Fix Wi-Fi collapsing whenever a Bluetooth device is connected
The primary fix is work Omarchy 4 already does, and does more safely than the record's commands. Confirmed on this machine: /usr/share/omarchy/bin/omarchy-network-band exists and omarchy network band runs, and it is in upstream v4.0.3 at bin/omarchy-network-band. Reading it, it checks the SSID is actually reachable on the requested band before pinning, and on failure it restores the previous 802-11-wireless.band and reconnects, with the comment "rather than leaving the machine stranded offline". The record's bare nmcli connection modify ... band a followed by nmcli connection up has neither guard, and the consequence, a laptop pinned to a band its AP does not serve with the setting surviving reboot, is missing from danger entirely. That is the main correction. The script also handles 6GHz, which NetworkManager has accepted since 1.44 and the record does not mention, and a 6 GHz pin is strictly better for Bluetooth coexistence than 5 GHz.
The record's coexistence analysis is right and I extended it. modinfo -p iwlwifi on kernel 7.1.9 on this machine still lists bt_coex_active, so the parameter exists. A GitHub code search of torvalds/linux master for bt_coex_active under drivers/net/wireless/intel/iwlwifi returns five files: iwl-modparams.h and iwl-drv.c for the declaration, dvm/main.c and dvm/lib.c where it has real effect, and mvm/mac80211.c where line 461 only logs iwlmvm doesn't allow to disable BT Coex, check bt_coex_active module parameter. The record named only dvm/main.c. More importantly, kernel 7.1 ships a third driver, iwlmld (present at /lib/modules/7.1.9-arch1-2/kernel/drivers/net/wireless/intel/iwlwifi/mld), which does not reference bt_coex_active at all and configures coexistence unconditionally in mld/coex.c via iwl_mld_send_bt_init_conf. So on BE200 class hardware the parameter is not even read to complain about. The record's operational conclusion held, but its hardware boundary was a generation out of date.
The A2DP advice is correct but partly redundant on Omarchy 4, and I said so rather than dropping it. Confirmed on this machine: ~/.config/wireplumber/wireplumber.conf.d/bluetooth-a2dp-autoconnect.conf is installed and sets bluez5.auto-connect = [ a2dp_sink a2dp_source ], and the same file is in v4.0.3 at config/wireplumber/wireplumber.conf.d/bluetooth-a2dp-autoconnect.conf. The profile name still resolves: spa_bt_profile_name in PipeWire's spa/plugins/bluez5/defs.h returns a2dp-sink for SPA_BT_PROFILE_A2DP_SINK, and pipewire 1:1.6.8-1 with wireplumber 0.5.15-1 is installed here, so pactl set-card-profile ... a2dp-sink is current rather than PulseAudio era advice.
The discovery advice is folklore on a stock Omarchy 4 machine and I cut it back. bluetoothctl show on this machine reports Discoverable: no on controller 3C:6A:A7:68:5C:62, and /usr/share/omarchy/default/systemd/user/bt-agent.service documents that the adapter is only pairable while the user has the Bluetooth panel open and scanning. Nothing scans in the background, so discoverable off and pairable off recover no airtime, and pairable off actively fights Omarchy's own pairing flow. omarchy-bluetooth-power states that BlueZ never persists these properties, so none of the three commands survive a reboot anyway. /etc/bluetooth/main.conf is stock bluez 5.87-2 with every section empty, so Omarchy sets nothing there.
The kernel branch checks out, with one redundancy removed. I read /usr/bin/omarchy-update-pacman-guard and /usr/share/omarchy/default/libalpm/hooks/00-omarchy-update-guard.hook: the guard only aborts when both -S and -u appear in the pacman command line, so pacman -S --needed linux-lts linux-lts-headers passes, and the record's ordering of omarchy update first is the correct way to avoid a partial upgrade. /usr/bin/limine-mkinitcpio does exist, owned by limine-mkinitcpio-hook, but 80-limine-efi-deploy.hook and 90-limine-mkinitcpio-remove-post.hook already deploy the entry during the install transaction, so the manual call is redundant and I dropped it. omarchy-setup-direct-boot exists in /usr/share/omarchy/bin, so the existing Direct Boot warning is real and I kept it and named the command. The record's 6.11 to 6.12 RTL8852CE claim is sourced but four kernel series stale against 7.1.9, so I reframed it as a precedent rather than as current advice. iw dev wlan0 was wrong throughout: this machine's only wireless interface is wlo1, with no net.ifnames=0 on the kernel cmdline.
