amdgpu_devic...r_accessible(9)

📦 version 7.3.1, in nixpkgs unstable

💖 NAME 💖

okay so first things first, this page is amdgpu_devic...r_accessible(9) from the Kernel API Manual, and it's literally about amdgpu_device_is_peer_accessible — the vibe check that tells you whether one GPU can reach another through the PCIe BAR. bestie, it's a peer-access checker, that's the whole personality 🦋

✨ SYNOPSIS ✨

bool amdgpu_device_is_peer_accessible (struct amdgpu_device *adev ,
struct amdgpu_device *peer_adev );

like, it takes two amdgpu devices and hands you back a bool. no cap. 💅

🎀 ARGUMENTS 🎀

  • adev — an amdgpu_device pointer. this is the one being peeked at, the main character of the question 👑
  • peer_adev — an amdgpu_device pointer too, but for the peer device, aka the one that's trying to access adev. it's the one knocking on the door 🙈

🌸 DESCRIPTION 🌸

so the way this works is, it returns true if peer_adev can access (DMA) adev through the PCIe BAR. that's the entire deal, fr fr.

and the conditions? both of them, not optional:

  • adev has to be "large BAR", and
  • the BAR has to match the DMA mask of peer_adev.

both boxes ticked or it's not giving accessible, periodt. ⭐

🔥 SEE ALSO 🔥

okay so this is the part where the manpage goes absolutely feral with the cross-references and honestly? iconic. every single one of these is real and you should look them up iykyk:

  • the kernel file itself: ./drivers/gpu/drm/amd/amdgpu/amdgpu_device.c

and then the whole friend group:

like i said. a whole friend group. ate and left no crumbs 🌈

and that's the page, bestie 🫶 — signed off from ./drivers/gpu/drm/amd/amdgpu, dated October 2026, still answering to the name amdgpu_devic...r_accessible(9). go check on your peers, they're probably accessible. mwah 💜✨

📦 where this page comes from