[PATCH v2 2/3] mtd: rawnand: rzn1: Add new NAND controller driver

Miquel Raynal miquel.raynal at bootlin.com
Fri Nov 26 06:06:24 PST 2021


Introduce Renesas RZ/N1x NAND controller driver which supports:
- All ONFI timing modes
- Different configurations of its internal ECC controller
- On-die (not tested) and software ECC support
- Several chips (not tested)
- Subpage accesses
- DMA and PIO

This controller was originally provided by Evatronix before being bought
by Cadence.

Signed-off-by: Miquel Raynal <miquel.raynal at bootlin.com>
---
 drivers/mtd/nand/raw/Kconfig                |    6 +
 drivers/mtd/nand/raw/Makefile               |    1 +
 drivers/mtd/nand/raw/rzn1-nand-controller.c | 1420 +++++++++++++++++++
 3 files changed, 1427 insertions(+)
 create mode 100644 drivers/mtd/nand/raw/rzn1-nand-controller.c

diff --git a/drivers/mtd/nand/raw/Kconfig b/drivers/mtd/nand/raw/Kconfig
index e1baed6c5b33..d3cfdfbeb021 100644
--- a/drivers/mtd/nand/raw/Kconfig
+++ b/drivers/mtd/nand/raw/Kconfig
@@ -467,6 +467,12 @@ config MTD_NAND_PL35X
 	  Enables support for PrimeCell SMC PL351 and PL353 NAND
 	  controller found on Zynq7000.
 
+config MTD_NAND_RZN1
+	tristate "Renesas RZ/N1D, RZ/N1S, RZ/N1L NAND controller"
+	depends on ARCH_RENESAS || COMPILE_TEST
+	help
+	  Enables support for Renesas RZ/N1x SoC family NAND controller.
+
 comment "Misc"
 
 config MTD_SM_COMMON
diff --git a/drivers/mtd/nand/raw/Makefile b/drivers/mtd/nand/raw/Makefile
index ee55f79e5e82..4d75d2520edf 100644
--- a/drivers/mtd/nand/raw/Makefile
+++ b/drivers/mtd/nand/raw/Makefile
@@ -60,6 +60,7 @@ obj-$(CONFIG_MTD_NAND_ARASAN)		+= arasan-nand-controller.o
 obj-$(CONFIG_MTD_NAND_INTEL_LGM)	+= intel-nand-controller.o
 obj-$(CONFIG_MTD_NAND_ROCKCHIP)		+= rockchip-nand-controller.o
 obj-$(CONFIG_MTD_NAND_PL35X)		+= pl35x-nand-controller.o
+obj-$(CONFIG_MTD_NAND_RZN1)		+= rzn1-nand-controller.o
 
 nand-objs := nand_base.o nand_legacy.o nand_bbt.o nand_timings.o nand_ids.o
 nand-objs += nand_onfi.o
diff --git a/drivers/mtd/nand/raw/rzn1-nand-controller.c b/drivers/mtd/nand/raw/rzn1-nand-controller.c
new file mode 100644
index 000000000000..c7a4f95d79c7
--- /dev/null
+++ b/drivers/mtd/nand/raw/rzn1-nand-controller.c
@@ -0,0 +1,1420 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Evatronix/Renesas RZ/N1D, RZ/N1S, RZ/N1L NAND flash controller driver
+ *
+ * Copyright (C) 2021 Schneider Electric
+ * Author: Miquel RAYNAL <miquel.raynal at bootlin.com>
+ */
+
+#include <linux/bitfield.h>
+#include <linux/clk.h>
+#include <linux/dma-mapping.h>
+#include <linux/interrupt.h>
+#include <linux/iopoll.h>
+#include <linux/module.h>
+#include <linux/mtd/mtd.h>
+#include <linux/mtd/rawnand.h>
+#include <linux/of.h>
+#include <linux/platform_device.h>
+#include <linux/slab.h>
+
+#define COMMAND_REG 0x00
+#define   COMMAND_SEQ(x) FIELD_PREP(GENMASK(5, 0), (x))
+#define     COMMAND_SEQ_10 COMMAND_SEQ(0x2A)
+#define     COMMAND_SEQ_12 COMMAND_SEQ(0x0C)
+#define     COMMAND_SEQ_18 COMMAND_SEQ(0x32)
+#define     COMMAND_SEQ_19 COMMAND_SEQ(0x13)
+#define     COMMAND_SEQ_GEN_IN COMMAND_SEQ_18
+#define     COMMAND_SEQ_GEN_OUT COMMAND_SEQ_19
+#define     COMMAND_SEQ_READ_PAGE COMMAND_SEQ_10
+#define     COMMAND_SEQ_WRITE_PAGE COMMAND_SEQ_12
+#define   COMMAND_INPUT_SEL_AHBS 0
+#define   COMMAND_INPUT_SEL_DMA BIT(6)
+#define   COMMAND_FIFO_SEL 0
+#define   COMMAND_DATA_SEL BIT(7)
+#define   COMMAND_0(x) FIELD_PREP(GENMASK(15, 8), (x))
+#define   COMMAND_1(x) FIELD_PREP(GENMASK(23, 16), (x))
+#define   COMMAND_2(x) FIELD_PREP(GENMASK(31, 24), (x))
+
+#define CONTROL_REG 0x04
+#define   CONTROL_CHECK_RB_LINE 0
+#define   CONTROL_ECC_BLOCK_SIZE(x) FIELD_PREP(GENMASK(2, 1), (x))
+#define     CONTROL_ECC_BLOCK_SIZE_256 CONTROL_ECC_BLOCK_SIZE(0)
+#define     CONTROL_ECC_BLOCK_SIZE_512 CONTROL_ECC_BLOCK_SIZE(1)
+#define     CONTROL_ECC_BLOCK_SIZE_1024 CONTROL_ECC_BLOCK_SIZE(2)
+#define   CONTROL_INT_EN BIT(4)
+#define   CONTROL_ECC_EN BIT(5)
+#define   CONTROL_BLOCK_SIZE(x) FIELD_PREP(GENMASK(7, 6), (x))
+#define     CONTROL_BLOCK_SIZE_32P CONTROL_BLOCK_SIZE(0)
+#define     CONTROL_BLOCK_SIZE_64P CONTROL_BLOCK_SIZE(1)
+#define     CONTROL_BLOCK_SIZE_128P CONTROL_BLOCK_SIZE(2)
+#define     CONTROL_BLOCK_SIZE_256P CONTROL_BLOCK_SIZE(3)
+
+#define STATUS_REG 0x8
+#define   MEM_RDY(cs, reg) (FIELD_GET(GENMASK(3, 0), (reg)) & BIT(cs))
+#define   CTRL_RDY(reg) (FIELD_GET(BIT(8), (reg)) == 0)
+
+#define ECC_CTRL_REG 0x18
+#define   ECC_CTRL_CAP(x) FIELD_PREP(GENMASK(2, 0), (x))
+#define     ECC_CTRL_CAP_2B ECC_CTRL_CAP(0)
+#define     ECC_CTRL_CAP_4B ECC_CTRL_CAP(1)
+#define     ECC_CTRL_CAP_8B ECC_CTRL_CAP(2)
+#define     ECC_CTRL_CAP_16B ECC_CTRL_CAP(3)
+#define     ECC_CTRL_CAP_24B ECC_CTRL_CAP(4)
+#define     ECC_CTRL_CAP_32B ECC_CTRL_CAP(5)
+#define   ECC_CTRL_ERR_THRESHOLD(x) FIELD_PREP(GENMASK(13, 8), (x))
+
+#define INT_MASK_REG 0x10
+#define INT_STATUS_REG 0x14
+#define   INT_CMD_END BIT(1)
+#define   INT_DMA_END BIT(3)
+#define   INT_MEM_RDY(cs) FIELD_PREP(GENMASK(11, 8), BIT(cs))
+#define   INT_DMA_ENDED BIT(3)
+#define   MEM_IS_RDY(cs, reg) (FIELD_GET(GENMASK(11, 8), (reg)) & BIT(cs))
+#define   DMA_HAS_ENDED(reg) FIELD_GET(BIT(3), (reg))
+
+#define ECC_OFFSET_REG 0x1C
+#define   ECC_OFFSET(x) FIELD_PREP(GENMASK(15, 0), (x))
