[参考译文] IWR6843AOPEVM：用于闪存/SPI 数据的 UART — 无法运行 SPI

你好。

启用 SPI 组件时、您是否能够跟踪代码失败的位置？  可能是您为 CLI 和 SPI 任务授予了相同级别的优先级、如果某项任务的设置顺序不正确、则可能会导致代码崩溃。  您是否能够在 CCS 调试中运行自定义演示来确定发生崩溃的位置？

此致、

Santosh

您好、Santosh、

对不起,正忙两天来解决它 — 也与 Opus 4.5 的帮助下,但没有成功.

我还检查了所有导线并进行了 GPIO Bang 测试、看看是否相应地发送了传输的数据。  
我将很快更新我的代码、以便您可以查看它。 不用说、我不确定我的代码是错误的还是在 IWR6843AOP 或 MCU 上正确的。 我们先从 IWR6843 开始  

============================================================================
TI mmWave Forum Post - Summary
============================================================================

SUBJECT: IWR6843AOP SPI Slave Mode - SPI_transfer() Blocks Indefinitely

DEVICE: IWR6843AOP (not standard ISK)
SDK: mmWave SDK 03.06.02.00-LTS
APPLICATION: Overhead 3D People Tracking Demo

============================================================================
PROBLEM DESCRIPTION
============================================================================

We are trying to use IWR6843AOP as an SPI slave to stream radar data to an 
ESP32 master. The SPI_transfer() call in slave mode BLOCKS indefinitely, 
causing the CLI to freeze after sensorStart.

SYMPTOMS:
- SPI_open() succeeds
- All SOC configuration calls succeed
- GPIO12 (data-ready) toggles correctly
- SPI_transfer() is called with valid parameters
- SPI_transfer() NEVER returns (no timeout, no error)
- CLI becomes unresponsive
- Master (ESP32) is ready and clocking, but slave doesn't respond

HARDWARE SETUP:
- IWR6843AOP (slave) <-> ESP32P4 (master)
- 4-wire SPI: MOSI, MISO, CLK, CS
- GPIO12 (INT) for data-ready signal
- All wires verified with oscilloscope

============================================================================
CONFIGURATION ATTEMPTS
============================================================================

We have tried ALL of the following combinations (all result in blocking):

1. SPI_PINMODE_4PIN_CS vs SPI_PINMODE_3PIN
2. chipSelect = 0 vs 1
3. csHold = 0 vs 1
4. dataSize = 8 vs 16
5. frameFormat = SPI_POL0_PHA0 vs SPI_POL0_PHA1
6. DMA mode vs polling mode
7. Different timeout values (10, 100, 1000 ticks)

All configurations result in the same behavior: SPI_transfer() blocks forever.

============================================================================
CODE REFERENCE
============================================================================

See attached code file (TI_FORUM_POST_CODE.txt) for:
1. Pinmux configuration (AOP variant)
2. SPI initialization code
3. SPI_transfer() call that blocks
4. All SOC setup calls we're using

The code follows the TI test example (xwr68xx/main.c) structure:
- SOC_selectDMARequestMapping()
- SOC_selectInterruptRequestMapping()
- SOC_SPIOutputCtrl() with 4-pin mode (param=0)

============================================================================
WORKAROUND
============================================================================

We are currently using GPIO bit-bang mode to work around this issue, which
works but is slow. We would prefer to use the hardware SPI driver.

============================================================================
QUESTIONS
============================================================================

1. Is there a known issue with SPI_transfer() blocking in slave mode on 
   IWR6843AOP?

2. Are there any additional SOC calls or configuration steps required for 
   AOP variant that differ from standard ISK?

3. Should we be using a different SPI instance (SPI-B instead of SPI-A)?

4. Is there a way to make SPI_transfer() non-blocking or check its status?

5. Are there any hardware requirements we might be missing (pull-up 
   resistors, clock speed limits, etc.)?

6. Is there a specific order of operations required before calling 
   SPI_transfer() in slave mode?

============================================================================
ADDITIONAL INFORMATION
============================================================================

- We verified the MISO wire works using GPIO bit-bang test
- The master (ESP32) is correctly configured and ready
- We can see CLK toggling on oscilloscope
- CS is asserted correctly
- The issue is specifically with SPI_transfer() blocking

We would appreciate any guidance on how to properly use SPI slave mode
on IWR6843AOP, or confirmation if this is a known limitation.

