[PATCH v8 07/28] spi: cadence-qspi: Add support for octal DTR flashes

[Pratyush Yadav <p.yadav@ti.com>]

Set up opcode extension and enable/disable DTR mode based on whether the
command is DTR or not.

xSPI flashes can have a 4-byte dummy address associated with some
commands like the Read Status Register command in octal DTR mode. Since
the flash does not support sending the dummy address, we can not use
automatic write completion polling in DTR mode. Further, no write
completion polling makes it impossible to use DAC mode for DTR writes.
In that mode, the controller does not know beforehand how long a write
will be and so it can de-assert Chip Select (CS#) at any time. Once CS#
is de-assert, the flash will go into burning phase. But since the
controller does not do write completion polling, it does not know when
the flash is busy and might send in writes while the flash is not ready.

So, disable write completion polling and make writes go through indirect
mode for DTR writes and let spi-mem take care of polling the SR.

Signed-off-by: Pratyush Yadav <p.yadav@ti.com>
---
 drivers/spi/cadence_qspi.c     |  39 ++++-
 drivers/spi/cadence_qspi.h     |  14 +-
 drivers/spi/cadence_qspi_apb.c | 286 ++++++++++++++++++++++++++++++---
 3 files changed, 313 insertions(+), 26 deletions(-)

diff --git a/drivers/spi/cadence_qspi.c b/drivers/spi/cadence_qspi.c
index a961193cdc..d1b3808c4d 100644
--- a/drivers/spi/cadence_qspi.c
+++ b/drivers/spi/cadence_qspi.c
@@ -43,20 +43,22 @@ static int cadence_spi_write_speed(struct udevice *bus, uint hz)
 	return 0;
 }
 
-static int cadence_spi_read_id(void *reg_base, u8 len, u8 *idcode)
+static int cadence_spi_read_id(struct cadence_spi_plat *plat, u8 len,
+			       u8 *idcode)
 {
 	struct spi_mem_op op = SPI_MEM_OP(SPI_MEM_OP_CMD(0x9F, 1),
 					  SPI_MEM_OP_NO_ADDR,
 					  SPI_MEM_OP_NO_DUMMY,
 					  SPI_MEM_OP_DATA_IN(len, idcode, 1));
 
-	return cadence_qspi_apb_command_read(reg_base, &op);
+	return cadence_qspi_apb_command_read(plat, &op);
 }
 
 /* Calibration sequence to determine the read data capture delay register */
 static int spi_calibration(struct udevice *bus, uint hz)
 {
 	struct cadence_spi_priv *priv = dev_get_priv(bus);
+	struct cadence_spi_plat *plat = dev_get_plat(bus);
 	void *base = priv->regbase;
 	unsigned int idcode = 0, temp = 0;
 	int err = 0, i, range_lo = -1, range_hi = -1;
@@ -71,7 +73,7 @@ static int spi_calibration(struct udevice *bus, uint hz)
 	cadence_qspi_apb_controller_enable(base);
 
 	/* read the ID which will be our golden value */
-	err = cadence_spi_read_id(base, 3, (u8 *)&idcode);
+	err = cadence_spi_read_id(plat, 3, (u8 *)&idcode);
 	if (err) {
 		puts("SF: Calibration failed (read)\n");
 		return err;
@@ -90,7 +92,7 @@ static int spi_calibration(struct udevice *bus, uint hz)
 		cadence_qspi_apb_controller_enable(base);
 
 		/* issue a RDID to get the ID value */
-		err = cadence_spi_read_id(base, 3, (u8 *)&temp);
+		err = cadence_spi_read_id(plat, 3, (u8 *)&temp);
 		if (err) {
 			puts("SF: Calibration failed (read)\n");
 			return err;
@@ -271,10 +273,14 @@ static int cadence_spi_mem_exec_op(struct spi_slave *spi,
 
 	switch (mode) {
 	case CQSPI_STIG_READ:
-		err = cadence_qspi_apb_command_read(base, op);
+		err = cadence_qspi_apb_command_read_setup(plat, op);
+		if (!err)
+			err = cadence_qspi_apb_command_read(plat, op);
 		break;
 	case CQSPI_STIG_WRITE:
-		err = cadence_qspi_apb_command_write(base, op);
+		err = cadence_qspi_apb_command_write_setup(plat, op);
+		if (!err)
+			err = cadence_qspi_apb_command_write(plat, op);
 		break;
 	case CQSPI_READ:
 		err = cadence_qspi_apb_read_setup(plat, op);
@@ -294,6 +300,26 @@ static int cadence_spi_mem_exec_op(struct spi_slave *spi,
 	return err;
 }
 
