openmixerGitHub →

The stagebox is not establishing

Symptom

The box is powered and cabled, but the mixer shows no stagebox: no reac-capture device band in the patchbay, no inputs, and discovery reports nothing.

Work through these in order

1. Is the transport running at all?

systemctl --user status reac-pw
journalctl --user -u reac-pw -n 100 --no-pager

Two failures show up here immediately:

  • Missing configuration. If ~/.config/reac-pw/reac-pw.env declares no interface, or box, the unit fails fast with a message telling you to set it from Setup → Adapters → REAC. That is by design; configure it there rather than writing the file.
  • Missing capabilities. The unit checks /usr/bin/reac-pw for cap_net_raw before it starts and refuses with an explicit message if it is gone. Confirm with getcap /usr/bin/reac-pw (expect cap_net_raw,cap_sys_nice=ep) and reinstall the reac-pw package if it is missing.

2. Is it the right interface, and is the interface right?

ip -br link

The REAC interface must be:

  • the one physically cabled to the box — unplug it and watch which one loses carrier;
  • UP, with carrier;
  • without an IP address. REAC is raw Layer-2; addressing does nothing for it.

Then check what the mixer was told:

cat ~/.config/reac-pw/reac-pw.env ~/.config/reac-pw/*.env

REAC_LIVE_IFACE and REAC_TX_IFACE must both name that interface.

3. Is something else acting as master?

Exactly one master is allowed on a REAC segment. If a Roland desk is also on the wire acting as master, or a second copy of the transport is running on the same interface, the box cannot settle — the visible result is a box that never establishes, or one that appears and disappears.

Take everything else off the segment and try again. See the single-master rule.

4. Is the switch in the way?

The most reliable test is to remove the switch: run a direct cable from the REAC interface to the box. If it establishes on a direct cable and not through the switch, the switch is doing something to Layer-2 traffic it does not recognise, or the port is a mirror destination. See wiring.

5. Is more than one box transmitting?

Only one stagebox per segment is tracked: the first box to complete its handshake owns the link and frames from any other box are ignored. If you have two boxes on one wire, the second one will never appear.

Enable the transport's counters to see it:

systemctl --user edit reac-pw     # [Service] Environment=REAC_DEBUG=1
systemctl --user restart reac-pw
journalctl --user -u reac-pw -f

An other-source count climbing means a second box is transmitting.

6. Give it longer

A cold box can take a noticeable while to complete its handshake — a real desk holds a cold box waiting for tens of seconds. If the log shows the box being recognised but never granted, the dwell can be lengthened with REACPW_GRANT_DWELL_S (whole seconds); see the transport's knobs.

Try this only after the cabling and single-master checks. It is a knob for a box that wants a longer wait, not a fix for a wiring fault.

7. Power-cycle the box

Last, and it does sometimes matter: power the box down, wait, power it up with the cable already in place, and watch the log.

If it establishes but sounds wrong

Then this page is done and you want granulated or stuttery box audio.

If it establishes but the inputs are wrong

Check what the transport RECOGNIZED, not what anyone declared: reac.box-model / reac.box-width / reac.box-source on the reac-playback node (pw-dump | grep reac.box), and the recognized box = … line in journalctl --user -u reac-pw. Since 2026-08-05 the box's own declaration is what sizes and numbers the inputs and REAC_BOX is ignored by the transport; a reac.box-model of none means no box has declared itself yet, which is a wire or link problem, never a configuration one. (Historically the declared model sized the inputs, so an S-1608 declared as an S-0808 gave you eight inputs and no explanation — which is why nothing declares it any more.) See supported boxes.