// SPDX-License-Identifier: GPL-2.0-only
//
// Copyright (C) 2020 NVIDIA CORPORATION.

#include <linux/clk.h>
#include <linux/completion.h>
#include <linux/delay.h>
#include <linux/dmaengine.h>
#include <linux/dma-mapping.h>
#include <linux/dmapool.h>
#include <linux/err.h>
#include <linux/interrupt.h>
#include <linux/io.h>
#include <linux/iopoll.h>
#include <linux/kernel.h>
#include <linux/kthread.h>
#include <linux/module.h>
#include <linux/platform_device.h>
#include <linux/pm_runtime.h>
#include <linux/of.h>
#include <linux/reset.h>
#include <linux/spi/spi.h>
#include <linux/acpi.h>
#include <linux/property.h>
#include <linux/sizes.h>

#define QSPI_COMMAND1				0x000
#define QSPI_BIT_LENGTH(x)			(((x) & 0x1f) << 0)
#define QSPI_PACKED				BIT(5)
#define QSPI_INTERFACE_WIDTH_MASK		(0x03 << 7)
#define QSPI_INTERFACE_WIDTH(x)			(((x) & 0x03) << 7)
#define QSPI_INTERFACE_WIDTH_SINGLE		QSPI_INTERFACE_WIDTH(0)
#define QSPI_INTERFACE_WIDTH_DUAL		QSPI_INTERFACE_WIDTH(1)
#define QSPI_INTERFACE_WIDTH_QUAD		QSPI_INTERFACE_WIDTH(2)
#define QSPI_SDR_DDR_SEL			BIT(9)
#define QSPI_TX_EN				BIT(11)
#define QSPI_RX_EN				BIT(12)
#define QSPI_CS_SW_VAL				BIT(20)
#define QSPI_CS_SW_HW				BIT(21)

#define QSPI_CS_POL_INACTIVE(n)			(1 << (22 + (n)))
#define QSPI_CS_POL_INACTIVE_MASK		(0xF << 22)
#define QSPI_CS_SEL_0				(0 << 26)
#define QSPI_CS_SEL_1				(1 << 26)
#define QSPI_CS_SEL_2				(2 << 26)
#define QSPI_CS_SEL_3				(3 << 26)
#define QSPI_CS_SEL_MASK			(3 << 26)
#define QSPI_CS_SEL(x)				(((x) & 0x3) << 26)

#define QSPI_CONTROL_MODE_0			(0 << 28)
#define QSPI_CONTROL_MODE_3			(3 << 28)
#define QSPI_CONTROL_MODE_MASK			(3 << 28)
#define QSPI_M_S				BIT(30)
#define QSPI_PIO				BIT(31)

#define QSPI_COMMAND2				0x004
#define QSPI_TX_TAP_DELAY(x)			(((x) & 0x3f) << 10)
#define QSPI_RX_TAP_DELAY(x)			(((x) & 0xff) << 0)

#define QSPI_CS_TIMING1				0x008
#define QSPI_SETUP_HOLD(setup, hold)		(((setup) << 4) | (hold))

#define QSPI_CS_TIMING2				0x00c
#define CYCLES_BETWEEN_PACKETS_0(x)		(((x) & 0x1f) << 0)
#define CS_ACTIVE_BETWEEN_PACKETS_0		BIT(5)

#define QSPI_TRANS_STATUS			0x010
#define QSPI_BLK_CNT(val)			(((val) >> 0) & 0xffff)
#define QSPI_RDY				BIT(30)

#define QSPI_FIFO_STATUS			0x014
#define QSPI_RX_FIFO_EMPTY			BIT(0)
#define QSPI_RX_FIFO_FULL			BIT(1)
#define QSPI_TX_FIFO_EMPTY			BIT(2)
#define QSPI_TX_FIFO_FULL			BIT(3)
#define QSPI_RX_FIFO_UNF			BIT(4)
#define QSPI_RX_FIFO_OVF			BIT(5)
#define QSPI_TX_FIFO_UNF			BIT(6)
#define QSPI_TX_FIFO_OVF			BIT(7)
#define QSPI_ERR				BIT(8)
#define QSPI_TX_FIFO_FLUSH			BIT(14)
#define QSPI_RX_FIFO_FLUSH			BIT(15)
#define QSPI_TX_FIFO_EMPTY_COUNT(val)		(((val) >> 16) & 0x7f)
#define QSPI_RX_FIFO_FULL_COUNT(val)		(((val) >> 23) & 0x7f)

#define QSPI_FIFO_ERROR				(QSPI_RX_FIFO_UNF | \
						 QSPI_RX_FIFO_OVF | \
						 QSPI_TX_FIFO_UNF | \
						 QSPI_TX_FIFO_OVF)
#define QSPI_FIFO_EMPTY				(QSPI_RX_FIFO_EMPTY | \
						 QSPI_TX_FIFO_EMPTY)

#define QSPI_TX_DATA				0x018
#define QSPI_RX_DATA				0x01c

#define QSPI_DMA_CTL				0x020
#define QSPI_TX_TRIG(n)				(((n) & 0x3) << 15)
#define QSPI_TX_TRIG_1				QSPI_TX_TRIG(0)
#define QSPI_TX_TRIG_4				QSPI_TX_TRIG(1)
#define QSPI_TX_TRIG_8				QSPI_TX_TRIG(2)
#define QSPI_TX_TRIG_16				QSPI_TX_TRIG(3)

#define QSPI_RX_TRIG(n)				(((n) & 0x3) << 19)
#define QSPI_RX_TRIG_1				QSPI_RX_TRIG(0)
#define QSPI_RX_TRIG_4				QSPI_RX_TRIG(1)
#define QSPI_RX_TRIG_8				QSPI_RX_TRIG(2)
#define QSPI_RX_TRIG_16				QSPI_RX_TRIG(3)

#define QSPI_DMA_EN				BIT(31)

#define QSPI_DMA_BLK				0x024
#define QSPI_DMA_BLK_SET(x)			(((x) & 0xffff) << 0)

#define QSPI_DMA_MEM_ADDRESS			0x028
#define QSPI_DMA_HI_ADDRESS			0x02c

#define QSPI_TX_FIFO				0x108
#define QSPI_RX_FIFO				0x188

#define QSPI_FIFO_DEPTH				64

#define QSPI_INTR_MASK				0x18c
#define QSPI_INTR_RX_FIFO_UNF_MASK		BIT(25)
#define QSPI_INTR_RX_FIFO_OVF_MASK		BIT(26)
#define QSPI_INTR_TX_FIFO_UNF_MASK		BIT(27)
#define QSPI_INTR_TX_FIFO_OVF_MASK		BIT(28)
#define QSPI_INTR_RDY_MASK			BIT(29)
#define QSPI_INTR_RX_TX_FIFO_ERR		(QSPI_INTR_RX_FIFO_UNF_MASK | \
						 QSPI_INTR_RX_FIFO_OVF_MASK | \
						 QSPI_INTR_TX_FIFO_UNF_MASK | \
						 QSPI_INTR_TX_FIFO_OVF_MASK)

#define QSPI_MISC_REG                           0x194
#define QSPI_NUM_DUMMY_CYCLE(x)			(((x) & 0xff) << 0)
#define QSPI_DUMMY_CYCLES_MAX			0xff

#define QSPI_CMB_SEQ_CMD			0x19c
#define QSPI_COMMAND_VALUE_SET(X)		(((x) & 0xFF) << 0)

#define QSPI_CMB_SEQ_CMD_CFG			0x1a0
#define QSPI_COMMAND_X1_X2_X4(x)		((((x) >> 1) & 0x3) << 13)
#define QSPI_COMMAND_X1_X2_X4_MASK		(0x03 << 13)
#define QSPI_COMMAND_SDR_DDR			BIT(12)
#define QSPI_COMMAND_SIZE_SET(x)		(((x) & 0xFF) << 0)

#define QSPI_GLOBAL_CONFIG			0X1a4
#define QSPI_CMB_SEQ_EN				BIT(0)
#define QSPI_TPM_WAIT_POLL_EN			BIT(1)

#define QSPI_CMB_SEQ_ADDR			0x1a8
#define QSPI_ADDRESS_VALUE_SET(X)		(((x) & 0xFFFF) << 0)

#define QSPI_CMB_SEQ_ADDR_CFG			0x1ac
#define QSPI_ADDRESS_X1_X2_X4(x)		((((x) >> 1) & 0x3) << 13)
#define QSPI_ADDRESS_X1_X2_X4_MASK		(0x03 << 13)
#define QSPI_ADDRESS_SDR_DDR			BIT(12)
#define QSPI_ADDRESS_SIZE_SET(x)		(((x) & 0xFF) << 0)

#define DATA_DIR_TX				BIT(0)
#define DATA_DIR_RX				BIT(1)

#define QSPI_DMA_TIMEOUT			(msecs_to_jiffies(1000))
#define DEFAULT_QSPI_DMA_BUF_LEN		SZ_64K

enum tegra_qspi_transfer_type {
	CMD_TRANSFER   = 0,
	ADDR_TRANSFER  = 1,
	DUMMY_TRANSFER = 2,
	DATA_TRANSFER  = 3
};

struct tegra_qspi_soc_data {
	bool cmb_xfer_capable;
	bool supports_tpm;
	bool has_ext_dma;
	unsigned int cs_count;
};

struct tegra_qspi_client_data {
	int tx_clk_tap_delay;
	int rx_clk_tap_delay;
};

struct tegra_qspi {
	struct device				*dev;
	struct spi_controller			*host;
	/* lock to protect data accessed by irq */
	spinlock_t				lock;

	struct clk				*clk;
	void __iomem				*base;
	phys_addr_t				phys;
	unsigned int				irq;
	struct work_struct			irq_work;
	struct workqueue_struct			*wq;
	/*
	 * Set by tegra_qspi_handle_timeout() while it drains the bottom
	 * half so tegra_qspi_isr() suppresses new queue_work() calls
	 * that would otherwise race the recovery path or the caller's
	 * cleanup of curr_xfer.
	 */
	bool					recovery_in_progress;

	u32					cur_speed;
	unsigned int				cur_pos;
	unsigned int				words_per_32bit;
	unsigned int				bytes_per_word;
	unsigned int				curr_dma_words;
	unsigned int				cur_direction;

	unsigned int				cur_rx_pos;
	unsigned int				cur_tx_pos;

	unsigned int				dma_buf_size;
	unsigned int				max_buf_size;
	bool					is_curr_dma_xfer;
	/*
	 * Cached "this PIO chunk completes the whole transfer" decision,
	 * computed by tegra_qspi_start_cpu_based_transfer() before it
	 * unmasks the IRQ. Used by the hard IRQ small-PIO fastpath in
	 * place of dereferencing curr_xfer->len, so the ISR cannot touch
	 * the spi_transfer object even on a late IRQ that races with the
	 * synchronous teardown path. Multi-chunk PIO transfers always go
	 * through the workqueue (this flag is only set on the final
	 * chunk), so the fastpath cannot recurse into
	 * tegra_qspi_start_cpu_based_transfer() from hard IRQ context.
	 */
	bool					is_last_pio_chunk;

	struct completion			rx_dma_complete;
	struct completion			tx_dma_complete;

	u32					tx_status;
	u32					rx_status;
	u32					status_reg;
	u32					trans_status;
	bool					is_packed;
	bool					use_dma;

