<feed xmlns='http://www.w3.org/2005/Atom'>
<title>open-iov.git/docs/05-operations/05-gpubreach-audit.md, branch master</title>
<subtitle>Open-IOV vGPU-IO: host-side GSP emulation for unmodified NVIDIA guest drivers
</subtitle>
<id>https://git.vgpu.io/open-iov.git/atom?h=master</id>
<link rel='self' href='https://git.vgpu.io/open-iov.git/atom?h=master'/>
<link rel='alternate' type='text/html' href='https://git.vgpu.io/open-iov.git/'/>
<updated>2026-10-21T07:29:12+00:00</updated>
<entry>
<title>docs: GPUBreach-class audit and threat model for the host write paths</title>
<updated>2026-10-21T07:29:12+00:00</updated>
<author>
<name>Arthur Hanson Rasmusson</name>
<email>arthur@vgpu.io</email>
</author>
<published>2026-10-21T07:29:12+00:00</published>
<link rel='alternate' type='text/html' href='https://git.vgpu.io/open-iov.git/commit/?id=0b307461d22a7bfa85d4ec87b80a0e5205fc3cd6'/>
<id>urn:sha1:0b307461d22a7bfa85d4ec87b80a0e5205fc3cd6</id>
<content type='text'>
Audit of every host-side write whose target derives from guest data,
against the GPUBreach (IEEE S&amp;P 2026) chain: GPU PTE corruption -&gt;
arbitrary device read/write -&gt; driver bugs -&gt; CPU root.

Findings: the host is structurally off that path (guest GPU page tables
host-emulated, physical GPU PTEs host-owned, every write aperture-bounded
by vfio_dma_rw to guest RAM or the vidmem shadow, no guest DMA to host
buffers), and the remaining wrong-but-valid write class is bounded by the
per-page resolution + refusal guards (FILL-TRUNC, FILL-PT-REFUSED,
LOWFILL-REFUSED, GRFILL-DROP, stalled-VA demand, sema_copy_dst_pin_guard,
DST-DIVERGE). Documents residual risks, what is not claimed (no isolation
or side-channel claim; physical Rowhammer is out of scope), and three
follow-ups.
</content>
</entry>
</feed>
