zephyr/drivers/spi/spi_ll_stm32.c

951 lines
25 KiB
C

/*
* Copyright (c) 2016 BayLibre, SAS
*
* SPDX-License-Identifier: Apache-2.0
*/
#define DT_DRV_COMPAT st_stm32_spi
#define LOG_LEVEL CONFIG_SPI_LOG_LEVEL
#include <logging/log.h>
LOG_MODULE_REGISTER(spi_ll_stm32);
#include <sys/util.h>
#include <kernel.h>
#include <soc.h>
#include <stm32_ll_spi.h>
#include <errno.h>
#include <drivers/spi.h>
#include <toolchain.h>
#ifdef CONFIG_SPI_STM32_DMA
#include <dt-bindings/dma/stm32_dma.h>
#include <drivers/dma.h>
#endif
#include <pinmux/stm32/pinmux_stm32.h>
#include <drivers/clock_control/stm32_clock_control.h>
#include <drivers/clock_control.h>
#include "spi_ll_stm32.h"
#define DEV_CFG(dev) \
(const struct spi_stm32_config * const)(dev->config)
#define DEV_DATA(dev) \
(struct spi_stm32_data * const)(dev->data)
/*
* Check for SPI_SR_FRE to determine support for TI mode frame format
* error flag, because STM32F1 SoCs do not support it and STM32CUBE
* for F1 family defines an unused LL_SPI_SR_FRE.
*/
#if defined(CONFIG_SOC_SERIES_STM32MP1X) || \
defined(CONFIG_SOC_SERIES_STM32H7X)
#define SPI_STM32_ERR_MSK (LL_SPI_SR_UDR | LL_SPI_SR_CRCE | LL_SPI_SR_MODF | \
LL_SPI_SR_OVR | LL_SPI_SR_TIFRE)
#else
#if defined(LL_SPI_SR_UDR)
#define SPI_STM32_ERR_MSK (LL_SPI_SR_UDR | LL_SPI_SR_CRCERR | LL_SPI_SR_MODF | \
LL_SPI_SR_OVR | LL_SPI_SR_FRE)
#elif defined(SPI_SR_FRE)
#define SPI_STM32_ERR_MSK (LL_SPI_SR_CRCERR | LL_SPI_SR_MODF | \
LL_SPI_SR_OVR | LL_SPI_SR_FRE)
#else
#define SPI_STM32_ERR_MSK (LL_SPI_SR_CRCERR | LL_SPI_SR_MODF | LL_SPI_SR_OVR)
#endif
#endif /* CONFIG_SOC_SERIES_STM32MP1X */
#ifdef CONFIG_SPI_STM32_DMA
/* dummy value used for transferring NOP when tx buf is null
* and use as dummy sink for when rx buf is null
*/
uint32_t dummy_rx_tx_buffer;
/* This function is executed in the interrupt context */
static void dma_callback(const struct device *dev, void *arg,
uint32_t channel, int status)
{
/* arg directly holds the spi device */
struct spi_stm32_data *data = arg;
if (status != 0) {
LOG_ERR("DMA callback error with channel %d.", channel);
data->status_flags |= SPI_STM32_DMA_ERROR_FLAG;
} else {
/* identify the origin of this callback */
if (channel == data->dma_tx.channel) {
/* this part of the transfer ends */
data->status_flags |= SPI_STM32_DMA_TX_DONE_FLAG;
} else if (channel == data->dma_rx.channel) {
/* this part of the transfer ends */
data->status_flags |= SPI_STM32_DMA_RX_DONE_FLAG;
} else {
LOG_ERR("DMA callback channel %d is not valid.",
channel);
data->status_flags |= SPI_STM32_DMA_ERROR_FLAG;
}
}
k_sem_give(&data->status_sem);
}
static int spi_stm32_dma_tx_load(const struct device *dev, const uint8_t *buf,
size_t len)
{
const struct spi_stm32_config *cfg = DEV_CFG(dev);
struct spi_stm32_data *data = DEV_DATA(dev);
struct dma_block_config *blk_cfg;
int ret;
/* remember active TX DMA channel (used in callback) */
struct stream *stream = &data->dma_tx;
blk_cfg = &stream->dma_blk_cfg;
/* prepare the block for this TX DMA channel */
memset(blk_cfg, 0, sizeof(struct dma_block_config));
blk_cfg->block_size = len;
/* tx direction has memory as source and periph as dest. */
if (buf == NULL) {
dummy_rx_tx_buffer = 0;
/* if tx buff is null, then sends NOP on the line. */
blk_cfg->source_address = (uint32_t)&dummy_rx_tx_buffer;
