[BUG] ESP32C6 DevKitM spi issue

[FelipeMdeO]

Description / Steps to reproduce the issue

Hello All,

I am trying use spi in esp32c6 DevKit M but I am facing issues.
I tried use default spi example, it fails some time, please look image below:

So I used logic analyzer to try debug whats happen and spi app has strange behavior. I cannot get clocks and cannot see MISO/MOSI bits.
Also, I implemented a app by myself and I had same behavior, I cannot see clocks and data using logic analyzer and oscilloscope.
Follow my app example:

#include <nuttx/config.h>
#include <nuttx/spi/spi_transfer.h>
#include <fcntl.h>
#include <unistd.h>
#include <stdio.h>
#include <stdint.h>

#include "spitool.h"

#define SPI_BUS 2  // SPI bus number
#define SPI_DEVICE 0  // SPI device (chip select)
#define SPI_MODE 0  // SPI mode (0, 1, 2, 3)
#define SPI_FREQUENCY 1000000  // SPI frequency (1 MHz)
#define SPI_BITS 8  // Number of bits per SPI word
#define SPI_DELAY 0  // Delay between transfers (in microseconds)

void send_spi_data(int fd, uint8_t *txbuffer, int length)
{
    struct spi_sequence_s seq;
    struct spi_trans_s trans;

    // Configure the transfer sequence
    seq.dev = SPIDEV_ID(SPIDEVTYPE_USER, SPI_DEVICE);
    seq.mode = SPI_MODE;
    seq.nbits = SPI_BITS;
    seq.frequency = SPI_FREQUENCY;
    seq.ntrans = 1;
    seq.trans = &trans;

    // Configure the transfer
    trans.deselect = true;
    trans.delay = SPI_DELAY;
    trans.nwords = length;
    trans.txbuffer = txbuffer;
    trans.rxbuffer = NULL;  // We are not receiving data

    // Perform the SPI transfer
    if (spidev_transfer(fd, &seq) < 0) {
        perror("SPI transfer failed");
    } else {
        printf("SPI transfer succeeded\n");
    }
}

int main(void)
{
    int fd;
    uint8_t txbuffer[] = {0xDE, 0xAD, 0xBE, 0xEF};  // Known data buffer

    // Open the SPI device
    fd = spidev_open(SPI_BUS);
    if (fd < 0) {
        perror("Failed to open SPI device");
        return -1;
    }

    while (1) {
        // Send data via SPI
        send_spi_data(fd, txbuffer, sizeof(txbuffer));

        // Wait for 1 second
        sleep(1);
    }

    // Close the SPI device
    close(fd);
    return 0;
}

Below, follow logic analyzer and oscilloscope captures:

Please, anyone can check it in your side?

On which OS does this issue occur?

[Linux]

What is the version of your OS?

Ubuntu 23.1

NuttX Version

master

Issue Architecture

[risc-v]

Issue Area

[Applications], [Drivers]

Verification

• I have verified before submitting the report.

[tmedicci]

Hi @FelipeMdeO ,

This feature is constantly tested by our internal CI and we found no related issues regarding SPI master. Let's try to debug to find the root cause of the issue you reported.

First of all, please try the attached firmware. You can flash it with:

esptool.py -c esp32c6 -p /dev/ttyUSB0 -b 921600 write_flash -fs detect -fm dio 0x0 nuttx.merged.bin

build_test_spi_master_esp32c6.zip

You don't need to make any external connection (CONFIG_ESPRESSIF_SPI_TEST_MODE is enabled). Just run the same commands and you should get the expected result.

That being said, some questions:

• Which defconfig did you use?
• Which pins did you connect externally?

Finally, can you please enable the following configs:

CONFIG_SYSTEM_NXDIAG=y
CONFIG_SYSTEM_NXDIAG_ESPRESSIF=y

and, on NuttX, run nxdiag --all and share your results, please?

[FelipeMdeO]

Hello @tmedicci ,

Which defconfig did you use?
./tools/configure.sh esp32c6-devkitm:spi

Which pins did you connect externally?

