blob: e75303727560a3e35471b4cce344e8f186db335c [file] [log] [blame]
/*
* Copyright (c) 2018, NXP
*
* SPDX-License-Identifier: Apache-2.0
*/
#define DT_DRV_COMPAT nxp_imx_lpspi
#include <errno.h>
#include <drivers/spi.h>
#include <drivers/clock_control.h>
#include <fsl_lpspi.h>
#include <logging/log.h>
#ifdef CONFIG_SPI_MCUX_LPSPI_DMA
#include <drivers/dma.h>
#endif
LOG_MODULE_REGISTER(spi_mcux_lpspi, CONFIG_SPI_LOG_LEVEL);
#include "spi_context.h"
#define CHIP_SELECT_COUNT 4
#define MAX_DATA_WIDTH 4096
struct spi_mcux_config {
LPSPI_Type *base;
const struct device *clock_dev;
clock_control_subsys_t clock_subsys;
void (*irq_config_func)(const struct device *dev);
uint32_t pcs_sck_delay;
uint32_t sck_pcs_delay;
uint32_t transfer_delay;
};
#ifdef CONFIG_SPI_MCUX_LPSPI_DMA
#define SPI_MCUX_LPSPI_DMA_ERROR_FLAG 0x01
#define SPI_MCUX_LPSPI_DMA_RX_DONE_FLAG 0x02
#define SPI_MCUX_LPSPI_DMA_TX_DONE_FLAG 0x04
#define SPI_MCUX_LPSPI_DMA_DONE_FLAG \
(SPI_MCUX_LPSPI_DMA_RX_DONE_FLAG | SPI_MCUX_LPSPI_DMA_TX_DONE_FLAG)
struct stream {
const struct device *dma_dev;
uint32_t channel; /* stores the channel for dma */
struct dma_config dma_cfg;
struct dma_block_config dma_blk_cfg;
};
#endif
struct spi_mcux_data {
const struct device *dev;
lpspi_master_handle_t handle;
struct spi_context ctx;
size_t transfer_len;
#ifdef CONFIG_SPI_MCUX_LPSPI_DMA
volatile uint32_t status_flags;
struct stream dma_rx;
struct stream dma_tx;
/* dummy value used for transferring NOP when tx buf is null */
uint32_t dummy_tx_buffer;
/* dummy value used to read RX data into when rx buf is null */
uint32_t dummy_rx_buffer;
#endif
};
static void spi_mcux_transfer_next_packet(const struct device *dev)
{
const struct spi_mcux_config *config = dev->config;
struct spi_mcux_data *data = dev->data;
LPSPI_Type *base = config->base;
struct spi_context *ctx = &data->ctx;
lpspi_transfer_t transfer;
status_t status;
if ((ctx->tx_len == 0) && (ctx->rx_len == 0)) {
/* nothing left to rx or tx, we're done! */
spi_context_cs_control(&data->ctx, false);
spi_context_complete(&data->ctx, 0);
return;
}
transfer.configFlags = kLPSPI_MasterPcsContinuous |
(ctx->config->slave << LPSPI_MASTER_PCS_SHIFT);
if (ctx->tx_len == 0) {
/* rx only, nothing to tx */
transfer.txData = NULL;
transfer.rxData = ctx->rx_buf;
transfer.dataSize = ctx->rx_len;
} else if (ctx->rx_len == 0) {
/* tx only, nothing to rx */
transfer.txData = (uint8_t *) ctx->tx_buf;
transfer.rxData = NULL;
transfer.dataSize = ctx->tx_len;
} else if (ctx->tx_len == ctx->rx_len) {
/* rx and tx are the same length */
transfer.txData = (uint8_t *) ctx->tx_buf;
transfer.rxData = ctx->rx_buf;
transfer.dataSize = ctx->tx_len;
} else if (ctx->tx_len > ctx->rx_len) {
/* Break up the tx into multiple transfers so we don't have to
* rx into a longer intermediate buffer. Leave chip select
* active between transfers.
*/
transfer.txData = (uint8_t *) ctx->tx_buf;
transfer.rxData = ctx->rx_buf;
transfer.dataSize = ctx->rx_len;
transfer.configFlags |= kLPSPI_MasterPcsContinuous;
} else {
/* Break up the rx into multiple transfers so we don't have to
* tx from a longer intermediate buffer. Leave chip select
* active between transfers.
