hz;
+	bcm_qspi_hw_set_parms(qspi, &parms);
+
+	if (has_bspi(qspi))
+		bcm_qspi_bspi_init(qspi);
+}
+
+static void bcm_qspi_hw_uninit(struct bcm_qspi *qspi)
+{
+	bcm_qspi_write(qspi, MSPI, MSPI_SPCR2, 0);
+	if (has_bspi(qspi))
+		bcm_qspi_write(qspi, MSPI, MSPI_WRITE_LOCK, 0);
+
+}
+
+static const struct of_device_id bcm_qspi_of_match[] = {
+	{ .compatible = "brcm,spi-bcm-qspi" },
+	{},
+};
+MODULE_DEVICE_TABLE(of, bcm_qspi_of_match);
+
+int bcm_qspi_probe(struct platform_device *pdev,
+		   struct bcm_qspi_soc_intc *soc_intc)
+{
+	struct device *dev = &pdev->dev;
+	struct bcm_qspi *qspi;
+	struct spi_master *master;
+	struct resource *res;
+	int irq, ret = 0, num_ints = 0;
+	u32 val;
+	const char *name = NULL;
+	int num_irqs = ARRAY_SIZE(qspi_irq_tab);
+
+	/* We only support device-tree instantiation */
+	if (!dev->of_node)
+		return -ENODEV;
+
+	if (!of_match_node(bcm_qspi_of_match, dev->of_node))
+		return -ENODEV;
+
+	master = spi_alloc_master(dev, sizeof(struct bcm_qspi));
+	if (!master) {
+		dev_err(dev, "error allocating spi_master\n");
+		return -ENOMEM;
+	}
+
+	qspi = spi_master_get_devdata(master);
+	qspi->pdev = pdev;
+	qspi->trans_pos.trans = NULL;
+	qspi->trans_pos.byte = 0;
+	qspi->master = master;
+
+	master->bus_num = -1;
+	master->mode_bits = SPI_CPHA | SPI_CPOL | SPI_RX_DUAL | SPI_RX_QUAD;
+	master->setup = bcm_qspi_setup;
+	master->transfer_one = bcm_qspi_transfer_one;
+	master->spi_flash_read = bcm_qspi_flash_read;
+	master->cleanup = bcm_qspi_cleanup;
+	master->dev.of_node = dev->of_node;
+	master->num_chipselect = NUM_CHIPSELECT;
+
+	qspi->big_endian = of_device_is_big_endian(dev->of_node);
+
+	if (!of_property_read_u32(dev->of_node, "num-cs", &val))
+		master->num_chipselect = val;
+
+	res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "hif_mspi");
+	if (!res)
+		res = platform_get_resource_byname(pdev, IORESOURCE_MEM,
+						   "mspi");
+
+	if (res) {
+		qspi->base[MSPI]  = devm_ioremap_resource(dev, res);
+		if (IS_ERR(qspi->base[MSPI])) {
+			ret = PTR_ERR(qspi->base[MSPI]);
+			goto qspi_probe_err;
+		}
+	} else {
+		goto qspi_probe_err;
+	}
+
+	res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "bspi");
+	if (res) {
+		qspi->base[BSPI]  = devm_ioremap_resource(dev, res);
+		if (IS_ERR(qspi->base[BSPI])) {
+			ret = PTR_ERR(qspi->base[BSPI]);
+			goto qspi_probe_err;
+		}
+		qspi->bspi_mode = true;
+	} else {
+		qspi->bspi_mode = false;
+	}
+
+	dev_info(dev, "using %smspi mode\n", qspi->bspi_mode ? "bspi-" : "");
+
+	res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "cs_reg");
+	if (res) {
+		qspi->base[CHIP_SELECT]  = devm_ioremap_resource(dev, res);
+		if (IS_ERR(qspi->base[CHIP_SELECT])) {
+			ret = PTR_ERR(qspi->base[CHIP_SELECT]);
+			goto qspi_probe_err;
+		}
+	}
+
+	qspi->dev_ids = kcalloc(num_irqs, sizeof(struct bcm_qspi_dev_id),
+				GFP_KERNEL);
+	if (!qspi->dev_ids) {
+		ret = -ENOMEM;
+		goto qspi_probe_err;
+	}
+
+	for (val = 0; val < num_irqs; val++) {
+		irq = -1;
+		name = qspi_irq_tab[val].irq_name;
+		if (qspi_irq_tab[val].irq_source == SINGLE_L2) {
+			/* get the l2 interrupts */
+			irq = platform_get_irq_byname(pdev, name);
+		} else if (!num_ints && soc_intc) {
+			/* all mspi, bspi intrs muxed to one L1 intr */
+			irq = platform_get_irq(pdev, 0);
+		}
+
+		if (irq  >= 0) {
+			ret = devm_request_irq(&pdev->dev, irq,
+					       qspi_irq_tab[val].irq_handler, 0,
+					       name,
+					       &qspi->dev_ids[val]);
+			if (ret < 0) {
+				dev_err(&pdev->dev, "IRQ %s not found\n", name);
+				goto qspi_probe_err;
+			}
+
+			qspi->dev_ids[val].dev = qspi;
+			qspi->dev_ids[val].irqp = &qspi_irq_tab[val];
+			num_ints++;
+			dev_dbg(&pdev->dev, "registered IRQ %s %d\n",
+				qspi_irq_tab[val].irq_name,
+				irq);
+		}
+	}
+
+	if (!num_ints) {
+		dev_err(&pdev->dev, "no IRQs registered, cannot init driver\n");
+		ret = -EINVAL;
+		goto qspi_probe_err;
+	}
+
+	/*
+	 * Some SoCs integrate spi controller (e.g., its interrupt bits)
+	 * in specific ways
+	 */
+	if (soc_intc) {
+		qspi->soc_intc = soc_intc;
+		soc_intc->bcm_qspi_int_set(soc_intc, MSPI_DONE, true);
+	} else {
+		qspi->soc_intc = NULL;
+	}
+
+	qspi->clk = devm_clk_get(&pdev->dev, NULL);
+	if (IS_ERR(qspi->clk)) {
+		dev_warn(dev, "unable to get clock\n");
+		ret = PTR_ERR(qspi->clk);
+		goto qspi_probe_err;
+	}
+
+	ret = clk_prepare_enable(qspi->clk);
+	if (ret) {
+		dev_err(dev, "failed to prepare clock\n");
+		goto qspi_probe_err;
+	}
+
+	qspi->base_clk = clk_get_rate(qspi->clk);
+	qspi->max_speed_hz = qspi->base_clk / (QSPI_SPBR_MIN * 2);
+
+	bcm_qspi_hw_init(qspi);
+	init_completion(&qspi->mspi_done);
+	init_completion(&qspi->bspi_done);
+	qspi->curr_cs = -1;
+
+	platform_set_drvdata(pdev, qspi);
+
+	qspi->xfer_mode.width = -1;
+	qspi->xfer_mode.addrlen = -1;
+	qspi->xfer_mode.hp = -1;
+
+	ret = devm_spi_register_master(&pdev->dev, master);
+	if (ret < 0) {
+		dev_err(dev, "can't register master\n");
+		goto qspi_reg_err;
+	}
+
+	return 0;
+
+qspi_reg_err:
+	bcm_qspi_hw_uninit(qspi);
+	clk_disable_unprepare(qspi->clk);
+qspi_probe_err:
+	spi_master_put(master);
+	kfree(qspi->dev_ids);
+	return ret;
+}
+/* probe function to be called by SoC specific platform driver probe */
+EXPORT_SYMBOL_GPL(bcm_qspi_probe);
+
+int bcm_qspi_remove(struct platform_device *pdev)
+{
+	struct bcm_qspi *qspi = platform_get_drvdata(pdev);
+
+	platform_set_drvdata(pdev, NULL);
+	bcm_qspi_hw_uninit(qspi);
+	clk_disable_unprepare(qspi->clk);
+	kfree(qspi->dev_ids);
+	spi_unregister_master(qspi->master);
+
+	return 0;
+}
+/* function to be called by SoC specific platform driver remove() */
+EXPORT_SYMBOL_GPL(bcm_qspi_remove);
+
+static int __maybe_unused bcm_qspi_suspend(struct device *dev)
+{
+	struct bcm_qspi *qspi = dev_get_drvdata(dev);
+
+	spi_master_suspend(qspi->master);
+	clk_disable(qspi->clk);
+	bcm_qspi_hw_uninit(qspi);
+
+	return 0;
+};
+
+static int __maybe_unused bcm_qspi_resume(struct device *dev)
+{
+	struct bcm_qspi *qspi = dev_get_drvdata(dev);
+	int ret = 0;
+
+	bcm_qspi_hw_init(qspi);
+	bcm_qspi_chip_select(qspi, qspi->curr_cs);
+	if (qspi->soc_intc)
+		/* enable MSPI interrupt */
+		qspi->soc_intc->bcm_qspi_int_set(qspi->soc_intc, MSPI_DONE,
+						 true);
+
+	ret = clk_enable(qspi->clk);
+	if (!ret)
+		spi_master_resume(qspi->master);
+
+	return ret;
+}
+
+SIMPLE_DEV_PM_OPS(bcm_qspi_pm_ops, bcm_qspi_suspend, bcm_qspi_resume);
+
+/* pm_ops to be called by SoC specific platform driver */
+EXPORT_SYMBOL_GPL(bcm_qspi_pm_ops);
+
+MODULE_AUTHOR("Kamal Dasu");
+MODULE_DESCRIPTION("Broadcom QSPI driver");
+MODULE_LICENSE("GPL v2");
+MODULE_ALIAS("platform:" DRIVER_NAME);
diff --git a/drivers/spi/spi-bcm-qspi.h b/drivers/spi/spi-bcm-qspi.h
new file mode 100644
index 000000000000..7abfc75a3860
--- /dev/null
+++ b/drivers/spi/spi-bcm-qspi.h
@@ -0,0 +1,115 @@
+/*
+ * Copyright 2016 Broadcom
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, version 2, as
+ * published by the Free Software Foundation (the "GPL").
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License version 2 (GPLv2) for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * version 2 (GPLv2) along with this source code.
+ */
+
+#ifndef __SPI_BCM_QSPI_H__
+#define __SPI_BCM_QSPI_H__
+
+#include <linux/types.h>
+#include <linux/io.h>
+
+/* BSPI interrupt masks */
+#define INTR_BSPI_LR_OVERREAD_MASK		BIT(4)
+#define INTR_BSPI_LR_SESSION_DONE_MASK		BIT(3)
+#define INTR_BSPI_LR_IMPATIENT_MASK		BIT(2)
+#define INTR_BSPI_LR_SESSION_ABORTED_MASK	BIT(1)
+#define INTR_BSPI_LR_FULLNESS_REACHED_MASK	BIT(0)
+
+#define BSPI_LR_INTERRUPTS_DATA		       \
+	(INTR_BSPI_LR_SESSION_DONE_MASK |	       \
+	 INTR_BSPI_LR_FULLNESS_REACHED_MASK)
+
+#define BSPI_LR_INTERRUPTS_ERROR               \
+	(INTR_BSPI_LR_OVERREAD_MASK |	       \
+	 INTR_BSPI_LR_IMPATIENT_MASK |	       \
+	 INTR_BSPI_LR_SESSION_ABORTED_MASK)
+
+#define BSPI_LR_INTERRUPTS_ALL                 \
+	(BSPI_LR_INTERRUPTS_ERROR |	       \
+	 BSPI_LR_INTERRUPTS_DATA)
+
+/* MSPI Interrupt masks */
+#define INTR_MSPI_HALTED_MASK			BIT(6)
+#define INTR_MSPI_DONE_MASK			BIT(5)
+
+#define MSPI_INTERRUPTS_ALL		       \
+	(INTR_MSPI_DONE_MASK |		       \
+	 INTR_MSPI_HALTED_MASK)
+
+#define QSPI_INTERRUPTS_ALL                    \
+	(MSPI_INTERRUPTS_ALL |		       \
+	 BSPI_LR_INTERRUPTS_ALL)
+
+struct platform_device;
+struct dev_pm_ops;
+
+enum {
+	MSPI_DONE = 0x1,
+	BSPI_DONE = 0x2,
+	BSPI_ERR = 0x4,
+	MSPI_BSPI_DONE = 0x7
+};
+
+struct bcm_qspi_soc_intc {
+	void (*bcm_qspi_int_ack)(struct bcm_qspi_soc_intc *soc_intc, int type);
+	void (*bcm_qspi_int_set)(struct bcm_qspi_soc_intc *soc_intc, int type,
+				 bool en);
+	u32 (*bcm_qspi_get_int_status)(struct bcm_qspi_soc_intc *soc_intc);
+};
+
+/* Read controller register*/
+static inline u32 bcm_qspi_readl(bool be, void __iomem *addr)
+{
+	if (be)
+		return ioread32be(addr);
+	else
+		return readl_relaxed(addr);
+}
+
+/* Write controller register*/
+static inline void bcm_qspi_writel(bool be,
+				   unsigned int data, void __iomem *addr)
+{
+	if (be)
+		iowrite32be(data, addr);
+	else
+		writel_relaxed(data, addr);
+}
+
+static inline u32 get_qspi_mask(int type)
+{
+	switch (type) {
+	case MSPI_DONE:
+		return INTR_MSPI_DONE_MASK;
+	case BSPI_DONE:
+		return BSPI_LR_INTERRUPTS_ALL;
+	case MSPI_BSPI_DONE:
+		return QSPI_INTERRUPTS_ALL;
+	case BSPI_ERR:
+		return BSPI_LR_INTERRUPTS_ERROR;
+	}
+
+	return 0;
+}
+
+/* The common driver functions to be called by the SoC platform driver */
+int bcm_qspi_probe(struct platform_device *pdev,
+		   struct bcm_qspi_soc_intc *soc_intc);
+int bcm_qspi_remove(struct platform_device *pdev);
+
+/* pm_ops used by the SoC platform driver called on PM suspend/resume */
+extern const struct dev_pm_ops bcm_qspi_pm_ops;
+
+#endif /* __SPI_BCM_QSPI_H__ */
diff --git a/drivers/spi/spi-brcmstb-qspi.c b/drivers/spi/spi-brcmstb-qspi.c
new file mode 100644
index 000000000000..c7df92e7cf6f
--- /dev/null
+++ b/drivers/spi/spi-brcmstb-qspi.c
@@ -0,0 +1,53 @@
+/*
+ * Copyright 2016 Broadcom
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License, version 2, as
+ * published by the Free Software Foundation (the "GPL").
+ *
+ * This program is distributed in the hope that it will be useful, but
+ * WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License version 2 (GPLv2) for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * version 2 (GPLv2) along with this source code.
+ */
+
+#include <linux/device.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/of.h>
+#include "spi-bcm-qspi.h"
+
+static const struct of_device_id brcmstb_qspi_of_match[] = {
+	{ .compatible = "brcm,spi-brcmstb-qspi" },
+	{ .compatible = "brcm,spi-brcmstb-mspi" },
+	{},
+};
+MODULE_DEVICE_TABLE(of, brcmstb_qspi_of_match);
+
+static int brcmstb_qspi_probe(struct platform_device *pdev)
+{
+	return bcm_qspi_probe(pdev, NULL);
+}
+
+static int brcmstb_qspi_remove(struct platform_device *pdev)
+{
+	return bcm_qspi_remove(pdev);
+}
+
+static struct platform_driver brcmstb_qspi_driver = {
+	.probe			= brcmstb_qspi_probe,
+	.remove			= brcmstb_qspi_remove,
+	.driver = {
+		.name		= "brcmstb_qspi",
+		.pm		= &bcm_qspi_pm_ops,
+		.of_match_table = brcmstb_qspi_of_match,
+	}
+};
+module_platform_driver(brcmstb_qspi_driver);
+
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Kamal Dasu");
+MODULE_DESCRIPTION("Broadcom SPI driver for settop SoC");
diff --git a/drivers/spi/spi-cavium-thunderx.c b/drivers/spi/spi-cavium-thunderx.c
new file mode 100644
index 000000000000..877937706240
--- /dev/null
+++ b/drivers/spi/spi-cavium-thunderx.c
@@ -0,0 +1,120 @@
+/*
+ * Cavium ThunderX SPI driver.
+ *
+ * Copyright (C) 2016 Cavium Inc.
+ * Authors: Jan Glauber <jglauber@cavium.com>
+ */
+
+#include <linux/module.h>
+#include <linux/pci.h>
+#include <linux/spi/spi.h>
+
+#include "spi-cavium.h"
+
+#define DRV_NAME "spi-thunderx"
+
+#define SYS_FREQ_DEFAULT 700000000 /* 700 Mhz */
+
+static int thunderx_spi_probe(struct pci_dev *pdev,
+			      const struct pci_device_id *ent)
+{
+	struct device *dev = &pdev->dev;
+	struct spi_master *master;
+	struct octeon_spi *p;
+	int ret;
+
+	master = spi_alloc_master(dev, sizeof(struct octeon_spi));
+	if (!master)
+		return -ENOMEM;
+
+	p = spi_master_get_devdata(master);
+
+	ret = pcim_enable_device(pdev);
+	if (ret)
+		goto error;
+
+	ret = pci_request_regions(pdev, DRV_NAME);
+	if (ret)
+		goto error;
+
+	p->register_base = pcim_iomap(pdev, 0, pci_resource_len(pdev, 0));
+	if (!p->register_base) {
+		ret = -EINVAL;
+		goto error;
+	}
+
+	p->regs.config = 0x1000;
+	p->regs.status = 0x1008;
+	p->regs.tx = 0x1010;
+	p->regs.data = 0x1080;
+
+	p->clk = devm_clk_get(dev, NULL);
+	if (IS_ERR(p->clk)) {
+		ret = PTR_ERR(p->clk);
+		goto error;
+	}
+
+	ret = clk_prepare_enable(p->clk);
+	if (ret)
+		goto error;
+
+	p->sys_freq = clk_get_rate(p->clk);
+	if (!p->sys_freq)
+		p->sys_freq = SYS_FREQ_DEFAULT;
+	dev_info(dev, "Set system clock to %u\n", p->sys_freq);
+
+	master->num_chipselect = 4;
+	master->mode_bits = SPI_CPHA | SPI_CPOL | SPI_CS_HIGH |
+			    SPI_LSB_FIRST | SPI_3WIRE;
+	master->transfer_one_message = octeon_spi_transfer_one_message;
+	master->bits_per_word_mask = SPI_BPW_MASK(8);
+	master->max_speed_hz = OCTEON_SPI_MAX_CLOCK_HZ;
+	master->dev.of_node = pdev->dev.of_node;
+
+	pci_set_drvdata(pdev, master);
+
+	ret = devm_spi_register_master(dev, master);
+	if (ret)
+		goto error;
+
+	return 0;
+
+error:
+	clk_disable_unprepare(p->clk);
+	spi_master_put(master);
+	return ret;
+}
+
+static void thunderx_spi_remove(struct pci_dev *pdev)
+{
+	struct spi_master *master = pci_get_drvdata(pdev);
+	struct octeon_spi *p;
+
+	p = spi_master_get_devdata(master);
+	if (!p)
+		return;
+
+	clk_disable_unprepare(p->clk);
+	/* Put everything in a known state. */
+	writeq(0, p->register_base + OCTEON_SPI_CFG(p));
+}
+
+static const struct pci_device_id thunderx_spi_pci_id_table[] = {
+	{ PCI_DEVICE(PCI_VENDOR_ID_CAVIUM, 0xa00b) },
+	{ 0, }
+};
+
+MODULE_DEVICE_TABLE(pci, thunderx_spi_pci_id_table);
+
+static struct pci_driver thunderx_spi_driver = {
+	.name		= DRV_NAME,
+	.id_table	= thunderx_spi_pci_id_table,
+	.probe		= thunderx_spi_probe,
+	.remove		= thunderx_spi_remove,
+};
+
+module_pci_driver(thunderx_spi_driver);
+
+MODULE_DESCRIPTION("Cavium, Inc. ThunderX SPI bus driver");
+MODULE_AUTHOR("Jan Glauber");
+MODULE_LICENSE("GPL");
diff --git a/drivers/spi/spi-cavium.h b/drivers/spi/spi-cavium.h
index 88c5f36e7ea7..1f91d61b745b 100644
--- a/drivers/spi/spi-cavium.h
+++ b/drivers/spi/spi-cavium.h
@@ -1,6 +1,8 @@
 #ifndef __SPI_CAVIUM_H
 #define __SPI_CAVIUM_H
 
