linux
/proc and /sys/... readers exposed as a Rust builtin. Linux-only — the table is registered empty
on macOS and Windows. Plan 18 / v0.15.1.
Backed by the procfs crate; numbers are bytes (kB fields are converted in Rust so callers don't
have to).
Host
linux.kernel()→{version, hostname, os_release, btime}linux.uptime()→{uptime_secs, idle_secs}linux.loadavg()→{one, five, fifteen, running, total, last_pid}linux.meminfo()→{total, free, available, buffers, cached, swap_total, swap_free, …}linux.netdev()→ per-interface RX/TX counterslinux.diskstats()→ per-block-device IO stats
CPU
linux.cpu_stat()→ aggregate/proc/statfirst row, jiffies ({user, nice, system, idle, iowait, irq, softirq, steal, guest, guest_nice})linux.cpu_stat_per_core()→ list of per-CPU rowslinux.cpu_percent(prev, curr)→{total_pct}— Lua-side delta math, no kernel call
Per-process
linux.proc_stat(pid)→/proc/<pid>/stat(state, ppid, utime, stime, vsize, rss_pages, num_threads, …)linux.proc_status(pid)→/proc/<pid>/status(name, uid, gid, vm_rss, vm_size, threads, …)
Errors surface as linux.X: … runtime errors. The process.list builtin gives you a
names-and-cmdlines view; combine it with linux.proc_stat for a top-style readout.