	/*
	 * Set while a streaming mapping created by
	 * tegra_qspi_dma_map_xfer() is outstanding, together with the
	 * length that was handed to dma_map_single(). is_curr_dma_xfer
	 * only records that the last armed chunk took the DMA path,
	 * which is not the same thing: a chunk whose mapping failed
	 * leaves the DMA path flagged with nothing mapped.
	 */
	bool					is_dma_mapped;
	unsigned int				dma_map_len;

	u32					command1_reg;
	u32					dma_control_reg;
	u32					def_command1_reg;
	u32					def_command2_reg;
	u32					spi_cs_timing1;
	u32					spi_cs_timing2;
	u8					dummy_cycles;

	struct completion			xfer_completion;
	struct spi_transfer			*curr_xfer;

	struct device				*rx_dma_dev;
	struct dma_chan				*rx_dma_chan;
	u32					*rx_dma_buf;
	dma_addr_t				rx_dma_phys;
	struct dma_async_tx_descriptor		*rx_dma_desc;

	struct device				*tx_dma_dev;
	struct dma_chan				*tx_dma_chan;
	u32					*tx_dma_buf;
	dma_addr_t				tx_dma_phys;
	struct dma_async_tx_descriptor		*tx_dma_desc;
	const struct tegra_qspi_soc_data	*soc_data;
};

static inline u32 tegra_qspi_readl(struct tegra_qspi *tqspi, unsigned long offset)
{
	return readl(tqspi->base + offset);
}

static inline void tegra_qspi_writel(struct tegra_qspi *tqspi, u32 value, unsigned long offset)
{
	writel(value, tqspi->base + offset);

	/* read back register to make sure that register writes completed */
	if (offset != QSPI_TX_FIFO)
		readl(tqspi->base + QSPI_COMMAND1);
}

static void tegra_qspi_mask_clear_irq(struct tegra_qspi *tqspi)
{
	u32 value;

	/* write 1 to clear status register */
	value = tegra_qspi_readl(tqspi, QSPI_TRANS_STATUS);
	tegra_qspi_writel(tqspi, value, QSPI_TRANS_STATUS);

	value = tegra_qspi_readl(tqspi, QSPI_INTR_MASK);
	if (!(value & QSPI_INTR_RDY_MASK)) {
		value |= (QSPI_INTR_RDY_MASK | QSPI_INTR_RX_TX_FIFO_ERR);
		tegra_qspi_writel(tqspi, value, QSPI_INTR_MASK);
	}

	/* clear fifo status error if any */
	value = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
	if (value & QSPI_ERR)
		tegra_qspi_writel(tqspi, QSPI_ERR | QSPI_FIFO_ERROR, QSPI_FIFO_STATUS);
}

static unsigned int
tegra_qspi_calculate_curr_xfer_param(struct tegra_qspi *tqspi, struct spi_transfer *t)
{
	unsigned int max_word, max_len, total_fifo_words;
	unsigned int remain_len = t->len - tqspi->cur_pos;
	unsigned int bits_per_word = t->bits_per_word;

	tqspi->bytes_per_word = DIV_ROUND_UP(bits_per_word, 8);

	/*
	 * Tegra QSPI controller supports packed or unpacked mode transfers.
	 * Packed mode is used for data transfers using 8, 16, or 32 bits per
	 * word with a minimum transfer of 1 word and for all other transfers
	 * unpacked mode will be used.
	 */

	if ((bits_per_word == 8 || bits_per_word == 16 ||
	     bits_per_word == 32) && t->len > 3) {
		tqspi->is_packed = true;
		tqspi->words_per_32bit = 32 / bits_per_word;
	} else {
		tqspi->is_packed = false;
		tqspi->words_per_32bit = 1;
	}

	if (tqspi->is_packed) {
		max_len = min(remain_len, tqspi->max_buf_size);
		tqspi->curr_dma_words = max_len / tqspi->bytes_per_word;
		total_fifo_words = (max_len + 3) / 4;
	} else {
		max_word = (remain_len - 1) / tqspi->bytes_per_word + 1;
		max_word = min(max_word, tqspi->max_buf_size / 4);
		tqspi->curr_dma_words = max_word;
		total_fifo_words = max_word;
	}

	return total_fifo_words;
}

static unsigned int
tegra_qspi_fill_tx_fifo_from_client_txbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
{
	unsigned int written_words, fifo_words_left, count;
	unsigned int len, tx_empty_count, max_n_32bit, i;
	u8 *tx_buf = (u8 *)t->tx_buf + tqspi->cur_tx_pos;
	u32 fifo_status;

	fifo_status = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
	tx_empty_count = QSPI_TX_FIFO_EMPTY_COUNT(fifo_status);

	if (tqspi->is_packed) {
		fifo_words_left = tx_empty_count * tqspi->words_per_32bit;
		written_words = min(fifo_words_left, tqspi->curr_dma_words);
		len = written_words * tqspi->bytes_per_word;
		max_n_32bit = DIV_ROUND_UP(len, 4);
		for (count = 0; count < max_n_32bit; count++) {
			u32 x = 0;

			for (i = 0; (i < 4) && len; i++, len--)
				x |= (u32)(*tx_buf++) << (i * 8);
			tegra_qspi_writel(tqspi, x, QSPI_TX_FIFO);
		}

		tqspi->cur_tx_pos += written_words * tqspi->bytes_per_word;
	} else {
		unsigned int write_bytes;
		u8 bytes_per_word = tqspi->bytes_per_word;

		max_n_32bit = min(tqspi->curr_dma_words, tx_empty_count);
		written_words = max_n_32bit;
		len = written_words * tqspi->bytes_per_word;
		if (len > t->len - tqspi->cur_pos)
			len = t->len - tqspi->cur_pos;
		write_bytes = len;
		for (count = 0; count < max_n_32bit; count++) {
			u32 x = 0;

			for (i = 0; len && (i < min(4, bytes_per_word)); i++, len--)
				x |= (u32)(*tx_buf++) << (i * 8);
			tegra_qspi_writel(tqspi, x, QSPI_TX_FIFO);
		}

		tqspi->cur_tx_pos += write_bytes;
	}

	return written_words;
}

static unsigned int
tegra_qspi_read_rx_fifo_to_client_rxbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
{
	u8 *rx_buf = (u8 *)t->rx_buf + tqspi->cur_rx_pos;
	unsigned int len, rx_full_count, count, i;
	unsigned int read_words = 0;
	u32 fifo_status, x;

	fifo_status = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
	rx_full_count = QSPI_RX_FIFO_FULL_COUNT(fifo_status);
	if (tqspi->is_packed) {
		len = tqspi->curr_dma_words * tqspi->bytes_per_word;
		for (count = 0; count < rx_full_count; count++) {
			x = tegra_qspi_readl(tqspi, QSPI_RX_FIFO);

			for (i = 0; len && (i < 4); i++, len--)
				*rx_buf++ = (x >> i * 8) & 0xff;
		}

		read_words += tqspi->curr_dma_words;
		tqspi->cur_rx_pos += tqspi->curr_dma_words * tqspi->bytes_per_word;
	} else {
		u32 rx_mask = ((u32)1 << t->bits_per_word) - 1;
		u8 bytes_per_word = tqspi->bytes_per_word;
		unsigned int read_bytes;

		len = rx_full_count * bytes_per_word;
		if (len > t->len - tqspi->cur_pos)
			len = t->len - tqspi->cur_pos;
		read_bytes = len;
		for (count = 0; count < rx_full_count; count++) {
			x = tegra_qspi_readl(tqspi, QSPI_RX_FIFO) & rx_mask;

			for (i = 0; len && (i < bytes_per_word); i++, len--)
				*rx_buf++ = (x >> (i * 8)) & 0xff;
		}

		read_words += rx_full_count;
		tqspi->cur_rx_pos += read_bytes;
	}

	return read_words;
}

static void
tegra_qspi_copy_client_txbuf_to_qspi_txbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
{
	/*
	 * In packed mode, each word in FIFO may contain multiple packets
	 * based on bits per word. So all bytes in each FIFO word are valid.
	 *
	 * In unpacked mode, each word in FIFO contains single packet and
	 * based on bits per word any remaining bits in FIFO word will be
	 * ignored by the hardware and are invalid bits.
	 */
	if (tqspi->is_packed) {
		tqspi->cur_tx_pos += tqspi->curr_dma_words * tqspi->bytes_per_word;
	} else {
		u8 *tx_buf = (u8 *)t->tx_buf + tqspi->cur_tx_pos;
		unsigned int i, count, consume, write_bytes;

		/*
		 * Fill tx_dma_buf to contain single packet in each word based
		 * on bits per word from SPI core tx_buf.
		 */
		consume = tqspi->curr_dma_words * tqspi->bytes_per_word;
		if (consume > t->len - tqspi->cur_pos)
			consume = t->len - tqspi->cur_pos;
		write_bytes = consume;
		for (count = 0; count < tqspi->curr_dma_words; count++) {
			u32 x = 0;

			for (i = 0; consume && (i < tqspi->bytes_per_word); i++, consume--)
				x |= (u32)(*tx_buf++) << (i * 8);
			tqspi->tx_dma_buf[count] = x;
		}

		tqspi->cur_tx_pos += write_bytes;
	}
}

static void
tegra_qspi_copy_qspi_rxbuf_to_client_rxbuf(struct tegra_qspi *tqspi, struct spi_transfer *t)
{
	if (tqspi->is_packed) {
		tqspi->cur_rx_pos += tqspi->curr_dma_words * tqspi->bytes_per_word;
	} else {
		unsigned char *rx_buf = t->rx_buf + tqspi->cur_rx_pos;
		u32 rx_mask = ((u32)1 << t->bits_per_word) - 1;
		unsigned int i, count, consume, read_bytes;

		/*
		 * Each FIFO word contains single data packet.
		 * Skip invalid bits in each FIFO word based on bits per word
		 * and align bytes while filling in SPI core rx_buf.
		 */
		consume = tqspi->curr_dma_words * tqspi->bytes_per_word;
		if (consume > t->len - tqspi->cur_pos)
			consume = t->len - tqspi->cur_pos;
		read_bytes = consume;
		for (count = 0; count < tqspi->curr_dma_words; count++) {
			u32 x = tqspi->rx_dma_buf[count] & rx_mask;

			for (i = 0; consume && (i < tqspi->bytes_per_word); i++, consume--)
				*rx_buf++ = (x >> (i * 8)) & 0xff;
		}

		tqspi->cur_rx_pos += read_bytes;
	}
}

static void tegra_qspi_dma_complete(void *args)
{
	struct completion *dma_complete = args;

	complete(dma_complete);
}

static int tegra_qspi_start_tx_dma(struct tegra_qspi *tqspi, struct spi_transfer *t, int len)
{
	dma_addr_t tx_dma_phys;

	reinit_completion(&tqspi->tx_dma_complete);

	if (tqspi->is_packed)
		tx_dma_phys = t->tx_dma;
	else
		tx_dma_phys = tqspi->tx_dma_phys;

	tqspi->tx_dma_desc = dmaengine_prep_slave_single(tqspi->tx_dma_chan, tx_dma_phys,
							 len, DMA_MEM_TO_DEV,
							 DMA_PREP_INTERRUPT |  DMA_CTRL_ACK);

	if (!tqspi->tx_dma_desc) {
		dev_err(tqspi->dev, "Unable to get TX descriptor\n");
		return -EIO;
	}

	tqspi->tx_dma_desc->callback = tegra_qspi_dma_complete;
	tqspi->tx_dma_desc->callback_param = &tqspi->tx_dma_complete;
	dmaengine_submit(tqspi->tx_dma_desc);
	dma_async_issue_pending(tqspi->tx_dma_chan);

	return 0;
}

static int tegra_qspi_start_rx_dma(struct tegra_qspi *tqspi, struct spi_transfer *t, int len)
{
	dma_addr_t rx_dma_phys;

	reinit_completion(&tqspi->rx_dma_complete);

	if (tqspi->is_packed)
		rx_dma_phys = t->rx_dma;
	else
		rx_dma_phys = tqspi->rx_dma_phys;

