u-boot/drivers/spi/mtk_spim.c
Tom Rini 15713fc855 mediatek: Include <linux/sizes.h> where needed
These files reference SZ_ macros without including <linux/sizes.h>,
correct this.

Signed-off-by: Tom Rini <trini@konsulko.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2022-11-10 10:08:55 -05:00

703 lines
18 KiB
C

// SPDX-License-Identifier: GPL-2.0
/*
* Copyright (C) 2022 MediaTek Inc. All Rights Reserved.
*
* Author: SkyLake.Huang <skylake.huang@mediatek.com>
*/
#include <clk.h>
#include <cpu_func.h>
#include <div64.h>
#include <dm.h>
#include <spi.h>
#include <spi-mem.h>
#include <stdbool.h>
#include <watchdog.h>
#include <dm/device.h>
#include <dm/device_compat.h>
#include <dm/devres.h>
#include <dm/pinctrl.h>
#include <linux/bitops.h>
#include <linux/completion.h>
#include <linux/dma-mapping.h>
#include <linux/io.h>
#include <linux/iopoll.h>
#include <linux/sizes.h>
#define SPI_CFG0_REG 0x0000
#define SPI_CFG1_REG 0x0004
#define SPI_TX_SRC_REG 0x0008
#define SPI_RX_DST_REG 0x000c
#define SPI_TX_DATA_REG 0x0010
#define SPI_RX_DATA_REG 0x0014
#define SPI_CMD_REG 0x0018
#define SPI_IRQ_REG 0x001c
#define SPI_STATUS_REG 0x0020
#define SPI_PAD_SEL_REG 0x0024
#define SPI_CFG2_REG 0x0028
#define SPI_TX_SRC_REG_64 0x002c
#define SPI_RX_DST_REG_64 0x0030
#define SPI_CFG3_IPM_REG 0x0040
#define SPI_CFG0_SCK_HIGH_OFFSET 0
#define SPI_CFG0_SCK_LOW_OFFSET 8
#define SPI_CFG0_CS_HOLD_OFFSET 16
#define SPI_CFG0_CS_SETUP_OFFSET 24
#define SPI_ADJUST_CFG0_CS_HOLD_OFFSET 0
#define SPI_ADJUST_CFG0_CS_SETUP_OFFSET 16
#define SPI_CFG1_CS_IDLE_OFFSET 0
#define SPI_CFG1_PACKET_LOOP_OFFSET 8
#define SPI_CFG1_PACKET_LENGTH_OFFSET 16
#define SPI_CFG1_GET_TICKDLY_OFFSET 29
#define SPI_CFG1_GET_TICKDLY_MASK GENMASK(31, 29)
#define SPI_CFG1_CS_IDLE_MASK 0xff
#define SPI_CFG1_PACKET_LOOP_MASK 0xff00
#define SPI_CFG1_PACKET_LENGTH_MASK 0x3ff0000
#define SPI_CFG1_IPM_PACKET_LENGTH_MASK GENMASK(31, 16)
#define SPI_CFG2_SCK_HIGH_OFFSET 0
#define SPI_CFG2_SCK_LOW_OFFSET 16
#define SPI_CFG2_SCK_HIGH_MASK GENMASK(15, 0)
#define SPI_CFG2_SCK_LOW_MASK GENMASK(31, 16)
#define SPI_CMD_ACT BIT(0)
#define SPI_CMD_RESUME BIT(1)
#define SPI_CMD_RST BIT(2)
#define SPI_CMD_PAUSE_EN BIT(4)
#define SPI_CMD_DEASSERT BIT(5)
#define SPI_CMD_SAMPLE_SEL BIT(6)
#define SPI_CMD_CS_POL BIT(7)
#define SPI_CMD_CPHA BIT(8)
#define SPI_CMD_CPOL BIT(9)
#define SPI_CMD_RX_DMA BIT(10)
#define SPI_CMD_TX_DMA BIT(11)
#define SPI_CMD_TXMSBF BIT(12)
#define SPI_CMD_RXMSBF BIT(13)
#define SPI_CMD_RX_ENDIAN BIT(14)
#define SPI_CMD_TX_ENDIAN BIT(15)