blk_cfg->source_addr_adj = DMA_ADDR_ADJ_NO_CHANGE;
} else {
blk_cfg->source_address = (uint32_t)buf;
if (data->dma_tx.src_addr_increment) {
blk_cfg->source_addr_adj = DMA_ADDR_ADJ_INCREMENT;
} else {
blk_cfg->source_addr_adj = DMA_ADDR_ADJ_NO_CHANGE;
}
}
blk_cfg->dest_address = (uint32_t)LL_SPI_DMA_GetRegAddr(cfg->spi);
/* fifo mode NOT USED there */
if (data->dma_tx.dst_addr_increment) {
blk_cfg->dest_addr_adj = DMA_ADDR_ADJ_INCREMENT;
} else {
blk_cfg->dest_addr_adj = DMA_ADDR_ADJ_NO_CHANGE;
}
/* give the fifo mode from the DT */
blk_cfg->fifo_mode_control = data->dma_tx.fifo_threshold;
/* direction is given by the DT */
stream->dma_cfg.head_block = blk_cfg;
/* give the client dev as arg, as the callback comes from the dma */
stream->dma_cfg.user_data = data;
/* pass our client origin to the dma: data->dma_tx.dma_channel */
ret = dma_config(data->dma_tx.dma_dev, data->dma_tx.channel,
&stream->dma_cfg);
/* the channel is the actual stream from 0 */
if (ret != 0) {
return ret;
}
/* gives the request ID to the dma mux */
return dma_start(data->dma_tx.dma_dev, data->dma_tx.channel);
}
static int spi_stm32_dma_rx_load(const struct device *dev, uint8_t *buf,
size_t len)
{
const struct spi_stm32_config *cfg = DEV_CFG(dev);
struct spi_stm32_data *data = DEV_DATA(dev);
struct dma_block_config *blk_cfg;
int ret;
/* retrieve active RX DMA channel (used in callback) */
struct stream *stream = &data->dma_rx;
blk_cfg = &stream->dma_blk_cfg;
/* prepare the block for this RX DMA channel */
memset(blk_cfg, 0, sizeof(struct dma_block_config));
blk_cfg->block_size = len;
/* rx direction has periph as source and mem as dest. */
if (buf == NULL) {
/* if rx buff is null, then write data to dummy address. */
blk_cfg->dest_address = (uint32_t)&dummy_rx_tx_buffer;
blk_cfg->dest_addr_adj = DMA_ADDR_ADJ_NO_CHANGE;
} else {
blk_cfg->dest_address = (uint32_t)buf;
if (data->dma_rx.dst_addr_increment) {
blk_cfg->dest_addr_adj = DMA_ADDR_ADJ_INCREMENT;
} else {
blk_cfg->dest_addr_adj = DMA_ADDR_ADJ_NO_CHANGE;
}
}
blk_cfg->source_address = (uint32_t)LL_SPI_DMA_GetRegAddr(cfg->spi);
if (data->dma_rx.src_addr_increment) {
blk_cfg->source_addr_adj = DMA_ADDR_ADJ_INCREMENT;
} else {
blk_cfg->source_addr_adj = DMA_ADDR_ADJ_NO_CHANGE;
}
/* give the fifo mode from the DT */
blk_cfg->fifo_mode_control = data->dma_rx.fifo_threshold;
/* direction is given by the DT */
stream->dma_cfg.head_block = blk_cfg;
stream->dma_cfg.user_data = data;
/* pass our client origin to the dma: data->dma_rx.channel */
ret = dma_config(data->dma_rx.dma_dev, data->dma_rx.channel,
&stream->dma_cfg);
/* the channel is the actual stream from 0 */
if (ret != 0) {
return ret;
}
/* gives the request ID to the dma mux */
return dma_start(data->dma_rx.dma_dev, data->dma_rx.channel);
}
static int spi_dma_move_buffers(const struct device *dev, size_t len)
{
struct spi_stm32_data *data = DEV_DATA(dev);
int ret;
size_t dma_segment_len;
dma_segment_len = len / data->dma_rx.dma_cfg.dest_data_size;
ret = spi_stm32_dma_rx_load(dev, data->ctx.rx_buf, dma_segment_len);
if (ret != 0) {
return ret;
}
dma_segment_len = len / data->dma_tx.dma_cfg.source_data_size;
ret = spi_stm32_dma_tx_load(dev, data->ctx.tx_buf, dma_segment_len);
return ret;
}
#endif /* CONFIG_SPI_STM32_DMA */
