Skip to main content

Set up Trusted Execution Environment (TEE)

info

In case the ParaTime you want to run does not require the use of a TEE (e.g. Intel SGX), you can skip setting up a TEE.

If the ParaTime is configured to run in a TEE (currently only Intel SGX), you must make sure that your system supports running SGX enclaves. This requires that your hardware has SGX support, that SGX support is enabled and that the additional driver and software components are properly installed and running.

BIOS Configuration​

To enable Intel SGX on your hardware, you also need to configure the BIOS. First, update the BIOS to the latest version with the latest microcode and then proceed with BIOS configuration as shown below. Note that some settings may not apply to your BIOS. In that case, configure only the relevant ones. Please set the BIOS settings as follows:

  • SGX: ENABLE
  • Hyper-Threading: DISABLE
  • Intel SpeedStep: DISABLE
  • Turbo Mode: DISABLE
  • CPU AES: ENABLE
  • SGX Auto MP Registration: ENABLE

Ensure Clock Synchronization​

Due to additional sanity checks within runtime enclaves, you should ensure that the node's local clock is synchronized (e.g. using NTP). If it is off by more than half a second you may experience unexpected runtime aborts.

Ensure Proper SGX Device Permissions​

Make sure that the user that is running the Oasis Node binary has access to the SGX device (e.g. /dev/sgx_enclave). This can usually be achieved by adding the user into the right group, for example in case the permissions of the SGX device are as follows:

crw-rw---- 1 root sgx 10, 125 Oct 28 09:28 /dev/sgx_enclave

and the user running Oasis Node is oasis, you can do:

sudo adduser oasis sgx

Failure to do so may result in the "permission denied OS error 13" during runtime startup.

If you are planning to run your node from an interactive session, make sure to log out for permissions to take effect.

AESM Service​

To allow execution of SGX enclaves, several Architectural Enclaves (AE) are involved (i.e. Launch Enclave, Provisioning Enclave, Provisioning Certificate Enclave, Quoting Enclave, Platform Services Enclaves).

Communication between application-spawned SGX enclaves and Intel-provided Architectural Enclaves is through Application Enclave Service Manager (AESM). AESM runs as a daemon and provides a socket through which applications can facilitate various SGX services such as launch approval, remote attestation quote signing, etc.

Oasis node requires the use of DCAP attestation. To see if your system supports it, run the following:

 cpuid -1  | grep "SGX"

and look for the following line:

      SGX_LC: SGX launch config supported      = true

DCAP Attestation​

Ubuntu 22.04+​

A convenient way to install the AESM service on Ubuntu 22.04 systems is to use the Intel's official Intel SGX APT repository.

First add Intel SGX APT repository to your system:

curl -fsSL https://download.01.org/intel-sgx/sgx_repo/ubuntu/intel-sgx-deb.key | sudo gpg --dearmor -o /usr/share/keyrings/intel-sgx-deb.gpg
echo "deb [arch=amd64 signed-by=/usr/share/keyrings/intel-sgx-deb.gpg] https://download.01.org/intel-sgx/sgx_repo/ubuntu $(lsb_release -cs) main" | sudo tee /etc/apt/sources.list.d/intel-sgx.list > /dev/null

And then install the sgx-aesm-service, libsgx-aesm-ecdsa-plugin, libsgx-aesm-quote-ex-plugin and libsgx-dcap-default-qpl packages:

sudo apt update
sudo apt install sgx-aesm-service libsgx-aesm-ecdsa-plugin libsgx-aesm-quote-ex-plugin libsgx-dcap-default-qpl

The AESM service should be up and running. To confirm that, use:

sudo systemctl status aesmd.service

Configuring the Quote Provider​

The Intel Quote Provider (libsgx-dcap-default-qpl) needs to be configured in order to use either the Intel PCS, the PCCS of your cloud service provider, or your own PCCS. The configuration file is located at /etc/sgx_default_qcnl.conf.

Make sure to always restart the aesmd.service after updating the configuration, via:

sudo systemctl restart aesmd.service

Intel PCS​

Using the Intel PCS is the simplest and most generic way, but it may be less reliable than using your own PCCS. Some cloud providers (see the following section) also require you to use their PCCS.

To use Intel PCS update the pccs_url value in /etc/sgx_default_qcnl.conf to the Intel PCS API URL:

  //PCCS server address
  "pccs_url": "https://api.trustedservices.intel.com/sgx/certification/v4/"
tip

In case there is an error in the QPL configuration file, attestation will refuse to work and the AESM service may produce unhelpful errors like the following:

Couldn't find the platform library. (null)

The only thing that needs to be changed is the pccs_url value above. Do not add any comments and/or modify punctuation as these could make the configuration file invalid.

Cloud Service Provider's PCCS​

Some cloud providers require you to use their PCCS.

  • Azure: See the Azure documentation for details on configuring the quote provider. The documentation contains an example of an Intel QPL configuration file that can be used.

  • Alibaba Cloud: See the Alibaba Cloud documentation for details on configuring the quote provider. The documentation shows the required sgx_default_qcnl.conf changes.

