Skip to content

Troubleshooting

mescon edited this page Nov 6, 2026 · 23 revisions

Troubleshooting

First move for anything at all: open logi-wheel or logi-wheel-gui. When either one cannot find a wheel, it says which check failed, what that means, and the one command that fixes it. A button runs that command. That covers the usual causes: nothing plugged in, a G923 still in console mode, the driver not loaded, another driver holding the wheel, missing permission rules.

From a git checkout there is also:

./tools/setup.sh doctor

It health-checks every layer: module built, loaded, bound, the udev rule (udev is the service that sets device permissions), permissions, tools. It names the exact fix for anything that fails.

Force feedback stopped when TrueForce was set up

Requires 0.34.0 or newer. It needs no configuration.

To reach Logitech's SDK (the software kit games use for TrueForce), the game must read the wheel over raw HID. Raw HID means the game talks to the USB device directly, instead of through the Linux input layer. PROTON_ENABLE_HIDRAW enables that. A game that drives force the older Windows way needs somewhere to send it on that path. The driver provides that.

Two side effects of raw HID remain. They happen because the game reads a control at rest by a different convention:

  • Pedals inverted: resting registers as fully pressed. Turn on the game's own "invert axis" option for those pedals.
  • A control stuck as held: for example a constant "up" that makes menus unusable. Rebinding it in the game usually clears it.

If either spoils a game, you may prefer force feedback over the game's own TrueForce. Then turn the raw path off for that one game in ~/.config/logi-wheel/games.conf. For example:

3058630  hidraw=0 tfsim=1 relay=ac-evo

You keep force feedback and get simulated TrueForce instead.

A harsh buzz, or an engine note an octave low

Requires 0.37.0 or newer. Update, and there is nothing to configure.

The wheel's TrueForce endpoint carries one packet per millisecond in total. So exactly one program streams to it at a time. When two do, the wheel's torque target alternates between their values every millisecond. That is a fixed 500 Hz buzz through the rim. Each side's samples also arrive at half rate, which is the octave. The driver stands aside when a program takes the stream. It carries its own force inside that program's packets. This project's streamers take a lease on the wheel, so they cannot collide with each other. The rule and the wire measurements behind it are in docs/TRUEFORCE_PROTOCOL.md. stream_yield in docs/SYSFS_API.md turns the driver's half of it off.