============================================================================

// ============================================================================
// TI mmWave Forum Post - SPI Slave Mode Issue on IWR6843AOP
// ============================================================================
// 
// ISSUE: SPI_transfer() in slave mode BLOCKS indefinitely, causing CLI to freeze
// after sensorStart. The master (ESP32) is ready and clocking, but SPI_transfer
// never returns.
//
// DEVICE: IWR6843AOP (not standard ISK)
// SDK: mmWave SDK 03.06.02.00-LTS
// MODE: SPI Slave, 4-pin with CS
// ============================================================================

// ----------------------------------------------------------------------------
// 1. PINMUX CONFIGURATION (IWR6843AOP - different from standard ISK!)
// ----------------------------------------------------------------------------
// Reference: pinmux_xwr68xx.h
// AOP variant uses different pad indices than standard ISK

static void SPI_DataInterface_configureSpiPinmux(void)
{
    /* SPIA_MOSI - PIN F2 / PADAD (AOP variant) */
    Pinmux_Set_OverrideCtrl(SOC_XWR68XX_PIND13_PADAD, 
                            PINMUX_OUTEN_RETAIN_HW_CTRL, 
                            PINMUX_INPEN_RETAIN_HW_CTRL);
    Pinmux_Set_FuncSel(SOC_XWR68XX_PIND13_PADAD, 
                       SOC_XWR68XX_PIND13_PADAD_SPIA_MOSI);

    /* SPIA_MISO - PIN D1 / PADAE (AOP variant) */
    Pinmux_Set_OverrideCtrl(SOC_XWR68XX_PINE14_PADAE, 
                            PINMUX_OUTEN_RETAIN_HW_CTRL, 
                            PINMUX_INPEN_RETAIN_HW_CTRL);
    Pinmux_Set_FuncSel(SOC_XWR68XX_PINE14_PADAE, 
                       SOC_XWR68XX_PINE14_PADAE_SPIA_MISO);

    /* SPIA_CLK - PIN D2 / PADAF (AOP variant) */
    Pinmux_Set_OverrideCtrl(SOC_XWR68XX_PINE13_PADAF, 
                            PINMUX_OUTEN_RETAIN_HW_CTRL, 
                            PINMUX_INPEN_RETAIN_HW_CTRL);
    Pinmux_Set_FuncSel(SOC_XWR68XX_PINE13_PADAF, 
                       SOC_XWR68XX_PINE13_PADAF_SPIA_CLK);

    /* SPIA_CSN - PIN C2 / PADAG (AOP variant) */
    Pinmux_Set_OverrideCtrl(SOC_XWR68XX_PINE15_PADAG, 
                            PINMUX_OUTEN_RETAIN_HW_CTRL, 
                            PINMUX_INPEN_RETAIN_HW_CTRL);
    Pinmux_Set_FuncSel(SOC_XWR68XX_PINE15_PADAG, 
                       SOC_XWR68XX_PINE15_PADAG_SPIA_CSN);
}

// ----------------------------------------------------------------------------
// 2. SPI INITIALIZATION (SLAVE MODE)
// ----------------------------------------------------------------------------

int32_t SPI_DataInterface_init(SOC_Handle socHandle)
{
    SPI_Params      spiParams;
    int32_t         errCode = 0;
    
    /* Step 1: Configure pinmux */
    SPI_DataInterface_configureSpiPinmux();
    
    /* Step 2: Initialize SPI driver */
    SPI_init();
    
    /* Step 3: CRITICAL - Map DMA and interrupt requests (per TI test code) */
    if (SOC_selectDMARequestMapping(socHandle, SOC_MODULE_SPIA, &errCode) < 0)
    {
        System_printf("SPI: ERROR - SOC_selectDMARequestMapping failed\n");
        return -3;
    }
    
    if (SOC_selectInterruptRequestMapping(socHandle, SOC_MODULE_SPIA, &errCode) < 0)
    {
        System_printf("SPI: ERROR - SOC_selectInterruptRequestMapping failed\n");
        return -4;
    }
    
    /* Step 4: Enable SPI output control - 4-pin mode (per TI test code line 243) */
    if (SOC_SPIOutputCtrl(socHandle, 0U, 0U, &errCode) < 0)
    {
        System_printf("SPI: ERROR - SOC_SPIOutputCtrl failed\n");
        return -5;
    }
    
    /* Step 5: Configure SPI parameters for SLAVE mode */
    SPI_Params_init(&spiParams);
    spiParams.mode              = SPI_SLAVE;
    spiParams.dataSize          = 16U;              /* 16-bit data */
    spiParams.frameFormat       = SPI_POL0_PHA1;    /* Mode 1: CPOL=0, CPHA=1 */
    spiParams.pinMode           = SPI_PINMODE_4PIN_CS;
    spiParams.shiftFormat       = SPI_MSB_FIRST;
    spiParams.transferMode      = SPI_MODE_BLOCKING;
    spiParams.transferTimeout   = 100;              /* 100 ticks timeout */
    spiParams.eccEnable         = 0;
    spiParams.csHold            = 1;                /* Per TI test for external devices */
    