	tqspi->rx_dma_desc = dmaengine_prep_slave_single(tqspi->rx_dma_chan, rx_dma_phys,
							 len, DMA_DEV_TO_MEM,
							 DMA_PREP_INTERRUPT |  DMA_CTRL_ACK);

	if (!tqspi->rx_dma_desc) {
		dev_err(tqspi->dev, "Unable to get RX descriptor\n");
		return -EIO;
	}

	tqspi->rx_dma_desc->callback = tegra_qspi_dma_complete;
	tqspi->rx_dma_desc->callback_param = &tqspi->rx_dma_complete;
	dmaengine_submit(tqspi->rx_dma_desc);
	dma_async_issue_pending(tqspi->rx_dma_chan);

	return 0;
}

static int tegra_qspi_flush_fifos(struct tegra_qspi *tqspi, bool atomic)
{
	void __iomem *addr = tqspi->base + QSPI_FIFO_STATUS;
	u32 val;

	val = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
	if ((val & QSPI_FIFO_EMPTY) == QSPI_FIFO_EMPTY)
		return 0;

	val |= QSPI_RX_FIFO_FLUSH | QSPI_TX_FIFO_FLUSH;
	tegra_qspi_writel(tqspi, val, QSPI_FIFO_STATUS);

	if (!atomic)
		return readl_relaxed_poll_timeout(addr, val,
						  (val & QSPI_FIFO_EMPTY) == QSPI_FIFO_EMPTY,
						  1000, 1000000);

	return readl_relaxed_poll_timeout_atomic(addr, val,
						 (val & QSPI_FIFO_EMPTY) == QSPI_FIFO_EMPTY,
						 1000, 1000000);
}

static void tegra_qspi_unmask_irq(struct tegra_qspi *tqspi)
{
	u32 intr_mask;

	intr_mask = tegra_qspi_readl(tqspi, QSPI_INTR_MASK);
	intr_mask &= ~(QSPI_INTR_RDY_MASK | QSPI_INTR_RX_TX_FIFO_ERR);
	tegra_qspi_writel(tqspi, intr_mask, QSPI_INTR_MASK);
}

static int tegra_qspi_dma_map_xfer(struct tegra_qspi *tqspi, struct spi_transfer *t)
{
	u8 *tx_buf = (u8 *)t->tx_buf + tqspi->cur_tx_pos;
	u8 *rx_buf = (u8 *)t->rx_buf + tqspi->cur_rx_pos;
	unsigned int len;

	len = DIV_ROUND_UP(tqspi->curr_dma_words * tqspi->bytes_per_word, 4) * 4;

	if (t->tx_buf) {
		t->tx_dma = dma_map_single(tqspi->tx_dma_dev, (void *)tx_buf, len, DMA_TO_DEVICE);
		if (dma_mapping_error(tqspi->tx_dma_dev, t->tx_dma))
			return -ENOMEM;
	}

	if (t->rx_buf) {
		t->rx_dma = dma_map_single(tqspi->rx_dma_dev, (void *)rx_buf, len, DMA_FROM_DEVICE);
		if (dma_mapping_error(tqspi->rx_dma_dev, t->rx_dma)) {
			if (t->tx_buf)
				dma_unmap_single(tqspi->tx_dma_dev, t->tx_dma,
						 len, DMA_TO_DEVICE);
			return -ENOMEM;
		}
	}

	tqspi->dma_map_len = len;
	tqspi->is_dma_mapped = true;

	return 0;
}

static void tegra_qspi_dma_unmap_xfer(struct tegra_qspi *tqspi, struct spi_transfer *t)
{
	if (!tqspi->is_dma_mapped)
		return;

	if (t->tx_buf)
		dma_unmap_single(tqspi->tx_dma_dev, t->tx_dma,
				 tqspi->dma_map_len, DMA_TO_DEVICE);
	if (t->rx_buf)
		dma_unmap_single(tqspi->rx_dma_dev, t->rx_dma,
				 tqspi->dma_map_len, DMA_FROM_DEVICE);

	tqspi->is_dma_mapped = false;
}

static int tegra_qspi_start_dma_based_transfer(struct tegra_qspi *tqspi, struct spi_transfer *t)
{
	struct dma_slave_config dma_sconfig = { 0 };
	dma_addr_t rx_dma_phys, tx_dma_phys;
	unsigned int len;
	u8 dma_burst;
	int ret = 0;
	u32 val;

	if (tqspi->is_packed) {
		ret = tegra_qspi_dma_map_xfer(tqspi, t);
		if (ret < 0)
			return ret;
	}

	val = QSPI_DMA_BLK_SET(tqspi->curr_dma_words - 1);
	tegra_qspi_writel(tqspi, val, QSPI_DMA_BLK);

	/*
	 * Reset the cached transfer status before unmasking the IRQ for
	 * this chunk. The cache must represent only the IRQ for THIS
	 * chunk; a stale RDY from the previous chunk of a multi-chunk
	 * transfer would otherwise mislead tegra_qspi_handle_timeout()
	 * into a false-positive recovery while the new chunk is still in
	 * flight. Pairs with smp_load_acquire() in
	 * tegra_qspi_handle_timeout(). The new chunk's IRQ cannot fire
	 * until QSPI_DMA_CTL is written below.
	 */
	smp_store_release(&tqspi->trans_status, 0);
	tegra_qspi_unmask_irq(tqspi);

	if (tqspi->is_packed)
		len = DIV_ROUND_UP(tqspi->curr_dma_words * tqspi->bytes_per_word, 4) * 4;
	else
		len = tqspi->curr_dma_words * 4;

	/* set attention level based on length of transfer */
	if (tqspi->soc_data->has_ext_dma) {
		val = 0;
		if (len & 0xf) {
			val |= QSPI_TX_TRIG_1 | QSPI_RX_TRIG_1;
			dma_burst = 1;
		} else if (((len) >> 4) & 0x1) {
			val |= QSPI_TX_TRIG_4 | QSPI_RX_TRIG_4;
			dma_burst = 4;
		} else {
			val |= QSPI_TX_TRIG_8 | QSPI_RX_TRIG_8;
			dma_burst = 8;
		}

		tegra_qspi_writel(tqspi, val, QSPI_DMA_CTL);
	}

	tqspi->dma_control_reg = val;

	dma_sconfig.device_fc = true;

	if (tqspi->cur_direction & DATA_DIR_TX) {
		if (tqspi->tx_dma_chan) {
			dma_sconfig.dst_addr = tqspi->phys + QSPI_TX_FIFO;
			dma_sconfig.dst_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
			dma_sconfig.dst_maxburst = dma_burst;
			ret = dmaengine_slave_config(tqspi->tx_dma_chan, &dma_sconfig);
			if (ret < 0) {
				dev_err(tqspi->dev, "failed DMA slave config: %d\n", ret);
				return ret;
			}

			tegra_qspi_copy_client_txbuf_to_qspi_txbuf(tqspi, t);
			ret = tegra_qspi_start_tx_dma(tqspi, t, len);
			if (ret < 0) {
				dev_err(tqspi->dev, "failed to starting TX DMA: %d\n", ret);
				return ret;
			}
		} else {
			if (tqspi->is_packed)
				tx_dma_phys = t->tx_dma;
			else
				tx_dma_phys = tqspi->tx_dma_phys;
			tegra_qspi_copy_client_txbuf_to_qspi_txbuf(tqspi, t);
			tegra_qspi_writel(tqspi, lower_32_bits(tx_dma_phys),
					  QSPI_DMA_MEM_ADDRESS);
			tegra_qspi_writel(tqspi, (upper_32_bits(tx_dma_phys) & 0xff),
					  QSPI_DMA_HI_ADDRESS);
		}
	}

	if (tqspi->cur_direction & DATA_DIR_RX) {
		if (tqspi->rx_dma_chan) {
			dma_sconfig.src_addr = tqspi->phys + QSPI_RX_FIFO;
			dma_sconfig.src_addr_width = DMA_SLAVE_BUSWIDTH_4_BYTES;
			dma_sconfig.src_maxburst = dma_burst;
			ret = dmaengine_slave_config(tqspi->rx_dma_chan, &dma_sconfig);
			if (ret < 0) {
				dev_err(tqspi->dev, "failed DMA slave config: %d\n", ret);
				return ret;
			}

			ret = tegra_qspi_start_rx_dma(tqspi, t, len);
			if (ret < 0) {
				dev_err(tqspi->dev, "failed to start RX DMA: %d\n", ret);
				if (tqspi->cur_direction & DATA_DIR_TX)
					dmaengine_terminate_all(tqspi->tx_dma_chan);
				return ret;
			}
		} else {
			if (tqspi->is_packed)
				rx_dma_phys = t->rx_dma;
			else
				rx_dma_phys = tqspi->rx_dma_phys;

			tegra_qspi_writel(tqspi, lower_32_bits(rx_dma_phys),
					  QSPI_DMA_MEM_ADDRESS);
			tegra_qspi_writel(tqspi, (upper_32_bits(rx_dma_phys) & 0xff),
					  QSPI_DMA_HI_ADDRESS);
		}
	}

	tegra_qspi_writel(tqspi, tqspi->command1_reg, QSPI_COMMAND1);

	/*
	 * WRITE_ONCE() pairs with READ_ONCE() in tegra_qspi_isr() and
	 * tegra_qspi_work_handler(); the flag is read lock-free across
	 * the hard-IRQ / process-context boundary so the annotation
	 * prevents compiler tearing and silences KCSAN.
	 */
	WRITE_ONCE(tqspi->is_curr_dma_xfer, true);
	tqspi->dma_control_reg = val;
	val |= QSPI_DMA_EN;
	tegra_qspi_writel(tqspi, val, QSPI_DMA_CTL);

	return ret;
}

static int tegra_qspi_start_cpu_based_transfer(struct tegra_qspi *qspi, struct spi_transfer *t)
{
	u32 val;
	unsigned int cur_words;

	if (qspi->cur_direction & DATA_DIR_TX)
		cur_words = tegra_qspi_fill_tx_fifo_from_client_txbuf(qspi, t);
	else
		cur_words = qspi->curr_dma_words;

	val = QSPI_DMA_BLK_SET(cur_words - 1);
	tegra_qspi_writel(qspi, val, QSPI_DMA_BLK);

	/*
	 * Snapshot whether this PIO chunk completes the whole transfer
	 * before unmasking the IRQ, so the hard IRQ small-PIO fastpath
	 * can decide whether to drain inline without dereferencing the
	 * spi_transfer object. cur_pos / curr_dma_words / bytes_per_word
	 * are stable here: they are written by
	 * tegra_qspi_calculate_curr_xfer_param() earlier in this code
	 * path. The IRQ cannot fire until the QSPI_COMMAND1 write below
	 * kicks the transfer off, so this store happens-before any ISR
	 * that observes the unmask.
	 */
	WRITE_ONCE(qspi->is_last_pio_chunk,
		   qspi->cur_pos + qspi->curr_dma_words * qspi->bytes_per_word >= t->len);

	/*
	 * Reset the cached transfer status before unmasking the IRQ for
	 * this chunk so the cache represents only the IRQ for THIS chunk;
	 * a stale RDY from the previous chunk would otherwise mislead
	 * tegra_qspi_handle_timeout() into a false-positive recovery
	 * while the new chunk is still in flight. Pairs with
	 * smp_load_acquire() in tegra_qspi_handle_timeout(). The new
	 * chunk's IRQ cannot fire until QSPI_COMMAND1 is written below.
	 */
	smp_store_release(&qspi->trans_status, 0);
	tegra_qspi_unmask_irq(qspi);

