Add G29 force-feedback doc and a new-lg4ff installer

BeamNG.drive 0.39 native lost G29 FFB because SDL3's HIDAPI lg4ff driver
advertises an effect-status query it cannot answer (12,201 failures per
session in beamng.log). DOCS-g29-ffb.md records the diagnosis, the narrow
SDL_JOYSTICK_HIDAPI_LG4FF=0 hint vs the broad one, and the launch-option
%command% bug that had gamemoderun passed as a game argument.

install_new_lg4ff.sh swaps the in-tree hid-logitech for new-lg4ff via DKMS
(dkms.conf installs it under the same name in updates/dkms/), regenerates
initramfs presets that bundle the module, and offers --status/--revert.
Passed install/status/revert in vmbox; host run pending.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SS8M5wk88nJTXxDWz61KKk
This commit is contained in:
Krzysztof Rudnicki 2026-09-14 21:41:46 +02:00
parent b47c6a1a96
commit c81e4a9446
3 changed files with 356 additions and 0 deletions

View File

@ -0,0 +1,42 @@
{
"intent": "Restore G29 force feedback in BeamNG.drive native Linux and give the wheel an installable/revertible new-lg4ff driver upgrade.",
"scope": [
"linux_configuration/fixes/install_new_lg4ff.sh (new)",
"linux_configuration/docs/DOCS-g29-ffb.md (new), docs/DOCS-g29-shifter.md (cross-link)",
"The SDL env var itself lives in steam-backlog-enforcer (_desktop_env.py, commit 0f5d234) and in the BeamNG Steam launch options; not in this repo"
],
"changes": [
"install_new_lg4ff.sh: installs new-lg4ff-dkms-git from AUR, verifies dkms placed hid-logitech under updates/dkms/, regenerates initramfs presets that bundle hid-logitech, reloads the module when the wheel is present; --status and --revert (pacman -Rns + depmod + reload)",
"DOCS-g29-ffb.md: the SDL3 hidapi lg4ff diagnosis (12,201 'Failed to get joystick3 FFB effect status' lines), narrow vs broad SDL hint, the missing %command% launch-option bug, new-lg4ff traps"
],
"verification": [
{
"command": "grep -c 'FFB effect status' ~/.local/share/BeamNG/BeamNG.drive/current/beamng.log (after relaunch with SDL_JOYSTICK_HIDAPI_LG4FF=0)",
"result": "pass",
"evidence": "0 (was 12201); haptic feature line 'Constant' (kernel path); user confirmed force + self-centring in a car"
},
{
"command": "python3 /tmp/g29_ff_probe.py (evdev FF_CONSTANT +/-0x7fff, 2 s each)",
"result": "pass",
"evidence": "user: 'it moved all the way to the left and to the right' - kernel driver drives the motor"
},
{
"command": "vmbox lg4ff: makepkg new-lg4ff-dkms-git; install_new_lg4ff.sh; --status; --revert",
"result": "pass",
"evidence": "install rc=0 (module file updates/dkms/hid-logitech.ko.zst), status rc=0, revert rc=0 (module file back to kernel/drivers/hid/, package removed). Passed in vmbox only; host run pending"
},
{
"command": "shellcheck linux_configuration/fixes/install_new_lg4ff.sh; pre-commit run --files <changed>",
"result": "pass",
"evidence": "shellcheck clean; all hooks Passed"
}
],
"risks": [
"Host initramfs bundles hid-logitech (autodetect); the script regenerates all presets with mkinitcpio -P, which touches /boot",
"modprobe -r hid_logitech while a game holds the evdev node drops that handle; run with games closed"
],
"rollback": [
"install_new_lg4ff.sh --revert (pacman -Rns new-lg4ff-dkms-git, depmod, mkinitcpio -P, module reload)",
"Validate: --status shows module file under kernel/drivers/hid/ and ff caps 300040000 0"
]
}

