[PATCH v2 06/10] drivers: spi: Add SPI controller driver for Octeon

[Daniel Schwierzeck <daniel.schwierzeck@gmail.com>]

Am Donnerstag, den 23.07.2020, 12:17 +0200 schrieb Stefan Roese:
> From: Suneel Garapati <sgarapati@marvell.com>
> 
> Adds support for SPI controllers found on Octeon II/III and Octeon TX
> TX2 SoC platforms.
> 
> Signed-off-by: Aaron Williams <awilliams@marvell.com>
> Signed-off-by: Suneel Garapati <sgarapati@marvell.com>
> Signed-off-by: Stefan Roese <sr@denx.de>
> Cc: Daniel Schwierzeck <daniel.schwierzeck@gmail.com>
> Cc: Aaron Williams <awilliams@marvell.com>
> Cc: Chandrakala Chavva <cchavva@marvell.com>
> Cc: Jagan Teki <jagan@amarulasolutions.com>
> 
> ---
> 
> Changes in v2:
> - Newly added to this series
> - Removed inclusion of "common.h"
> - Added "depends on DM_PCI" to Kconfig
> - Tested on MIPS Octeon and ARM Octeon TX2
> - Fixed issues with Octeon TX2 registration. Now only one driver is
>   registered and the "ops" is overwritten in the Octeon TX2 case.
> - Use dev_get_driver_data() to get the driver data struct
> - Removed "struct pci_device_id" definition and U_BOOT_PCI_DEVICE()
>   as its not needed for the PCI based probing on Octeon TX2
> 
>  drivers/spi/Kconfig      |   8 +
>  drivers/spi/Makefile     |   1 +
>  drivers/spi/octeon_spi.c | 647 +++++++++++++++++++++++++++++++++++++++
>  3 files changed, 656 insertions(+)
>  create mode 100644 drivers/spi/octeon_spi.c
> 
> diff --git a/drivers/spi/Kconfig b/drivers/spi/Kconfig
> index 30d808d7bb..3fc2d0674a 100644
> --- a/drivers/spi/Kconfig
> +++ b/drivers/spi/Kconfig
> @@ -240,6 +240,14 @@ config NXP_FSPI
>  	  Enable the NXP FlexSPI (FSPI) driver. This driver can be used to
>  	  access the SPI NOR flash on platforms embedding this NXP IP core.
>  
> +config OCTEON_SPI
> +	bool "Octeon SPI driver"
> +	depends on DM_PCI && (ARCH_OCTEON || ARCH_OCTEONTX || ARCH_OCTEONTX2)
> +	help
> +	  Enable the Octeon SPI driver. This driver can be used to
> +	  access the SPI NOR flash on Octeon II/III and OcteonTX/TX2
> +	  SoC platforms.
> +
>  config OMAP3_SPI
>  	bool "McSPI driver for OMAP"
>  	help
> diff --git a/drivers/spi/Makefile b/drivers/spi/Makefile
> index 4e7461771f..b5c9ff1af8 100644
> --- a/drivers/spi/Makefile
> +++ b/drivers/spi/Makefile
> @@ -43,6 +43,7 @@ obj-$(CONFIG_MXC_SPI) += mxc_spi.o
>  obj-$(CONFIG_MXS_SPI) += mxs_spi.o
>  obj-$(CONFIG_NXP_FSPI) += nxp_fspi.o
>  obj-$(CONFIG_ATCSPI200_SPI) += atcspi200_spi.o
> +obj-$(CONFIG_OCTEON_SPI) += octeon_spi.o
>  obj-$(CONFIG_OMAP3_SPI) += omap3_spi.o
>  obj-$(CONFIG_PIC32_SPI) += pic32_spi.o
>  obj-$(CONFIG_PL022_SPI) += pl022_spi.o
> diff --git a/drivers/spi/octeon_spi.c b/drivers/spi/octeon_spi.c
> new file mode 100644
> index 0000000000..2fb39e444c
> --- /dev/null
> +++ b/drivers/spi/octeon_spi.c
> @@ -0,0 +1,647 @@
> +// SPDX-License-Identifier:    GPL-2.0
> +/*
> + * Copyright (C) 2018 Marvell International Ltd.
