fix: RBAC bypass vulnerabilities in model access (backport #4270) (#4285)

Closes security gaps where RBAC checks could be bypassed:

o Inference router: Added RBAC enforcement in the fallback
  path to ensure access control is applied consistently.

o Model listing: Dynamic models fetched via provider_data were returned
  without RBAC checks. Added filtering to ensure users only see models
  they have permission to access.

Both fixes create temporary ModelWithOwner objects for RBAC validation,
maintaining security through consistent access control enforcement.

Closes: #4269
<hr>This is an automatic backport of pull request #4270 done by
[Mergify](https://mergify.com).

Signed-off-by: Derek Higgins <derekh@redhat.com>
Signed-off-by: Charlie Doern <cdoern@redhat.com>
Co-authored-by: Derek Higgins <derekh@redhat.com>

llama_stack/core/routers/inference.py

@@ -49,10 +49,17 @@ from llama_stack.apis.inference import (
 )
 from llama_stack.apis.models import Model, ModelType
 from llama_stack.apis.telemetry import MetricEvent, MetricInResponse, Telemetry
+from llama_stack.core.access_control.access_control import is_action_allowed
+from llama_stack.core.datatypes import ModelWithOwner
+from llama_stack.core.request_headers import get_authenticated_user
 from llama_stack.log import get_logger
 from llama_stack.models.llama.llama3.chat_format import ChatFormat
 from llama_stack.models.llama.llama3.tokenizer import Tokenizer
-from llama_stack.providers.datatypes import HealthResponse, HealthStatus, RoutingTable
+from llama_stack.providers.datatypes import (
+    HealthResponse,
+    HealthStatus,
+    RoutingTable,
+)
 from llama_stack.providers.utils.inference.inference_store import InferenceStore
 from llama_stack.providers.utils.telemetry.tracing import enqueue_event, get_current_span
 
@@ -186,15 +193,41 @@ class InferenceRouter(Inference):
             provider = await self.routing_table.get_provider_impl(model.identifier)
             return provider, model.provider_resource_id
 
+        # Handles cases where clients use the provider format directly
+        return await self._get_provider_by_fallback(model_id, expected_model_type)
+
+    async def _get_provider_by_fallback(self, model_id: str, expected_model_type: str) -> tuple[Inference, str]:
+        """
+        Handle fallback case where model_id is in provider_id/provider_resource_id format.
+        """
         splits = model_id.split("/", maxsplit=1)
         if len(splits) != 2:
             raise ModelNotFoundError(model_id)
 
         provider_id, provider_resource_id = splits
+
+        # Check if provider exists
         if provider_id not in self.routing_table.impls_by_provider_id:
             logger.warning(f"Provider {provider_id} not found for model {model_id}")
             raise ModelNotFoundError(model_id)
 
+        # Create a temporary model object for RBAC check
+        temp_model = ModelWithOwner(
+            identifier=model_id,
+            provider_id=provider_id,
+            provider_resource_id=provider_resource_id,
+            model_type=expected_model_type,
+            metadata={},  # Empty metadata for temporary object
+        )
+
+        # Perform RBAC check
+        user = get_authenticated_user()
+        if not is_action_allowed(self.routing_table.policy, "read", temp_model, user):
+            logger.debug(
+                f"Access denied to model '{model_id}' via fallback path for user {user.principal if user else 'anonymous'}"
+            )
+            raise ModelNotFoundError(model_id)
+
         return self.routing_table.impls_by_provider_id[provider_id], provider_resource_id
 
     async def openai_completion(