scx_bpf_pick_any_cpu_node(9)

📦 version 7.3.1, in nixpkgs unstable

💅 scx_bpf_pick_any_cpu_node(9)

okay so real talk, this is a page straight out of the Kernel API Manual — it's scx_bpf_pick_any_cpu_node(9) and bestie, it's literally the "gimme a CPU, any CPU" function. Let's get into it ✨

🌸 SYNOPSIS 🌸

__bpf_kfunc s32 scx_bpf_pick_any_cpu_node (const struct cpumask
*cpus_allowed , s32 node , u64 flags , const struct bpf_prog_aux *aux );

that's the whole vibe, no cap 🫶

🎀 ARGUMENTS 🎀

  • cpus_allowed — the allowed cpumask, aka the guest list for your CPU party 💌
  • node — the target NUMA node, like the specific venue you're tryna book first
  • flags — your SCX_PICK_IDLE_* flags, the little switches that change how the search behaves
  • aux — the implicit BPF argument that gives you access to bpf_prog_aux, and it's hidden from BPF progs — so like, don't go looking for it, it's not for you 🙈

🔥 DESCRIPTION 🔥

So here's what it does: it picks and claims an idle cpu inside cpus_allowed. If there's literally no idle one available, it'll just pick any CPU in cpus_allowed anyway. It is guaranteed to succeed and returns the number of the picked idle cpu as long as cpus_allowed isn't empty. And if cpus_allowed is empty? You get -EBUSY back. 💀

The search starts at node and then hops to the other online NUMA nodes in order of increasing distance — unless you specified SCX_PICK_IDLE_IN_NODE, in which case the search is limited to the target node, regardless of the CPU idle state. periodt. 📍

⚠️ caveat time, bestie: if ops.update_idle is implemented and SCX_OPS_KEEP_BUILTIN_IDLE is not set, this function can't tell which CPUs are idle and will always pick any CPU. like it just gives up on the whole idle-detection thing 😭

🦋 SEE ALSO 🦋

Kernel file ./kernel/sched/ext/idle.c, plus the whole friend group:

and that's the page!! 🧚 this one ships from ./kernel/sched/ext, dated October 2026 — catch you on the next manpage bestie, mwah 💖👑

📦 where this page comes from