+static bool cadence_spi_mem_supports_op(struct spi_slave *slave,
+					const struct spi_mem_op *op)
+{
+	bool all_true, all_false;
+
+	all_true = op->cmd.dtr && op->addr.dtr && op->dummy.dtr &&
+		   op->data.dtr;
+	all_false = !op->cmd.dtr && !op->addr.dtr && !op->dummy.dtr &&
+		    !op->data.dtr;
+
+	/* Mixed DTR modes not supported. */
+	if (!(all_true || all_false))
+		return false;
+
+	if (all_true)
+		return spi_mem_dtr_supports_op(slave, op);
+	else
+		return spi_mem_default_supports_op(slave, op);
+}
+
 static int cadence_spi_of_to_plat(struct udevice *bus)
 {
 	struct cadence_spi_plat *plat = dev_get_plat(bus);
@@ -350,6 +376,7 @@ static int cadence_spi_of_to_plat(struct udevice *bus)
 
 static const struct spi_controller_mem_ops cadence_spi_mem_ops = {
 	.exec_op = cadence_spi_mem_exec_op,
+	.supports_op = cadence_spi_mem_supports_op,
 };
 
 static const struct dm_spi_ops cadence_spi_ops = {
diff --git a/drivers/spi/cadence_qspi.h b/drivers/spi/cadence_qspi.h
index 5c745541a6..49b401168f 100644
--- a/drivers/spi/cadence_qspi.h
+++ b/drivers/spi/cadence_qspi.h
@@ -36,6 +36,12 @@ struct cadence_spi_plat {
 	u32		tsd2d_ns;
 	u32		tchsh_ns;
 	u32		tslch_ns;
+
+	/* Transaction protocol parameters. */
+	u8		inst_width;
+	u8		addr_width;
+	u8		data_width;
+	bool		dtr;
 };
 
 struct cadence_spi_priv {
@@ -59,9 +65,13 @@ void cadence_qspi_apb_controller_enable(void *reg_base_addr);
 void cadence_qspi_apb_controller_disable(void *reg_base_addr);
 void cadence_qspi_apb_dac_mode_enable(void *reg_base);
 
-int cadence_qspi_apb_command_read(void *reg_base_addr,
+int cadence_qspi_apb_command_read_setup(struct cadence_spi_plat *plat,
+					const struct spi_mem_op *op);
+int cadence_qspi_apb_command_read(struct cadence_spi_plat *plat,
 				  const struct spi_mem_op *op);
-int cadence_qspi_apb_command_write(void *reg_base_addr,
+int cadence_qspi_apb_command_write_setup(struct cadence_spi_plat *plat,
+					 const struct spi_mem_op *op);
+int cadence_qspi_apb_command_write(struct cadence_spi_plat *plat,
 				   const struct spi_mem_op *op);
 
 int cadence_qspi_apb_read_setup(struct cadence_spi_plat *plat,
diff --git a/drivers/spi/cadence_qspi_apb.c b/drivers/spi/cadence_qspi_apb.c
index 92e57730bd..c36a652211 100644
--- a/drivers/spi/cadence_qspi_apb.c
+++ b/drivers/spi/cadence_qspi_apb.c
@@ -51,7 +51,7 @@
 #define CQSPI_STIG_DATA_LEN_MAX			8
 
 #define CQSPI_DUMMY_CLKS_PER_BYTE		8
-#define CQSPI_DUMMY_BYTES_MAX			4
+#define CQSPI_DUMMY_CLKS_MAX			31
 
 /****************************************************************************
  * Controller's configuration and status register (offset from QSPI_BASE)
@@ -65,6 +65,8 @@
 #define	CQSPI_REG_CONFIG_XIP_IMM		BIT(18)
 #define	CQSPI_REG_CONFIG_CHIPSELECT_LSB		10
 #define	CQSPI_REG_CONFIG_BAUD_LSB		19
+#define CQSPI_REG_CONFIG_DTR_PROTO		BIT(24)
+#define CQSPI_REG_CONFIG_DUAL_OPCODE		BIT(30)
 #define	CQSPI_REG_CONFIG_IDLE_LSB		31
 #define	CQSPI_REG_CONFIG_CHIPSELECT_MASK	0xF
 #define	CQSPI_REG_CONFIG_BAUD_MASK		0xF
@@ -83,6 +85,7 @@
 
