GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux kernel BPF verifier: regsafe() wrongly prunes states when old range is BEYOND_PKT_END

CVSS 7.8CVE-2026-43030Kernel, userspace & hypervisorcurated

Impact

When the old register state has range == BEYOND_PKT_END and the current state has a finite range, regsafe() could return true, so the verifier prunes a state with a valid packet range instead of exploring it. State pruning that treats an unexplored path as equivalent is how a program with unsafe packet accesses gets accepted, which puts packet-pointer reads and writes outside their verified bounds. This is the verifier's own safety logic, so it affects every host that lets a workload load BPF programs - including the GPU nodes running BPF-based CNI and observability. The record gives only the one-line fix and no exploit detail; treat the exposure as unsafe packet access rather than a demonstrated escalation chain.

Who can reach it

Local user able to load a BPF program touching packet pointers, which in practice means CAP_BPF/CAP_SYS_ADMIN or a network-program loader exposed to the workload. Not remotely reachable on its own.

What to do

Update the host kernel to a build carrying the stable fix commits below and reboot each node. Until patched, do not grant CAP_BPF or CAP_NET_ADMIN to tenant workloads and keep unprivileged BPF disabled. No vendor advisory beyond the kernel commits is present in the record.

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.