+
+#define ECC_STAT_REG 0x20
+#define   ECC_STAT_CORRECTABLE(cs, reg) (FIELD_GET(GENMASK(3, 0), (reg)) & BIT(cs))
+#define   ECC_STAT_UNCORRECTABLE(cs, reg) (FIELD_GET(GENMASK(11, 8), (reg)) & BIT(cs))
+
+#define ADDR0_COL_REG 0x24
+#define   ADDR0_COL(x) FIELD_PREP(GENMASK(15, 0), (x))
+
+#define ADDR0_ROW_REG 0x28
+#define   ADDR0_ROW(x) FIELD_PREP(GENMASK(23, 0), (x))
+
+#define ADDR1_COL_REG 0x2C
+#define   ADDR1_COL(x) FIELD_PREP(GENMASK(15, 0), (x))
+
+#define ADDR1_ROW_REG 0x30
+#define   ADDR1_ROW(x) FIELD_PREP(GENMASK(23, 0), (x))
+
+#define FIFO_DATA_REG 0x38
+
+#define DATA_REG 0x3C
+
+#define DATA_REG_SIZE_REG 0x40
+
+#define DMA_ADDR_LOW_REG 0x64
+
+#define DMA_ADDR_HIGH_REG 0x68
+
+#define DMA_CNT_REG 0x6C
+
+#define DMA_CTRL_REG 0x70
+#define   DMA_CTRL_INCREMENT_BURST_4 0
+#define   DMA_CTRL_REGISTER_MANAGED_MODE 0
+#define   DMA_CTRL_START BIT(7)
+
+#define MEM_CTRL_REG 0x80
+#define   MEM_CTRL_CS(cs) FIELD_PREP(GENMASK(1, 0), (cs))
+#define   MEM_CTRL_DIS_WP(cs) FIELD_PREP(GENMASK(11, 8), BIT((cs)))
+
+#define DATA_SIZE_REG 0x84
+#define   DATA_SIZE(x) FIELD_PREP(GENMASK(14, 0), (x))
+
+#define TIMINGS_ASYN_REG 0x88
+#define   TIMINGS_ASYN_TRWP(x) FIELD_PREP(GENMASK(3, 0), max((x), 1U) - 1)
+#define   TIMINGS_ASYN_TRWH(x) FIELD_PREP(GENMASK(7, 4), max((x), 1U) - 1)
+
+#define TIM_SEQ0_REG 0x90
+#define   TIM_SEQ0_TCCS(x) FIELD_PREP(GENMASK(5, 0), max((x), 1U) - 1)
+#define   TIM_SEQ0_TADL(x) FIELD_PREP(GENMASK(13, 8), max((x), 1U) - 1)
+#define   TIM_SEQ0_TRHW(x) FIELD_PREP(GENMASK(21, 16), max((x), 1U) - 1)
+#define   TIM_SEQ0_TWHR(x) FIELD_PREP(GENMASK(29, 24), max((x), 1U) - 1)
+
+#define TIM_SEQ1_REG 0x94
+#define   TIM_SEQ1_TWB(x) FIELD_PREP(GENMASK(5, 0), max((x), 1U) - 1)
+#define   TIM_SEQ1_TRR(x) FIELD_PREP(GENMASK(13, 8), max((x), 1U) - 1)
+#define   TIM_SEQ1_TWW(x) FIELD_PREP(GENMASK(21, 16), max((x), 1U) - 1)
+
+#define TIM_GEN_SEQ0_REG 0x98
+#define   TIM_GEN_SEQ0_D0(x) FIELD_PREP(GENMASK(5, 0), max((x), 1U) - 1)
+#define   TIM_GEN_SEQ0_D1(x) FIELD_PREP(GENMASK(13, 8), max((x), 1U) - 1)
+#define   TIM_GEN_SEQ0_D2(x) FIELD_PREP(GENMASK(21, 16), max((x), 1U) - 1)
+#define   TIM_GEN_SEQ0_D3(x) FIELD_PREP(GENMASK(29, 24), max((x), 1U) - 1)
+
+#define TIM_GEN_SEQ1_REG 0x9c
+#define   TIM_GEN_SEQ1_D4(x) FIELD_PREP(GENMASK(5, 0), max((x), 1U) - 1)
+#define   TIM_GEN_SEQ1_D5(x) FIELD_PREP(GENMASK(13, 8), max((x), 1U) - 1)
+#define   TIM_GEN_SEQ1_D6(x) FIELD_PREP(GENMASK(21, 16), max((x), 1U) - 1)
+#define   TIM_GEN_SEQ1_D7(x) FIELD_PREP(GENMASK(29, 24), max((x), 1U) - 1)
+
+#define TIM_GEN_SEQ2_REG 0xA0
+#define   TIM_GEN_SEQ2_D8(x) FIELD_PREP(GENMASK(5, 0), max((x), 1U) - 1)
+#define   TIM_GEN_SEQ2_D9(x) FIELD_PREP(GENMASK(13, 8), max((x), 1U) - 1)
+#define   TIM_GEN_SEQ2_D10(x) FIELD_PREP(GENMASK(21, 16), max((x), 1U) - 1)
+#define   TIM_GEN_SEQ2_D11(x) FIELD_PREP(GENMASK(29, 24), max((x), 1U) - 1)
+
+#define FIFO_INIT_REG 0xB4
+#define   FIFO_INIT BIT(0)
+
+#define FIFO_STATE_REG 0xB4
+#define   FIFO_STATE_R_EMPTY(reg) FIELD_GET(BIT(0), (reg))
+#define   FIFO_STATE_W_FULL(reg) FIELD_GET(BIT(1), (reg))
+#define   FIFO_STATE_C_EMPTY(reg) FIELD_GET(BIT(2), (reg))
+#define   FIFO_STATE_R_FULL(reg) FIELD_GET(BIT(6), (reg))
+#define   FIFO_STATE_W_EMPTY(reg) FIELD_GET(BIT(7), (reg))
+
+#define GEN_SEQ_CTRL_REG 0xB8
+#define   GEN_SEQ_CMD0_EN BIT(0)
+#define   GEN_SEQ_CMD1_EN BIT(1)
+#define   GEN_SEQ_CMD2_EN BIT(2)
+#define   GEN_SEQ_CMD3_EN BIT(3)
+#define   GEN_SEQ_COL_A0(x) FIELD_PREP(GENMASK(5, 4), min((x), 2U))
+#define   GEN_SEQ_COL_A1(x) FIELD_PREP(GENMASK(7, 6), min((x), 2U))
+#define   GEN_SEQ_ROW_A0(x) FIELD_PREP(GENMASK(9, 8), min((x), 3U))
+#define   GEN_SEQ_ROW_A1(x) FIELD_PREP(GENMASK(11, 10), min((x), 3U))
+#define   GEN_SEQ_DATA_EN BIT(12)
+#define   GEN_SEQ_DELAY_EN(x) FIELD_PREP(GENMASK(14, 13), (x))
+#define     GEN_SEQ_DELAY0_EN GEN_SEQ_DELAY_EN(1)
+#define     GEN_SEQ_DELAY1_EN GEN_SEQ_DELAY_EN(2)
+#define   GEN_SEQ_IMD_SEQ BIT(15)
+#define   GEN_SEQ_COMMAND_3(x) FIELD_PREP(GENMASK(26, 16), (x))
+
+#define DMA_TLVL_REG 0x114
+#define   DMA_TLVL(x) FIELD_PREP(GENMASK(7, 0), (x))
+#define   DMA_TLVL_MAX DMA_TLVL(0xFF)
+
+#define TIM_GEN_SEQ3_REG 0x134
+#define   TIM_GEN_SEQ3_D12(x) FIELD_PREP(GENMASK(5, 0), max((x), 1U) - 1)
+
+#define ECC_CNT_REG 0x14C
+#define   ECC_CNT(cs, reg) FIELD_GET(GENMASK(5, 0), (reg) >> ((cs) * 8))
+
+#define RZN1_CS_NUM 4
+
+#define TO_CYCLES64(ps, period_ns) ((unsigned int)DIV_ROUND_UP_ULL(div_u64(ps, 1000), \
+								   period_ns))
+
+struct rzn1_nand_chip_sel {
+	unsigned int cs;
+};
+
+struct rzn1_nand_chip {
+	struct nand_chip chip;
+	struct list_head node;
+	int selected_die;
+	u32 ctrl;
+	unsigned int nsels;
+	u32 control;
+	u32 ecc_ctrl;
+	u32 timings_asyn;
+	u32 tim_seq0;
+	u32 tim_seq1;
+	u32 tim_gen_seq0;
+	u32 tim_gen_seq1;
+	u32 tim_gen_seq2;
+	u32 tim_gen_seq3;
+	struct rzn1_nand_chip_sel sels[];
+};
+
+struct rzn1_nandc {
+	struct nand_controller controller;
+	struct device *dev;
+	void __iomem *regs;
+	struct clk *hclk;
+	struct clk *eclk;
+	unsigned long assigned_cs;
+	struct list_head chips;
+	struct nand_chip *selected_chip;
+	struct completion complete;
+	bool use_polling;
+	u8 *buf;
+	unsigned int buf_sz;
+};
+
+struct rzn1_op {
+	u32 command;
+	u32 addr0_col;
+	u32 addr0_row;