 #define	CQSPI_REG_WR_INSTR			0x08
 #define	CQSPI_REG_WR_INSTR_OPCODE_LSB		0
+#define CQSPI_REG_WR_INSTR_TYPE_ADDR_LSB	12
 #define	CQSPI_REG_WR_INSTR_TYPE_DATA_LSB	16
 
 #define	CQSPI_REG_DELAY				0x0C
@@ -120,6 +123,9 @@
 #define	CQSPI_REG_SDRAMLEVEL_RD_MASK		0xFFFF
 #define	CQSPI_REG_SDRAMLEVEL_WR_MASK		0xFFFF
 
+#define CQSPI_REG_WR_COMPLETION_CTRL		0x38
+#define CQSPI_REG_WR_DISABLE_AUTO_POLL		BIT(14)
+
 #define	CQSPI_REG_IRQSTATUS			0x40
 #define	CQSPI_REG_IRQMASK			0x44
 
@@ -166,6 +172,11 @@
 #define	CQSPI_REG_CMDWRITEDATALOWER		0xA8
 #define	CQSPI_REG_CMDWRITEDATAUPPER		0xAC
 
+#define CQSPI_REG_OP_EXT_LOWER			0xE0
+#define CQSPI_REG_OP_EXT_READ_LSB		24
+#define CQSPI_REG_OP_EXT_WRITE_LSB		16
+#define CQSPI_REG_OP_EXT_STIG_LSB		0
+
 #define CQSPI_REG_IS_IDLE(base)					\
 	((readl(base + CQSPI_REG_CONFIG) >>		\
 		CQSPI_REG_CONFIG_IDLE_LSB) & 0x1)
@@ -203,6 +214,75 @@ void cadence_qspi_apb_dac_mode_enable(void *reg_base)
 	writel(reg, reg_base + CQSPI_REG_CONFIG);
 }
 
+static unsigned int cadence_qspi_calc_dummy(const struct spi_mem_op *op,
+					    bool dtr)
+{
+	unsigned int dummy_clk;
+
+	dummy_clk = op->dummy.nbytes * (8 / op->dummy.buswidth);
+	if (dtr)
+		dummy_clk /= 2;
+
+	return dummy_clk;
+}
+
+static u32 cadence_qspi_calc_rdreg(struct cadence_spi_plat *plat)
+{
+	u32 rdreg = 0;
+
+	rdreg |= plat->inst_width << CQSPI_REG_RD_INSTR_TYPE_INSTR_LSB;
+	rdreg |= plat->addr_width << CQSPI_REG_RD_INSTR_TYPE_ADDR_LSB;
+	rdreg |= plat->data_width << CQSPI_REG_RD_INSTR_TYPE_DATA_LSB;
+
+	return rdreg;
+}
+
+static int cadence_qspi_buswidth_to_inst_type(u8 buswidth)
+{
+	switch (buswidth) {
+	case 0:
+	case 1:
+		return CQSPI_INST_TYPE_SINGLE;
+
+	case 2:
+		return CQSPI_INST_TYPE_DUAL;
+
+	case 4:
+		return CQSPI_INST_TYPE_QUAD;
+
+	case 8:
+		return CQSPI_INST_TYPE_OCTAL;
+
+	default:
+		return -ENOTSUPP;
+	}
+}
+
+static int cadence_qspi_set_protocol(struct cadence_spi_plat *plat,
+				     const struct spi_mem_op *op)
+{
+	int ret;
+
+	plat->dtr = op->data.dtr && op->cmd.dtr && op->addr.dtr;
+
+	ret = cadence_qspi_buswidth_to_inst_type(op->cmd.buswidth);
+	if (ret < 0)
+		return ret;
+	plat->inst_width = ret;
+
+	ret = cadence_qspi_buswidth_to_inst_type(op->addr.buswidth);
+	if (ret < 0)
+		return ret;
+	plat->addr_width = ret;
+
+	ret = cadence_qspi_buswidth_to_inst_type(op->data.buswidth);
+	if (ret < 0)
+		return ret;
+	plat->data_width = ret;
+
+	return 0;
+}
+
 /* Return 1 if idle, otherwise return 0 (busy). */
 static unsigned int cadence_qspi_wait_idle(void *reg_base)
 {
@@ -434,21 +514,109 @@ static int cadence_qspi_apb_exec_flash_cmd(void *reg_base,
 	return 0;
 }
 
