sgx: add automated DCAP registration using in-cluster PCCS caching

.github/workflows/lib-build.yaml

@@ -35,6 +35,7 @@ jobs:
           - openssl-qat-engine
           - sgx-sdk-demo
           - sgx-aesmd-demo
+          - sgx-dcap-infra
           - dsa-dpdk-dmadevtest
           - intel-npu-demo
         builder: [buildah, docker]

demo/sgx-dcap-infra/Dockerfile

@@ -0,0 +1,59 @@
+FROM ubuntu:24.04 AS builder
+
+RUN apt update && \
+    env DEBIAN_FRONTEND=noninteractive apt install -y \
+    build-essential \
+    curl \
+    libcurl4-openssl-dev
+
+WORKDIR /opt/intel
+
+ARG SGX_SDK_URL=https://download.01.org/intel-sgx/sgx-linux/2.27/distro/ubuntu24.04-server/sgx_linux_x64_sdk_2.27.100.1.bin
+
+RUN curl -sSLfO ${SGX_SDK_URL} \
+ && export SGX_SDK_INSTALLER=$(basename $SGX_SDK_URL) \
+ && chmod +x $SGX_SDK_INSTALLER \
+ && ./$SGX_SDK_INSTALLER --prefix /opt/intel \
+ && rm $SGX_SDK_INSTALLER
+
+ARG DCAP_VERSION=DCAP_1.24
+ARG DCAP_TARBALL_SHA256="c9295f5fd3f489b2fbd5f0d33836b09420976506ac834bc9c9a401f4a6a1204a"
+
+RUN curl -sSLfO https://github.com/intel/confidential-computing.tee.dcap/archive/$DCAP_VERSION.tar.gz && \
+    echo "$DCAP_TARBALL_SHA256 $DCAP_VERSION.tar.gz" | sha256sum -c - && \
+    tar xzf $DCAP_VERSION.tar.gz && mv confidential-computing.tee.dcap-* SGXDataCenterAttestationPrimitives
+
+WORKDIR SGXDataCenterAttestationPrimitives/tools/PCKRetrievalTool
+
+RUN sed -e 's:sys/firmware/efi:run:g' -i App/utility.cpp \
+    && make
+
+FROM ubuntu:24.04
+
+WORKDIR /opt/intel/sgx-pck-id-retrieval-tool/
+COPY --from=builder /opt/intel/SGXDataCenterAttestationPrimitives/tools/PCKRetrievalTool/PCKIDRetrievalTool .
+
+RUN ln -sf /lib/x86_64-linux-gnu/libsgx_id_enclave.signed.so.1 && \
+    ln -sf /lib/x86_64-linux-gnu/libsgx_pce.signed.so.1
+
+ARG SGX_SDK_VERSION=2_27_100
+RUN apt update && apt install -y --no-install-recommends curl ca-certificates gpg \
+    && echo "deb [arch=amd64 signed-by=/usr/share/keyrings/intel-sgx.gpg] https://download.01.org/intel-sgx/sgx_repo/ubuntu noble main" | \
+    tee -a /etc/apt/sources.list.d/intel-sgx.list \
+    && curl -s https://download.01.org/intel-sgx/sgx_repo/ubuntu/intel-sgx-deb.key | \
+    gpg --dearmor --output /usr/share/keyrings/intel-sgx.gpg \
+    && curl -sSLf https://download.01.org/intel-sgx/sgx_repo/ubuntu/apt_preference_files/99sgx_${SGXSDK_VERSION}_noble_custom_version.cfg | \
+    tee -a /etc/apt/preferences.d/99sgx_sdk \
+    && apt update \
+    && apt install -y --no-install-recommends \
+       tdx-qgs \
+       libsgx-ae-pce \
+       libsgx-ae-id-enclave \
+       libsgx-ra-uefi \
+       libsgx-dcap-default-qpl
+
+RUN rm /etc/qgs.conf
+
+COPY dcap-registration-flow /usr/bin
+
+ENTRYPOINT ["/opt/intel/tdx-qgs/qgs", "--no-daemon", "-n=4"]