+	u32 addr1_col;
+	u32 addr1_row;
+	u32 data_size;
+	u32 ecc_offset;
+	u32 gen_seq_ctrl;
+	u8 *buf;
+	bool read;
+	unsigned int len;
+};
+
+static inline struct rzn1_nandc *to_rzn1_nandc(struct nand_controller *ctrl)
+{
+	return container_of(ctrl, struct rzn1_nandc, controller);
+}
+
+static inline struct rzn1_nand_chip *to_rzn1_nand(struct nand_chip *chip)
+{
+	return container_of(chip, struct rzn1_nand_chip, chip);
+}
+
+static inline unsigned int to_nandc_cs(struct rzn1_nand_chip *nand)
+{
+	return nand->sels[nand->selected_die].cs;
+}
+
+static void rzn1_nandc_dis_correction(struct rzn1_nandc *nandc)
+{
+	u32 control;
+
+	control = readl_relaxed(nandc->regs + CONTROL_REG);
+	control &= ~CONTROL_ECC_EN;
+	writel_relaxed(control, nandc->regs + CONTROL_REG);
+}
+
+static void rzn1_nandc_en_correction(struct rzn1_nandc *nandc)
+{
+	u32 control;
+
+	control = readl_relaxed(nandc->regs + CONTROL_REG);
+	control |= CONTROL_ECC_EN;
+	writel_relaxed(control, nandc->regs + CONTROL_REG);
+}
+
+static void rzn1_nandc_clear_status(struct rzn1_nandc *nandc)
+{
+	writel_relaxed(0, nandc->regs + INT_STATUS_REG);
+	writel_relaxed(0, nandc->regs + ECC_STAT_REG);
+	writel_relaxed(0, nandc->regs + ECC_CNT_REG);
+}
+
+static void rzn1_nandc_dis_interrupts(struct rzn1_nandc *nandc)
+{
+	writel_relaxed(0, nandc->regs + INT_MASK_REG);
+}
+
+static void rzn1_nandc_en_interrupts(struct rzn1_nandc *nandc, u32 val)
+{
+	if (!nandc->use_polling)
+		writel_relaxed(val, nandc->regs + INT_MASK_REG);
+}
+
+static void rzn1_nandc_clear_fifo(struct rzn1_nandc *nandc)
+{
+	writel_relaxed(FIFO_INIT, nandc->regs + FIFO_INIT_REG);
+}
+
+static void rzn1_nandc_select_target(struct nand_chip *chip, int die_nr)
+{
+	struct rzn1_nand_chip *rzn1_nand = to_rzn1_nand(chip);
+	struct rzn1_nandc *nandc = to_rzn1_nandc(chip->controller);
+	unsigned int cs = rzn1_nand->sels[die_nr].cs;
+
+	if (chip == nandc->selected_chip && die_nr == rzn1_nand->selected_die)
+		return;
+
+	rzn1_nandc_clear_status(nandc);
+	writel_relaxed(MEM_CTRL_CS(cs) | MEM_CTRL_DIS_WP(cs), nandc->regs + MEM_CTRL_REG);
+	writel_relaxed(rzn1_nand->control, nandc->regs + CONTROL_REG);
+	writel_relaxed(rzn1_nand->ecc_ctrl, nandc->regs + ECC_CTRL_REG);
+	writel_relaxed(rzn1_nand->timings_asyn, nandc->regs + TIMINGS_ASYN_REG);
+	writel_relaxed(rzn1_nand->tim_seq0, nandc->regs + TIM_SEQ0_REG);
+	writel_relaxed(rzn1_nand->tim_seq1, nandc->regs + TIM_SEQ1_REG);
+	writel_relaxed(rzn1_nand->tim_gen_seq0, nandc->regs + TIM_GEN_SEQ0_REG);
+	writel_relaxed(rzn1_nand->tim_gen_seq1, nandc->regs + TIM_GEN_SEQ1_REG);
+	writel_relaxed(rzn1_nand->tim_gen_seq2, nandc->regs + TIM_GEN_SEQ2_REG);
+	writel_relaxed(rzn1_nand->tim_gen_seq3, nandc->regs + TIM_GEN_SEQ3_REG);
+
+	nandc->selected_chip = chip;
+	rzn1_nand->selected_die = die_nr;
+}
+
+static void rzn1_nandc_trigger_op(struct rzn1_nandc *nandc, struct rzn1_op *rop)
+{
+	writel_relaxed(rop->addr0_col, nandc->regs + ADDR0_COL_REG);
+	writel_relaxed(rop->addr0_row, nandc->regs + ADDR0_ROW_REG);
+	writel_relaxed(rop->addr1_col, nandc->regs + ADDR1_COL_REG);
+	writel_relaxed(rop->addr1_row, nandc->regs + ADDR1_ROW_REG);
+	writel_relaxed(rop->ecc_offset, nandc->regs + ECC_OFFSET_REG);
+	writel_relaxed(rop->gen_seq_ctrl, nandc->regs + GEN_SEQ_CTRL_REG);
+	writel_relaxed(DATA_SIZE(rop->len), nandc->regs + DATA_SIZE_REG);
+	writel_relaxed(rop->command, nandc->regs + COMMAND_REG);
+}
+
+static void rzn1_nandc_trigger_dma(struct rzn1_nandc *nandc)
+{
+	writel_relaxed(DMA_CTRL_INCREMENT_BURST_4 |
+		       DMA_CTRL_REGISTER_MANAGED_MODE |
+		       DMA_CTRL_START, nandc->regs + DMA_CTRL_REG);
+}
+
+static irqreturn_t rzn1_nandc_irq_handler(int irq, void *private)
+{
+	struct rzn1_nandc *nandc = private;
+
+	rzn1_nandc_dis_interrupts(nandc);
+	complete(&nandc->complete);
+
+	return IRQ_HANDLED;
+}
+
+static int rzn1_nandc_wait_end_of_op(struct rzn1_nandc *nandc,
+				     struct nand_chip *chip)
+{
+	struct rzn1_nand_chip *rzn1_nand = to_rzn1_nand(chip);
+	unsigned int cs = to_nandc_cs(rzn1_nand);
+	u32 status;
+	int ret;
+
+	ret = readl_poll_timeout(nandc->regs + STATUS_REG, status,
+				 MEM_RDY(cs, status) && CTRL_RDY(status),
+				 1, 100000);
+	if (ret)
+		dev_err(nandc->dev, "Operation timed out, status: 0x%08x\n",
+			status);
+
+	return ret;
+}
+
+static int rzn1_nandc_wait_end_of_io(struct rzn1_nandc *nandc,
+				     struct nand_chip *chip)
+{
+	int timeout_ms = 1000;
+	int ret;
+
+	if (nandc->use_polling) {
+		struct rzn1_nand_chip *rzn1_nand = to_rzn1_nand(chip);
+		unsigned int cs = to_nandc_cs(rzn1_nand);
+		u32 status;
+
+		ret = readl_poll_timeout(nandc->regs + INT_STATUS_REG, status,
+					 MEM_IS_RDY(cs, status) &
+					 DMA_HAS_ENDED(status),
+					 0, timeout_ms * 1000);
+	} else {
+		ret = wait_for_completion_timeout(&nandc->complete,
+						  msecs_to_jiffies(timeout_ms));
+		if (!ret)
+			ret = -ETIMEDOUT;
+		else
+			ret = 0;
+	}
+
+	return ret;
+}
+
+static int rzn1_read_page_hw_ecc(struct nand_chip *chip, u8 *buf,
+				 int oob_required, int page)
+{
+	struct rzn1_nandc *nandc = to_rzn1_nandc(chip->controller);
+	struct mtd_info *mtd = nand_to_mtd(chip);
+	struct rzn1_nand_chip *rzn1_nand = to_rzn1_nand(chip);
+	unsigned int cs = to_nandc_cs(rzn1_nand);
+	struct rzn1_op rop = {