> + *
> + * https://spdx.org/licenses
> + */
> +
> +#include <clk.h>
> +#include <dm.h>
> +#include <malloc.h>
> +#include <spi.h>
> +#include <spi-mem.h>
> +#include <watchdog.h>
> +#include <asm/io.h>
> +#include <asm/unaligned.h>
> +#include <linux/bitfield.h>
> +#include <linux/compat.h>
> +#include <linux/delay.h>
> +
> +#define OCTEON_SPI_MAX_BYTES	9
> +#define OCTEON_SPI_MAX_CLOCK_HZ	50000000
> +
> +#define OCTEON_SPI_NUM_CS	4
> +
> +#define OCTEON_SPI_CS_VALID(cs)	((cs) < OCTEON_SPI_NUM_CS)
> +
> +#define MPI_CFG			0x0000
> +#define MPI_STS			0x0008
> +#define MPI_TX			0x0010
> +#define MPI_XMIT		0x0018
> +#define MPI_WIDE_DAT		0x0040
> +#define MPI_IO_CTL		0x0048
> +#define MPI_DAT(X)		(0x0080 + ((X) << 3))
> +#define MPI_WIDE_BUF(X)		(0x0800 + ((X) << 3))
> +#define MPI_CYA_CFG		0x1000
> +#define MPI_CLKEN		0x1080
> +
> +#define MPI_CFG_ENABLE		BIT_ULL(0)
> +#define MPI_CFG_IDLELO		BIT_ULL(1)
> +#define MPI_CFG_CLK_CONT	BIT_ULL(2)
> +#define MPI_CFG_WIREOR		BIT_ULL(3)
> +#define MPI_CFG_LSBFIRST	BIT_ULL(4)
> +#define MPI_CFG_CS_STICKY	BIT_ULL(5)
> +#define MPI_CFG_CSHI		BIT_ULL(7)
> +#define MPI_CFG_IDLECLKS	GENMASK_ULL(9, 8)
> +#define MPI_CFG_TRITX		BIT_ULL(10)
> +#define MPI_CFG_CSLATE		BIT_ULL(11)
> +#define MPI_CFG_CSENA0		BIT_ULL(12)
> +#define MPI_CFG_CSENA1		BIT_ULL(13)
> +#define MPI_CFG_CSENA2		BIT_ULL(14)
> +#define MPI_CFG_CSENA3		BIT_ULL(15)
> +#define MPI_CFG_CLKDIV		GENMASK_ULL(28, 16)
> +#define MPI_CFG_LEGACY_DIS	BIT_ULL(31)
> +#define MPI_CFG_IOMODE		GENMASK_ULL(35, 34)
> +#define MPI_CFG_TB100_EN	BIT_ULL(49)
> +
> +#define MPI_DAT_DATA		GENMASK_ULL(7, 0)
> +
> +#define MPI_STS_BUSY		BIT_ULL(0)
> +#define MPI_STS_MPI_INTR	BIT_ULL(1)
> +#define MPI_STS_RXNUM		GENMASK_ULL(12, 8)
> +
> +#define MPI_TX_TOTNUM		GENMASK_ULL(4, 0)
> +#define MPI_TX_TXNUM		GENMASK_ULL(12, 8)
> +#define MPI_TX_LEAVECS		BIT_ULL(16)
> +#define MPI_TX_CSID		GENMASK_ULL(21, 20)
> +
> +#define MPI_XMIT_TOTNUM		GENMASK_ULL(10, 0)
> +#define MPI_XMIT_TXNUM		GENMASK_ULL(30, 20)
> +#define MPI_XMIT_BUF_SEL	BIT_ULL(59)
> +#define MPI_XMIT_LEAVECS	BIT_ULL(60)
> +#define MPI_XMIT_CSID		GENMASK_ULL(62, 61)
> +
> +enum {
> +	PROBE_PCI = 0,		/* PCI based probing */
> +	PROBE_DT,		/* DT based probing */
> +};
> +
> +/* Used on Octeon TX2 */
