nx_cgroup_v2_probe.nx

nx_cgroup_v2_probe.nx -- Linux cgroups v2 host detection. Stage A foundation per NISHI_ELDER_AI_OFF_DOCKER_2026_05_20.md. To enforce nx_budget ceilings via the host kernel (memory.max, pids.max, cpu.weight), the substrate must FIRST detect whether cgroups v2 is available on the host. This primitive is the inspection-only foundation; mutation primitives (mkdir cgroup, write memory.max, attach pid) ship in a later session ON A LINUX HOST where their smokes can actually run. Detection works by attempting to read /sys/fs/cgroup/cgroup.controllers: - File exists + readable: cgroups v2 is mounted; returns 1 - File missing / unreadable: cgroups v2 unavailable; returns 0 Host coverage: Linux with cgroup v2 -> available()=1 Linux with cgroup v1 -> available()=0 (no v2 root) Cygwin / Windows-mounted bash -> available()=0 (no /sys at all) macOS -> available()=0 (no cgroups) qemu-user-static -> passthrough to host kernel; if host has cgroups v2, returns 1 HONEST scope this V1: - Read-only probe; no mutation - Single canonical file check (cgroup.controllers) - No delegated-controllers parsing (queued) - No process-self-cgroup parsing (queued; reads /proc/self/cgroup) - No mutation primitives (queued for Linux-host session) genealogy_id: linux_cgroup_v2_2018 + systemd_cgroup_delegation_2017 + cardinal_2026-05-20_elder_ai_off_docker lineage_id: substrate_cgroup_v2_probe_v1 nx_capability_manifest: variant_class: cgroup_v2_probe variant_id: cgroup_v2_probe_v1_read_only requires_isa: [rv64imac, x86_64, aarch64] requires_syscalls: [openat, read, close, mmap] requires_ram_min_b: 4096 tier_floor: NX_TIER_INF_MOBILE