+		.command = COMMAND_INPUT_SEL_DMA | COMMAND_0(NAND_CMD_READ0) |
+			   COMMAND_2(NAND_CMD_READSTART) | COMMAND_FIFO_SEL |
+			   COMMAND_SEQ_READ_PAGE,
+		.addr0_row = page,
+		.len = mtd->writesize,
+		.ecc_offset = ECC_OFFSET(mtd->writesize + 2),
+	};
+	unsigned int max_bitflips = 0;
+	dma_addr_t dma_addr;
+	u32 ecc_stat;
+	int bf, ret, i;
+
+	/* Prepare controller */
+	rzn1_nandc_select_target(chip, chip->cur_cs);
+	rzn1_nandc_clear_status(nandc);
+	reinit_completion(&nandc->complete);
+	rzn1_nandc_en_interrupts(nandc, INT_DMA_ENDED);
+	rzn1_nandc_en_correction(nandc);
+
+	/* Configure DMA */
+	dma_addr = dma_map_single(nandc->dev, nandc->buf, mtd->writesize,
+				  DMA_FROM_DEVICE);
+	writel(dma_addr, nandc->regs + DMA_ADDR_LOW_REG);
+	writel(mtd->writesize, nandc->regs + DMA_CNT_REG);
+	writel(DMA_TLVL_MAX, nandc->regs + DMA_TLVL_REG);
+
+	rzn1_nandc_trigger_op(nandc, &rop);
+	rzn1_nandc_trigger_dma(nandc);
+
+	ret = rzn1_nandc_wait_end_of_io(nandc, chip);
+	dma_unmap_single(nandc->dev, dma_addr, mtd->writesize, DMA_FROM_DEVICE);
+	rzn1_nandc_dis_correction(nandc);
+	if (ret) {
+		dev_err(nandc->dev, "Read page operation never ending\n");
+		return ret;
+	}
+
+	ecc_stat = readl_relaxed(nandc->regs + ECC_STAT_REG);
+
+	if (oob_required || ECC_STAT_UNCORRECTABLE(cs, ecc_stat)) {
+		ret = nand_change_read_column_op(chip, mtd->writesize,
+						 chip->oob_poi, mtd->oobsize,
+						 false);
+		if (ret)
+			return ret;
+	}
+
+	if (ECC_STAT_UNCORRECTABLE(cs, ecc_stat)) {
+		for (i = 0; i < chip->ecc.steps; i++) {
+			unsigned int off = i * chip->ecc.size;
+			unsigned int eccoff = i * chip->ecc.bytes;
+
+			bf = nand_check_erased_ecc_chunk(nandc->buf + off,
+							 chip->ecc.size,
+							 chip->oob_poi + 2 + eccoff,
+							 chip->ecc.bytes,
+							 NULL, 0,
+							 chip->ecc.strength);
+			if (bf < 0) {
+				mtd->ecc_stats.failed++;
+			} else {
+				mtd->ecc_stats.corrected += bf;
+				max_bitflips = max_t(unsigned int, max_bitflips, bf);
+			}
+		}
+	} else if (ECC_STAT_CORRECTABLE(cs, ecc_stat)) {
+		bf = ECC_CNT(cs, readl_relaxed(nandc->regs + ECC_CNT_REG));
+		/*
+		 * The number of bitflips is an approximation given the fact
+		 * that this controller does not provide per-chunk details but
+		 * only gives statistics on the entire page.
+		 */
+		mtd->ecc_stats.corrected += bf;
+	}
+
+	memcpy(buf, nandc->buf, mtd->writesize);
+
+	return 0;
+}
+
+static int rzn1_read_subpage_hw_ecc(struct nand_chip *chip, u32 req_offset,
+				    u32 req_len, u8 *bufpoi, int page)
+{
+	struct rzn1_nandc *nandc = to_rzn1_nandc(chip->controller);
+	struct mtd_info *mtd = nand_to_mtd(chip);
+	struct rzn1_nand_chip *rzn1_nand = to_rzn1_nand(chip);
+	unsigned int cs = to_nandc_cs(rzn1_nand);
+	unsigned int page_off = round_down(req_offset, chip->ecc.size);
+	unsigned int real_len = round_up(req_offset + req_len - page_off,
+					 chip->ecc.size);
+	unsigned int start_chunk = page_off / chip->ecc.size;
+	unsigned int nchunks = real_len / chip->ecc.size;
+	unsigned int ecc_off = 2 + (start_chunk * chip->ecc.bytes);
+	struct rzn1_op rop = {
+		.command = COMMAND_INPUT_SEL_AHBS | COMMAND_0(NAND_CMD_READ0) |
+			   COMMAND_2(NAND_CMD_READSTART) | COMMAND_FIFO_SEL |
+			   COMMAND_SEQ_READ_PAGE,
+		.addr0_row = page,
+		.addr0_col = page_off,
+		.len = real_len,
+		.ecc_offset = ECC_OFFSET(mtd->writesize + ecc_off),
+	};
+	unsigned int max_bitflips = 0, i;
+	u32 ecc_stat;
+	int bf, ret;
+
+	/* Prepare controller */
+	rzn1_nandc_select_target(chip, chip->cur_cs);
+	rzn1_nandc_clear_status(nandc);
+	rzn1_nandc_en_correction(nandc);
+	rzn1_nandc_trigger_op(nandc, &rop);
+
+	while (!FIFO_STATE_C_EMPTY(readl(nandc->regs + FIFO_STATE_REG)))
+		cpu_relax();
+
+	while (FIFO_STATE_R_EMPTY(readl(nandc->regs + FIFO_STATE_REG)))
+		cpu_relax();
+
+	ioread32_rep(nandc->regs + FIFO_DATA_REG, bufpoi + page_off,
+		     real_len / 4);
+
+	if (!FIFO_STATE_R_EMPTY(readl(nandc->regs + FIFO_STATE_REG))) {
+		dev_err(nandc->dev, "Clearing residual data in the read FIFO\n");
+		rzn1_nandc_clear_fifo(nandc);
+	}
+
+	ret = rzn1_nandc_wait_end_of_op(nandc, chip);
+	rzn1_nandc_dis_correction(nandc);
+	if (ret) {
+		dev_err(nandc->dev, "Read subpage operation never ending\n");
+		return ret;
+	}
+
+	ecc_stat = readl_relaxed(nandc->regs + ECC_STAT_REG);
+
+	if (ECC_STAT_UNCORRECTABLE(cs, ecc_stat)) {
+		ret = nand_change_read_column_op(chip, mtd->writesize,
+						 chip->oob_poi, mtd->oobsize,
+						 false);
+		if (ret)
+			return ret;
+
+		for (i = start_chunk; i < nchunks; i++) {
+			unsigned int dataoff = i * chip->ecc.size;
+			unsigned int eccoff = 2 + (i * chip->ecc.bytes);
+
+			bf = nand_check_erased_ecc_chunk(bufpoi + dataoff,
+							 chip->ecc.size,
+							 chip->oob_poi + eccoff,
+							 chip->ecc.bytes,
+							 NULL, 0,
+							 chip->ecc.strength);
+			if (bf < 0) {
+				mtd->ecc_stats.failed++;
+			} else {
+				mtd->ecc_stats.corrected += bf;
+				max_bitflips = max_t(unsigned int, max_bitflips, bf);
+			}
+		}
+	} else if (ECC_STAT_CORRECTABLE(cs, ecc_stat)) {
+		bf = ECC_CNT(cs, readl_relaxed(nandc->regs + ECC_CNT_REG));
+		/*
+		 * The number of bitflips is an approximation given the fact
+		 * that this controller does not provide per-chunk details but
+		 * only gives statistics on the entire page.
