fix(pod): css miss due to container cgroup path check errors, and unified resource release

[sisttri]

No description provided.

[huhong-web]

Code Review for PR #253: fix(pod): css miss due to container cgroup path check errors, and unified resource release

1. Copyright Check

• [PASS] All modified files have proper copyright headers

2. Golang Style Check

• [PASS] Code follows Go style conventions
• [PASS] Error handling is proper
• [PASS] Variable naming follows Go conventions

3. Abstract Interface Check

• [PASS] Uses existing BPF and context interfaces properly

4. Architecture Check

• [PASS] Follows existing patterns for resource management
• [PASS] Proper cleanup with cancel functions and reader closing

5. Concurrency Safety Check

• [PASS] Uses sync.Map for concurrent access (line 76)
• [PASS] Context cancellation pattern is correct
• [PASS] Global variables are properly protected

6. Code Quality Issues

Issue 1: Global mutable state pattern

File: internal/pod/container_css_default.go:76-80

var (
    cgroupCssMetaDataMap      sync.Map
    _cgroupCssBpfInternal     *bpf.BPF
    _cgroupCssBpfCancelFunc   context.CancelFunc
    _cgroupCssBpfReader       bpf.PerfEventReader
)

Global mutable state is a common pattern in this codebase, but consider encapsulating these in a struct for better testability and lifecycle management.

Issue 2: Error handling in WalkDir callback

File: internal/pod/container_css_default.go:197-202

if err != nil {
    if path != realRoot {
        return nil // ignore error for container destroy, but not for root path
    }
    return err
}

The error handling is reasonable, but consider adding a comment explaining why container destroy errors should be ignored (e.g., race condition with container deletion).

Issue 3: Regex matching without pre-compilation

File: internal/pod/container_css_default.go:211-214

if !kubeletContainerIDRegexp.MatchString(d.Name()) {
    return nil
}

If kubeletContainerIDRegexp is used frequently in WalkDir callbacks, ensure it's pre-compiled (which it appears to be based on the name). If not, consider using regexp.MustCompile at package level.

7. Resource Management

• [PASS] Proper cleanup with _cgroupCssBpfCancelFunc
• [PASS] Reader is properly closed in release function
• [PASS] BPF resource is properly released

8. Testing Considerations

• The PR modifies core cgroup handling logic
• Consider adding unit tests for:
  • cgroupRootNotify with various error conditions
  • kubeletContainerIDRegexp matching patterns
  • cgroupCssRelease cleanup sequence

Summary

This PR fixes a critical issue where container cgroup path checks could cause CSS misses. The implementation is solid with proper resource cleanup and error handling.

Recommendation: Approve with suggestions for documentation and testing improvements.