> +void board_acquire_flash_arb(bool acquire);
> +
> +struct octeon_spi_data {
> +	int probe;
> +	u32 reg_offs;
> +};
> +
> +/* Local driver data structure */
> +struct octeon_spi {
> +	void __iomem *base;	/* Register base address */
> +	struct clk clk;
> +	u32 clkdiv;		/* Clock divisor for device speed */
> +};
> +
> +static u64 octeon_spi_set_mpicfg(struct udevice *dev)
> +{
> +	struct dm_spi_slave_platdata *slave = dev_get_parent_platdata(dev);
> +	struct udevice *bus = dev_get_parent(dev);
> +	struct octeon_spi *priv = dev_get_priv(bus);
> +	u64 mpi_cfg;
> +	uint max_speed = slave->max_hz;
> +	bool cpha, cpol;
> +
> +	if (!max_speed)
> +		max_speed = 12500000;
> +	if (max_speed > OCTEON_SPI_MAX_CLOCK_HZ)
> +		max_speed = OCTEON_SPI_MAX_CLOCK_HZ;
> +
> +	debug("\n slave params %d %d %d\n", slave->cs,
> +	      slave->max_hz, slave->mode);
> +	cpha = !!(slave->mode & SPI_CPHA);
> +	cpol = !!(slave->mode & SPI_CPOL);
> +
> +	mpi_cfg = FIELD_PREP(MPI_CFG_CLKDIV, priv->clkdiv & 0x1fff) |
> +		FIELD_PREP(MPI_CFG_CSHI, !!(slave->mode & SPI_CS_HIGH)) |
> +		FIELD_PREP(MPI_CFG_LSBFIRST, !!(slave->mode & SPI_LSB_FIRST)) |
> +		FIELD_PREP(MPI_CFG_WIREOR, !!(slave->mode & SPI_3WIRE)) |
> +		FIELD_PREP(MPI_CFG_IDLELO, cpha != cpol) |
> +		FIELD_PREP(MPI_CFG_CSLATE, cpha) |
> +		MPI_CFG_CSENA0 | MPI_CFG_CSENA1 |
> +		MPI_CFG_CSENA2 | MPI_CFG_CSENA1 |
> +		MPI_CFG_ENABLE;
> +
> +	debug("\n mpi_cfg %llx\n", mpi_cfg);
> +	return mpi_cfg;
> +}
> +
> +/**
> + * Wait until the SPI bus is ready
> + *
> + * @param	dev	SPI device to wait for
> + */
> +static void octeon_spi_wait_ready(struct udevice *dev)
> +{
> +	struct udevice *bus = dev_get_parent(dev);
> +	struct octeon_spi *priv = dev_get_priv(bus);
> +	void *base = priv->base;
> +	u64 mpi_sts;
> +
> +	do {
> +		mpi_sts = readq(base + MPI_STS);
> +		WATCHDOG_RESET();
> +	} while (mpi_sts & MPI_STS_BUSY);
> +
> +	debug("%s(%s)\n", __func__, dev->name);
> +}
> +
> +/**
> + * Claim the bus for a slave device
> + *
> + * @param	dev	SPI bus
> + *
> + * @return	0 for success, -EINVAL if chip select is invalid
> + */
> +static int octeon_spi_claim_bus(struct udevice *dev)
> +{
> +	struct udevice *bus = dev_get_parent(dev);
> +	struct octeon_spi *priv = dev_get_priv(bus);
> +	void *base = priv->base;