+		 */
+		mtd->ecc_stats.corrected += bf;
+	}
+
+	return 0;
+}
+
+static int rzn1_write_page_hw_ecc(struct nand_chip *chip, const u8 *buf,
+				  int oob_required, int page)
+{
+	struct rzn1_nandc *nandc = to_rzn1_nandc(chip->controller);
+	struct mtd_info *mtd = nand_to_mtd(chip);
+	struct rzn1_nand_chip *rzn1_nand = to_rzn1_nand(chip);
+	unsigned int cs = to_nandc_cs(rzn1_nand);
+	struct rzn1_op rop = {
+		.command = COMMAND_INPUT_SEL_DMA | COMMAND_0(NAND_CMD_SEQIN) |
+			   COMMAND_1(NAND_CMD_PAGEPROG) | COMMAND_FIFO_SEL |
+			   COMMAND_SEQ_WRITE_PAGE,
+		.addr0_row = page,
+		.len = mtd->writesize,
+		.ecc_offset = ECC_OFFSET(mtd->writesize + 2),
+	};
+	dma_addr_t dma_addr;
+	int ret;
+
+	memcpy(nandc->buf, buf, mtd->writesize);
+
+	/* Prepare controller */
+	rzn1_nandc_select_target(chip, chip->cur_cs);
+	rzn1_nandc_clear_status(nandc);
+	reinit_completion(&nandc->complete);
+	rzn1_nandc_en_interrupts(nandc, INT_MEM_RDY(cs));
+	rzn1_nandc_en_correction(nandc);
+
+	/* Configure DMA */
+	dma_addr = dma_map_single(nandc->dev, (void *)nandc->buf, mtd->writesize,
+				  DMA_TO_DEVICE);
+	writel(dma_addr, nandc->regs + DMA_ADDR_LOW_REG);
+	writel(mtd->writesize, nandc->regs + DMA_CNT_REG);
+	writel(DMA_TLVL_MAX, nandc->regs + DMA_TLVL_REG);
+
+	rzn1_nandc_trigger_op(nandc, &rop);
+	rzn1_nandc_trigger_dma(nandc);
+
+	ret = rzn1_nandc_wait_end_of_io(nandc, chip);
+	dma_unmap_single(nandc->dev, dma_addr, mtd->writesize, DMA_TO_DEVICE);
+	rzn1_nandc_dis_correction(nandc);
+	if (ret) {
+		dev_err(nandc->dev, "Write page operation never ending\n");
+		return ret;
+	}
+
+	if (!oob_required)
+		return 0;
+
+	return nand_change_write_column_op(chip, mtd->writesize, chip->oob_poi,
+					   mtd->oobsize, false);
+}
+
+static int rzn1_write_subpage_hw_ecc(struct nand_chip *chip, u32 req_offset,
+				     u32 req_len, const u8 *bufpoi,
+				     int oob_required, int page)
+{
+	struct rzn1_nandc *nandc = to_rzn1_nandc(chip->controller);
+	struct mtd_info *mtd = nand_to_mtd(chip);
+	unsigned int page_off = round_down(req_offset, chip->ecc.size);
+	unsigned int real_len = round_up(req_offset + req_len - page_off,
+					 chip->ecc.size);
+	unsigned int start_chunk = page_off / chip->ecc.size;
+	unsigned int ecc_off = 2 + (start_chunk * chip->ecc.bytes);
+	struct rzn1_op rop = {
+		.command = COMMAND_INPUT_SEL_AHBS | COMMAND_0(NAND_CMD_SEQIN) |
+			   COMMAND_1(NAND_CMD_PAGEPROG) | COMMAND_FIFO_SEL |
+			   COMMAND_SEQ_WRITE_PAGE,
+		.addr0_row = page,
+		.addr0_col = page_off,
+		.len = real_len,
+		.ecc_offset = ECC_OFFSET(mtd->writesize + ecc_off),
+	};
+	int ret;
+
+	/* Prepare controller */
+	rzn1_nandc_select_target(chip, chip->cur_cs);
+	rzn1_nandc_clear_status(nandc);
+	rzn1_nandc_en_correction(nandc);
+	rzn1_nandc_trigger_op(nandc, &rop);
+
+	while (FIFO_STATE_W_FULL(readl(nandc->regs + FIFO_STATE_REG)))
+		cpu_relax();
+
+	iowrite32_rep(nandc->regs + FIFO_DATA_REG, bufpoi + page_off,
+		      real_len / 4);
+
+	while (!FIFO_STATE_W_EMPTY(readl(nandc->regs + FIFO_STATE_REG)))
+		cpu_relax();
+
+	ret = rzn1_nandc_wait_end_of_op(nandc, chip);
+	rzn1_nandc_dis_correction(nandc);
+	if (ret) {
+		dev_err(nandc->dev, "Write subpage operation never ending\n");
+		return ret;
+	}
+
+	return 0;
+}
+
+/*
+ * This controller is simple enough and thus does not need to use the parser
+ * provided by the core, instead, handle every situation here.
+ */
+static int rzn1_nandc_exec_op(struct nand_chip *chip,
+			      const struct nand_operation *op, bool check_only)
+{
+	struct rzn1_nandc *nandc = to_rzn1_nandc(chip->controller);
+	const struct nand_op_instr *instr = NULL;
+	struct rzn1_op rop = {
+		.command = COMMAND_INPUT_SEL_AHBS,
+		.gen_seq_ctrl = GEN_SEQ_IMD_SEQ,
+	};
+	unsigned int cmd_phase = 0, addr_phase = 0, data_phase = 0,
+		delay_phase = 0, delays = 0;
+	unsigned int op_id, col_addrs, row_addrs, naddrs, remainder, words, i;
+	const u8 *addrs;
+	u32 last_bytes;
+	int ret;
+
+	if (!check_only)
+		rzn1_nandc_select_target(chip, op->cs);
+
+	for (op_id = 0; op_id < op->ninstrs; op_id++) {
+		instr = &op->instrs[op_id];
+
+		nand_op_trace("  ", instr);
+
+		switch (instr->type) {
+		case NAND_OP_CMD_INSTR:
+			switch (cmd_phase++) {
+			case 0:
+				rop.command |= COMMAND_0(instr->ctx.cmd.opcode);
+				rop.gen_seq_ctrl |= GEN_SEQ_CMD0_EN;
+				break;
+			case 1:
+				rop.gen_seq_ctrl |= GEN_SEQ_COMMAND_3(instr->ctx.cmd.opcode);
+				rop.gen_seq_ctrl |= GEN_SEQ_CMD3_EN;
+				if (addr_phase == 0)
+					addr_phase = 1;
+				break;
+			case 2:
+				rop.command |= COMMAND_2(instr->ctx.cmd.opcode);
+				rop.gen_seq_ctrl |= GEN_SEQ_CMD2_EN;
+				if (addr_phase <= 1)
+					addr_phase = 2;
+				break;
+			case 3:
+				rop.command |= COMMAND_1(instr->ctx.cmd.opcode);
+				rop.gen_seq_ctrl |= GEN_SEQ_CMD1_EN;
+				if (addr_phase <= 1)
+					addr_phase = 2;
+				if (delay_phase == 0)
+					delay_phase = 1;
+				if (data_phase == 0)
+					data_phase = 1;
+				break;
+			default:
+				return -EOPNOTSUPP;
+			}
+			break;
+
+		case NAND_OP_ADDR_INSTR:
+			addrs = instr->ctx.addr.addrs;
+			naddrs = instr->ctx.addr.naddrs;
+			if (naddrs > 5)
+				return -EOPNOTSUPP;
+
+			col_addrs = min(2U, naddrs);
+			row_addrs = naddrs > 2 ? naddrs - col_addrs : 0;
+
+			switch (addr_phase++) {
+			case 0:
+				for (i = 0; i < col_addrs; i++)
+					rop.addr0_col |= addrs[i] << (i * 8);
+				rop.gen_seq_ctrl |= GEN_SEQ_COL_A0(col_addrs);
+
+				for (i = 0; i < row_addrs; i++)
+					rop.addr0_row |= addrs[2 + i] << (i * 8);
+				rop.gen_seq_ctrl |= GEN_SEQ_ROW_A0(row_addrs);
+
+				if (cmd_phase == 0)
+					cmd_phase = 1;
+				break;
+			case 1:
+				for (i = 0; i < col_addrs; i++)
+					rop.addr1_col |= addrs[i] << (i * 8);
+				rop.gen_seq_ctrl |= GEN_SEQ_COL_A1(col_addrs);
+
+				for (i = 0; i < row_addrs; i++)
+					rop.addr1_row |= addrs[2 + i] << (i * 8);
+				rop.gen_seq_ctrl |= GEN_SEQ_ROW_A1(row_addrs);
+
+				if (cmd_phase <= 1)
+					cmd_phase = 2;
+				break;
+			default:
+				return -EOPNOTSUPP;
+			}
+			break;
+
+		case NAND_OP_DATA_IN_INSTR:
+			rop.read = true;
+			fallthrough;
+		case NAND_OP_DATA_OUT_INSTR:
+			rop.gen_seq_ctrl |= GEN_SEQ_DATA_EN;
+			rop.buf = instr->ctx.data.buf.in;
+			rop.len = instr->ctx.data.len;
+			rop.command |= COMMAND_FIFO_SEL;
+
+			switch (data_phase++) {
+			case 0:
+				if (cmd_phase <= 2)
+					cmd_phase = 3;
+				if (addr_phase <= 1)
+					addr_phase = 2;
+				if (delay_phase == 0)
+					delay_phase = 1;
+				break;
+			default:
+				return -EOPNOTSUPP;
+			}
+			break;
+
+		case NAND_OP_WAITRDY_INSTR:
+			switch (delay_phase++) {
+			case 0:
+				rop.gen_seq_ctrl |= GEN_SEQ_DELAY0_EN;
+
+				if (cmd_phase <= 2)
+					cmd_phase = 3;
+				break;
+			case 1:
+				rop.gen_seq_ctrl |= GEN_SEQ_DELAY1_EN;
+
+				if (cmd_phase <= 3)
+					cmd_phase = 4;
+				if (data_phase == 0)
+					data_phase = 1;
+				break;
+			default:
+				return -EOPNOTSUPP;
+			}
+			break;
+		}
+	}
+
+	/*
+	 * Sequence 19 is generic and dedicated to write operations.