+#include <linux/clk.h>
+
 #define OCTEON_SPI_MAX_BYTES 9
 #define OCTEON_SPI_MAX_CLOCK_HZ 16000000
 
@@ -17,6 +19,7 @@ struct octeon_spi {
 	u64 cs_enax;
 	int sys_freq;
 	struct octeon_spi_regs regs;
+	struct clk *clk;
 };
 
 #define OCTEON_SPI_CFG(x)	(x->regs.config)
diff --git a/drivers/spi/spi-dw.c b/drivers/spi/spi-dw.c
index c09bb745693a..27960e46135d 100644
--- a/drivers/spi/spi-dw.c
+++ b/drivers/spi/spi-dw.c
@@ -283,7 +283,6 @@ static int dw_spi_transfer_one(struct spi_master *master,
 	struct chip_data *chip = spi_get_ctldata(spi);
 	u8 imask = 0;
 	u16 txlevel = 0;
-	u16 clk_div;
 	u32 cr0;
 	int ret;
 
@@ -298,13 +297,13 @@ static int dw_spi_transfer_one(struct spi_master *master,
 	spi_enable_chip(dws, 0);
 
 	/* Handle per transfer options for bpw and speed */
-	if (transfer->speed_hz != chip->speed_hz) {
-		/* clk_div doesn't support odd number */
-		clk_div = (dws->max_freq / transfer->speed_hz + 1) & 0xfffe;
-
-		chip->speed_hz = transfer->speed_hz;
-		chip->clk_div = clk_div;
-
+	if (transfer->speed_hz != dws->current_freq) {
+		if (transfer->speed_hz != chip->speed_hz) {
+			/* clk_div doesn't support odd number */
+			chip->clk_div = (DIV_ROUND_UP(dws->max_freq, transfer->speed_hz) + 1) & 0xfffe;
+			chip->speed_hz = transfer->speed_hz;
+		}
+		dws->current_freq = transfer->speed_hz;
 		spi_set_clk(dws, chip->clk_div);
 	}
 	if (transfer->bits_per_word == 8) {
diff --git a/drivers/spi/spi-dw.h b/drivers/spi/spi-dw.h
index 61bc3cbab38d..c21ca02f8ec5 100644
--- a/drivers/spi/spi-dw.h
+++ b/drivers/spi/spi-dw.h
@@ -123,6 +123,7 @@ struct dw_spi {
 	u8			n_bytes;	/* current is a 1/2 bytes op */
 	u32			dma_width;
 	irqreturn_t		(*transfer_handler)(struct dw_spi *dws);
+	u32			current_freq;	/* frequency in hz */
 
 	/* DMA info */
 	int			dma_inited;
diff --git a/drivers/spi/spi-fsl-dspi.c b/drivers/spi/spi-fsl-dspi.c
index 9e9dadb52b3d..35c0dd945668 100644
--- a/drivers/spi/spi-fsl-dspi.c
+++ b/drivers/spi/spi-fsl-dspi.c
@@ -159,7 +159,7 @@ struct fsl_dspi {
 	u8			cs;
 	u16			void_write_data;
 	u32			cs_change;
-	struct fsl_dspi_devtype_data *devtype_data;
+	const struct fsl_dspi_devtype_data *devtype_data;
 
 	wait_queue_head_t	waitq;
 	u32			waitflags;
@@ -624,10 +624,13 @@ static int dspi_resume(struct device *dev)
 {
 	struct spi_master *master = dev_get_drvdata(dev);
 	struct fsl_dspi *dspi = spi_master_get_devdata(master);
+	int ret;
 
 	pinctrl_pm_select_default_state(dev);
 
-	clk_prepare_enable(dspi->clk);
+	ret = clk_prepare_enable(dspi->clk);
+	if (ret)
+		return ret;
 	spi_master_resume(master);
 
 	return 0;
@@ -651,8 +654,6 @@ static int dspi_probe(struct platform_device *pdev)
 	struct resource *res;
 	void __iomem *base;
 	int ret = 0, cs_num, bus_num;
-	const struct of_device_id *of_id =
-			of_match_device(fsl_dspi_dt_ids, &pdev->dev);
 
 	master = spi_alloc_master(&pdev->dev, sizeof(struct fsl_dspi));
 	if (!master)
@@ -686,7 +687,7 @@ static int dspi_probe(struct platform_device *pdev)
 	}
 	master->bus_num = bus_num;
 
-	dspi->devtype_data = (struct fsl_dspi_devtype_data *)of_id->data;
+	dspi->devtype_data = of_device_get_match_data(&pdev->dev);
 	if (!dspi->devtype_data) {
 		dev_err(&pdev->dev, "can't get devtype_data\n");
 		ret = -EFAULT;
@@ -728,7 +729,9 @@ static int dspi_probe(struct platform_device *pdev)
 		dev_err(&pdev->dev, "unable to get clock\n");
 		goto out_master_put;
 	}
-	clk_prepare_enable(dspi->clk);
+	ret = clk_prepare_enable(dspi->clk);
+	if (ret)
+		goto out_master_put;
 
 	master->max_speed_hz =
 		clk_get_rate(dspi->clk) / dspi->devtype_data->max_clock_factor;
@@ -760,7 +763,6 @@ static int dspi_remove(struct platform_device *pdev)
 	/* Disconnect from the SPI framework */
 	clk_disable_unprepare(dspi->clk);
 	spi_unregister_master(dspi->master);
-	spi_master_put(dspi->master);
 
 	return 0;
 }
diff --git a/drivers/spi/spi-fsl-espi.c b/drivers/spi/spi-fsl-espi.c
index 8d85a3c343da..7451585a080e 100644
--- a/drivers/spi/spi-fsl-espi.c
+++ b/drivers/spi/spi-fsl-espi.c
@@ -12,7 +12,6 @@
 #include <linux/err.h>
 #include <linux/fsl_devices.h>
 #include <linux/interrupt.h>
-#include <linux/irq.h>
 #include <linux/module.h>
 #include <linux/mm.h>
 #include <linux/of.h>
@@ -27,40 +26,29 @@
 #include "spi-fsl-lib.h"
 
 /* eSPI Controller registers */
-struct fsl_espi_reg {
-	__be32 mode;		/* 0x000 - eSPI mode register */
-	__be32 event;		/* 0x004 - eSPI event register */
-	__be32 mask;		/* 0x008 - eSPI mask register */
-	__be32 command;		/* 0x00c - eSPI command register */
-	__be32 transmit;	/* 0x010 - eSPI transmit FIFO access register*/
-	__be32 receive;		/* 0x014 - eSPI receive FIFO access register*/
-	u8 res[8];		/* 0x018 - 0x01c reserved */
-	__be32 csmode[4];	/* 0x020 - 0x02c eSPI cs mode register */
-};
+#define ESPI_SPMODE	0x00	/* eSPI mode register */
+#define ESPI_SPIE	0x04	/* eSPI event register */
+#define ESPI_SPIM	0x08	/* eSPI mask register */
+#define ESPI_SPCOM	0x0c	/* eSPI command register */
+#define ESPI_SPITF	0x10	/* eSPI transmit FIFO access register*/
+#define ESPI_SPIRF	0x14	/* eSPI receive FIFO access register*/
+#define ESPI_SPMODE0	0x20	/* eSPI cs0 mode register */
 
-struct fsl_espi_transfer {
-	const void *tx_buf;
-	void *rx_buf;
-	unsigned len;
-	unsigned n_tx;
-	unsigned n_rx;
-	unsigned actual_length;
-	int status;
-};
+#define ESPI_SPMODEx(x)	(ESPI_SPMODE0 + (x) * 4)
 
 /* eSPI Controller mode register definitions */
-#define SPMODE_ENABLE		(1 << 31)
-#define SPMODE_LOOP		(1 << 30)
+#define SPMODE_ENABLE		BIT(31)
+#define SPMODE_LOOP		BIT(30)
 #define SPMODE_TXTHR(x)		((x) << 8)
 #define SPMODE_RXTHR(x)		((x) << 0)
 
 /* eSPI Controller CS mode register definitions */
-#define CSMODE_CI_INACTIVEHIGH	(1 << 31)
-#define CSMODE_CP_BEGIN_EDGECLK	(1 << 30)
-#define CSMODE_REV		(1 << 29)
-#define CSMODE_DIV16		(1 << 28)
+#define CSMODE_CI_INACTIVEHIGH	BIT(31)
+#define CSMODE_CP_BEGIN_EDGECLK	BIT(30)
+#define CSMODE_REV		BIT(29)
+#define CSMODE_DIV16		BIT(28)
 #define CSMODE_PM(x)		((x) << 24)
-#define CSMODE_POL_1		(1 << 20)
+#define CSMODE_POL_1		BIT(20)
 #define CSMODE_LEN(x)		((x) << 16)
 #define CSMODE_BEF(x)		((x) << 12)
 #define CSMODE_AFT(x)		((x) << 8)
@@ -72,29 +60,114 @@ struct fsl_espi_transfer {
 		| CSMODE_AFT(0) | CSMODE_CG(1))
 
 /* SPIE register values */
-#define	SPIE_NE		0x00000200	/* Not empty */
-#define	SPIE_NF		0x00000100	/* Not full */
-
-/* SPIM register values */
-#define	SPIM_NE		0x00000200	/* Not empty */
-#define	SPIM_NF		0x00000100	/* Not full */
 #define SPIE_RXCNT(reg)     ((reg >> 24) & 0x3F)
 #define SPIE_TXCNT(reg)     ((reg >> 16) & 0x3F)
+#define	SPIE_TXE		BIT(15)	/* TX FIFO empty */
+#define	SPIE_DON		BIT(14)	/* TX done */
+#define	SPIE_RXT		BIT(13)	/* RX FIFO threshold */
+#define	SPIE_RXF		BIT(12)	/* RX FIFO full */
+#define	SPIE_TXT		BIT(11)	/* TX FIFO threshold*/
+#define	SPIE_RNE		BIT(9)	/* RX FIFO not empty */
+#define	SPIE_TNF		BIT(8)	/* TX FIFO not full */
+
+/* SPIM register values */
+#define	SPIM_TXE		BIT(15)	/* TX FIFO empty */
+#define	SPIM_DON		BIT(14)	/* TX done */
+#define	SPIM_RXT		BIT(13)	/* RX FIFO threshold */
+#define	SPIM_RXF		BIT(12)	/* RX FIFO full */
+#define	SPIM_TXT		BIT(11)	/* TX FIFO threshold*/
+#define	SPIM_RNE		BIT(9)	/* RX FIFO not empty */
+#define	SPIM_TNF		BIT(8)	/* TX FIFO not full */
 
 /* SPCOM register values */
 #define SPCOM_CS(x)		((x) << 30)
+#define SPCOM_DO		BIT(28) /* Dual output */
+#define SPCOM_TO		BIT(27) /* TX only */
+#define SPCOM_RXSKIP(x)		((x) << 16)
 #define SPCOM_TRANLEN(x)	((x) << 0)
+
 #define	SPCOM_TRANLEN_MAX	0x10000	/* Max transaction length */
 