  • Other cloud providers: If you are using a different cloud service provider, consult their specific documentation for the appropriate PCCS configuration and guidance on configuring the quote provider, or use one of the other PCCS options.

Own PCCS​

It is also possible to run PCCS yourself. Follow official Intel instructions on how to setup your own PCCS.

DCAP Attestation Docker​

Alternatively, an easy way to install and run the AESM service on a Docker-enabled system is to use our AESM container image.

Executing the following command should (always) pull the latest version of our AESMD Docker container, map the SGX devices and /var/run/aesmd directory and ensure AESM is running in the background (also automatically started on boot):

docker run \
  --pull always \
  --detach \
  --restart always \
  --device /dev/sgx_enclave \
  --device /dev/sgx_provision \
  --volume /var/run/aesmd:/var/run/aesmd \
  --name aesmd \
  ghcr.io/oasisprotocol/aesmd-dcap:master

By default, the Intel Quote Provider in the docker container is configured to use the Intel PCS endpoint. To override the Intel Quote Provider configuration within the container mount your own custom configuration using the volume flag.

docker run \
  --pull always \
  --detach \
  --restart always \
  --device /dev/sgx_enclave \
  --device /dev/sgx_provision \
  --volume /var/run/aesmd:/var/run/aesmd \
  --volume /etc/sgx_default_qcnl.conf:/etc/sgx_default_qcnl.conf \
  --name aesmd \
  ghcr.io/oasisprotocol/aesmd-dcap:master

The default Intel Quote Provider config is available in Intel SGX Github repository.

Multi-socket Systems​

Note that platform provisioning for multi-socket systems (e.g. systems with multiple CPUs) is more complex, especially if one is using a hypervisor and running SGX workloads inside guest VMs. In this case additional provisioning may be required to be performed on the host.

Note that the system must be booted in UEFI mode for provisioning to work as the provisioning process uses UEFI variables to communicate with the BIOS. In addition the SGX Auto MP Registration BIOS configuration setting should be set to enabled.

Ubuntu 22.04+​

To provision and register your multi-socket system you need to install the Intel SGX Multi-Package Registration Agent Service as follows (assuming Intel's SGX apt repository has been added as discussed above):

sudo apt install sgx-ra-service

After boot, the log in /var/log/mpa_registration.log should indicate successful registration. If an error is reported, make sure that you have enabled SGX Auto MP Registration in the BIOS as mentioned above. You can also perform re-provisioning by rebooting and setting the SGX Factory Reset option to enabled.

VMware vSphere 8.0+​

In order to enable SGX remote attestation on VMware vSphere-based systems, please follow the vSphere guide.

Migrate from EPID Attestation to DCAP Attestation​

EPID attestation will be discontinued in 2025 and will no longer be available on any processors. All nodes using EPID attestation must migrate to DCAP attestation.

For transitioning to the DCAP attestation, follow these steps:

  1. See if your system supports DCAP attestation. If your hardware does not support DCAP attestation, you'll need to migrate your node to newer hardware.
  2. Transition to DCAP attestation:
  • In case you are running AESM service on Docker follow these instructions.
  • Otherwise manually configure AESM service to use DCAP attestation:
    1. Remove any leftover EPID attestation packages. If running on Ubuntu 22.04 run the following command:
      sudo apt remove libsgx-aesm-launch-plugin libsgx-aesm-epid-plugin
    1. Configure AESM service to use DCAP attestation
    2. Restart the AESM service. If running on Ubuntu 22.04 run the following command:
    sudo systemctl restart aesmd.service
  1. Configure the Quote Provider.
  2. Use the attestation tool to test if your settings are correct.
  3. Restart your compute node and verify that node is ready.

Check SGX Setup​

In order to make sure that your SGX setup is working, you can use the sgx-detect tool from the sgxs-tools Rust package.

There are no pre-built packages for it, so you will need to compile it yourself.

Install Dependencies​

Make sure you have the following installed on your system:

On Fedora, you can install all the above with:

sudo dnf install gcc protobuf-compiler pkg-config openssl-devel

On Ubuntu, you can install all the above with:

sudo apt install gcc protobuf-compiler pkg-config libssl-dev

Install Rust​

We follow Rust upstream's recommendation on using rustup to install and manage Rust versions.

caution

rustup cannot be installed alongside a distribution packaged Rust version. You will need to remove it (if it's present) before you can start using rustup.

Install rustup by running:

curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh
tip

If you want to avoid directly executing a shell script fetched the internet, you can also download rustup-init executable for your platform and run it manually. This will run rustup-init which will download and install the latest stable version of Rust on your system.

Build and Install sgxs-tools​

cargo install sgxs-tools

Run sgx-detect Tool​

After the installation completes, run sgx-detect to make sure that everything is set up correctly:

sudo $(which sgx-detect)
tip

If you don't run the sgx-detect tool as root, it might not have the necessary permissions to access the SGX kernel device.

