#16141
Issues/omacom/omarchy/#16141

linux-omarchy 7.2.8-9: AMD 400-series xHCI dies after a Realtek RTL8761BU bluetooth dongle reset, takes keyboard/mouse with it

Openosszoiopened 19h ago

System details

AMD Ryzen 9 5900X, NVIDIA RTX 5080, Omarchy 4.0.6-1, linux-omarchy 7.2.8-9 (was 7.2.5-3)

What's wrong?

since going from linux-omarchy 7.2.5-3 to 7.2.8-9 the AMD 400 series chipset USB controller (1022:43d5 at 02:00.0) dies after a while and takes every USB device on it down with it.. keyboard, mouse receiver, webcam, mic, all gone until a reboot or an unbind/bind of the controller

the trigger is a Realtek RTL8761BU bluetooth dongle (0bda:a725) sitting on that controller.. btusb hits a tx timeout, resets the dongle, the dongle stops answering descriptor reads and then the whole xHCI goes:

[   35.666798] Bluetooth: hci0: command 0x2005 tx timeout
[   35.666831] Bluetooth: hci0: Resetting usb device.
[   35.772018] usb 1-2.1.4: reset full-speed USB device number 9 using xhci_hcd
[   51.456020] usb 1-2.1.4: device descriptor read/64, error -110
[   67.326026] usb 1-2.1.4: device descriptor read/64, error -110
[  104.100020] usb 1-2.1.4: Device not responding to setup address.
[  109.348012] usb 1-2.1.4: device not accepting address 9, error -71
[  109.350009] usb 1-2.1.4: WARN: invalid context state for evaluate context command.
[  110.003335] xhci_hcd 0000:02:00.0: ERROR unknown event type 2
[  126.239517] xhci_hcd 0000:02:00.0: Abort failed to stop command ring: -110
[  126.274926] xhci_hcd 0000:02:00.0: xHCI host controller not responding, assume dead
[  126.276496] xhci_hcd 0000:02:00.0: HC died; cleaning up

the dongle was already a bit flaky on 7.2.5, same tx timeout + reset every now and then, but it always came back on its own:

[  113.475781] Bluetooth: hci0: command 0x2005 tx timeout
[  113.475798] Bluetooth: hci0: Resetting usb device.
[  115.634024] usb 1-2.1.4: reset full-speed USB device number 9 using xhci_hcd
[  115.864008] Bluetooth: hci1: RTL: examining hci_ver=0a hci_rev=dfc6 lmp_ver=0a lmp_subver=d922

this is what the journal has per boot:

kernel     boots                     dongle resets   controller died
7.2.5-3    5 (sep 16 -> oct 21)      2-14 each       never
7.2.8-9    1st boot after update     5               after ~2 min
7.2.8-9    2nd boot                  61              after ~49 min

dongle firmware is the same in every boot (rtl8761bu, fw version 0xdfc6d922) so it's not the linux-firmware bump from the same update.. amd-ucode also went 20260810 -> 20260916 in that update and I haven't ruled that one out

recovering without a reboot works: unplug the dongle, then echo 0000:02:00.0 > /sys/bus/pci/drivers/xhci_hcd/unbind and the same into bind.. everything comes back

I haven't tried arch's stock linux 7.2.11 or going back to 7.2.5 with the dongle plugged in yet, so for now this is a correlation across 7 boots, not a bisect

Activity

0

No activity yet.