This is the normal Tegra SPI driver modified to work with the QSPI controller in Tegra210. It does not do 2x/4x transfers or any other QSPI protocol.
Author: Yen Lin yelin@nvidia.com Signed-off-by: Yen Lin yelin@nvidia.com Signed-off-by: Tom Warren twarren@nvidia.com --- Changes in v2: - Drop defconfig and pinmux files, this is a driver-only patch. - When pinmux tables have been updated for P2371/P2571, another - patch will be sent to enable the QSPI driver on those boards.
drivers/spi/Kconfig | 5 + drivers/spi/Makefile | 1 + drivers/spi/tegra210_qspi.c | 411 ++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 417 insertions(+) create mode 100644 drivers/spi/tegra210_qspi.c
diff --git a/drivers/spi/Kconfig b/drivers/spi/Kconfig index 8e04fce..168f31d 100644 --- a/drivers/spi/Kconfig +++ b/drivers/spi/Kconfig @@ -115,6 +115,11 @@ config TEGRA20_SLINK be used to access the SPI NOR flash on platforms embedding this nVidia Tegra20/Tegra30 IP cores.
+config TEGRA210_QSPI + bool "nVidia Tegra210 QSPI driver" + help + Enable the Tegra Quad-SPI (QSPI) driver for T210. + config XILINX_SPI bool "Xilinx SPI driver" help diff --git a/drivers/spi/Makefile b/drivers/spi/Makefile index de241be..209a41e 100644 --- a/drivers/spi/Makefile +++ b/drivers/spi/Makefile @@ -46,6 +46,7 @@ obj-$(CONFIG_SH_QSPI) += sh_qspi.o obj-$(CONFIG_TEGRA114_SPI) += tegra114_spi.o obj-$(CONFIG_TEGRA20_SFLASH) += tegra20_sflash.o obj-$(CONFIG_TEGRA20_SLINK) += tegra20_slink.o +obj-$(CONFIG_TEGRA210_QSPI) += tegra210_qspi.o obj-$(CONFIG_TI_QSPI) += ti_qspi.o obj-$(CONFIG_XILINX_SPI) += xilinx_spi.o obj-$(CONFIG_ZYNQ_SPI) += zynq_spi.o diff --git a/drivers/spi/tegra210_qspi.c b/drivers/spi/tegra210_qspi.c new file mode 100644 index 0000000..876034d --- /dev/null +++ b/drivers/spi/tegra210_qspi.c @@ -0,0 +1,411 @@ +/* + * NVIDIA Tegra210 QSPI controller driver + * (C) Copyright 2015 + * NVIDIA Corporation <www.nvidia.com> + * + * SPDX-License-Identifier: GPL-2.0+ + */ + +#include <common.h> +#include <dm.h> +#include <asm/io.h> +#include <asm/arch/clock.h> +#include <asm/arch-tegra/clk_rst.h> +#include <spi.h> +#include <fdtdec.h> +#include "tegra_spi.h" + +DECLARE_GLOBAL_DATA_PTR; + +/* COMMAND1 */ +#define QSPI_CMD1_GO (1 << 31) +#define QSPI_CMD1_M_S (1 << 30) +#define QSPI_CMD1_MODE_MASK 0x3 +#define QSPI_CMD1_MODE_SHIFT 28 +#define QSPI_CMD1_CS_SEL_MASK 0x3 +#define QSPI_CMD1_CS_SEL_SHIFT 26 +#define QSPI_CMD1_CS_POL_INACTIVE0 (1 << 22) +#define QSPI_CMD1_CS_SW_HW (1 << 21) +#define QSPI_CMD1_CS_SW_VAL (1 << 20) +#define QSPI_CMD1_IDLE_SDA_MASK 0x3 +#define QSPI_CMD1_IDLE_SDA_SHIFT 18 +#define QSPI_CMD1_BIDIR (1 << 17) +#define QSPI_CMD1_LSBI_FE (1 << 16) +#define QSPI_CMD1_LSBY_FE (1 << 15) +#define QSPI_CMD1_BOTH_EN_BIT (1 << 14) +#define QSPI_CMD1_BOTH_EN_BYTE (1 << 13) +#define QSPI_CMD1_RX_EN (1 << 12) +#define QSPI_CMD1_TX_EN (1 << 11) +#define QSPI_CMD1_PACKED (1 << 5) +#define QSPI_CMD1_BITLEN_MASK 0x1F +#define QSPI_CMD1_BITLEN_SHIFT 0 + +/* COMMAND2 */ +#define QSPI_CMD2_TX_CLK_TAP_DELAY (1 << 6) +#define QSPI_CMD2_TX_CLK_TAP_DELAY_MASK (0x3F << 6) +#define QSPI_CMD2_RX_CLK_TAP_DELAY (1 << 0) +#define QSPI_CMD2_RX_CLK_TAP_DELAY_MASK (0x3F << 0) + +/* TRANSFER STATUS */ +#define QSPI_XFER_STS_RDY (1 << 30) + +/* FIFO STATUS */ +#define QSPI_FIFO_STS_CS_INACTIVE (1 << 31) +#define QSPI_FIFO_STS_FRAME_END (1 << 30) +#define QSPI_FIFO_STS_RX_FIFO_FLUSH (1 << 15) +#define QSPI_FIFO_STS_TX_FIFO_FLUSH (1 << 14) +#define QSPI_FIFO_STS_ERR (1 << 8) +#define QSPI_FIFO_STS_TX_FIFO_OVF (1 << 7) +#define QSPI_FIFO_STS_TX_FIFO_UNR (1 << 6) +#define QSPI_FIFO_STS_RX_FIFO_OVF (1 << 5) +#define QSPI_FIFO_STS_RX_FIFO_UNR (1 << 4) +#define QSPI_FIFO_STS_TX_FIFO_FULL (1 << 3) +#define QSPI_FIFO_STS_TX_FIFO_EMPTY (1 << 2) +#define QSPI_FIFO_STS_RX_FIFO_FULL (1 << 1) +#define QSPI_FIFO_STS_RX_FIFO_EMPTY (1 << 0) + +#define QSPI_TIMEOUT 1000 + +struct qspi_regs { + u32 command1; /* 000:QSPI_COMMAND1 register */ + u32 command2; /* 004:QSPI_COMMAND2 register */ + u32 timing1; /* 008:QSPI_CS_TIM1 register */ + u32 timing2; /* 00c:QSPI_CS_TIM2 register */ + u32 xfer_status;/* 010:QSPI_TRANS_STATUS register */ + u32 fifo_status;/* 014:QSPI_FIFO_STATUS register */ + u32 tx_data; /* 018:QSPI_TX_DATA register */ + u32 rx_data; /* 01c:QSPI_RX_DATA register */ + u32 dma_ctl; /* 020:QSPI_DMA_CTL register */ + u32 dma_blk; /* 024:QSPI_DMA_BLK register */ + u32 rsvd[56]; /* 028-107 reserved */ + u32 tx_fifo; /* 108:QSPI_FIFO1 register */ + u32 rsvd2[31]; /* 10c-187 reserved */ + u32 rx_fifo; /* 188:QSPI_FIFO2 register */ + u32 spare_ctl; /* 18c:QSPI_SPARE_CTRL register */ +}; + +struct tegra210_qspi_priv { + struct qspi_regs *regs; + unsigned int freq; + unsigned int mode; + int periph_id; + int valid; + int last_transaction_us; +}; + +static int tegra210_qspi_ofdata_to_platdata(struct udevice *bus) +{ + struct tegra_spi_platdata *plat = bus->platdata; + const void *blob = gd->fdt_blob; + int node = bus->of_offset; + + plat->base = dev_get_addr(bus); + plat->periph_id = clock_decode_periph_id(blob, node); + + if (plat->periph_id == PERIPH_ID_NONE) { + debug("%s: could not decode periph id %d\n", __func__, + plat->periph_id); + return -FDT_ERR_NOTFOUND; + } + + /* Use 500KHz as a suitable default */ + plat->frequency = fdtdec_get_int(blob, node, "spi-max-frequency", + 500000); + plat->deactivate_delay_us = fdtdec_get_int(blob, node, + "spi-deactivate-delay", 0); + debug("%s: base=%#08lx, periph_id=%d, max-frequency=%d, deactivate_delay=%d\n", + __func__, plat->base, plat->periph_id, plat->frequency, + plat->deactivate_delay_us); + + return 0; +} + +static int tegra210_qspi_probe(struct udevice *bus) +{ + struct tegra_spi_platdata *plat = dev_get_platdata(bus); + struct tegra210_qspi_priv *priv = dev_get_priv(bus); + + priv->regs = (struct qspi_regs *)plat->base; + + priv->last_transaction_us = timer_get_us(); + priv->freq = plat->frequency; + priv->periph_id = plat->periph_id; + + return 0; +} + +static int tegra210_qspi_claim_bus(struct udevice *bus) +{ + struct tegra210_qspi_priv *priv = dev_get_priv(bus); + struct qspi_regs *regs = priv->regs; + + /* Change SPI clock to correct frequency, PLLP_OUT0 source */ + clock_start_periph_pll(priv->periph_id, CLOCK_ID_PERIPH, priv->freq); + + /* Clear stale status here */ + setbits_le32(®s->fifo_status, + QSPI_FIFO_STS_ERR | + QSPI_FIFO_STS_TX_FIFO_OVF | + QSPI_FIFO_STS_TX_FIFO_UNR | + QSPI_FIFO_STS_RX_FIFO_OVF | + QSPI_FIFO_STS_RX_FIFO_UNR | + QSPI_FIFO_STS_TX_FIFO_FULL | + QSPI_FIFO_STS_TX_FIFO_EMPTY | + QSPI_FIFO_STS_RX_FIFO_FULL | + QSPI_FIFO_STS_RX_FIFO_EMPTY); + debug("%s: FIFO STATUS = %08x\n", __func__, readl(®s->fifo_status)); + + /* Set master mode and sw controlled CS */ + setbits_le32(®s->command1, QSPI_CMD1_M_S | QSPI_CMD1_CS_SW_HW | + (priv->mode << QSPI_CMD1_MODE_SHIFT)); + debug("%s: COMMAND1 = %08x\n", __func__, readl(®s->command1)); + + return 0; +} + +/** + * Activate the CS by driving it LOW + * + * @param slave Pointer to spi_slave to which controller has to + * communicate with + */ +static void spi_cs_activate(struct udevice *dev) +{ + struct udevice *bus = dev->parent; + struct tegra_spi_platdata *pdata = dev_get_platdata(bus); + struct tegra210_qspi_priv *priv = dev_get_priv(bus); + + /* If it's too soon to do another transaction, wait */ + if (pdata->deactivate_delay_us && + priv->last_transaction_us) { + ulong delay_us; /* The delay completed so far */ + delay_us = timer_get_us() - priv->last_transaction_us; + if (delay_us < pdata->deactivate_delay_us) + udelay(pdata->deactivate_delay_us - delay_us); + } + + clrbits_le32(&priv->regs->command1, QSPI_CMD1_CS_SW_VAL); +} + +/** + * Deactivate the CS by driving it HIGH + * + * @param slave Pointer to spi_slave to which controller has to + * communicate with + */ +static void spi_cs_deactivate(struct udevice *dev) +{ + struct udevice *bus = dev->parent; + struct tegra_spi_platdata *pdata = dev_get_platdata(bus); + struct tegra210_qspi_priv *priv = dev_get_priv(bus); + + setbits_le32(&priv->regs->command1, QSPI_CMD1_CS_SW_VAL); + + /* Remember time of this transaction so we can honour the bus delay */ + if (pdata->deactivate_delay_us) + priv->last_transaction_us = timer_get_us(); + + debug("Deactivate CS, bus '%s'\n", bus->name); +} + +static int tegra210_qspi_xfer(struct udevice *dev, unsigned int bitlen, + const void *data_out, void *data_in, + unsigned long flags) +{ + struct udevice *bus = dev->parent; + struct tegra210_qspi_priv *priv = dev_get_priv(bus); + struct qspi_regs *regs = priv->regs; + u32 reg, tmpdout, tmpdin = 0; + const u8 *dout = data_out; + u8 *din = data_in; + int num_bytes; + int ret; + + debug("%s: slave %u:%u dout %p din %p bitlen %u\n", + __func__, bus->seq, spi_chip_select(dev), dout, din, bitlen); + if (bitlen % 8) + return -1; + num_bytes = bitlen / 8; + + ret = 0; + + /* clear all error status bits */ + reg = readl(®s->fifo_status); + writel(reg, ®s->fifo_status); + + /* flush RX/TX FIFOs */ + setbits_le32(®s->fifo_status, + (QSPI_FIFO_STS_RX_FIFO_FLUSH | + QSPI_FIFO_STS_TX_FIFO_FLUSH)); + while ((readl(®s->fifo_status) & + (QSPI_FIFO_STS_RX_FIFO_FLUSH | + QSPI_FIFO_STS_TX_FIFO_FLUSH))) + ; + + /* + * Notes: + * 1. don't set LSBY_FE, so no need to swap bytes from/to TX/RX FIFOs; + * 2. don't set RX_EN and TX_EN yet. + * (SW needs to make sure that while programming the blk_size, + * tx_en and rx_en bits must be zero) + * [TODO] I (Yen Lin) have problems when both RX/TX EN bits are set + * i.e., both dout and din are not NULL. + */ + clrsetbits_le32(®s->command1, + (QSPI_CMD1_LSBI_FE | QSPI_CMD1_LSBY_FE | + QSPI_CMD1_RX_EN | QSPI_CMD1_TX_EN), + (spi_chip_select(dev) << QSPI_CMD1_CS_SEL_SHIFT)); + + /* set xfer size to 1 block (32 bits) */ + writel(0, ®s->dma_blk); + + if (flags & SPI_XFER_BEGIN) + spi_cs_activate(dev); + + /* handle data in 32-bit chunks */ + while (num_bytes > 0) { + int bytes; + int tm; + + tmpdout = 0; + bytes = (num_bytes > 4) ? 4 : num_bytes; + + if (dout != NULL) { + memcpy((void *)&tmpdout, (void *)dout, bytes); + dout += bytes; + num_bytes -= bytes; + writel(tmpdout, ®s->tx_fifo); + setbits_le32(®s->command1, QSPI_CMD1_TX_EN); + } + + if (din != NULL) + setbits_le32(®s->command1, QSPI_CMD1_RX_EN); + + /* clear ready bit */ + setbits_le32(®s->xfer_status, QSPI_XFER_STS_RDY); + + clrsetbits_le32(®s->command1, + QSPI_CMD1_BITLEN_MASK << QSPI_CMD1_BITLEN_SHIFT, + (bytes * 8 - 1) << QSPI_CMD1_BITLEN_SHIFT); + /* Need to stabilize other reg bit before GO bit set */ + udelay(2); + setbits_le32(®s->command1, QSPI_CMD1_GO); + udelay(1); + + /* + * Wait for SPI transmit FIFO to empty, or to time out. + * The RX FIFO status will be read and cleared last + */ + for (tm = 0; tm < QSPI_TIMEOUT; ++tm) { + u32 fifo_status, xfer_status; + + xfer_status = readl(®s->xfer_status); + if (!