GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux ccp/SEV: zero-length cert chain in PDH export hits a BUG_ON on CONFIG_DEBUG_VIRTUAL kernels

CVE-2026-80652Kernel, userspace & hypervisorcurated

Impact

The SEV PDH export ioctl on /dev/sev did not treat a zero-length userspace cert-chain buffer as a length query, so it took the physical address of ZERO_SIZE_PTR. On kernels built with CONFIG_DEBUG_VIRTUAL=y that hits a BUG_ON in __phys_addr and takes the host down - a syzkaller-reproduced local panic from a single ioctl. The commit message states the bug is benign on ordinary production kernels, where the address computation is just arithmetic and the PSP errors out before writing, so the exposure is limited to debug-configured builds. On a confidential-computing GPU host a panic here kills every guest on the node, and SEV hosts are exactly the ones you cannot cheaply drain.

Who can reach it

Local, from any process that can open /dev/sev and issue the PDH export ioctl. On a correctly permissioned host that is the VMM or a management daemon rather than a tenant; authentication is whatever guards that device node. Only kernels built with CONFIG_DEBUG_VIRTUAL=y are known to panic.

What to do

Fixed in the mainline Linux kernel ccp/PSP driver; take the distro kernel update and reboot the host. SEV/PSP paths are not meaningfully live-patchable, so this is a drain-and-reboot for confidential-VM nodes. Interim mitigation on unpatched hosts: confirm CONFIG_DEBUG_VIRTUAL is off in your production kernel config and audit which accounts hold /dev/sev. The record names no fixed release version, only the stable commits.

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.