 #define AUTOSUSPEND_TIMEOUT 2000
 
+static inline u32 fsl_espi_read_reg(struct mpc8xxx_spi *mspi, int offset)
+{
+	return ioread32be(mspi->reg_base + offset);
+}
+
+static inline u8 fsl_espi_read_reg8(struct mpc8xxx_spi *mspi, int offset)
+{
+	return ioread8(mspi->reg_base + offset);
+}
+
+static inline void fsl_espi_write_reg(struct mpc8xxx_spi *mspi, int offset,
+				      u32 val)
+{
+	iowrite32be(val, mspi->reg_base + offset);
+}
+
+static inline void fsl_espi_write_reg8(struct mpc8xxx_spi *mspi, int offset,
+				       u8 val)
+{
+	iowrite8(val, mspi->reg_base + offset);
+}
+
+static void fsl_espi_copy_to_buf(struct spi_message *m,
+				 struct mpc8xxx_spi *mspi)
+{
+	struct spi_transfer *t;
+	u8 *buf = mspi->local_buf;
+
+	list_for_each_entry(t, &m->transfers, transfer_list) {
+		if (t->tx_buf)
+			memcpy(buf, t->tx_buf, t->len);
+		else
+			memset(buf, 0, t->len);
+		buf += t->len;
+	}
+}
+
+static void fsl_espi_copy_from_buf(struct spi_message *m,
+				   struct mpc8xxx_spi *mspi)
+{
+	struct spi_transfer *t;
+	u8 *buf = mspi->local_buf;
+
+	list_for_each_entry(t, &m->transfers, transfer_list) {
+		if (t->rx_buf)
+			memcpy(t->rx_buf, buf, t->len);
+		buf += t->len;
+	}
+}
+
+static int fsl_espi_check_message(struct spi_message *m)
+{
+	struct mpc8xxx_spi *mspi = spi_master_get_devdata(m->spi->master);
+	struct spi_transfer *t, *first;
+
+	if (m->frame_length > SPCOM_TRANLEN_MAX) {
+		dev_err(mspi->dev, "message too long, size is %u bytes\n",
+			m->frame_length);
+		return -EMSGSIZE;
+	}
+
+	first = list_first_entry(&m->transfers, struct spi_transfer,
+				 transfer_list);
+	list_for_each_entry(t, &m->transfers, transfer_list) {
+		if (first->bits_per_word != t->bits_per_word ||
+		    first->speed_hz != t->speed_hz) {
+			dev_err(mspi->dev, "bits_per_word/speed_hz should be the same for all transfers\n");
+			return -EINVAL;
+		}
+	}
+
+	return 0;
+}
+
 static void fsl_espi_change_mode(struct spi_device *spi)
 {
 	struct mpc8xxx_spi *mspi = spi_master_get_devdata(spi->master);
 	struct spi_mpc8xxx_cs *cs = spi->controller_state;
-	struct fsl_espi_reg *reg_base = mspi->reg_base;
-	__be32 __iomem *mode = &reg_base->csmode[spi->chip_select];
-	__be32 __iomem *espi_mode = &reg_base->mode;
 	u32 tmp;
 	unsigned long flags;
 
@@ -102,10 +175,11 @@ static void fsl_espi_change_mode(struct spi_device *spi)
 	local_irq_save(flags);
 
 	/* Turn off SPI unit prior changing mode */
-	tmp = mpc8xxx_spi_read_reg(espi_mode);
-	mpc8xxx_spi_write_reg(espi_mode, tmp & ~SPMODE_ENABLE);
-	mpc8xxx_spi_write_reg(mode, cs->hw_mode);
-	mpc8xxx_spi_write_reg(espi_mode, tmp);
+	tmp = fsl_espi_read_reg(mspi, ESPI_SPMODE);
+	fsl_espi_write_reg(mspi, ESPI_SPMODE, tmp & ~SPMODE_ENABLE);
+	fsl_espi_write_reg(mspi, ESPI_SPMODEx(spi->chip_select),
+			      cs->hw_mode);
+	fsl_espi_write_reg(mspi, ESPI_SPMODE, tmp);
 
 	local_irq_restore(flags);
 }
@@ -131,27 +205,15 @@ static u32 fsl_espi_tx_buf_lsb(struct mpc8xxx_spi *mpc8xxx_spi)
 	return data;
 }
 
-static int fsl_espi_setup_transfer(struct spi_device *spi,
+static void fsl_espi_setup_transfer(struct spi_device *spi,
 					struct spi_transfer *t)
 {
 	struct mpc8xxx_spi *mpc8xxx_spi = spi_master_get_devdata(spi->master);
-	int bits_per_word = 0;
+	int bits_per_word = t ? t->bits_per_word : spi->bits_per_word;
+	u32 hz = t ? t->speed_hz : spi->max_speed_hz;
 	u8 pm;
-	u32 hz = 0;
 	struct spi_mpc8xxx_cs *cs = spi->controller_state;
 
-	if (t) {
-		bits_per_word = t->bits_per_word;
-		hz = t->speed_hz;
-	}
-
-	/* spi_transfer level calls that work per-word */
-	if (!bits_per_word)
-		bits_per_word = spi->bits_per_word;
-
-	if (!hz)
-		hz = spi->max_speed_hz;
-
 	cs->rx_shift = 0;
 	cs->tx_shift = 0;
 	cs->get_rx = mpc8xxx_spi_rx_buf_u32;
@@ -169,12 +231,10 @@ static int fsl_espi_setup_transfer(struct spi_device *spi,
 	mpc8xxx_spi->get_rx = cs->get_rx;
 	mpc8xxx_spi->get_tx = cs->get_tx;
 
-	bits_per_word = bits_per_word - 1;
-
 	/* mask out bits we are going to set */
 	cs->hw_mode &= ~(CSMODE_LEN(0xF) | CSMODE_DIV16 | CSMODE_PM(0xF));
 
-	cs->hw_mode |= CSMODE_LEN(bits_per_word);
+	cs->hw_mode |= CSMODE_LEN(bits_per_word - 1);
 
 	if ((mpc8xxx_spi->spibrg / hz) > 64) {
 		cs->hw_mode |= CSMODE_DIV16;
@@ -196,36 +256,16 @@ static int fsl_espi_setup_transfer(struct spi_device *spi,
 	cs->hw_mode |= CSMODE_PM(pm);
 
 	fsl_espi_change_mode(spi);
-	return 0;
-}
-
-static int fsl_espi_cpu_bufs(struct mpc8xxx_spi *mspi, struct spi_transfer *t,
-		unsigned int len)
-{
-	u32 word;
-	struct fsl_espi_reg *reg_base = mspi->reg_base;
-
-	mspi->count = len;
-
-	/* enable rx ints */
-	mpc8xxx_spi_write_reg(&reg_base->mask, SPIM_NE);
-
-	/* transmit word */
-	word = mspi->get_tx(mspi);
-	mpc8xxx_spi_write_reg(&reg_base->transmit, word);
-
-	return 0;
 }
 
 static int fsl_espi_bufs(struct spi_device *spi, struct spi_transfer *t)
 {
 	struct mpc8xxx_spi *mpc8xxx_spi = spi_master_get_devdata(spi->master);
-	struct fsl_espi_reg *reg_base = mpc8xxx_spi->reg_base;
-	unsigned int len = t->len;
+	u32 word;
 	int ret;
 
 	mpc8xxx_spi->len = t->len;
-	len = roundup(len, 4) / 4;
+	mpc8xxx_spi->count = roundup(t->len, 4) / 4;
 
 	mpc8xxx_spi->tx = t->tx_buf;
 	mpc8xxx_spi->rx = t->rx_buf;
@@ -233,17 +273,15 @@ static int fsl_espi_bufs(struct spi_device *spi, struct spi_transfer *t)
 	reinit_completion(&mpc8xxx_spi->done);
 
 	/* Set SPCOM[CS] and SPCOM[TRANLEN] field */
-	if (t->len > SPCOM_TRANLEN_MAX) {
-		dev_err(mpc8xxx_spi->dev, "Transaction length (%d)"
-				" beyond the SPCOM[TRANLEN] field\n", t->len);
-		return -EINVAL;
-	}
-	mpc8xxx_spi_write_reg(&reg_base->command,
+	fsl_espi_write_reg(mpc8xxx_spi, ESPI_SPCOM,
 		(SPCOM_CS(spi->chip_select) | SPCOM_TRANLEN(t->len - 1)));
 
-	ret = fsl_espi_cpu_bufs(mpc8xxx_spi, t, len);
-	if (ret)
-		return ret;
+	/* enable rx ints */
+	fsl_espi_write_reg(mpc8xxx_spi, ESPI_SPIM, SPIM_RNE);
+
+	/* transmit word */
+	word = mpc8xxx_spi->get_tx(mpc8xxx_spi);
+	fsl_espi_write_reg(mpc8xxx_spi, ESPI_SPITF, word);
 
 	/* Won't hang up forever, SPI bus sometimes got lost interrupts... */
 	ret = wait_for_completion_timeout(&mpc8xxx_spi->done, 2 * HZ);
@@ -253,230 +291,76 @@ static int fsl_espi_bufs(struct spi_device *spi, struct spi_transfer *t)
 			mpc8xxx_spi->count);
 
 	/* disable rx ints */
-	mpc8xxx_spi_write_reg(&reg_base->mask, 0);
+	fsl_espi_write_reg(mpc8xxx_spi, ESPI_SPIM, 0);
 
-	return mpc8xxx_spi->count;
+	return mpc8xxx_spi->count > 0 ? -EMSGSIZE : 0;
 }
 
-static inline void fsl_espi_addr2cmd(unsigned int addr, u8 *cmd)
-{
-	if (cmd) {
-		cmd[1] = (u8)(addr >> 16);
-		cmd[2] = (u8)(addr >> 8);
-		cmd[3] = (u8)(addr >> 0);
-	}
-}
-
-static inline unsigned int fsl_espi_cmd2addr(u8 *cmd)
-{
-	if (cmd)
-		return cmd[1] << 16 | cmd[2] << 8 | cmd[3] << 0;
-
-	return 0;
-}
-
-static void fsl_espi_do_trans(struct spi_message *m,
-				struct fsl_espi_transfer *tr)
+static int fsl_espi_trans(struct spi_message *m, struct spi_transfer *trans)
 {
+	struct mpc8xxx_spi *mspi = spi_master_get_devdata(m->spi->master);
 	struct spi_device *spi = m->spi;
-	struct mpc8xxx_spi *mspi = spi_master_get_devdata(spi->master);
-	struct fsl_espi_transfer *espi_trans = tr;
-	struct spi_message message;
-	struct spi_transfer *t, *first, trans;
-	int status = 0;
-
-	spi_message_init(&message);
-	memset(&trans, 0, sizeof(trans));
-
-	first = list_first_entry(&m->transfers, struct spi_transfer,
-			transfer_list);
-	list_for_each_entry(t, &m->transfers, transfer_list) {
-		if ((first->bits_per_word != t->bits_per_word) ||
-			(first->speed_hz != t->speed_hz)) {
-			espi_trans->status = -EINVAL;
-			dev_err(mspi->dev,
-				"bits_per_word/speed_hz should be same for the same SPI transfer\n");
-			return;
-		}
-
-		trans.speed_hz = t->speed_hz;
-		trans.bits_per_word = t->bits_per_word;
-		trans.delay_usecs = max(first->delay_usecs, t->delay_usecs);
-	}
-
-	trans.len = espi_trans->len;
-	trans.tx_buf = espi_trans->tx_buf;
-	trans.rx_buf = espi_trans->rx_buf;
-	spi_message_add_tail(&trans, &message);
-
-	list_for_each_entry(t, &message.transfers, transfer_list) {
-		if (t->bits_per_word || t->speed_hz) {
-			status = -EINVAL;
-
-			status = fsl_espi_setup_transfer(spi, t);
-			if (status < 0)
-				break;
-		}
+	int ret;
 
-		if (t->len)
-			status = fsl_espi_bufs(spi, t);
+	fsl_espi_copy_to_buf(m, mspi);
+	fsl_espi_setup_transfer(spi, trans);
 
-		if (status) {
-			status = -EMSGSIZE;
-			break;
-		}
+	ret = fsl_espi_bufs(spi, trans);
 
-		if (t->delay_usecs)
-			udelay(t->delay_usecs);
-	}
+	if (trans->delay_usecs)
+		udelay(trans->delay_usecs);
 
-	espi_trans->status = status;
 	fsl_espi_setup_transfer(spi, NULL);
-}
-
-static void fsl_espi_cmd_trans(struct spi_message *m,
-				struct fsl_espi_transfer *trans, u8 *rx_buff)
-{
-	struct spi_transfer *t;
-	u8 *local_buf;
-	int i = 0;
-	struct fsl_espi_transfer *espi_trans = trans;
-
-	local_buf = kzalloc(SPCOM_TRANLEN_MAX, GFP_KERNEL);
-	if (!local_buf) {
-		espi_trans->status = -ENOMEM;
-		return;
-	}
-
-	list_for_each_entry(t, &m->transfers, transfer_list) {
-		if (t->tx_buf) {
-			memcpy(local_buf + i, t->tx_buf, t->len);
-			i += t->len;
-		}
-	}
-
-	espi_trans->tx_buf = local_buf;
-	espi_trans->rx_buf = local_buf;
-	fsl_espi_do_trans(m, espi_trans);
 
-	espi_trans->actual_length = espi_trans->len;
-	kfree(local_buf);
-}
-
-static void fsl_espi_rw_trans(struct spi_message *m,
-				struct fsl_espi_transfer *trans, u8 *rx_buff)
-{
-	struct fsl_espi_transfer *espi_trans = trans;
-	unsigned int total_len = espi_trans->len;
-	struct spi_transfer *t;
-	u8 *local_buf;
-	u8 *rx_buf = rx_buff;
-	unsigned int trans_len;
-	unsigned int addr;
-	unsigned int tx_only;
-	unsigned int rx_pos = 0;
-	unsigned int pos;
-	int i, loop;
-
-	local_buf = kzalloc(SPCOM_TRANLEN_MAX, GFP_KERNEL);
-	if (!local_buf) {
-		espi_trans->status = -ENOMEM;
-		return;
-	}
-
-	for (pos = 0, loop = 0; pos < total_len; pos += trans_len, loop++) {
-		trans_len = total_len - pos;
-
-		i = 0;
-		tx_only = 0;
-		list_for_each_entry(t, &m->transfers, transfer_list) {
-			if (t->tx_buf) {
-				memcpy(local_buf + i, t->tx_buf, t->len);
-				i += t->len;
-				if (!t->rx_buf)
-					tx_only += t->len;
-			}
-		}
-
-		/* Add additional TX bytes to compensate SPCOM_TRANLEN_MAX */
-		if (loop > 0)
-			trans_len += tx_only;
-
-		if (trans_len > SPCOM_TRANLEN_MAX)
-			trans_len = SPCOM_TRANLEN_MAX;
-
-		/* Update device offset */
-		if (pos > 0) {
-			addr = fsl_espi_cmd2addr(local_buf);
-			addr += rx_pos;
-			fsl_espi_addr2cmd(addr, local_buf);
-		}
-
-		espi_trans->len = trans_len;
-		espi_trans->tx_buf = local_buf;
-		espi_trans->rx_buf = local_buf;
-		fsl_espi_do_trans(m, espi_trans);
-
-		/* If there is at least one RX byte then copy it to rx_buf */
-		if (tx_only < SPCOM_TRANLEN_MAX)
-			memcpy(rx_buf + rx_pos, espi_trans->rx_buf + tx_only,
-					trans_len - tx_only);
-
-		rx_pos += trans_len - tx_only;
-
-		if (loop > 0)
-			espi_trans->actual_length += espi_trans->len - tx_only;
-		else
-			espi_trans->actual_length += espi_trans->len;
-	}
+	if (!ret)
+		fsl_espi_copy_from_buf(m, mspi);
 
-	kfree(local_buf);
+	return ret;
 }
 
 static int fsl_espi_do_one_msg(struct spi_master *master,
 			       struct spi_message *m)
 {
-	struct spi_transfer *t;
-	u8 *rx_buf = NULL;
-	unsigned int n_tx = 0;
-	unsigned int n_rx = 0;
-	unsigned int xfer_len = 0;
-	struct fsl_espi_transfer espi_trans;
+	struct mpc8xxx_spi *mspi = spi_master_get_devdata(m->spi->master);
+	unsigned int delay_usecs = 0;
+	struct spi_transfer *t, trans = {};
+	int ret;
+
+	ret = fsl_espi_check_message(m);
+	if (ret)
+		goto out;
 
 	list_for_each_entry(t, &m->transfers, transfer_list) {
-		if (t->tx_buf)
-			n_tx += t->len;
-		if (t->rx_buf) {
-			n_rx += t->len;
-			rx_buf = t->rx_buf;
-		}
-		if ((t->tx_buf) || (t->rx_buf))
-			xfer_len += t->len;
+		if (t->delay_usecs > delay_usecs)
+			delay_usecs = t->delay_usecs;
 	}
 
-	espi_trans.n_tx = n_tx;
-	espi_trans.n_rx = n_rx;
-	espi_trans.len = xfer_len;
-	espi_trans.actual_length = 0;
-	espi_trans.status = 0;
+	t = list_first_entry(&m->transfers, struct spi_transfer,
+			     transfer_list);
+
+	trans.len = m->frame_length;
+	trans.speed_hz = t->speed_hz;
+	trans.bits_per_word = t->bits_per_word;
+	trans.delay_usecs = delay_usecs;
+	trans.tx_buf = mspi->local_buf;
+	trans.rx_buf = mspi->local_buf;
+
+	if (trans.len)
+		ret = fsl_espi_trans(m, &trans);
 