Both forum sources still resolve, HTTP 200 for https://bbs.archlinux.org/viewtopic.php?id=287090 and ...id=302036, and both raw kernel URLs return 200, so nothing is removed. Severity stays medium and frequency stays common, because unlike the power save record the distribution does not remove the underlying cause, it only gives a safer tool. Not exercised: this workstation has an Intel Wireless-AC 9560 CNVi part (8086:a370, subsystem 8086:0034) paired with an Intel 9460/9560 Jefferson Peak Bluetooth controller on USB 8087:0aaa, which is exactly the shared combo part the cause describes, but wlo1 is down and the machine runs on eno2, and I was instructed not to touch NetworkManager, the radios, Bluetooth or audio. So I did not associate on 2.4 GHz, did not measure any throughput collapse or ping latency, did not pair a Bluetooth audio device (no bluez_card.* exists in pactl list cards here), and did not test any modprobe option, band pin or LTS kernel boot.
Pinning a band can leave the machine offline. 802-11-wireless.band a on a profile whose AP has no 5 GHz radio means there is nothing to associate to, and the setting persists across reboots, so a laptop with no wired fallback has no way back on to the network until you clear it. Check the AP answers on the band first, or use omarchy network band, which refuses a band the SSID is not reachable on and reverts plus reconnects if the radio cannot come back up.
Installing an additional kernel and rebooting into it is safe as long as you keep the current one installed and the Limine menu reachable. Do not enable Omarchy's Direct Boot with omarchy-setup-direct-boot while you are testing, because it points firmware straight at the Omarchy UKI and you would then have no way to select the other entry without going through the firmware boot menu.
Symptom
The moment Bluetooth headphones, a mouse or a controller connect, 2.4 GHz Wi-Fi falls apart: pages stop loading, YouTube buffers forever, ping latency jumps from 5 ms to hundreds of ms with packet loss, and downloads drop to a trickle. Disconnect the Bluetooth device and everything is instantly normal again. Sometimes it is symmetrical: the mouse stutters and the headset drops out while a large download runs.
Cause
Bluetooth and 2.4 GHz Wi-Fi share the same ISM band, and on almost every laptop they share the same combo chip and the same antenna. The chip's coexistence arbiter has to time-slice between them, and when the arbitration is poor (a firmware regression, a laptop whose antenna wiring the driver cannot detect, or a headset running the airtime-hungry HFP/SCO profile), one side starves the other. This is a hardware-arbitration problem, not a configuration error, so the reliable fixes are about getting off the shared band rather than tuning software.
Fix
Find the real interface name first. Omarchy 4 uses predictable names such as wlo1 or wlp3s0, and wlan0 does not exist:
iw dev | awk '/Interface/ {print $2}'
Use that name in place of wlo1 below. Confirm you are actually on 2.4 GHz:
iw dev wlo1 link | grep -i freq # 2412-2484 MHz = 2.4 GHz, 5xxx = 5 GHz
**The fix that works: move Wi-Fi off 2.4 GHz.**
On Omarchy 4, use the command the distribution already ships for this. It checks that the SSID is actually reachable on the band before pinning, and if the radio cannot come back up it puts the previous setting back and reconnects, so a wrong guess does not leave you offline:
omarchy network band # show current band, what is available, what is pinned
omarchy network band 5 # pin 5 GHz
omarchy network band 6 # pin 6 GHz if the AP offers it
omarchy network band auto # unpin
On plain Arch, or if you want to see the underlying property, NetworkManager takes 802-11-wireless.band as "a" for 5 GHz, "bg" for 2.4 GHz, and since 1.44 "6GHz":
nmcli -g 802-11-wireless.band connection show "<SSID>"
nmcli connection modify "<SSID>" 802-11-wireless.band a
nmcli connection up "<SSID>"
iw dev wlo1 link | grep -i freq
Check that the AP answers on the band before you pin it. If it does not, nmcli connection up fails and the profile stays pinned to a band with nothing to associate to:
nmcli -f SSID,FREQ device wifi list --rescan no | grep "<SSID>"
If the pin leaves you offline, put it back:
nmcli connection modify "<SSID>" 802-11-wireless.band ""
nmcli connection up "<SSID>"
If the two bands have separate SSIDs, connect to the 5 GHz one and give it a higher autoconnect priority:
nmcli connection modify "<SSID-5G>" connection.autoconnect-priority 10
If you must stay on 2.4 GHz, move the AP to channel 1 or 11 rather than leaving it on auto, so Bluetooth's adaptive frequency hopping has clear room away from your channel.