#define SPI_CMD_FINISH_IE BIT(16)
#define SPI_CMD_PAUSE_IE BIT(17)
#define SPI_CMD_IPM_NONIDLE_MODE BIT(19)
#define SPI_CMD_IPM_SPIM_LOOP BIT(21)
#define SPI_CMD_IPM_GET_TICKDLY_OFFSET 22
#define SPI_CMD_IPM_GET_TICKDLY_MASK GENMASK(24, 22)
#define PIN_MODE_CFG(x) ((x) / 2)
#define SPI_CFG3_IPM_PIN_MODE_OFFSET 0
#define SPI_CFG3_IPM_HALF_DUPLEX_DIR BIT(2)
#define SPI_CFG3_IPM_HALF_DUPLEX_EN BIT(3)
#define SPI_CFG3_IPM_XMODE_EN BIT(4)
#define SPI_CFG3_IPM_NODATA_FLAG BIT(5)
#define SPI_CFG3_IPM_CMD_BYTELEN_OFFSET 8
#define SPI_CFG3_IPM_ADDR_BYTELEN_OFFSET 12
#define SPI_CFG3_IPM_DUMMY_BYTELEN_OFFSET 16
#define SPI_CFG3_IPM_CMD_PIN_MODE_MASK GENMASK(1, 0)
#define SPI_CFG3_IPM_CMD_BYTELEN_MASK GENMASK(11, 8)
#define SPI_CFG3_IPM_ADDR_BYTELEN_MASK GENMASK(15, 12)
#define SPI_CFG3_IPM_DUMMY_BYTELEN_MASK GENMASK(19, 16)
#define MT8173_SPI_MAX_PAD_SEL 3
#define MTK_SPI_PAUSE_INT_STATUS 0x2
#define MTK_SPI_IDLE 0
#define MTK_SPI_PAUSED 1
#define MTK_SPI_MAX_FIFO_SIZE 32U
#define MTK_SPI_PACKET_SIZE 1024
#define MTK_SPI_IPM_PACKET_SIZE SZ_64K
#define MTK_SPI_IPM_PACKET_LOOP SZ_256
#define MTK_SPI_32BITS_MASK 0xffffffff
#define DMA_ADDR_EXT_BITS 36
#define DMA_ADDR_DEF_BITS 32
#define CLK_TO_US(freq, clkcnt) DIV_ROUND_UP((clkcnt), (freq) / 1000000)
/* struct mtk_spim_capability
* @enhance_timing: Some IC design adjust cfg register to enhance time accuracy
* @dma_ext: Some IC support DMA addr extension
* @ipm_design: The IPM IP design improves some features, and supports dual/quad mode
* @support_quad: Whether quad mode is supported
*/
struct mtk_spim_capability {
bool enhance_timing;
bool dma_ext;
bool ipm_design;
bool support_quad;
};
/* struct mtk_spim_priv
* @base: Base address of the spi controller
* @state: Controller state
* @sel_clk: Pad clock
* @spi_clk: Core clock
* @xfer_len: Current length of data for transfer
* @hw_cap: Controller capabilities
* @tick_dly: Used to postpone SPI sampling time
* @sample_sel: Sample edge of MISO
* @dev: udevice of this spi controller
* @tx_dma: Tx DMA address
* @rx_dma: Rx DMA address
*/
struct mtk_spim_priv {
void __iomem *base;
u32 state;
struct clk sel_clk, spi_clk;
u32 xfer_len;
struct mtk_spim_capability hw_cap;
u32 tick_dly;
u32 sample_sel;
struct device *dev;
dma_addr_t tx_dma;
dma_addr_t rx_dma;
};
static void mtk_spim_reset(struct mtk_spim_priv *priv)
{
/* set the software reset bit in SPI_CMD_REG. */
setbits_le32(priv->base + SPI_CMD_REG, SPI_CMD_RST);
clrbits_le32(priv->base + SPI_CMD_REG, SPI_CMD_RST);
}
static int mtk_spim_hw_init(struct spi_slave *slave)
