CVE-2026-63864

Quoted from the kernel CNA's own scoring file, word for word. A highlighted metric carries the most severe value CVSS v3.1 defines for it, which is what pushes the score up. source

AV:LAttack vector, local

The bug is reached only through the bpf() syscall during BPF_PROG_LOAD verification (check_cfg/visit_insn), which is a local kernel interface per kernel CVSS guidance for BPF.

AC:LAttack complexity, low

An attacker can craft a BPF program using tail calls and subprograms to hit visit_tailcall_insn(), and can pressure kernel memory (parallel BPF loads/maps) to make the small jt allocation fail while verification continues.

PR:NPrivileges required, none

Exploitation occurs at BPF_PROG_LOAD time via bpf(); on systems with kernel.unprivileged_bpf_disabled=0, unprivileged users can load socket-filter programs with PROG_ARRAY maps and tail calls without CAP_BPF or root.

UI:NUser interaction, none

No victim interaction is required; the attacker triggers the flaw by loading a malicious BPF program through their own bpf() syscall.

S:UScope, unchanged

Impact is a BPF verifier bypass leading to kernel memory corruption and local privilege escalation within the same kernel security boundary, not a cross-authority escape such as VM guest-to-host.

C:HConfidentiality impact, high

Incorrect tail-call stack liveness lets verifier-approved programs perform out-of-bounds stack reads at runtime, exposing adjacent kernel stack memory and enabling arbitrary kernel memory disclosure primitives.

I:HIntegrity impact, high

The verifier bypass permits out-of-bounds stack writes and type confusion on stack slots, which can be leveraged for control-flow hijacking and arbitrary kernel code execution from BPF context.

A:HAvailability impact, high

Out-of-bounds BPF stack access can trigger kernel oops/panic during program execution, and repeated exploitation can be used for denial of service even when full exploitation is not attempted.