การแก้ไขข้อบกพร่องย้อนหลังด้วยดัมป์แกนหน่วยความจำ

Welcome! In this lesson, we'll explore post-mortem debugging. This powerful technique lets you investigate software failures after they've occurred, without needing to reproduce the issue live.

It's incredibly useful when you can't attach a debugger directly to a crashing application, especially in production environments.

The cornerstone of post-mortem debugging is the core dump. Think of it as a snapshot of a program's entire memory space and CPU state at the exact moment it crashed.

• It's a file generated by the operating system.
• It contains critical information about the program's execution.
• It helps you understand why a crash happened.

Core dumps are typically generated when a program encounters a severe, unhandled error that causes it to terminate unexpectedly. Common scenarios include:

• Segmentation Faults (Segfaults): Accessing invalid memory.
• Unhandled Exceptions: Language-specific errors not caught by the program.
• Assertion Failures: When a program's internal assumptions are violated.
• Program Crashes: Any abrupt, abnormal termination.

On Linux systems, core dump generation might be disabled by default or limited in size. You can enable it:

• Temporarily: Use ulimit -c unlimited in your shell session.
• System-wide: Modify /etc/sysctl.conf (e.g., kernel.core_pattern to specify output path and filename format).

Without proper configuration, your system might not save core dumps when crashes occur.

A core dump is packed with forensic data. It typically includes:

• Memory Image: A copy of the program's entire virtual memory.
• CPU Register Values: The state of the CPU registers at the crash time.
• Stack Trace: The sequence of function calls leading up to the crash.
• Process Information: Process ID, signal that caused the crash, executable path.
• Loaded Libraries: Information about shared libraries linked to the program.

To make sense of a core dump, you need specialized tools. Some popular ones include:

• GDB (GNU Debugger): Widely used for C/C++ programs on Linux/Unix.
• WinDbg: Microsoft's powerful debugger for Windows applications.
• jstack/jmap: For Java applications, these tools can extract thread dumps and memory maps that act as a form of 'core dump'.
• Delve: A debugger for Go programs.

We'll focus on GDB as a common example.

Let's look at a simple C program that will intentionally cause a segmentation fault. This will generate a core dump file if your system is configured to do so.

Try compiling and running this code:

After the program crashes and creates a core dump (e.g., core or core.PID), you can load it into GDB. Assuming your executable is a.out:

gdb ./a.out core
bt

The bt (backtrace) command is essential. It shows the call stack leading to the crash, helping you pinpoint the exact function and line number where the error occurred.

Once you have the backtrace, you can navigate the stack frames (e.g., using frame N where N is the frame number). Then, you can inspect variable values at that point in time:

• print variable_name: Shows the value of a specific variable.
• info locals: Lists all local variables in the current stack frame and their values.

This helps you understand the state of the program's data when it crashed.

Great work! You've learned about the power of post-mortem debugging using core dumps.

• Core dumps are memory snapshots of crashed programs.
• They contain vital info like stack traces and variable states.
• Tools like GDB help analyze them without live access.

This technique is indispensable for debugging hard-to-reproduce or production-only issues, enabling you to fix problems even after the event.