ESP32-C5 SDIO slave not receiving data from STM32 SPI master (MOSI/CMD issue)

[KimChungJae]

ESP32-C5 SDIO slave not receiving data from STM32 SPI master (MOSI/CMD issue)

Environment

ESP32 Side:

• Chip: ESP32-C5 (eco2)
• Firmware: ESP-AT v5.1.0.0-dev (factory_ESP32C5-SDIO.bin)
• Compile: Feb 5 2026 13:16:02
• ESP-IDF: v5.5.1-833-gcc569cbd80-dirty
• Mode: SDIO Slave (1-bit mode)

Host Side:

• MCU: STM32H750VBT6
• Interface: SPI6 as SDSPI master (software SDIO emulation)
• Clock: 468 kHz (Prescaler 256)
• Library: esp-at/examples/at_sdspi_host (ported to STM32H7)

Hardware Connection

STM32H750 Pin	Function	ESP32-C5 GPIO	SDIO Signal	Pull-up
PG13	SPI6_SCK	GPIO9	CLK	No
PG14	SPI6_MOSI	GPIO10	CMD	10kΩ ✅
PG12	SPI6_MISO	GPIO8	DAT0	10kΩ ✅
PA8	GPIO_EXTI8	GPIO7	DAT1 (HS)	10kΩ ✅
PA15	GPIO_Output	GPIO13	DAT3 (CS)	10kΩ ✅

Verified: All connections confirmed with multimeter, pull-up resistors present.

Problem Description

Symptom 1: MISO works, but MOSI doesn't

MISO (ESP32 → STM32): ✅ Working

STM32 receives data from ESP32:
- Pattern 1: 0xA5 0xA5 0xA5 0xA5 0xA5 0xA5 0xA5 0xA5
- Pattern 2: 0x3C 0xD1 0x05 0x90 0x61 0x74 0x05 0x90

This proves the hardware connection is good!

MOSI (STM32 → ESP32): ❌ Not working

STM32 sends:
- Idle clocks: 0xFF 0xFF 0xFF 0xFF ... (20 bytes)
- CMD0: 0x40 0x00 0x00 0x00 0x00 0x95 0xFF 0xFF ... (15 bytes)

ESP32 response: Nothing (timeout)

Error from STM32 sdspi library:
E sdspi_transaction: Not found response
E sdspi_transaction: CMD0 response error, expect 0x1, response 0

Symptom 2: ESP32 always logs error during initialization

ESP32 UART log:

I (933) main_task: Calling app_main()
I (933) at-workaround: no external 32k oscillator, disable it now.
Waiting 3 seconds for STM32 host to boot...
Starting AT interface...
I (4001) at-init: at param mode: 1
...
I (6941) at-init: module_name: ESP32C5-SDIO
I (6944) at-init: max tx power=78, ret=0
...
I (6960) at-init: v5.1.0.0-dev (unknown)
E (6964) at-sdio: invalid data:0 or len:-1      ← Always occurs!
I (6968) wifi:mode : sta (3c:dc:75:84:51:58) + softAP (3c:dc:75:84:51:59)

Error location: main/interface/sdio/at_sdio_task.c

• Line 50 in at_sdio_write_data(): if (len < 0 || data == NULL)
• Line 86 in at_sdio_read_data(): if (data == NULL || len < 0)

Timing: This error occurs 23ms after SDIO initialization, regardless of when STM32 sends CMD0.

Symptom 3: ESP32 stops reading MOSI after error

After the "invalid data" error appears:

• ✅ ESP32 continues to send data on MISO (STM32 receives it)
• ❌ ESP32 stops receiving data on MOSI (ignores STM32's CMD0)

Theory: After the error, ESP32 enters an error state and disables MOSI/CMD reception.

What We Tried

1. Timing adjustments ⏱️

Attempt 1: Added 3-second delay in app_main() before esp_at_init()

// C:\ESP\esp-at\main\app_main.c
void app_main(void)
{
    esp_at_main_preprocess();
    ESP_ERROR_CHECK(nvs_flash_init());
    ESP_ERROR_CHECK(esp_at_netif_init());
    ESP_ERROR_CHECK(esp_event_loop_create_default());
    
    printf("Waiting 3 seconds for STM32 host to boot...\n");
    vTaskDelay(pdMS_TO_TICKS(3000));  // ← Added delay
    printf("Starting AT interface...\n");
    
    esp_at_init();
}

Result: "invalid data" error still occurs at the same timing (E (6964))

Attempt 2: STM32 waits 4.8s ~ 8s after ESP32 reset
Result: No difference, CMD0 still fails

2. SPI clock speed 🔄

• Tested: 15 MHz, 468 kHz
• Result: No difference

3. Hardware verification ✅

• Multimeter: All connections verified, continuity OK
• Oscilloscope: MISO shows data, MOSI shows clock but ESP32 doesn't respond
• Pull-ups: Confirmed 10kΩ on CMD, DAT0, DAT1, DAT3