	WRITE_ONCE(qspi->is_curr_dma_xfer, false);
	val = qspi->command1_reg;
	val |= QSPI_PIO;
	tegra_qspi_writel(qspi, val, QSPI_COMMAND1);

	return 0;
}

static void tegra_qspi_deinit_dma(struct tegra_qspi *tqspi)
{
	if (tqspi->tx_dma_buf) {
		dma_free_coherent(tqspi->tx_dma_dev, tqspi->dma_buf_size,
				  tqspi->tx_dma_buf, tqspi->tx_dma_phys);
		tqspi->tx_dma_buf = NULL;
	}

	if (tqspi->tx_dma_chan) {
		dma_release_channel(tqspi->tx_dma_chan);
		tqspi->tx_dma_chan = NULL;
	}

	if (tqspi->rx_dma_buf) {
		dma_free_coherent(tqspi->rx_dma_dev, tqspi->dma_buf_size,
				  tqspi->rx_dma_buf, tqspi->rx_dma_phys);
		tqspi->rx_dma_buf = NULL;
	}

	if (tqspi->rx_dma_chan) {
		dma_release_channel(tqspi->rx_dma_chan);
		tqspi->rx_dma_chan = NULL;
	}
}

static int tegra_qspi_init_dma(struct tegra_qspi *tqspi)
{
	struct dma_chan *dma_chan;
	dma_addr_t dma_phys;
	u32 *dma_buf;
	int err;

	if (tqspi->soc_data->has_ext_dma) {
		dma_chan = dma_request_chan(tqspi->dev, "rx");
		if (IS_ERR(dma_chan)) {
			err = PTR_ERR(dma_chan);
			goto err_out;
		}

		tqspi->rx_dma_chan = dma_chan;
		tqspi->rx_dma_dev = dmaengine_get_dma_device(tqspi->rx_dma_chan);

		dma_chan = dma_request_chan(tqspi->dev, "tx");
		if (IS_ERR(dma_chan)) {
			err = PTR_ERR(dma_chan);
			goto err_out;
		}

		tqspi->tx_dma_chan = dma_chan;
		tqspi->tx_dma_dev = dmaengine_get_dma_device(tqspi->tx_dma_chan);
	} else {
		if (!device_iommu_mapped(tqspi->dev)) {
			dev_warn(tqspi->dev,
				 "IOMMU not enabled in device-tree, falling back to PIO mode\n");
			return 0;
		}

		tqspi->rx_dma_dev = tqspi->dev;
		tqspi->tx_dma_dev = tqspi->dev;
	}

	dma_buf = dma_alloc_coherent(tqspi->rx_dma_dev, tqspi->dma_buf_size, &dma_phys, GFP_KERNEL);
	if (!dma_buf) {
		err = -ENOMEM;
		goto err_out;
	}

	tqspi->rx_dma_buf = dma_buf;
	tqspi->rx_dma_phys = dma_phys;

	dma_buf = dma_alloc_coherent(tqspi->tx_dma_dev, tqspi->dma_buf_size, &dma_phys, GFP_KERNEL);
	if (!dma_buf) {
		err = -ENOMEM;
		goto err_out;
	}

	tqspi->tx_dma_buf = dma_buf;
	tqspi->tx_dma_phys = dma_phys;
	tqspi->use_dma = true;

	return 0;

err_out:
	tegra_qspi_deinit_dma(tqspi);

	if (err != -EPROBE_DEFER) {
		dev_err(tqspi->dev, "cannot use DMA: %d\n", err);
		dev_err(tqspi->dev, "falling back to PIO\n");
		return 0;
	}

	return err;
}

static u32 tegra_qspi_setup_transfer_one(struct spi_device *spi, struct spi_transfer *t,
					 bool is_first_of_msg)
{
	struct tegra_qspi *tqspi = spi_controller_get_devdata(spi->controller);
	struct tegra_qspi_client_data *cdata = spi->controller_data;
	u32 command1, command2, speed = t->speed_hz;
	u8 bits_per_word = t->bits_per_word;
	u32 tx_tap = 0, rx_tap = 0;
	unsigned long flags;
	int req_mode;

	if (!has_acpi_companion(tqspi->dev) && speed != tqspi->cur_speed) {
		clk_set_rate(tqspi->clk, speed);
		tqspi->cur_speed = speed;
	}

	spin_lock_irqsave(&tqspi->lock, flags);
	tqspi->cur_pos = 0;
	tqspi->cur_rx_pos = 0;
	tqspi->cur_tx_pos = 0;
	tqspi->curr_xfer = t;
	/*
	 * Pairs with smp_load_acquire() in tegra_qspi_handle_timeout().
	 * Clearing the cached trans_status before unmasking the IRQ for
	 * the new transfer prevents a stale RDY bit from the previous
	 * transfer fooling the timeout handler into a false recovery.
	 */
	smp_store_release(&tqspi->trans_status, 0);
	spin_unlock_irqrestore(&tqspi->lock, flags);

	if (is_first_of_msg) {
		tegra_qspi_mask_clear_irq(tqspi);

		command1 = tqspi->def_command1_reg;
		command1 |= QSPI_CS_SEL(spi_get_chipselect(spi, 0));
		command1 |= QSPI_BIT_LENGTH(bits_per_word - 1);

		command1 &= ~QSPI_CONTROL_MODE_MASK;
		req_mode = spi->mode & 0x3;
		if (req_mode == SPI_MODE_3)
			command1 |= QSPI_CONTROL_MODE_3;
		else
			command1 |= QSPI_CONTROL_MODE_0;

		if (spi->mode & SPI_CS_HIGH)
			command1 |= QSPI_CS_SW_VAL;
		else
			command1 &= ~QSPI_CS_SW_VAL;
		tegra_qspi_writel(tqspi, command1, QSPI_COMMAND1);

		if (cdata && cdata->tx_clk_tap_delay)
			tx_tap = cdata->tx_clk_tap_delay;

		if (cdata && cdata->rx_clk_tap_delay)
			rx_tap = cdata->rx_clk_tap_delay;

		command2 = QSPI_TX_TAP_DELAY(tx_tap) | QSPI_RX_TAP_DELAY(rx_tap);
		if (command2 != tqspi->def_command2_reg)
			tegra_qspi_writel(tqspi, command2, QSPI_COMMAND2);

	} else {
		command1 = tqspi->command1_reg;
		command1 &= ~QSPI_BIT_LENGTH(~0);
		command1 |= QSPI_BIT_LENGTH(bits_per_word - 1);
	}

	command1 &= ~QSPI_SDR_DDR_SEL;

	return command1;
}

static int tegra_qspi_start_transfer_one(struct spi_device *spi,
					 struct spi_transfer *t, u32 command1)
{
	struct tegra_qspi *tqspi = spi_controller_get_devdata(spi->controller);
	unsigned int total_fifo_words;
	u8 bus_width = 0;
	int ret;

	total_fifo_words = tegra_qspi_calculate_curr_xfer_param(tqspi, t);

	command1 &= ~QSPI_PACKED;
	if (tqspi->is_packed)
		command1 |= QSPI_PACKED;
	tegra_qspi_writel(tqspi, command1, QSPI_COMMAND1);

	tqspi->cur_direction = 0;

	command1 &= ~(QSPI_TX_EN | QSPI_RX_EN);
	if (t->rx_buf) {
		command1 |= QSPI_RX_EN;
		tqspi->cur_direction |= DATA_DIR_RX;
		bus_width = t->rx_nbits;
	}

	if (t->tx_buf) {
		command1 |= QSPI_TX_EN;
		tqspi->cur_direction |= DATA_DIR_TX;
		bus_width = t->tx_nbits;
	}

	command1 &= ~QSPI_INTERFACE_WIDTH_MASK;

	if (bus_width == SPI_NBITS_QUAD)
		command1 |= QSPI_INTERFACE_WIDTH_QUAD;
	else if (bus_width == SPI_NBITS_DUAL)
		command1 |= QSPI_INTERFACE_WIDTH_DUAL;
	else
		command1 |= QSPI_INTERFACE_WIDTH_SINGLE;

	tqspi->command1_reg = command1;

	tegra_qspi_writel(tqspi, QSPI_NUM_DUMMY_CYCLE(tqspi->dummy_cycles), QSPI_MISC_REG);

	ret = tegra_qspi_flush_fifos(tqspi, false);
	if (ret < 0)
		return ret;

	if (tqspi->use_dma && total_fifo_words > QSPI_FIFO_DEPTH)
		ret = tegra_qspi_start_dma_based_transfer(tqspi, t);
	else
		ret = tegra_qspi_start_cpu_based_transfer(tqspi, t);

	return ret;
}

static struct tegra_qspi_client_data *tegra_qspi_parse_cdata_dt(struct spi_device *spi)
{
	struct tegra_qspi_client_data *cdata;
	struct tegra_qspi *tqspi = spi_controller_get_devdata(spi->controller);

	cdata = devm_kzalloc_obj(tqspi->dev, *cdata);
	if (!cdata)
		return NULL;

	device_property_read_u32(&spi->dev, "nvidia,tx-clk-tap-delay",
				 &cdata->tx_clk_tap_delay);
	device_property_read_u32(&spi->dev, "nvidia,rx-clk-tap-delay",
				 &cdata->rx_clk_tap_delay);

	return cdata;
}

static int tegra_qspi_setup(struct spi_device *spi)
{
	struct tegra_qspi *tqspi = spi_controller_get_devdata(spi->controller);
	struct tegra_qspi_client_data *cdata = spi->controller_data;
	unsigned long flags;
	u32 val;
	int ret;

	ret = pm_runtime_resume_and_get(tqspi->dev);
	if (ret < 0) {
		dev_err(tqspi->dev, "failed to get runtime PM: %d\n", ret);
		return ret;
	}

	if (!cdata) {
		cdata = tegra_qspi_parse_cdata_dt(spi);
		spi->controller_data = cdata;
	}
	spin_lock_irqsave(&tqspi->lock, flags);

	/* keep default cs state to inactive */
	val = tqspi->def_command1_reg;
	val |= QSPI_CS_SEL(spi_get_chipselect(spi, 0));
	if (spi->mode & SPI_CS_HIGH)
		val &= ~QSPI_CS_POL_INACTIVE(spi_get_chipselect(spi, 0));
	else
		val |= QSPI_CS_POL_INACTIVE(spi_get_chipselect(spi, 0));

	tqspi->def_command1_reg = val;
	tegra_qspi_writel(tqspi, tqspi->def_command1_reg, QSPI_COMMAND1);

	spin_unlock_irqrestore(&tqspi->lock, flags);

	pm_runtime_put_autosuspend(tqspi->dev);

	return 0;
}

static void tegra_qspi_dump_regs(struct tegra_qspi *tqspi)
{
	dev_dbg(tqspi->dev, "============ QSPI REGISTER DUMP ============\n");
	dev_dbg(tqspi->dev, "Command1:    0x%08x | Command2:    0x%08x\n",
		tegra_qspi_readl(tqspi, QSPI_COMMAND1),
		tegra_qspi_readl(tqspi, QSPI_COMMAND2));
	dev_dbg(tqspi->dev, "DMA_CTL:     0x%08x | DMA_BLK:     0x%08x\n",
		tegra_qspi_readl(tqspi, QSPI_DMA_CTL),
		tegra_qspi_readl(tqspi, QSPI_DMA_BLK));
	dev_dbg(tqspi->dev, "INTR_MASK:  0x%08x | MISC: 0x%08x\n",
		tegra_qspi_readl(tqspi, QSPI_INTR_MASK),
		tegra_qspi_readl(tqspi, QSPI_MISC_REG));
	dev_dbg(tqspi->dev, "TRANS_STAT:  0x%08x | FIFO_STATUS: 0x%08x\n",
		tegra_qspi_readl(tqspi, QSPI_TRANS_STATUS),
		tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS));
}

static void tegra_qspi_reset(struct tegra_qspi *tqspi)
{
	if (device_reset(tqspi->dev) < 0) {
		dev_warn_once(tqspi->dev, "device reset failed\n");
		tegra_qspi_mask_clear_irq(tqspi);
	}
}

static void tegra_qspi_handle_error(struct tegra_qspi *tqspi)
{
	dev_err(tqspi->dev, "error in transfer, fifo status 0x%08x\n", tqspi->status_reg);
	tegra_qspi_dump_regs(tqspi);
	tegra_qspi_flush_fifos(tqspi, true);
	tegra_qspi_reset(tqspi);
}

static void tegra_qspi_transfer_end(struct spi_device *spi)
{
	struct tegra_qspi *tqspi = spi_controller_get_devdata(spi->controller);
	int cs_val = (spi->mode & SPI_CS_HIGH) ? 0 : 1;

	if (cs_val)
		tqspi->command1_reg |= QSPI_CS_SW_VAL;
	else
		tqspi->command1_reg &= ~QSPI_CS_SW_VAL;
	tegra_qspi_writel(tqspi, tqspi->command1_reg, QSPI_COMMAND1);
	tegra_qspi_writel(tqspi, tqspi->def_command1_reg, QSPI_COMMAND1);
}

static irqreturn_t handle_cpu_based_xfer(struct tegra_qspi *tqspi);
static irqreturn_t handle_dma_based_xfer(struct tegra_qspi *tqspi);

