Pre-flight checklist
Upstream / vanilla kernel check
I was unable to test with a vanilla/upstream kernel
Kernel variant
linux-cachyos (EEVDF, Clang)
System information (cachyos-bugreport.sh)
Manual system information (if cachyos-bugreport.sh is unavailable)
Hardware / System
- Laptop: Framework (Alder Lake-P iGPU present)
- eGPU enclosure: Razer Core X (Thunderbolt 3)
- GPU: AMD RX 6800 XT (Navi 21, PCI ID `1002:73bf`)
- Thunderbolt authorization: `boltctl` shows authorized at 40 Gb/s
Please fill in the following (I can provide exact command outputs if needed):
- Kernel version (`uname -r`): 6.19.10-1
- CPU model: 12th Gen Intel(R) Core(TM) i7-1260P
- GPU model & driver:
- iGPU: Intel Corporation Alder Lake-P Integrated Graphics Controller (rev 0c): i915
- eGPU: AMD RX 6800 XT (Navi 21, `1002:73bf`), driver: amdgpu
- Architecture (e.g., `x86-64`, `x86-64-v3`, `x86-64-v4`, `znver4`): x86_64
- Active SCX scheduler (if any): none
Bug description
After upgrading to linux-cachyos 6.19.10-1, my Thunderbolt eGPU becomes unusable. The AMD GPU initializes, then the PCIe hotplug port reports Link Down and Card not present, followed by amdgpu: device lost from bus!. The device ends up unbound (no “Kernel driver in use: amdgpu” in lspci).
Downgrading the kernel to linux-cachyos 6.19.6-2 immediately restores correct behavior.
Steps to reproduce
- Connect and authorize Razer Core X eGPU
- Boot linux-cachyos 6.19.10-1
- Observe that the external monitor on the eGPU stays black / eGPU disappears
Expected behavior
lspci -nnk -s 81:00.0 shows Kernel driver in use: amdgpu, eGPU remains present, and Vulkan enumerates the AMD GPU
Actual behavior
- pciehp: Slot(6): Link Down
- pciehp: Slot(6): Card not present
- amdgpu 0000:81:00.0: device lost from bus!
Logs / stack traces
1383:Apr 04 11:43:18 cachyos-x8664 kernel: pcieport 0000:00:07.3: pciehp: Slot(6): Link Down
1384:Apr 04 11:43:18 cachyos-x8664 kernel: pcieport 0000:00:07.3: pciehp: Slot(6): Card not present
1398:Apr 04 11:43:18 cachyos-x8664 kernel: amdgpu 0000:81:00.0: amdgpu: amdgpu: finishing device.
1399:Apr 04 11:43:18 cachyos-x8664 kernel: amdgpu 0000:81:00.0: amdgpu: device lost from bus!
...
1421:Apr 04 11:43:19 cachyos-x8664 kernel: amdgpu 0000:81:00.0: amdgpu: device lost from bus!
Additional system information
No response
Pre-flight checklist
Upstream / vanilla kernel check
I was unable to test with a vanilla/upstream kernel
Kernel variant
linux-cachyos (EEVDF, Clang)
System information (cachyos-bugreport.sh)
Manual system information (if cachyos-bugreport.sh is unavailable)
Bug description
After upgrading to
linux-cachyos 6.19.10-1, my Thunderbolt eGPU becomes unusable. The AMD GPU initializes, then the PCIe hotplug port reportsLink DownandCard not present, followed byamdgpu: device lost from bus!. The device ends up unbound (no “Kernel driver in use: amdgpu” inlspci).Downgrading the kernel to
linux-cachyos 6.19.6-2immediately restores correct behavior.Steps to reproduce
Expected behavior
lspci -nnk -s 81:00.0 shows Kernel driver in use: amdgpu, eGPU remains present, and Vulkan enumerates the AMD GPU
Actual behavior
Logs / stack traces
Additional system information
No response