+	 * Sequence 18 is also generic and works for all other operations.
+	 */
+	if (rop.buf && !rop.read)
+		rop.command |= COMMAND_SEQ_GEN_OUT;
+	else
+		rop.command |= COMMAND_SEQ_GEN_IN;
+
+	if (delays > 1) {
+		dev_err(nandc->dev, "Cannot handle more than one wait delay\n");
+		return -EOPNOTSUPP;
+	}
+
+	if (check_only)
+		return 0;
+
+	rzn1_nandc_trigger_op(nandc, &rop);
+
+	words = rop.len / sizeof(u32);
+	remainder = rop.len % sizeof(u32);
+	if (rop.buf && rop.read) {
+		while (!FIFO_STATE_C_EMPTY(readl(nandc->regs + FIFO_STATE_REG)))
+			cpu_relax();
+
+		while (FIFO_STATE_R_EMPTY(readl(nandc->regs + FIFO_STATE_REG)))
+			cpu_relax();
+
+		ioread32_rep(nandc->regs + FIFO_DATA_REG, rop.buf, words);
+		if (remainder) {
+			last_bytes = readl_relaxed(nandc->regs + FIFO_DATA_REG);
+			memcpy(rop.buf + (words * sizeof(u32)), &last_bytes,
+			       remainder);
+		}
+
+		if (!FIFO_STATE_R_EMPTY(readl(nandc->regs + FIFO_STATE_REG))) {
+			dev_warn(nandc->dev,
+				 "Clearing residual data in the read FIFO\n");
+			rzn1_nandc_clear_fifo(nandc);
+		}
+	} else if (rop.len && !rop.read) {
+		while (FIFO_STATE_W_FULL(readl(nandc->regs + FIFO_STATE_REG)))
+			cpu_relax();
+
+		iowrite32_rep(nandc->regs + FIFO_DATA_REG, rop.buf,
+			      DIV_ROUND_UP(rop.len, 4));
+
+		if (remainder) {
+			last_bytes = 0;
+			memcpy(&last_bytes, rop.buf + (words * sizeof(u32)), remainder);
+			writel_relaxed(last_bytes, nandc->regs + FIFO_DATA_REG);
+		}
+
+		while (!FIFO_STATE_W_EMPTY(readl(nandc->regs + FIFO_STATE_REG)))
+			cpu_relax();
+	}
+
+	ret = rzn1_nandc_wait_end_of_op(nandc, chip);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
+static int rzn1_nandc_setup_interface(struct nand_chip *chip, int chipnr,
+				      const struct nand_interface_config *conf)
+{
+	struct rzn1_nand_chip *rzn1_nand = to_rzn1_nand(chip);
+	struct rzn1_nandc *nandc = to_rzn1_nandc(chip->controller);
+	unsigned int period_ns = 1000000000 / clk_get_rate(nandc->eclk);
+	const struct nand_sdr_timings *sdr;
+	unsigned int cyc, cle, ale, bef_dly, ca_to_data;
+
+	sdr = nand_get_sdr_timings(conf);
+	if (IS_ERR(sdr))
+		return PTR_ERR(sdr);
+
+	if (sdr->tRP_min != sdr->tWP_min || sdr->tREH_min != sdr->tWH_min) {
+		dev_err(nandc->dev, "Read and write hold times must be identical\n");
+		return -EINVAL;
+	}
+
+	if (chipnr < 0)
+		return 0;
+
+	rzn1_nand->timings_asyn =
+		TIMINGS_ASYN_TRWP(TO_CYCLES64(sdr->tRP_min, period_ns)) |
+		TIMINGS_ASYN_TRWH(TO_CYCLES64(sdr->tREH_min, period_ns));
+	rzn1_nand->tim_seq0 =
+		TIM_SEQ0_TCCS(TO_CYCLES64(sdr->tCCS_min, period_ns)) |
+		TIM_SEQ0_TADL(TO_CYCLES64(sdr->tADL_min, period_ns)) |
+		TIM_SEQ0_TRHW(TO_CYCLES64(sdr->tRHW_min, period_ns)) |
+		TIM_SEQ0_TWHR(TO_CYCLES64(sdr->tWHR_min, period_ns));
+	rzn1_nand->tim_seq1 =
+		TIM_SEQ1_TWB(TO_CYCLES64(sdr->tWB_max, period_ns)) |
+		TIM_SEQ1_TRR(TO_CYCLES64(sdr->tRR_min, period_ns)) |
+		TIM_SEQ1_TWW(TO_CYCLES64(sdr->tWW_min, period_ns));
+
+	cyc = sdr->tDS_min + sdr->tDH_min;
+	cle = sdr->tCLH_min + sdr->tCLS_min;
+	ale = sdr->tALH_min + sdr->tALS_min;
+	bef_dly = sdr->tWB_max - sdr->tDH_min;
+	ca_to_data = sdr->tWHR_min + sdr->tREA_max - sdr->tDH_min;
+
+	/*
+	 * D0 = CMD -> ADDR = tCLH + tCLS - 1 cycle
+	 * D1 = CMD -> CMD = tCLH + tCLS - 1 cycle
+	 * D2 = CMD -> DLY = tWB - tDH
+	 * D3 = CMD -> DATA = tWHR + tREA - tDH
+	 */
+	rzn1_nand->tim_gen_seq0 =
+		TIM_GEN_SEQ0_D0(TO_CYCLES64(cle - cyc, period_ns)) |
+		TIM_GEN_SEQ0_D1(TO_CYCLES64(cle - cyc, period_ns)) |
+		TIM_GEN_SEQ0_D2(TO_CYCLES64(bef_dly, period_ns)) |
+		TIM_GEN_SEQ0_D3(TO_CYCLES64(ca_to_data, period_ns));
+
+	/*
+	 * D4 = ADDR -> CMD = tALH + tALS - 1 cyle
+	 * D5 = ADDR -> ADDR = tALH + tALS - 1 cyle
+	 * D6 = ADDR -> DLY = tWB - tDH
+	 * D7 = ADDR -> DATA = tWHR + tREA - tDH
+	 */
+	rzn1_nand->tim_gen_seq1 =
+		TIM_GEN_SEQ1_D4(TO_CYCLES64(ale - cyc, period_ns)) |
+		TIM_GEN_SEQ1_D5(TO_CYCLES64(ale - cyc, period_ns)) |
+		TIM_GEN_SEQ1_D6(TO_CYCLES64(bef_dly, period_ns)) |
+		TIM_GEN_SEQ1_D7(TO_CYCLES64(ca_to_data, period_ns));
+
+	/*
+	 * D8 = DLY -> DATA = tRR + tREA
+	 * D9 = DLY -> CMD = tRR
+	 * D10 = DATA -> CMD = tCLH + tCLS - 1 cycle
+	 * D11 = DATA -> DLY = tWB - tDH
+	 */
+	rzn1_nand->tim_gen_seq2 =
+		TIM_GEN_SEQ2_D8(TO_CYCLES64(sdr->tRR_min + sdr->tREA_max, period_ns)) |
+		TIM_GEN_SEQ2_D9(TO_CYCLES64(sdr->tRR_min, period_ns)) |
+		TIM_GEN_SEQ2_D10(TO_CYCLES64(cle - cyc, period_ns)) |
+		TIM_GEN_SEQ2_D11(TO_CYCLES64(bef_dly, period_ns));
+
+	/* D12 = DATA -> END = tCLH - tDH */
+	rzn1_nand->tim_gen_seq3 =
+		TIM_GEN_SEQ3_D12(TO_CYCLES64(sdr->tCLH_min - sdr->tDH_min, period_ns));
+
+	return 0;
+}
+
+static int rzn1_nandc_ooblayout_ecc(struct mtd_info *mtd, int section,
+				    struct mtd_oob_region *oobregion)
+{
+	struct nand_chip *chip = mtd_to_nand(mtd);
+	unsigned int eccbytes = round_up(chip->ecc.bytes, 4) * chip->ecc.steps;
+
+	if (section)
+		return -ERANGE;
+
+	oobregion->offset = 2;
+	oobregion->length = eccbytes;
+
+	return 0;
+}
+
+static int rzn1_nandc_ooblayout_free(struct mtd_info *mtd, int section,
+				     struct mtd_oob_region *oobregion)
+{
+	struct nand_chip *chip = mtd_to_nand(mtd);
+	unsigned int eccbytes = round_up(chip->ecc.bytes, 4) * chip->ecc.steps;
+
+	if (section)
+		return -ERANGE;
+
+	oobregion->offset = 2 + eccbytes;
+	oobregion->length = mtd->oobsize - oobregion->offset;
+
+	return 0;
+}
+
+static const struct mtd_ooblayout_ops rzn1_nandc_ooblayout_ops = {
+	.ecc = rzn1_nandc_ooblayout_ecc,
+	.free = rzn1_nandc_ooblayout_free,
+};
+
+static int rzn1_nandc_hw_ecc_controller_init(struct nand_chip *chip)
+{
+	struct rzn1_nand_chip *rzn1_nand = to_rzn1_nand(chip);
+	struct mtd_info *mtd = nand_to_mtd(chip);
+	struct rzn1_nandc *nandc = to_rzn1_nandc(chip->controller);
+
+	if (mtd->writesize > SZ_16K) {
+		dev_err(nandc->dev, "Unsupported page size\n");