> +	u64 mpi_cfg;
> +
> +	debug("\n\n%s(%s)\n", __func__, dev->name);
> +	if (!OCTEON_SPI_CS_VALID(spi_chip_select(dev)))
> +		return -EINVAL;
> +
> +	if (IS_ENABLED(CONFIG_ARCH_OCTEONTX2))
> +		board_acquire_flash_arb(true);
> +
> +	mpi_cfg = readq(base + MPI_CFG);
> +	mpi_cfg &= ~MPI_CFG_TRITX;
> +	mpi_cfg |= MPI_CFG_ENABLE;
> +	writeq(mpi_cfg, base + MPI_CFG);
> +	mpi_cfg = readq(base + MPI_CFG);
> +	udelay(5);	/** Wait for bus to settle */
> +
> +	return 0;
> +}
> +
> +/**
> + * Release the bus to a slave device
> + *
> + * @param	dev	SPI bus
> + *
> + * @return	0 for success, -EINVAL if chip select is invalid
> + */
> +static int octeon_spi_release_bus(struct udevice *dev)
> +{
> +	struct udevice *bus = dev_get_parent(dev);
> +	struct octeon_spi *priv = dev_get_priv(bus);
> +	void *base = priv->base;
> +	u64 mpi_cfg;
> +
> +	debug("%s(%s)\n\n", __func__, dev->name);
> +	if (!OCTEON_SPI_CS_VALID(spi_chip_select(dev)))
> +		return -EINVAL;
> +
> +	if (IS_ENABLED(CONFIG_ARCH_OCTEONTX2))
> +		board_acquire_flash_arb(false);
> +
> +	mpi_cfg = readq(base + MPI_CFG);
> +	mpi_cfg &= ~MPI_CFG_ENABLE;
> +	writeq(mpi_cfg, base + MPI_CFG);
> +	mpi_cfg = readq(base + MPI_CFG);
> +	udelay(1);
> +
> +	return 0;
> +}
> +
> +static int octeon_spi_xfer(struct udevice *dev, unsigned int bitlen,
> +			   const void *dout, void *din, unsigned long flags)
> +{
> +	struct udevice *bus = dev_get_parent(dev);
> +	struct octeon_spi *priv = dev_get_priv(bus);
> +	void *base = priv->base;
> +	u64 mpi_tx;
> +	u64 mpi_cfg;
> +	u64 wide_dat = 0;
> +	int len = bitlen / 8;
> +	int i;
> +	const u8 *tx_data = dout;
> +	u8 *rx_data = din;
> +	int cs = spi_chip_select(dev);
> +
> +	if (!OCTEON_SPI_CS_VALID(cs))
> +		return -EINVAL;
> +
> +	debug("\n %s(%s, %u, %p, %p, 0x%lx), cs: %d\n",
> +	      __func__, dev->name, bitlen, dout, din, flags, cs);
> +
> +	mpi_cfg = octeon_spi_set_mpicfg(dev);
> +	if (mpi_cfg != readq(base + MPI_CFG)) {
> +		writeq(mpi_cfg, base + MPI_CFG);
> +		mpi_cfg = readq(base + MPI_CFG);
> +		udelay(10);
> +	}
> +
> +	debug("\n mpi_cfg upd %llx\n", mpi_cfg);
> +
> +	/*
> +	 * Start by writing and reading 8 bytes at a time. While we can support
> +	 * up to 10, it's easier to just use 8 with the MPI_WIDE_DAT register.
