ib_register_device(9)

📦 version 7.3.1, in nixpkgs unstable

💅 ib_register_device — like, the Kernel API Manual girlies' favorite little registration ritual ✨

okay so bestie, you know how every low-level driver out there is just dying to be seen by the IB core? yeah. this is the function that makes it official. let's get into it 🫶

✨ SYNOPSIS ✨

int ib_register_device (struct ib_device *device , const char *name ,
       struct device *dma_device );

like, three args, one int back. no cap, that's the whole vibe 🌸

🎀 ARGUMENTS 🎀

  • device — the actual device you're registering. your main character. 👑
  • name — the device name, and it has to be unique within the device's net namespace. plot twist: it can contain a '%', and if it does, a unique index gets tacked onto the device name you passed in. cute, right? 🦋
  • dma_device — a pointer to a DMA-capable device. here's the tea ☕: if this is NULL, then the IB device itself gets used instead. and in that case the caller should fully set up the ibdev for DMA — which usually means reaching for dma_virt_ops. iykyk 💀

💖 DESCRIPTION 💖

so the low-level drivers use ib_register_device to register their devices with the IB core. literally that simple. and once a device is added, all registered clients get a callback for it — everyone's invited to the group chat 📲✨

one thing though: device must be allocated with ib_alloc_device. non-negotiable, bestie.

okay so this is the slightly unhinged part 🙈 — if the driver uses ops.dealloc_driver and calls any ib_unregister_device asynchronously, then the device pointer may become freed as soon as this function returns. like, immediately. so don't go touching it after, fr fr 😭

👑 SEE ALSO 👑

Kernel file ./drivers/infiniband/core/device.c — and the whole extended friendship circle:

mwah 💋 this lil page lives at ./drivers/infiniband/core, and it's stamped October 2026 — so if you're reading this in the future, hiii, you're doing amazing, go register that device bestie 💜🌈

📦 where this page comes from