Common symptoms

Symptom Fix
The wheel never shows up in lsusb, or appears and vanishes Read the kernel log first: sudo dmesg | grep -i usb | tail -n 30. Lines like disabled by hub (EMI?), device descriptor read/64, error -71 or unable to enumerate USB device mean the USB link itself failed, before any driver ran. The usual cause is a front-panel port or a hub. Move the wheel to a port on the back of the computer. Then power it off and on (issue #141)
logi-dd is gone / where is the app? The apps are logi-wheel and logi-wheel-gui. The logi-dd names remain as symlinks, so existing scripts and launch options keep working. See Installation
No wheel_* files, no FFB (wheel grabbed by hid-generic) The module did not bind. Confirm it is loaded: lsmod | grep hid_logitech_dd. Replug the wheel. Check dmesg. sudo logi-rebind-wheel moves it across without a replug
Settings writes return "permission denied" The udev rule that grants your desktop user access is missing or not applied. Run sudo udevadm control --reload-rules && sudo udevadm trigger. Then replug the wheel. No group membership is needed. The apps report this one themselves
Force feedback pulls the wrong way A native game and a Wine/Proton game can want opposite signs. Toggle Invert constant force in logi-wheel (the wheel_ffb_constant_sign attribute)
A game stops seeing the wheel after a driver reload Restart Steam completely. Its device list goes stale across reloads
Steering and force feedback stop as soon as a session starts, though the wheel worked in the game's menus Raw HID is on with Logitech's stock TrueForce library, and no proxy sits in front of it. The library takes the wheel and never gets its rotation answered. Launch through logi-launch. It stages a proxy that answers, and it refuses raw HID when it cannot. By hand: tools/install-tf-shim.sh --proxy. Also check that Steam Input is disabled for the game. Its virtual gamepad gives the same symptom
Rotation snaps to 90° (45° each way) when a sim starts Not your setup, and not the wheel. Logitech's TrueForce SDK asks G HUB what rotation your wheel is set to. Under Proton nothing answers, so it falls back to 90, the minimum of the wheel's legal 90 to 2700 range. On the direct-drive wheels the range really is written. The driver puts it back by itself (wheel_range_restore, on by default). On the G923 the wheel is never actually changed, so there is nothing to put back. The rim keeps its full travel (check a game's config screen and you will see it). The game clamps its own steering. The fix there is the game's own steering-lock or rotation setting. See issue #27 for the full trail
Steering feels off-center Hold the rim physically straight. Then run echo 1 > "$H/wheel_calibrate_here", or use the Calibrate action in logi-wheel
Rumble shakes the steering instead of buzzing the rim Check that cat "$H/wheel_texture_route" says tf. Texture belongs on the haptic channel
No FFB in a DirectInput game (Le Mans Ultimate etc.) Launch it with logi-launch %command%. That runs the logi-ffb proxy and the telemetry relay for you. Turn off Steam Input. See Force Feedback in Games
No force at all in the next game after playing a DirectInput sim Requires 0.37.0 or newer. There, logi-ffb hands the wheel's force gain back when it exits. The gain is a device-wide setting that nothing else resets. On an older version, replug the wheel to clear it
Lights, engine texture or settings land on the other wheel Requires 0.37.0 or newer. There, every piece addresses a wheel by its own identity, and a session undoes only what it armed. On a mixed rig, logi-launch --wheel dd / --wheel g923 picks which wheel a game's session is for. See Force Feedback in Games
Pedals read as fully pressed when released (in game) A side effect of PROTON_ENABLE_HIDRAW=1. The game reads the wheel's raw HID reports, which use the opposite convention to the Linux input layer. Turn on the game's own "invert axis" for those pedals. Or unset the variable, if you do not need TrueForce. Nothing is wrong with the wheel.
TrueForce does not turn on in a game that supports it Check three things. Logitech's TrueForce files are installed for that game (Setup page; /tmp/logi-launch.log says whether raw HID was granted and the proxy staged). The game launches with logi-launch %command%, which sets the raw-HID mode and stages the SDK forwarder (a hand-set PROTON_ENABLE_HIDRAW=1 remains a fallback). Steam Input is disabled for the game
TrueForce worked, then never again after a game crash or hard kill The wheel's session latched. A session killed before its teardown reached the wheel leaves the next SDK session opening fine (SetForceMode succeeds) but never streaming. Power-cycle the wheel base before relaunching. LOGI_TF_REARM=1 enables an experimental pre-launch re-arm in logi-launch. It is off by default and unvalidated
No engine texture and dark rev lights in AC EVO A leftover LOGI_ESCAPE_RELAY=0 in the game's launch options starves both. Remove it. logi-launch %command% starts the telemetry bridge and the texture merge itself
Wheel oscillates hard during a load screen Keep hands clear during AC EVO map loads. The driver damps game-uploaded springs (wheel_spring_damping, default 25) to keep the wheel's over-torque failsafe from tripping
G923: pedals act pressed when released (a game rewinds a replay, throttle stuck on) The wheel sends its pedals inverted. Nothing in HID can mark an axis that way, so every layer passes it on. Fixed in the driver: update, and a released pedal reads 0. If you had already inverted these axes inside a game, clear that setting. Otherwise the two corrections cancel out. See G923
G923: no force feedback in a game Remove PROTON_ENABLE_HIDRAW=1 from the game's launch options. That recipe is for the RS50 / G PRO and breaks the G923. Use logi-launch %command% as the only launch option, or none at all. Steam Input off. See G923
G923: another driver grabbed the wheel The packaged udev rule reclaims c266/c267/c26e automatically. Replug and check dmesg. If you run berarma's new-lg4ff, note that this driver blacklists it. The two race for the same PIDs
The light strip stays dark, but every setting saves fine Check which onboard profile is active. The wheel stores five. A profile can keep the strip off entirely. LED writes still report success on such a profile, so the driver looks broken while the wheel does what the profile says. Switch profiles in the app and try again. Custom colours need 0.30.0 or newer, and the symptom is identical, so update first
G923: settings missing from the app (LIGHTSYNC, curves, profile slots, mode toggle) Not a fault. The apps hide what a wheel does not have, and those are direct-drive features. The G923's four settings, desktop profiles and the tests are all there. See G923
TrueForce stops working in Assetto Corsa EVO after a game update A stale Proton prefix, not the driver. Delete the input* files and video.videosettings under <prefix>/drive_c/users/steamuser/Saved Games/ACE/. Start the game again; the game rebuilds them. Deleting the whole prefix also works, but loses your lap times and settings. Windows users report the same thing after the same update. Found by simonr2k4
RS50 Xbox edition: no input device at all The wheel is in console mode (046d:c275). There it speaks the console's own protocol and has no HID++ interface to bind. Requires 0.38.0 or newer. Install usb_modeswitch and replug. The switch to 046d:c276 is automatic on every plug-in. By hand, or on a system without systemd: sudo logi-wheel-modeswitch
G923 Xbox edition: no input device at all The wheel is in console mode (046d:c26d). There it speaks the console's own protocol and has no HID++ interface to bind. Install usb_modeswitch and replug. The switch to 046d:c26e is automatic on every plug-in. By hand, or on a system without systemd: sudo logi-wheel-modeswitch. doctor reports the rule and the helper on their own lines. If it never switches, the out-of-tree xone driver may be claiming the wheel first

