Esp32 Crash Review

ESP32 Crash Review

Analyze crash reports, Guru Meditation errors, and panic dumps to determine root cause.

When to Use

• After observing a Guru Meditation error.
• When a device is stuck in a crash-reboot loop.
• When analyzing crash dumps from field devices.

Steps

1. Identify the Exception Type

2. Extract Key Information

From the panic output, collect:

Guru Meditation Error: Core [0/1] panic'ed ([ExceptionType]).
Exception was unhandled.
Core [N] register dump:
PC      : 0x........  PS      : 0x........
A0      : 0x........  A1      : 0x........  ...
EXCVADDR: 0x........    ← The address that caused the fault

PC (Program Counter) → Where the crash happened. EXCVADDR (Exception Address) → What address was accessed illegally. A1 → Stack pointer at time of crash.

3. Decode the Backtrace

# Using addr2line
xtensa-esp32-elf-addr2line -pfiaC -e build/project.elf ADDR1 ADDR2 ADDR3

# Using idf.py
idf.py monitor  # Automatically decodes backtraces

# Using ESP-IDF monitor with ELF
python -m esp_idf_monitor --port /dev/ttyUSB0 --baud 115200 build/project.elf

4. Analyze Common Patterns

NULL Pointer Dereference

• EXCVADDR: 0x00000000 or small value (< 0x100)
• Cause: Using an uninitialized pointer or a handle that failed to initialize.
• Fix: Add NULL checks before dereferencing.

Stack Overflow

• Signs: IllegalInstruction or LoadProhibited with A1 near the end of stack space.
• Cause: Task stack too small, deep recursion, large local variables.
• Fix: Increase task stack size, reduce local buffer sizes, check recursion depth.

ISR Violation

• Signs: Crash inside a FreeRTOS API that is not *FromISR.
• Cause: Calling blocking functions from an interrupt handler.
• Fix: Use ISR-safe API variants only.

Memory Corruption

• Signs: IllegalInstruction at a random address, corrupted register values.
• Cause: Buffer overflow, use-after-free, heap corruption.
• Fix: Enable heap poisoning (CONFIG_HEAP_POISONING_COMPREHENSIVE), use AddressSanitizer.

Watchdog Reset (not Guru Meditation but related)

• Signs: TG0WDT_SYS_RESET or TG1WDT_SYS_RESET in reset reason.
• Cause: Task not yielding, blocking the idle task or system tasks.
• Fix: Add yields, reduce critical section length, feed watchdog.

5. Recommend Diagnostic Steps

Based on the analysis, suggest:

1. Immediate fix — if root cause is clear.
2. Diagnostic configuration — enable features that help narrow down:
   • CONFIG_HEAP_POISONING_COMPREHENSIVE=y
   • CONFIG_HEAP_TASK_TRACKING=y
   • CONFIG_FREERTOS_WATCHPOINT_END_OF_STACK=y
   • CONFIG_ESP_SYSTEM_PANIC_PRINT_REBOOT=y
   • CONFIG_COMPILER_STACK_CHECK=y
3. Reproduction steps — minimal steps to trigger the crash again.

6. Generate Crash Report

# Crash Analysis Report

## Exception
- **Type**: [ExceptionType]
- **Core**: [0/1]
- **PC**: [address] → [decoded function]
- **EXCVADDR**: [address]

## Backtrace
| # | Address | Function | File:Line |
|---|---------|----------|-----------|
| 0 | 0x... | function_name | file.c:123 |

## Root Cause
[Analysis of most likely cause]

## Classification
[NULL pointer / Stack overflow / ISR violation / Memory corruption / Other]

## Recommended Fix
[Specific code change or configuration]

## Prevention
[How to prevent this class of bug in the future]

Source: agodianel/esp32-claude-workbench — distributed by TomeVault.