systemd-units（系统单元）

Systemd Units

Overview

This skill provides guidance for creating and hardening systemd service unit files following modern Linux service management best practices. It covers proper service type selection, dependency management, security sandboxing, and filesystem isolation to create reliable, secure services.

When to Use This Skill

Use this skill when:

• Creating new systemd service units for web applications, APIs, background workers, or daemons
• Hardening existing service files with security and sandboxing options
• Converting traditional init scripts or manual process management to systemd
• Troubleshooting service startup, restart, or permission issues
• Implementing proper service dependencies and ordering

Workflow Decision Tree

User request involves systemd service?
│
├─ Creating a new service?
│  │
│  ├─ Web application/API → Use "Creating a New Service" workflow
│  │                        → Start from assets/basic-webapp.service or assets/hardened-webapp.service
│  │
│  ├─ Background worker/daemon → Use "Creating a New Service" workflow
│  │                            → Start from assets/background-worker.service
│  │
│  └─ One-time initialization → Use "Creating a New Service" workflow
│                              → Start from assets/oneshot-init.service
│
└─ Hardening existing service?
   │
   └─ Use "Hardening an Existing Service" workflow
      → Reference references/systemd_options.md for security options
      → Compare against assets/hardened-webapp.service for patterns

Creating a New Service

When creating a new systemd service, follow these steps:

1. Select the Appropriate Template

Start with the most relevant template from assets/:

• basic-webapp.service: Simple web application without heavy sandboxing (development/internal use)
• hardened-webapp.service: Production web application with full security hardening
• background-worker.service: Queue processor, scheduled task, or background daemon
• oneshot-init.service: One-time initialization script or setup task

Copy the template and customize the following sections in order:

2. Configure [Unit] Section

Update the service metadata and dependencies:

[Unit]
Description=Clear description of what this service does
Documentation=https://example.com/docs or man:program(8)
After=network-online.target database.service
Wants=network-online.target
Requires=database.service

Key decisions:

• After=: List services this must start after (ordering dependency)
• Wants=: Soft dependencies (service will start even if these fail)
• Requires=: Hard dependencies (service fails if these fail)
• For web services, always include After=network-online.target and Wants=network-online.target

3. Configure Service Type and Execution

Select the appropriate service type based on the application's capabilities:

Recommended service types (in order of preference):

1. Type=notify: Application supports sd_notify protocol (best option for reliability)
2. Type=exec: Standard long-running process (good default)
3. Type=oneshot: One-time execution that exits (initialization scripts)

Avoid Type=simple (poor error detection) and avoid Type=forking (deprecated pattern).

[Service]
Type=exec
ExecStart=/usr/bin/node /opt/webapp/server.js
WorkingDirectory=/opt/webapp

Always use absolute paths for ExecStart= and related commands.

4. Configure User and Environment

Set the execution user and environment variables:

# Option 1: Dynamic user (recommended for security)
DynamicUser=yes

# Option 2: Specific user/group
User=webapp
Group=webapp

# Environment configuration
Environment="NODE_ENV=production"
EnvironmentFile=-/etc/webapp/webapp.env

Best practice: Use DynamicUser=yes for services that don't need a specific user. For secrets, use EnvironmentFile= with restricted permissions instead of embedding credentials in the service file.

5. Configure Restart Policy

Set appropriate restart behavior:

Restart=on-failure    # For web apps (restart on crashes)
RestartSec=10        # Wait 10 seconds before restart
TimeoutStartSec=30   # Fail if startup takes >30 seconds
TimeoutStopSec=30    # Force kill if graceful stop takes >30 seconds

Common patterns:

• Web applications: Restart=on-failure with RestartSec=10
• Critical workers: Restart=always with RestartSec=5
• Oneshot tasks: Omit Restart= (default is no restart)

6. Apply Security Hardening

For production services, apply progressive security hardening:

Start with these baseline options:

# Filesystem protection
ProtectSystem=strict          # Entire filesystem read-only except specified paths
ProtectHome=yes              # Hide /home directories
PrivateTmp=yes               # Private /tmp namespace
ReadWritePaths=/var/lib/myapp /var/log/myapp  # Writable paths whitelist
NoNewPrivileges=yes          # Prevent privilege escalation