(xfer_status & QSPI_XFER_STS_RDY)) + continue; + + fifo_status = readl(®s->fifo_status); + if (fifo_status & QSPI_FIFO_STS_ERR) { + debug("%s: got a fifo error: ", __func__); + if (fifo_status & QSPI_FIFO_STS_TX_FIFO_OVF) + debug("tx FIFO overflow "); + if (fifo_status & QSPI_FIFO_STS_TX_FIFO_UNR) + debug("tx FIFO underrun "); + if (fifo_status & QSPI_FIFO_STS_RX_FIFO_OVF) + debug("rx FIFO overflow "); + if (fifo_status & QSPI_FIFO_STS_RX_FIFO_UNR) + debug("rx FIFO underrun "); + if (fifo_status & QSPI_FIFO_STS_TX_FIFO_FULL) + debug("tx FIFO full "); + if (fifo_status & QSPI_FIFO_STS_TX_FIFO_EMPTY) + debug("tx FIFO empty "); + if (fifo_status & QSPI_FIFO_STS_RX_FIFO_FULL) + debug("rx FIFO full "); + if (fifo_status & QSPI_FIFO_STS_RX_FIFO_EMPTY) + debug("rx FIFO empty "); + debug("\n"); + break; + } + + if (!(fifo_status & QSPI_FIFO_STS_RX_FIFO_EMPTY)) { + tmpdin = readl(®s->rx_fifo); + if (din != NULL) { + memcpy(din, &tmpdin, bytes); + din += bytes; + num_bytes -= bytes; + } + } + break; + } + + if (tm >= QSPI_TIMEOUT) + ret = tm; + + /* clear ACK RDY, etc. bits */ + writel(readl(®s->fifo_status), ®s->fifo_status); + } + + if (flags & SPI_XFER_END) + spi_cs_deactivate(dev); + + debug("%s: transfer ended. Value=%08x, fifo_status = %08x\n", + __func__, tmpdin, readl(®s->fifo_status)); + + if (ret) { + printf("%s: timeout during SPI transfer, tm %d\n", + __func__, ret); + return -1; + } + + return ret; +} + +static int tegra210_qspi_set_speed(struct udevice *bus, uint speed) +{ + struct tegra_spi_platdata *plat = bus->platdata; + struct tegra210_qspi_priv *priv = dev_get_priv(bus); + + if (speed > plat->frequency) + speed = plat->frequency; + priv->freq = speed; + debug("%s: regs=%p, speed=%d\n", __func__, priv->regs, priv->freq); + + return 0; +} + +static int tegra210_qspi_set_mode(struct udevice *bus, uint mode) +{ + struct tegra210_qspi_priv *priv = dev_get_priv(bus); + + priv->mode = mode; + debug("%s: regs=%p, mode=%d\n", __func__, priv->regs, priv->mode); + + return 0; +} + +static const struct dm_spi_ops tegra210_qspi_ops = { + .claim_bus = tegra210_qspi_claim_bus, + .xfer = tegra210_qspi_xfer, + .set_speed = tegra210_qspi_set_speed, + .set_mode = tegra210_qspi_set_mode, + /* + * cs_info is not needed, since we require all chip selects to be + * in the device tree explicitly + */ +}; + +static const struct udevice_id tegra210_qspi_ids[] = { + { .compatible = "nvidia,tegra210-qspi" }, + { } +}; + +U_BOOT_DRIVER(tegra210_qspi) = { + .name = "tegra210-qspi", + .id = UCLASS_SPI, + .of_match = tegra210_qspi_ids, + .ops = &tegra210_qspi_ops, + .ofdata_to_platdata = tegra210_qspi_ofdata_to_platdata, + .platdata_auto_alloc_size = sizeof(struct tegra_spi_platdata), + .priv_auto_alloc_size = sizeof(struct tegra210_qspi_priv), + .per_child_auto_alloc_size = sizeof(struct spi_slave), + .probe = tegra210_qspi_probe, +};