/* Value to shift out when no application data needs transmitting. */
#define SPI_STM32_TX_NOP 0x00
static bool spi_stm32_transfer_ongoing(struct spi_stm32_data *data)
{
return spi_context_tx_on(&data->ctx) || spi_context_rx_on(&data->ctx);
}
static int spi_stm32_get_err(SPI_TypeDef *spi)
{
uint32_t sr = LL_SPI_ReadReg(spi, SR);
if (sr & SPI_STM32_ERR_MSK) {
LOG_ERR("%s: err=%d", __func__,
sr & (uint32_t)SPI_STM32_ERR_MSK);
/* OVR error must be explicitly cleared */
if (LL_SPI_IsActiveFlag_OVR(spi)) {
LL_SPI_ClearFlag_OVR(spi);
}
return -EIO;
}
return 0;
}
/* Shift a SPI frame as master. */
static void spi_stm32_shift_m(SPI_TypeDef *spi, struct spi_stm32_data *data)
{
uint16_t tx_frame = SPI_STM32_TX_NOP;
uint16_t rx_frame;
while (!ll_func_tx_is_empty(spi)) {
/* NOP */
}
#if defined(CONFIG_SOC_SERIES_STM32MP1X) || \
defined(CONFIG_SOC_SERIES_STM32H7X)
/* With the STM32MP1 and the STM32H7, if the device is the SPI master,
* we need to enable the start of the transfer with
* LL_SPI_StartMasterTransfer(spi)
*/
if (LL_SPI_GetMode(spi) == LL_SPI_MODE_MASTER) {
LL_SPI_StartMasterTransfer(spi);
while (!LL_SPI_IsActiveMasterTransfer(spi)) {
/* NOP */
}
}
#endif
if (SPI_WORD_SIZE_GET(data->ctx.config->operation) == 8) {
if (spi_context_tx_buf_on(&data->ctx)) {
tx_frame = UNALIGNED_GET((uint8_t *)(data->ctx.tx_buf));
}
LL_SPI_TransmitData8(spi, tx_frame);
/* The update is ignored if TX is off. */
spi_context_update_tx(&data->ctx, 1, 1);
} else {
if (spi_context_tx_buf_on(&data->ctx)) {
tx_frame = UNALIGNED_GET((uint16_t *)(data->ctx.tx_buf));
}
LL_SPI_TransmitData16(spi, tx_frame);
/* The update is ignored if TX is off. */
spi_context_update_tx(&data->ctx, 2, 1);
}
while (!ll_func_rx_is_not_empty(spi)) {
/* NOP */
}
if (SPI_WORD_SIZE_GET(data->ctx.config->operation) == 8) {
rx_frame = LL_SPI_ReceiveData8(spi);
if (spi_context_rx_buf_on(&data->ctx)) {
UNALIGNED_PUT(rx_frame, (uint8_t *)data->ctx.rx_buf);
}
spi_context_update_rx(&data->ctx, 1, 1);
} else {
rx_frame = LL_SPI_ReceiveData16(spi);
if (spi_context_rx_buf_on(&data->ctx)) {
UNALIGNED_PUT(rx_frame, (uint16_t *)data->ctx.rx_buf);
}
spi_context_update_rx(&data->ctx, 2, 1);
}
}
/* Shift a SPI frame as slave. */
static void spi_stm32_shift_s(SPI_TypeDef *spi, struct spi_stm32_data *data)
{
if (ll_func_tx_is_empty(spi) && spi_context_tx_on(&data->ctx)) {
uint16_t tx_frame;
if (SPI_WORD_SIZE_GET(data->ctx.config->operation) == 8) {
tx_frame = UNALIGNED_GET((uint8_t *)(data->ctx.tx_buf));
LL_SPI_TransmitData8(spi, tx_frame);
spi_context_update_tx(&data->ctx, 1, 1);
} else {
tx_frame = UNALIGNED_GET((uint16_t *)(data->ctx.tx_buf));
LL_SPI_TransmitData16(spi, tx_frame);
spi_context_update_tx(&data->ctx, 2, 1);
}
} else {
ll_func_disable_int_tx_empty(spi);
}
if (ll_func_rx_is_not_empty(spi) &&
spi_context_rx_buf_on(&data->ctx)) {
uint16_t rx_frame;
if (SPI_WORD_SIZE_GET(data->ctx.config->operation) == 8) {
rx_frame = LL_SPI_ReceiveData8(spi);
UNALIGNED_PUT(rx_frame, (uint8_t *)data->ctx.rx_buf);
spi_context_update_rx(&data->ctx, 1, 1);
} else {
rx_frame = LL_SPI_ReceiveData16(spi);
UNALIGNED_PUT(rx_frame, (uint16_t *)data->ctx.rx_buf);
spi_context_update_rx(&data->ctx, 2, 1);
}
}
}
/*
* Without a FIFO, we can only shift out one frame's worth of SPI
* data, and read the response back.
