Skip to main content
Version: Next

Multi-Platform Patching

Copa also supports patching multi-platform container images, streamlining the process of securing applications deployed across diverse hardware platforms. This guide explains how Copa handles multi-platform images and how you can use this feature.

Usage​

To patch a multi-platform image, you can use the copa patch command with the --report flag pointing to a directory (which tells Copa this will be a multi-platform patch) along with flags to specify your image, and desired output tag.

Create vulnerability reports for multi-platform images​

Before you can patch a multi-platform image, you need to generate vulnerability reports for each platform architecture. You can do this using trivy using --platform flag to specify the architecture. Below is an example of how to generate vulnerability reports for a multi-platform image like nginx:1.25.0.

export IMAGE=docker.io/library/nginx:1.25.0 # Replace with your multi-platform image

mkdir -p reports

trivy image --vuln-type os --scanners vuln --ignore-unfixed \
  -f json -o reports/amd64.json \
  --platform linux/amd64 $IMAGE
trivy image --vuln-type os --scanners vuln --ignore-unfixed \
  -f json -o reports/arm64.json \
  --platform linux/arm64 $IMAGE

This will create two JSON files in the reports directory, one for each architecture (amd64 and arm64).

Patching Multi-Platform Images​

To patch a multi-platform image, you can use the copa patch command with the --image flag to specify the multi-platform image, the --report flag to point to the directory containing your vulnerability reports, and optionally a --tag for the final patched image.

copa patch \
  --image $IMAGE \
  --report reports

Key Flags for Multi-Platform Patching:

  • --report <directory_path>: Specifies the directory containing platform-specific vulnerability reports.
  • --tag <final_tag> (optional): The tag for the final, reassembled multi-platform manifest (e.g., 1.0-patched).
  • --push (optional): If included, Copa pushes the final multi-platform manifest to the registry.
  • --ignore-errors (optional, default: false): When false (default), Copa warns about errors and fails if any platform encounters an error. When true, Copa warns about errors but continues processing other platforms.

Things to Keep in Mind​

If you don't provide a --report flag pointing to a directory, Copa will not perform multi-platform patching and will instead only patch the image for the architecture of the host machine.

If --push is not specified, the individual patched images will be saved locally, and you can push them to your registry later using docker push and then docker manifest create/push to create the multi-platform manifest.

Emulation and QEMU for Cross-Platform Patching ⚙️​

When patching an image for an architecture different from your host machine's architecture (e.g., patching an arm64 image on an amd64 machine), Copa relies on emulation. This is often necessary for multi-platform image patching, as you might not have native hardware for every architecture you intend to patch.

Copa leverages BuildKit, which in turn can use QEMU for emulation. QEMU is a generic and open-source machine emulator and virtualizer. When BuildKit detects that it needs to execute binaries for a foreign architecture, it can use QEMU user-mode emulation to run those commands.

Why Emulation is Needed​

  • Running Package Managers: To apply patches, Copa needs to execute the package manager (like apk, apt, yum) inside the environment of the target image's architecture. If you're on an amd64 host trying to patch an arm64 image, the arm64 package manager won't run natively. QEMU bridges this gap.
  • Ensuring Correctness: Emulation helps ensure that the patches are applied in an environment that closely mirrors the target architecture, reducing the chances of incompatibilities.

Setting up QEMU​

Docker Desktop (macOS and Windows) comes pre-configured with QEMU emulation support and requires no additional setup.

For Linux hosts or when using BuildKit outside of Docker Desktop, your host system (where the copa command and BuildKit daemon are running) needs to have QEMU static binaries registered with the kernel's binfmt_misc handler. This allows the kernel to automatically invoke QEMU when it encounters a binary for a foreign architecture.

Installation Steps (Linux/Non-Docker Desktop environments):

One way to set this up, especially in Dockerized environments or on Linux hosts, is to use the multiarch/qemu-user-static image:

  1. Ensure your kernel supports binfmt_misc: Most modern Linux kernels do.

  2. Register QEMU handlers: You can do this by running the multiarch/qemu-user-static Docker image with privileged mode:

    docker run --rm --privileged multiarch/qemu-user-static --reset -p yes