*/
transfer.txData = (uint8_t *) ctx->tx_buf;
transfer.rxData = ctx->rx_buf;
transfer.dataSize = ctx->tx_len;
transfer.configFlags |= kLPSPI_MasterPcsContinuous;
}
if (!(ctx->tx_count <= 1 && ctx->rx_count <= 1)) {
transfer.configFlags |= kLPSPI_MasterPcsContinuous;
}
data->transfer_len = transfer.dataSize;
status = LPSPI_MasterTransferNonBlocking(base, &data->handle,
&transfer);
if (status != kStatus_Success) {
LOG_ERR("Transfer could not start");
}
}
static void spi_mcux_isr(const struct device *dev)
{
const struct spi_mcux_config *config = dev->config;
struct spi_mcux_data *data = dev->data;
LPSPI_Type *base = config->base;
LPSPI_MasterTransferHandleIRQ(base, &data->handle);
}
static void spi_mcux_master_transfer_callback(LPSPI_Type *base,
lpspi_master_handle_t *handle, status_t status, void *userData)
{
struct spi_mcux_data *data = userData;
spi_context_update_tx(&data->ctx, 1, data->transfer_len);
spi_context_update_rx(&data->ctx, 1, data->transfer_len);
spi_mcux_transfer_next_packet(data->dev);
}
static int spi_mcux_configure(const struct device *dev,
const struct spi_config *spi_cfg)
{
const struct spi_mcux_config *config = dev->config;
struct spi_mcux_data *data = dev->data;
LPSPI_Type *base = config->base;
lpspi_master_config_t master_config;
uint32_t clock_freq;
uint32_t word_size;
if (spi_context_configured(&data->ctx, spi_cfg)) {
/* This configuration is already in use */
return 0;
}
if (spi_cfg->operation & SPI_HALF_DUPLEX) {
LOG_ERR("Half-duplex not supported");
return -ENOTSUP;
}
LPSPI_MasterGetDefaultConfig(&master_config);
if (spi_cfg->slave > CHIP_SELECT_COUNT) {
LOG_ERR("Slave %d is greater than %d",
spi_cfg->slave,
CHIP_SELECT_COUNT);
return -EINVAL;
}
word_size = SPI_WORD_SIZE_GET(spi_cfg->operation);
if (word_size > MAX_DATA_WIDTH) {
LOG_ERR("Word size %d is greater than %d",
word_size, MAX_DATA_WIDTH);
return -EINVAL;
}
master_config.bitsPerFrame = word_size;
master_config.cpol =
(SPI_MODE_GET(spi_cfg->operation) & SPI_MODE_CPOL)
? kLPSPI_ClockPolarityActiveLow
: kLPSPI_ClockPolarityActiveHigh;
master_config.cpha =
(SPI_MODE_GET(spi_cfg->operation) & SPI_MODE_CPHA)
? kLPSPI_ClockPhaseSecondEdge
: kLPSPI_ClockPhaseFirstEdge;
master_config.direction =
(spi_cfg->operation & SPI_TRANSFER_LSB)
? kLPSPI_LsbFirst
: kLPSPI_MsbFirst;
master_config.baudRate = spi_cfg->frequency;
master_config.pcsToSckDelayInNanoSec = config->pcs_sck_delay;
master_config.lastSckToPcsDelayInNanoSec = config->sck_pcs_delay;
master_config.betweenTransferDelayInNanoSec = config->transfer_delay;
if (clock_control_get_rate(config->clock_dev, config->clock_subsys,
&clock_freq)) {
return -EINVAL;
}
LPSPI_MasterInit(base, &master_config, clock_freq);
LPSPI_MasterTransferCreateHandle(base, &data->handle,
spi_mcux_master_transfer_callback,
data);
LPSPI_SetDummyData(base, 0);
data->ctx.config = spi_cfg;
return 0;
}
#ifdef CONFIG_SPI_MCUX_LPSPI_DMA
/* This function is executed in the interrupt context */
static void spi_mcux_dma_callback(const struct device *dev, void *arg,
uint32_t channel, int status)
{
/* arg directly holds the spi device */
struct spi_mcux_data *data = arg;
if (status != 0) {
LOG_ERR("DMA callback error with channel %d.", channel);
data->status_flags |= SPI_MCUX_LPSPI_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_MCUX_LPSPI_DMA_TX_DONE_FLAG;
