fix(net): allow SSO apps egress to keycloak backend (Cilium 1.20 CVE-2025-30162)

App SSO broke cluster-wide after the Cilium 1.16.5->1.20.1 upgrade: the CVE-2025-30162 fix (in
>=1.16.8/1.17.2) makes the shared Gateway's L7 proxy re-evaluate a pod's egress policy against
Keycloak's REAL backend cluster identity (enforce_policy_on_l7lb) when it hairpins to keycloak.home.
'toEntities: world' matches the Gateway VIP at L3/L4 but not the resolved keycloak identity at L7, so
OIDC discovery/token 403'd ('Access denied') for every app with a NetworkPolicy (gitea/nextcloud/
semaphore/headlamp/crafty...); Grafana survived only because monitoring has no NetworkPolicy.

Fix: explicit toEndpoints egress allow to the keycloak backend for SSO apps.
- catalog/render.py: emit it for auth.mode in {oidc,forward-auth}; re-rendered 8 app goldens.
- platform/headlamp + platform/crafty: hand-written policies get the same allow.
Split-horizon design unchanged. Verified live: headlamp-ns -> keycloak discovery 403 -> 200.

Co-Authored-By: Claude Opus 4.8 <[email protected]>

apps/wave-0/gitea.yaml

@@ -185,6 +185,10 @@ spec:
   egress:
   - toEntities:
     - world
+  - toEndpoints:
+    - matchLabels:
+        io.kubernetes.pod.namespace: keycloak
+        app: keycloak
 ---
 # HTTPRoute gitea-3000 (R-TLS/R-DNS)
 apiVersion: gateway.networking.k8s.io/v1

apps/wave-0/jackett.yaml

@@ -184,6 +184,10 @@ spec:
   egress:
   - toEntities:
     - world
+  - toEndpoints:
+    - matchLabels:
+        io.kubernetes.pod.namespace: keycloak
+        app: keycloak
 ---
 # HTTPRoute jackett-9117 (R-TLS/R-DNS)
 apiVersion: gateway.networking.k8s.io/v1

apps/wave-0/pihole.yaml

@@ -140,6 +140,10 @@ spec:
   egress:
   - toEntities:
     - world
+  - toEndpoints:
+    - matchLabels:
+        io.kubernetes.pod.namespace: keycloak
+        app: keycloak
 ---
 # HTTPRoute pihole-80 (R-TLS/R-DNS)
 apiVersion: gateway.networking.k8s.io/v1

apps/wave-0/qbittorrent.yaml

@@ -184,6 +184,10 @@ spec:
   egress:
   - toEntities:
     - world
+  - toEndpoints:
+    - matchLabels:
+        io.kubernetes.pod.namespace: keycloak
+        app: keycloak
 ---
 # HTTPRoute qbittorrent-8080 (R-TLS/R-DNS)
 apiVersion: gateway.networking.k8s.io/v1

apps/wave-1/nextcloud.yaml

@@ -205,6 +205,10 @@ spec:
   egress:
   - toEntities:
     - world
+  - toEndpoints:
+    - matchLabels:
+        io.kubernetes.pod.namespace: keycloak
+        app: keycloak
 ---
 # HTTPRoute nextcloud-80 (R-TLS/R-DNS)
 apiVersion: gateway.networking.k8s.io/v1

apps/wave-1/radarr.yaml

@@ -206,6 +206,10 @@ spec:
   egress:
   - toEntities:
     - world
+  - toEndpoints:
+    - matchLabels:
+        io.kubernetes.pod.namespace: keycloak
+        app: keycloak
 ---
 # HTTPRoute radarr-7878 (R-TLS/R-DNS)
 apiVersion: gateway.networking.k8s.io/v1

apps/wave-1/semaphore.yaml

@@ -209,6 +209,10 @@ spec:
   egress:
   - toEntities:
     - world
+  - toEndpoints:
+    - matchLabels:
+        io.kubernetes.pod.namespace: keycloak
+        app: keycloak
 ---
 # HTTPRoute semaphore-3000 (R-TLS/R-DNS)
 apiVersion: gateway.networking.k8s.io/v1

apps/wave-1/sonarr.yaml

@@ -206,6 +206,10 @@ spec:
   egress:
   - toEntities:
     - world
+  - toEndpoints:
+    - matchLabels:
+        io.kubernetes.pod.namespace: keycloak
+        app: keycloak
 ---
 # HTTPRoute sonarr-8989 (R-TLS/R-DNS)
 apiVersion: gateway.networking.k8s.io/v1

