intel_wakeref_auto(9)

📦 version 7.3.1, in nixpkgs unstable

💅 intel_wakeref_auto(9) 💅

okay so this is a page straight out of the Kernel API Manual, and bestie, it's a tiny one but it matters. Buckle up 🎀

✨ NAME ✨

intel_wakeref_auto — delay the runtime-pm autosuspend 🦋

✨ SYNOPSIS ✨

void intel_wakeref_auto (struct intel_wakeref_auto *wf , unsigned long
timeout );

like, that's the whole vibe. two args, no cap ✨

✨ ARGUMENTS ✨

  • wf — the wakeref 💖 that's it, that's the description, she's literally just the wakeref.
  • timeout — the relative timeout, measured in jiffies ⏳ (yes, jiffies, not seconds, don't come for me)

✨ DESCRIPTION ✨

okay so here's the tea ☕ the runtime-pm core does this thing where, after the last wakeref gets released, it waits a lil' suspend delay before it actually triggers runtime suspend of the device. and that delay? it's configurable via sysfs, with, like, little regard to the device characteristics 🙈 which is lowkey not it.

so instead, the plan is: tune the autosuspend based on our HW knowledge 👑 we know our hardware, so we should get a say, right? and that's exactly what intel_wakeref_auto does — it delays the sleep by the supplied timeout. slay 💅

and here's the iconic part: pass timeout = 0 to cancel a previous autosuspend by executing the suspend immediately ⚡ not me cancelling my own nap, honestly. zero means go, periodt 🔥

✨ SEE ALSO ✨

the kernel file ./drivers/gpu/drm/i915/intel_wakeref.h — go read her, she's the main character 📖 and also all her friends:

like, the whole friend group 🫶

ngl the vibes here are immaculate but remember: timeout is in jiffies, and 0 is not "no timeout", it's "do it now". don't get it twisted 💀

and that's the page, bestie!! 🌸 brought to you by the good people at ./drivers/gpu/drm/i915, dated October 2026, and yes this is intel_wakeref_auto(9) signing off — mwah 💜🧚✨

📦 where this page comes from