{
struct udevice *bus = dev_get_parent(slave->dev);
struct mtk_spim_priv *priv = dev_get_priv(bus);
u16 cpha, cpol;
u32 reg_val;
cpha = slave->mode & SPI_CPHA ? 1 : 0;
cpol = slave->mode & SPI_CPOL ? 1 : 0;
if (priv->hw_cap.enhance_timing) {
if (priv->hw_cap.ipm_design) {
/* CFG3 reg only used for spi-mem,
* here write to default value
*/
writel(0x0, priv->base + SPI_CFG3_IPM_REG);
clrsetbits_le32(priv->base + SPI_CMD_REG,
SPI_CMD_IPM_GET_TICKDLY_MASK,
priv->tick_dly <<
SPI_CMD_IPM_GET_TICKDLY_OFFSET);
} else {
clrsetbits_le32(priv->base + SPI_CFG1_REG,
SPI_CFG1_GET_TICKDLY_MASK,
priv->tick_dly <<
SPI_CFG1_GET_TICKDLY_OFFSET);
}
}
reg_val = readl(priv->base + SPI_CMD_REG);
if (priv->hw_cap.ipm_design) {
/* SPI transfer without idle time until packet length done */
reg_val |= SPI_CMD_IPM_NONIDLE_MODE;
if (slave->mode & SPI_LOOP)
reg_val |= SPI_CMD_IPM_SPIM_LOOP;
else
reg_val &= ~SPI_CMD_IPM_SPIM_LOOP;
}
if (cpha)
reg_val |= SPI_CMD_CPHA;
else
reg_val &= ~SPI_CMD_CPHA;
if (cpol)
reg_val |= SPI_CMD_CPOL;
else
reg_val &= ~SPI_CMD_CPOL;
/* set the mlsbx and mlsbtx */
if (slave->mode & SPI_LSB_FIRST) {
reg_val &= ~SPI_CMD_TXMSBF;
reg_val &= ~SPI_CMD_RXMSBF;
} else {
reg_val |= SPI_CMD_TXMSBF;
reg_val |= SPI_CMD_RXMSBF;
}
/* do not reverse tx/rx endian */
reg_val &= ~SPI_CMD_TX_ENDIAN;
reg_val &= ~SPI_CMD_RX_ENDIAN;
if (priv->hw_cap.enhance_timing) {
/* set CS polarity */
if (slave->mode & SPI_CS_HIGH)
reg_val |= SPI_CMD_CS_POL;
else
reg_val &= ~SPI_CMD_CS_POL;
if (priv->sample_sel)
reg_val |= SPI_CMD_SAMPLE_SEL;
else
reg_val &= ~SPI_CMD_SAMPLE_SEL;
}
/* disable dma mode */
reg_val &= ~(SPI_CMD_TX_DMA | SPI_CMD_RX_DMA);
/* disable deassert mode */
reg_val &= ~SPI_CMD_DEASSERT;
writel(reg_val, priv->base + SPI_CMD_REG);
return 0;
}
static void mtk_spim_prepare_transfer(struct mtk_spim_priv *priv,
u32 speed_hz)
{
u32 spi_clk_hz, div, sck_time, cs_time, reg_val;
spi_clk_hz = clk_get_rate(&priv->spi_clk);
if (speed_hz <= spi_clk_hz / 4)
div = DIV_ROUND_UP(spi_clk_hz, speed_hz);
else
div = 4;
sck_time = (div + 1) / 2;
cs_time = sck_time * 2;
if (priv->hw_cap.enhance_timing) {
reg_val = ((sck_time - 1) & 0xffff)
<< SPI_CFG2_SCK_HIGH_OFFSET;
reg_val |= ((sck_time - 1) & 0xffff)
<< SPI_CFG2_SCK_LOW_OFFSET;
writel(reg_val, priv->base + SPI_CFG2_REG);
reg_val = ((cs_time - 1) & 0xffff)
<< SPI_ADJUST_CFG0_CS_HOLD_OFFSET;
reg_val |= ((cs_time - 1) & 0xffff)
<< SPI_ADJUST_CFG0_CS_SETUP_OFFSET;
writel(reg_val, priv->base + SPI_CFG0_REG);
} else {
reg_val = ((sck_time - 1) & 0xff)
<< SPI_CFG0_SCK_HIGH_OFFSET;