/**
 * tegra_qspi_handle_timeout - Handle transfer timeout with hardware check
 * @tqspi: QSPI controller instance
 *
 * When wait_for_completion_timeout() expires the hardware may still have
 * finished the current chunk. Drain the pending bottom half and, if the
 * whole transfer really did complete during the drain, consume the
 * completion and report success.
 *
 * When the bottom half advanced the transfer by only one chunk of a
 * multi-chunk DMA/PIO transfer without signalling xfer_completion, a
 * fallback that ran handle_{cpu,dma}_based_xfer() here would race with
 * the DMA engine already moving the next chunk into the client buffer
 * (spi_finalize_current_message() would then release the buffer while
 * the controller is still writing memory). Fake completion is therefore
 * only attempted when the current chunk is the last chunk of the
 * transfer; multi-chunk continuation timeouts return -ETIMEDOUT and
 * let the caller reset the controller.
 *
 * Returns: 0 if the transfer completed, -ETIMEDOUT otherwise.
 */
static int tegra_qspi_handle_timeout(struct tegra_qspi *tqspi)
{
	struct spi_transfer *t;
	unsigned long flags;
	bool is_last_chunk;
	bool lost_irq_snapshot = false;
	irqreturn_t ret;
	int retval;
	u32 status, refreshed;
	u32 lost_fifo_status = 0;
	u32 lost_tx_status = 0;
	u32 lost_rx_status = 0;

	/*
	 * Snapshot both the ISR cache and (if the cache is empty) the
	 * live status registers BEFORE entering recovery. The recovery
	 * path calls tegra_qspi_mask_clear_irq() below, which performs
	 * W1Cs on QSPI_TRANS_STATUS and on the QSPI_FIFO_STATUS error
	 * bits: a lost-IRQ recovery must capture the current FIFO error
	 * state before the mask erases it.
	 *
	 * Cache-live-cache retry: if the initial cache load returns zero
	 * we fall back to a live QSPI_TRANS_STATUS read, and if that also
	 * returns zero we retry the cache once more. That closes the
	 * interleaving where an ISR on another CPU publishes trans_status
	 * with release semantics and then W1Cs the hardware between our
	 * cache load and our live load: the second cache load observes
	 * the now-visible release and we correctly classify the transfer
	 * as complete rather than reporting a false timeout.
	 *
	 * The trans_status cache is reset to zero in
	 * tegra_qspi_start_{cpu,dma}_based_transfer() before unmasking
	 * the IRQ for every chunk, so a stale RDY from the previous
	 * chunk of a multi-chunk transfer cannot survive into this
	 * check.
	 */
	status = smp_load_acquire(&tqspi->trans_status);
	if (!status) {
		status = tegra_qspi_readl(tqspi, QSPI_TRANS_STATUS);
		if (!status) {
			/* Retry cache; pairs with release in ISR post-store. */
			status = smp_load_acquire(&tqspi->trans_status);
		} else {
			/*
			 * Live register shows RDY but the ISR cache is
			 * empty: either the ISR ran and cleared HW between
			 * our two loads (the cache retry above would have
			 * observed it, so we would not be here), or the IRQ
			 * was genuinely lost. Snapshot the live FIFO error
			 * status now so tegra_qspi_mask_clear_irq() below
			 * does not W1C it away before the manual handler
			 * downstream can see it.
			 */
			lost_fifo_status = tegra_qspi_readl(tqspi,
							    QSPI_FIFO_STATUS);
			lost_tx_status = lost_fifo_status &
					 (QSPI_TX_FIFO_UNF | QSPI_TX_FIFO_OVF);
			lost_rx_status = lost_fifo_status &
					 (QSPI_RX_FIFO_OVF | QSPI_RX_FIFO_UNF);
			lost_irq_snapshot = true;
		}
	}

	/*
	 * Enter recovery unconditionally. Every expired
	 * wait_for_completion_timeout() must serialise against a delayed
	 * ISR or worker before the caller runs dma_stop() +
	 * device_reset() + curr_xfer clear: publishing
	 * recovery_in_progress under tqspi->lock, masking the controller
	 * IRQ, calling synchronize_irq() to drain any in-flight ISR
	 * (including the small-PIO hard-IRQ fastpath), and finally
	 * cancel_work_sync() to drain the workqueue gives us that
	 * serialisation regardless of whether the hardware finished. A
	 * genuine hardware timeout still ends up as -ETIMEDOUT further
	 * down, but only after ISR and workqueue activity are quiesced.
	 *
	 * cancel_work_sync() cancels a pending worker without executing
	 * it and waits for a currently running one to finish; the
	 * recovery_in_progress guard checked inside tegra_qspi_isr()
	 * under tqspi->lock is atomic with its queue_work() and small-PIO
	 * fastpath dispatch decisions, so no new bottom-half work is
	 * enqueued once we publish the flag.
	 *
	 * tegra_qspi_mask_clear_irq() is idempotent: its read-modify-write
	 * of QSPI_INTR_MASK and W1C of QSPI_TRANS_STATUS / FIFO error
	 * status all tolerate a double-write, so it is safe whether or
	 * not the ISR has already run for this transfer.
	 */
	spin_lock_irqsave(&tqspi->lock, flags);
	WRITE_ONCE(tqspi->recovery_in_progress, true);
	spin_unlock_irqrestore(&tqspi->lock, flags);

	tegra_qspi_mask_clear_irq(tqspi);
	synchronize_irq(tqspi->irq);
	cancel_work_sync(&tqspi->irq_work);

	if (try_wait_for_completion(&tqspi->xfer_completion)) {
		retval = 0;
		goto out;
	}

	/*
	 * Re-check the cache after the drain: the worker we just drained
	 * may have published a completion status the entry snapshot did
	 * not observe (for example the ISR fired on another CPU after we
	 * loaded the cache but before we masked).
	 */
	refreshed = smp_load_acquire(&tqspi->trans_status);
	if (refreshed) {
		status = refreshed;
		/*
		 * The drain observed a real ISR completion. Prefer the
		 * ISR-published status_reg / tx_status / rx_status over
		 * the pre-drain lost-IRQ snapshot, which may have been
		 * taken after the ISR W1C-ed the hardware error bits and
		 * therefore holds cleared values.
		 */
		lost_irq_snapshot = false;
	}

	if (!(status & QSPI_RDY)) {
		retval = -ETIMEDOUT;
		goto out;
	}

	/*
	 * If the ISR never ran (lost IRQ path) publish the FIFO error
	 * snapshot we captured before mask_clear_irq() so the manual
	 * handler downstream has fresh error state rather than stale
	 * fields from a previous chunk's ISR.
	 */
	if (lost_irq_snapshot) {
		WRITE_ONCE(tqspi->status_reg, lost_fifo_status);
		WRITE_ONCE(tqspi->tx_status, lost_tx_status);
		WRITE_ONCE(tqspi->rx_status, lost_rx_status);
	}

	/*
	 * The bottom half did not signal full completion. Either the work
	 * ran and advanced the transfer by one chunk (possibly arming the
	 * next chunk of a multi-chunk transfer), or it was cancelled
	 * before it could run, or the current chunk really did not
	 * complete. Only fake completion when the current chunk is the
	 * last chunk of the transfer; otherwise the DMA engine may still
	 * be moving the next chunk into memory, and returning 0 here would
	 * let seq_xfer clear curr_xfer and finalise the message while the
	 * hardware is still writing.
	 *
	 * The last-chunk arithmetic mirrors tegra_qspi_start_cpu_based_
	 * transfer(), which uses cur_pos + curr_dma_words * bytes_per_word
	 * >= t->len to set is_last_pio_chunk before arming the IRQ.
	 */
	spin_lock_irqsave(&tqspi->lock, flags);
	t = tqspi->curr_xfer;
	if (!t) {
		/* CPU-path handler already cleared curr_xfer */
		spin_unlock_irqrestore(&tqspi->lock, flags);
		retval = 0;
		goto out;
	}
	is_last_chunk = (tqspi->cur_pos +
			 tqspi->curr_dma_words * tqspi->bytes_per_word) >= t->len;
	spin_unlock_irqrestore(&tqspi->lock, flags);

	if (!is_last_chunk) {
		retval = -ETIMEDOUT;
		goto out;
	}

	dev_warn_ratelimited(tqspi->dev,
			     "QSPI interrupt timeout, but transfer complete\n");

	if (!READ_ONCE(tqspi->is_curr_dma_xfer))
		ret = handle_cpu_based_xfer(tqspi);
	else
		ret = handle_dma_based_xfer(tqspi);

	retval = (ret == IRQ_HANDLED) ? 0 : -EIO;

out:
	/*
	 * The drained bottom half may have unmasked the controller IRQ
	 * to arm the next chunk of a multi-chunk transfer. Re-mask and
	 * synchronize before clearing recovery_in_progress so that no
	 * lingering ISR can queue fresh work behind the caller's back
	 * (the caller's dma_stop() + device_reset() + curr_xfer clear
	 * runs immediately after we return on the error path).
	 */
	tegra_qspi_mask_clear_irq(tqspi);
	synchronize_irq(tqspi->irq);
	WRITE_ONCE(tqspi->recovery_in_progress, false);
	return retval;
}

static u32 tegra_qspi_cmd_config(bool is_ddr, u8 bus_width, u8 len)
{
	u32 cmd_config = 0;