-	if (!rx_buf)
-		fsl_espi_cmd_trans(m, &espi_trans, NULL);
-	else
-		fsl_espi_rw_trans(m, &espi_trans, rx_buf);
+	m->actual_length = ret ? 0 : trans.len;
+out:
+	if (m->status == -EINPROGRESS)
+		m->status = ret;
 
-	m->actual_length = espi_trans.actual_length;
-	m->status = espi_trans.status;
 	spi_finalize_current_message(master);
-	return 0;
+
+	return ret;
 }
 
 static int fsl_espi_setup(struct spi_device *spi)
 {
 	struct mpc8xxx_spi *mpc8xxx_spi;
-	struct fsl_espi_reg *reg_base;
-	int retval;
-	u32 hw_mode;
 	u32 loop_mode;
 	struct spi_mpc8xxx_cs *cs = spi_get_ctldata(spi);
 
@@ -491,13 +375,11 @@ static int fsl_espi_setup(struct spi_device *spi)
 	}
 
 	mpc8xxx_spi = spi_master_get_devdata(spi->master);
-	reg_base = mpc8xxx_spi->reg_base;
 
 	pm_runtime_get_sync(mpc8xxx_spi->dev);
 
-	hw_mode = cs->hw_mode; /* Save original settings */
-	cs->hw_mode = mpc8xxx_spi_read_reg(
-			&reg_base->csmode[spi->chip_select]);
+	cs->hw_mode = fsl_espi_read_reg(mpc8xxx_spi,
+					   ESPI_SPMODEx(spi->chip_select));
 	/* mask out bits we are going to set */
 	cs->hw_mode &= ~(CSMODE_CP_BEGIN_EDGECLK | CSMODE_CI_INACTIVEHIGH
 			 | CSMODE_REV);
@@ -510,21 +392,17 @@ static int fsl_espi_setup(struct spi_device *spi)
 		cs->hw_mode |= CSMODE_REV;
 
 	/* Handle the loop mode */
-	loop_mode = mpc8xxx_spi_read_reg(&reg_base->mode);
+	loop_mode = fsl_espi_read_reg(mpc8xxx_spi, ESPI_SPMODE);
 	loop_mode &= ~SPMODE_LOOP;
 	if (spi->mode & SPI_LOOP)
 		loop_mode |= SPMODE_LOOP;
-	mpc8xxx_spi_write_reg(&reg_base->mode, loop_mode);
+	fsl_espi_write_reg(mpc8xxx_spi, ESPI_SPMODE, loop_mode);
 
-	retval = fsl_espi_setup_transfer(spi, NULL);
+	fsl_espi_setup_transfer(spi, NULL);
 
 	pm_runtime_mark_last_busy(mpc8xxx_spi->dev);
 	pm_runtime_put_autosuspend(mpc8xxx_spi->dev);
 
-	if (retval < 0) {
-		cs->hw_mode = hw_mode; /* Restore settings */
-		return retval;
-	}
 	return 0;
 }
 
@@ -536,12 +414,10 @@ static void fsl_espi_cleanup(struct spi_device *spi)
 	spi_set_ctldata(spi, NULL);
 }
 
-void fsl_espi_cpu_irq(struct mpc8xxx_spi *mspi, u32 events)
+static void fsl_espi_cpu_irq(struct mpc8xxx_spi *mspi, u32 events)
 {
-	struct fsl_espi_reg *reg_base = mspi->reg_base;
-
 	/* We need handle RX first */
-	if (events & SPIE_NE) {
+	if (events & SPIE_RNE) {
 		u32 rx_data, tmp;
 		u8 rx_data_8;
 		int rx_nr_bytes = 4;
@@ -551,7 +427,7 @@ void fsl_espi_cpu_irq(struct mpc8xxx_spi *mspi, u32 events)
 		if (SPIE_RXCNT(events) < min(4, mspi->len)) {
 			ret = spin_event_timeout(
 				!(SPIE_RXCNT(events =
-				mpc8xxx_spi_read_reg(&reg_base->event)) <
+				fsl_espi_read_reg(mspi, ESPI_SPIE)) <
 						min(4, mspi->len)),
 						10000, 0); /* 10 msec */
 			if (!ret)
@@ -560,10 +436,10 @@ void fsl_espi_cpu_irq(struct mpc8xxx_spi *mspi, u32 events)
 		}
 
 		if (mspi->len >= 4) {
-			rx_data = mpc8xxx_spi_read_reg(&reg_base->receive);
+			rx_data = fsl_espi_read_reg(mspi, ESPI_SPIRF);
 		} else if (mspi->len <= 0) {
 			dev_err(mspi->dev,
-				"unexpected RX(SPIE_NE) interrupt occurred,\n"
+				"unexpected RX(SPIE_RNE) interrupt occurred,\n"
 				"(local rxlen %d bytes, reg rxlen %d bytes)\n",
 				min(4, mspi->len), SPIE_RXCNT(events));
 			rx_nr_bytes = 0;
@@ -572,7 +448,8 @@ void fsl_espi_cpu_irq(struct mpc8xxx_spi *mspi, u32 events)
 			tmp = mspi->len;
 			rx_data = 0;
 			while (tmp--) {
-				rx_data_8 = in_8((u8 *)&reg_base->receive);
+				rx_data_8 = fsl_espi_read_reg8(mspi,
+							       ESPI_SPIRF);
 				rx_data |= (rx_data_8 << (tmp * 8));
 			}
 
@@ -585,30 +462,24 @@ void fsl_espi_cpu_irq(struct mpc8xxx_spi *mspi, u32 events)
 			mspi->get_rx(rx_data, mspi);
 	}
 
-	if (!(events & SPIE_NF)) {
+	if (!(events & SPIE_TNF)) {
 		int ret;
 
 		/* spin until TX is done */
-		ret = spin_event_timeout(((events = mpc8xxx_spi_read_reg(
-				&reg_base->event)) & SPIE_NF), 1000, 0);
+		ret = spin_event_timeout(((events = fsl_espi_read_reg(
+				mspi, ESPI_SPIE)) & SPIE_TNF), 1000, 0);
 		if (!ret) {
-			dev_err(mspi->dev, "tired waiting for SPIE_NF\n");
-
-			/* Clear the SPIE bits */
-			mpc8xxx_spi_write_reg(&reg_base->event, events);
+			dev_err(mspi->dev, "tired waiting for SPIE_TNF\n");
 			complete(&mspi->done);
 			return;
 		}
 	}
 
-	/* Clear the events */
-	mpc8xxx_spi_write_reg(&reg_base->event, events);
-
 	mspi->count -= 1;
 	if (mspi->count) {
 		u32 word = mspi->get_tx(mspi);
 
-		mpc8xxx_spi_write_reg(&reg_base->transmit, word);
+		fsl_espi_write_reg(mspi, ESPI_SPITF, word);
 	} else {
 		complete(&mspi->done);
 	}
@@ -617,20 +488,21 @@ void fsl_espi_cpu_irq(struct mpc8xxx_spi *mspi, u32 events)
 static irqreturn_t fsl_espi_irq(s32 irq, void *context_data)
 {
 	struct mpc8xxx_spi *mspi = context_data;
-	struct fsl_espi_reg *reg_base = mspi->reg_base;
-	irqreturn_t ret = IRQ_NONE;
 	u32 events;
 
 	/* Get interrupt events(tx/rx) */
-	events = mpc8xxx_spi_read_reg(&reg_base->event);
-	if (events)
-		ret = IRQ_HANDLED;
+	events = fsl_espi_read_reg(mspi, ESPI_SPIE);
+	if (!events)
+		return IRQ_NONE;
 
 	dev_vdbg(mspi->dev, "%s: events %x\n", __func__, events);
 
 	fsl_espi_cpu_irq(mspi, events);
 
-	return ret;
+	/* Clear the events */
+	fsl_espi_write_reg(mspi, ESPI_SPIE, events);
+
+	return IRQ_HANDLED;
 }
 
 #ifdef CONFIG_PM
@@ -638,12 +510,11 @@ static int fsl_espi_runtime_suspend(struct device *dev)
 {
 	struct spi_master *master = dev_get_drvdata(dev);
 	struct mpc8xxx_spi *mpc8xxx_spi = spi_master_get_devdata(master);
-	struct fsl_espi_reg *reg_base = mpc8xxx_spi->reg_base;
 	u32 regval;
 
-	regval = mpc8xxx_spi_read_reg(&reg_base->mode);
+	regval = fsl_espi_read_reg(mpc8xxx_spi, ESPI_SPMODE);
 	regval &= ~SPMODE_ENABLE;
-	mpc8xxx_spi_write_reg(&reg_base->mode, regval);
+	fsl_espi_write_reg(mpc8xxx_spi, ESPI_SPMODE, regval);
 
 	return 0;
 }
@@ -652,39 +523,35 @@ static int fsl_espi_runtime_resume(struct device *dev)
 {
 	struct spi_master *master = dev_get_drvdata(dev);
 	struct mpc8xxx_spi *mpc8xxx_spi = spi_master_get_devdata(master);
-	struct fsl_espi_reg *reg_base = mpc8xxx_spi->reg_base;
 	u32 regval;
 
-	regval = mpc8xxx_spi_read_reg(&reg_base->mode);
+	regval = fsl_espi_read_reg(mpc8xxx_spi, ESPI_SPMODE);
 	regval |= SPMODE_ENABLE;
-	mpc8xxx_spi_write_reg(&reg_base->mode, regval);
+	fsl_espi_write_reg(mpc8xxx_spi, ESPI_SPMODE, regval);
 
 	return 0;
 }
 #endif
 
-static size_t fsl_espi_max_transfer_size(struct spi_device *spi)
+static size_t fsl_espi_max_message_size(struct spi_device *spi)
 {
 	return SPCOM_TRANLEN_MAX;
 }
 
-static struct spi_master * fsl_espi_probe(struct device *dev,
-		struct resource *mem, unsigned int irq)
+static int fsl_espi_probe(struct device *dev, struct resource *mem,
+			  unsigned int irq)
 {
 	struct fsl_spi_platform_data *pdata = dev_get_platdata(dev);
 	struct spi_master *master;
 	struct mpc8xxx_spi *mpc8xxx_spi;
-	struct fsl_espi_reg *reg_base;
 	struct device_node *nc;
 	const __be32 *prop;
 	u32 regval, csmode;
-	int i, len, ret = 0;
+	int i, len, ret;
 
 	master = spi_alloc_master(dev, sizeof(struct mpc8xxx_spi));
-	if (!master) {
-		ret = -ENOMEM;
-		goto err;
-	}
+	if (!master)
+		return -ENOMEM;
 
 	dev_set_drvdata(dev, master);
 
@@ -695,18 +562,23 @@ static struct spi_master * fsl_espi_probe(struct device *dev,
 	master->cleanup = fsl_espi_cleanup;
 	master->transfer_one_message = fsl_espi_do_one_msg;
 	master->auto_runtime_pm = true;
-	master->max_transfer_size = fsl_espi_max_transfer_size;
+	master->max_message_size = fsl_espi_max_message_size;
 
 	mpc8xxx_spi = spi_master_get_devdata(master);
 
+	mpc8xxx_spi->local_buf =
+		devm_kmalloc(dev, SPCOM_TRANLEN_MAX, GFP_KERNEL);
+	if (!mpc8xxx_spi->local_buf) {
+		ret = -ENOMEM;
+		goto err_probe;
+	}
+
 	mpc8xxx_spi->reg_base = devm_ioremap_resource(dev, mem);
 	if (IS_ERR(mpc8xxx_spi->reg_base)) {
 		ret = PTR_ERR(mpc8xxx_spi->reg_base);
 		goto err_probe;
 	}
 
-	reg_base = mpc8xxx_spi->reg_base;
-
 	/* Register for SPI Interrupt */
 	ret = devm_request_irq(dev, mpc8xxx_spi->irq, fsl_espi_irq,
 			  0, "fsl_espi", mpc8xxx_spi);
@@ -719,10 +591,10 @@ static struct spi_master * fsl_espi_probe(struct device *dev,
 	}
 
 	/* SPI controller initializations */
-	mpc8xxx_spi_write_reg(&reg_base->mode, 0);
-	mpc8xxx_spi_write_reg(&reg_base->mask, 0);
-	mpc8xxx_spi_write_reg(&reg_base->command, 0);
-	mpc8xxx_spi_write_reg(&reg_base->event, 0xffffffff);
+	fsl_espi_write_reg(mpc8xxx_spi, ESPI_SPMODE, 0);
+	fsl_espi_write_reg(mpc8xxx_spi, ESPI_SPIM, 0);
+	fsl_espi_write_reg(mpc8xxx_spi, ESPI_SPCOM, 0);
+	fsl_espi_write_reg(mpc8xxx_spi, ESPI_SPIE, 0xffffffff);
 
 	/* Init eSPI CS mode register */
 	for_each_available_child_of_node(master->dev.of_node, nc) {
@@ -747,7 +619,7 @@ static struct spi_master * fsl_espi_probe(struct device *dev,
 			csmode &= ~(CSMODE_AFT(0xf));
 			csmode |= CSMODE_AFT(be32_to_cpup(prop));
 		}
-		mpc8xxx_spi_write_reg(&reg_base->csmode[i], csmode);
+		fsl_espi_write_reg(mpc8xxx_spi, ESPI_SPMODEx(i), csmode);
 
 		dev_info(dev, "cs=%d, init_csmode=0x%x\n", i, csmode);
 	}
@@ -755,7 +627,7 @@ static struct spi_master * fsl_espi_probe(struct device *dev,
 	/* Enable SPI interface */
 	regval = pdata->initial_spmode | SPMODE_INIT_VAL | SPMODE_ENABLE;
 
-	mpc8xxx_spi_write_reg(&reg_base->mode, regval);
+	fsl_espi_write_reg(mpc8xxx_spi, ESPI_SPMODE, regval);
 
 	pm_runtime_set_autosuspend_delay(dev, AUTOSUSPEND_TIMEOUT);
 	pm_runtime_use_autosuspend(dev);
@@ -767,12 +639,13 @@ static struct spi_master * fsl_espi_probe(struct device *dev,
 	if (ret < 0)
 		goto err_pm;
 
-	dev_info(dev, "at 0x%p (irq = %d)\n", reg_base, mpc8xxx_spi->irq);
+	dev_info(dev, "at 0x%p (irq = %d)\n", mpc8xxx_spi->reg_base,
+		 mpc8xxx_spi->irq);
 
 	pm_runtime_mark_last_busy(dev);
 	pm_runtime_put_autosuspend(dev);
 
-	return master;
+	return 0;
 
 err_pm:
 	pm_runtime_put_noidle(dev);
@@ -780,8 +653,7 @@ err_pm:
 	pm_runtime_set_suspended(dev);
 err_probe:
 	spi_master_put(master);
-err:
-	return ERR_PTR(ret);
+	return ret;
 }
 
 static int of_fsl_espi_get_chipselects(struct device *dev)
@@ -807,10 +679,9 @@ static int of_fsl_espi_probe(struct platform_device *ofdev)
 {
 	struct device *dev = &ofdev->dev;
 	struct device_node *np = ofdev->dev.of_node;
-	struct spi_master *master;
 	struct resource mem;
 	unsigned int irq;
-	int ret = -ENOMEM;
+	int ret;
 
 	ret = of_mpc8xxx_spi_probe(ofdev);
 	if (ret)
@@ -818,28 +689,17 @@ static int of_fsl_espi_probe(struct platform_device *ofdev)
 
 	ret = of_fsl_espi_get_chipselects(dev);
 	if (ret)
-		goto err;
+		return ret;
 
 	ret = of_address_to_resource(np, 0, &mem);
 	if (ret)
-		goto err;
+		return ret;
 
 	irq = irq_of_parse_and_map(np, 0);
-	if (!irq) {
-		ret = -EINVAL;
-		goto err;
-	}
-
-	master = fsl_espi_probe(dev, &mem, irq);
-	if (IS_ERR(master)) {
-		ret = PTR_ERR(master);
-		goto err;
-	}
-
-	return 0;
+	if (!irq)
+		return -EINVAL;
 