reg_val |= ((sck_time - 1) & 0xff) << SPI_CFG0_SCK_LOW_OFFSET;
reg_val |= ((cs_time - 1) & 0xff) << SPI_CFG0_CS_HOLD_OFFSET;
reg_val |= ((cs_time - 1) & 0xff) << SPI_CFG0_CS_SETUP_OFFSET;
writel(reg_val, priv->base + SPI_CFG0_REG);
}
reg_val = readl(priv->base + SPI_CFG1_REG);
reg_val &= ~SPI_CFG1_CS_IDLE_MASK;
reg_val |= ((cs_time - 1) & 0xff) << SPI_CFG1_CS_IDLE_OFFSET;
writel(reg_val, priv->base + SPI_CFG1_REG);
}
/**
* mtk_spim_setup_packet() - setup packet format.
* @priv: controller priv
*
* This controller sents/receives data in packets. The packet size is
* configurable.
*
* This function calculates the maximum packet size available for current
* data, and calculates the number of packets required to sent/receive data
* as much as possible.
*/
static void mtk_spim_setup_packet(struct mtk_spim_priv *priv)
{
u32 packet_size, packet_loop, reg_val;
/* Calculate maximum packet size */
if (priv->hw_cap.ipm_design)
packet_size = min_t(u32,
priv->xfer_len,
MTK_SPI_IPM_PACKET_SIZE);
else
packet_size = min_t(u32,
priv->xfer_len,
MTK_SPI_PACKET_SIZE);
/* Calculates number of packets to sent/receive */
packet_loop = priv->xfer_len / packet_size;
reg_val = readl(priv->base + SPI_CFG1_REG);
if (priv->hw_cap.ipm_design)
reg_val &= ~SPI_CFG1_IPM_PACKET_LENGTH_MASK;
else
reg_val &= ~SPI_CFG1_PACKET_LENGTH_MASK;
reg_val |= (packet_size - 1) << SPI_CFG1_PACKET_LENGTH_OFFSET;
reg_val &= ~SPI_CFG1_PACKET_LOOP_MASK;
reg_val |= (packet_loop - 1) << SPI_CFG1_PACKET_LOOP_OFFSET;
writel(reg_val, priv->base + SPI_CFG1_REG);
}
static void mtk_spim_enable_transfer(struct mtk_spim_priv *priv)
{
u32 cmd;
cmd = readl(priv->base + SPI_CMD_REG);
if (priv->state == MTK_SPI_IDLE)
cmd |= SPI_CMD_ACT;
else
cmd |= SPI_CMD_RESUME;
writel(cmd, priv->base + SPI_CMD_REG);
}
static bool mtk_spim_supports_op(struct spi_slave *slave,
const struct spi_mem_op *op)
{
struct udevice *bus = dev_get_parent(slave->dev);
struct mtk_spim_priv *priv = dev_get_priv(bus);
if (op->cmd.buswidth == 0 || op->cmd.buswidth > 4 ||
op->addr.buswidth > 4 || op->dummy.buswidth > 4 ||
op->data.buswidth > 4)
return false;
if (!priv->hw_cap.support_quad && (op->cmd.buswidth > 2 ||
op->addr.buswidth > 2 || op->dummy.buswidth > 2 ||
op->data.buswidth > 2))
return false;
if (op->addr.nbytes && op->dummy.nbytes &&
op->addr.buswidth != op->dummy.buswidth)
return false;
if (op->addr.nbytes + op->dummy.nbytes > 16)
return false;
if (op->data.nbytes > MTK_SPI_IPM_PACKET_SIZE) {
if (op->data.nbytes / MTK_SPI_IPM_PACKET_SIZE >
MTK_SPI_IPM_PACKET_LOOP ||