+static int cadence_qspi_setup_opcode_ext(struct cadence_spi_plat *plat,
+					 const struct spi_mem_op *op,
+					 unsigned int shift)
+{
+	unsigned int reg;
+	u8 ext;
+
+	if (op->cmd.nbytes != 2)
+		return -EINVAL;
+
+	/* Opcode extension is the LSB. */
+	ext = op->cmd.opcode & 0xff;
+
+	reg = readl(plat->regbase + CQSPI_REG_OP_EXT_LOWER);
+	reg &= ~(0xff << shift);
+	reg |= ext << shift;
+	writel(reg, plat->regbase + CQSPI_REG_OP_EXT_LOWER);
+
+	return 0;
+}
+
+static int cadence_qspi_enable_dtr(struct cadence_spi_plat *plat,
+				   const struct spi_mem_op *op,
+				   unsigned int shift,
+				   bool enable)
+{
+	unsigned int reg;
+	int ret;
+
+	reg = readl(plat->regbase + CQSPI_REG_CONFIG);
+
+	if (enable) {
+		reg |= CQSPI_REG_CONFIG_DTR_PROTO;
+		reg |= CQSPI_REG_CONFIG_DUAL_OPCODE;
+
+		/* Set up command opcode extension. */
+		ret = cadence_qspi_setup_opcode_ext(plat, op, shift);
+		if (ret)
+			return ret;
+	} else {
+		reg &= ~CQSPI_REG_CONFIG_DTR_PROTO;
+		reg &= ~CQSPI_REG_CONFIG_DUAL_OPCODE;
+	}
+
+	writel(reg, plat->regbase + CQSPI_REG_CONFIG);
+
+	return 0;
+}
+
+int cadence_qspi_apb_command_read_setup(struct cadence_spi_plat *plat,
+					const struct spi_mem_op *op)
+{
+	int ret;
+	unsigned int reg;
+
+	ret = cadence_qspi_set_protocol(plat, op);
+	if (ret)
+		return ret;
+
+	ret = cadence_qspi_enable_dtr(plat, op, CQSPI_REG_OP_EXT_STIG_LSB,
+				      plat->dtr);
+	if (ret)
+		return ret;
+
+	reg = cadence_qspi_calc_rdreg(plat);
+	writel(reg, plat->regbase + CQSPI_REG_RD_INSTR);
+
+	return 0;
+}
+
 /* For command RDID, RDSR. */
-int cadence_qspi_apb_command_read(void *reg_base, const struct spi_mem_op *op)
+int cadence_qspi_apb_command_read(struct cadence_spi_plat *plat,
+				  const struct spi_mem_op *op)
 {
+	void *reg_base = plat->regbase;
 	unsigned int reg;
 	unsigned int read_len;
 	int status;
 	unsigned int rxlen = op->data.nbytes;
 	void *rxbuf = op->data.buf.in;
+	unsigned int dummy_clk;
+	u8 opcode;
 
 	if (rxlen > CQSPI_STIG_DATA_LEN_MAX || !rxbuf) {
 		printf("QSPI: Invalid input arguments rxlen %u\n", rxlen);
 		return -EINVAL;
 	}
 
-	reg = op->cmd.opcode << CQSPI_REG_CMDCTRL_OPCODE_LSB;
+	if (plat->dtr)
+		opcode = op->cmd.opcode >> 8;
+	else
+		opcode = op->cmd.opcode;
+
+	reg = opcode << CQSPI_REG_CMDCTRL_OPCODE_LSB;
+
+	/* Set up dummy cycles. */
+	dummy_clk = cadence_qspi_calc_dummy(op, plat->dtr);
+	if (dummy_clk > CQSPI_DUMMY_CLKS_MAX)
+		return -ENOTSUPP;
+
+	if (dummy_clk)
+		reg |= (dummy_clk & CQSPI_REG_CMDCTRL_DUMMY_MASK)
+		     << CQSPI_REG_CMDCTRL_DUMMY_LSB;
 
 	reg |= (0x1 << CQSPI_REG_CMDCTRL_RD_EN_LSB);
 
