Skip to content

Commit

Permalink
Lots of K64F SPI fixes (#315)
Browse files Browse the repository at this point in the history
* Update DSPI driver to latest version from NXP

* Other updates to SDK files

* Fix double DMA transactions, improve dummy byte support

* Simplify dummy byte setting

* Also fix SPI driver bug

* Fix infinite hang with interrupt SPI

* Better fix for SPI hang

* Remove earlier fix
  • Loading branch information
multiplemonomials authored Jul 28, 2024
1 parent de0c404 commit 5354079
Show file tree
Hide file tree
Showing 13 changed files with 1,463 additions and 519 deletions.
10 changes: 6 additions & 4 deletions drivers/include/drivers/SPI.h
Original file line number Diff line number Diff line change
Expand Up @@ -480,7 +480,7 @@ class SPI : private NonCopyable<SPI> {
typename std::enable_if<std::is_integral<WordT>::value, int>::type
transfer(const WordT *tx_buffer, int tx_length, CacheAlignedBuffer<WordT> &rx_buffer, int rx_length, const event_callback_t &callback, int event = SPI_EVENT_COMPLETE)
{
MBED_ASSERT(rx_length <= static_cast<int>(rx_buffer.capacity()));
MBED_ASSERT(rx_length <= static_cast<int>(rx_buffer.capacity() * sizeof(WordT)));
return transfer_internal(tx_buffer, tx_length, rx_buffer.data(), rx_length, callback, event);
}

Expand All @@ -489,7 +489,7 @@ class SPI : private NonCopyable<SPI> {
typename std::enable_if<std::is_integral<WordT>::value, int>::type
transfer(const std::nullptr_t tx_buffer, int tx_length, CacheAlignedBuffer<WordT> &rx_buffer, int rx_length, const event_callback_t &callback, int event = SPI_EVENT_COMPLETE)
{
MBED_ASSERT(rx_length <= static_cast<int>(rx_buffer.capacity()));
MBED_ASSERT(rx_length <= static_cast<int>(rx_buffer.capacity() * sizeof(WordT)));
return transfer_internal(tx_buffer, tx_length, rx_buffer.data(), rx_length, callback, event);
}
template<typename WordT>
Expand Down Expand Up @@ -528,7 +528,7 @@ class SPI : private NonCopyable<SPI> {
typename std::enable_if<std::is_integral<WordT>::value, int>::type
transfer_and_wait(const WordT *tx_buffer, int tx_length, CacheAlignedBuffer<WordT> &rx_buffer, int rx_length, rtos::Kernel::Clock::duration_u32 timeout = rtos::Kernel::wait_for_u32_forever)
{
MBED_ASSERT(rx_length <= static_cast<int>(rx_buffer.capacity()));
MBED_ASSERT(rx_length <= static_cast<int>(rx_buffer.capacity() * sizeof(WordT)));
return transfer_and_wait_internal(tx_buffer, tx_length, rx_buffer.data(), rx_length, timeout);
}

Expand All @@ -537,7 +537,7 @@ class SPI : private NonCopyable<SPI> {
typename std::enable_if<std::is_integral<WordT>::value, int>::type
transfer_and_wait(const std::nullptr_t tx_buffer, int tx_length, CacheAlignedBuffer<WordT> &rx_buffer, int rx_length, rtos::Kernel::Clock::duration_u32 timeout = rtos::Kernel::wait_for_u32_forever)
{
MBED_ASSERT(rx_length <= static_cast<int>(rx_buffer.capacity()));
MBED_ASSERT(rx_length <= static_cast<int>(rx_buffer.capacity() * sizeof(WordT)));
return transfer_and_wait_internal(tx_buffer, tx_length, rx_buffer.data(), rx_length, timeout);
}
template<typename WordT>
Expand All @@ -549,6 +549,8 @@ class SPI : private NonCopyable<SPI> {

/**
* @brief Abort the on-going SPI transfer, if any, and continue with transfers in the queue, if any.
*
* @note If a transfer is aborted, its callback will not be called at all.
*/
void abort_transfer();

Expand Down
3 changes: 3 additions & 0 deletions hal/include/hal/spi_api.h
Original file line number Diff line number Diff line change
Expand Up @@ -235,6 +235,9 @@ void spi_free(spi_t *obj);
*
* Set the number of bits per frame, configure clock polarity and phase, shift order and master/slave mode.
* The default bit order is MSB.
*
* This function shall NOT modify the SPI frequency if the SPI frequency has previously been set.
*
* @param[in,out] obj The SPI object to configure
* @param[in] bits The number of bits per frame
* @param[in] mode The SPI mode (clock polarity, phase, and shift direction)
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -38,6 +38,7 @@ target_sources(mbed-mcu-k64f
drivers/fsl_cmp.c
drivers/fsl_cmt.c
drivers/fsl_crc.c
drivers/fsl_common_arm.c
drivers/fsl_dac.c
drivers/fsl_dmamux.c
drivers/fsl_dspi.c
Expand Down

Large diffs are not rendered by default.