op->data.nbytes % MTK_SPI_IPM_PACKET_SIZE != 0)
return false;
}
return true;
}
static void mtk_spim_setup_dma_xfer(struct mtk_spim_priv *priv,
const struct spi_mem_op *op)
{
writel((u32)(priv->tx_dma & MTK_SPI_32BITS_MASK),
priv->base + SPI_TX_SRC_REG);
if (priv->hw_cap.dma_ext)
writel((u32)(priv->tx_dma >> 32),
priv->base + SPI_TX_SRC_REG_64);
if (op->data.dir == SPI_MEM_DATA_IN) {
writel((u32)(priv->rx_dma & MTK_SPI_32BITS_MASK),
priv->base + SPI_RX_DST_REG);
if (priv->hw_cap.dma_ext)
writel((u32)(priv->rx_dma >> 32),
priv->base + SPI_RX_DST_REG_64);
}
}
static int mtk_spim_transfer_wait(struct spi_slave *slave,
const struct spi_mem_op *op)
{
struct udevice *bus = dev_get_parent(slave->dev);
struct mtk_spim_priv *priv = dev_get_priv(bus);
u32 sck_l, sck_h, spi_bus_clk, clk_count, reg;
ulong us = 1;
int ret = 0;
if (op->data.dir == SPI_MEM_NO_DATA)
clk_count = 32;
else
clk_count = op->data.nbytes;
spi_bus_clk = clk_get_rate(&priv->spi_clk);
sck_l = readl(priv->base + SPI_CFG2_REG) >> SPI_CFG2_SCK_LOW_OFFSET;
sck_h = readl(priv->base + SPI_CFG2_REG) & SPI_CFG2_SCK_HIGH_MASK;
do_div(spi_bus_clk, sck_l + sck_h + 2);
us = CLK_TO_US(spi_bus_clk, clk_count * 8);
us += 1000 * 1000; /* 1s tolerance */
if (us > UINT_MAX)
us = UINT_MAX;
ret = readl_poll_timeout(priv->base + SPI_STATUS_REG, reg,
reg & 0x1, us);
if (ret < 0) {
dev_err(priv->dev, "transfer timeout, val: 0x%lx\n", us);
return -ETIMEDOUT;
}
return 0;
}
static int mtk_spim_exec_op(struct spi_slave *slave,
const struct spi_mem_op *op)
{
struct udevice *bus = dev_get_parent(slave->dev);
struct mtk_spim_priv *priv = dev_get_priv(bus);
u32 reg_val, nio = 1, tx_size;
char *tx_tmp_buf;
char *rx_tmp_buf;
int i, ret = 0;
mtk_spim_reset(priv);
mtk_spim_hw_init(slave);
mtk_spim_prepare_transfer(priv, slave->max_hz);
reg_val = readl(priv->base + SPI_CFG3_IPM_REG);
/* opcode byte len */
reg_val &= ~SPI_CFG3_IPM_CMD_BYTELEN_MASK;
reg_val |= 1 << SPI_CFG3_IPM_CMD_BYTELEN_OFFSET;
/* addr & dummy byte len */
if (op->addr.nbytes || op->dummy.nbytes)
reg_val |= (op->addr.nbytes + op->dummy.nbytes) <<
SPI_CFG3_IPM_ADDR_BYTELEN_OFFSET;
/* data byte len */
if (!op->data.nbytes) {
reg_val |= SPI_CFG3_IPM_NODATA_FLAG;
writel(0, priv->base + SPI_CFG1_REG);
} else {
reg_val &= ~SPI_CFG3_IPM_NODATA_FLAG;
priv->xfer_len = op->data.nbytes;
mtk_spim_setup_packet(priv);
}
if (op->addr.nbytes || op->dummy.nbytes) {
if (op->addr.buswidth == 1 || op->dummy.buswidth == 1)
reg_val |= SPI_CFG3_IPM_XMODE_EN;
else
reg_val &= ~SPI_CFG3_IPM_XMODE_EN;
}
if (op->addr.buswidth == 2 ||
op->dummy.buswidth == 2 ||
