v3d_submit_csd_ioctl(9)

📦 version 7.3.1, in nixpkgs unstable

💅 v3d_submit_csd_ioctl(9) — the Kernel API Manual page, bestie

Okay so picture this: it's the Kernel API Manual and the page is literally v3d_submit_csd_ioctl(9) — yes the (9) is giving kernel-section energy, we love that for us. She submits a CSD (compute shader) job to the V3D. That's the whole name, no notes. 🎀

✨ SYNOPSIS ✨

int v3d_submit_csd_ioctl (struct drm_device *dev , void *data , struct
drm_file *file_priv );

That's the whole vibe check: she takes three things and hands you back an int. Iconic, minimal, periodt. 🦋

🌸 ARGUMENTS 🌸

  • dev — the DRM device. Like, the actual graphics device you're talking to. Main character.
  • data — the ioctl argument. The payload, the little gift bag of stuff coming from userspace.
  • file_priv — the DRM file for this fd. Your file descriptor's whole personality in one struct, iykyk.

🔥 DESCRIPTION 🔥

So here's the tea: userspace hands over the register setup for the CSD. That's it — that's the ask.

And do we validate it? Bestie. No. We don't need to validate it, because the CSD is behind the MMU. 🛡️ The MMU's got her, she's protected, we're chilling. So userspace just provides the register setup and we're not out here double-checking it, no cap.

Lowkey important to clock: the "we don't need to validate" is genuinely the stated reason here — the CSD sits behind the MMU. Not me making excuses, that's literally the manpage. 💀

👑 SEE ALSO 👑

Kernel file ./drivers/gpu/drm/v3d/v3d_submit.c — that's the source of truth, go read her.

And the whole friend group of related calls:

Squad's all here, they ate. 🫶

🧚 cute little sign-off 🧚

And that's the page!! Source/version corner, as promised: she lives in ./drivers/gpu/drm/v3d, she's dated October 2026, and the version string is just her own name back at you — v3d_submit_csd_ioctl(9). 🌈 Full circle, we love a self-referential queen. Mwah, go submit some compute shaders responsibly 💖✨

📦 where this page comes from