scx_bpf_select_cpu_dfl(9)

📦 version 7.3.1, in nixpkgs unstable

💅 scx_bpf_select_cpu_dfl(9) — like, the default CPU picker, bestie

okay so this is a Kernel API Manual page, section 9 (kernel API, iykyk), and it's about scx_bpf_select_cpu_dfl — literally the built-in, off-the-shelf version of ops.select_cpu(). if you're writing an sched_ext scheduler and you don't wanna hand-roll your own CPU picking logic, this is your girl 🎀

✨ NAME ✨

scx_bpf_select_cpu_dfl — the default implementation of ops.select_cpu(), no cap 🫶

✨ SYNOPSIS ✨

__bpf_kfunc s32 scx_bpf_select_cpu_dfl (struct task_struct *p , s32
prev_cpu , u64 wake_flags , bool *is_idle , const struct bpf_prog_aux
*aux );

✨ ARGUMENTS ✨

  • p — the task_struct you're picking a CPU for. the main character, honestly 👑
  • prev_cpu — the CPU that p was hanging out on before. its last known location 🗺️
  • wake_flags — the SCX_WAKE_* flags. the vibe check for the wakeup 🌸
  • is_idle — an out parameter, bestie. it tells you whether the CPU that got returned is idle — so like, is it a good candidate for direct dispatching or not 🤔
  • aux — an implicit BPF argument used to access bpf_prog_aux, and it's hidden from BPF progs 🙈 don't worry about it, it's not for you

✨ DESCRIPTION ✨

so basically you can call this from ops.select_cpu, from ops.enqueue, or even from an unlocked context like a BPF test_run call — as long as built-in CPU selection is enabled. and "enabled" here means: ops.update_idle is missing, or SCX_OPS_KEEP_BUILTIN_IDLE is set. one of those two, periodt ✨

it returns the picked CPU, and *is_idle tells you whether that picked CPU is currently idle — which lowkey makes it a great candidate for direct dispatching 🔥

✨ SEE ALSO ✨

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

anyway that's the whole page, mwah 💋 this one came from ./kernel/sched/ext and it's dated October 2026 — go pick some CPUs, bestie, you got this 🦋💖

📦 where this page comes from