($H above is the wheel's sysfs directory. sysfs is the kernel's settings file tree. Set it with H=$(ls -d /sys/class/hidraw/*/device/wheel_range | head -1 | xargs dirname). On a G923 there are no wheel_* files. Its classic attributes are listed in the Sysfs API Reference.)

Behaviors that are not bugs

  • The wheel self-centers when nothing is running (compat mode, onboard profiles). This is firmware behavior. It is the same on Windows with G HUB running.
  • Compat mode boots at 90°. This is the factory default. Set your range once. Go to desktop mode first (wheel_profile=0), then set wheel_range=900. See Configuring the Wheel.
  • Strong forces when reversing in sims that model self-aligning torque. This is real-car physics coming through a direct-drive motor. It is identical on Windows. Lower wheel_strength if it is too much.
  • The rev strip goes dark when telemetry stops. This is deliberate. The display follows the game's live data. It darkens when telemetry is lost, instead of freezing at the last level.

Reporting a bug

One command collects everything:

logi-wheel --report

In the apps it is the same report. Press b on the terminal app's Info page. Or click the Collect button on the window's Info page, which copies the report to the clipboard and saves a copy. All three print the same thing: versions, which wheels are bound and to what, every wheel setting and its value, the simulated-TrueForce configuration, and which udev rules are installed.

It deliberately leaves out the three things that identify you and help diagnose nothing: your wheel's serial number, your profile names and your lighting slot names. They are shown as withheld, not printed. The dmesg command the report suggests filters the serial line out too. That matters because the driver logs the serial at probe (when it first finds the wheel).

If you would rather assemble it yourself, an issue needs:

  • your kernel version and distribution,
  • the driver and app versions. All four binaries answer --version with the build stamp in parentheses. dmesg shows the same stamp for the module. ./tools/setup.sh doctor prints them together, and fails when an app and the module were not built from the same source,
  • dmesg | grep -iE 'rs50|g pro|g923' output,
  • cat "$H/wheel_firmware" output (RS50 / G PRO). Firmware-dependent behavior (feature index drift, settings quirks) is tracked against it.

The Info / Testing page (where logi-wheel and logi-wheel-gui open) shows the wheel's name, serial and firmware, plus the app and driver versions, in one copyable place.

Help the project by racing

The fastest way to improve the driver is to race and report:

  • Any sim from the "expected" list in Force Feedback in Games working (or not) moves the compatibility matrix. One sentence and a wheel_firmware output is enough.
  • Real G PRO owners: feel reports on texture routing (wheel_texture_route tf vs kf) and the rev-light control (wheel_rev_level) are the top items the project cannot test itself.
  • G923 PlayStation owners: force feedback and simulated TrueForce are hardware-verified. Feel reports from real races, and from any game where something is off, are what tunes them. See G923.
  • G923 Xbox owners: force feedback and TrueForce are confirmed on a real unit. Rev lights are new. The one thing still unconfirmed in a game is whether the strip follows engine RPM while driving. See G923.

Clone this wiki locally