@@ -475,15 +643,39 @@ int cadence_qspi_apb_command_read(void *reg_base, const struct spi_mem_op *op)
 	return 0;
 }
 
+int cadence_qspi_apb_command_write_setup(struct cadence_spi_plat *plat,
+					 const struct spi_mem_op *op)
+{
+	int ret;
+	unsigned int reg;
+
+	ret = cadence_qspi_set_protocol(plat, op);
+	if (ret)
+		return ret;
+
+	ret = cadence_qspi_enable_dtr(plat, op, CQSPI_REG_OP_EXT_STIG_LSB,
+				      plat->dtr);
+	if (ret)
+		return ret;
+
+	reg = cadence_qspi_calc_rdreg(plat);
+	writel(reg, plat->regbase + CQSPI_REG_RD_INSTR);
+
+	return 0;
+}
+
 /* For commands: WRSR, WREN, WRDI, CHIP_ERASE, BE, etc. */
-int cadence_qspi_apb_command_write(void *reg_base, const struct spi_mem_op *op)
+int cadence_qspi_apb_command_write(struct cadence_spi_plat *plat,
+				   const struct spi_mem_op *op)
 {
 	unsigned int reg = 0;
 	unsigned int wr_data;
 	unsigned int wr_len;
 	unsigned int txlen = op->data.nbytes;
 	const void *txbuf = op->data.buf.out;
+	void *reg_base = plat->regbase;
 	u32 addr;
+	u8 opcode;
 
 	/* Reorder address to SPI bus order if only transferring address */
 	if (!txlen) {
@@ -499,7 +691,12 @@ int cadence_qspi_apb_command_write(void *reg_base, const struct spi_mem_op *op)
 		return -EINVAL;
 	}
 
-	reg |= op->cmd.opcode << CQSPI_REG_CMDCTRL_OPCODE_LSB;
+	if (plat->dtr)
+		opcode = op->cmd.opcode >> 8;
+	else
+		opcode = op->cmd.opcode;
+
+	reg |= opcode << CQSPI_REG_CMDCTRL_OPCODE_LSB;
 
 	if (txlen) {
 		/* writing data = yes */
@@ -533,29 +730,39 @@ int cadence_qspi_apb_read_setup(struct cadence_spi_plat *plat,
 	unsigned int rd_reg;
 	unsigned int dummy_clk;
 	unsigned int dummy_bytes = op->dummy.nbytes;
+	int ret;
+	u8 opcode;
+
+	ret = cadence_qspi_set_protocol(plat, op);
+	if (ret)
+		return ret;
+
+	ret = cadence_qspi_enable_dtr(plat, op, CQSPI_REG_OP_EXT_READ_LSB,
+				      plat->dtr);
+	if (ret)
+		return ret;
 
 	/* Setup the indirect trigger address */
 	writel(plat->trigger_address,
 	       plat->regbase + CQSPI_REG_INDIRECTTRIGGER);
 
 	/* Configure the opcode */
-	rd_reg = op->cmd.opcode << CQSPI_REG_RD_INSTR_OPCODE_LSB;
+	if (plat->dtr)
+		opcode = op->cmd.opcode >> 8;
+	else
+		opcode = op->cmd.opcode;
 
-	if (op->data.buswidth == 8)
-		/* Instruction and address at DQ0, data at DQ0-7. */
-		rd_reg |= CQSPI_INST_TYPE_OCTAL << CQSPI_REG_RD_INSTR_TYPE_DATA_LSB;
-	else if (op->data.buswidth == 4)
-		/* Instruction and address at DQ0, data at DQ0-3. */
-		rd_reg |= CQSPI_INST_TYPE_QUAD << CQSPI_REG_RD_INSTR_TYPE_DATA_LSB;
+	rd_reg = opcode << CQSPI_REG_RD_INSTR_OPCODE_LSB;
+	rd_reg |= cadence_qspi_calc_rdreg(plat);
 
