GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux i40e: unvalidated TC index in a VF config-queues message reads past vf->ch[]

CVSS 8.8CVE-2025-39971Kernel, userspace & hypervisorcurated

Impact

i40e_vc_config_queues_msg() iterated vf->ch[idx] without checking idx against the number of active or initialized traffic classes, so a virtual function could drive the PF driver to index outside that array using a value it supplies in a virtchnl message. The attacker side here is a VF - a tenant VM or container holding a passed-through i40e virtual function - and the corrupted side is the host PF driver, which is exactly the trust boundary SR-IOV is supposed to hold. On a GPU node, i40e VFs are a common way to give tenants line-rate access to the data network, and a host-side PF driver fault takes the node's networking (and the node) with it. The CVSS vector marks changed scope, consistent with a guest reaching host state.

Who can reach it

A local guest or container that owns an i40e virtual function and can send virtchnl messages to the PF. No host privilege needed; requires that SR-IOV VFs on Intel X710/XL710-class NICs be handed to untrusted tenants.

What to do

Update to a stable kernel carrying the idx range check and reboot each affected node - a driver fix with no runtime toggle, so drain and reboot. Interim mitigation is to stop assigning i40e VFs to untrusted tenants on unpatched 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.