usb_device.reset() wedges the Arctis Nova 7 dongle until physical replug

[CaWu009]

Summary

configure_virtual_sinks() calls self.usb_device.reset() unconditionally before kernel_detach(). On the Arctis Nova 7 dongle (1038:22a1, bcdDevice 2.31) that reset leaves the device in a state where usbhid can no longer read the HID descriptor on interface 3:

usbhid 1-3:1.3: can't add hid device: -71
usbhid 1-3:1.3: probe with driver usbhid failed with error -71

The hidraw node for that interface disappears, headsetcontrol stops seeing the headset too, and every subsequent SET_REPORT times out. Only a physical replug recovers it — a service restart does not, because the reset happens again on each start.

Environment

• linux-arctis-manager 2.4.1 (pipx, Python 3.14.6)
• Fedora, kernel 7.1.5
• SteelSeries Arctis Nova 7, 1038:22a1, bcdDevice 2.31

Still present on v2.5.0-beta3 — the reset() call is unchanged there.

Evidence

The same -71 appears from kernel_attach() during teardown(), so once init has failed the interface is left with no driver bound at all:

$ for i in /sys/bus/usb/devices/1-3:1.*; do echo "$i: $(basename $(readlink $i/driver))"; done
/sys/bus/usb/devices/1-3:1.0: snd-usb-audio
/sys/bus/usb/devices/1-3:1.1: snd-usb-audio
/sys/bus/usb/devices/1-3:1.2: snd-usb-audio
/sys/bus/usb/devices/1-3:1.3: (none)
/sys/bus/usb/devices/1-3:1.4: usbhid
/sys/bus/usb/devices/1-3:1.5: usbhid

With the reset skipped, init succeeds reliably whenever the headset is powered on — verified with pyusb against the same interface, SET_REPORT completing in ~2 ms.

Suggested fix

Make the reset opt-in, or skip it for known-affected product IDs. Local workaround in use here:

if os.environ.get('LAM_SKIP_USB_RESET', '1') != '1':
    try:
        self.usb_device.reset()
    except usb.core.USBError as e:
        self.logger.warning(f"Error resetting USB device (non-fatal): {e}")

Happy to open a PR if you have a preference for how it should be gated (env var, device config flag, or product-ID allowlist).

[maacel]

I can reproduce a very similar issue on an Arctis Nova 7 Gen 2 (1038:227e) using linux-arctis-manager 2.4.1 on CachyOS/Arch.

My failure looks like this:

CoreEngine: Kernel driver active on interface 3, detaching...
CoreEngine: Claimed interface 3
CoreEngine: Initializing device...
CoreEngine: Error sending command: [Errno 110] Operation timed out
CoreEngine: Unrecoverable USB error (errno 5): [Errno 5] Input/Output Error
CoreEngine: Device init failed after connect
CoreEngine: Re-attaching kernel driver...
CoreEngine: Error re-attaching kernel driver: [Errno 2] Entity not found

After the failure:

1-4:1.0 snd-usb-audio
1-4:1.1 snd-usb-audio
1-4:1.2 snd-usb-audio
1-4:1.3 NO DRIVER
1-4:1.4 usbhid
1-4:1.5 usbhid

The dongle remains visible as:

1038:227e SteelSeries ApS Arctis Nova 7 Gen 2

Normal USB audio continues to work.

I can usually recover without physically unplugging the dongle:

systemctl --user stop arctis-manager
echo '1-4:1.3' | sudo tee /sys/bus/usb/drivers/usbhid/bind
systemctl --user start arctis-manager

This has recovered the device multiple times.

I also tested booting/restarting with the headset powered both ON and OFF. In my case the problem occurs intermittently regardless of the headset power state.

Firmware for both headset and dongle has been updated with SteelSeries GG under Windows.

This looks very similar to #61, but on the Nova 7 Gen 2 PID 227e. I would be happy to test the proposed skip-reset workaround or a patched build.