*
* TODO: support 16-bit data frames.
*/
static int spi_stm32_shift_frames(SPI_TypeDef *spi, struct spi_stm32_data *data)
{
uint16_t operation = data->ctx.config->operation;
if (SPI_OP_MODE_GET(operation) == SPI_OP_MODE_MASTER) {
spi_stm32_shift_m(spi, data);
} else {
spi_stm32_shift_s(spi, data);
}
return spi_stm32_get_err(spi);
}
static void spi_stm32_complete(struct spi_stm32_data *data, SPI_TypeDef *spi,
int status)
{
#ifdef CONFIG_SPI_STM32_INTERRUPT
ll_func_disable_int_tx_empty(spi);
ll_func_disable_int_rx_not_empty(spi);
ll_func_disable_int_errors(spi);
#endif
spi_context_cs_control(&data->ctx, false);
#if DT_HAS_COMPAT_STATUS_OKAY(st_stm32_spi_fifo)
/* Flush RX buffer */
while (ll_func_rx_is_not_empty(spi)) {
(void) LL_SPI_ReceiveData8(spi);
}
#endif
if (LL_SPI_GetMode(spi) == LL_SPI_MODE_MASTER) {
while (ll_func_spi_is_busy(spi)) {
/* NOP */
}
}
/* BSY flag is cleared when MODF flag is raised */
if (LL_SPI_IsActiveFlag_MODF(spi)) {
LL_SPI_ClearFlag_MODF(spi);
}
ll_func_disable_spi(spi);
#ifdef CONFIG_SPI_STM32_INTERRUPT
spi_context_complete(&data->ctx, status);
#endif
}
#ifdef CONFIG_SPI_STM32_INTERRUPT
static void spi_stm32_isr(const struct device *dev)
{
const struct spi_stm32_config *cfg = dev->config;
struct spi_stm32_data *data = dev->data;
SPI_TypeDef *spi = cfg->spi;
int err;
err = spi_stm32_get_err(spi);
if (err) {
spi_stm32_complete(data, spi, err);
return;
}
if (spi_stm32_transfer_ongoing(data)) {
err = spi_stm32_shift_frames(spi, data);
}
if (err || !spi_stm32_transfer_ongoing(data)) {
spi_stm32_complete(data, spi, err);
}
}
#endif
static int spi_stm32_configure(const struct device *dev,
const struct spi_config *config)
{
const struct spi_stm32_config *cfg = DEV_CFG(dev);
struct spi_stm32_data *data = DEV_DATA(dev);
const uint32_t scaler[] = {
LL_SPI_BAUDRATEPRESCALER_DIV2,
LL_SPI_BAUDRATEPRESCALER_DIV4,
LL_SPI_BAUDRATEPRESCALER_DIV8,
LL_SPI_BAUDRATEPRESCALER_DIV16,
LL_SPI_BAUDRATEPRESCALER_DIV32,
LL_SPI_BAUDRATEPRESCALER_DIV64,
LL_SPI_BAUDRATEPRESCALER_DIV128,
LL_SPI_BAUDRATEPRESCALER_DIV256
};
SPI_TypeDef *spi = cfg->spi;
uint32_t clock;
int br;
if (spi_context_configured(&data->ctx, config)) {
/* Nothing to do */
return 0;
}
if ((SPI_WORD_SIZE_GET(config->operation) != 8)
&& (SPI_WORD_SIZE_GET(config->operation) != 16)) {
return -ENOTSUP;
}
if (clock_control_get_rate(device_get_binding(STM32_CLOCK_CONTROL_NAME),
(clock_control_subsys_t) &cfg->pclken, &clock) < 0) {
LOG_ERR("Failed call clock_control_get_rate");
return -EIO;
}
for (br = 1 ; br <= ARRAY_SIZE(scaler) ; ++br) {
uint32_t clk = clock >> br;
if (clk <= config->frequency) {
break;
}
}
if (br > ARRAY_SIZE(scaler)) {
LOG_ERR("Unsupported frequency %uHz, max %uHz, min %uHz",
config->frequency,
clock >> 1,
clock >> ARRAY_SIZE(scaler));
return -EINVAL;
}
LL_SPI_Disable(spi);
LL_SPI_SetBaudRatePrescaler(spi, scaler[br - 1]);
if (SPI_MODE_GET(config->operation) & SPI_MODE_CPOL) {