 	writel(op->addr.val, plat->regbase + CQSPI_REG_INDIRECTRDSTARTADDR);
 
 	if (dummy_bytes) {
-		if (dummy_bytes > CQSPI_DUMMY_BYTES_MAX)
-			dummy_bytes = CQSPI_DUMMY_BYTES_MAX;
-
 		/* Convert to clock cycles. */
-		dummy_clk = dummy_bytes * CQSPI_DUMMY_CLKS_PER_BYTE;
+		dummy_clk = cadence_qspi_calc_dummy(op, plat->dtr);
+
+		if (dummy_clk > CQSPI_DUMMY_CLKS_MAX)
+			return -ENOTSUPP;
 
 		if (dummy_clk)
 			rd_reg |= (dummy_clk & CQSPI_REG_RD_INSTR_DUMMY_MASK)
@@ -682,17 +889,52 @@ int cadence_qspi_apb_write_setup(struct cadence_spi_plat *plat,
 				 const struct spi_mem_op *op)
 {
 	unsigned int reg;
+	int ret;
+	u8 opcode;
+
+	ret = cadence_qspi_set_protocol(plat, op);
+	if (ret)
+		return ret;
+
+	ret = cadence_qspi_enable_dtr(plat, op, CQSPI_REG_OP_EXT_WRITE_LSB,
+				      plat->dtr);
+	if (ret)
+		return ret;
 
 	/* Setup the indirect trigger address */
 	writel(plat->trigger_address,
 	       plat->regbase + CQSPI_REG_INDIRECTTRIGGER);
 
 	/* Configure the opcode */
-	reg = op->cmd.opcode << CQSPI_REG_WR_INSTR_OPCODE_LSB;
+	if (plat->dtr)
+		opcode = op->cmd.opcode >> 8;
+	else
+		opcode = op->cmd.opcode;
+
+	reg = opcode << CQSPI_REG_WR_INSTR_OPCODE_LSB;
+	reg |= plat->data_width << CQSPI_REG_WR_INSTR_TYPE_DATA_LSB;
+	reg |= plat->addr_width << CQSPI_REG_WR_INSTR_TYPE_ADDR_LSB;
 	writel(reg, plat->regbase + CQSPI_REG_WR_INSTR);
 
+	reg = cadence_qspi_calc_rdreg(plat);
+	writel(reg, plat->regbase + CQSPI_REG_RD_INSTR);
+
 	writel(op->addr.val, plat->regbase + CQSPI_REG_INDIRECTWRSTARTADDR);
 
+	if (plat->dtr) {
+		/*
+		 * Some flashes like the cypress Semper flash expect a 4-byte
+		 * dummy address with the Read SR command in DTR mode, but this
+		 * controller does not support sending address with the Read SR
+		 * command. So, disable write completion polling on the
+		 * controller's side. spi-nor will take care of polling the
+		 * status register.
+		 */
+		reg = readl(plat->regbase + CQSPI_REG_WR_COMPLETION_CTRL);
+		reg |= CQSPI_REG_WR_DISABLE_AUTO_POLL;
+		writel(reg, plat->regbase + CQSPI_REG_WR_COMPLETION_CTRL);
+	}
+
 	reg = readl(plat->regbase + CQSPI_REG_SIZE);
 	reg &= ~CQSPI_REG_SIZE_ADDRESS_MASK;
 	reg |= (op->addr.nbytes - 1);
@@ -787,7 +1029,15 @@ int cadence_qspi_apb_write_execute(struct cadence_spi_plat *plat,
 	const void *buf = op->data.buf.out;
 	size_t len = op->data.nbytes;
 
-	if (plat->use_dac_mode && (to + len < plat->ahbsize)) {
+	/*
+	 * Some flashes like the Cypress Semper flash expect a dummy 4-byte
+	 * address (all 0s) with the read status register command in DTR mode.
+	 * But this controller does not support sending dummy address bytes to
+	 * the flash when it is polling the write completion register in DTR
+	 * mode. So, we can not use direct mode when in DTR mode for writing
+	 * data.
+	 */
+	if (!plat->dtr && plat->use_dac_mode && (to + len < plat->ahbsize)) {
 		memcpy_toio(plat->ahbbase + to, buf, len);
 		if (!cadence_qspi_wait_idle(plat->regbase))
 			return -EIO;
-- 
2.30.0