> +	 */
> +	while (len > 8) {
> +		if (tx_data) {
> +			wide_dat = get_unaligned((u64 *)tx_data);
> +			debug("  tx: %016llx \t", (unsigned long long)wide_dat);
> +			tx_data += 8;
> +			writeq(wide_dat, base + MPI_WIDE_DAT);
> +		}
> +
> +		mpi_tx = FIELD_PREP(MPI_TX_CSID, cs) |
> +			FIELD_PREP(MPI_TX_LEAVECS, 1) |
> +			FIELD_PREP(MPI_TX_TXNUM, tx_data ? 8 : 0) |
> +			FIELD_PREP(MPI_TX_TOTNUM, 8);
> +		writeq(mpi_tx, base + MPI_TX);
> +
> +		octeon_spi_wait_ready(dev);
> +
> +		debug("\n ");
> +
> +		if (rx_data) {
> +			wide_dat = readq(base + MPI_WIDE_DAT);
> +			debug("  rx: %016llx\t", (unsigned long long)wide_dat);
> +			*(u64 *)rx_data = wide_dat;
> +			rx_data += 8;
> +		}
> +		len -= 8;
> +	}
> +
> +	debug("\n ");
> +
> +	/* Write and read the rest of the data */
> +	if (tx_data) {
> +		for (i = 0; i < len; i++) {
> +			debug("  tx: %02x\n", *tx_data);
> +			writeq(*tx_data++, base + MPI_DAT(i));
> +		}
> +	}
> +
> +	mpi_tx = FIELD_PREP(MPI_TX_CSID, cs) |
> +		FIELD_PREP(MPI_TX_LEAVECS, !(flags & SPI_XFER_END)) |
> +		FIELD_PREP(MPI_TX_TXNUM, tx_data ? len : 0) |
> +		FIELD_PREP(MPI_TX_TOTNUM, len);
> +	writeq(mpi_tx, base + MPI_TX);
> +
> +	octeon_spi_wait_ready(dev);
> +
> +	debug("\n ");
> +
> +	if (rx_data) {
> +		for (i = 0; i < len; i++) {
> +			*rx_data = readq(base + MPI_DAT(i)) & 0xff;
> +			debug("  rx: %02x\n", *rx_data);
> +			rx_data++;
> +		}
> +	}
> +
> +	return 0;
> +}
> +
> +static int octeontx2_spi_xfer(struct udevice *dev, unsigned int bitlen,
> +			      const void *dout, void *din, unsigned long flags)
> +{
> +	struct udevice *bus = dev_get_parent(dev);
> +	struct octeon_spi *priv = dev_get_priv(bus);
> +	void *base = priv->base;
> +	u64 mpi_xmit;
> +	u64 mpi_cfg;
> +	u64 wide_dat = 0;
> +	int len = bitlen / 8;
> +	int rem;
> +	int i;
> +	const u8 *tx_data = dout;
> +	u8 *rx_data = din;
> +	int cs = spi_chip_select(dev);
> +
> +	if (!OCTEON_SPI_CS_VALID(cs))
> +		return -EINVAL;
> +
> +	debug("\n %s(%s, %u, %p, %p, 0x%lx), cs: %d\n",
> +	      __func__, dev->name, bitlen, dout, din, flags, cs);
> +
> +	mpi_cfg = octeon_spi_set_mpicfg(dev);
> +
> +	mpi_cfg |= MPI_CFG_TRITX | MPI_CFG_LEGACY_DIS | MPI_CFG_CS_STICKY |
> +		MPI_CFG_TB100_EN;
> +
> +	mpi_cfg &= ~MPI_CFG_IOMODE;
> +	if (flags & (SPI_TX_DUAL | SPI_RX_DUAL))
> +		mpi_cfg |= FIELD_PREP(MPI_CFG_IOMODE, 2);
> +	if (flags & (SPI_TX_QUAD | SPI_RX_QUAD))
> +		mpi_cfg |= FIELD_PREP(MPI_CFG_IOMODE, 3);
> +
> +	if (mpi_cfg != readq(base + MPI_CFG)) {
> +		writeq(mpi_cfg, base + MPI_CFG);
> +		mpi_cfg = readq(base + MPI_CFG);
> +		udelay(10);
> +	}
> +
> +	debug("\n mpi_cfg upd %llx\n\n", mpi_cfg);
> +
> +	/* Start by writing or reading 1024 bytes at a time. */
> +	while (len > 1024) {
> +		if (tx_data) {
> +			/* 8 bytes per iteration */
> +			for (i = 0; i < 128; i++) {
