When using any external device over Thunderbolt (Linux kernel 5.0.9), it simple crashes after a while: xhci_hcd 0000:1f:00.0: xHCI host controller not responding, assume dead
When using the same device over USB 3.0, it works without any issues.
I already tried blacklisting any power management (using TLP blacklisting + udev), but this doesn't prevent the crashes from happening.
Any other workarounds I haven't tried? Firmware is up-to-date (flashed on Windows 10) and BIOS as well.
Hi Alberto, thanks for your reply.
What is the model of the NUC? NUC7i7BNH
What is the model of the monitor? Non - headless computer
What is the video port the monitor is using? N/A
What type of video cable are you using? N/A
Are you using adapters, converters or receivers? USB 3.1 C
Just to make sure, the current OS installed in the NUC is it Linux or Windows*? Linux 5.0.9
I have updated the BIOS and firmware(s) to the latest version you provided over Windows 10 (UEFI) by using a separated drive.
Both are reporting the latest versions are successfully installed.
I'm running Arch Linux (kernel 5.0.9) - Intel doesn't seem to support fwupdate (yet) - that's why I have used Windows 10 to upgrade the firmware.
The peripheral is an ICY BOX IB-3740-C31 and works fine when being connected over USB 3.0, but really slow and thus crashing over USB 3.1/Thunderbolt.
If more debug info is needed, please let me know.
On both kernel versions I have crashes and slow file operations, the actual filesystem didn't make a differ: XFS, ext4 and ZFS. The same device works perfectly when being connected over USB 3.0.
It really seems the controller is dead, I need to shutdown the NUC (power-cycle) to recover it from it's deadlock.
For some reason it also locks the connected device, so I need to power-cycle this as well.
The same issue occured with this device: https://www.startech.com/HDD/Enclosures/USB-3-0-Enclosure-for-3-5-SATA-Hard-Drives-with-UASP~S3510BM...
All being connected over PCI-e/Thunderbolt, when using the same device(s) over USB 3.0, no issues and performance lost occurs, although I only receive half amount of (possible) speed.
Thanks for investigating this issue.