LL_SPI_SetClockPolarity(spi, LL_SPI_POLARITY_HIGH);
} else {
LL_SPI_SetClockPolarity(spi, LL_SPI_POLARITY_LOW);
}
if (SPI_MODE_GET(config->operation) & SPI_MODE_CPHA) {
LL_SPI_SetClockPhase(spi, LL_SPI_PHASE_2EDGE);
} else {
LL_SPI_SetClockPhase(spi, LL_SPI_PHASE_1EDGE);
}
LL_SPI_SetTransferDirection(spi, LL_SPI_FULL_DUPLEX);
if (config->operation & SPI_TRANSFER_LSB) {
LL_SPI_SetTransferBitOrder(spi, LL_SPI_LSB_FIRST);
} else {
LL_SPI_SetTransferBitOrder(spi, LL_SPI_MSB_FIRST);
}
LL_SPI_DisableCRC(spi);
if (config->cs || !IS_ENABLED(CONFIG_SPI_STM32_USE_HW_SS)) {
#if defined(CONFIG_SOC_SERIES_STM32MP1X) || \
defined(CONFIG_SOC_SERIES_STM32H7X)
if (SPI_OP_MODE_GET(config->operation) == SPI_OP_MODE_MASTER) {
if (LL_SPI_GetNSSPolarity(spi) == LL_SPI_NSS_POLARITY_LOW)
LL_SPI_SetInternalSSLevel(spi, LL_SPI_SS_LEVEL_HIGH);
}
#endif
LL_SPI_SetNSSMode(spi, LL_SPI_NSS_SOFT);
} else {
if (config->operation & SPI_OP_MODE_SLAVE) {
LL_SPI_SetNSSMode(spi, LL_SPI_NSS_HARD_INPUT);
} else {
LL_SPI_SetNSSMode(spi, LL_SPI_NSS_HARD_OUTPUT);
}
}
if (config->operation & SPI_OP_MODE_SLAVE) {
LL_SPI_SetMode(spi, LL_SPI_MODE_SLAVE);
} else {
LL_SPI_SetMode(spi, LL_SPI_MODE_MASTER);
}
if (SPI_WORD_SIZE_GET(config->operation) == 8) {
LL_SPI_SetDataWidth(spi, LL_SPI_DATAWIDTH_8BIT);
} else {
LL_SPI_SetDataWidth(spi, LL_SPI_DATAWIDTH_16BIT);
}
#if DT_HAS_COMPAT_STATUS_OKAY(st_stm32_spi_fifo)
ll_func_set_fifo_threshold_8bit(spi);
#endif
#ifndef CONFIG_SOC_SERIES_STM32F1X
LL_SPI_SetStandard(spi, LL_SPI_PROTOCOL_MOTOROLA);
#endif
/* At this point, it's mandatory to set this on the context! */
data->ctx.config = config;
spi_context_cs_configure(&data->ctx);
LOG_DBG("Installed config %p: freq %uHz (div = %u),"
" mode %u/%u/%u, slave %u",
config, clock >> br, 1 << br,
(SPI_MODE_GET(config->operation) & SPI_MODE_CPOL) ? 1 : 0,
(SPI_MODE_GET(config->operation) & SPI_MODE_CPHA) ? 1 : 0,
(SPI_MODE_GET(config->operation) & SPI_MODE_LOOP) ? 1 : 0,
config->slave);
return 0;
}
static int spi_stm32_release(const struct device *dev,
const struct spi_config *config)
{
struct spi_stm32_data *data = DEV_DATA(dev);
spi_context_unlock_unconditionally(&data->ctx);
return 0;
}
static int transceive(const struct device *dev,
const struct spi_config *config,
const struct spi_buf_set *tx_bufs,
const struct spi_buf_set *rx_bufs,
bool asynchronous, struct k_poll_signal *signal)
{
const struct spi_stm32_config *cfg = DEV_CFG(dev);
struct spi_stm32_data *data = DEV_DATA(dev);
SPI_TypeDef *spi = cfg->spi;
int ret;
if (!tx_bufs && !rx_bufs) {
return 0;
}
#ifndef CONFIG_SPI_STM32_INTERRUPT
if (asynchronous) {
return -ENOTSUP;
}
#endif
spi_context_lock(&data->ctx, asynchronous, signal, config);
ret = spi_stm32_configure(dev, config);
if (ret) {
return ret;
}
/* Set buffers info */
spi_context_buffers_setup(&data->ctx, tx_bufs, rx_bufs, 1);
#if DT_HAS_COMPAT_STATUS_OKAY(st_stm32_spi_fifo)