> +				wide_dat = get_unaligned((u64 *)tx_data);
> +				debug("  tx: %016llx \t",
> +				      (unsigned long long)wide_dat);
> +				if ((i % 4) == 3)
> +					debug("\n");
> +				tx_data += 8;
> +				writeq(wide_dat, base + MPI_WIDE_BUF(i));
> +			}
> +		}
> +
> +		mpi_xmit = FIELD_PREP(MPI_XMIT_CSID, cs) | MPI_XMIT_LEAVECS |
> +			FIELD_PREP(MPI_XMIT_TXNUM, tx_data ? 1024 : 0) |
> +			FIELD_PREP(MPI_XMIT_TOTNUM, 1024);
> +		writeq(mpi_xmit, base + MPI_XMIT);
> +
> +		octeon_spi_wait_ready(dev);
> +
> +		debug("\n ");
> +
> +		if (rx_data) {
> +			/* 8 bytes per iteration */
> +			for (i = 0; i < 128; i++) {
> +				wide_dat = readq(base + MPI_WIDE_BUF(i));
> +				debug("  rx: %016llx\t",
> +				      (unsigned long long)wide_dat);
> +				if ((i % 4) == 3)
> +					debug("\n");
> +				*(u64 *)rx_data = wide_dat;
> +				rx_data += 8;
> +			}
> +		}
> +		len -= 1024;
> +	}
> +
> +	if (tx_data) {
> +		rem = len % 8;
> +		/* 8 bytes per iteration */
> +		for (i = 0; i < len / 8; i++) {
> +			wide_dat = get_unaligned((u64 *)tx_data);
> +			debug("  tx: %016llx \t",
> +			      (unsigned long long)wide_dat);
> +			if ((i % 4) == 3)
> +				debug("\n");
> +			tx_data += 8;
> +			writeq(wide_dat, base + MPI_WIDE_BUF(i));
> +		}
> +		if (rem) {
> +			memcpy(&wide_dat, tx_data, rem);
> +			debug("  rtx: %016llx\t", wide_dat);
> +			writeq(wide_dat, base + MPI_WIDE_BUF(i));
> +		}
> +	}
> +
> +	mpi_xmit = FIELD_PREP(MPI_XMIT_CSID, cs) |
> +		FIELD_PREP(MPI_XMIT_LEAVECS, !(flags & SPI_XFER_END)) |
> +		FIELD_PREP(MPI_XMIT_TXNUM, tx_data ? len : 0) |
> +		FIELD_PREP(MPI_XMIT_TOTNUM, len);
> +	writeq(mpi_xmit, base + MPI_XMIT);
> +
> +	octeon_spi_wait_ready(dev);
> +
> +	debug("\n ");
> +
> +	if (rx_data) {
> +		rem = len % 8;
> +		/* 8 bytes per iteration */
> +		for (i = 0; i < len / 8; i++) {
> +			wide_dat = readq(base + MPI_WIDE_BUF(i));
> +			debug("  rx: %016llx\t",
> +			      (unsigned long long)wide_dat);
> +			if ((i % 4) == 3)
> +				debug("\n");
> +			*(u64 *)rx_data = wide_dat;
> +			rx_data += 8;
> +		}
> +		if (rem) {
> +			wide_dat = readq(base + MPI_WIDE_BUF(i));
> +			debug("  rrx: %016llx\t",
> +			      (unsigned long long)wide_dat);
> +			memcpy(rx_data, &wide_dat, rem);
> +			rx_data += rem;
> +		}
> +	}
> +
> +	return 0;
> +}
> +
> +static bool octeon_spi_supports_op(struct spi_slave *slave,
> +				   const struct spi_mem_op *op)
> +{
> +	/* For now, support only below combinations
> +	 * 1-1-1
> +	 * 1-1-2 1-2-2
> +	 * 1-1-4 1-4-4
> +	 */
> +	if (op->cmd.buswidth != 1)
> +		return false;
> +	return true;
> +}
> +
> +static int octeon_spi_exec_op(struct spi_slave *slave,
> +			      const struct spi_mem_op *op)
> +{
> +	unsigned long flags = SPI_XFER_BEGIN;
> +	const void *tx;
> +	void *rx;
> +	u8 opcode, *buf;
> +	u8 *addr;
> +	int i, temp, ret;
> +