-err:
-	return ret;
+	return fsl_espi_probe(dev, &mem, irq);
 }
 
 static int of_fsl_espi_remove(struct platform_device *dev)
@@ -873,27 +733,26 @@ static int of_fsl_espi_resume(struct device *dev)
 	struct fsl_spi_platform_data *pdata = dev_get_platdata(dev);
 	struct spi_master *master = dev_get_drvdata(dev);
 	struct mpc8xxx_spi *mpc8xxx_spi;
-	struct fsl_espi_reg *reg_base;
 	u32 regval;
 	int i, ret;
 
 	mpc8xxx_spi = spi_master_get_devdata(master);
-	reg_base = mpc8xxx_spi->reg_base;
 
 	/* SPI controller initializations */
-	mpc8xxx_spi_write_reg(&reg_base->mode, 0);
-	mpc8xxx_spi_write_reg(&reg_base->mask, 0);
-	mpc8xxx_spi_write_reg(&reg_base->command, 0);
-	mpc8xxx_spi_write_reg(&reg_base->event, 0xffffffff);
+	fsl_espi_write_reg(mpc8xxx_spi, ESPI_SPMODE, 0);
+	fsl_espi_write_reg(mpc8xxx_spi, ESPI_SPIM, 0);
+	fsl_espi_write_reg(mpc8xxx_spi, ESPI_SPCOM, 0);
+	fsl_espi_write_reg(mpc8xxx_spi, ESPI_SPIE, 0xffffffff);
 
 	/* Init eSPI CS mode register */
 	for (i = 0; i < pdata->max_chipselect; i++)
-		mpc8xxx_spi_write_reg(&reg_base->csmode[i], CSMODE_INIT_VAL);
+		fsl_espi_write_reg(mpc8xxx_spi, ESPI_SPMODEx(i),
+				      CSMODE_INIT_VAL);
 
 	/* Enable SPI interface */
 	regval = pdata->initial_spmode | SPMODE_INIT_VAL | SPMODE_ENABLE;
 
-	mpc8xxx_spi_write_reg(&reg_base->mode, regval);
+	fsl_espi_write_reg(mpc8xxx_spi, ESPI_SPMODE, regval);
 
 	ret = pm_runtime_force_resume(dev);
 	if (ret < 0)
diff --git a/drivers/spi/spi-fsl-lib.h b/drivers/spi/spi-fsl-lib.h
index 84f5dcb7a897..2925c8089fd9 100644
--- a/drivers/spi/spi-fsl-lib.h
+++ b/drivers/spi/spi-fsl-lib.h
@@ -23,13 +23,14 @@
 /* SPI/eSPI Controller driver's private data. */
 struct mpc8xxx_spi {
 	struct device *dev;
-	void *reg_base;
+	void __iomem *reg_base;
 
 	/* rx & tx bufs from the spi_transfer */
 	const void *tx;
 	void *rx;
 #if IS_ENABLED(CONFIG_SPI_FSL_ESPI)
 	int len;
+	u8 *local_buf;
 #endif
 
 	int subblock;
diff --git a/drivers/spi/spi-imx.c b/drivers/spi/spi-imx.c
index f63cb30f9010..deb782f6556c 100644
--- a/drivers/spi/spi-imx.c
+++ b/drivers/spi/spi-imx.c
@@ -186,17 +186,19 @@ static unsigned int spi_imx_clkdiv_1(unsigned int fin,
 
 /* MX1, MX31, MX35, MX51 CSPI */
 static unsigned int spi_imx_clkdiv_2(unsigned int fin,
-		unsigned int fspi)
+		unsigned int fspi, unsigned int *fres)
 {
 	int i, div = 4;
 
 	for (i = 0; i < 7; i++) {
 		if (fspi * div >= fin)
-			return i;
+			goto out;
 		div <<= 1;
 	}
 
-	return 7;
+out:
+	*fres = fin / div;
+	return i;
 }
 
 static int spi_imx_bytes_per_word(const int bpw)
@@ -453,6 +455,9 @@ static void mx51_ecspi_reset(struct spi_imx_data *spi_imx)
 #define MX31_CSPISTATUS		0x14
 #define MX31_STATUS_RR		(1 << 3)
 
+#define MX31_CSPI_TESTREG	0x1C
+#define MX31_TEST_LBC		(1 << 14)
+
 /* These functions also work for the i.MX35, but be aware that
  * the i.MX35 has a slightly different register layout for bits
  * we do not use here.
@@ -482,9 +487,11 @@ static int mx31_config(struct spi_device *spi, struct spi_imx_config *config)
 {
 	struct spi_imx_data *spi_imx = spi_master_get_devdata(spi->master);
 	unsigned int reg = MX31_CSPICTRL_ENABLE | MX31_CSPICTRL_MASTER;
+	unsigned int clk;
 
-	reg |= spi_imx_clkdiv_2(spi_imx->spi_clk, config->speed_hz) <<
+	reg |= spi_imx_clkdiv_2(spi_imx->spi_clk, config->speed_hz, &clk) <<
 		MX31_CSPICTRL_DR_SHIFT;
+	spi_imx->spi_bus_clk = clk;
 
 	if (is_imx35_cspi(spi_imx)) {
 		reg |= (config->bpw - 1) << MX35_CSPICTRL_BL_SHIFT;
@@ -506,6 +513,13 @@ static int mx31_config(struct spi_device *spi, struct spi_imx_config *config)
 
 	writel(reg, spi_imx->base + MXC_CSPICTRL);
 
+	reg = readl(spi_imx->base + MX31_CSPI_TESTREG);
+	if (spi->mode & SPI_LOOP)
+		reg |= MX31_TEST_LBC;
+	else
+		reg &= ~MX31_TEST_LBC;
+	writel(reg, spi_imx->base + MX31_CSPI_TESTREG);
+
 	return 0;
 }
 
@@ -625,9 +639,12 @@ static int mx1_config(struct spi_device *spi, struct spi_imx_config *config)
 {
 	struct spi_imx_data *spi_imx = spi_master_get_devdata(spi->master);
 	unsigned int reg = MX1_CSPICTRL_ENABLE | MX1_CSPICTRL_MASTER;
+	unsigned int clk;
 
-	reg |= spi_imx_clkdiv_2(spi_imx->spi_clk, config->speed_hz) <<
+	reg |= spi_imx_clkdiv_2(spi_imx->spi_clk, config->speed_hz, &clk) <<
 		MX1_CSPICTRL_DR_SHIFT;
+	spi_imx->spi_bus_clk = clk;
+
 	reg |= config->bpw - 1;
 
 	if (spi->mode & SPI_CPHA)
@@ -1179,7 +1196,7 @@ static int spi_imx_probe(struct platform_device *pdev)
 	spi_imx->bitbang.master->prepare_message = spi_imx_prepare_message;
 	spi_imx->bitbang.master->unprepare_message = spi_imx_unprepare_message;
 	spi_imx->bitbang.master->mode_bits = SPI_CPOL | SPI_CPHA | SPI_CS_HIGH;
-	if (is_imx51_ecspi(spi_imx))
+	if (is_imx35_cspi(spi_imx) || is_imx51_ecspi(spi_imx))
 		spi_imx->bitbang.master->mode_bits |= SPI_LOOP;
 
 	init_completion(&spi_imx->xfer_done);
@@ -1251,6 +1268,12 @@ static int spi_imx_probe(struct platform_device *pdev)
 		goto out_clk_put;
 	}
 
+	if (!master->cs_gpios) {
+		dev_err(&pdev->dev, "No CS GPIOs available\n");
+		ret = -EINVAL;
+		goto out_clk_put;
+	}
+
 	for (i = 0; i < master->num_chipselect; i++) {
 		if (!gpio_is_valid(master->cs_gpios[i]))
 			continue;
diff --git a/drivers/spi/spi-iproc-qspi.c b/drivers/spi/spi-iproc-qspi.c
new file mode 100644
index 000000000000..be6ccb204a66
--- /dev/null
+++ b/drivers/spi/spi-iproc-qspi.c
@@ -0,0 +1,163 @@
+/*
+ * Copyright 2016 Broadcom Limited
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ */
+
+#include <linux/device.h>
+#include <linux/io.h>
+#include <linux/ioport.h>
+#include <linux/module.h>
+#include <linux/of.h>
+#include <linux/of_address.h>
+#include <linux/platform_device.h>
+#include <linux/slab.h>
+
+#include "spi-bcm-qspi.h"
+
+#define INTR_BASE_BIT_SHIFT			0x02
+#define INTR_COUNT				0x07
+
+struct bcm_iproc_intc {
+	struct bcm_qspi_soc_intc soc_intc;
+	struct platform_device *pdev;
+	void __iomem *int_reg;
+	void __iomem *int_status_reg;
+	spinlock_t soclock;
+	bool big_endian;
+};
+
+static u32 bcm_iproc_qspi_get_l2_int_status(struct bcm_qspi_soc_intc *soc_intc)
+{
+	struct bcm_iproc_intc *priv =
+			container_of(soc_intc, struct bcm_iproc_intc, soc_intc);
+	void __iomem *mmio = priv->int_status_reg;
+	int i;
+	u32 val = 0, sts = 0;
+
+	for (i = 0; i < INTR_COUNT; i++) {
+		if (bcm_qspi_readl(priv->big_endian, mmio + (i * 4)))
+			val |= 1UL << i;
+	}
+
+	if (val & INTR_MSPI_DONE_MASK)
+		sts |= MSPI_DONE;
+
+	if (val & BSPI_LR_INTERRUPTS_ALL)
+		sts |= BSPI_DONE;
+
+	if (val & BSPI_LR_INTERRUPTS_ERROR)
+		sts |= BSPI_ERR;
+
+	return sts;
+}
+
+static void bcm_iproc_qspi_int_ack(struct bcm_qspi_soc_intc *soc_intc, int type)
+{
+	struct bcm_iproc_intc *priv =
+			container_of(soc_intc, struct bcm_iproc_intc, soc_intc);
+	void __iomem *mmio = priv->int_status_reg;
+	u32 mask = get_qspi_mask(type);
+	int i;
+
+	for (i = 0; i < INTR_COUNT; i++) {
+		if (mask & (1UL << i))
+			bcm_qspi_writel(priv->big_endian, 1, mmio + (i * 4));
+	}
+}
+
+static void bcm_iproc_qspi_int_set(struct bcm_qspi_soc_intc *soc_intc, int type,
+				   bool en)
+{
+	struct bcm_iproc_intc *priv =
+			container_of(soc_intc, struct bcm_iproc_intc, soc_intc);
+	void __iomem *mmio = priv->int_reg;
+	u32 mask = get_qspi_mask(type);
+	u32 val;
+	unsigned long flags;
+
+	spin_lock_irqsave(&priv->soclock, flags);
+
+	val = bcm_qspi_readl(priv->big_endian, mmio);
+
+	if (en)
+		val = val | (mask << INTR_BASE_BIT_SHIFT);
+	else
+		val = val & ~(mask << INTR_BASE_BIT_SHIFT);
+
+	bcm_qspi_writel(priv->big_endian, val, mmio);
+
+	spin_unlock_irqrestore(&priv->soclock, flags);
+}
+
+static int bcm_iproc_probe(struct platform_device *pdev)
+{
+	struct device *dev = &pdev->dev;
+	struct bcm_iproc_intc *priv;
+	struct bcm_qspi_soc_intc *soc_intc;
+	struct resource *res;
+
+	priv = devm_kzalloc(dev, sizeof(*priv), GFP_KERNEL);
+	if (!priv)
+		return -ENOMEM;
+	soc_intc = &priv->soc_intc;
+	priv->pdev = pdev;
+
+	spin_lock_init(&priv->soclock);
+
+	res = platform_get_resource_byname(pdev, IORESOURCE_MEM, "intr_regs");
+	priv->int_reg = devm_ioremap_resource(dev, res);
+	if (IS_ERR(priv->int_reg))
+		return PTR_ERR(priv->int_reg);
+
+	res = platform_get_resource_byname(pdev, IORESOURCE_MEM,
+					   "intr_status_reg");
+	priv->int_status_reg = devm_ioremap_resource(dev, res);
+	if (IS_ERR(priv->int_status_reg))
+		return PTR_ERR(priv->int_status_reg);
+
+	priv->big_endian = of_device_is_big_endian(dev->of_node);
+
+	bcm_iproc_qspi_int_ack(soc_intc, MSPI_BSPI_DONE);
+	bcm_iproc_qspi_int_set(soc_intc, MSPI_BSPI_DONE, false);
+
+	soc_intc->bcm_qspi_int_ack = bcm_iproc_qspi_int_ack;
+	soc_intc->bcm_qspi_int_set = bcm_iproc_qspi_int_set;
+	soc_intc->bcm_qspi_get_int_status = bcm_iproc_qspi_get_l2_int_status;
+
+	return bcm_qspi_probe(pdev, soc_intc);
+}
+
+static int bcm_iproc_remove(struct platform_device *pdev)
+{
+	return bcm_qspi_remove(pdev);
+}
+
+static const struct of_device_id bcm_iproc_of_match[] = {
+	{ .compatible = "brcm,spi-nsp-qspi" },
+	{ .compatible = "brcm,spi-ns2-qspi" },
+	{},
+};
+MODULE_DEVICE_TABLE(of, bcm_iproc_of_match);
+
+static struct platform_driver bcm_iproc_driver = {
+	.probe			= bcm_iproc_probe,
+	.remove			= bcm_iproc_remove,
+	.driver = {
+		.name		= "bcm_iproc",
+		.pm		= &bcm_qspi_pm_ops,
+		.of_match_table = bcm_iproc_of_match,
+	}
+};
+module_platform_driver(bcm_iproc_driver);
+
+MODULE_LICENSE("GPL v2");
+MODULE_AUTHOR("Kamal Dasu");
+MODULE_DESCRIPTION("SPI flash driver for Broadcom iProc SoCs");
diff --git a/drivers/spi/spi-jcore.c b/drivers/spi/spi-jcore.c
new file mode 100644
index 000000000000..f8117b80fa22
--- /dev/null
+++ b/drivers/spi/spi-jcore.c
@@ -0,0 +1,231 @@
+/*
+ * J-Core SPI controller driver
+ *
+ * Copyright (C) 2012-2016 Smart Energy Instruments, Inc.
+ *
+ * Current version by Rich Felker
+ * Based loosely on initial version by Oleksandr G Zhadan
+ *
+ */
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/errno.h>
+#include <linux/module.h>
+#include <linux/platform_device.h>
+#include <linux/spi/spi.h>
+#include <linux/clk.h>
+#include <linux/err.h>
+#include <linux/io.h>
+#include <linux/of.h>
+#include <linux/delay.h>
+
+#define DRV_NAME	"jcore_spi"
+
+#define CTRL_REG	0x0
+#define DATA_REG	0x4
+
+#define JCORE_SPI_CTRL_XMIT		0x02
+#define JCORE_SPI_STAT_BUSY		0x02
+#define JCORE_SPI_CTRL_LOOP		0x08
+#define JCORE_SPI_CTRL_CS_BITS		0x15
+
+#define JCORE_SPI_WAIT_RDY_MAX_LOOP	2000000
+
+struct jcore_spi {
+	struct spi_master *master;
+	void __iomem *base;
+	unsigned int cs_reg;
+	unsigned int speed_reg;
+	unsigned int speed_hz;
+	unsigned int clock_freq;
+};
+
+static int jcore_spi_wait(void __iomem *ctrl_reg)
+{
+	unsigned timeout = JCORE_SPI_WAIT_RDY_MAX_LOOP;
+
+	do {
+		if (!(readl(ctrl_reg) & JCORE_SPI_STAT_BUSY))
+			return 0;
+		cpu_relax();
+	} while (--timeout);
+
+	return -EBUSY;
+}
+
+static void jcore_spi_program(struct jcore_spi *hw)
+{
+	void __iomem *ctrl_reg = hw->base + CTRL_REG;
+
+	if (jcore_spi_wait(ctrl_reg))
+		dev_err(hw->master->dev.parent,
+			"timeout waiting to program ctrl reg.\n");
+
+	writel(hw->cs_reg | hw->speed_reg, ctrl_reg);
+}
+
+static void jcore_spi_chipsel(struct spi_device *spi, bool value)
+{
+	struct jcore_spi *hw = spi_master_get_devdata(spi->master);
+	u32 csbit = 1U << (2 * spi->chip_select);
+
+	dev_dbg(hw->master->dev.parent, "chipselect %d\n", spi->chip_select);
+
+	if (value)
+		hw->cs_reg |= csbit;
+	else
+		hw->cs_reg &= ~csbit;
+
+	jcore_spi_program(hw);
+}
+
+static void jcore_spi_baudrate(struct jcore_spi *hw, int speed)
+{
+	if (speed == hw->speed_hz) return;
+	hw->speed_hz = speed;
+	if (speed >= hw->clock_freq / 2)
+		hw->speed_reg = 0;
+	else
+		hw->speed_reg = ((hw->clock_freq / 2 / speed) - 1) << 27;
+	jcore_spi_program(hw);
+	dev_dbg(hw->master->dev.parent, "speed=%d reg=0x%x\n",
+		speed, hw->speed_reg);
+}
+
+static int jcore_spi_txrx(struct spi_master *master, struct spi_device *spi,
+			  struct spi_transfer *t)
+{
+	struct jcore_spi *hw = spi_master_get_devdata(master);
+
+	void __iomem *ctrl_reg = hw->base + CTRL_REG;
+	void __iomem *data_reg = hw->base + DATA_REG;
+	u32 xmit;
+
+	/* data buffers */
+	const unsigned char *tx;
+	unsigned char *rx;
+	unsigned int len;
+	unsigned int count;
+
+	jcore_spi_baudrate(hw, t->speed_hz);
+
+	xmit = hw->cs_reg | hw->speed_reg | JCORE_SPI_CTRL_XMIT;
+	tx = t->tx_buf;
+	rx = t->rx_buf;
+	len = t->len;
+
+	for (count = 0; count < len; count++) {
+		if (jcore_spi_wait(ctrl_reg))
+			break;
+
+		writel(tx ? *tx++ : 0, data_reg);
+		writel(xmit, ctrl_reg);
+
+		if (jcore_spi_wait(ctrl_reg))
+			break;
+
+		if (rx)
+			*rx++ = readl(data_reg);
+	}
+
+	spi_finalize_current_transfer(master);
+
+	if (count < len)
+		return -EREMOTEIO;
+
+	return 0;
+}
+
+static int jcore_spi_probe(struct platform_device *pdev)
+{
+	struct device_node *node = pdev->dev.of_node;
+	struct jcore_spi *hw;
+	struct spi_master *master;
+	struct resource *res;
+	u32 clock_freq;
+	struct clk *clk;
+	int err = -ENODEV;
+
+	master = spi_alloc_master(&pdev->dev, sizeof(struct jcore_spi));
+	if (!master)
+		return err;
+
+	/* Setup the master state. */
+	master->num_chipselect = 3;
+	master->mode_bits = SPI_CPOL | SPI_CPHA | SPI_CS_HIGH;
+	master->transfer_one = jcore_spi_txrx;
+	master->set_cs = jcore_spi_chipsel;
+	master->dev.of_node = node;
+	master->bus_num = pdev->id;
+
+	hw = spi_master_get_devdata(master);
+	hw->master = master;
+	platform_set_drvdata(pdev, hw);
+
+	/* Find and map our resources */
+	res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
+	if (!res)
+		goto exit_busy;
+	if (!devm_request_mem_region(&pdev->dev, res->start,
+				     resource_size(res), pdev->name))
+		goto exit_busy;
+	hw->base = devm_ioremap_nocache(&pdev->dev, res->start,
+					resource_size(res));
+	if (!hw->base)
+		goto exit_busy;
+
+	/*
+	 * The SPI clock rate controlled via a configurable clock divider
+	 * which is applied to the reference clock. A 50 MHz reference is
+	 * most suitable for obtaining standard SPI clock rates, but some
+	 * designs may have a different reference clock, and the DT must
+	 * make the driver aware so that it can properly program the
+	 * requested rate. If the clock is omitted, 50 MHz is assumed.
+	 */
+	clock_freq = 50000000;
+	clk = devm_clk_get(&pdev->dev, "ref_clk");
+	if (!IS_ERR_OR_NULL(clk)) {
+		if (clk_enable(clk) == 0)
+			clock_freq = clk_get_rate(clk);
+		else
+			dev_warn(&pdev->dev, "could not enable ref_clk\n");
+	}
+	hw->clock_freq = clock_freq;
+
+	/* Initialize all CS bits to high. */
+	hw->cs_reg = JCORE_SPI_CTRL_CS_BITS;
+	jcore_spi_baudrate(hw, 400000);
+
+	/* Register our spi controller */
+	err = devm_spi_register_master(&pdev->dev, master);
+	if (err)
+		goto exit;
+
+	return 0;
+
+exit_busy:
+	err = -EBUSY;
+exit:
+	spi_master_put(master);
+	return err;
+}
+
+static const struct of_device_id jcore_spi_of_match[] = {
+	{ .compatible = "jcore,spi2" },
+	{},
+};
+
+static struct platform_driver jcore_spi_driver = {
+	.probe = jcore_spi_probe,
+	.driver = {
+		.name = DRV_NAME,
+		.of_match_table = jcore_spi_of_match,
+	},
+};
+
+module_platform_driver(jcore_spi_driver);
+
+MODULE_DESCRIPTION("J-Core SPI driver");
+MODULE_AUTHOR("Rich Felker <dalias@libc.org>");
+MODULE_LICENSE("GPL");
+MODULE_ALIAS("platform:" DRV_NAME);
diff --git a/drivers/spi/spi-loopback-test.c b/drivers/spi/spi-loopback-test.c
index faa1b133b0f1..50e620f4e8fe 100644
--- a/drivers/spi/spi-loopback-test.c
+++ b/drivers/spi/spi-loopback-test.c
@@ -405,7 +405,7 @@ struct rx_ranges {
 	u8 *end;
 };
 