/* Flush RX buffer */
while (ll_func_rx_is_not_empty(spi)) {
(void) LL_SPI_ReceiveData8(spi);
}
#endif
LL_SPI_Enable(spi);
/* This is turned off in spi_stm32_complete(). */
spi_context_cs_control(&data->ctx, true);
#ifdef CONFIG_SPI_STM32_INTERRUPT
ll_func_enable_int_errors(spi);
if (rx_bufs) {
ll_func_enable_int_rx_not_empty(spi);
}
ll_func_enable_int_tx_empty(spi);
ret = spi_context_wait_for_completion(&data->ctx);
#else
do {
ret = spi_stm32_shift_frames(spi, data);
} while (!ret && spi_stm32_transfer_ongoing(data));
spi_stm32_complete(data, spi, ret);
#ifdef CONFIG_SPI_SLAVE
if (spi_context_is_slave(&data->ctx) && !ret) {
ret = data->ctx.recv_frames;
}
#endif /* CONFIG_SPI_SLAVE */
#endif
spi_context_release(&data->ctx, ret);
return ret;
}
#ifdef CONFIG_SPI_STM32_DMA
static int wait_dma_rx_tx_done(const struct device *dev)
{
struct spi_stm32_data *data = DEV_DATA(dev);
int res = -1;
while (1) {
res = k_sem_take(&data->status_sem, K_MSEC(1000));
if (res != 0) {
return res;
}
if (data->status_flags & SPI_STM32_DMA_ERROR_FLAG) {
return -EIO;
}
if (data->status_flags & SPI_STM32_DMA_DONE_FLAG) {
return 0;
}
}
return res;
}
static int transceive_dma(const struct device *dev,
const struct spi_config *config,
const struct spi_buf_set *tx_bufs,
const struct spi_buf_set *rx_bufs,
bool asynchronous, struct k_poll_signal *signal)
{
const struct spi_stm32_config *cfg = DEV_CFG(dev);
struct spi_stm32_data *data = DEV_DATA(dev);
SPI_TypeDef *spi = cfg->spi;
int ret;
if (!tx_bufs && !rx_bufs) {
return 0;
}
if (asynchronous) {
return -ENOTSUP;
}
spi_context_lock(&data->ctx, asynchronous, signal, config);
k_sem_reset(&data->status_sem);
ret = spi_stm32_configure(dev, config);
if (ret != 0) {
return ret;
}
/* Set buffers info */
spi_context_buffers_setup(&data->ctx, tx_bufs, rx_bufs, 1);
/* This is turned off in spi_stm32_complete(). */
spi_context_cs_control(&data->ctx, true);
LL_SPI_Enable(spi);
while (data->ctx.rx_len > 0 || data->ctx.tx_len > 0) {
size_t dma_len;
if (data->ctx.rx_len == 0) {
dma_len = data->ctx.tx_len;
} else if (data->ctx.tx_len == 0) {
dma_len = data->ctx.rx_len;
} else {
dma_len = MIN(data->ctx.tx_len, data->ctx.rx_len);
}
data->status_flags = 0;
ret = spi_dma_move_buffers(dev, dma_len);
if (ret != 0) {
break;
}
LL_SPI_EnableDMAReq_RX(spi);
LL_SPI_EnableDMAReq_TX(spi);
ret = wait_dma_rx_tx_done(dev);
if (ret != 0) {
break;
}
#ifdef SPI_SR_FTLVL
while (LL_SPI_GetTxFIFOLevel(spi) > 0) {
}
#endif
/* wait until TX buffer is really empty */
while (LL_SPI_IsActiveFlag_TXE(spi) == 0) {
}
/* wait until hardware is really ready */
while (LL_SPI_IsActiveFlag_BSY(spi) == 1) {
}
LL_SPI_DisableDMAReq_TX(spi);
LL_SPI_DisableDMAReq_RX(spi);
spi_context_update_tx(&data->ctx, 1, dma_len);
spi_context_update_rx(&data->ctx, 1, dma_len);
}
LL_SPI_Disable(spi);
LL_SPI_DisableDMAReq_TX(spi);
LL_SPI_DisableDMAReq_RX(spi);