+		return -EINVAL;
+	}
+
+	switch (chip->ecc.size) {
+	case SZ_256:
+		rzn1_nand->control |= CONTROL_ECC_BLOCK_SIZE_256;
+		break;
+	case SZ_512:
+		rzn1_nand->control |= CONTROL_ECC_BLOCK_SIZE_512;
+		break;
+	case SZ_1K:
+		rzn1_nand->control |= CONTROL_ECC_BLOCK_SIZE_1024;
+		break;
+	default:
+		dev_err(nandc->dev, "Unsupported ECC chunk size\n");
+		return -EINVAL;
+	}
+
+	switch (chip->ecc.strength) {
+	case 2:
+		chip->ecc.bytes = 4;
+		rzn1_nand->ecc_ctrl |= ECC_CTRL_CAP_2B;
+		break;
+	case 4:
+		chip->ecc.bytes = 7;
+		rzn1_nand->ecc_ctrl |= ECC_CTRL_CAP_4B;
+		break;
+	case 8:
+		chip->ecc.bytes = 14;
+		rzn1_nand->ecc_ctrl |= ECC_CTRL_CAP_8B;
+		break;
+	case 16:
+		chip->ecc.bytes = 28;
+		rzn1_nand->ecc_ctrl |= ECC_CTRL_CAP_16B;
+		break;
+	case 24:
+		chip->ecc.bytes = 42;
+		rzn1_nand->ecc_ctrl |= ECC_CTRL_CAP_24B;
+		break;
+	case 32:
+		chip->ecc.bytes = 56;
+		rzn1_nand->ecc_ctrl |= ECC_CTRL_CAP_32B;
+		break;
+	default:
+		dev_err(nandc->dev, "Unsupported ECC strength\n");
+		return -EINVAL;
+	}
+
+	rzn1_nand->ecc_ctrl |= ECC_CTRL_ERR_THRESHOLD(chip->ecc.strength);
+
+	mtd_set_ooblayout(mtd, &rzn1_nandc_ooblayout_ops);
+	chip->ecc.steps = mtd->writesize / chip->ecc.size;
+	chip->ecc.read_page = rzn1_read_page_hw_ecc;
+	chip->ecc.read_subpage = rzn1_read_subpage_hw_ecc;
+	chip->ecc.write_page = rzn1_write_page_hw_ecc;
+	chip->ecc.write_subpage = rzn1_write_subpage_hw_ecc;
+
+	return 0;
+}
+
+static int rzn1_nandc_ecc_init(struct nand_chip *chip)
+{
+	struct nand_ecc_ctrl *ecc = &chip->ecc;
+	const struct nand_ecc_props *requirements =
+		nanddev_get_ecc_requirements(&chip->base);
+	struct rzn1_nandc *nandc = to_rzn1_nandc(chip->controller);
+	int ret;
+
+	if (ecc->engine_type != NAND_ECC_ENGINE_TYPE_NONE &&
+	    (!ecc->size || !ecc->strength)) {
+		if (requirements->step_size && requirements->strength) {
+			ecc->size = requirements->step_size;
+			ecc->strength = requirements->strength;
+		} else {
+			dev_err(nandc->dev, "No minimum ECC strength\n");
+			return -EINVAL;
+		}
+	}
+
+	switch (ecc->engine_type) {
+	case NAND_ECC_ENGINE_TYPE_ON_HOST:
+		ret = rzn1_nandc_hw_ecc_controller_init(chip);
+		if (ret)
+			return ret;
+		break;
+	case NAND_ECC_ENGINE_TYPE_NONE:
+	case NAND_ECC_ENGINE_TYPE_SOFT:
+	case NAND_ECC_ENGINE_TYPE_ON_DIE:
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int rzn1_nandc_attach_chip(struct nand_chip *chip)
+{
+	struct rzn1_nand_chip *rzn1_nand = to_rzn1_nand(chip);
+	struct rzn1_nandc *nandc = to_rzn1_nandc(chip->controller);
+	struct mtd_info *mtd = nand_to_mtd(chip);
+	struct nand_memory_organization *memorg = nanddev_get_memorg(&chip->base);
+	int ret;
+
+	/* Do not store BBT bits in the OOB section as it is not protected */
+	if (chip->bbt_options & NAND_BBT_USE_FLASH)
+		chip->bbt_options |= NAND_BBT_NO_OOB;
+
+	if (mtd->writesize <= 512) {
+		dev_err(nandc->dev, "Small page devices not supported\n");
+		return -EINVAL;
+	}
+
+	rzn1_nand->control |= CONTROL_CHECK_RB_LINE | CONTROL_INT_EN;
+
+	switch (memorg->pages_per_eraseblock) {
+	case 32:
+		rzn1_nand->control |= CONTROL_BLOCK_SIZE_32P;
+		break;
+	case 64:
+		rzn1_nand->control |= CONTROL_BLOCK_SIZE_64P;
+		break;
+	case 128:
+		rzn1_nand->control |= CONTROL_BLOCK_SIZE_128P;
+		break;
+	case 256:
+		rzn1_nand->control |= CONTROL_BLOCK_SIZE_256P;
+		break;
+	default:
+		dev_err(nandc->dev, "Unsupported memory organization\n");
+		return -EINVAL;
+	}
+
+	chip->options |= NAND_SUBPAGE_READ;
+
+	ret = rzn1_nandc_ecc_init(chip);
+	if (ret) {
+		dev_err(nandc->dev, "ECC initialization failed (%d)\n", ret);
+		return ret;
+	}
+
+	/* Force an update of the configuration registers */
+	rzn1_nand->selected_die = -1;
+
+	return 0;
+}
+
+static const struct nand_controller_ops rzn1_nandc_ops = {
+	.attach_chip = rzn1_nandc_attach_chip,
+	.exec_op = rzn1_nandc_exec_op,
+	.setup_interface = rzn1_nandc_setup_interface,
+};
+
+static int rzn1_nandc_alloc_dma_buf(struct rzn1_nandc *nandc,
+				    struct mtd_info *new_mtd)
+{
+	unsigned int max_len = new_mtd->writesize + new_mtd->oobsize;
+	struct rzn1_nand_chip *entry, *temp;
+	struct nand_chip *chip;
+	struct mtd_info *mtd;
+
+	list_for_each_entry_safe(entry, temp, &nandc->chips, node) {
+		chip = &entry->chip;
+		mtd = nand_to_mtd(chip);
+		max_len = max(max_len, mtd->writesize + mtd->oobsize);
+	}
+
+	if (nandc->buf && nandc->buf_sz < max_len) {
+		devm_kfree(nandc->dev, nandc->buf);
+		nandc->buf = NULL;
+	}
+
+	if (!nandc->buf) {
+		nandc->buf_sz = max_len;
+		nandc->buf = devm_kmalloc(nandc->dev, max_len, GFP_KERNEL | GFP_DMA);
+		if (!nandc->buf)
+			return -ENOMEM;
+	}
+
+	return 0;
+}
+
+static int rzn1_nandc_chip_init(struct rzn1_nandc *nandc, struct device_node *np)
+{
+	struct rzn1_nand_chip *rzn1_nand;
+	struct mtd_info *mtd;
+	struct nand_chip *chip;
+	int nsels, ret, i;
+	u32 cs;
+
+	nsels = of_property_count_elems_of_size(np, "reg", sizeof(u32));
+	if (nsels <= 0) {
+		ret = (nsels < 0) ? nsels : -EINVAL;
+		dev_err(nandc->dev, "Invalid reg property (%d)\n", ret);
+		return ret;
+	}
+
+	/* Alloc the driver's NAND chip structure */
+	rzn1_nand = devm_kzalloc(nandc->dev, struct_size(rzn1_nand, sels, nsels),
+				 GFP_KERNEL);
+	if (!rzn1_nand)
+		return -ENOMEM;
+
+	rzn1_nand->nsels = nsels;
+	rzn1_nand->selected_die = -1;
+
+	for (i = 0; i < nsels; i++) {
+		ret = of_property_read_u32_index(np, "reg", i, &cs);
+		if (ret) {
+			dev_err(nandc->dev, "Incomplete reg property (%d)\n", ret);
+			return ret;
+		}
+
+		if (cs >= RZN1_CS_NUM) {
+			dev_err(nandc->dev, "Invalid reg property (%d)\n", cs);
+			return -EINVAL;
+		}
+
+		if (test_and_set_bit(cs, &nandc->assigned_cs)) {
+			dev_err(nandc->dev, "CS %d already assigned\n", cs);
+			return -EINVAL;
+		}
+
+		/*
+		 * No need to check for RB or WP properties, there is a 1:1
+		 * mandatory mapping with the CS.