-int rx_ranges_cmp(void *priv, struct list_head *a, struct list_head *b)
+static int rx_ranges_cmp(void *priv, struct list_head *a, struct list_head *b)
 {
 	struct rx_ranges *rx_a = list_entry(a, struct rx_ranges, list);
 	struct rx_ranges *rx_b = list_entry(b, struct rx_ranges, list);
diff --git a/drivers/spi/spi-meson-spifc.c b/drivers/spi/spi-meson-spifc.c
index 2465259f6241..616566e793c6 100644
--- a/drivers/spi/spi-meson-spifc.c
+++ b/drivers/spi/spi-meson-spifc.c
@@ -442,6 +442,7 @@ static const struct dev_pm_ops meson_spifc_pm_ops = {
 
 static const struct of_device_id meson_spifc_dt_match[] = {
 	{ .compatible = "amlogic,meson6-spifc", },
+	{ .compatible = "amlogic,meson-gxbb-spifc", },
 	{ },
 };
 MODULE_DEVICE_TABLE(of, meson_spifc_dt_match);
diff --git a/drivers/spi/spi-pic32-sqi.c b/drivers/spi/spi-pic32-sqi.c
index c41abddab318..bd1c6b53283f 100644
--- a/drivers/spi/spi-pic32-sqi.c
+++ b/drivers/spi/spi-pic32-sqi.c
@@ -253,15 +253,13 @@ static struct ring_desc *ring_desc_get(struct pic32_sqi *sqi)
 		return NULL;
 
 	rdesc = list_first_entry(&sqi->bd_list_free, struct ring_desc, list);
-	list_del(&rdesc->list);
-	list_add_tail(&rdesc->list, &sqi->bd_list_used);
+	list_move_tail(&rdesc->list, &sqi->bd_list_used);
 	return rdesc;
 }
 
 static void ring_desc_put(struct pic32_sqi *sqi, struct ring_desc *rdesc)
 {
-	list_del(&rdesc->list);
-	list_add(&rdesc->list, &sqi->bd_list_free);
+	list_move(&rdesc->list, &sqi->bd_list_free);
 }
 
 static int pic32_sqi_one_transfer(struct pic32_sqi *sqi,
diff --git a/drivers/spi/spi-pxa2xx-dma.c b/drivers/spi/spi-pxa2xx-dma.c
index db3ae1dd829e..04f3eecf5cf3 100644
--- a/drivers/spi/spi-pxa2xx-dma.c
+++ b/drivers/spi/spi-pxa2xx-dma.c
@@ -23,7 +23,7 @@
 static void pxa2xx_spi_dma_transfer_complete(struct driver_data *drv_data,
 					     bool error)
 {
-	struct spi_message *msg = drv_data->cur_msg;
+	struct spi_message *msg = drv_data->master->cur_msg;
 
 	/*
 	 * It is possible that one CPU is handling ROR interrupt and other
@@ -76,7 +76,8 @@ static struct dma_async_tx_descriptor *
 pxa2xx_spi_dma_prepare_one(struct driver_data *drv_data,
 			   enum dma_transfer_direction dir)
 {
-	struct chip_data *chip = drv_data->cur_chip;
+	struct chip_data *chip =
+		spi_get_ctldata(drv_data->master->cur_msg->spi);
 	struct spi_transfer *xfer = drv_data->cur_transfer;
 	enum dma_slave_buswidth width;
 	struct dma_slave_config cfg;
@@ -146,7 +147,7 @@ irqreturn_t pxa2xx_spi_dma_transfer(struct driver_data *drv_data)
 int pxa2xx_spi_dma_prepare(struct driver_data *drv_data, u32 dma_burst)
 {
 	struct dma_async_tx_descriptor *tx_desc, *rx_desc;
-	int err = 0;
+	int err;
 
 	tx_desc = pxa2xx_spi_dma_prepare_one(drv_data, DMA_MEM_TO_DEV);
 	if (!tx_desc) {
diff --git a/drivers/spi/spi-pxa2xx.c b/drivers/spi/spi-pxa2xx.c
index 87150a1049bd..dd7b5b47291d 100644
--- a/drivers/spi/spi-pxa2xx.c
+++ b/drivers/spi/spi-pxa2xx.c
@@ -28,6 +28,7 @@
 #include <linux/spi/spi.h>
 #include <linux/delay.h>
 #include <linux/gpio.h>
+#include <linux/gpio/consumer.h>
 #include <linux/slab.h>
 #include <linux/clk.h>
 #include <linux/pm_runtime.h>
@@ -62,6 +63,13 @@ MODULE_ALIAS("platform:pxa2xx-spi");
 				| QUARK_X1000_SSCR1_TFT		\
 				| SSCR1_SPH | SSCR1_SPO | SSCR1_LBM)
 
+#define CE4100_SSCR1_CHANGE_MASK (SSCR1_TTELP | SSCR1_TTE | SSCR1_SCFR \
+				| SSCR1_ECRA | SSCR1_ECRB | SSCR1_SCLKDIR \
+				| SSCR1_SFRMDIR | SSCR1_RWOT | SSCR1_TRAIL \
+				| SSCR1_IFS | SSCR1_STRF | SSCR1_EFWR \
+				| CE4100_SSCR1_RFT | CE4100_SSCR1_TFT | SSCR1_MWDS \
+				| SSCR1_SPH | SSCR1_SPO | SSCR1_LBM)
+
 #define LPSS_GENERAL_REG_RXTO_HOLDOFF_DISABLE	BIT(24)
 #define LPSS_CS_CONTROL_SW_MODE			BIT(0)
 #define LPSS_CS_CONTROL_CS_HIGH			BIT(1)
@@ -175,6 +183,8 @@ static u32 pxa2xx_spi_get_ssrc1_change_mask(const struct driver_data *drv_data)
 	switch (drv_data->ssp_type) {
 	case QUARK_X1000_SSP:
 		return QUARK_X1000_SSCR1_CHANGE_MASK;
+	case CE4100_SSP:
+		return CE4100_SSCR1_CHANGE_MASK;
 	default:
 		return SSCR1_CHANGE_MASK;
 	}
@@ -186,6 +196,8 @@ pxa2xx_spi_get_rx_default_thre(const struct driver_data *drv_data)
 	switch (drv_data->ssp_type) {
 	case QUARK_X1000_SSP:
 		return RX_THRESH_QUARK_X1000_DFLT;
+	case CE4100_SSP:
+		return RX_THRESH_CE4100_DFLT;
 	default:
 		return RX_THRESH_DFLT;
 	}
@@ -199,6 +211,9 @@ static bool pxa2xx_spi_txfifo_full(const struct driver_data *drv_data)
 	case QUARK_X1000_SSP:
 		mask = QUARK_X1000_SSSR_TFL_MASK;
 		break;
+	case CE4100_SSP:
+		mask = CE4100_SSSR_TFL_MASK;
+		break;
 	default:
 		mask = SSSR_TFL_MASK;
 		break;
@@ -216,6 +231,9 @@ static void pxa2xx_spi_clear_rx_thre(const struct driver_data *drv_data,
 	case QUARK_X1000_SSP:
 		mask = QUARK_X1000_SSCR1_RFT;
 		break;
+	case CE4100_SSP:
+		mask = CE4100_SSCR1_RFT;
+		break;
 	default:
 		mask = SSCR1_RFT;
 		break;
@@ -230,6 +248,9 @@ static void pxa2xx_spi_set_rx_thre(const struct driver_data *drv_data,
 	case QUARK_X1000_SSP:
 		*sccr1_reg |= QUARK_X1000_SSCR1_RxTresh(threshold);
 		break;
+	case CE4100_SSP:
+		*sccr1_reg |= CE4100_SSCR1_RxTresh(threshold);
+		break;
 	default:
 		*sccr1_reg |= SSCR1_RxTresh(threshold);
 		break;
@@ -316,7 +337,7 @@ static void lpss_ssp_select_cs(struct driver_data *drv_data,
 
 	value = __lpss_ssp_read_priv(drv_data, config->reg_cs_ctrl);
 
-	cs = drv_data->cur_msg->spi->chip_select;
+	cs = drv_data->master->cur_msg->spi->chip_select;
 	cs <<= config->cs_sel_shift;
 	if (cs != (value & config->cs_sel_mask)) {
 		/*
@@ -355,10 +376,11 @@ static void lpss_ssp_cs_control(struct driver_data *drv_data, bool enable)
 
 static void cs_assert(struct driver_data *drv_data)
 {
-	struct chip_data *chip = drv_data->cur_chip;
+	struct chip_data *chip =
+		spi_get_ctldata(drv_data->master->cur_msg->spi);
 
 	if (drv_data->ssp_type == CE4100_SSP) {
-		pxa2xx_spi_write(drv_data, SSSR, drv_data->cur_chip->frm);
+		pxa2xx_spi_write(drv_data, SSSR, chip->frm);
 		return;
 	}
 
@@ -378,7 +400,8 @@ static void cs_assert(struct driver_data *drv_data)
 
 static void cs_deassert(struct driver_data *drv_data)
 {
-	struct chip_data *chip = drv_data->cur_chip;
+	struct chip_data *chip =
+		spi_get_ctldata(drv_data->master->cur_msg->spi);
 
 	if (drv_data->ssp_type == CE4100_SSP)
 		return;
@@ -508,7 +531,7 @@ static int u32_reader(struct driver_data *drv_data)
 
 void *pxa2xx_spi_next_transfer(struct driver_data *drv_data)
 {
-	struct spi_message *msg = drv_data->cur_msg;
+	struct spi_message *msg = drv_data->master->cur_msg;
 	struct spi_transfer *trans = drv_data->cur_transfer;
 
 	/* Move to next transfer */
@@ -529,8 +552,7 @@ static void giveback(struct driver_data *drv_data)
 	struct spi_message *msg;
 	unsigned long timeout;
 
-	msg = drv_data->cur_msg;
-	drv_data->cur_msg = NULL;
+	msg = drv_data->master->cur_msg;
 	drv_data->cur_transfer = NULL;
 
 	last_transfer = list_last_entry(&msg->transfers, struct spi_transfer,
@@ -575,13 +597,13 @@ static void giveback(struct driver_data *drv_data)
 			cs_deassert(drv_data);
 	}
 
-	drv_data->cur_chip = NULL;
 	spi_finalize_current_message(drv_data->master);
 }
 
 static void reset_sccr1(struct driver_data *drv_data)
 {
-	struct chip_data *chip = drv_data->cur_chip;
+	struct chip_data *chip =
+		spi_get_ctldata(drv_data->master->cur_msg->spi);
 	u32 sccr1_reg;
 
 	sccr1_reg = pxa2xx_spi_read(drv_data, SSCR1) & ~drv_data->int_cr1;
@@ -589,6 +611,9 @@ static void reset_sccr1(struct driver_data *drv_data)
 	case QUARK_X1000_SSP:
 		sccr1_reg &= ~QUARK_X1000_SSCR1_RFT;
 		break;
+	case CE4100_SSP:
+		sccr1_reg &= ~CE4100_SSCR1_RFT;
+		break;
 	default:
 		sccr1_reg &= ~SSCR1_RFT;
 		break;
@@ -610,7 +635,7 @@ static void int_error_stop(struct driver_data *drv_data, const char* msg)
 
 	dev_err(&drv_data->pdev->dev, "%s\n", msg);
 