dma_stop(data->dma_rx.dma_dev, data->dma_rx.channel);
dma_stop(data->dma_tx.dma_dev, data->dma_tx.channel);
spi_stm32_complete(data, spi, ret);
spi_context_release(&data->ctx, ret);
return ret;
}
#endif /* CONFIG_SPI_STM32_DMA */
static int spi_stm32_transceive(const struct device *dev,
const struct spi_config *config,
const struct spi_buf_set *tx_bufs,
const struct spi_buf_set *rx_bufs)
{
#ifdef CONFIG_SPI_STM32_DMA
struct spi_stm32_data *data = DEV_DATA(dev);
if ((data->dma_tx.dma_name != NULL)
&& (data->dma_rx.dma_name != NULL)) {
return transceive_dma(dev, config, tx_bufs, rx_bufs,
false, NULL);
}
#endif /* CONFIG_SPI_STM32_DMA */
return transceive(dev, config, tx_bufs, rx_bufs, false, NULL);
}
#ifdef CONFIG_SPI_ASYNC
static int spi_stm32_transceive_async(const struct device *dev,
const struct spi_config *config,
const struct spi_buf_set *tx_bufs,
const struct spi_buf_set *rx_bufs,
struct k_poll_signal *async)
{
return transceive(dev, config, tx_bufs, rx_bufs, true, async);
}
#endif /* CONFIG_SPI_ASYNC */
static const struct spi_driver_api api_funcs = {
.transceive = spi_stm32_transceive,
#ifdef CONFIG_SPI_ASYNC
.transceive_async = spi_stm32_transceive_async,
#endif
.release = spi_stm32_release,
};
static int spi_stm32_init(const struct device *dev)
{
struct spi_stm32_data *data __attribute__((unused)) = dev->data;
const struct spi_stm32_config *cfg = dev->config;
int err;
__ASSERT_NO_MSG(device_get_binding(STM32_CLOCK_CONTROL_NAME));
if (clock_control_on(device_get_binding(STM32_CLOCK_CONTROL_NAME),
(clock_control_subsys_t) &cfg->pclken) != 0) {
LOG_ERR("Could not enable SPI clock");
return -EIO;
}
/* Configure dt provided device signals when available */
err = stm32_dt_pinctrl_configure(cfg->pinctrl_list,
cfg->pinctrl_list_size,
(uint32_t)cfg->spi);
if (err < 0) {
LOG_ERR("SPI pinctrl setup failed (%d)", err);
return err;
}
#ifdef CONFIG_SPI_STM32_INTERRUPT
cfg->irq_config(dev);
#endif
#ifdef CONFIG_SPI_STM32_DMA
if (data->dma_tx.dma_name != NULL) {
/* Get the binding to the DMA device */
data->dma_tx.dma_dev = device_get_binding(data->dma_tx.dma_name);
if (!data->dma_tx.dma_dev) {
LOG_ERR("%s device not found", data->dma_tx.dma_name);
return -ENODEV;
}
}
if (data->dma_rx.dma_name != NULL) {
data->dma_rx.dma_dev = device_get_binding(data->dma_rx.dma_name);
if (!data->dma_rx.dma_dev) {
LOG_ERR("%s device not found", data->dma_rx.dma_name);
return -ENODEV;
}
}
#endif /* CONFIG_SPI_STM32_DMA */
spi_context_unlock_unconditionally(&data->ctx);
return 0;
}
#ifdef CONFIG_SPI_STM32_INTERRUPT
#define STM32_SPI_IRQ_HANDLER_DECL(id) \
static void spi_stm32_irq_config_func_##id(const struct device *dev)
#define STM32_SPI_IRQ_HANDLER_FUNC(id) \
.irq_config = spi_stm32_irq_config_func_##id,
#define STM32_SPI_IRQ_HANDLER(id) \
static void spi_stm32_irq_config_func_##id(const struct device *dev) \
{ \
IRQ_CONNECT(DT_INST_IRQN(id), \
DT_INST_IRQ(id, priority), \
