[Bug 221422] New: xhci_hcd host controller dies on Intel NUC8i5BEH with USB audio interfaces (POD HD500X, Mbox 2)

[Bug 221422] New: xhci_hcd host controller dies on Intel NUC8i5BEH with USB audio interfaces (POD HD500X, Mbox 2)

[bugzilla-daemon]

https://bugzilla.kernel.org/show_bug.cgi?id=221422

            Bug ID: 221422
           Summary: xhci_hcd host controller dies on Intel NUC8i5BEH with
                    USB audio interfaces (POD HD500X, Mbox 2)
           Product: Drivers
           Version: 2.5
          Hardware: All
                OS: Linux
            Status: NEW
          Severity: normal
          Priority: P3
         Component: USB
          Assignee: drivers_usb@kernel-bugs.kernel.org
          Reporter: hmoller@gmail.com
        Regression: No

Hardware:
- Intel NUC8i5BEH (Bean Canyon)
- BIOS versions tested: 0092, 0095, 0098
- CPU: Intel i5-8259U

Affected USB audio devices:
- Line 6 POD HD500X (driver: snd_usb_podhd)
- Digidesign Mbox 2

Kernel:
- Reproduced on Ubuntu 24.04 kernels 6.8, 6.14, 6.17
- Same Linux kernel works correctly on another machine (ASUS Zenbook 8th gen
Intel) with the same USB audio devices

Problem:
On this NUC, USB audio streaming eventually causes xHCI host controller
failure.
Sometimes it also happens on device disconnect.

Typical sequence:
- device enumerates correctly
- driver attaches successfully
- after some time with active audio, or on disconnect, I get:
  - timeout: still 16 active urbs..
  - xhci_hcd: xHCI host controller not responding
  - xhci_hcd: HC died; cleaning up
  - USB disconnect of the audio device

Important observation:
This is not limited to one interface:
- Line 6 POD HD500X
- Digidesign Mbox 2

Both interfaces work correctly on another laptop with the same Linux kernel, so
this appears to be platform-specific to the NUC8i5BEH/xHCI implementation.

What I already tested:
- multiple USB cables
- direct connection and hub
- USB-A and USB-C ports
- multiple kernels (6.8 / 6.14 / 6.17)
- BIOS 0092, 0095 and 0098
- intel_iommu=off
- processor.max_cstate=1 intel_idle.max_cstate=1
- different audio buffer settings (larger buffer (64x3) minimice or disapear
(64x5) problem)
- reproduced both with active audio streaming and on disconnect

Expected behavior:
Even if the device/stream fails, the xHCI host controller should not die
completely.

Observed behavior:
The host controller dies and all devices on that bus disconnect.

To make them work again need to:
echo -n "0000:c2:00.4" | sudo tee /sys/bus/pci/drivers/xhci_hcd/unbind
echo -n "0000:c2:00.4" | sudo tee /sys/bus/pci/drivers/xhci_hcd/bind

Same kernel, same USB audio devices, different hardware:
works on ASUS Zenbook i5 8th gen, fails on Intel NUC8i5BEH.
This may be a hardware/firmware-specific xHCI issue on the NUC platform rather
than a generic kernel regression, but I am reporting it here because the kernel
might still be able to add a quirk or workaround.

$uname -a
Linux nuc8i5beh 6.17.0-22-generic #22~24.04.1-Ubuntu SMP PREEMPT_DYNAMIC Thu
Mar 26 15:25:54 UTC 2 x86_64 x86_64 x86_64 GNU/Linux

$ sudo dmidecode -s bios-version
BECFL357.86A.0098.2026.0204.1428

$ cat /proc/cmdline
BOOT_IMAGE=/boot/vmlinuz-6.17.0-22-generic
root=UUID=b116f7d7-5bb8-44d5-a83c-527892de0889 ro quiet splash threadirqs
preempt=full vt.handoff=7

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching the assignee of the bug.