Original file line number Diff line number Diff line change
@@ -0,0 +1,233 @@
/*
* Copyright (c) 2015-2016, Freescale Semiconductor, Inc.
* Copyright 2016-2021 NXP
* All rights reserved.
*
* SPDX-License-Identifier: BSD-3-Clause
*/

#include "fsl_common.h"

/* Component ID definition, used by tools. */
#ifndef FSL_COMPONENT_ID
#define FSL_COMPONENT_ID "platform.drivers.common_arm"
#endif

#ifndef __GIC_PRIO_BITS
#if defined(ENABLE_RAM_VECTOR_TABLE)
uint32_t InstallIRQHandler(IRQn_Type irq, uint32_t irqHandler)
{
#ifdef __VECTOR_TABLE
#undef __VECTOR_TABLE
#endif

/* Addresses for VECTOR_TABLE and VECTOR_RAM come from the linker file */
#if defined(__CC_ARM) || defined(__ARMCC_VERSION)
extern uint32_t Image$$VECTOR_ROM$$Base[];
extern uint32_t Image$$VECTOR_RAM$$Base[];
extern uint32_t Image$$RW_m_data$$Base[];

#define __VECTOR_TABLE Image$$VECTOR_ROM$$Base
#define __VECTOR_RAM Image$$VECTOR_RAM$$Base
#define __RAM_VECTOR_TABLE_SIZE (((uint32_t)Image$$RW_m_data$$Base - (uint32_t)Image$$VECTOR_RAM$$Base))
#elif defined(__ICCARM__)
extern uint32_t __RAM_VECTOR_TABLE_SIZE[];
extern uint32_t __VECTOR_TABLE[];
extern uint32_t __VECTOR_RAM[];
#elif defined(__GNUC__)
extern uint32_t __VECTOR_TABLE[];
extern uint32_t __VECTOR_RAM[];
extern uint32_t __RAM_VECTOR_TABLE_SIZE_BYTES[];
uint32_t __RAM_VECTOR_TABLE_SIZE = (uint32_t)(__RAM_VECTOR_TABLE_SIZE_BYTES);
#endif /* defined(__CC_ARM) || defined(__ARMCC_VERSION) */
uint32_t n;
uint32_t ret;
uint32_t irqMaskValue;

irqMaskValue = DisableGlobalIRQ();
if (SCB->VTOR != (uint32_t)__VECTOR_RAM)
{
/* Copy the vector table from ROM to RAM */
for (n = 0; n < ((uint32_t)__RAM_VECTOR_TABLE_SIZE) / sizeof(uint32_t); n++)
{
__VECTOR_RAM[n] = __VECTOR_TABLE[n];
}
/* Point the VTOR to the position of vector table */
SCB->VTOR = (uint32_t)__VECTOR_RAM;
}

ret = __VECTOR_RAM[(int32_t)irq + 16];
/* make sure the __VECTOR_RAM is noncachable */
__VECTOR_RAM[(int32_t)irq + 16] = irqHandler;

EnableGlobalIRQ(irqMaskValue);

return ret;
}
#endif /* ENABLE_RAM_VECTOR_TABLE. */
#endif /* __GIC_PRIO_BITS. */

#if (defined(FSL_FEATURE_SOC_SYSCON_COUNT) && (FSL_FEATURE_SOC_SYSCON_COUNT > 0))

/*
* When FSL_FEATURE_POWERLIB_EXTEND is defined to non-zero value,
* powerlib should be used instead of these functions.
*/
#if !(defined(FSL_FEATURE_POWERLIB_EXTEND) && (FSL_FEATURE_POWERLIB_EXTEND != 0))

/*
* When the SYSCON STARTER registers are discontinuous, these functions are
* implemented in fsl_power.c.
*/
#if !(defined(FSL_FEATURE_SYSCON_STARTER_DISCONTINUOUS) && FSL_FEATURE_SYSCON_STARTER_DISCONTINUOUS)

void EnableDeepSleepIRQ(IRQn_Type interrupt)
{
uint32_t intNumber = (uint32_t)interrupt;

uint32_t index = 0;

while (intNumber >= 32u)
{
index++;
intNumber -= 32u;
}

SYSCON->STARTERSET[index] = 1UL << intNumber;
(void)EnableIRQ(interrupt); /* also enable interrupt at NVIC */
}

void DisableDeepSleepIRQ(IRQn_Type interrupt)
{
uint32_t intNumber = (uint32_t)interrupt;