spi_stm32_isr, DEVICE_DT_INST_GET(id), 0); \
irq_enable(DT_INST_IRQN(id)); \
}
#else
#define STM32_SPI_IRQ_HANDLER_DECL(id)
#define STM32_SPI_IRQ_HANDLER_FUNC(id)
#define STM32_SPI_IRQ_HANDLER(id)
#endif
#define DMA_CHANNEL_CONFIG(id, dir) \
DT_INST_DMAS_CELL_BY_NAME(id, dir, channel_config)
#define DMA_FEATURES(id, dir) \
DT_INST_DMAS_CELL_BY_NAME(id, dir, features)
#define SPI_DMA_CHANNEL_INIT(index, dir, dir_cap, src_dev, dest_dev) \
.dma_name = DT_INST_DMAS_LABEL_BY_NAME(index, dir), \
.channel = \
DT_INST_DMAS_CELL_BY_NAME(index, dir, channel), \
.dma_cfg = { \
.dma_slot = \
DT_INST_DMAS_CELL_BY_NAME(index, dir, slot), \
.channel_direction = STM32_DMA_CONFIG_DIRECTION( \
DMA_CHANNEL_CONFIG(index, dir)), \
.source_data_size = STM32_DMA_CONFIG_##src_dev##_DATA_SIZE( \
DMA_CHANNEL_CONFIG(index, dir)), \
.dest_data_size = STM32_DMA_CONFIG_##dest_dev##_DATA_SIZE( \
DMA_CHANNEL_CONFIG(index, dir)), \
.source_burst_length = 1, /* SINGLE transfer */ \
.dest_burst_length = 1, /* SINGLE transfer */ \
.channel_priority = STM32_DMA_CONFIG_PRIORITY( \
DMA_CHANNEL_CONFIG(index, dir)),\
.dma_callback = dma_callback, \
.block_count = 2, \
}, \
.src_addr_increment = STM32_DMA_CONFIG_##src_dev##_ADDR_INC( \
DMA_CHANNEL_CONFIG(index, dir)), \
.dst_addr_increment = STM32_DMA_CONFIG_##dest_dev##_ADDR_INC( \
DMA_CHANNEL_CONFIG(index, dir)), \
.fifo_threshold = STM32_DMA_FEATURES_FIFO_THRESHOLD( \
DMA_FEATURES(index, dir)), \
#if CONFIG_SPI_STM32_DMA
#define SPI_DMA_CHANNEL(id, dir, DIR, src, dest) \
.dma_##dir = { \
COND_CODE_1(DT_INST_DMAS_HAS_NAME(id, dir), \
(SPI_DMA_CHANNEL_INIT(id, dir, DIR, src, dest)),\
(NULL)) \
},
#define SPI_DMA_STATUS_SEM(id) \
.status_sem = Z_SEM_INITIALIZER( \
spi_stm32_dev_data_##id.status_sem, 0, 1),
#else
#define SPI_DMA_CHANNEL(id, dir, DIR, src, dest)
#define SPI_DMA_STATUS_SEM(id)
#endif
#define STM32_SPI_INIT(id) \
STM32_SPI_IRQ_HANDLER_DECL(id); \
\
static const struct soc_gpio_pinctrl spi_pins_##id[] = \
ST_STM32_DT_INST_PINCTRL(id, 0); \
\
static const struct spi_stm32_config spi_stm32_cfg_##id = { \
.spi = (SPI_TypeDef *) DT_INST_REG_ADDR(id), \
.pclken = { \
.enr = DT_INST_CLOCKS_CELL(id, bits), \
.bus = DT_INST_CLOCKS_CELL(id, bus) \
}, \
.pinctrl_list = spi_pins_##id, \
.pinctrl_list_size = ARRAY_SIZE(spi_pins_##id), \
STM32_SPI_IRQ_HANDLER_FUNC(id) \
}; \
\
static struct spi_stm32_data spi_stm32_dev_data_##id = { \
SPI_CONTEXT_INIT_LOCK(spi_stm32_dev_data_##id, ctx), \
SPI_CONTEXT_INIT_SYNC(spi_stm32_dev_data_##id, ctx), \
SPI_DMA_CHANNEL(id, rx, RX, PERIPHERAL, MEMORY) \
SPI_DMA_CHANNEL(id, tx, TX, MEMORY, PERIPHERAL) \
SPI_DMA_STATUS_SEM(id) \
}; \
\
DEVICE_DT_INST_DEFINE(id, &spi_stm32_init, device_pm_control_nop, \
&spi_stm32_dev_data_##id, &spi_stm32_cfg_##id, \
POST_KERNEL, CONFIG_SPI_INIT_PRIORITY, \
&api_funcs); \
\
STM32_SPI_IRQ_HANDLER(id)
DT_INST_FOREACH_STATUS_OKAY(STM32_SPI_INIT)