When everything works, you should get output similar to the following (some things depend on hardware features so your output may differ):

Detecting SGX, this may take a minute...
✔  SGX instruction set
  ✔  CPU support
  ✔  CPU configuration
  ✔  Enclave attributes
  ✔  Enclave Page Cache
  SGX features
    ✔  SGX2  ✔  EXINFO  ✔  ENCLV  ✔  OVERSUB  ✔  KSS
    Total EPC size: 92.8MiB
✘  Flexible launch control
  ✔  CPU support
  ？ CPU configuration
  ✘  Able to launch production mode enclave
✔  SGX system software
  ✔  SGX kernel device (/dev/sgx_enclave)
  ✘  libsgx_enclave_common
  ✔  AESM service
  ✔  Able to launch enclaves
    ✔  Debug mode
    ✘  Production mode
    ✔  Production mode (Intel whitelisted)

The important part is the checkbox under Able to launch enclaves in both Debug mode and Production mode (Intel whitelisted).

In case you encounter errors, see the list of common SGX installation issues for help.

Oasis Attestation tool​

To test if your settings are correct, you may use the attestation tool (binary) for testing remote attestation against Intel SGX's development server.

Troubleshooting​

See the general troubleshooting section, before proceeding with ParaTime node-specific troubleshooting.

AESM could not be contacted​

If running sgx-detect --verbose reports:

🕮  SGX system software > AESM service
AESM could not be contacted. AESM is needed for launching enclaves and generating attestations.

Please check your AESM installation.

debug: error communicating with aesm
debug: cause: Connection refused (os error 111)

More information: https://edp.fortanix.com/docs/installation/help/#aesm-service

Ensure you have completed all the necessary installation steps outlined in DCAP Attestation section.

AESM: error 30​

If you are encountering the following error message in your node's logs:

failed to initialize TEE: error while getting quote info from AESMD: aesm: error 30

Ensure you have all required SGX driver libraries installed as listed in DCAP Attestation section.

Permission Denied When Accessing SGX Kernel Device​

If running sgx-detect --verbose reports:

🕮  SGX system software > SGX kernel device
Permission denied while opening the SGX device (/dev/sgx/enclave, /dev/sgx or
/dev/isgx). Make sure you have the necessary permissions to create SGX enclaves.
If you are running in a container, make sure the device permissions are
correctly set on the container.

debug: Error opening device: Permission denied (os error 13)
debug: cause: Permission denied (os error 13)

Ensure you are running the sgx-detect tool as root via:

sudo $(which sgx-detect) --verbose

Error Opening SGX Kernel Device​

If running sgx-detect --verbose reports:

🕮  SGX system software > SGX kernel device
The SGX device (/dev/sgx/enclave, /dev/sgx or /dev/isgx) could not be opened:
"/dev" mounted with `noexec` option.

debug: Error opening device: "/dev" mounted with `noexec` option
debug: cause: "/dev" mounted with `noexec` option

Ensure /dev is NOT Mounted with the noexec Option​

Some Linux distributions mount /dev with the noexec mount option. If that is the case, it will prevent the enclave loader from mapping executable pages.

Ensure your /dev (i.e. devtmpfs) is not mounted with the noexec option. To check that, use:

cat /proc/mounts | grep devtmpfs

To temporarily remove the noexec mount option for /dev, run:

sudo mount -o remount,exec /dev

To permanently remove the noexec mount option for /dev, add the following to the system's /etc/fstab file:

devtmpfs        /dev        devtmpfs    defaults,exec 0 0
info

This is the recommended way to modify mount options for virtual (i.e. API) file system as described in systemd's API File Systems documentation.

Unable to Launch Enclaves: Operation not permitted​

If running sgx-detect --verbose reports:

🕮  SGX system software > Able to launch enclaves > Debug mode
The enclave could not be launched.

debug: failed to load report enclave
debug: cause: failed to load report enclave
debug: cause: Failed to map enclave into memory.
debug: cause: Operation not permitted (os error 1)

Ensure your system's /dev is NOT mounted with the noexec mount option.

Unable to Launch Enclaves: Invalid argument​

If running sgx-detect --verbose reports:

🕮  SGX system software > Able to launch enclaves > Debug mode
The enclave could not be launched.

debug: failed to load report enclave
debug: cause: Failed to call EINIT.
debug: cause: I/O ctl failed.
debug: cause: Invalid argument (os error 22)

This may be related to a bug in the Linux kernel when attempting to run enclaves on certain hardware configurations. Upgrading the Linux kernel to a version equal to or greater than 6.5.0 may solve the issue.

Couldn't find the platform library​

If AESMD service log reports:

[read_persistent_data ../qe_logic.cpp:1084] Couldn't find the platform library. (null)
[get_platform_quote_cert_data ../qe_logic.cpp:438] Couldn't load the platform library. (null)

It may be that the DCAP quote provider is misconfigured or the configuration file is not a valid JSON file but is malformed. Double-check that its configuration file (e.g. /etc/sgx_default_qcnl.conf) is correct.