+		 */
+		rzn1_nand->sels[i].cs = cs;
+	}
+
+	chip = &rzn1_nand->chip;
+	chip->controller = &nandc->controller;
+	nand_set_flash_node(chip, np);
+
+	mtd = nand_to_mtd(chip);
+	mtd->dev.parent = nandc->dev;
+	if (!mtd->name) {
+		dev_err(nandc->dev, "Missing MTD label\n");
+		return -EINVAL;
+	}
+
+	ret = nand_scan(chip, rzn1_nand->nsels);
+	if (ret) {
+		dev_err(nandc->dev, "Failed to scan the NAND chip (%d)\n", ret);
+		return ret;
+	}
+
+	ret = rzn1_nandc_alloc_dma_buf(nandc, mtd);
+	if (ret)
+		goto cleanup_nand;
+
+	ret = mtd_device_register(mtd, NULL, 0);
+	if (ret) {
+		dev_err(nandc->dev, "Failed to register MTD device (%d)\n", ret);
+		goto cleanup_nand;
+	}
+
+	list_add_tail(&rzn1_nand->node, &nandc->chips);
+
+	return 0;
+
+cleanup_nand:
+	nand_cleanup(chip);
+
+	return ret;
+}
+
+static void rzn1_nandc_chips_cleanup(struct rzn1_nandc *nandc)
+{
+	struct rzn1_nand_chip *entry, *temp;
+	struct nand_chip *chip;
+	int ret;
+
+	list_for_each_entry_safe(entry, temp, &nandc->chips, node) {
+		chip = &entry->chip;
+		ret = mtd_device_unregister(nand_to_mtd(chip));
+		WARN_ON(ret);
+		nand_cleanup(chip);
+		list_del(&entry->node);
+	}
+}
+
+static int rzn1_nandc_chips_init(struct rzn1_nandc *nandc)
+{
+	struct device_node *np;
+	int ret;
+
+	for_each_child_of_node(nandc->dev->of_node, np) {
+		ret = rzn1_nandc_chip_init(nandc, np);
+		if (ret) {
+			of_node_put(np);
+			goto cleanup_chips;
+		}
+	}
+
+	return 0;
+
+cleanup_chips:
+	rzn1_nandc_chips_cleanup(nandc);
+
+	return ret;
+}
+
+static int rzn1_nandc_probe(struct platform_device *pdev)
+{
+	struct rzn1_nandc *nandc;
+	int irq, ret;
+
+	nandc = devm_kzalloc(&pdev->dev, sizeof(*nandc), GFP_KERNEL);
+	if (!nandc)
+		return -ENOMEM;
+
+	nandc->dev = &pdev->dev;
+	nand_controller_init(&nandc->controller);
+	nandc->controller.ops = &rzn1_nandc_ops;
+	INIT_LIST_HEAD(&nandc->chips);
+	init_completion(&nandc->complete);
+
+	nandc->regs = devm_platform_ioremap_resource(pdev, 0);
+	if (IS_ERR(nandc->regs))
+		return PTR_ERR(nandc->regs);
+
+	nandc->hclk = devm_clk_get(&pdev->dev, "hclk");
+	if (IS_ERR(nandc->hclk))
+		return PTR_ERR(nandc->hclk);
+
+	nandc->eclk = devm_clk_get(&pdev->dev, "eclk");
+	if (IS_ERR(nandc->eclk))
+		return PTR_ERR(nandc->eclk);
+
+	ret = clk_prepare_enable(nandc->hclk);
+	if (ret)
+		return ret;
+
+	ret = clk_prepare_enable(nandc->eclk);
+	if (ret)
+		goto disable_hclk;
+
+	rzn1_nandc_dis_interrupts(nandc);
+	irq = platform_get_irq_optional(pdev, 0);
+	if (irq == -EPROBE_DEFER) {
+		ret = irq;
+		goto disable_eclk;
+	} else if (irq < 0) {
+		dev_info(&pdev->dev, "No IRQ found, fallback to polling\n");
+		nandc->use_polling = true;
+	} else {
+		ret = devm_request_irq(&pdev->dev, irq, rzn1_nandc_irq_handler, 0,
+				       "rzn1-nand-controller", nandc);
+		if (ret < 0)
+			goto disable_eclk;
+	}
+
+	ret = dma_set_mask(&pdev->dev, DMA_BIT_MASK(32));
+	if (ret)
+		goto disable_eclk;
+
+	rzn1_nandc_clear_fifo(nandc);
+
+	platform_set_drvdata(pdev, nandc);
+
+	ret = rzn1_nandc_chips_init(nandc);
+	if (ret)
+		goto disable_eclk;
+
+	return 0;
+
+disable_eclk:
+	clk_disable_unprepare(nandc->eclk);
+disable_hclk:
+	clk_disable_unprepare(nandc->hclk);
+
+	return ret;
+}
+
+static int rzn1_nandc_remove(struct platform_device *pdev)
+{
+	struct rzn1_nandc *nandc = platform_get_drvdata(pdev);
+
+	rzn1_nandc_chips_cleanup(nandc);
+
+	clk_disable_unprepare(nandc->eclk);
+	clk_disable_unprepare(nandc->hclk);
+
+	return 0;
+}
+
+static const struct of_device_id rzn1_nandc_id_table[] = {
+	{ .compatible = "renesas,rzn1-nand-controller" },
+	{} /* sentinel */
+};
+MODULE_DEVICE_TABLE(of, nandc_id_table);
+
+static struct platform_driver rzn1_nandc_driver = {
+	.driver = {
+		.name = "renesas-nandc",
+		.of_match_table = of_match_ptr(rzn1_nandc_id_table),
+	},
+	.probe = rzn1_nandc_probe,
+	.remove = rzn1_nandc_remove,
+};
+module_platform_driver(rzn1_nandc_driver);
+
+MODULE_AUTHOR("Miquel Raynal <miquel.raynal at bootlin.com>");
+MODULE_DESCRIPTION("Renesas RZ/N1x NAND flash controller driver");
+MODULE_LICENSE("GPL");
-- 
2.27.0




More information about the linux-arm-kernel mailing list