GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux KVM SEV: sub-page command buffer on SNP hosts triggers RMP faults and host panic

UnscoredCVE-2026-80853Kernel, userspace & hypervisorcurated

Impact

When KVM {de,en}crypts memory of an SEV or SEV-ES guest on an SNP-enabled host, it used a sub-page temporary buffer, so the page handed to AMD firmware also held unrelated kernel allocations. Transferring a page to Firmware ownership makes it inaccessible to software, so any other kernel code touching the rest of that page takes an RMP violation #PF - the bug report shows a panic arriving through an unrelated getxattr path. The result is a full host panic on a confidential-computing hypervisor: every guest on the box dies at once, and on a GPU host that is every tenant sharing those GPUs plus a cold reboot rather than a drain. This is a host stability failure driven by normal SEV guest lifecycle operations, not an escape.

Who can reach it

Local. Triggered by KVM servicing SEV/SEV-ES memory encrypt/decrypt commands on an SNP-enabled AMD host; a tenant able to run or manipulate an SEV guest can drive the code path, but the fault lands on whichever kernel allocation happens to share the page, so timing decides whether it panics. No remote or unauthenticated path.

What to do

Apply the stable kernel fix (linked commits) and reboot each affected hypervisor node - this is a kernel change, so drain the node and reboot; there is no runtime toggle. The record names no fixed release version, only the commits. Short of patching, the only avoidance is not running SEV/SEV-ES guests on SNP-enabled hosts.

References

Related entries

All Kernel, userspace & hypervisor entries

This entry is curated: imported from vendor advisories with machine assistance, not yet individually verified. Confirm against your vendor's advisory before acting, and report anything wrong.