**Keep the headset on A2DP rather than HFP.** HFP and its SCO link keeps the radio at constant duty, where A2DP bursts. Omarchy 4 already asks WirePlumber to prefer the A2DP profiles on connect, in ~/.config/wireplumber/wireplumber.conf.d/bluetooth-a2dp-autoconnect.conf:
monitor.bluez.rules = [
{
matches = [ { device.name = "~bluez_card.*" } ]
actions = { update-props = { bluez5.auto-connect = [ a2dp_sink a2dp_source ] } }
}
]
WirePlumber still switches to HFP when an application opens the headset microphone, so check and override the live profile when you are not on a call:
pactl list cards | grep -E 'Name: bluez_card|Active Profile'
pactl set-card-profile bluez_card.XX_XX_XX_XX_XX_XX a2dp-sink
**Do not bother turning off discovery on a stock Omarchy 4 machine.** Active scanning does hop across the whole band, but Omarchy does not scan in the background. bluetoothctl show reports Discoverable: no, and bt-agent.service only auto-accepts pairing while you have the Bluetooth panel open and scanning. If you have started a scan by hand, stop that one scan. None of these properties persist across a reboot, because BlueZ never saves them:
bluetoothctl scan off
Leave pairable alone. Turning it off breaks Omarchy's own pairing flow and buys no airtime.
**On old Intel cards only**, the coexistence arbiter can be turned off. bt_coex_active is still declared in iwlwifi/iwl-drv.c and modinfo -p iwlwifi still lists it on kernel 7.1, but check which driver your card uses before spending time on it:
basename "$(readlink -f /sys/class/net/wlo1/device/driver)"
ls /sys/module | grep -E '^iwl(mvm|mld|dvm)$'
echo 'options iwlwifi bt_coex_active=0' | sudo tee /etc/modprobe.d/iwlwifi-coex.conf
sudo reboot
This only does anything under iwldvm, the 5000 and 6000 series cards. Under iwlmvm (7260 and newer through AX210) the only thing that reads the parameter is mvm/mac80211.c, and all it does is log iwlmvm doesn't allow to disable BT Coex, check bt_coex_active module parameter. Under iwlmld, which kernel 7.1 uses for BE200 class cards, nothing reads it at all and coexistence is configured unconditionally in mld/coex.c. So on any recent laptop this is a dead end, despite the amount of forum advice that says otherwise.
**If this started right after a kernel update**, suspect a coexistence regression rather than your setup. Realtek RTL8852CE users tracked exactly this to the 6.11 to 6.12 jump, which is the cited precedent rather than current advice against kernel 7.1. Test the LTS kernel, which installs alongside the one you have:
omarchy update
sudo pacman -S --needed linux-lts linux-lts-headers
sudo reboot
# pick the LTS entry from the Limine menu
Run omarchy update first. pacman -S on its own uses the sync databases as they stand, so a fully updated system is the only state in which it cannot pull a partial upgrade. The Omarchy ALPM guard does not block this command, because it only aborts when both -S and -u are present. Pacman's own Limine hooks deploy the new entry, so there is no need to run limine-mkinitcpio by hand.
Verify
With the Bluetooth device connected and playing audio, measure latency and loss against the router rather than against the internet:
ping -i 0.2 -c 300 <router-ip>
Latency should stay in single or low double digit milliseconds with no loss. Then confirm the band actually moved:
omarchy network band # Omarchy 4
iw dev wlo1 link | grep -i freq # any Arch: expect 5xxx or 6xxx MHz
A large download should hold its speed with the headset in use. If you pinned a band, reboot once and check it reassociated on that band by itself rather than falling back.
Sources
- https://bbs.archlinux.org/viewtopic.php?id=287090
- https://bbs.archlinux.org/viewtopic.php?id=302036
- https://raw.githubusercontent.com/torvalds/linux/master/drivers/net/wireless/intel/iwlwifi/iwl-drv.c
- https://raw.githubusercontent.com/torvalds/linux/master/drivers/net/wireless/intel/iwlwifi/dvm/main.c
- https://networkmanager.dev/docs/api/latest/settings-802-11-wireless.html
- https://github.com/omacom/omarchy/blob/v4.0.3/bin/omarchy-network-band
- https://github.com/omacom/omarchy/blob/v4.0.3/config/wireplumber/wireplumber.conf.d/bluetooth-a2dp-autoconnect.conf
- https://raw.githubusercontent.com/torvalds/linux/master/drivers/net/wireless/intel/iwlwifi/mvm/mac80211.c
- https://raw.githubusercontent.com/torvalds/linux/master/drivers/net/wireless/intel/iwlwifi/mld/coex.c
- https://raw.githubusercontent.com/PipeWire/pipewire/master/spa/plugins/bluez5/defs.h
- https://man.archlinux.org/man/nm-settings-nmcli.5