(void)DisableIRQ(interrupt); /* also disable interrupt at NVIC */
uint32_t index = 0;

while (intNumber >= 32u)
{
index++;
intNumber -= 32u;
}

SYSCON->STARTERCLR[index] = 1UL << intNumber;
}
#endif /* FSL_FEATURE_SYSCON_STARTER_DISCONTINUOUS */
#endif /* FSL_FEATURE_POWERLIB_EXTEND */
#endif /* FSL_FEATURE_SOC_SYSCON_COUNT */

#if defined(SDK_DELAY_USE_DWT) && defined(DWT)
/* Use WDT. */
static void enableCpuCycleCounter(void)
{
/* Make sure the DWT trace fucntion is enabled. */
if (CoreDebug_DEMCR_TRCENA_Msk != (CoreDebug_DEMCR_TRCENA_Msk & CoreDebug->DEMCR))
{
CoreDebug->DEMCR |= CoreDebug_DEMCR_TRCENA_Msk;
}

/* CYCCNT not supported on this device. */
assert(DWT_CTRL_NOCYCCNT_Msk != (DWT->CTRL & DWT_CTRL_NOCYCCNT_Msk));

/* Read CYCCNT directly if CYCCENT has already been enabled, otherwise enable CYCCENT first. */
if (DWT_CTRL_CYCCNTENA_Msk != (DWT_CTRL_CYCCNTENA_Msk & DWT->CTRL))
{
DWT->CTRL |= DWT_CTRL_CYCCNTENA_Msk;
}
}

static uint32_t getCpuCycleCount(void)
{
return DWT->CYCCNT;
}
#else /* defined(SDK_DELAY_USE_DWT) && defined(DWT) */
/* Use software loop. */
#if defined(__CC_ARM) /* This macro is arm v5 specific */
/* clang-format off */
__ASM static void DelayLoop(uint32_t count)
{
loop
SUBS R0, R0, #1
CMP R0, #0
BNE loop
BX LR
}
/* clang-format on */
#elif defined(__ARMCC_VERSION) || defined(__ICCARM__) || defined(__GNUC__)
/* Cortex-M0 has a smaller instruction set, SUBS isn't supported in thumb-16 mode reported from __GNUC__ compiler,
* use SUB and CMP here for compatibility */
static void DelayLoop(uint32_t count)
{
__ASM volatile(" MOV R0, %0" : : "r"(count));
__ASM volatile(
"loop: \n"
#if defined(__GNUC__) && !defined(__ARMCC_VERSION)
" SUB R0, R0, #1 \n"
#else
" SUBS R0, R0, #1 \n"
#endif
" CMP R0, #0 \n"

" BNE loop \n"
:
:
: "r0");
}
#endif /* defined(__CC_ARM) */
#endif /* defined(SDK_DELAY_USE_DWT) && defined(DWT) */

/*!
* @brief Delay at least for some time.
* Please note that, if not uses DWT, this API will use while loop for delay, different run-time environments have
* effect on the delay time. If precise delay is needed, please enable DWT delay. The two parmeters delayTime_us and
* coreClock_Hz have limitation. For example, in the platform with 1GHz coreClock_Hz, the delayTime_us only supports
* up to 4294967 in current code. If long time delay is needed, please implement a new delay function.
*
* @param delayTime_us Delay time in unit of microsecond.
* @param coreClock_Hz Core clock frequency with Hz.
*/
void SDK_DelayAtLeastUs(uint32_t delayTime_us, uint32_t coreClock_Hz)
{
uint64_t count;

if (delayTime_us > 0U)
{
count = USEC_TO_COUNT(delayTime_us, coreClock_Hz);

assert(count <= UINT32_MAX);

#if defined(SDK_DELAY_USE_DWT) && defined(DWT) /* Use DWT for better accuracy */

enableCpuCycleCounter();
/* Calculate the count ticks. */
count += getCpuCycleCount();

if (count > UINT32_MAX)
{
count -= UINT32_MAX;
/* Wait for cyccnt overflow. */
while (count < getCpuCycleCount())
{
}
}

/* Wait for cyccnt reach count value. */
while (count > getCpuCycleCount())
{
}
#else
/* Divide value may be different in various environment to ensure delay is precise.
* Every loop count includes three instructions, due to Cortex-M7 sometimes executes
* two instructions in one period, through test here set divide 1.5. Other M cores use
* divide 4. By the way, divide 1.5 or 4 could let the count lose precision, but it does
* not matter because other instructions outside while loop is enough to fill the time.
*/
#if (__CORTEX_M == 7)
count = count / 3U * 2U;
#else
count = count / 4U;
#endif
DelayLoop((uint32_t)count);
#endif /* defined(SDK_DELAY_USE_DWT) && defined(DWT) */
}
}
Loading

0 comments on commit 5354079

Please sign in to comment.