	/* Extract Command configuration and value */
	if (is_ddr)
		cmd_config |= QSPI_COMMAND_SDR_DDR;
	else
		cmd_config &= ~QSPI_COMMAND_SDR_DDR;

	cmd_config |= QSPI_COMMAND_X1_X2_X4(bus_width);
	cmd_config |= QSPI_COMMAND_SIZE_SET((len * 8) - 1);

	return cmd_config;
}

static u32 tegra_qspi_addr_config(bool is_ddr, u8 bus_width, u8 len)
{
	u32 addr_config = 0;

	if (is_ddr)
		addr_config |= QSPI_ADDRESS_SDR_DDR;
	else
		addr_config &= ~QSPI_ADDRESS_SDR_DDR;

	addr_config |= QSPI_ADDRESS_X1_X2_X4(bus_width);
	addr_config |= QSPI_ADDRESS_SIZE_SET((len * 8) - 1);

	return addr_config;
}

static void tegra_qspi_dma_stop(struct tegra_qspi *tqspi)
{
	u32 value;

	if ((tqspi->cur_direction & DATA_DIR_TX) && tqspi->tx_dma_chan)
		dmaengine_terminate_all(tqspi->tx_dma_chan);

	if ((tqspi->cur_direction & DATA_DIR_RX) && tqspi->rx_dma_chan)
		dmaengine_terminate_all(tqspi->rx_dma_chan);

	value = tegra_qspi_readl(tqspi, QSPI_DMA_CTL);
	value &= ~QSPI_DMA_EN;
	tegra_qspi_writel(tqspi, value, QSPI_DMA_CTL);
}

static void tegra_qspi_pio_stop(struct tegra_qspi *tqspi)
{
	u32 value;

	value = tegra_qspi_readl(tqspi, QSPI_COMMAND1);
	value &= ~QSPI_PIO;
	tegra_qspi_writel(tqspi, value, QSPI_COMMAND1);
}

static int tegra_qspi_combined_seq_xfer(struct tegra_qspi *tqspi,
					struct spi_message *msg)
{
	bool is_first_msg = true;
	struct spi_transfer *xfer;
	struct spi_device *spi = msg->spi;
	u8 transfer_phase = 0;
	u32 cmd1 = 0;
	int ret = 0;
	u32 address_value = 0;
	u32 cmd_config = 0, addr_config = 0;
	u8 cmd_value = 0, val = 0;
	unsigned long flags;

	/* Enable Combined sequence mode */
	val = tegra_qspi_readl(tqspi, QSPI_GLOBAL_CONFIG);
	if (spi->mode & SPI_TPM_HW_FLOW) {
		if (tqspi->soc_data->supports_tpm)
			val |= QSPI_TPM_WAIT_POLL_EN;
		else
			return -EIO;
	}
	val |= QSPI_CMB_SEQ_EN;
	tegra_qspi_writel(tqspi, val, QSPI_GLOBAL_CONFIG);
	/* Process individual transfer list */
	list_for_each_entry(xfer, &msg->transfers, transfer_list) {
		switch (transfer_phase) {
		case CMD_TRANSFER:
			/* X1 SDR mode */
			cmd_config = tegra_qspi_cmd_config(false, xfer->tx_nbits,
							   xfer->len);
			cmd_value = *((const u8 *)(xfer->tx_buf));
			break;
		case ADDR_TRANSFER:
			/* X1 SDR mode */
			addr_config = tegra_qspi_addr_config(false, xfer->tx_nbits,
							     xfer->len);
			address_value = *((const u32 *)(xfer->tx_buf));
			break;
		case DUMMY_TRANSFER:
			if (xfer->dummy_data) {
				tqspi->dummy_cycles = xfer->len * 8 / xfer->tx_nbits;
				break;
			}
			transfer_phase++;
			fallthrough;
		case DATA_TRANSFER:
			/* Program Command, Address value in register */
			tegra_qspi_writel(tqspi, cmd_value, QSPI_CMB_SEQ_CMD);
			tegra_qspi_writel(tqspi, address_value,
					  QSPI_CMB_SEQ_ADDR);
			/* Program Command and Address config in register */
			tegra_qspi_writel(tqspi, cmd_config,
					  QSPI_CMB_SEQ_CMD_CFG);
			tegra_qspi_writel(tqspi, addr_config,
					  QSPI_CMB_SEQ_ADDR_CFG);

			reinit_completion(&tqspi->xfer_completion);
			cmd1 = tegra_qspi_setup_transfer_one(spi, xfer,
							     is_first_msg);
			ret = tegra_qspi_start_transfer_one(spi, xfer,
							    cmd1);

			if (ret < 0) {
				dev_err(tqspi->dev, "Failed to start transfer-one: %d\n",
					ret);
				return ret;
			}

			is_first_msg = false;
			ret = wait_for_completion_timeout
					(&tqspi->xfer_completion,
					QSPI_DMA_TIMEOUT);

			if (ret == 0) {
				/*
				 * Check if hardware completed the transfer even though
				 * workqueue was delayed. If so, process completion and
				 * continue.
				 */
				ret = tegra_qspi_handle_timeout(tqspi);
				if (ret < 0) {
					/* Real timeout - clean up and fail */
					WARN_ON_ONCE(1);
					dev_err(tqspi->dev, "transfer timeout\n");

					/* Abort transfer by resetting pio/dma bit */
					if (tqspi->is_curr_dma_xfer) {
						tegra_qspi_dma_stop(tqspi);
						tegra_qspi_dma_unmap_xfer(tqspi, xfer);
					} else {
						tegra_qspi_pio_stop(tqspi);
					}

					/* Reset controller if timeout happens */
					tegra_qspi_reset(tqspi);

					ret = -EIO;
					goto exit;
				}
			}

			if (tqspi->tx_status ||  tqspi->rx_status) {
				dev_err(tqspi->dev, "QSPI Transfer failed\n");
				tqspi->tx_status = 0;
				tqspi->rx_status = 0;
				ret = -EIO;
				goto exit;
			}
			break;
		default:
			ret = -EINVAL;
			goto exit;
		}
		msg->actual_length += xfer->len;
		if (!xfer->cs_change && transfer_phase == DATA_TRANSFER) {
			tegra_qspi_transfer_end(spi);
			spi_transfer_delay_exec(xfer);
		}
		spin_lock_irqsave(&tqspi->lock, flags);
		tqspi->curr_xfer = NULL;
		spin_unlock_irqrestore(&tqspi->lock, flags);
		transfer_phase++;
	}
	ret = 0;

exit:
	spin_lock_irqsave(&tqspi->lock, flags);
	tqspi->curr_xfer = NULL;
	spin_unlock_irqrestore(&tqspi->lock, flags);
	msg->status = ret;

	return ret;
}

static int tegra_qspi_non_combined_seq_xfer(struct tegra_qspi *tqspi,
					    struct spi_message *msg)
{
	struct spi_device *spi = msg->spi;
	struct spi_transfer *transfer;
	bool is_first_msg = true;
	int ret = 0, val = 0;
	unsigned long flags;

	msg->status = 0;
	msg->actual_length = 0;
	tqspi->tx_status = 0;
	tqspi->rx_status = 0;

	/* Disable Combined sequence mode */
	val = tegra_qspi_readl(tqspi, QSPI_GLOBAL_CONFIG);
	val &= ~QSPI_CMB_SEQ_EN;
	if (tqspi->soc_data->supports_tpm)
		val &= ~QSPI_TPM_WAIT_POLL_EN;
	tegra_qspi_writel(tqspi, val, QSPI_GLOBAL_CONFIG);
	list_for_each_entry(transfer, &msg->transfers, transfer_list) {
		struct spi_transfer *xfer = transfer;
		u8 dummy_bytes = 0;
		u32 cmd1;

		tqspi->dummy_cycles = 0;
		/*
		 * Tegra QSPI hardware supports dummy bytes transfer after actual transfer
		 * bytes based on programmed dummy clock cycles in the QSPI_MISC register.
		 * So, check if the next transfer is dummy data transfer and program dummy
		 * clock cycles along with the current transfer and skip next transfer.
		 */
		if (!list_is_last(&xfer->transfer_list, &msg->transfers)) {
			struct spi_transfer *next_xfer;

			next_xfer = list_next_entry(xfer, transfer_list);
			if (next_xfer->dummy_data) {
				u32 dummy_cycles = next_xfer->len * 8 / next_xfer->tx_nbits;

				if (dummy_cycles <= QSPI_DUMMY_CYCLES_MAX) {
					tqspi->dummy_cycles = dummy_cycles;
					dummy_bytes = next_xfer->len;
					transfer = next_xfer;
				}
			}
		}

		reinit_completion(&tqspi->xfer_completion);

		cmd1 = tegra_qspi_setup_transfer_one(spi, xfer, is_first_msg);

		ret = tegra_qspi_start_transfer_one(spi, xfer, cmd1);
		if (ret < 0) {
			dev_err(tqspi->dev, "failed to start transfer: %d\n", ret);
			goto complete_xfer;
		}

		ret = wait_for_completion_timeout(&tqspi->xfer_completion,
						  QSPI_DMA_TIMEOUT);
		if (ret == 0) {
			/*
			 * Check if hardware completed the transfer even though
			 * workqueue was delayed. If so, process completion and
			 * continue.
			 */
			ret = tegra_qspi_handle_timeout(tqspi);
			if (ret < 0) {
				/* Real timeout - clean up and fail */
				WARN_ON(1);
				dev_err(tqspi->dev, "transfer timeout\n");

				if (tqspi->is_curr_dma_xfer) {
					tegra_qspi_dma_stop(tqspi);
					tegra_qspi_dma_unmap_xfer(tqspi, xfer);
				}

				tegra_qspi_handle_error(tqspi);
				ret = -EIO;
				goto complete_xfer;
			}
		}

		if (tqspi->tx_status ||  tqspi->rx_status) {
			tegra_qspi_handle_error(tqspi);
			ret = -EIO;
			goto complete_xfer;
		}

		msg->actual_length += xfer->len + dummy_bytes;

complete_xfer:
		spin_lock_irqsave(&tqspi->lock, flags);
		tqspi->curr_xfer = NULL;
		spin_unlock_irqrestore(&tqspi->lock, flags);

		if (ret < 0) {
			tegra_qspi_transfer_end(spi);
			spi_transfer_delay_exec(xfer);
			goto exit;
		}

		if (list_is_last(&xfer->transfer_list, &msg->transfers)) {
			/* de-activate CS after last transfer only when cs_change is not set */
			if (!xfer->cs_change) {
				tegra_qspi_transfer_end(spi);
				spi_transfer_delay_exec(xfer);
			}
		} else if (xfer->cs_change) {
			 /* de-activated CS between the transfers only when cs_change is set */
			tegra_qspi_transfer_end(spi);
			spi_transfer_delay_exec(xfer);
		}
	}

	ret = 0;
exit:
	msg->status = ret;

	return ret;
}

static bool tegra_qspi_validate_cmb_seq(struct tegra_qspi *tqspi,
					struct spi_message *msg)
{
	int transfer_count = 0;
	struct spi_transfer *xfer;

	list_for_each_entry(xfer, &msg->transfers, transfer_list) {
		transfer_count++;
	}
	if (!tqspi->soc_data->cmb_xfer_capable)
		return false;
	if (transfer_count > 4 || transfer_count < 3)
		return false;
	xfer = list_first_entry(&msg->transfers, typeof(*xfer),
				transfer_list);
	if (xfer->len > 2)
		return false;
	xfer = list_next_entry(xfer, transfer_list);
	if (xfer->len > 4 || xfer->len < 3)
		return false;
	xfer = list_next_entry(xfer, transfer_list);
	if (transfer_count == 4) {
		if (xfer->dummy_data != 1)
			return false;
		if ((xfer->len * 8 / xfer->tx_nbits) > QSPI_DUMMY_CYCLES_MAX)
			return false;
		xfer = list_next_entry(xfer, transfer_list);
	}
	if (!tqspi->soc_data->has_ext_dma && xfer->len > (QSPI_FIFO_DEPTH << 2))
		return false;

	return true;
}

static int tegra_qspi_transfer_one_message(struct spi_controller *host,
					   struct spi_message *msg)
{
	struct tegra_qspi *tqspi = spi_controller_get_devdata(host);
	int ret;

	if (tegra_qspi_validate_cmb_seq(tqspi, msg))
		ret = tegra_qspi_combined_seq_xfer(tqspi, msg);
	else
		ret = tegra_qspi_non_combined_seq_xfer(tqspi, msg);

	spi_finalize_current_message(host);

	return ret;
}

static irqreturn_t handle_cpu_based_xfer(struct tegra_qspi *tqspi)
{
	struct spi_transfer *t;
	unsigned long flags;

	spin_lock_irqsave(&tqspi->lock, flags);
	t = tqspi->curr_xfer;

	if (!t) {
		spin_unlock_irqrestore(&tqspi->lock, flags);
		return IRQ_HANDLED;
	}

	if (tqspi->tx_status ||  tqspi->rx_status) {
		tegra_qspi_handle_error(tqspi);
		complete(&tqspi->xfer_completion);
		goto exit;
	}

	if (tqspi->cur_direction & DATA_DIR_RX)
		tegra_qspi_read_rx_fifo_to_client_rxbuf(tqspi, t);

	if (tqspi->cur_direction & DATA_DIR_TX)
		tqspi->cur_pos = tqspi->cur_tx_pos;
	else
		tqspi->cur_pos = tqspi->cur_rx_pos;

	if (tqspi->cur_pos == t->len) {
		complete(&tqspi->xfer_completion);
		goto exit;
	}

	tegra_qspi_calculate_curr_xfer_param(tqspi, t);
	tegra_qspi_start_cpu_based_transfer(tqspi, t);
exit:
	tqspi->curr_xfer = NULL;
	spin_unlock_irqrestore(&tqspi->lock, flags);
	return IRQ_HANDLED;
}

static irqreturn_t handle_dma_based_xfer(struct tegra_qspi *tqspi)
{
	struct spi_transfer *t;
	unsigned int total_fifo_words;
	unsigned long flags;
	long wait_status;
	int num_errors = 0;