demo/sgx-dcap-infra/dcap-registration-flow

@@ -0,0 +1,25 @@
+#!/usr/bin/env bash
+
+set -u
+
+if [ ! -x "${PWD}"/PCKIDRetrievalTool ]; then
+	echo "dcap-registration-flow: PCKIDRetrievalTool must be in the workingDir and executable"
+	exit 1
+fi
+
+echo "Waiting for the PCCS to be ready ..."
+
+if ! curl --retry 20 --retry-delay 30 -k ${PCCS_URL}/sgx/certification/v4/rootcacrl &> /dev/null; then
+	echo "ERROR: PCCS pod didn't become ready after 20 minutes"
+	exit 1
+fi
+
+echo "PCCS is online, proceeding ..."
+
+ARGS="-user_token ${USER_TOKEN} -url ${PCCS_URL} -use_secure_cert ${SECURE_CERT}"
+
+echo "Calling PCKIDRetrievalTool ..."
+
+./PCKIDRetrievalTool ${ARGS}
+
+sleep infinity

deployments/sgx_dcap/base/kustomization.yaml

@@ -0,0 +1,11 @@
+resources:
+- node-services.yaml
+generatorOptions:
+  disableNameSuffixHash: true
+
+# required .env.pccs-credentials keys:
+# USER_TOKEN=
+secretGenerator:
+- name: pccs-credentials
+  envs:
+  - .env.pccs-credentials

deployments/sgx_dcap/base/node-services.yaml

@@ -0,0 +1,90 @@
+# TODO
+# cert-manager / service-ca certificates
+# CURL_CA_BUNDLE once ^ is available
+# NFD (TDX) nodeSelector
+# cpu and memory resources/limits
+# split service name and port (PCCS URL)
+apiVersion: apps/v1
+kind: DaemonSet
+metadata:
+  name: intel-dcap-node-infra
+spec:
+  selector:
+    matchLabels:
+      app: dcap-node-infra
+  template:
+    metadata:
+      annotations:
+        qcnl-conf: '{"pccs_url": "https://pccs-service:8042/sgx/certification/v4/", "use_secure_cert": false, "pck_cache_expire_hours": 168}'
+      labels:
+        app: dcap-node-infra
+        pccs-secure-cert: 'false'
+    spec:
+      automountServiceAccountToken: false
+      initContainers:
+      - name: platform-registration
+        image: intel/sgx-dcap-infra:devel
+        restartPolicy: Always
+        workingDir: "/opt/intel/sgx-pck-id-retrieval-tool/"
+        command: ['/usr/bin/dcap-registration-flow']
+        env:
+          - name: PCCS_URL
+            value: "https://pccs-service:8042"
+          - name: SECURE_CERT
+            valueFrom:
+              fieldRef:
+                fieldPath: metadata.labels['pccs-secure-cert']
+        envFrom:
+          - secretRef:
+              name: pccs-credentials
+        securityContext:
+          readOnlyRootFilesystem: true
+          allowPrivilegeEscalation: false
+          capabilities:
+            drop:
+              - ALL
+            add:
+              - LINUX_IMMUTABLE
+        resources:
+          limits:
+            sgx.intel.com/registration: 1
+      containers:
+      - name: tdx-qgs
+        image: intel/sgx-dcap-infra:devel
+        securityContext:
+          readOnlyRootFilesystem: true
+          allowPrivilegeEscalation: false
+          capabilities:
+            drop:
+              - ALL
+        resources:
+          limits:
+            sgx.intel.com/qe: 1
+        imagePullPolicy: IfNotPresent
+        env:
+        - name: QCNL_CONF_PATH
+          value: "/run/dcap/qcnl_conf"
+        - name: XDG_CACHE_HOME
+          value: "/run/dcap/cache"
+        volumeMounts:
+        - name: dcap-qcnl-cache
+          mountPath: /run/dcap/cache
+        - name: qgs-socket
+          mountPath: /var/run/tdx-qgs
+        - name: qcnl-config
+          mountPath: /run/dcap/
+          readOnly: true
+      volumes:
+      - name: dcap-qcnl-cache
+        emptyDir:
+          sizeLimit: 50Mi
+      - name: qgs-socket
+        hostPath:
+          path: /var/run/tdx-qgs
+          type: DirectoryOrCreate
+      - name: qcnl-config
+        downwardAPI:
+          items:
+          - path: "qcnl_conf"
+            fieldRef:
+              fieldPath: metadata.annotations['qcnl-conf']