op->data.buswidth == 2)
nio = 2;
else if (op->addr.buswidth == 4 ||
op->dummy.buswidth == 4 ||
op->data.buswidth == 4)
nio = 4;
reg_val &= ~SPI_CFG3_IPM_CMD_PIN_MODE_MASK;
reg_val |= PIN_MODE_CFG(nio) << SPI_CFG3_IPM_PIN_MODE_OFFSET;
reg_val |= SPI_CFG3_IPM_HALF_DUPLEX_EN;
if (op->data.dir == SPI_MEM_DATA_IN)
reg_val |= SPI_CFG3_IPM_HALF_DUPLEX_DIR;
else
reg_val &= ~SPI_CFG3_IPM_HALF_DUPLEX_DIR;
writel(reg_val, priv->base + SPI_CFG3_IPM_REG);
tx_size = 1 + op->addr.nbytes + op->dummy.nbytes;
if (op->data.dir == SPI_MEM_DATA_OUT)
tx_size += op->data.nbytes;
tx_size = max(tx_size, (u32)32);
/* Fill up tx data */
tx_tmp_buf = kzalloc(tx_size, GFP_KERNEL);
if (!tx_tmp_buf) {
ret = -ENOMEM;
goto exit;
}
tx_tmp_buf[0] = op->cmd.opcode;
if (op->addr.nbytes) {
for (i = 0; i < op->addr.nbytes; i++)
tx_tmp_buf[i + 1] = op->addr.val >>
(8 * (op->addr.nbytes - i - 1));
}
if (op->dummy.nbytes)
memset(tx_tmp_buf + op->addr.nbytes + 1, 0xff,
op->dummy.nbytes);
if (op->data.nbytes && op->data.dir == SPI_MEM_DATA_OUT)
memcpy(tx_tmp_buf + op->dummy.nbytes + op->addr.nbytes + 1,
op->data.buf.out, op->data.nbytes);
/* Finish filling up tx data */
priv->tx_dma = dma_map_single(tx_tmp_buf, tx_size, DMA_TO_DEVICE);
if (dma_mapping_error(priv->dev, priv->tx_dma)) {
ret = -ENOMEM;
goto tx_free;
}
if (op->data.dir == SPI_MEM_DATA_IN) {
if (!IS_ALIGNED((size_t)op->data.buf.in, 4)) {
rx_tmp_buf = kzalloc(op->data.nbytes, GFP_KERNEL);
if (!rx_tmp_buf) {
ret = -ENOMEM;
goto tx_unmap;
}
} else {
rx_tmp_buf = op->data.buf.in;
}
priv->rx_dma = dma_map_single(rx_tmp_buf, op->data.nbytes,
DMA_FROM_DEVICE);
if (dma_mapping_error(priv->dev, priv->rx_dma)) {
ret = -ENOMEM;
goto rx_free;
}
}
reg_val = readl(priv->base + SPI_CMD_REG);
reg_val |= SPI_CMD_TX_DMA;
if (op->data.dir == SPI_MEM_DATA_IN)
reg_val |= SPI_CMD_RX_DMA;
writel(reg_val, priv->base + SPI_CMD_REG);
mtk_spim_setup_dma_xfer(priv, op);
mtk_spim_enable_transfer(priv);
/* Wait for the interrupt. */
ret = mtk_spim_transfer_wait(slave, op);
if (ret)
goto rx_unmap;
if (op->data.dir == SPI_MEM_DATA_IN &&
!IS_ALIGNED((size_t)op->data.buf.in, 4))
memcpy(op->data.buf.in, rx_tmp_buf, op->data.nbytes);
rx_unmap:
/* spi disable dma */
reg_val = readl(priv->base + SPI_CMD_REG);
reg_val &= ~SPI_CMD_TX_DMA;
if (op->data.dir == SPI_MEM_DATA_IN)
reg_val &= ~SPI_CMD_RX_DMA;
writel(reg_val, priv->base + SPI_CMD_REG);
writel(0, priv->base + SPI_TX_SRC_REG);
writel(0, priv->base + SPI_RX_DST_REG);
if (op->data.dir == SPI_MEM_DATA_IN)
dma_unmap_single(priv->rx_dma,
