Intel Software Guard Extensions (SGX) device plugin for Kubernetes



The Intel SGX device plugin and related components allow workloads to use Intel SGX on platforms with SGX Flexible Launch Control enabled, e.g.,:

  • 3rd Generation Intel® Xeon® Scalable Platform, code-named “Ice Lake”

  • Intel® Xeon® E3

  • Intel® NUC Kit NUC7CJYH

The SGX solution comes in three parts:

This README covers setting up all three components.

Modes and Configuration options

The SGX plugin can take a number of command line arguments, summarised in the following table:

Flag Argument Meaning
-enclave-limit int the number of containers per node allowed to use /dev/sgx/enclave (default: 20)
-provision-limit int the number of containers per node allowed to use /dev/sgx/provision (default: 20)

The plugin also accepts a number of other arguments related to logging. Please use the -h option to see the complete list of logging related options.


The following sections cover how to obtain, build and install the necessary Kubernetes SGX specific components.

They can be installed either using a DaemonSet or running ‘by hand’ on each node.


The component has the same basic dependancies as the generic plugin framework dependencies.

The SGX device plugin requires Linux Kernel SGX drivers to be available. These drivers are available in Linux 5.11. The SGX DCAP out-of-tree driver v1.41 is also known to work.

The hardware platform must support SGX Flexible Launch Control.

Backwards compatibility note

The SGX device nodes have changed from /dev/sgx/[enclave|provision] to /dev/sgx_[enclave|provision] in v4x RFC patches according to the LKML feedback.

Backwards compatibility is provided by adding /dev/sgx directory volume mount to containers. This assumes the cluster admin has installed the udev rules provided below to make the old device nodes as symlinks to the new device nodes.

Note: the symlinks become visible in all containers requesting SGX resources but are potentially dangling links if the device the corresponding device resource is not requested.

$ cat /etc/udev/rules/9*.rules
$ sudo udevadm trigger
$ ls -la /dev/sgx/*
lrwxrwxrwx 1 root root 14 Nov 18 01:01 /dev/sgx/enclave -> ../sgx_enclave
lrwxrwxrwx 1 root root 16 Nov 18 01:01 /dev/sgx/provision -> ../sgx_provision

The backwards compatibility will be removed in the next release (v0.20) and from the main development branch once the SGX SDK and DCAP releases default to the new devices.

Pre-built images

Pre-built images are available on Docker Hub. These images are automatically built and uploaded to the hub from the latest main branch of this repository.

Release tagged images of the components are also available on Docker Hub, tagged with their release version numbers in the format x.y.z, corresponding to the branches and releases in this repository. Thus the easiest way to deploy Intel SGX components in your cluster is to follow the steps below.

The deployment YAML files supplied with the components in this repository use the images with the devel tag by default. If you do not build your own local images, your Kubernetes cluster may pull down the devel images from Docker Hub by default.

<RELEASE_VERSION> needs to be substituted with the desired release version, e.g. v0.19.0 or main.

Deploy node-feature-discovery

$ kubectl apply -k<RELEASE_VERSION>

Note: The default nfd-worker.conf assumes that the in-tree driver is used and enabled (CONFIG_X86_SGX=y). If the SGX DCAP out-of-tree driver is used, the kConfig match in nfd-worker.conf must be removed.

Deploy cert-manager

$ kubectl apply -f

Deploy Intel Device plugin operator

$ kubectl apply -k<RELEASE_VERSION>

Note: See the operator deployment details for setting it up on systems behind proxies.

Deploy SGX device plugin with the operator

$ kubectl apply -f

Getting the source code

$ export INTEL_DEVICE_PLUGINS_SRC=/path/to/intel-device-plugins-for-kubernetes

Deploying as a DaemonSet

To deploy the plugin as a DaemonSet, you first need to build a container image for the plugin and ensure that is visible to your nodes.

Build the plugin and EPC source images

The following will use docker to build a local container images called intel/intel-sgx-plugin and intel/intel-sgx-initcontainer with the tag devel. The image build tool can be changed from the default docker by setting the BUILDER argument to the Makefile.

$ make intel-sgx-plugin
Successfully tagged intel/intel-sgx-plugin:devel
$ make intel-sgx-initcontainer
Successfully tagged intel/intel-sgx-initcontainer:devel

Deploy the DaemonSet

There are two alternative ways to deploy SGX device plugin.

