iommufd_device_detach(9)

📦 version 7.3.1, in nixpkgs unstable

💅 iommufd_device_detach — kernel API manual, but make it make sense

okay so bestie, we're in the Kernel API Manual today and the star of the show is iommufd_device_detach(9). yes it's a kernel function, no it's not scary, i promise 🫶

✨ NAME ✨

iommufd_device_detach — disconnect a device/pasid from an iommu_domain. that's the whole vibe: it's the breakup function 💔

🎀 SYNOPSIS 🎀

void iommufd_device_detach (struct iommufd_device *idev , ioasid_t pasid );

literally returns void, so no "did it work?" drama. it just does the thing. slay.

🦋 ARGUMENTS 🦋

  • idev — the device to detach. this is the one we're breaking up with 👋
  • pasid — the pasid to detach. also getting the goodbye text, no cap.

🌸 DESCRIPTION 🌸

so this is the undo button for iommufd_device_attach — it literally reverses it. it disconnects the idev from the pt_id it was previously attached to, and after that the device goes right back to a blocked DMA translation. like, it's grounded. no DMA for you until you get attached again 🚫

and real talk: pasid is IOMMU_NO_PASID when there's no pasid usage happening. iykyk, that's the no-pasid mood.

👑 SEE ALSO 👑

the kernel file lives at ./drivers/iommu/iommufd/device.c and honestly it's the main character of this section. go read it. then go read its friends:

that's the whole group chat, fr fr 💬

⚠️ quick vibe check: this is the undo of iommufd_device_attach, and once you detach, the device is back to blocked DMA translation. that's the documented behaviour — don't come for me, i'm just the messenger 💀

and that's a wrap, bestie 🎀 this page came from the source/version line ./drivers/iommu/iommufd and it's dated October 2026, so like… future-coded of it honestly ✨ periodt.

📦 where this page comes from