op->data.nbytes, DMA_FROM_DEVICE);
rx_free:
if (op->data.dir == SPI_MEM_DATA_IN &&
!IS_ALIGNED((size_t)op->data.buf.in, 4))
kfree(rx_tmp_buf);
tx_unmap:
dma_unmap_single(priv->tx_dma,
tx_size, DMA_TO_DEVICE);
tx_free:
kfree(tx_tmp_buf);
exit:
return ret;
}
static int mtk_spim_adjust_op_size(struct spi_slave *slave,
struct spi_mem_op *op)
{
int opcode_len;
if (!op->data.nbytes)
return 0;
if (op->data.dir != SPI_MEM_NO_DATA) {
opcode_len = 1 + op->addr.nbytes + op->dummy.nbytes;
if (opcode_len + op->data.nbytes > MTK_SPI_IPM_PACKET_SIZE) {
op->data.nbytes = MTK_SPI_IPM_PACKET_SIZE - opcode_len;
/* force data buffer dma-aligned. */
op->data.nbytes -= op->data.nbytes % 4;
}
}
return 0;
}
static int mtk_spim_get_attr(struct mtk_spim_priv *priv, struct udevice *dev)
{
int ret;
priv->hw_cap.enhance_timing = dev_read_bool(dev, "enhance_timing");
priv->hw_cap.dma_ext = dev_read_bool(dev, "dma_ext");
priv->hw_cap.ipm_design = dev_read_bool(dev, "ipm_design");
priv->hw_cap.support_quad = dev_read_bool(dev, "support_quad");
ret = dev_read_u32(dev, "tick_dly", &priv->tick_dly);
if (ret < 0)
dev_err(priv->dev, "tick dly not set.\n");
ret = dev_read_u32(dev, "sample_sel", &priv->sample_sel);
if (ret < 0)
dev_err(priv->dev, "sample sel not set.\n");
return ret;
}
static int mtk_spim_probe(struct udevice *dev)
{
struct mtk_spim_priv *priv = dev_get_priv(dev);
int ret;
priv->base = (void __iomem *)devfdt_get_addr(dev);
if (!priv->base)
return -EINVAL;
mtk_spim_get_attr(priv, dev);
ret = clk_get_by_name(dev, "sel-clk", &priv->sel_clk);
if (ret < 0) {
dev_err(dev, "failed to get sel-clk\n");
return ret;
}
ret = clk_get_by_name(dev, "spi-clk", &priv->spi_clk);
if (ret < 0) {
dev_err(dev, "failed to get spi-clk\n");
return ret;
}
clk_enable(&priv->sel_clk);
clk_enable(&priv->spi_clk);
return 0;
}
static int mtk_spim_set_speed(struct udevice *dev, uint speed)
{
return 0;
}
static int mtk_spim_set_mode(struct udevice *dev, uint mode)
{
return 0;
}
static const struct spi_controller_mem_ops mtk_spim_mem_ops = {
.adjust_op_size = mtk_spim_adjust_op_size,
.supports_op = mtk_spim_supports_op,
.exec_op = mtk_spim_exec_op
};
static const struct dm_spi_ops mtk_spim_ops = {
.mem_ops = &mtk_spim_mem_ops,
.set_speed = mtk_spim_set_speed,
.set_mode = mtk_spim_set_mode,
};
static const struct udevice_id mtk_spim_ids[] = {
{ .compatible = "mediatek,ipm-spi" },
{}
};
U_BOOT_DRIVER(mtk_spim) = {
.name = "mtk_spim",
.id = UCLASS_SPI,
.of_match = mtk_spim_ids,
.ops = &mtk_spim_ops,
.priv_auto = sizeof(struct mtk_spim_priv),
.probe = mtk_spim_probe,
};