View File

@ -0,0 +1,105 @@
# Logitech G29: no force feedback in BeamNG.drive (native Linux)
Files: `fixes/install_new_lg4ff.sh` (optional driver upgrade),
`~/src/steam-backlog-enforcer/steam_backlog_enforcer/_desktop_env.py` (the
host-level env var). Shifter companion: `DOCS-g29-shifter.md`.
## Symptom
BeamNG.drive 0.39.x, **native Linux build** (not Proton): the G29 steers,
pedals and buttons work, but the wheel is limp — no resistance, no
self-centring. Other games with FFB are unaffected.
## Cause (measured 2026-09-14, BeamNG 0.39.4.0, SDL 3.4.12 static)
`~/.local/share/BeamNG/BeamNG.drive/current/beamng.log`:
```
D|logHapticFeatures| Device joystick3 supports the following ffb features:
Constant Sine Triangle SawToothUp SawToothDown Spring Damper Friction Status Query
I|inputRegistry| Created ffbid 0, ... on device joystick3 with axis id=0
E|updateFrame| Failed to get joystick3 FFB effect status: (x12,201 per session)
```
That feature list is SDL3's **HIDAPI lg4ff** driver (`SDL_hidapi_lg4ff.c`),
which talks to the wheel over hidraw and bypasses the kernel `hid-logitech`
driver — the kernel driver advertises only constant/gain/autocenter
(`capabilities/ff` = `300040000 0`). The hidapi haptic backend advertises a
status query it cannot answer; BeamNG polls it every frame, gets an error, and
treats the effect as dead. Known regression:
<https://www.beamng.com/threads/ffb-not-working-on-linux.110807/>.
Not involved: the shifter HID-BPF (input direction only), hidraw permissions
(SDL had already opened the device), Proton (not used).
## Fix: make SDL use the kernel driver
Two env vars exist; they are not interchangeable:
| var | scope | use |
| ----------------------------- | ------------------------------------- | -------------------------------------------------------------------------- |
| `SDL_JOYSTICK_HIDAPI_LG4FF=0` | only the Logitech-wheel HIDAPI driver | **host-level**: set for every game, gamepads keep HIDAPI (BT rumble, LEDs) |
| `SDL_JOYSTICK_HIDAPI=0` | all of SDL's HIDAPI joystick drivers | forum-proven fallback, per-game only |
Both names verified against strings in `BinLinux/BeamNG.drive.x64`.
1. **Per game** — Steam → BeamNG.drive → Properties → Launch Options:
```
SDL_JOYSTICK_HIDAPI_LG4FF=0 LSFG_PROCESS=BeamNG.Drive gamemoderun %command%
```
The previous value had no `%command%`, so Steam appended
`LSFG_PROCESS=... gamemoderun ./BinLinux/BeamNG.drive.x64` as _arguments_
to the game (visible in `pgrep -a BeamNG.drive.x64`) and neither gamemode
nor LSFG was ever applied.
2. **Host level** — Steam on this machine is started by
`steam-backlog-enforcer` through `sudo -u kuhy env <allowlist> steam …`.
`sudo`'s `env_reset` makes that list the entire environment Steam and every
game sees, so `/etc/environment`, `~/.profile` or `~/.pam_environment`
never reach it. The variable lives in `desktop_env_args()` in the
enforcer's `_desktop_env.py`. Check it landed:
```bash
tr '\0' '\n' < /proc/$(pgrep -x steam)/environ | grep LG4FF
```
A missing line means Steam predates the enforcer change — it takes effect
on the next Steam launch (`systemctl restart steam-backlog-enforcer` kills
a running game; check `pgrep -f BeamNG.drive.x64` first).
## Verify
- Game-independent motor check, BeamNG closed:
`fftest /dev/input/by-id/usb-Logitech_G29_Driving_Force_Racing_Wheel-event-joystick`
— effects 0 (constant) and 4 (autocenter) must move the wheel.
- In a fresh `beamng.log` the haptic feature line becomes the kernel-style
list and `grep -c 'FFB effect status' beamng.log` drops to 0. A non-zero
count with working force is fine; a zero count with no force is not.
- The gate is physical: drive a car, feel resistance and centring.
## Optional: new-lg4ff instead of the in-tree driver
`fixes/install_new_lg4ff.sh` installs `new-lg4ff-dkms-git` (AUR). Its
`dkms.conf` sets `DEST_MODULE_NAME="hid-logitech"`, so DKMS places it in
`updates/dkms/` and it shadows the in-tree module _by name_ — no blacklist,
no `modules-load.d`. Once loaded it registers as `hid_logitech_new`. Adds
spring/damper/friction/periodic effects and the `spring_level` /
`damper_level` / `friction_level` / `gain` / `autocenter` sysfs knobs.
Traps the script handles:
- `mkinitcpio`'s `autodetect` hook bundles `hid-logitech` into the initramfs
because the wheel is plugged in at build time; a stale in-tree copy there
would load first at boot. The script regenerates the presets when an image
carries the module.
- `dkms status <name>` is a prefix filter that prints _everything_ on a miss;
always grep `dkms status` output.
- USB interface 1 of the G29 (`0003:046D:C24F.0007`) is unbound by design.
```bash
~/src/testsAndMisc/linux_configuration/fixes/install_new_lg4ff.sh # install (runs yay; needs a TTY)
~/src/testsAndMisc/linux_configuration/fixes/install_new_lg4ff.sh --status # which module owns the wheel + ff caps
~/src/testsAndMisc/linux_configuration/fixes/install_new_lg4ff.sh --revert # pacman -Rns + reload, no reboot
```

