seaweedfs/weed/s3api/policy_engine/README_POLICY_ENGINE.md

SeaweedFS Policy Evaluation Engine

This document describes the comprehensive policy evaluation engine that has been added to SeaweedFS, providing AWS S3-compatible policy support while maintaining full backward compatibility with existing identities.json configuration.

Overview

The policy engine provides:

• Full AWS S3 policy compatibility - JSON policies with conditions, wildcards, and complex logic
• Backward compatibility - Existing identities.json continues to work unchanged
• Bucket policies - Per-bucket access control policies
• IAM policies - User and group-level policies
• Condition evaluation - IP restrictions, time-based access, SSL-only, etc.
• AWS-compliant evaluation order - Explicit Deny > Explicit Allow > Default Deny

Architecture

Files Created

1. policy_engine/types.go - Core policy data structures and validation
2. policy_engine/conditions.go - Condition evaluators (StringEquals, IpAddress, etc.)
3. policy_engine/engine.go - Main policy evaluation engine
4. policy_engine/integration.go - Integration with existing IAM system
5. policy_engine/engine_test.go - Comprehensive tests
6. policy_engine/examples.go - Usage examples and documentation (excluded from builds)
7. policy_engine/wildcard_matcher.go - Optimized wildcard pattern matching
8. policy_engine/wildcard_matcher_test.go - Wildcard matching tests

Key Components

PolicyEngine
├── Bucket Policies (per-bucket JSON policies)
├── User Policies (converted from identities.json + new IAM policies)
├── Condition Evaluators (IP, time, string, numeric, etc.)
└── Evaluation Logic (AWS-compliant precedence)

Backward Compatibility

Existing identities.json (No Changes Required)

Your existing configuration continues to work exactly as before:

{
  "identities": [
    {
      "name": "readonly_user",
      "credentials": [{"accessKey": "key123", "secretKey": "secret123"}],
      "actions": ["Read:public-bucket/*", "List:public-bucket"]
    }
  ]
}

Legacy actions are automatically converted to AWS-style policies:

• Read:bucket/* → s3:GetObject on arn:aws:s3:::bucket/*
• Write:bucket → s3:PutObject, s3:DeleteObject on arn:aws:s3:::bucket/*
• Admin → s3:* on arn:aws:s3:::*

New Capabilities

1. Bucket Policies

Set bucket-level policies using standard S3 API:

# Set bucket policy
curl -X PUT "http://localhost:8333/bucket?policy" \
     -H "Authorization: AWS access_key:signature" \
     -d '{
       "Version": "2012-10-17",
       "Statement": [
         {
           "Effect": "Allow",
           "Principal": "*",
           "Action": "s3:GetObject",
           "Resource": "arn:aws:s3:::bucket/*"
         }
       ]
     }'

# Get bucket policy
curl "http://localhost:8333/bucket?policy"

# Delete bucket policy
curl -X DELETE "http://localhost:8333/bucket?policy"

2. Advanced Conditions

Support for all AWS condition operators:

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Principal": "*",
      "Action": "s3:GetObject",
      "Resource": "arn:aws:s3:::secure-bucket/*",
      "Condition": {
        "IpAddress": {
          "aws:SourceIp": ["192.168.1.0/24", "10.0.0.0/8"]
        },
        "Bool": {
          "aws:SecureTransport": "true"
        },
        "DateGreaterThan": {
          "aws:CurrentTime": "2023-01-01T00:00:00Z"
        }
      }
    }
  ]
}

3. Supported Condition Operators

• String: StringEquals, StringNotEquals, StringLike, StringNotLike
• Numeric: NumericEquals, NumericLessThan, NumericGreaterThan, etc.
• Date: DateEquals, DateLessThan, DateGreaterThan, etc.
• IP: IpAddress, NotIpAddress (supports CIDR notation)
• Boolean: Bool
• ARN: ArnEquals, ArnLike
• Null: Null

4. Condition Keys

Standard AWS condition keys are supported:

• aws:CurrentTime - Current request time
• aws:SourceIp - Client IP address
• aws:SecureTransport - Whether HTTPS is used
• aws:UserAgent - Client user agent
• s3:x-amz-acl - Requested ACL
• s3:VersionId - Object version ID
• And many more...

Policy Evaluation

Evaluation Order (AWS-Compatible)

1. Explicit Deny - If any policy explicitly denies access → DENY
2. Explicit Allow - If any policy explicitly allows access → ALLOW
3. Default Deny - If no policy matches → DENY

Policy Sources (Evaluated Together)

1. Bucket Policies - Stored per-bucket, highest priority
2. User Policies - Converted from identities.json + new IAM policies
3. Legacy IAM - For backward compatibility (lowest priority)

Examples

Public Read Bucket

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Sid": "PublicRead",
      "Effect": "Allow",
      "Principal": "*",
      "Action": "s3:GetObject",
      "Resource": "arn:aws:s3:::public-bucket/*"
    }
  ]
}

IP-Restricted Bucket

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Allow",
      "Principal": "*",
      "Action": ["s3:GetObject", "s3:PutObject"],
      "Resource": "arn:aws:s3:::secure-bucket/*",
      "Condition": {
        "IpAddress": {
          "aws:SourceIp": "192.168.1.0/24"
        }
      }
    }
  ]
}

SSL-Only Access

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Effect": "Deny",
      "Principal": "*",
      "Action": "s3:*",
      "Resource": ["arn:aws:s3:::ssl-bucket/*", "arn:aws:s3:::ssl-bucket"],
      "Condition": {
        "Bool": {
          "aws:SecureTransport": "false"
        }
      }
    }
  ]
}

Integration

For Existing SeaweedFS Users

1. No changes required - Your existing setup continues to work
2. Optional enhancement - Add bucket policies for fine-grained control
3. Gradual migration - Move to full AWS policies over time

For New Users

1. Start with either identities.json or AWS-style policies
2. Use bucket policies for complex access patterns
3. Full feature parity with AWS S3 policies

Testing

Run the policy engine tests:

# Core policy tests
go test -v -run TestPolicyEngine

# Condition evaluator tests  
go test -v -run TestConditionEvaluators

# Legacy compatibility tests
go test -v -run TestConvertIdentityToPolicy

# Validation tests
go test -v -run TestPolicyValidation

Performance

• Compiled patterns - Regex patterns are pre-compiled for fast matching
• Cached policies - Policies are cached in memory with TTL
• Early termination - Evaluation stops on first explicit deny
• Minimal overhead - Backward compatibility with minimal performance impact

Migration Path

Phase 1: Backward Compatible (Current)

• Keep existing identities.json unchanged
• Add bucket policies as needed
• Legacy actions automatically converted to AWS policies

Phase 2: Enhanced (Optional)

• Add advanced conditions to policies
• Use full AWS S3 policy features
• Maintain backward compatibility

Phase 3: Full Migration (Future)

• Migrate to pure IAM policies
• Use AWS CLI/SDK for policy management
• Complete AWS S3 feature parity

Compatibility

• ✅ Full backward compatibility with existing identities.json
• ✅ AWS S3 API compatibility for bucket policies
• ✅ Standard condition operators and keys
• ✅ Proper evaluation precedence (Deny > Allow > Default Deny)
• ✅ Performance optimized with caching and compiled patterns

The policy engine provides a seamless upgrade path from SeaweedFS's existing simple IAM system to full AWS S3-compatible policies, giving you the best of both worlds: simplicity for basic use cases and power for complex enterprise scenarios.