4. DAT1 (HS) interrupt implementation

• Configured PA8 as GPIO_EXTI with falling edge trigger
• Implemented semaphore handler in sdspi_port.c
• Result: No change

5. DMA settings

• Tried: DMA_CIRCULAR, DMA_NORMAL modes
• Tried: Polling mode instead of DMA
• Result: No difference (hardware signals confirmed with oscilloscope)

Code References

STM32 SDSPI Port (working part)

Based on esp-at/examples/at_sdspi_host/STM32 example, ported to STM32H7:

• User/Src/sdspi_port.c: Hardware abstraction layer
• User/Src/sdspi_host.c: SDSPI protocol implementation (unchanged from example)
• User/Src/esp_sdspi.c: High-level API

Key function: at_spi_transmit()

esp_err_t at_spi_transmit(void* tx_buff, void* rx_buff, uint32_t len)
{
    // Cache operations for STM32H7
    SCB_CleanDCache_by_Addr((uint32_t*)tx_buff, len);
    
    // SPI transfer (tried both DMA and Polling modes)
    HAL_SPI_TransmitReceive(&hspi6, tx_buff, rx_buff, len, 1000);
    
    SCB_InvalidateDCache_by_Addr((uint32_t*)rx_buff, len);
    
    return ESP_OK;
}

Debug Logs

STM32 Debug Output:

[SDSPI] Resetting ESP32-C5...
[SDSPI] Wait 4.8s (ESP32 boot 2s + app delay 3s)...
TX[20]: FF FF FF FF FF FF FF FF (CS=1)    ← Idle clocks
RX[20]: A5 A5 A5 A5 A5 A5 A5 A5            ← ESP32 responds!
TX[15]: 40 00 00 00 00 95 FF FF (CS=0)    ← CMD0
RX[15]: 40 00 00 00 00 95 FF FF            ← Echo back (no ESP32 response)
E sdspi_transaction: Not found response
E sdspi_transaction: CMD0 response error, expect 0x1, response 0

ESP32 Debug Output:

I (933) main_task: Calling app_main()
Waiting 3 seconds for STM32 host to boot...
Starting AT interface...
I (4001) at-init: at param mode: 1
I (6941) at-init: module_name: ESP32C5-SDIO
E (6964) at-sdio: invalid data:0 or len:-1    ← Always occurs
I (6968) wifi:mode : sta (3c:dc:75:84:51:58) + softAP (3c:dc:75:84:51:59)

Questions

1. What causes "invalid data:0 or len:-1" error?

Looking at at_sdio_task.c:

static int32_t at_sdio_write_data(uint8_t *data, int32_t len)
{
    if (len < 0 || data == NULL) {
        ESP_LOGE(TAG, "invalid data:%p or len:%d", data, len);  // ← Line 50
        return -1;
    }
    // ...
}

Who calls this function with NULL/invalid args during initialization?

• Is this expected during AT core initialization?
• Should we ignore this error?

2. Why does ESP32 stop receiving MOSI after the error?

static void at_sdio_task(void *params)
{
    ulTaskNotifyTake(pdTRUE, portMAX_DELAY);
    
    for (;;) {
        esp_err_t ret = sdio_slave_recv(&handle, &addr, &size, portMAX_DELAY);  // ← Should wait forever
        // ...
    }
}

The task uses portMAX_DELAY, so it should wait forever for data from STM32. But after the "invalid data" error, ESP32 doesn't receive CMD0 from STM32.

Does the error state disable MOSI/CMD pin reception?

3. Is ESP32-C5 SDIO slave tested with SPI-based hosts?

The example at_sdspi_host is based on STM32F1 with hardware SDIO controller. We're using STM32H7 with SPI6 emulating SDIO.

Is software SDIO emulation (SPI) tested/supported with ESP32-C5?

Workaround Attempts

We tried modifying at_sdio_task.c to ignore the error:

// Commented out error log
if (len < 0 || data == NULL) {
    // ESP_LOGE(TAG, "invalid data:%p or len:%d", data, len);  // Commented
    return -1;
}

Result: Error log disappears, but ESP32 still doesn't receive CMD0.

Request

1. Root cause analysis: Why does "invalid data" error occur during initialization?
2. Workaround: How to make ESP32 receive MOSI/CMD data after initialization?
3. Confirmation: Is ESP32-C5 SDIO slave + STM32 SPI master a supported configuration?

We've spent over a month on this issue. Any help would be greatly appreciated!

Additional Information

• STM32 MISO reception works (proves hardware is OK)
• STM32 can read data from ESP32 (0xA5, 0x3C patterns)
• Only MOSI direction fails (ESP32 can't read from STM32)
• Problem is not timing-related (tried many variations)
• Problem is not clock speed-related (tested 468kHz ~ 15MHz)

Thank you!