The first approach involves deployment of the SGX DaemonSet YAML and node-feature-discovery with the necessary configuration.

There is a kustomization for deploying everything:

$ kubectl apply -k ${INTEL_DEVICE_PLUGINS_SRC}/deployments/sgx_plugin/overlays/epc-nfd/

The second approach has a lesser deployment footprint. It does not deploy NFD, but a helper daemonset that creates'true' node label and advertises EPC capacity to the API server.

The following kustomization is used for this approach:

$ kubectl apply -k ${INTEL_DEVICE_PLUGINS_SRC}/deployments/sgx_plugin/overlays/epc-register/

Verify SGX device plugin is registered:

Verification of the plugin deployment and detection of SGX hardware can be confirmed by examining the resource allocations on the nodes:

$ kubectl describe node <node name> | grep
             20        98566144  20    20        98566144  20    1           1        400         400  1           1

Deploying by hand

For development purposes, it is sometimes convenient to deploy the plugin ‘by hand’ on a node. In this case, you do not need to build the complete container image, and can build just the plugin.

Build SGX device plugin

$ make sgx_plugin

Deploy SGX plugin

Deploy the plugin on a node by running it as root. The below is just an example - modify the paramaters as necessary for your setup:

$ sudo -E ${INTEL_DEVICE_PLUGINS_SRC}/cmd/sgx_plugin/sgx_plugin -enclave-limit 50 -provision-limit 1 -v 2
I0626 20:33:01.414446  964346 server.go:219] Start server for provision at: /var/lib/kubelet/device-plugins/
I0626 20:33:01.414640  964346 server.go:219] Start server for enclave at: /var/lib/kubelet/device-plugins/
I0626 20:33:01.417315  964346 server.go:237] Device plugin for provision registered
I0626 20:33:01.417748  964346 server.go:237] Device plugin for enclave registered

SGX device plugin demos

SGX ECDSA Remote Attestation

The SGX remote attestation allows a relying party to verify that the software is running inside an Intel® SGX enclave on a platform that has the trusted computing base up to date.

The demo guides to run an SGX DCAP/ECDSA quote generation in on a single-node kubernetes cluster using Intel® reference SGX PCK Certificate Cache Service (PCCS) that is configured to service localhost connections.

Read more about SGX Remote Attestation.

Remote Attestation Prerequisites

For the SGX ECDSA Remote Attestation demo to work, the platform must be correctly registered and a PCCS running.

For documentation to set up Intel® reference PCCS, refer to: Intel® Software Guard Extensions (Intel® SGX) Services and Intel® Software Guard Extensions SDK for Linux

Furthermore, the Kubernetes cluster must be set up according the instructions above.

Build the image

The demo uses container images build from Intel® SGX SDK and DCAP releases.

To build the demo images:

$ make sgx-aesmd-demo
Successfully tagged intel/sgx-aesmd-demo:devel
$ make sgx-sdk-demo
Successfully tagged intel/sgx-sdk-demo:devel
Deploy the pods

The demo runs Intel aesmd (architectural enclaves service daemon) that is responsible for generating SGX quotes for workloads. It is deployed with hostNetwork: true to allow connections to localhost PCCS.

$ kubectl apply -k<RELEASE_VERSION>
$ kubectl get pods
  NAME                     READY     STATUS    RESTARTS   AGE
  intel-sgx-aesmd-mrnm8                1/1     Running   0          3h47m
  sgxdeviceplugin-sample-z5dcq-llwlw   1/1     Running   0          28m

The sample application runs SGX DCAP Quote Generation sample:

$ kubectl apply -k<RELEASE_VERSION>
$ kubectl get pods
  NAME                                 READY   STATUS      RESTARTS   AGE
  intel-sgx-aesmd-mrnm8                1/1     Running     0          3h55m
  ecdsa-quote-intelsgx-demo-job-vtq84  0/1     Completed   0          4s
  sgxdeviceplugin-sample-z5dcq-llwlw   1/1     Running     0          35m
$ kubectl logs ecdsa-quote-intelsgx-demo-job-vtq84

  Step1: Call sgx_qe_get_target_info:succeed!
  Step2: Call create_app_report:succeed!
  Step3: Call sgx_qe_get_quote_size:succeed!
  Step4: Call sgx_qe_get_quote:succeed!cert_key_type = 0x5

Note: The deployment example above uses kustomize that is available in kubectl since Kubernetes v1.14 release.