	/*
	 * Snapshot curr_xfer under the lock before the (potentially long)
	 * DMA waits below. The timeout path can clear tqspi->curr_xfer
	 * concurrently; using the local copy keeps the subsequent dma_unmap
	 * and FIFO-drain steps consistent with the transfer that actually
	 * started, and lets us bail safely if cleanup already happened.
	 */
	spin_lock_irqsave(&tqspi->lock, flags);
	t = tqspi->curr_xfer;
	spin_unlock_irqrestore(&tqspi->lock, flags);
	if (!t)
		return IRQ_HANDLED;

	if (tqspi->cur_direction & DATA_DIR_TX) {
		if (tqspi->tx_status) {
			if (tqspi->tx_dma_chan)
				dmaengine_terminate_all(tqspi->tx_dma_chan);
			num_errors++;
		} else if (tqspi->tx_dma_chan) {
			wait_status = wait_for_completion_interruptible_timeout(
				&tqspi->tx_dma_complete, QSPI_DMA_TIMEOUT);
			if (wait_status <= 0) {
				dmaengine_terminate_all(tqspi->tx_dma_chan);
				dev_err(tqspi->dev, "failed TX DMA transfer\n");
				num_errors++;
			}
		}
	}

	if (tqspi->cur_direction & DATA_DIR_RX) {
		if (tqspi->rx_status) {
			if (tqspi->rx_dma_chan)
				dmaengine_terminate_all(tqspi->rx_dma_chan);
			num_errors++;
		} else if (tqspi->rx_dma_chan) {
			wait_status = wait_for_completion_interruptible_timeout(
				&tqspi->rx_dma_complete, QSPI_DMA_TIMEOUT);
			if (wait_status <= 0) {
				dmaengine_terminate_all(tqspi->rx_dma_chan);
				dev_err(tqspi->dev, "failed RX DMA transfer\n");
				num_errors++;
			}
		}
	}

	spin_lock_irqsave(&tqspi->lock, flags);

	if (num_errors) {
		tegra_qspi_dma_unmap_xfer(tqspi, t);
		tegra_qspi_handle_error(tqspi);
		complete(&tqspi->xfer_completion);
		goto exit;
	}

	if (tqspi->cur_direction & DATA_DIR_RX)
		tegra_qspi_copy_qspi_rxbuf_to_client_rxbuf(tqspi, t);

	if (tqspi->cur_direction & DATA_DIR_TX)
		tqspi->cur_pos = tqspi->cur_tx_pos;
	else
		tqspi->cur_pos = tqspi->cur_rx_pos;

	if (tqspi->cur_pos == t->len) {
		tegra_qspi_dma_unmap_xfer(tqspi, t);
		complete(&tqspi->xfer_completion);
		goto exit;
	}

	tegra_qspi_dma_unmap_xfer(tqspi, t);

	/* continue transfer in current message */
	total_fifo_words = tegra_qspi_calculate_curr_xfer_param(tqspi, t);
	if (total_fifo_words > QSPI_FIFO_DEPTH)
		num_errors = tegra_qspi_start_dma_based_transfer(tqspi, t);
	else
		num_errors = tegra_qspi_start_cpu_based_transfer(tqspi, t);

exit:
	spin_unlock_irqrestore(&tqspi->lock, flags);
	return IRQ_HANDLED;
}

/**
 * tegra_qspi_work_handler - Workqueue handler for interrupt bottom-half
 * @work: work_struct embedded in tegra_qspi
 *
 * Runs in process context and can sleep (needed for DMA completion waits).
 * Runs on any CPU in the WQ_UNBOUND pool, so the bottom half can migrate off
 * the interrupt-taking CPU that the previous threaded IRQ pinned to
 * (irq_thread() calls set_cpus_allowed_ptr() with the IRQ affinity mask).
 *
 * The hard IRQ handler has already:
 * - Verified this is our interrupt (QSPI_RDY was set)
 * - Cached FIFO status in tqspi->status_reg
 * - Parsed tx_status / rx_status from FIFO status
 * - Masked further interrupts
 */
static void tegra_qspi_work_handler(struct work_struct *work)
{
	struct tegra_qspi *tqspi = container_of(work, struct tegra_qspi, irq_work);
	unsigned long flags;

	spin_lock_irqsave(&tqspi->lock, flags);

	/*
	 * tegra_qspi_handle_timeout() sets recovery_in_progress under
	 * tqspi->lock and then calls cancel_work_sync(), so any running
	 * worker is drained and tegra_qspi_isr() cannot enqueue a new
	 * one while recovery runs. The curr_xfer NULL check catches the
	 * case where the timeout path already tore the transfer down
	 * before this work got a chance to run.
	 */
	if (!tqspi->curr_xfer) {
		spin_unlock_irqrestore(&tqspi->lock, flags);
		return;
	}

	spin_unlock_irqrestore(&tqspi->lock, flags);

	/*
	 * Lock is released here but handlers safely re-check curr_xfer under
	 * lock before dereferencing.
	 * DMA handler also needs to sleep in wait_for_completion_*(), which
	 * cannot be done while holding spinlock.
	 */
	if (!READ_ONCE(tqspi->is_curr_dma_xfer))
		handle_cpu_based_xfer(tqspi);
	else
		handle_dma_based_xfer(tqspi);
}

/**
 * tegra_qspi_isr - Hard IRQ handler
 * @irq: IRQ number
 * @context_data: QSPI controller instance
 *
 * Runs in hard IRQ context with minimal latency. Cannot sleep.
 *
 * Tegra QSPI uses a dedicated, non-shared GIC SPI line on every SoC that
 * uses this driver. The handler always returns IRQ_HANDLED and always
 * acknowledges/re-masks the controller IRQ, so the level-triggered line
 * cannot stay asserted and trip the kernel spurious-IRQ detector into
 * disabling the line. On a stray IRQ where curr_xfer is NULL (e.g. the
 * timeout path has already torn the transfer down) the FIFO/status
 * processing and bottom-half scheduling are skipped because there is no
 * transfer to drive forward.
 *
 * Return: IRQ_HANDLED.
 */
static irqreturn_t tegra_qspi_isr(int irq, void *context_data)
{
	struct tegra_qspi *tqspi = context_data;
	u32 status_reg, trans_status;
	u32 tx_status = 0, rx_status = 0;

	if (!READ_ONCE(tqspi->curr_xfer)) {
		tegra_qspi_mask_clear_irq(tqspi);
		return IRQ_HANDLED;
	}

	spin_lock(&tqspi->lock);
	status_reg = tegra_qspi_readl(tqspi, QSPI_FIFO_STATUS);
	trans_status = tegra_qspi_readl(tqspi, QSPI_TRANS_STATUS);

	if (tqspi->cur_direction & DATA_DIR_TX) {
		tx_status = status_reg & (QSPI_TX_FIFO_UNF | QSPI_TX_FIFO_OVF);
		WRITE_ONCE(tqspi->tx_status, tx_status);
	}

	if (tqspi->cur_direction & DATA_DIR_RX) {
		rx_status = status_reg & (QSPI_RX_FIFO_OVF | QSPI_RX_FIFO_UNF);
		WRITE_ONCE(tqspi->rx_status, rx_status);
	}

	WRITE_ONCE(tqspi->status_reg, status_reg);
	/*
	 * Publish trans_status with release semantics before we clear
	 * the hardware status in tegra_qspi_mask_clear_irq() below. That
	 * ordering matters for the lock-free cache read in
	 * tegra_qspi_handle_timeout(): if the timeout path sees the
	 * released trans_status it also observes the matching status_reg
	 * / tx_status / rx_status; if it does not yet see the released
	 * value it falls back to a live QSPI_TRANS_STATUS read, and that
	 * live read still returns QSPI_RDY because we have not cleared
	 * the register yet. Reversing this order would open a window
	 * where the cache is still zero but the hardware bit has already
	 * been cleared, making the fallback report a false timeout.
	 */
	smp_store_release(&tqspi->trans_status, trans_status);

	tegra_qspi_mask_clear_irq(tqspi);

	/*
	 * If tegra_qspi_handle_timeout() is draining the bottom half,
	 * skip queueing new work. The flag is set under tqspi->lock and
	 * queue_work() below happens while we still hold the lock, so
	 * the guard is atomic with the queue decision. Any ISR that had
	 * already passed this check is drained by the synchronize_irq()
	 * call that tegra_qspi_handle_timeout() issues after publishing
	 * the flag.
	 */
	if (READ_ONCE(tqspi->recovery_in_progress)) {
		spin_unlock(&tqspi->lock);
		return IRQ_HANDLED;
	}

