maybe_create_worker(9)

📦 version 7.3.1, in nixpkgs unstable

🌸 maybe_create_worker(9) — kernel API manual, bestie edition 🌸

okay so real talk, this is a page from the Kernel API Manual and it's genuinely one of those little functions that's doing SO much behind the scenes. it's giving main character energy in kernel/workqueue.c. let's get into it 💅

✨ NAME ✨

maybe_create_worker — literally spawns a new worker when it's necessary, and only when it's necessary 😌

✨ SYNOPSIS ✨

void maybe_create_worker (struct worker_pool *pool );

✨ ARGUMENTS ✨

  • pool — the pool you want a fresh new worker for 🫶 that's it, that's the whole argument list, we love a one-arg icon

✨ DESCRIPTION ✨

so the whole vibe: create a new worker for pool if necessary. that "if necessary" is doing a lot of heavy lifting, so listen up 👑

on return from this function, pool is guaranteed to have at least one idle worker. no cap, that's the promise. 🎀

but here's the dramatic part 😭 — if creating that new worker takes longer than MAYDAY_INTERVAL, a mayday gets sent out to all rescuers that have works scheduled on pool, so they can go resolve a possible allocation deadlock. yes, an actual mayday. this function has a distress signal and honestly? iconic. 🚨

and on return, two things are guaranteed:

  • need_to_create_worker is false
  • may_start_working is true

both of those, every time. periodt. ✨

LOCKING: raw_spin_lock_irq(pool->lock) — and heads up, that lock may be released and regrabbed multiple times during this. also it does GFP_KERNEL allocations. and the big one: called only from manager. so don't go calling it from random places, bestie, that's not your lane 🙈

✨ SEE ALSO ✨

go check out the kernel file ./kernel/workqueue.c, and honestly just take the whole friend group while you're at it:

that's the whole page, bestie, no notes 🦋 and for the record this came out of the ./kernel source, dated October 2026, under the Kernel API Manual. go off, workqueue 💜✨

📦 where this page comes from