LOG_DBG("DMA TX Block Complete");
} else if (channel == data->dma_rx.channel) {
/* this part of the transfer ends */
data->status_flags |= SPI_MCUX_LPSPI_DMA_RX_DONE_FLAG;
LOG_DBG("DMA RX Block Complete");
} else {
LOG_ERR("DMA callback channel %d is not valid.",
channel);
data->status_flags |= SPI_MCUX_LPSPI_DMA_ERROR_FLAG;
}
}
spi_context_complete(&data->ctx, 0);
}
static int spi_mcux_dma_tx_load(const struct device *dev, const uint8_t *buf, size_t len)
{
const struct spi_mcux_config *cfg = dev->config;
struct spi_mcux_data *data = dev->data;
struct dma_block_config *blk_cfg;
LPSPI_Type *base = cfg->base;
/* 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));
if (buf == NULL) {
/* Treat the transfer as a peripheral to peripheral one, so that DMA
* reads from this address each time
*/
blk_cfg->source_address = (uint32_t)&data->dummy_tx_buffer;
stream->dma_cfg.channel_direction = PERIPHERAL_TO_PERIPHERAL;
} else {
/* tx direction has memory as source and periph as dest. */
blk_cfg->source_address = (uint32_t)buf;
stream->dma_cfg.channel_direction = MEMORY_TO_PERIPHERAL;
}
/* Enable scatter/gather */
blk_cfg->source_gather_en = 1;
/* Dest is LPSPI tx fifo */
blk_cfg->dest_address = LPSPI_GetTxRegisterAddress(base);
blk_cfg->block_size = len;
/* Transfer 1 byte each DMA loop */
stream->dma_cfg.source_burst_length = 1;
stream->dma_cfg.head_block = &stream->dma_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 */
return dma_config(data->dma_tx.dma_dev, data->dma_tx.channel,
&stream->dma_cfg);
}
static int spi_mcux_dma_rx_load(const struct device *dev, uint8_t *buf,
size_t len)
{
const struct spi_mcux_config *cfg = dev->config;
struct spi_mcux_data *data = dev->data;
struct dma_block_config *blk_cfg;
LPSPI_Type *base = cfg->base;
/* 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));
if (buf == NULL) {
/* Treat the transfer as a peripheral to peripheral one, so that DMA
* reads from this address each time
*/
blk_cfg->dest_address = (uint32_t)&data->dummy_rx_buffer;
stream->dma_cfg.channel_direction = PERIPHERAL_TO_PERIPHERAL;
} else {
/* rx direction has periph as source and mem as dest. */
blk_cfg->dest_address = (uint32_t)buf;
stream->dma_cfg.channel_direction = PERIPHERAL_TO_MEMORY;
}
blk_cfg->block_size = len;
/* Enable scatter/gather */
blk_cfg->dest_scatter_en = 1;
/* Source is LPSPI rx fifo */
blk_cfg->source_address = LPSPI_GetRxRegisterAddress(base);
stream->dma_cfg.source_burst_length = 1;
stream->dma_cfg.head_block = blk_cfg;
stream->dma_cfg.user_data = data;
/* pass our client origin to the dma: data->dma_rx.channel */
return dma_config(data->dma_rx.dma_dev, data->dma_rx.channel,
&stream->dma_cfg);
}
static int wait_dma_rx_tx_done(const struct device *dev)
{
struct spi_mcux_data *data = dev->data;
int ret = -1;
while (1) {
ret = spi_context_wait_for_completion(&data->ctx);
if (ret) {
LOG_DBG("Timed out waiting for SPI context to complete");
return ret;
}
if (data->status_flags & SPI_MCUX_LPSPI_DMA_ERROR_FLAG) {
return -EIO;
}
if ((data->status_flags & SPI_MCUX_LPSPI_DMA_DONE_FLAG) ==
SPI_MCUX_LPSPI_DMA_DONE_FLAG) {
LOG_DBG("DMA block completed");
return 0;
}
}
}
static int transceive_dma(const struct device *dev,
const struct spi_config *spi_cfg,