catalog/render.py

@@ -527,6 +527,28 @@ def r_internet_egress(s: dict) -> dict:
     switch (a cluster-wide singleton keyed on the egress-class pod label) decide the PATH; this
     policy decides that north-south egress is ALLOWED at all.
     """
+    egress = [{"toEntities": ["world"]}]
+    # SSO apps (oidc/forward-auth) reach Keycloak at keycloak.home, which coredns split-horizon
+    # resolves to the shared Gateway VIP (an L2-announced LoadBalancer IP Cilium classifies as
+    # `world`), so the pod hairpins THROUGH the Gateway. Under Cilium >=1.16.8/1.17.2 (the
+    # CVE-2025-30162 fix, present in 1.20), the Gateway's L7 proxy re-evaluates THIS pod's egress
+    # policy against Keycloak's REAL backend cluster identity (enforce_policy_on_l7lb) — `world`
+    # matches the VIP at L3/L4 but NOT the resolved keycloak identity at L7, so OIDC discovery/
+    # token 403s ("Access denied") without an explicit allow to the keycloak backend. Scope it to
+    # the apps that actually talk to Keycloak (auth mode oidc or forward-auth).
+    if s.get("auth", {}).get("mode") in ("oidc", "forward-auth"):
+        egress.append(
+            {
+                "toEndpoints": [
+                    {
+                        "matchLabels": {
+                            "io.kubernetes.pod.namespace": "keycloak",
+                            "app": "keycloak",
+                        }
+                    }
+                ]
+            }
+        )
     return {
         "apiVersion": "cilium.io/v2",
         "kind": "CiliumNetworkPolicy",
@@ -537,7 +559,7 @@ def r_internet_egress(s: dict) -> dict:
         },
         "spec": {
             "endpointSelector": {"matchLabels": {"app.kubernetes.io/name": s["name"]}},
-            "egress": [{"toEntities": ["world"]}],
+            "egress": egress,
         },
     }
 

platform/crafty/networkpolicy.yaml

@@ -80,7 +80,10 @@ spec:
 #     Bedrock binaries) when the operator creates/updates a server.
 #   - The oauth2-proxy sidecar must reach Keycloak at keycloak.home.keithjackson.dev, which coredns
 #     split-horizon resolves to the Gateway VIP (an L2-announced LoadBalancer IP Cilium classifies as
-#     `world`) — so `world` egress covers both the jar downloads AND the OIDC path (same as headlamp).
+#     `world`); the sidecar hairpins THROUGH the shared Gateway. `world` covers the jar downloads and
+#     the VIP at L3/L4, but under Cilium >=1.16.8/1.17.2 (CVE-2025-30162 fix, in 1.20) the Gateway's
+#     L7 proxy re-checks this egress against Keycloak's REAL backend identity (enforce_policy_on_l7lb),
+#     so the OIDC path 403s ("Access denied") without the explicit keycloak toEndpoints allow below.
 #   - The sidecar -> Crafty hop is pod-local (127.0.0.1:8443) and is NOT subject to network policy.
 apiVersion: cilium.io/v2
 kind: CiliumNetworkPolicy
@@ -96,3 +99,7 @@ spec:
   egress:
     - toEntities:
         - world
+    - toEndpoints:
+        - matchLabels:
+            io.kubernetes.pod.namespace: keycloak
+            app: keycloak

platform/headlamp/networkpolicy.yaml

@@ -53,9 +53,12 @@ spec:
 # Egress allow: Headlamp's backend must reach (a) the kube-apiserver — it forwards each user's OIDC
 # id_token there for per-user RBAC — and (b) Keycloak for the OIDC auth-code/token/JWKS exchange.
 # Keycloak is reached at keycloak.home.keithjackson.dev, which coredns split-horizon resolves to the
-# Gateway VIP (an L2-announced LoadBalancer IP that Cilium classifies as `world`), so `world` egress
+# Gateway VIP (an L2-announced LoadBalancer IP that Cilium classifies as `world`); the pod hairpins
-# covers the Keycloak path (same as the gitea OIDC bootstrap). DNS is already allowed by the
+# THROUGH the shared Gateway. `world` matches the VIP at L3/L4, but under Cilium >=1.16.8/1.17.2 (the
-# default-deny baseline above.
+# CVE-2025-30162 fix, present in 1.20) the Gateway's L7 proxy re-evaluates this egress against
+# Keycloak's REAL backend cluster identity (enforce_policy_on_l7lb) — so `world` alone 403s the OIDC
+# path ("Access denied"). The explicit toEndpoints allow to the keycloak backend below is required.
+# DNS is already allowed by the default-deny baseline above.
 apiVersion: cilium.io/v2
 kind: CiliumNetworkPolicy
 metadata:
@@ -71,3 +74,7 @@ spec:
     - toEntities:
         - kube-apiserver
         - world
+    - toEndpoints:
+        - matchLabels:
+            io.kubernetes.pod.namespace: keycloak
+            app: keycloak