SAGE安装和运行

@misc{https://doi.org/10.48550/arxiv.2209.03125,
  doi = {10.48550/ARXIV.2209.03125},
  url = {https://arxiv.org/abs/2209.03125},
  author = {Ivanov, Andrei and Rothenberger, Benjamin and Dethise, Arnaud and Canini, Marco and Hoefler, Torsten and Perrig, Adrian},
  keywords = {Cryptography and Security (cs.CR), FOS: Computer and information sciences, FOS: Computer and information sciences},
  title = {SAGE: Software-based Attestation for GPU Execution},
  publisher = {arXiv},
  year = {2022},
  copyright = {arXiv.org perpetual, non-exclusive license}
}

引用如上，paper title: SAGE:software-based attestation for GPU execution

 github链接是：GitHub - spcl/sage

第一部分：SGX

Prerequisites installation instructions (~30-60 min)

1. Install SGX SDK (this may take around 30-60 minutes depending on the hardware). Follow installation instructions here (default configuration, Ubuntu 20.04): GitHub - intel/linux-sgx: Intel SGX for Linux*

2. Install NVCC

sudo apt update
sudo apt install nvidia-cuda-toolkit

上述都没有任何问题，很顺利。

SAGE Compilation instructions (~5 min)

Before compiling make sure that the environment variables for SGX SDK are set:

source ${sgx-sdk-install-path}/environment

Compile in simulation mode using:

make SGX_MODE=SIM

Run using:

./app

上述仍然没有任何问题。

End-to-end build inside container (~30-60 min)

podman run -it --rm -v ./:/SGX -w /SGX --security-opt label=disable docker.io/nvidia/cuda:11.8.0-devel-ubuntu20.04 /bin/bash -c "\
    export DEBIAN_FRONTEND=noninteractive && \
    apt-get update && \
    apt-get install -y build-essential ocaml ocamlbuild automake autoconf libtool wget python-is-python3 libssl-dev git cmake perl && \
    apt-get install -y libssl-dev libcurl4-openssl-dev protobuf-compiler libprotobuf-dev debhelper cmake reprepro unzip pkgconf libboost-dev libboost-system-dev libboost-thread-dev lsb-release libsystemd0 && \
    git clone -b sgx_2.19 https://github.com/intel/linux-sgx.git ; \
    cd linux-sgx && make preparation && \
    cp external/toolset/ubuntu20.04/* /usr/local/bin && \
    make sdk_install_pkg && \
    cd /SGX && \
    (echo yes | ./linux/installer/bin/sgx_linux_x64_sdk_2.19.100.3.bin) && \
    source sgxsdk/environment && \
    export LD_LIBRARY_PATH=/usr/local/cuda/lib64/:$LD_LIBRARY_PATH && \
    export LIBRARY_PATH=/usr/local/cuda/lib64/:$LIBRARY_PATH && \
    make SGX_MODE=SIM && \
    exec bash"

首先，podman安装在我的另一篇文章里写到了，这里不赘述。

其次，podman根本run不成功，因此我把上述代码改成了

docker run -it --rm -v ./:/SGX -w /SGX --security-opt label=disable docker.io/nvidia/cuda:11.8.0-devel-ubuntu20.04 /bin/bash

仍然显示timeout

解决办法：去到hub.docker.com/r/nvidia/cuda这个网站，搜索我的cuda版本11.4，按照网站里给的命令

docker pull nvidia/cuda:11.4.3-devel-ubuntu20.04

先拉取镜像。

在拉取过程中，时而会卡，我切换了网络，多试几次就好了。

完成之后可以输入docker image，看到确实已经存在了nvidia/cuda。

此时重复run的这行代码，但是请注意需要改为

docker run -it --rm -v ./:/SGX -w /SGX --security-opt label=disable docker.io/nvidia/cuda:11.4.3-devel-ubuntu20.04 /bin/bash

立刻就进入容器了。

我个人认为没必要一行运行这么多命令，所以我分步进行下述操作。

我没有提到的命令行都是没有问题的。

git clone -b sgx_2.19 https://github.com/intel/linux-sgx.git

上述这行需要删除掉“-b sgx_2.19”

(echo yes | ./linux/installer/bin/sgx_linux_x64_sdk_2.19.100.3.bin)

上述这行显示没有这个文件。

我需要先

cd /SGX/linux-sgx/linux/installer/bin/

再

(echo yes | ./sgx_linux_x64_sdk_2.19.100.3.bin)

成功之后会显示

The SDK package can be found in /SGX/linux-sgx/linux/installer/bin/sgxsdk

make SGX_MODE=SIM && \

上述这行显示没有这个文件。

我需要先

cd /SGX/linux-sgx/SampleCode/SampleEnclave

再继续上述make操作即可。

运行./app就显示SGX内的样例代码已经成功运行起来。

第一部分结束。