<?xml version="1.0" encoding="utf-8"?>
<feed xmlns="http://www.w3.org/2005/Atom">
  <title>Lei Zhou — updates</title>
  <subtitle>Mixed criticality, platform security and Physical AI: new case studies, library notes and research updates.</subtitle>
  <link href="https://zlhk100.github.io/feed.xml" rel="self"/>
  <link href="https://zlhk100.github.io/"/>
  <id>https://zlhk100.github.io/</id>
  <updated>2026-11-15T18:00:00-05:00</updated>
  <author><name>Lei Zhou</name><uri>https://zlhk100.github.io/</uri></author>
  <rights>Written content CC BY 4.0 unless noted</rights>
  <entry>
    <title>From virgl to zero-copy gfxstream</title>
    <link href="https://zlhk100.github.io/library/zero-copy-gpu-virtualisation.html"/>
    <id>https://zlhk100.github.io/library/zero-copy-gpu-virtualisation.html</id>
    <updated>2026-11-15T18:00:00-05:00</updated>
    <summary>Moving an Android Automotive guest from virgl to gfxstream on an Arm SoC whose hypervisor cannot share host memory at runtime: three compatibility contracts, zero copy from guest memory, and per-frame attribution of virtualisation overhead.</summary>
  </entry>
  <entry>
    <title>Case study: GPU virtualisation for a software-defined vehicle cockpit</title>
    <link href="https://zlhk100.github.io/work/gpu-virtualisation-cockpit.html"/>
    <id>https://zlhk100.github.io/work/gpu-virtualisation-cockpit.html</id>
    <updated>2026-11-15T17:55:00-05:00</updated>
    <summary>Moving an Android Automotive guest from virgl to gfxstream, with Vulkan, GLES through ANGLE and zero-copy rendering from guest memory, then optimising it to the customer’s performance targets.</summary>
  </entry>
  <entry>
    <title>From chain of trust to chain of timing</title>
    <link href="https://zlhk100.github.io/library/chain-of-timing.html"/>
    <id>https://zlhk100.github.io/library/chain-of-timing.html</id>
    <updated>2026-11-15T12:00:00-05:00</updated>
    <summary>A position paper for Physical AI: access, trust and time contracts at every boundary on a shared SoC, the SPDM/IDE trust chain as a template for a chain of timing, where timing breaks hop by hop, and proposals by community.</summary>
  </entry>
  <entry>
    <title>From specification to silicon: the SPDM/IDE trust chain</title>
    <link href="https://zlhk100.github.io/library/spdm-ide-trust-chain.html"/>
    <id>https://zlhk100.github.io/library/spdm-ide-trust-chain.html</id>
    <updated>2026-11-15T11:00:00-05:00</updated>
    <summary>How DICE identity, SPDM attestation and IDE link protection compose into a verifiable trust chain for confidential computing. First published on LinkedIn in April 2026; now the canonical copy, with dated corrections.</summary>
  </entry>
  <entry>
    <title>Case study: QNX and Linux cockpit displays on one PC</title>
    <link href="https://zlhk100.github.io/work/vcockpit.html"/>
    <id>https://zlhk100.github.io/work/vcockpit.html</id>
    <updated>2026-11-14T09:00:00-05:00</updated>
    <summary>vcockpit composes a QNX 8.0 instrument cluster and a Linux guest, each in its own virtual machine, into one cockpit display on a PC, so cockpit software can be built and integrated before vehicle hardware exists. Open source, with a codelab.</summary>
  </entry>
  <entry>
    <title>VirtIO and hypervisors: how virtual devices work</title>
    <link href="https://zlhk100.github.io/library/virtio-hypervisors.html"/>
    <id>https://zlhk100.github.io/library/virtio-hypervisors.html</id>
    <updated>2026-11-07T18:00:00-05:00</updated>
    <summary>A technical reference on device virtualisation: emulation, VirtIO and assignment; virtqueues and notifications; KVM with vhost, vhost-user and vDPA; Xen grants and IOREQ; QNX, L4Re and Gunyah; VFIO; pKVM and Arm CCA; and virtio-msg.</summary>
  </entry>
  <entry>
    <title>Mixed-criticality on shared silicon: a primer for Physical AI</title>
    <link href="https://zlhk100.github.io/library/mixed-criticality-primer.html"/>
    <id>https://zlhk100.github.io/library/mixed-criticality-primer.html</id>
    <updated>2026-11-09T12:00:00-05:00</updated>
    <summary>Where tail latency comes from when AI inference and real-time control share an SoC, why partial isolation falls short, two isolation architectures, measurement discipline, and the EU compliance dates that make it urgent.</summary>
  </entry>
  <entry>
    <title>Case study: TEE trusted services for an automotive telematics unit</title>
    <link href="https://zlhk100.github.io/work/tee-telematics.html"/>
    <id>https://zlhk100.github.io/work/tee-telematics.html</id>
    <updated>2026-11-09T08:45:00-05:00</updated>
    <summary>Hardened TLS, secure storage and secure logging in OP-TEE for a 5G telematics control unit, toward UN R155 and ISO/SAE 21434 compliance.</summary>
  </entry>
  <entry>
    <title>Case study: secure video streaming for in-flight entertainment</title>
    <link href="https://zlhk100.github.io/work/secure-video-ife.html"/>
    <id>https://zlhk100.github.io/work/secure-video-ife.html</id>
    <updated>2026-11-09T12:00:00-05:00</updated>
    <summary>Widevine L1 DRM and aviation-qualified secure boot on NXP i.MX8, designed for certification from day one.</summary>
  </entry>
</feed>