-	drv_data->cur_msg->state = ERROR_STATE;
+	drv_data->master->cur_msg->state = ERROR_STATE;
 	tasklet_schedule(&drv_data->pump_transfers);
 }
 
@@ -623,7 +648,7 @@ static void int_transfer_complete(struct driver_data *drv_data)
 		pxa2xx_spi_write(drv_data, SSTO, 0);
 
 	/* Update total byte transferred return count actual bytes read */
-	drv_data->cur_msg->actual_length += drv_data->len -
+	drv_data->master->cur_msg->actual_length += drv_data->len -
 				(drv_data->rx_end - drv_data->rx);
 
 	/* Transfer delays and chip select release are
@@ -631,7 +656,7 @@ static void int_transfer_complete(struct driver_data *drv_data)
 	 */
 
 	/* Move to next transfer */
-	drv_data->cur_msg->state = pxa2xx_spi_next_transfer(drv_data);
+	drv_data->master->cur_msg->state = pxa2xx_spi_next_transfer(drv_data);
 
 	/* Schedule transfer tasklet */
 	tasklet_schedule(&drv_data->pump_transfers);
@@ -746,7 +771,7 @@ static irqreturn_t ssp_int(int irq, void *dev_id)
 	if (!(status & mask))
 		return IRQ_NONE;
 
-	if (!drv_data->cur_msg) {
+	if (!drv_data->master->cur_msg) {
 
 		pxa2xx_spi_write(drv_data, SSCR0,
 				 pxa2xx_spi_read(drv_data, SSCR0)
@@ -905,7 +930,8 @@ static unsigned int ssp_get_clk_div(struct driver_data *drv_data, int rate)
 static unsigned int pxa2xx_ssp_get_clk_div(struct driver_data *drv_data,
 					   int rate)
 {
-	struct chip_data *chip = drv_data->cur_chip;
+	struct chip_data *chip =
+		spi_get_ctldata(drv_data->master->cur_msg->spi);
 	unsigned int clk_div;
 
 	switch (drv_data->ssp_type) {
@@ -934,25 +960,23 @@ static void pump_transfers(unsigned long data)
 {
 	struct driver_data *drv_data = (struct driver_data *)data;
 	struct spi_master *master = drv_data->master;
-	struct spi_message *message = NULL;
-	struct spi_transfer *transfer = NULL;
-	struct spi_transfer *previous = NULL;
-	struct chip_data *chip = NULL;
-	u32 clk_div = 0;
-	u8 bits = 0;
-	u32 speed = 0;
+	struct spi_message *message = master->cur_msg;
+	struct chip_data *chip = spi_get_ctldata(message->spi);
+	u32 dma_thresh = chip->dma_threshold;
+	u32 dma_burst = chip->dma_burst_size;
+	u32 change_mask = pxa2xx_spi_get_ssrc1_change_mask(drv_data);
+	struct spi_transfer *transfer;
+	struct spi_transfer *previous;
+	u32 clk_div;
+	u8 bits;
+	u32 speed;
 	u32 cr0;
 	u32 cr1;
-	u32 dma_thresh = drv_data->cur_chip->dma_threshold;
-	u32 dma_burst = drv_data->cur_chip->dma_burst_size;
-	u32 change_mask = pxa2xx_spi_get_ssrc1_change_mask(drv_data);
 	int err;
 	int dma_mapped;
 
 	/* Get current state information */
-	message = drv_data->cur_msg;
 	transfer = drv_data->cur_transfer;
-	chip = drv_data->cur_chip;
 
 	/* Handle for abort */
 	if (message->state == ERROR_STATE) {
@@ -1146,17 +1170,12 @@ static int pxa2xx_spi_transfer_one_message(struct spi_master *master,
 {
 	struct driver_data *drv_data = spi_master_get_devdata(master);
 
-	drv_data->cur_msg = msg;
 	/* Initial message state*/
-	drv_data->cur_msg->state = START_STATE;
-	drv_data->cur_transfer = list_entry(drv_data->cur_msg->transfers.next,
+	msg->state = START_STATE;
+	drv_data->cur_transfer = list_entry(msg->transfers.next,
 						struct spi_transfer,
 						transfer_list);
 
-	/* prepare to setup the SSP, in pump_transfers, using the per
-	 * chip configuration */
-	drv_data->cur_chip = spi_get_ctldata(drv_data->cur_msg->spi);
-
 	/* Mark as busy and launch transfers */
 	tasklet_schedule(&drv_data->pump_transfers);
 	return 0;
@@ -1176,9 +1195,26 @@ static int pxa2xx_spi_unprepare_transfer(struct spi_master *master)
 static int setup_cs(struct spi_device *spi, struct chip_data *chip,
 		    struct pxa2xx_spi_chip *chip_info)
 {
+	struct driver_data *drv_data = spi_master_get_devdata(spi->master);
 	int err = 0;
 
-	if (chip == NULL || chip_info == NULL)
+	if (chip == NULL)
+		return 0;
+
+	if (drv_data->cs_gpiods) {
+		struct gpio_desc *gpiod;
+
+		gpiod = drv_data->cs_gpiods[spi->chip_select];
+		if (gpiod) {
+			chip->gpio_cs = desc_to_gpio(gpiod);
+			chip->gpio_cs_inverted = spi->mode & SPI_CS_HIGH;
+			gpiod_set_value(gpiod, chip->gpio_cs_inverted);
+		}
+
+		return 0;
+	}
+
+	if (chip_info == NULL)
 		return 0;
 
 	/* NOTE: setup() can be called multiple times, possibly with
@@ -1213,7 +1249,7 @@ static int setup_cs(struct spi_device *spi, struct chip_data *chip,
 
 static int setup(struct spi_device *spi)
 {
-	struct pxa2xx_spi_chip *chip_info = NULL;
+	struct pxa2xx_spi_chip *chip_info;
 	struct chip_data *chip;
 	const struct lpss_config *config;
 	struct driver_data *drv_data = spi_master_get_devdata(spi->master);
@@ -1225,6 +1261,11 @@ static int setup(struct spi_device *spi)
 		tx_hi_thres = 0;
 		rx_thres = RX_THRESH_QUARK_X1000_DFLT;
 		break;
+	case CE4100_SSP:
+		tx_thres = TX_THRESH_CE4100_DFLT;
+		tx_hi_thres = 0;
+		rx_thres = RX_THRESH_CE4100_DFLT;
+		break;
 	case LPSS_LPT_SSP:
 	case LPSS_BYT_SSP:
 	case LPSS_BSW_SSP:
@@ -1309,6 +1350,10 @@ static int setup(struct spi_device *spi)
 				   | (QUARK_X1000_SSCR1_TxTresh(tx_thres)
 				   & QUARK_X1000_SSCR1_TFT);
 		break;
+	case CE4100_SSP:
+		chip->threshold = (CE4100_SSCR1_RxTresh(rx_thres) & CE4100_SSCR1_RFT) |
+			(CE4100_SSCR1_TxTresh(tx_thres) & CE4100_SSCR1_TFT);
+		break;
 	default:
 		chip->threshold = (SSCR1_RxTresh(rx_thres) & SSCR1_RFT) |
 			(SSCR1_TxTresh(tx_thres) & SSCR1_TFT);
@@ -1352,7 +1397,8 @@ static void cleanup(struct spi_device *spi)
 	if (!chip)
 		return;
 
-	if (drv_data->ssp_type != CE4100_SSP && gpio_is_valid(chip->gpio_cs))
+	if (drv_data->ssp_type != CE4100_SSP && !drv_data->cs_gpiods &&
+	    gpio_is_valid(chip->gpio_cs))
 		gpio_free(chip->gpio_cs);
 
 	kfree(chip);
@@ -1530,7 +1576,7 @@ static int pxa2xx_spi_probe(struct platform_device *pdev)
 	struct driver_data *drv_data;
 	struct ssp_device *ssp;
 	const struct lpss_config *config;
-	int status;
+	int status, count;
 	u32 tmp;
 
 	platform_info = dev_get_platdata(dev);
@@ -1630,15 +1676,20 @@ static int pxa2xx_spi_probe(struct platform_device *pdev)
 	pxa2xx_spi_write(drv_data, SSCR0, 0);
 	switch (drv_data->ssp_type) {
 	case QUARK_X1000_SSP:
-		tmp = QUARK_X1000_SSCR1_RxTresh(RX_THRESH_QUARK_X1000_DFLT)
-		      | QUARK_X1000_SSCR1_TxTresh(TX_THRESH_QUARK_X1000_DFLT);
+		tmp = QUARK_X1000_SSCR1_RxTresh(RX_THRESH_QUARK_X1000_DFLT) |
+		      QUARK_X1000_SSCR1_TxTresh(TX_THRESH_QUARK_X1000_DFLT);
 		pxa2xx_spi_write(drv_data, SSCR1, tmp);
 
 		/* using the Motorola SPI protocol and use 8 bit frame */
-		pxa2xx_spi_write(drv_data, SSCR0,
-				 QUARK_X1000_SSCR0_Motorola
-				 | QUARK_X1000_SSCR0_DataSize(8));
+		tmp = QUARK_X1000_SSCR0_Motorola | QUARK_X1000_SSCR0_DataSize(8);
+		pxa2xx_spi_write(drv_data, SSCR0, tmp);
 		break;
+	case CE4100_SSP:
+		tmp = CE4100_SSCR1_RxTresh(RX_THRESH_CE4100_DFLT) |
+		      CE4100_SSCR1_TxTresh(TX_THRESH_CE4100_DFLT);
+		pxa2xx_spi_write(drv_data, SSCR1, tmp);
+		tmp = SSCR0_SCR(2) | SSCR0_Motorola | SSCR0_DataSize(8);
+		pxa2xx_spi_write(drv_data, SSCR0, tmp);
 	default:
 		tmp = SSCR1_RxTresh(RX_THRESH_DFLT) |
 		      SSCR1_TxTresh(TX_THRESH_DFLT);
@@ -1669,6 +1720,39 @@ static int pxa2xx_spi_probe(struct platform_device *pdev)
 	}
 	master->num_chipselect = platform_info->num_chipselect;
 
+	count = gpiod_count(&pdev->dev, "cs");
+	if (count > 0) {
+		int i;
+
+		master->num_chipselect = max_t(int, count,
+			master->num_chipselect);
+
+		drv_data->cs_gpiods = devm_kcalloc(&pdev->dev,
+			master->num_chipselect, sizeof(struct gpio_desc *),
+			GFP_KERNEL);
+		if (!drv_data->cs_gpiods) {
+			status = -ENOMEM;
+			goto out_error_clock_enabled;
+		}
+
+		for (i = 0; i < master->num_chipselect; i++) {
+			struct gpio_desc *gpiod;
+
+			gpiod = devm_gpiod_get_index(dev, "cs", i,
+						     GPIOD_OUT_HIGH);
+			if (IS_ERR(gpiod)) {
+				/* Means use native chip select */
+				if (PTR_ERR(gpiod) == -ENOENT)
+					continue;
+
+				status = (int)PTR_ERR(gpiod);
+				goto out_error_clock_enabled;
+			} else {
+				drv_data->cs_gpiods[i] = gpiod;
+			}
+		}
+	}
+
 	tasklet_init(&drv_data->pump_transfers, pump_transfers,
 		     (unsigned long)drv_data);
 
@@ -1742,7 +1826,7 @@ static int pxa2xx_spi_suspend(struct device *dev)
 {
 	struct driver_data *drv_data = dev_get_drvdata(dev);
 	struct ssp_device *ssp = drv_data->ssp;
-	int status = 0;
+	int status;
 
 	status = spi_master_suspend(drv_data->master);
 	if (status != 0)
@@ -1759,7 +1843,7 @@ static int pxa2xx_spi_resume(struct device *dev)
 {
 	struct driver_data *drv_data = dev_get_drvdata(dev);
 	struct ssp_device *ssp = drv_data->ssp;
-	int status = 0;
+	int status;
 
 	/* Enable the SSP clock */
 	if (!pm_runtime_suspended(dev))
diff --git a/drivers/spi/spi-pxa2xx.h b/drivers/spi/spi-pxa2xx.h
index d217ad55cc12..ce31b8199bb3 100644
--- a/drivers/spi/spi-pxa2xx.h
+++ b/drivers/spi/spi-pxa2xx.h
@@ -53,9 +53,7 @@ struct driver_data {
 	atomic_t dma_running;
 
 	/* Current message transfer state info */
-	struct spi_message *cur_msg;
 	struct spi_transfer *cur_transfer;
-	struct chip_data *cur_chip;
 	size_t len;
 	void *tx;
 	void *tx_end;
@@ -68,6 +66,9 @@ struct driver_data {
 	void (*cs_control)(u32 command);
 
 	void __iomem *lpss_base;
+
+	/* GPIOs for chip selects */
+	struct gpio_desc **cs_gpiods;
 };
 
 struct chip_data {
diff --git a/drivers/spi/spi-qup.c b/drivers/spi/spi-qup.c
index 7f1555621f8e..1bfa889b8427 100644
--- a/drivers/spi/spi-qup.c
+++ b/drivers/spi/spi-qup.c
@@ -982,8 +982,10 @@ static int spi_qup_suspend(struct device *device)
 	if (ret)
 		return ret;
 
-	clk_disable_unprepare(controller->cclk);
-	clk_disable_unprepare(controller->iclk);
+	if (!pm_runtime_suspended(device)) {
+		clk_disable_unprepare(controller->cclk);
+		clk_disable_unprepare(controller->iclk);
+	}
 	return 0;
 }
 
diff --git a/drivers/spi/spi-rspi.c b/drivers/spi/spi-rspi.c
index 818843336932..a816f07e168e 100644
--- a/drivers/spi/spi-rspi.c
+++ b/drivers/spi/spi-rspi.c
@@ -295,14 +295,24 @@ static int rspi_set_config_register(struct rspi_data *rspi, int access_size)
 static int rspi_rz_set_config_register(struct rspi_data *rspi, int access_size)
 {
 	int spbr;
+	int div = 0;
+	unsigned long clksrc;
 
 	/* Sets output mode, MOSI signal, and (optionally) loopback */
 	rspi_write8(rspi, rspi->sppcr, RSPI_SPPCR);
 
+	clksrc = clk_get_rate(rspi->clk);
+	while (div < 3) {
+		if (rspi->max_speed_hz >= clksrc/4) /* 4=(CLK/2)/2 */
+			break;
+		div++;
+		clksrc /= 2;
+	}
+
 	/* Sets transfer bit rate */
-	spbr = DIV_ROUND_UP(clk_get_rate(rspi->clk),
-			    2 * rspi->max_speed_hz) - 1;
+	spbr = DIV_ROUND_UP(clksrc, 2 * rspi->max_speed_hz) - 1;
 	rspi_write8(rspi, clamp(spbr, 0, 255), RSPI_SPBR);
+	rspi->spcmd |= div << 2;
 
 	/* Disable dummy transmission, set byte access */
 	rspi_write8(rspi, SPDCR_SPLBYTE, RSPI_SPDCR);
diff --git a/drivers/spi/spi-sc18is602.c b/drivers/spi/spi-sc18is602.c
index 36af4d48a700..f63714ffb62f 100644
--- a/drivers/spi/spi-sc18is602.c
+++ b/drivers/spi/spi-sc18is602.c
@@ -23,6 +23,7 @@
 #include <linux/pm_runtime.h>
 #include <linux/of.h>
 #include <linux/platform_data/sc18is602.h>
+#include <linux/gpio/consumer.h>
 
 enum chips { sc18is602, sc18is602b, sc18is603 };
 
@@ -50,6 +51,8 @@ struct sc18is602 {
 	u8			buffer[SC18IS602_BUFSIZ + 1];
 	int			tlen;	/* Data queued for tx in buffer */
 	int			rindex;	/* Receive data index in buffer */
+
+	struct gpio_desc	*reset;
 };
 
 static int sc18is602_wait_ready(struct sc18is602 *hw, int len)
@@ -257,6 +260,12 @@ static int sc18is602_probe(struct i2c_client *client,
 	hw = spi_master_get_devdata(master);
 	i2c_set_clientdata(client, hw);
 
