ESP32-S3 Hangs on boot when UART is printing

[HamzaHajeir]

Board

Espressif ESP32-S3-DevKitC-1-N8 (8 MB QD)

Device Description

Official board by Espressif
Programming and powering by the UART port.

Hardware Configuration

Unconnected to any external peripherals.

Version

v2.0.9

IDE Name

PlatformIO

Operating System

Windows 11

Flash frequency

40MHz (Default one)

PSRAM enabled

yes

Upload speed

115200

Description

When data rushes to UART at application startup the MCU hangs.

This is observed in printing to UART within global constructors, wherein the Serial began in an earlier one.

This issue wasn't observed with normal ESP32.

The amount of data being successfully printed depends on the introduced delay (delayMs in the example sketch).

With no delay it printed ~155 Bytes, and with a delay of uS it printed ~182 Bytes, finally run with not problems when delay was 100uS.

Sketch

#include <Arduino.h>

int delayMs = 10;
class dummy {
  public:
  dummy () {
    Serial.begin(115200);
    for(int i=0;i<200;i++) {
      Serial.printf("0", i);
      delayMicroseconds(delayMs);
    }
  }
};
dummy d;
void setup() {
  // put your setup code here, to run once:
  pinMode(LED_BUILTIN, OUTPUT);
  Serial.begin(115200);
  Serial.printf("Hello World!\n");
}

void loop() {
  // put your main code here, to run repeatedly:
  digitalWrite(LED_BUILTIN, HIGH);
  delay(1000);
  digitalWrite(LED_BUILTIN, LOW);
  delay(1000);
}

Debug Message

rst:0x1 (POWERON),boot:0x8 (SPI_FAST_FLASH_BOOT)
SPIWP:0xee
mode:DIO, clock div:1
load:0x3fce3808,len:0x44c
load:0x403c9700,len:0xbe4
load:0x403cc700,len:0x2a68
entry 0x403c98d4
00000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000000ESP-ROM:esp32s3-20210327
Build:Mar 27 2021
rst:0x10 (RTCWDT_RTC_RST),boot:0x8 (SPI_FAST_FLASH_BOOT)
SPIWP:0xee
mode:DIO, clock div:1
load:0x3fce3808,len:0x44c
load:0x403c9700,len:0xbe4
load:0x403cc700,len:0x2a68
entry 0x403c98d4

Other Steps to Reproduce

No response

I have checked existing issues, online documentation and the Troubleshooting Guide

• I confirm I have checked existing issues, online documentation and Troubleshooting guide.

[brightproject]

Now I did one "research" about resetting the microcontroller, and found the reason, this is the installed core 0 in the arduino IDE.
Tell me, are you flashing the microcontroller through an arduino IDE?
What do you have selected in the board settings?
Additional link.

[HamzaHajeir]

I'm not sure what that issue brings to us, I use PlatformIO. The software does work ok after setup() is succeeded to be called.

…

On Mon, Jul 3, 2023, 19:04 Vladislav ***@***.***> wrote: Now I did one "research <#8374 (comment)>" about resetting the microcontroller, and found the reason, this is the installed core 0 in the arduino IDE. Tell me, are you flashing the microcontroller through an arduino IDE? What do you have selected in the board settings? — Reply to this email directly, view it on GitHub <#8377 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AH3O7J3CTLVNCBFKZAFQIPTXOLUQZANCNFSM6AAAAAAZ3PYC6Q> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[brightproject]

The software does work ok after setup() is succeeded to be called.

Show me your compilation settings.
Are you specifying which core 0 or core 1 to use?

[HamzaHajeir]

I don't override any such configuration. And other than the exact situation described OP, everything works as expected.

…

On Mon, Jul 3, 2023, 20:05 Vladislav ***@***.***> wrote: The software does work ok after setup() is succeeded to be called. Show me your compilation settings. Are you specifying which core 0 or core 1 to use? — Reply to this email directly, view it on GitHub <#8377 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AH3O7J5LJRP37UBJBAGLOR3XOL3VXANCNFSM6AAAAAAZ3PYC6Q> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[rhvarrier]

I'm having the same issue. I'm using the following compilation settings

[SuGlider]

Issue confirmed. Currently under analysis.

[SuGlider]

After investigating this issue, I found out a few things.
As conclusion, I see that it is due to the way how IDF UART API, used in Arduino, works.
The IDF UART driver freezes, apparently, depending of the "timing" of its execution.
It can't be fixed at this time.

A few facts:
1- The 1st stage boot loader will set the Watchdog Timer to reset the ESP32-S3 after about 8 seconds.
This WDT resets the S3 with the message "rst:0x10 (RTCWDT_RTC_RST)"

2- IDF UART driver needs FreeRTOS resources to be set in order to run correctly.

3- C++ constructors are executed before the main() function is executed. This creates a problem because the IDF Driver can't run well and the program ends up with potential "inconsistencies".

I have tested the same sketch running direct IDF UART writing functions, and it fails in the same way:

#include "driver/uart.h"

class dummy {
  public:
  dummy () {
    Serial.begin(115200);
    for(int i=0;i<200;i++) {
      char c = '0';
      uart_write_bytes(0, &c, 1);
      delayMicroseconds(delayMs);
    }
  }
};

As a workaround, I used printf() instead of Serial.printf() which uses some ROM functions instead of IDF UART drivers.
So, in case it is necessary to print something in a C++ Constructor, this is the way to do it.

class dummy {
  public:
  dummy () {
    for(int i=0;i<200;i++) {
      printf("%d", i%10);
      delayMicroseconds(delayMs);
    }
    printf("\nConstructor finished\n");
  }
};