Method authorizeTask

src/slave/slave.cpp:9694–9722

As a principle, we do not need to re-authorize actions that have already been authorized by the master. However, we re-authorize the RUN_TASK action on the agent even though the master has already authorized it because: a) in cases where hosts have heterogeneous user-account configurations, it makes sense to set the ACL on the agent instead of on the master b) compared to other actions such as kil

//    framework, it's a greater security risk if malicious tasks are launched
//    as a superuser on the agent.
Future<bool> Slave::authorizeTask(
    const TaskInfo& task,
    const FrameworkInfo& frameworkInfo)
{
  if (authorizer.isNone()) {
    return true;
  }

  // Authorize the task.
  authorization::Request request;

  if (frameworkInfo.has_principal()) {
    request.mutable_subject()->set_value(frameworkInfo.principal());
  }

  request.set_action(authorization::RUN_TASK);

  authorization::Object* object = request.mutable_object();

  object->mutable_task_info()->CopyFrom(task);
  object->mutable_framework_info()->CopyFrom(frameworkInfo);

  LOG(INFO)
    << "Authorizing framework principal '"
    << (frameworkInfo.has_principal() ? frameworkInfo.principal() : "ANY")
    << "' to launch task " << task.task_id();

  return authorizer.get()->authorized(request);
}


Future<bool> Slave::authorizeSandboxAccess(