The verified BPF loader performs post‑verification instruction rewrites that can take O(n²) time. A privileged process can submit a program that contains 131,072 no‑op jumps followed by a return. Verification completes quickly, but the rewrite passes then iterate over each no‑op, allocating and moving auxiliary data. Because this work never checks for signals or reschedules, a SIGKILL sent after load cannot terminate the process until all rewrites finish, effectively blocking the kernel from responding to user requests and potentially causing a denial of service.

The flaw exists in the Linux kernel BPF subsystem, affecting any system that allows privileged users to load eBPF programs. All kernel versions that have not applied the patch in the advisory are vulnerable. The issue is limited to the verifier path; kernel builds that disable BPF or restrict verifiable programs to unprivileged contexts are not impacted.

No CVSS score is currently published, and the EPSS score is unavailable. The vulnerability is listed as not in CISA KEV. Because the attack requires privileged access to load a malformed BPF program, the exploitability is restricted to systems with users who can run privileged bpf_prog_load calls. Nevertheless, the high runtime cost and lack of preemption can be leveraged by an attacker holding such privileges to exhaust CPU cycles and prevent task termination, leading to a denial‑of‑service condition.