Stop the whole machine freezing a couple of minutes after a Bluetooth headset connects
Issue #12405 read in full. It supports the freeze timing (95 to 120 s), the QCA9377 0cf3:e500 hardware, kernel 7.2.3 with bluez 5.87 and pipewire 1.6.8, the xhci_hcd 420 to 96 interrupt drop, the WirePlumber A2DP-only workaround verbatim, and the kernel maintainer's request to A/B test 7.2.7 builds. Three defects. The symptom's spa.bluez5: Failure in Bluetooth audio transport log line appears nowhere in the issue or the linked forum thread, and the reporter states nothing reaches the log, so it was removed as fabricated. The cause said the conclusion came from no kernel trace, but the CachyOS thread the issue links has five pstore-captured hard LOCKUP panics with the identical stack btusb_work -> __set_isoc_interface -> usb_set_interface -> xhci_urb_dequeue on 7.0.x and 7.1.x when a SCO link is set up or torn down, so the cause was rewritten to cite it. The fix claimed Omarchy restores bluetooth-a2dp-autoconnect.conf: on 4.0.4-1 that file lives in /usr/share/omarchy/config/wireplumber and is copied to ~/.config at install, and no script in /usr/share/omarchy/bin or migrations touches wireplumber config again (only the ASUS mixer installer), so the claim was dropped. Confirmed locally: wireplumber 0.5.17 defines bluetooth.autoswitch-to-headset-profile. Note the original report was on Omarchy 3.8.5 with the Arch kernel, not linux-omarchy. Not exercised: no Bluetooth headset was connected and no WirePlumber change was made.
Each freeze is a hard power-off, which can lose unsaved work and, rarely, corrupt files being written. Apply the workaround before reconnecting the headset.
Symptom
After the kernel 7.x / BlueZ 5.87 / PipeWire 1.6.8 updates, connecting a Bluetooth headset that has a microphone hard-freezes the laptop about 2 minutes after audio starts (95 to 120 seconds in one report). Only holding the power button recovers it and nothing is written to the journal, even with the soft and hard lockup panics enabled. Before the update the same headset worked. A related report freezes or panics when a headset voice call (HFP/HSP) starts or ends, for example on hanging up a Discord call.
Cause
Not root-caused upstream, but it is tied to the Bluetooth voice (SCO) path of btusb on kernel 7.x. A CachyOS user captured five identical hard LOCKUP panics through pstore on 7.0.x and 7.1.x with the stack btusb_work -> __set_isoc_interface -> usb_set_interface -> xhci_urb_dequeue, triggered when a HFP/HSP connection is set up or torn down (CSR 0a12:0001 dongle). The Omarchy report (Qualcomm Atheros QCA9377, 0cf3:e500, kernel 7.2.3) shows the same shape: restricting the headset to A2DP stops the freeze and drops xhci_hcd interrupts from about 420 to about 96 per second. Both reporters treat it as an upstream kernel regression.
Fix
Make Bluetooth headsets playback-only so the SCO path is never opened. Put this in its own file beside Omarchy's bluetooth-a2dp-autoconnect.conf rather than editing that one, so your change stays separate:
mkdir -p ~/.config/wireplumber/wireplumber.conf.d
# ~/.config/wireplumber/wireplumber.conf.d/52-bluetooth-a2dp-only.conf
monitor.bluez.properties = {
bluez5.roles = [ a2dp_sink a2dp_source ]
}
wireplumber.settings = {
bluetooth.autoswitch-to-headset-profile = false
}
systemctl --user restart wireplumber
The cost is that the headset microphone stops working. Use the laptop or a USB mic for calls.
To get HFP back, test a newer kernel when one ships and remove the file. Omarchy's kernel maintainer asked affected users to A/B test 7.2.7 builds in the issue. Check what you run:
uname -r
pacman -Q bluez pipewire wireplumber
Verify
pactl list cards | grep -A20 bluez_card | grep -E 'headset-head-unit|a2dp' lists only A2DP profiles. Play audio through the headset for 15 minutes without a freeze.