const struct spi_buf_set *tx_bufs,
const struct spi_buf_set *rx_bufs,
bool asynchronous,
struct k_poll_signal *sig)
{
const struct spi_mcux_config *config = dev->config;
struct spi_mcux_data *data = dev->data;
LPSPI_Type *base = config->base;
int ret;
size_t dma_size;
spi_context_lock(&data->ctx, asynchronous, sig, spi_cfg);
ret = spi_mcux_configure(dev, spi_cfg);
if (ret) {
goto out;
}
spi_context_buffers_setup(&data->ctx, tx_bufs, rx_bufs, 1);
spi_context_cs_control(&data->ctx, true);
/* DMA is fast enough watermarks are not required */
LPSPI_SetFifoWatermarks(base, 0U, 0U);
/* Send each spi buf via DMA, updating context as DMA completes */
while (data->ctx.rx_len > 0 || data->ctx.tx_len > 0) {
/* Clear status flags */
data->status_flags = 0U;
/* Load dma blocks of equal length */
dma_size = MIN(data->ctx.tx_len, data->ctx.rx_len);
if (dma_size == 0) {
dma_size = MAX(data->ctx.tx_len, data->ctx.rx_len);
}
ret = spi_mcux_dma_tx_load(dev, data->ctx.tx_buf, dma_size);
if (ret != 0) {
goto out;
}
ret = spi_mcux_dma_rx_load(dev, data->ctx.rx_buf, dma_size);
if (ret != 0) {
goto out;
}
/* Start DMA */
ret = dma_start(data->dma_tx.dma_dev, data->dma_tx.channel);
if (ret != 0) {
goto out;
}
ret = dma_start(data->dma_rx.dma_dev, data->dma_rx.channel);
if (ret != 0) {
goto out;
}
/* Enable DMA Requests */
LPSPI_EnableDMA(base, kLPSPI_TxDmaEnable | kLPSPI_RxDmaEnable);
/* Wait for DMA to finish */
ret = wait_dma_rx_tx_done(dev);
if (ret != 0) {
goto out;
}
while ((LPSPI_GetStatusFlags(base) & kLPSPI_ModuleBusyFlag)) {
/* wait until module is idle */
}
/* Disable DMA */
LPSPI_DisableDMA(base, kLPSPI_TxDmaEnable | kLPSPI_RxDmaEnable);
/* Update SPI contexts with amount of data we just sent */
spi_context_update_tx(&data->ctx, 1, dma_size);
spi_context_update_rx(&data->ctx, 1, dma_size);
}
spi_context_cs_control(&data->ctx, false);
out:
spi_context_release(&data->ctx, ret);
return ret;
}
#endif
static int transceive(const struct device *dev,
const struct spi_config *spi_cfg,
const struct spi_buf_set *tx_bufs,
const struct spi_buf_set *rx_bufs,
bool asynchronous,
struct k_poll_signal *signal)
{
struct spi_mcux_data *data = dev->data;
int ret;
spi_context_lock(&data->ctx, asynchronous, signal, spi_cfg);
ret = spi_mcux_configure(dev, spi_cfg);
if (ret) {
goto out;
}
spi_context_buffers_setup(&data->ctx, tx_bufs, rx_bufs, 1);
spi_context_cs_control(&data->ctx, true);
spi_mcux_transfer_next_packet(dev);
ret = spi_context_wait_for_completion(&data->ctx);
out:
spi_context_release(&data->ctx, ret);
return ret;
}
static int spi_mcux_transceive(const struct device *dev,
const struct spi_config *spi_cfg,
const struct spi_buf_set *tx_bufs,
const struct spi_buf_set *rx_bufs)
{
#ifdef CONFIG_SPI_MCUX_LPSPI_DMA
return transceive_dma(dev, spi_cfg, tx_bufs, rx_bufs, false, NULL);
#endif /* CONFIG_SPI_MCUX_LPSPI_DMA */
return transceive(dev, spi_cfg, tx_bufs, rx_bufs, false, NULL);
}
#ifdef CONFIG_SPI_ASYNC
static int spi_mcux_transceive_async(const struct device *dev,
const struct spi_config *spi_cfg,
const struct spi_buf_set *tx_bufs,
const struct spi_buf_set *rx_bufs,
struct k_poll_signal *async)
{
return transceive(dev, spi_cfg, tx_bufs, rx_bufs, true, async);
}
#endif /* CONFIG_SPI_ASYNC */
static int spi_mcux_release(const struct device *dev,