Brown - CS PIN 19
Red - FSPIQ 2
Yellow - CLK 6
Orange - FSPID 7

[FelipeMdeO]

About nxdiag output:
Follow it:

nsh> nxdiag --all
Nxdiag Report:

NuttX RTOS info:
	Hostname: 
	Release: 12.6.0-RC1
	Build: 3c05da536a-dirty Aug 12 2024 15:01:44
	Arch: risc-v
	Defconfig: esp32c6-devkitm:spi-dirty

Host system OS:
	Ubuntu 24.04 LTS Linux felipe-moura 6.8.0-40-generic #40-Ubuntu SMP PREEMPT_DYNAMIC Fri Jul  5 10:34:03 UTC 2024 x86_64 x86_64

Espressif specific information:

Bootloader version:
	esp32: Bootloader image not found
	esp32s2: Bootloader image not found
	esp32s3: Bootloader image not found
	esp32c2: Bootloader image not found
	esp32c3: Bootloader image not found
	esp32c6: Bootloader image not found
	esp32h2: Bootloader image not found

Toolchain version:
	clang: Not found
	gcc: gcc (Ubuntu 13.2.0-23ubuntu4) 13.2.0
	xtensa-esp32-elf-gcc: xtensa-esp32-elf-gcc (crosstool-NG esp-12.2.0_20230208) 12.2.0
	xtensa-esp32s2-elf-gcc: Not found
	xtensa-esp32s3-elf-gcc: xtensa-esp32s3-elf-gcc (crosstool-NG esp-12.2.0_20230208) 12.2.0
	riscv32-esp-elf-gcc: Not found
	riscv64-unknown-elf-gcc: Not found

Esptool version: 4.8.dev4

HAL version: sync/release_v5.1.c-nuttx-20230814-65-g8e8e62cf6a

[FelipeMdeO]

Hello @tmedicci,

about test your firmware, It isn't work properly also, I cannot get spi signals properly, however Send and Receive are equal even it aren't connected one with other (MISO wasn't connected to MOSI), I think your version is running in loopback situation. Please look video attached:

Screencast.from.2024-08-12.15-30-07.mp4

[tmedicci]

nxdiag

Yes, it's a loopback. You will not be able to measure anything, the pins are connected internally and the result was as expected: same content sent is successfully read.

Hello @tmedicci ,

Which defconfig did you use? ./tools/configure.sh esp32c6-devkitm:spi

Which pins did you connect externally?

Brown - CS PIN 19 Red - FSPIQ 2 Yellow - CLK 6 Orange - FSPID 7

I didn't get: in this test aren't you connecting MOSI with MISO? (pins 2 and 7 connected together) to test the loopback? If not, what is the slave device?

[FelipeMdeO]

Hello @tmedicci,

my early tests I keep MISO ando MOSI "floating" only connected to my logic analyzer, I believe I should at least see the clocks, right?

[tmedicci]

Oh, I see... Can you confirm that the loopback test is working as expected? (connecting pins 2 and 7)

@eren-terzioglu , can you check about the clock line, please?

[FelipeMdeO]

Hello @tmedicci, @eren-terzioglu .

I am facing issue using external loopback too. I connected directly, without "protoboard" and keep issue.
Loop my setup:

Look for result:

[eren-terzioglu]

Hi,
By default, spi example sends one word at a time and uses bus 2 but we should input them just in case. Could you please test with spi exch -b 2 -x 1 AB. AB might change. I tested with [build_test_spi_master_esp32c6.zip](https://github.com/user-attachments/files/16582907/build_test_spi_master_esp32c6.zip) and seems it is working

[eren-terzioglu]

We reproduced the issue, investigating it. Thanks for reporting.

[FelipeMdeO]

Hello @tmedicci and @eren-terzioglu . First of all thank for help.

Just to organize my last posts:

1 - Difference between data sent and received by spi
I just realized that this problem only occurs when I change the CS pin from 10 to any other pin. I need to change this pin because pin 10 is not exposed on the esp32c6 devkit m header.

2 - I can't see spi clock signals using a logic analyzer.

[FelipeMdeO]

Hello @eren-terzioglu
The "issue 1" not happen when I check option "SPI software CS", maybe this is expected behavior.

[acassis]

@eren-terzioglu @tmedicci in fact the documentation is confusing: https://nuttx.apache.org/docs/latest/applications/system/spi/index.html it doesn't say we need to connect MOSI to MISO (it is obvious for us, but not for someone outside the project). Also the text is citing "I2C Build Configuration" that is not related to that matter. I'll submit a fix to the documentation.

[acassis]

Done: #12913

[FelipeMdeO]

Hello @xiaoxiang781216, this topic is about spi issue. It wasn't solved yet. I think this topic must be kept open.

[eren-terzioglu]

Hello @xiaoxiang781216, this topic is about spi issue. It wasn't solved yet. I think this topic must be kept open.

Created #12924

[xiaoxiang781216]

Let me reopen this issue and link to #12924