> +	if (op->cmd.buswidth != 1)
> +		return -ENOTSUPP;
> +
> +	/* Send CMD */
> +	i = 0;
> +	opcode = op->cmd.opcode;
> +
> +	if (!op->data.nbytes && !op->addr.nbytes && !op->dummy.nbytes)
> +		flags |= SPI_XFER_END;
> +
> +	ret = octeontx2_spi_xfer(slave->dev, 8, (void *)&opcode, NULL, flags);
> +	if (ret < 0)
> +		return ret;
> +
> +	/* Send Address and dummy */
> +	if (op->addr.nbytes) {
> +		/* Alloc buffer for address+dummy */
> +		buf = (u8 *)calloc(1, op->addr.nbytes + op->dummy.nbytes);
> +		if (!buf) {
> +			printf("%s Out of memory\n", __func__);
> +			return -ENOMEM;
> +		}
> +		addr = (u8 *)&op->addr.val;
> +		for (temp = 0; temp < op->addr.nbytes; temp++)
> +			buf[i++] = *(u8 *)(addr + op->addr.nbytes - 1 - temp);
> +		for (temp = 0; temp < op->dummy.nbytes; temp++)
> +			buf[i++] = 0xff;
> +		if (op->addr.buswidth == 2)
> +			flags |= SPI_RX_DUAL;
> +		if (op->addr.buswidth == 4)
> +			flags |= SPI_RX_QUAD;
> +
> +		if (!op->data.nbytes)
> +			flags |= SPI_XFER_END;
> +		ret = octeontx2_spi_xfer(slave->dev, i * 8, (void *)buf, NULL,
> +					 flags);
> +		free(buf);
> +		if (ret < 0)
> +			return ret;
> +	}
> +	if (!op->data.nbytes)
> +		return 0;
> +
> +	/* Send/Receive Data */
> +	flags |= SPI_XFER_END;
> +	if (op->data.buswidth == 2)
> +		flags |= SPI_RX_DUAL;
> +	if (op->data.buswidth == 4)
> +		flags |= SPI_RX_QUAD;
> +
> +	rx = (op->data.dir == SPI_MEM_DATA_IN) ? op->data.buf.in : NULL;
> +	tx = (op->data.dir == SPI_MEM_DATA_OUT) ? op->data.buf.out : NULL;
> +
> +	ret = octeontx2_spi_xfer(slave->dev, (op->data.nbytes * 8), tx, rx,
> +				 flags);
> +	return ret;
> +}
> +
> +static const struct spi_controller_mem_ops octeontx2_spi_mem_ops = {
> +	.supports_op = octeon_spi_supports_op,
> +	.exec_op = octeon_spi_exec_op,
> +};
> +
> +/**
> + * Set the speed of the SPI bus
> + *
> + * @param	bus	bus to set
> + * @param	max_hz	maximum speed supported
> + */
> +static int octeon_spi_set_speed(struct udevice *bus, uint max_hz)
> +{
> +	struct octeon_spi *priv = dev_get_priv(bus);
> +	ulong clk_rate;
> +	u32 calc_hz;
> +
> +	if (max_hz > OCTEON_SPI_MAX_CLOCK_HZ)
> +		max_hz = OCTEON_SPI_MAX_CLOCK_HZ;
> +
> +	clk_rate = clk_get_rate(&priv->clk);
> +	if (IS_ERR_VALUE(clk_rate))
> +		return -EINVAL;
> +
> +	debug("%s(%s, %u, %lu)\n", __func__, bus->name, max_hz, clk_rate);
> +
> +	priv->clkdiv = clk_rate / (2 * max_hz);
> +	while (1) {
> +		calc_hz = clk_rate / (2 * priv->clkdiv);
> +		if (calc_hz <= max_hz)
> +			break;
> +		priv->clkdiv += 1;
> +	}
> +
> +	if (priv->clkdiv > 8191)
> +		return -EINVAL;
> +
> +	debug("%s: clkdiv=%d\n", __func__, priv->clkdiv);
> +
> +	return 0;
> +}
> +
> +static int octeon_spi_set_mode(struct udevice *bus, uint mode)
> +{
> +	/* We don't set it here */
> +	return 0;
> +}
> +
> +static const struct dm_spi_ops octeon_spi_ops = {