const struct spi_config *spi_cfg)
{
struct spi_mcux_data *data = dev->data;
spi_context_unlock_unconditionally(&data->ctx);
return 0;
}
static int spi_mcux_init(const struct device *dev)
{
int err;
const struct spi_mcux_config *config = dev->config;
struct spi_mcux_data *data = dev->data;
config->irq_config_func(dev);
err = spi_context_cs_configure_all(&data->ctx);
if (err < 0) {
return err;
}
spi_context_unlock_unconditionally(&data->ctx);
data->dev = dev;
#ifdef CONFIG_SPI_MCUX_LPSPI_DMA
if (!device_is_ready(data->dma_tx.dma_dev)) {
LOG_ERR("%s device is not ready", data->dma_tx.dma_dev->name);
return -ENODEV;
}
if (!device_is_ready(data->dma_rx.dma_dev)) {
LOG_ERR("%s device is not ready", data->dma_rx.dma_dev->name);
return -ENODEV;
}
#endif /* CONFIG_SPI_MCUX_LPSPI_DMA */
spi_context_unlock_unconditionally(&data->ctx);
return 0;
}
static const struct spi_driver_api spi_mcux_driver_api = {
.transceive = spi_mcux_transceive,
#ifdef CONFIG_SPI_ASYNC
.transceive_async = spi_mcux_transceive_async,
#endif
.release = spi_mcux_release,
};
#ifdef CONFIG_SPI_MCUX_LPSPI_DMA
#define SPI_DMA_CHANNELS(n) \
.dma_tx = { \
.dma_dev = DEVICE_DT_GET(DT_INST_DMAS_CTLR_BY_NAME(n, tx)), \
.channel = \
DT_INST_DMAS_CELL_BY_NAME(n, tx, mux), \
.dma_cfg = { \
.channel_direction = MEMORY_TO_PERIPHERAL, \
.dma_callback = spi_mcux_dma_callback, \
.source_data_size = 1, \
.dest_data_size = 1, \
.block_count = 1, \
.dma_slot = DT_INST_DMAS_CELL_BY_NAME(n, tx, source) \
} \
}, \
.dma_rx = { \
.dma_dev = DEVICE_DT_GET(DT_INST_DMAS_CTLR_BY_NAME(n, rx)), \
.channel = \
DT_INST_DMAS_CELL_BY_NAME(n, rx, mux), \
.dma_cfg = { \
.channel_direction = PERIPHERAL_TO_MEMORY, \
.dma_callback = spi_mcux_dma_callback, \
.source_data_size = 1, \
.dest_data_size = 1, \
.block_count = 1, \
.dma_slot = DT_INST_DMAS_CELL_BY_NAME(n, rx, source) \
} \
}
#else
#define SPI_DMA_CHANNELS(n)
#endif
#define SPI_MCUX_LPSPI_INIT(n) \
static void spi_mcux_config_func_##n(const struct device *dev); \
\
static const struct spi_mcux_config spi_mcux_config_##n = { \
.base = (LPSPI_Type *) DT_INST_REG_ADDR(n), \
.clock_dev = DEVICE_DT_GET(DT_INST_CLOCKS_CTLR(n)), \
.clock_subsys = \
(clock_control_subsys_t)DT_INST_CLOCKS_CELL(n, name), \
.irq_config_func = spi_mcux_config_func_##n, \
.pcs_sck_delay = UTIL_AND( \
DT_INST_NODE_HAS_PROP(n, pcs_sck_delay), \
DT_INST_PROP(n, pcs_sck_delay)), \
.sck_pcs_delay = UTIL_AND( \
DT_INST_NODE_HAS_PROP(n, sck_pcs_delay), \
DT_INST_PROP(n, sck_pcs_delay)), \
.transfer_delay = UTIL_AND( \
DT_INST_NODE_HAS_PROP(n, transfer_delay), \
DT_INST_PROP(n, transfer_delay)), \
}; \
\
static struct spi_mcux_data spi_mcux_data_##n = { \
SPI_CONTEXT_INIT_LOCK(spi_mcux_data_##n, ctx), \
SPI_CONTEXT_INIT_SYNC(spi_mcux_data_##n, ctx), \
SPI_CONTEXT_CS_GPIOS_INITIALIZE(DT_DRV_INST(n), ctx) \
SPI_DMA_CHANNELS(n) \
}; \
\
DEVICE_DT_INST_DEFINE(n, &spi_mcux_init, NULL, \
&spi_mcux_data_##n, \
&spi_mcux_config_##n, POST_KERNEL, \
CONFIG_SPI_INIT_PRIORITY, \
&spi_mcux_driver_api); \
\
static void spi_mcux_config_func_##n(const struct device *dev) \
{ \
IRQ_CONNECT(DT_INST_IRQN(n), DT_INST_IRQ(n, priority), \
spi_mcux_isr, DEVICE_DT_INST_GET(n), 0); \
\
irq_enable(DT_INST_IRQN(n)); \
}
DT_INST_FOREACH_STATUS_OKAY(SPI_MCUX_LPSPI_INIT)