CVE-2026-93147

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 malformed input is eBPF bytecode containing BPF_JMP|BPF_CALL with imm=BPF_FUNC_get_smp_processor_id, loaded via bpf_prog_load() (bpf(2) BPF_PROG_LOAD). s390 bpf_jit_insn() then inlines that call as LY from get_lowcore()->cpu_nr; no remote protocol carries this bytecode.

AC:LAttack complexity, low

The attacker sets BPF_REG_0 (s390 %r14) with BPF_LD_IMM64, then emits the helper call; bpf_jit_insn() always compiles it to LY (E358), which writes only bits 32-63 and leaves the attacker-chosen high half intact. bpf_opt_remove_dead_code() does not delete that store, and there is no race.

PR:LPrivileges required, low

bpf_prog_load() allows BPF_PROG_TYPE_SOCKET_FILTER; sk_filter_func_proto() reaches bpf_base_func_proto(), which returns bpf_get_raw_smp_processor_id_proto for BPF_FUNC_get_smp_processor_id before its CAP_BPF gate, and BPF_MAP_TYPE_ARRAY is an unprivileged map. CAP_BPF is also granted by bpf_token_capable() in a delegated user namespace.

UI:NUser interaction, none

The attacker loads the crafted program with bpf(BPF_PROG_LOAD) and executes it themselves via BPF_PROG_TEST_RUN (sk_filter_prog_ops.test_run = bpf_prog_test_run_skb) or SO_ATTACH_BPF on their own socket; no other user must mount, open, or otherwise act.

S:UScope, unchanged

LY leaving stale high bits in BPF_REG_0 lets the JITed program read and write the host kernel through a verifier-approved PTR_TO_MAP_VALUE. That stays inside the same kernel authority; it is not a KVM guest-to-host or IOMMU bypass.

C:HConfidentiality impact, high

do_refine_retval_range() sets R0’s 64-bit range to [0, nr_cpu_ids-1] after BPF_FUNC_get_smp_processor_id, so check_map_access() allows adding it to a map-value pointer. LY preserves attacker high bits, so the subsequent BPF_LDX reads kernel memory far outside the map.

I:HIntegrity impact, high

The same forged pointer is a BPF_STX destination: the store lands at map_value + ((old_r0 & 0xFFFFFFFF00000000) | cpu_nr). That is an attacker-controlled kernel write usable for corruption and control-flow hijack.

A:HAvailability impact, high

Using that out-of-bounds map-value pointer (map_value plus the uncleared high 32 bits of %r14 after LY of cpu_nr) faults on unmapped addresses and oopses or panics the kernel.