Requirements Authority Module

Overview

Requirements-driven approach to declaring the expected RBAC test results referenced by Patrole. These requirements express the intention behind the policy. A high-level YAML syntax is used to concisely and clearly map each policy action to the list of associated roles.

Note

The Custom Requirements File is required to use this validation approach and, currently, must be manually generated.

This validation approach can be toggled on by setting the [patrole].test_custom_requirements configuration option to True; see Patrole Configuration Guide for more information.

When to use

This RbacAuthority class can be used to achieve a requirements-driven approach to validating an OpenStack cloud’s RBAC implementation. Using this approach, Patrole computes expected test results by performing lookups against a Custom Requirements File which precisely defines the cloud’s RBAC requirements.

This validation approach should be used when:

  • The cloud has heavily customized policy files that require careful validation against one’s requirements.

    Heavily customized policy files can contain relatively nuanced/technical syntax that impinges upon the goal of using a clear and concise syntax present in the Custom Requirements File to drive RBAC validation.

  • The cloud has non-OpenStack services that require RBAC validation but which don’t leverage the oslo.policy framework.

    Services like Contrail that are present in an OpenStack-based cloud that interface with OpenStack services like Neutron also require RBAC validation. The requirements-driven approach to RBAC validation is framework-agnostic and so can work with any policy engine.

  • Expected results are captured as clear-cut, unambiguous requirements.

    Validating a cloud’s RBAC against high-level, clear-cut requirements is a valid use case. Relying on oslo.policy validating customized policy files is not sufficient to satisfy this use case.

As mentioned above, the trade-off with this approach is having to manually generate the Custom Requirements File. There is currently no tooling to automatically do this.

Custom Requirements File

File path of the YAML file that defines your RBAC requirements. This file must be located on the same host that Patrole runs on. The YAML file should be written as follows:

<service_foo>:
  <logical_or_example>:
    - <allowed_role_1>
    - <allowed_role_2>
  <logical_and_example>:
    - <allowed_role_3>, <allowed_role_4>
<service_bar>:
  <logical_not_example>:
    - <!disallowed_role_5>

Where:

  • service - the service that is being tested (Cinder, Nova, etc.).

  • api_action - the policy action that is being tested. Examples:

    • volume:create

    • os_compute_api:servers:start

    • add_image

  • allowed_role - the oslo.policy role that is allowed to perform the API.

Each item under logical_or_example is “logical OR”-ed together. Each role in the comma-separated string under logical_and_example is “logical AND”-ed together. And each item prefixed with “!” under logical_not_example is “logical negated”.

Note

The custom requirements file only allows policy actions to be mapped to the associated roles that define it. Complex oslo.policy constructs like literals or GenericChecks are not supported. For more information, reference the oslo.policy documentation.

Examples

Items within api_action are considered as logical or, so you may read:

<service_foo>:
  # "api_action_a: allowed_role_1 or allowed_role_2 or allowed_role_3"
  <api_action_a>:
    - <allowed_role_1>
    - <allowed_role_2>
    - <allowed_role_3>

as <allowed_role_1> or <allowed_role_2> or <allowed_role_3>.

Roles within comma-separated items are considered as logic and, so you may read:

<service_foo>:
  # "api_action_a: (allowed_role_1 and allowed_role_2) or allowed_role_3"
  <api_action_a>:
    - <allowed_role_1>, <allowed_role_2>
    - <allowed_role_3>

as <allowed_role_1> and <allowed_role_2> or <allowed_role_3>.

Also negative roles may be defined with an exclamation mark ahead of role:

<service_foo>:
  # "api_action_a: (allowed_role_1 and allowed_role_2 and not
  # disallowed_role_4) or allowed_role_3"
  <api_action_a>:
    - <allowed_role_1>, <allowed_role_2>, !<disallowed_role_4>
    - <allowed_role_3>

This example must be read as <allowed_role_1> and <allowed_role_2> and not <disallowed_role_4> or <allowed_role_3>.