    /* Polling mode (no DMA) */
    spiParams.dmaEnable         = 0;
    spiParams.dmaHandle         = NULL;
    
    /* Slave mode chip select */
    spiParams.u.slaveParams.chipSelect = 0;
    
    /* Step 6: Open SPI */
    gSpiHandle = SPI_open(0, &spiParams);  /* SPI instance 0 = SPI-A */
    if (gSpiHandle == NULL)
    {
        System_printf("SPI: ERROR - SPI_open failed!\n");
        return -2;
    }
    
    System_printf("SPI: SPI-A opened in SLAVE mode (4-PIN, polling, 16-bit, Mode1)\n");
    return 0;
}

// ----------------------------------------------------------------------------
// 3. SPI TRANSFER (THIS IS WHERE IT BLOCKS!)
// ----------------------------------------------------------------------------

int32_t SPI_DataInterface_sendData(uint8_t* data, uint32_t length, uint32_t frameNumber)
{
    SPI_Transaction spiTransaction;
    SPI_DataHeader  spiHeader;
    bool            transferOk;
    
    /* Prepare SPI header with sync word */
    spiHeader.syncWord = 0xAA55AA55;
    spiHeader.dataLength = length;
    spiHeader.frameNumber = frameNumber;
    spiHeader.reserved = 0;
    
    /* Copy header to TX buffer */
    memcpy(gSpiTxBuffer, &spiHeader, sizeof(SPI_DataHeader));
    
    /* Setup SPI transaction */
    memset(&spiTransaction, 0, sizeof(SPI_Transaction));
    spiTransaction.count = sizeof(SPI_DataHeader);  /* 16 bytes */
    spiTransaction.txBuf = gSpiTxBuffer;
    spiTransaction.rxBuf = gSpiRxBuffer;
    spiTransaction.slaveIndex = 0;
    
    /* Assert GPIO HIGH to signal master that data is ready */
    GPIO_write(DATA_READY_GPIO_PIN, 1);
    
    System_printf("SPI: Starting transfer (GPIO=HIGH)...\n");
    System_flush();  /* This prints successfully */
    
    /* PROBLEM: This call BLOCKS indefinitely, even though:
     *   - Master (ESP32) is ready and clocking
     *   - CS is asserted
     *   - CLK is toggling
     *   - Timeout is set to 100 ticks
     * 
     * The CLI freezes here - no further prints, no timeout, nothing.
     */
    transferOk = SPI_transfer(gSpiHandle, &spiTransaction);
    
    /* This line is NEVER reached! */
    System_printf("SPI: Transfer completed (result=%d)\n", transferOk);
    
    GPIO_write(DATA_READY_GPIO_PIN, 0);
    
    return transferOk ? 0 : -4;
}

// ----------------------------------------------------------------------------
// 4. WHAT WE'VE TRIED (all failed):
// ----------------------------------------------------------------------------
// ✓ SOC_SPIOutputCtrl() - called correctly (4-pin mode, param=0)
// ✓ SOC_selectDMARequestMapping() - called correctly
// ✓ SOC_selectInterruptRequestMapping() - called correctly
// ✓ Pinmux configuration - verified for AOP variant
// ✓ chipSelect = 0 vs 1 - tried both, same result
// ✓ SPI_PINMODE_3PIN vs 4PIN - tried both, same result
// ✓ csHold = 0 vs 1 - tried both, same result
// ✓ dataSize = 8 vs 16 - tried both, same result
// ✓ frameFormat = SPI_POL0_PHA0 vs SPI_POL0_PHA1 - tried both
// ✓ DMA vs polling mode - tried both, same result
//
// WORKAROUND: Using GPIO bit-bang instead (works, but slow)
// ----------------------------------------------------------------------------

// ----------------------------------------------------------------------------
// 5. QUESTIONS FOR TI:
// ----------------------------------------------------------------------------
// 1. Is there a known issue with SPI_transfer() blocking in slave mode?
// 2. Are there any additional SOC calls required for IWR6843AOP?
// 3. Should we use a different SPI instance or configuration?
// 4. Is there a way to make SPI_transfer() non-blocking or check status?
// 5. Are there any hardware requirements (pull-ups, clock speed, etc.)?
// ----------------------------------------------------------------------------