# Device and kernel protection
PrivateDevices=yes           # Restrict device access
ProtectKernelTunables=yes    # Prevent /proc/sys, /sys writes
ProtectKernelModules=yes     # Prevent kernel module loading
ProtectControlGroups=yes     # Protect cgroup filesystem

For internet-facing services, add:

# Network restrictions
RestrictAddressFamilies=AF_INET AF_INET6  # Only IPv4/IPv6

# Capability restrictions
CapabilityBoundingSet=       # Remove all capabilities

# System call filtering
SystemCallFilter=@system-service
SystemCallArchitectures=native

# Memory protection
MemoryDenyWriteExecute=yes   # W^X protection
LockPersonality=yes

Testing strategy: Start with maximum restrictions. If the service fails, use journalctl -xeu <service> to identify which restriction caused the failure, then selectively relax only that restriction.

Use systemd-analyze security <service> to verify the security posture after configuration.

7. Configure [Install] Section

Set the target for service enablement:

[Install]
WantedBy=multi-user.target   # Most common target (non-graphical multi-user system)

Common targets:

• multi-user.target: Standard for most services
• graphical.target: Services requiring graphical environment
• default.target: Alias for the default system target

8. Install and Test the Service

Place the service file and test:

# Copy to system directory
sudo cp myapp.service /etc/systemd/system/

# Reload systemd to recognize new service
sudo systemctl daemon-reload

# Start and check status
sudo systemctl start myapp
sudo systemctl status myapp

# View logs
journalctl -xeu myapp

# Enable to start on boot
sudo systemctl enable myapp

Hardening an Existing Service

When hardening an existing service file, follow these steps:

1. Analyze Current Configuration

Read the existing service file and identify security gaps:

# Check current security exposure
systemd-analyze security <service-name>

# Review current configuration
systemctl cat <service-name>

Look for:

• Missing sandboxing options (ProtectSystem, PrivateDevices, etc.)
• Overly permissive user (running as root unnecessarily)
• Missing filesystem restrictions
• No capability boundaries

2. Create Drop-in Override

Rather than modifying the vendor-supplied service file directly, create a drop-in override:

sudo systemctl edit <service-name>

This creates /etc/systemd/system/<service-name>.service.d/override.conf and preserves vendor updates.

3. Apply Security Options Progressively

Add security options in stages, testing after each stage:

Stage 1: Basic isolation

[Service]
# Filesystem protection
ProtectSystem=strict
ProtectHome=yes
PrivateTmp=yes
ReadWritePaths=/var/lib/myapp  # Add paths service needs to write

# Basic device protection
PrivateDevices=yes

Test: sudo systemctl restart <service> and check journalctl -xeu <service>

Stage 2: Kernel and capability restrictions

[Service]
# Kernel protection
ProtectKernelTunables=yes
ProtectKernelModules=yes
ProtectKernelLogs=yes
ProtectControlGroups=yes

# Remove capabilities
NoNewPrivileges=yes
CapabilityBoundingSet=

Test again.

Stage 3: Network and system call filtering

[Service]
# Network restrictions (adjust for service needs)
RestrictAddressFamilies=AF_INET AF_INET6 AF_UNIX

# System call filtering
SystemCallFilter=@system-service
SystemCallArchitectures=native

# Memory protection
MemoryDenyWriteExecute=yes
LockPersonality=yes

Test and verify functionality.

4. Handle Common Restriction Failures

If the service fails after adding restrictions:

Error: "Permission denied" accessing filesystem

• Add the path to ReadWritePaths= or relax ProtectSystem= to full instead of strict

Error: "Operation not permitted" for network operations

• Check RestrictAddressFamilies= includes needed protocols
• For Unix domain sockets, add AF_UNIX

Error: System call blocked

• Review journalctl for blocked syscall name
• Add exception: SystemCallFilter=@system-service <syscall-name>

Error: Cannot access devices

• Add specific device to DeviceAllow= instead of removing PrivateDevices=

5. Verify Hardening

After applying all restrictions:

# Check security score (lower is better)
systemd-analyze security <service-name>

# Verify service functionality
sudo systemctl restart <service-name>
sudo systemctl status <service-name>

# Test application-level functionality
# (Make requests, check logs, verify operations)

Best Practices

Service Type Selection

1. Prefer Type=notify for services that can implement sd_notify protocol - provides reliable startup verification
2. Use Type=exec for standard long-running processes - good error detection
3. Avoid Type=simple - provides no startup verification and poor error handling
4. Avoid Type=forking - deprecated pattern, use Type=notify or Type=exec instead

Dependency Management

1. Separate ordering from requirements: Use both After= and Requires=/Wants=

- After= controls startup sequence - Requires=/Wants= controls dependency relationships

1. Use Wants= for soft dependencies (tolerate failures)
2. Use Requires= for hard dependencies (fail if dependency fails)
3. For network services: Always include After=network-online.target and Wants=network-online.target

Security Hardening

1. Use DynamicUser=yes unless the service needs a specific user
2. Start with maximum restrictions and relax selectively - more effective than progressive hardening
3. Separate secrets from configuration - use EnvironmentFile= for sensitive values
4. Always set NoNewPrivileges=yes to prevent privilege escalation
5. Remove all capabilities by default with CapabilityBoundingSet=, add back only what's needed
6. Use systemd-analyze security to verify hardening effectiveness

Filesystem Access

1. Prefer ProtectSystem=strict with explicit ReadWritePaths= whitelist
2. Always enable PrivateTmp=yes unless sharing /tmp is explicitly required
3. Enable ProtectHome=yes unless home directory access is needed
4. Create dedicated directories under /var/lib/ for application data

Restart and Recovery

1. Web applications: Restart=on-failure with RestartSec=10
2. Critical workers: Restart=always with RestartSec=5
3. Set reasonable timeouts: TimeoutStartSec=30 and TimeoutStopSec=30 (adjust based on application)
4. Define custom success codes with SuccessExitStatus= if application uses non-zero exits for success

Logging and Debugging

1. Use structured logging: StandardOutput=journal and StandardError=journal
2. Set SyslogIdentifier= for easier log filtering
3. Debug with: journalctl -xeu <service> (shows extended info and follows)
4. Check dependencies: systemctl list-dependencies <service>

Resources

references/systemd_options.md

Comprehensive reference documentation for all systemd unit and service options. Read this file when:

• Looking up specific option syntax or behavior
• Understanding security/sandboxing options in detail
• Reviewing all available service types and their tradeoffs
• Finding additional hardening options beyond the templates

Use grep to search for specific options:

grep -i "ProtectSystem" references/systemd_options.md

assets/ Templates

Production-ready service file templates:

• basic-webapp.service: Starting point for web applications without heavy sandboxing
• hardened-webapp.service: Fully hardened web application with maximum security
• background-worker.service: Queue processor or background daemon with security hardening
• oneshot-init.service: One-time initialization script pattern

Use these as starting points and customize for specific application needs.

Common Troubleshooting

Service fails to start after hardening:

1. Check logs: journalctl -xeu <service>
2. Look for "Permission denied" or "Operation not permitted" errors
3. Identify the blocked operation (filesystem, syscall, network)
4. Selectively relax the relevant restriction
5. Retest and verify

Service starts but behaves incorrectly:

1. Verify WorkingDirectory= is correct
2. Check environment variables are loaded (EnvironmentFile=)
3. Verify filesystem paths are accessible (ReadWritePaths=)
4. Check user/group has appropriate permissions on files

Service restarts repeatedly:

1. Check logs for crash reason: journalctl -xeu <service>
2. Verify ExecStart= path is correct and executable
3. Check if application crashes due to missing dependencies
4. Consider increasing RestartSec= to prevent rapid restart loops
5. Check if Type= matches application behavior

Cannot access network after hardening:

1. Verify RestrictAddressFamilies= includes required protocols (IPv4: AF_INET, IPv6: AF_INET6, Unix sockets: AF_UNIX)
2. Check if firewall or network namespace is blocking access
3. For localhost-only services, ensure AF_INET is included