Linux Coredump

Korean README

• Example project for generating, analyzing, and managing core dumps for Linux C/C++ (gcc) programs.

---

File Overview

• main.cpp

  • A simple C++ example that intentionally causes a segmentation fault.

    int main() {
      std::string *ptr = NULL;
      ptr->clear(); // crash here
      return 0;
    }

  • Debug build example:

    g++ -g -O0 -Wall -Wextra -o hello main.cpp

  • Using AddressSanitizer / UBSanitizer:

    g++ -g -O0 -Wall -Wextra -fsanitize=address,undefined -fno-omit-frame-pointer -o hello main.cpp

  • AddressSanitizer (ASan): Detects memory errors at runtime

    • Detectable issues:
      • Heap/stack buffer overflow
      • Use-after-free
      • Double free
      • Memory leaks
      • Stack overflow
    • Characteristics:
      • Monitors memory access at runtime
      • Provides precise stack traces
      • Has performance overhead (~2–3x)

  • UndefinedBehaviorSanitizer (UBSan): Detects undefined behavior in C++

    • Detectable issues:
      • Integer overflow
      • Invalid casts
      • Null pointer dereference
      • Out-of-bounds access
      • Misaligned memory access
      • Invalid enum values
    • Characteristics:
      • Effective for detecting logic errors
      • Lighter than ASan
      • Can warn before crash

---

• setup_core_dump_systemwide.sh
  • Configures the system to generate core dump files globally (requires sudo).

  sudo ./setup_core_dump_systemwide.sh

  • Core file name pattern: core.<exe>.<pid>.<time>
    • <exe> : executable name
    • <pid> : process ID
    • <time> : epoch time
      • Convert with:

        date -d @<time> '+%Y-%m-%d %H:%M:%S %Z'

---

• run_hello_with_core.sh
  • Executes the hello program and generates a core dump.
  • Requires system-wide core dump configuration.
  • Set executable path before use:

  WORKDIR="/home/jaytwo/workspace/coredump-workspace"
  EXEC="${WORKDIR}/hello"

---

• run_hello_with_core_daemon.sh
  • Runs the hello program repeatedly in daemon mode.
    • Automatically restarts on exit
    • Enables core dumps and ASAN/UBSAN
    • Log file is rotated when exceeding 10MB (keeps last 10,000 lines)

    MAX_LOG_SIZE=10485760 # 10MB
    MAX_LOG_LINES=10000
    if [ -f "${LOG}" ] && [ $(stat -c%s "${LOG}") -ge $MAX_LOG_SIZE ]; then
        tail -n $MAX_LOG_LINES "${LOG}" > "${LOG}.tmp" && mv "${LOG}.tmp" "${LOG}"
        echo "[$(date '+%F %T')] log trimmed to last $MAX_LOG_LINES lines" >> "${LOG}"
    fi

  • Configure executable path:

    WORKDIR="/home/jaytwo/workspace/coredump-workspace"
    EXEC="${WORKDIR}/hello"

---

• gdb_hello_core.sh
  • Script to analyze core dump using gdb.

  ./gdb_hello_core.sh <core_dump_file>

  • Set executable path:

  WORKDIR="/home/jaytwo/workspace/coredump-workspace"
  EXEC="${WORKDIR}/hello"

---

• list_core_with_time.sh
  • Lists core files in the current directory with human-readable timestamps.

---

Usage Example

1. Configure core dump system

   sudo ./setup_core_dump_systemwide.sh

2. Build executable with debug symbols

   g++ -g -O0 -Wall -Wextra -o hello main.cpp

3. Generate core dump

   ./run_hello_with_core.sh

4. List core dump files

   ./list_core_with_time.sh

5. Analyze core dump

   ./gdb_hello_core.sh core.hello.<pid>.<time>

   • gdb commands:
     • r : run
     • bt full : full backtrace with local variables
     • list : show source lines

6. Run as daemon

   ./run_hello_with_core_daemon.sh

---

Automatic Startup Setup

• Installation guide: INSTALL.md