deployments/sgx_dcap/kustomization.yaml

@@ -0,0 +1,3 @@
+resources:
+  - base
+  - pccs

deployments/sgx_dcap/pccs/kustomization.yaml

@@ -0,0 +1,25 @@
+resources:
+- pccs.yaml
+- service.yaml
+generatorOptions:
+  disableNameSuffixHash: true
+
+# self-signed TLS certs for pccs-tls:
+# openssl req -x509 -sha256 -nodes -days 365 -newkey rsa:2048 -keyout private.pem -out file.crt -subj "/C=US/ST=Denial/L=Springfield/O=Dis/CN=www.example.com"
+# token hashesh follow (with 'helloworld' changed to the desired secret tokens):
+# echo -n helloworld | sha512sum | tr -d '[:space:]-'
+# where helloworld is then used as the USER_TOKEN in intel-dcap-node-infra deployment:
+#
+# required .env.pccs-tokens keys:
+# PCS_API_KEY=
+# PCCS_USER_TOKEN_HASH=
+# PCCS_ADMIN_TOKEN_HASH=
+secretGenerator:
+- name: pccs-tokens
+  envs:
+  - .env.pccs-tokens
+- name: pccs-tls
+  type: "kubernetes.io/tls"
+  files:
+  - tls.key=private.pem
+  - tls.crt=file.crt

deployments/sgx_dcap/pccs/pccs.yaml

@@ -0,0 +1,53 @@
+# TODO
+# cert-manager / service-ca certificates
+# add PCCS nodeSelector
+# cpu and memory resources/limits
+# make HTTPS_PORT configurable via env
+# fix proxy setting label
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: intel-dcap-pccs
+spec:
+  replicas: 1
+  selector:
+    matchLabels:
+      app: pccs
+  template:
+    metadata:
+      labels:
+        app: pccs
+        trustedservices.intel.com/cache: pccs
+    spec:
+      containers:
+        - name: pccs
+          image: quay.io/redhat-user-workloads/ose-osc-tenant/osc-pccs@sha256:de64fc7b13aaa7e466e825d62207f77e7c63a4f9da98663c3ab06abc45f2334d
+          ports:
+            - containerPort: 8042
+              name: pccs-port
+          volumeMounts:
+            - name: pccs-cache
+              mountPath: /run/pccs
+            - name: pccs-tls
+              mountPath: /opt/app-root/src/intel/pccs/ssl_key
+              readOnly: true
+          env:
+            - name: PCCS_FILL_MODE
+              value: "REQ"
+            - name: CLUSTER_HTTPS_PROXY
+              value: ""
+          envFrom:
+            - secretRef:
+                name: pccs-tokens
+      volumes:
+        - name: pccs-cache
+          emptyDir:
+            sizeLimit: 50Mi
+        - name: pccs-tls
+          secret:
+            secretName: pccs-tls
+            items:
+            - key: tls.key
+              path: private.pem
+            - key: tls.crt
+              path: file.crt

deployments/sgx_dcap/pccs/service.yaml

@@ -0,0 +1,12 @@
+apiVersion: v1
+kind: Service
+metadata:
+  name: pccs-service
+spec:
+  selector:
+    trustedservices.intel.com/cache: pccs
+  ports:
+  - name: pccs
+    protocol: TCP
+    port: 8042
+    targetPort: pccs-port

deployments/sgx_plugin/overlays/dcap-infra-resources/add-args.yaml

@@ -0,0 +1,11 @@
+apiVersion: apps/v1
+kind: DaemonSet
+metadata:
+  name: intel-sgx-plugin
+spec:
+  template:
+    spec:
+      containers:
+      - name: intel-sgx-plugin
+        args:
+          - "-dcap-infra-resources"

deployments/sgx_plugin/overlays/dcap-infra-resources/kustomization.yaml

@@ -0,0 +1,7 @@
+resources:
+  - ../../base
+patches:
+  - path: add-args.yaml
+    target:
+      kind: DaemonSet
+