+	/* assert reset and then release */
+	hw->reset = devm_gpiod_get_optional(dev, "reset", GPIOD_OUT_HIGH);
+	if (IS_ERR(hw->reset))
+		return PTR_ERR(hw->reset);
+	gpiod_set_value_cansleep(hw->reset, 0);
+
 	hw->master = master;
 	hw->client = client;
 	hw->dev = dev;
diff --git a/drivers/spi/spi-st-ssc4.c b/drivers/spi/spi-st-ssc4.c
index a56eca0e95a6..e54b59638458 100644
--- a/drivers/spi/spi-st-ssc4.c
+++ b/drivers/spi/spi-st-ssc4.c
@@ -175,10 +175,7 @@ static int spi_st_transfer_one(struct spi_master *master,
 
 static void spi_st_cleanup(struct spi_device *spi)
 {
-	int cs = spi->cs_gpio;
-
-	if (gpio_is_valid(cs))
-		devm_gpio_free(&spi->dev, cs);
+	gpio_free(spi->cs_gpio);
 }
 
 /* the spi->mode bits understood by this driver: */
@@ -201,14 +198,15 @@ static int spi_st_setup(struct spi_device *spi)
 		return -EINVAL;
 	}
 
-	if (devm_gpio_request(&spi->dev, cs, dev_name(&spi->dev))) {
+	ret = gpio_request(cs, dev_name(&spi->dev));
+	if (ret) {
 		dev_err(&spi->dev, "could not request gpio:%d\n", cs);
-		return -EINVAL;
+		return ret;
 	}
 
 	ret = gpio_direction_output(cs, spi->mode & SPI_CS_HIGH);
 	if (ret)
-		return ret;
+		goto out_free_gpio;
 
 	spi_st_clk = clk_get_rate(spi_st->clk);
 
@@ -217,7 +215,8 @@ static int spi_st_setup(struct spi_device *spi)
 	if (sscbrg < 0x07 || sscbrg > BIT(16)) {
 		dev_err(&spi->dev,
 			"baudrate %d outside valid range %d\n", sscbrg, hz);
-		return -EINVAL;
+		ret = -EINVAL;
+		goto out_free_gpio;
 	}
 
 	spi_st->baud = spi_st_clk / (2 * sscbrg);
@@ -266,6 +265,10 @@ static int spi_st_setup(struct spi_device *spi)
 	 readl_relaxed(spi_st->base + SSC_RBUF);
 
 	 return 0;
+
+out_free_gpio:
+	gpio_free(cs);
+	return ret;
 }
 
 /* Interrupt fired when TX shift register becomes empty */
diff --git a/drivers/spi/spi-ti-qspi.c b/drivers/spi/spi-ti-qspi.c
index ac0b072815a3..caeac66a3977 100644
--- a/drivers/spi/spi-ti-qspi.c
+++ b/drivers/spi/spi-ti-qspi.c
@@ -41,6 +41,8 @@ struct ti_qspi_regs {
 };
 
 struct ti_qspi {
+	struct completion	transfer_complete;
+
 	/* list synchronization */
 	struct mutex            list_lock;
 
@@ -54,6 +56,9 @@ struct ti_qspi {
 
 	struct ti_qspi_regs     ctx_reg;
 
+	dma_addr_t		mmap_phys_base;
+	struct dma_chan		*rx_chan;
+
 	u32 spi_max_frequency;
 	u32 cmd;
 	u32 dc;
@@ -379,6 +384,72 @@ static int qspi_transfer_msg(struct ti_qspi *qspi, struct spi_transfer *t,
 	return 0;
 }
 
+static void ti_qspi_dma_callback(void *param)
+{
+	struct ti_qspi *qspi = param;
+
+	complete(&qspi->transfer_complete);
+}
+
+static int ti_qspi_dma_xfer(struct ti_qspi *qspi, dma_addr_t dma_dst,
+			    dma_addr_t dma_src, size_t len)
+{
+	struct dma_chan *chan = qspi->rx_chan;
+	struct dma_device *dma_dev = chan->device;
+	dma_cookie_t cookie;
+	enum dma_ctrl_flags flags = DMA_CTRL_ACK | DMA_PREP_INTERRUPT;
+	struct dma_async_tx_descriptor *tx;
+	int ret;
+
+	tx = dma_dev->device_prep_dma_memcpy(chan, dma_dst, dma_src,
+					     len, flags);
+	if (!tx) {
+		dev_err(qspi->dev, "device_prep_dma_memcpy error\n");
+		return -EIO;
+	}
+
+	tx->callback = ti_qspi_dma_callback;
+	tx->callback_param = qspi;
+	cookie = tx->tx_submit(tx);
+
+	ret = dma_submit_error(cookie);
+	if (ret) {
+		dev_err(qspi->dev, "dma_submit_error %d\n", cookie);
+		return -EIO;
+	}
+
+	dma_async_issue_pending(chan);
+	ret = wait_for_completion_timeout(&qspi->transfer_complete,
+					  msecs_to_jiffies(len));
+	if (ret <= 0) {
+		dmaengine_terminate_sync(chan);
+		dev_err(qspi->dev, "DMA wait_for_completion_timeout\n");
+		return -ETIMEDOUT;
+	}
+
+	return 0;
+}
+
+static int ti_qspi_dma_xfer_sg(struct ti_qspi *qspi, struct sg_table rx_sg,
+			       loff_t from)
+{
+	struct scatterlist *sg;
+	dma_addr_t dma_src = qspi->mmap_phys_base + from;
+	dma_addr_t dma_dst;
+	int i, len, ret;
+
+	for_each_sg(rx_sg.sgl, sg, rx_sg.nents, i) {
+		dma_dst = sg_dma_address(sg);
+		len = sg_dma_len(sg);
+		ret = ti_qspi_dma_xfer(qspi, dma_dst, dma_src, len);
+		if (ret)
+			return ret;
+		dma_src += len;
+	}
+
+	return 0;
+}
+
 static void ti_qspi_enable_memory_map(struct spi_device *spi)
 {
 	struct ti_qspi  *qspi = spi_master_get_devdata(spi->master);
@@ -426,7 +497,7 @@ static void ti_qspi_setup_mmap_read(struct spi_device *spi,
 		      QSPI_SPI_SETUP_REG(spi->chip_select));
 }
 
-static int ti_qspi_spi_flash_read(struct  spi_device *spi,
+static int ti_qspi_spi_flash_read(struct spi_device *spi,
 				  struct spi_flash_read_message *msg)
 {
 	struct ti_qspi *qspi = spi_master_get_devdata(spi->master);
@@ -437,9 +508,23 @@ static int ti_qspi_spi_flash_read(struct  spi_device *spi,
 	if (!qspi->mmap_enabled)
 		ti_qspi_enable_memory_map(spi);
 	ti_qspi_setup_mmap_read(spi, msg);
-	memcpy_fromio(msg->buf, qspi->mmap_base + msg->from, msg->len);
+
+	if (qspi->rx_chan) {
+		if (msg->cur_msg_mapped) {
+			ret = ti_qspi_dma_xfer_sg(qspi, msg->rx_sg, msg->from);
+			if (ret)
+				goto err_unlock;
+		} else {
+			dev_err(qspi->dev, "Invalid address for DMA\n");
+			ret = -EIO;
+			goto err_unlock;
+		}
+	} else {
+		memcpy_fromio(msg->buf, qspi->mmap_base + msg->from, msg->len);
+	}
 	msg->retlen = msg->len;
 
+err_unlock:
 	mutex_unlock(&qspi->list_lock);
 
 	return ret;
@@ -536,6 +621,7 @@ static int ti_qspi_probe(struct platform_device *pdev)
 	struct device_node *np = pdev->dev.of_node;
 	u32 max_freq;
 	int ret = 0, num_cs, irq;
+	dma_cap_mask_t mask;
 
 	master = spi_alloc_master(&pdev->dev, sizeof(*qspi));
 	if (!master)
@@ -550,6 +636,7 @@ static int ti_qspi_probe(struct platform_device *pdev)
 	master->dev.of_node = pdev->dev.of_node;
 	master->bits_per_word_mask = SPI_BPW_MASK(32) | SPI_BPW_MASK(16) |
 				     SPI_BPW_MASK(8);
+	master->spi_flash_read = ti_qspi_spi_flash_read;
 
 	if (!of_property_read_u32(np, "num-cs", &num_cs))
 		master->num_chipselect = num_cs;
@@ -592,17 +679,6 @@ static int ti_qspi_probe(struct platform_device *pdev)
 		goto free_master;
 	}
 
-	if (res_mmap) {
-		qspi->mmap_base = devm_ioremap_resource(&pdev->dev,
-							res_mmap);
-		master->spi_flash_read = ti_qspi_spi_flash_read;
-		if (IS_ERR(qspi->mmap_base)) {
-			dev_err(&pdev->dev,
-				"falling back to PIO mode\n");
-			master->spi_flash_read = NULL;
-		}
-	}
-	qspi->mmap_enabled = false;
 
 	if (of_property_read_bool(np, "syscon-chipselects")) {
 		qspi->ctrl_base =
@@ -633,11 +709,37 @@ static int ti_qspi_probe(struct platform_device *pdev)
 	if (!of_property_read_u32(np, "spi-max-frequency", &max_freq))
 		qspi->spi_max_frequency = max_freq;
 
-	ret = devm_spi_register_master(&pdev->dev, master);
-	if (ret)
-		goto free_master;
+	dma_cap_zero(mask);
+	dma_cap_set(DMA_MEMCPY, mask);
 
-	return 0;
+	qspi->rx_chan = dma_request_chan_by_mask(&mask);
+	if (!qspi->rx_chan) {
+		dev_err(qspi->dev,
+			"No Rx DMA available, trying mmap mode\n");
+		ret = 0;
+		goto no_dma;
+	}
+	master->dma_rx = qspi->rx_chan;
+	init_completion(&qspi->transfer_complete);
+	if (res_mmap)
+		qspi->mmap_phys_base = (dma_addr_t)res_mmap->start;
+
+no_dma:
+	if (!qspi->rx_chan && res_mmap) {
+		qspi->mmap_base = devm_ioremap_resource(&pdev->dev, res_mmap);
+		if (IS_ERR(qspi->mmap_base)) {
+			dev_info(&pdev->dev,
+				 "mmap failed with error %ld using PIO mode\n",
+				 PTR_ERR(qspi->mmap_base));
+			qspi->mmap_base = NULL;
+			master->spi_flash_read = NULL;
+		}
+	}
+	qspi->mmap_enabled = false;
+
+	ret = devm_spi_register_master(&pdev->dev, master);
+	if (!ret)
+		return 0;
 
 free_master:
 	spi_master_put(master);
@@ -656,6 +758,9 @@ static int ti_qspi_remove(struct platform_device *pdev)
 	pm_runtime_put_sync(&pdev->dev);
 	pm_runtime_disable(&pdev->dev);
 
+	if (qspi->rx_chan)
+		dma_release_channel(qspi->rx_chan);
+
 	return 0;
 }
 
diff --git a/drivers/spi/spi-txx9.c b/drivers/spi/spi-txx9.c
index 7492ea346b43..51759d3fd45f 100644
--- a/drivers/spi/spi-txx9.c
+++ b/drivers/spi/spi-txx9.c
@@ -346,7 +346,7 @@ static int txx9spi_probe(struct platform_device *dev)
 		c->clk = NULL;
 		goto exit;
 	}
-	ret = clk_enable(c->clk);
+	ret = clk_prepare_enable(c->clk);
 	if (ret) {
 		c->clk = NULL;
 		goto exit;
@@ -395,7 +395,7 @@ static int txx9spi_probe(struct platform_device *dev)
 exit_busy:
 	ret = -EBUSY;
 exit:
-	clk_disable(c->clk);
+	clk_disable_unprepare(c->clk);
 	spi_master_put(master);
 	return ret;
 }
@@ -406,7 +406,7 @@ static int txx9spi_remove(struct platform_device *dev)
 	struct txx9spi *c = spi_master_get_devdata(master);
 
 	flush_work(&c->work);
-	clk_disable(c->clk);
+	clk_disable_unprepare(c->clk);
 	return 0;
 }
 
diff --git a/drivers/spi/spi-xlp.c b/drivers/spi/spi-xlp.c
index 8f04feca6ee3..4071a729eb2f 100644
--- a/drivers/spi/spi-xlp.c
+++ b/drivers/spi/spi-xlp.c
@@ -11,6 +11,7 @@
  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  * GNU General Public License for more details.
  */
+#include <linux/acpi.h>
 #include <linux/clk.h>
 #include <linux/kernel.h>
 #include <linux/module.h>
@@ -405,8 +406,9 @@ static int xlp_spi_probe(struct platform_device *pdev)
 	clk = devm_clk_get(&pdev->dev, NULL);
 	if (IS_ERR(clk)) {
 		dev_err(&pdev->dev, "could not get spi clock\n");
-		return -ENODEV;
+		return PTR_ERR(clk);
 	}
+
 	xspi->spi_clk = clk_get_rate(clk);
 
 	master = spi_alloc_master(&pdev->dev, 0);
@@ -437,6 +439,14 @@ static int xlp_spi_probe(struct platform_device *pdev)
 	return 0;
 }
 
+#ifdef CONFIG_ACPI
+static const struct acpi_device_id xlp_spi_acpi_match[] = {
+	{ "BRCM900D", 0 },
+	{ },
+};
+MODULE_DEVICE_TABLE(acpi, xlp_spi_acpi_match);
+#endif
+
 static const struct of_device_id xlp_spi_dt_id[] = {
 	{ .compatible = "netlogic,xlp832-spi" },
 	{ },
@@ -447,6 +457,7 @@ static struct platform_driver xlp_spi_driver = {
 	.driver = {
 		.name	= "xlp-spi",
 		.of_match_table = xlp_spi_dt_id,
+		.acpi_match_table = ACPI_PTR(xlp_spi_acpi_match),
 	},
 };
 module_platform_driver(xlp_spi_driver);
diff --git a/drivers/spi/spi.c b/drivers/spi/spi.c
index 200ca228d885..8146ccd35a1a 100644
--- a/drivers/spi/spi.c
+++ b/drivers/spi/spi.c
@@ -37,6 +37,7 @@
 #include <linux/kthread.h>
 #include <linux/ioport.h>
 #include <linux/acpi.h>
+#include <linux/highmem.h>
 
 #define CREATE_TRACE_POINTS
 #include <trace/events/spi.h>
@@ -709,6 +710,13 @@ static int spi_map_buf(struct spi_master *master, struct device *dev,
 {
 	const bool vmalloced_buf = is_vmalloc_addr(buf);
 	unsigned int max_seg_size = dma_get_max_seg_size(dev);
+#ifdef CONFIG_HIGHMEM
+	const bool kmap_buf = ((unsigned long)buf >= PKMAP_BASE &&
+				(unsigned long)buf < (PKMAP_BASE +
+					(LAST_PKMAP * PAGE_SIZE)));
+#else
+	const bool kmap_buf = false;
+#endif
 	int desc_len;
 	int sgs;
 	struct page *vm_page;
@@ -716,7 +724,7 @@ static int spi_map_buf(struct spi_master *master, struct device *dev,
 	size_t min;
 	int i, ret;
 
-	if (vmalloced_buf) {
+	if (vmalloced_buf || kmap_buf) {
 		desc_len = min_t(int, max_seg_size, PAGE_SIZE);
 		sgs = DIV_ROUND_UP(len + offset_in_page(buf), desc_len);
 	} else if (virt_addr_valid(buf)) {
@@ -732,10 +740,13 @@ static int spi_map_buf(struct spi_master *master, struct device *dev,
 
 	for (i = 0; i < sgs; i++) {
 
-		if (vmalloced_buf) {
+		if (vmalloced_buf || kmap_buf) {
 			min = min_t(size_t,
 				    len, desc_len - offset_in_page(buf));
-			vm_page = vmalloc_to_page(buf);
+			if (vmalloced_buf)
+				vm_page = vmalloc_to_page(buf);
+			else
+				vm_page = kmap_to_page(buf);
 			if (!vm_page) {
 				sg_free_table(sgt);
 				return -ENOMEM;