> +	.claim_bus	= octeon_spi_claim_bus,
> +	.release_bus	= octeon_spi_release_bus,
> +	.set_speed	= octeon_spi_set_speed,
> +	.set_mode	= octeon_spi_set_mode,
> +	.xfer		= octeon_spi_xfer,
> +};
> +
> +static const struct dm_spi_ops octeontx2_spi_ops = {
> +	.claim_bus	= octeon_spi_claim_bus,
> +	.release_bus	= octeon_spi_release_bus,
> +	.set_speed	= octeon_spi_set_speed,
> +	.set_mode	= octeon_spi_set_mode,
> +	.xfer		= octeontx2_spi_xfer,
> +	.mem_ops	= &octeontx2_spi_mem_ops,
> +};
> +
> +static int octeon_spi_probe(struct udevice *dev)
> +{
> +	struct octeon_spi *priv = dev_get_priv(dev);
> +	const struct octeon_spi_data *data;
> +	int ret;
> +
> +	data = (const struct octeon_spi_data *)dev_get_driver_data(dev);
> +	if (data->probe == PROBE_PCI) {
> +		pci_dev_t bdf = dm_pci_get_bdf(dev);
> +
> +		debug("SPI PCI device: %x\n", bdf);
> +		priv->base = dm_pci_map_bar(dev, PCI_BASE_ADDRESS_0,
> +					    PCI_REGION_MEM);
> +	} else {
> +		priv->base = dev_remap_addr(dev);
> +	}
> +
> +	priv->base += data->reg_offs;
> +
> +	/* Octeon TX2 needs a different ops struct */
> +	if (device_is_compatible(dev, "cavium,thunderx-spi")) {
> +		/*
> +		 * "ops" is const and can't be written directly. So we need
> +		 * to write the Octeon TX2 ops value using this trick
> +		 */
> +		writeq((u64)&octeontx2_spi_ops, (void *)&dev->driver->ops);
> +	}

can't you simply add a xfer() function pointer to "struct
octeon_spi_data" and assign the according xfer function to it? Then
in octeon_spi_xfer() you can simply call that function pointer. With
this you only need one instance of "struct dm_spi_ops" and don't need
this ugly hack ;) Maybe you can add some common code to
octeon_spi_xfer() itself and reduce the size of the SoC specific xfer
functions.

> +
> +	ret = clk_get_by_index(dev, 0, &priv->clk);
> +	if (ret < 0)
> +		return ret;
> +
> +	ret = clk_enable(&priv->clk);
> +	if (ret)
> +		return ret;
> +
> +	debug("SPI bus %s %d at %p\n", dev->name, dev->seq, priv->base);
> +
> +	return 0;
> +}
> +
> +static const struct octeon_spi_data spi_octeon_data = {
> +	.probe = PROBE_DT,
> +	.reg_offs = 0x0000,
> +};
> +
> +static const struct octeon_spi_data spi_octeontx_data = {
> +	.probe = PROBE_PCI,
> +	.reg_offs = 0x1000,
> +};
> +
> +static const struct udevice_id octeon_spi_ids[] = {
> +	/* MIPS Octeon */
> +	{ .compatible = "cavium,octeon-3010-spi",
> +	  .data = (ulong)&spi_octeon_data },
> +	/* ARM Octeon TX / TX2 */
> +	{ .compatible = "cavium,thunder-8190-spi",
> +	  .data = (ulong)&spi_octeontx_data },
> +	{ }
> +};
> +
> +U_BOOT_DRIVER(octeon_spi) = {
> +	.name			= "spi_octeon",
> +	.id			= UCLASS_SPI,
> +	.of_match		= octeon_spi_ids,
> +	.probe			= octeon_spi_probe,
> +	.priv_auto_alloc_size	= sizeof(struct octeon_spi),
> +	.ops			= &octeon_spi_ops,
> +};
-- 
- Daniel