	/*
	 * Small-PIO fastpath: drain the FIFO inline only when this chunk
	 * completes the entire outstanding transfer and no error bit was
	 * latched, to avoid workqueue scheduling latency for TPM-style
	 * short reads.
	 *
	 * The "last chunk" decision is computed and cached as a scalar by
	 * tegra_qspi_start_cpu_based_transfer() before it unmasks the IRQ,
	 * so the hard-IRQ fastpath never dereferences the spi_transfer
	 * pointer here. That keeps the ISR safe against any teardown race
	 * where the synchronous path could clear curr_xfer concurrently.
	 *
	 * The fastpath dispatch decision is made while still holding
	 * tqspi->lock, so the recovery_in_progress guard above covers it
	 * atomically with queue_work() below: an ISR that reaches the
	 * fastpath cannot race a tegra_qspi_handle_timeout() that
	 * subsequently observes recovery_in_progress == true, because
	 * that path calls synchronize_irq() before proceeding. We drop
	 * the lock before calling handle_cpu_based_xfer() so it can take
	 * tqspi->lock internally without deadlocking.
	 *
	 * Multi-chunk PIO continuation stays on the workqueue so that
	 * tegra_qspi_start_cpu_based_transfer() can re-arm the IRQ from
	 * process context. DMA transfers also stay on the workqueue
	 * because their completion path sleeps on the DMA engine.
	 * tegra_qspi_handle_error() -> device_reset() can sleep, so the
	 * fastpath only runs when both status words are clean.
	 */
	if (!READ_ONCE(tqspi->is_curr_dma_xfer) &&
	    READ_ONCE(tqspi->is_last_pio_chunk) &&
	    !tx_status && !rx_status) {
		spin_unlock(&tqspi->lock);
		return handle_cpu_based_xfer(tqspi);
	}

	queue_work(tqspi->wq, &tqspi->irq_work);

	spin_unlock(&tqspi->lock);

	return IRQ_HANDLED;
}

static struct tegra_qspi_soc_data tegra210_qspi_soc_data = {
	.has_ext_dma = true,
	.cmb_xfer_capable = false,
	.supports_tpm = false,
	.cs_count = 1,
};

static struct tegra_qspi_soc_data tegra186_qspi_soc_data = {
	.has_ext_dma = true,
	.cmb_xfer_capable = true,
	.supports_tpm = false,
	.cs_count = 1,
};

static struct tegra_qspi_soc_data tegra234_qspi_soc_data = {
	.has_ext_dma = false,
	.cmb_xfer_capable = true,
	.supports_tpm = true,
	.cs_count = 1,
};

static struct tegra_qspi_soc_data tegra241_qspi_soc_data = {
	.has_ext_dma = true,
	.cmb_xfer_capable = true,
	.supports_tpm = true,
	.cs_count = 4,
};

static const struct of_device_id tegra_qspi_of_match[] = {
	{
		.compatible = "nvidia,tegra210-qspi",
		.data	    = &tegra210_qspi_soc_data,
	}, {
		.compatible = "nvidia,tegra186-qspi",
		.data	    = &tegra186_qspi_soc_data,
	}, {
		.compatible = "nvidia,tegra194-qspi",
		.data	    = &tegra186_qspi_soc_data,
	}, {
		.compatible = "nvidia,tegra234-qspi",
		.data	    = &tegra234_qspi_soc_data,
	}, {
		.compatible = "nvidia,tegra241-qspi",
		.data	    = &tegra241_qspi_soc_data,
	},
	{}
};

MODULE_DEVICE_TABLE(of, tegra_qspi_of_match);

#ifdef CONFIG_ACPI
static const struct acpi_device_id tegra_qspi_acpi_match[] = {
	{
		.id = "NVDA1213",
		.driver_data = (kernel_ulong_t)&tegra210_qspi_soc_data,
	}, {
		.id = "NVDA1313",
		.driver_data = (kernel_ulong_t)&tegra186_qspi_soc_data,
	}, {
		.id = "NVDA1413",
		.driver_data = (kernel_ulong_t)&tegra234_qspi_soc_data,
	}, {
		.id = "NVDA1513",
		.driver_data = (kernel_ulong_t)&tegra241_qspi_soc_data,
	},
	{}
};

MODULE_DEVICE_TABLE(acpi, tegra_qspi_acpi_match);
#endif

static int tegra_qspi_probe(struct platform_device *pdev)
{
	struct spi_controller	*host;
	struct tegra_qspi	*tqspi;
	struct resource		*r;
	int ret, qspi_irq;
	int bus_num;

	host = devm_spi_alloc_host(&pdev->dev, sizeof(*tqspi));
	if (!host)
		return -ENOMEM;

	platform_set_drvdata(pdev, host);
	tqspi = spi_controller_get_devdata(host);

	host->mode_bits = SPI_MODE_0 | SPI_MODE_3 | SPI_CS_HIGH |
			  SPI_TX_DUAL | SPI_RX_DUAL | SPI_TX_QUAD | SPI_RX_QUAD;
	host->bits_per_word_mask = SPI_BPW_MASK(32) | SPI_BPW_MASK(16) | SPI_BPW_MASK(8);
	host->flags = SPI_CONTROLLER_HALF_DUPLEX;
	host->setup = tegra_qspi_setup;
	host->transfer_one_message = tegra_qspi_transfer_one_message;
	host->num_chipselect = 1;
	host->auto_runtime_pm = true;

	bus_num = of_alias_get_id(pdev->dev.of_node, "spi");
	if (bus_num >= 0)
		host->bus_num = bus_num;

	tqspi->host = host;
	tqspi->dev = &pdev->dev;
	spin_lock_init(&tqspi->lock);

	tqspi->soc_data = device_get_match_data(&pdev->dev);
	host->num_chipselect = tqspi->soc_data->cs_count;
	tqspi->base = devm_platform_get_and_ioremap_resource(pdev, 0, &r);
	if (IS_ERR(tqspi->base))
		return PTR_ERR(tqspi->base);

	tqspi->phys = r->start;
	qspi_irq = platform_get_irq(pdev, 0);
	if (qspi_irq < 0)
		return qspi_irq;
	tqspi->irq = qspi_irq;

	if (!has_acpi_companion(tqspi->dev)) {
		tqspi->clk = devm_clk_get(&pdev->dev, "qspi");
		if (IS_ERR(tqspi->clk)) {
			ret = PTR_ERR(tqspi->clk);
			dev_err(&pdev->dev, "failed to get clock: %d\n", ret);
			return ret;
		}

	}

	tqspi->max_buf_size = QSPI_FIFO_DEPTH << 2;
	tqspi->dma_buf_size = DEFAULT_QSPI_DMA_BUF_LEN;

	ret = tegra_qspi_init_dma(tqspi);
	if (ret < 0)
		return ret;

	if (tqspi->use_dma)
		tqspi->max_buf_size = tqspi->dma_buf_size;

	init_completion(&tqspi->tx_dma_complete);
	init_completion(&tqspi->rx_dma_complete);
	init_completion(&tqspi->xfer_completion);

	/*
	 * Set autosuspend delay to 500ms. Testing shows this value eliminates
	 * suspend/resume overhead during burst operations while allowing quick
	 * suspension during idle. For longer operations, the overhead is negligible.
	 */
	pm_runtime_set_autosuspend_delay(&pdev->dev, 500);
	pm_runtime_use_autosuspend(&pdev->dev);

	pm_runtime_enable(&pdev->dev);
	ret = pm_runtime_resume_and_get(&pdev->dev);
	if (ret < 0) {
		dev_err(&pdev->dev, "failed to get runtime PM: %d\n", ret);
		goto exit_pm_disable;
	}

	if (device_reset(tqspi->dev) < 0)
		dev_warn_once(tqspi->dev, "device reset failed\n");

	tqspi->def_command1_reg = QSPI_M_S | QSPI_CS_SW_HW |  QSPI_CS_SW_VAL;
	tegra_qspi_writel(tqspi, tqspi->def_command1_reg, QSPI_COMMAND1);
	tqspi->spi_cs_timing1 = tegra_qspi_readl(tqspi, QSPI_CS_TIMING1);
	tqspi->spi_cs_timing2 = tegra_qspi_readl(tqspi, QSPI_CS_TIMING2);
	tqspi->def_command2_reg = tegra_qspi_readl(tqspi, QSPI_COMMAND2);

	pm_runtime_put_autosuspend(&pdev->dev);

	tqspi->wq = alloc_workqueue("%s", WQ_HIGHPRI | WQ_UNBOUND, 0,
				    dev_name(&pdev->dev));
	if (!tqspi->wq) {
		dev_err(&pdev->dev, "failed to allocate workqueue\n");
		ret = -ENOMEM;
		goto exit_pm_disable;
	}

	INIT_WORK(&tqspi->irq_work, tegra_qspi_work_handler);

	ret = request_irq(tqspi->irq, tegra_qspi_isr, 0,
			  dev_name(&pdev->dev), tqspi);
	if (ret < 0) {
		dev_err(&pdev->dev, "failed to request IRQ#%u: %d\n", tqspi->irq, ret);
		goto exit_destroy_wq;
	}

	ret = spi_register_controller(host);
	if (ret < 0) {
		dev_err(&pdev->dev, "failed to register host: %d\n", ret);
		goto exit_free_irq;
	}

	return 0;

exit_free_irq:
	free_irq(tqspi->irq, tqspi);
exit_destroy_wq:
	destroy_workqueue(tqspi->wq);
exit_pm_disable:
	pm_runtime_dont_use_autosuspend(&pdev->dev);
	pm_runtime_force_suspend(&pdev->dev);
	tegra_qspi_deinit_dma(tqspi);
	return ret;
}

static void tegra_qspi_remove(struct platform_device *pdev)
{
	struct spi_controller *host = platform_get_drvdata(pdev);
	struct tegra_qspi *tqspi = spi_controller_get_devdata(host);

	/*
	 * Tear down in reverse order of probe() so that the controller stops
	 * accepting transfers before the IRQ is released, no new work can be
	 * queued after the IRQ is freed, and any work already queued is
	 * drained while the clocks are still running.
	 */
	spi_unregister_controller(host);
	free_irq(tqspi->irq, tqspi);
	destroy_workqueue(tqspi->wq);
	pm_runtime_dont_use_autosuspend(&pdev->dev);
	pm_runtime_force_suspend(&pdev->dev);
	tegra_qspi_deinit_dma(tqspi);
}

static int tegra_qspi_suspend(struct device *dev)
{
	struct spi_controller *host = dev_get_drvdata(dev);

	return spi_controller_suspend(host);
}

static int tegra_qspi_resume(struct device *dev)
{
	struct spi_controller *host = dev_get_drvdata(dev);
	struct tegra_qspi *tqspi = spi_controller_get_devdata(host);
	int ret;

	ret = pm_runtime_resume_and_get(dev);
	if (ret < 0) {
		dev_err(dev, "failed to get runtime PM: %d\n", ret);
		return ret;
	}

	tegra_qspi_writel(tqspi, tqspi->command1_reg, QSPI_COMMAND1);
	tegra_qspi_writel(tqspi, tqspi->def_command2_reg, QSPI_COMMAND2);
	pm_runtime_put_autosuspend(dev);

	return spi_controller_resume(host);
}

static int tegra_qspi_runtime_suspend(struct device *dev)
{
	struct spi_controller *host = dev_get_drvdata(dev);
	struct tegra_qspi *tqspi = spi_controller_get_devdata(host);

	/* Runtime pm disabled with ACPI */
	if (has_acpi_companion(tqspi->dev))
		return 0;
	/* flush all write which are in PPSB queue by reading back */
	tegra_qspi_readl(tqspi, QSPI_COMMAND1);

	clk_disable_unprepare(tqspi->clk);

	return 0;
}

static int tegra_qspi_runtime_resume(struct device *dev)
{
	struct spi_controller *host = dev_get_drvdata(dev);
	struct tegra_qspi *tqspi = spi_controller_get_devdata(host);
	int ret;

	/* Runtime pm disabled with ACPI */
	if (has_acpi_companion(tqspi->dev))
		return 0;
	ret = clk_prepare_enable(tqspi->clk);
	if (ret < 0)
		dev_err(tqspi->dev, "failed to enable clock: %d\n", ret);

	return ret;
}

static const struct dev_pm_ops tegra_qspi_pm_ops = {
	RUNTIME_PM_OPS(tegra_qspi_runtime_suspend, tegra_qspi_runtime_resume, NULL)
	SYSTEM_SLEEP_PM_OPS(tegra_qspi_suspend, tegra_qspi_resume)
};

static struct platform_driver tegra_qspi_driver = {
	.driver = {
		.name		= "tegra-qspi",
		.pm		= pm_ptr(&tegra_qspi_pm_ops),
		.of_match_table	= tegra_qspi_of_match,
		.acpi_match_table = ACPI_PTR(tegra_qspi_acpi_match),
	},
	.probe =	tegra_qspi_probe,
	.remove =	tegra_qspi_remove,
};
module_platform_driver(tegra_qspi_driver);

MODULE_ALIAS("platform:qspi-tegra");
MODULE_DESCRIPTION("NVIDIA Tegra QSPI Controller Driver");
MODULE_AUTHOR("Sowjanya Komatineni <skomatineni@nvidia.com>");
MODULE_LICENSE("GPL v2"