GPU VulnDB

Database/Kernel, userspace & hypervisor

Linux kernel BPF verifier: byte swap keeps a stale register link, letting programs pass with out-of-bounds access

CVSS 7.8CVE-2026-43070Kernel, userspace & hypervisorcurated

Impact

The verifier does not break the scalar ID link between two registers when one of them is byte-swapped with BPF_END. Bounds it learns from a check on the swapped register are then wrongly applied to the linked register. That gives false confidence in the register's value and can let a program that accesses memory out of bounds pass verification. Bugs in this class are a common route from BPF load rights to kernel memory corruption on the host. The real exposure depends on who on the node is allowed to load BPF programs.

Who can reach it

Local user or container that can load BPF programs. Most distributions disable unprivileged BPF by default, so in practice this means holders of CAP_BPF or CAP_SYS_ADMIN, such as privileged pods and observability or networking agents. The default container seccomp profile blocks bpf() for ordinary tenant pods.

What to do

Install a kernel that includes the upstream stable fix, then drain and reboot each node. Until then, keep kernel.unprivileged_bpf_disabled set and do not grant CAP_BPF or CAP_SYS_ADMIN to tenant workloads.

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.