Stop an Intel Bluetooth controller autosuspending when Omarchy's modprobe option and the btusb option change nothing
Confirmed on this workstation (systemd 261.2-1): /usr/lib/udev/hwdb.d/60-autosuspend-chromiumos.hwdb lists 8087:07DC, 0A2A, 0AAA, 0026, 0025 under ID_AUTOSUSPEND=1, 60-autosuspend.rules writes power/control="auto" only on add, and the Intel 8087:0aaa controller reads ID_AUTOSUSPEND=1, control=auto, runtime_status=suspended. Issue #12095 and its comments support the symptom, the 0x2013 timeout and blocked reset worker, the 61- udev rule working, and the TLP layer, and PR #12950 (open) ships the TLP USB_DENYLIST drop-in. The fix holds. What does not hold is the claim that the per-device write wins over usbcore.autosuspend. The issue's test wrote -1 to the parameter at runtime and reloaded btusb, but the USB device was not re-enumerated, and the kernel doc says the parameter only affects devices detected afterwards, so the controller kept its 2000 ms delay. The rule writes no delay for these IDs, and the kernel doc says a negative delay means never autosuspend, so a command-line usbcore.autosuspend=-1 would likely keep a non-TLP controller awake. Nobody has tested that. The title and cause are corrected to what the evidence supports. The udev rule stays the recommended fix because it is narrow and independent of the command line. Not exercised: no rule installed, no reconnect test.
Keeping the Bluetooth controller powered costs a little idle battery. A rule with no vendor and product match would pin every USB device on.
Symptom
A Bluetooth LE mouse or keyboard (Logitech MX Master 3S, ERGO K860 and similar) will not reconnect after the laptop sits idle or resumes. Omarchy's Bluetooth panel times out connecting, and the device only comes back after Forget and a fresh pair. The journal shows Bluetooth: hci0: command 0x2013 tx timeout then Bluetooth: hci0: Resetting usb device., and on some machines INFO: task kworker/... blocked for more than 122 seconds with __usb_queue_reset_device in the trace. Omarchy's /etc/modprobe.d/omarchy-usb-autosuspend.conf, writing -1 to /sys/module/usbcore/parameters/autosuspend at runtime, and options btusb enable_autosuspend=n made no difference.
Cause
systemd ships /usr/lib/udev/hwdb.d/60-autosuspend-chromiumos.hwdb, the Chromium OS autosuspend allowlist, which matches on every system and marks many Intel Bluetooth USB IDs (8087:0025, 8087:0026, 8087:0a2a, 8087:0aaa and others) with ID_AUTOSUSPEND=1. /usr/lib/udev/rules.d/60-autosuspend.rules then writes power/control=auto for that device when it is added. The device's idle delay comes from usbcore.autosuspend at enumeration, which is 2 seconds because Omarchy's modprobe.d file cannot apply to the built-in usbcore. btusb enable_autosuspend=n does not help because the udev rule has already allowed runtime suspend. Writing the usbcore parameter at runtime does not help either, because per the kernel documentation it only affects devices detected afterwards, and reloading btusb does not re-enumerate the USB device. A usbcore.autosuspend=-1 on the kernel command line would give the controller a negative delay at boot, which the kernel documentation says prevents autosuspend, but that has not been tested on this hardware. With the controller runtime-suspended, the BLE reconnect from a bonded device does not complete. Confirmed on this workstation (systemd 261.2-1): the hwdb entry and the rule line are present, and its Intel 8087:0aaa controller reads ID_AUTOSUSPEND=1, power/control=auto, runtime_status=suspended. The reconnect failure and the blocked reset worker are from the reporters on issue #12095, not reproduced here. TLP adds a further layer: its udev hook rewrites power/control and the delay after udev rules run, so TLP users need a TLP-side exclusion as well, and no kernel parameter protects them.
Fix
**1. Find the controller and its state.** No root needed:
for i in $(grep -l '^DRIVER=btusb' /sys/bus/usb/devices/*/uevent 2>/dev/null); do
d=$(dirname "$i"); d=${d%%:*}; echo "$d"
done | sort -u | while read -r d; do
echo "$d $(cat $d/idVendor):$(cat $d/idProduct) control=$(cat $d/power/control) status=$(cat $d/power/runtime_status)"
udevadm info -q property "$d" | grep ID_AUTOSUSPEND
done
control=auto with ID_AUTOSUSPEND=1 is this problem.
**2. Pin it on with a rule that runs after systemd's.** Use the vendor and product IDs printed above, in the lowercase form sysfs shows:
sudo install -Dm644 /dev/stdin /etc/udev/rules.d/61-bluetooth-no-autosuspend.rules <<'EOF'
ACTION=="add|bind|change", SUBSYSTEM=="usb", ATTR{idVendor}=="8087", ATTR{idProduct}=="0aaa", TEST=="power/control", ATTR{power/control}="on"
EOF
sudo udevadm control --reload
sudo udevadm trigger --action=change --subsystem-match=usb --attr-match=idVendor=8087
60-autosuspend.rules only acts on add, so the change trigger applies the new rule without undoing it. Rules in /etc/udev/rules.d/ and /usr/lib/udev/rules.d/ are ordered together by file name, so 61- runs after 60-.
**3. If TLP is installed**, also exclude the controller in a TLP drop-in, because TLP re-applies auto after udev:
# /etc/tlp.d/10-bluetooth.conf
USB_DENYLIST="8087:0aaa"
sudo systemctl restart tlp.service
Then reconnect the device. If it was already wedged, remove and re-pair it once.
Verify
Re-run step 1: the controller shows control=on status=active. udevadm test /sys/bus/usb/devices/<dev> 2>&1 | grep power/control lists the 60-autosuspend.rules write followed by the 61-bluetooth-no-autosuspend.rules write. The BLE device reconnects after an idle period and after a suspend and resume without re-pairing.