drm_sched_start(9)

📦 version 7.3.1, in nixpkgs unstable

✨ drm_sched_start(9) ✨

okay so this is a page from the Kernel API Manual, and it's about drm_sched_start — the lil glow-up routine your GPU scheduler runs after a reset 💅

🌸 NAME 🌸

drm_sched_start — recover jobs after a reset 🫶

🎀 SYNOPSIS 🎀

void drm_sched_start (struct drm_gpu_scheduler *sched , int errno );

🦋 ARGUMENTS 🦋

  • sched — the scheduler instance, bestie. like, which scheduler are we reviving ✨
  • errno — the error to set on the pending fences. yeah, you gotta tell 'em what went wrong 😭

💖 DESCRIPTION 💖

So here's the tea ☕: this function is typically used for reset recovery — peep the docu of drm_sched_backend_ops.timedout_job for the full details, iykyk.

🚨 Do NOT call it for scheduler startup. Not the vibe. The scheduler is already fully operational after drm_sched_init succeeded — it does not need you to come in and start it like that. periodt.

And because it's only</> for reset recovery, drivers must not call this function in their struct drm_sched_backend_ops.timedout_job callback when they skip a reset using enum drm_gpu_sched_stat.DRM_GPU_SCHED_STAT_NO_HANG. like, if you're skipping the reset, you're skipping the whole thing — no sneaky drm_sched_start on the way out 🙈

🌈 SEE ALSO 🌈

the whole friend group, fr fr 👑

and that's the whole page, bestie 🍓 this one lives at ./drivers/gpu/drm/scheduler, dated October 2026 — mwah, go recover some jobs 💜🧚

📦 where this page comes from