View File

@ -0,0 +1,209 @@
#!/bin/bash
# ============================================================================
# Logitech G29: replace the in-tree hid-logitech FFB driver with new-lg4ff
#
# new-lg4ff (github.com/berarma/new-lg4ff) adds spring/damper/friction and
# periodic effects the in-tree lg4ff lacks (it only has constant force). Its
# dkms.conf installs the module AS hid-logitech under updates/dkms/, so it
# shadows the in-tree copy by name: no blacklist, no modules-load drop-in.
# Why this matters for BeamNG and the SDL env vars: ../docs/DOCS-g29-ffb.md.
#
# install_new_lg4ff.sh install from AUR, reload, regen initramfs
# install_new_lg4ff.sh --status which module owns the wheel + FFB caps
# install_new_lg4ff.sh --revert remove the package, back to in-tree
#
# Run as the desktop user (yay refuses root); sudo is used where needed.
# The initramfs regen is required: mkinitcpio's autodetect hook bundles
# hid-logitech because the wheel is plugged in at build time, and a stale
# in-tree copy in the initramfs would win at boot over the DKMS one.
# ============================================================================
set -euo pipefail
SCRIPT_NAME="$(basename "$0")"
SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
readonly SCRIPT_NAME SCRIPT_DIR
readonly AUR_PKG="new-lg4ff-dkms-git"
readonly DKMS_NAME="new-lg4ff"
readonly MODULE="hid_logitech"
readonly VID_PID="046D:C24F"
readonly SHIFTER_FIX="$SCRIPT_DIR/fix_g29_shifter.sh"
readonly PACKAGES=(dkms joyutils)
MODE="install"
usage() {
sed -n '4,16p' "$0" | sed 's/^# \{0,3\}//'
exit 0
}
log() { echo "[$SCRIPT_NAME] $*"; }
die() {
echo "[$SCRIPT_NAME] ERROR: $*" >&2
exit 1
}
install_packages() {
local -a missing=()
local pkg
for pkg in "${PACKAGES[@]}"; do
pacman -Q "$pkg" &>/dev/null || missing+=("$pkg")
done
if ((${#missing[@]})); then
log "installing: ${missing[*]}"
sudo pacman -S --needed --noconfirm "${missing[@]}"
fi
# Headers for the RUNNING kernel, or dkms builds for a kernel nobody boots.
[[ -d "/usr/lib/modules/$(uname -r)/build" ]] ||
die "no headers for $(uname -r): install the matching linux*-headers"
}
# Path of the module modprobe would load: updates/dkms/... means new-lg4ff.
module_path() { modinfo -F filename "$MODULE" 2>/dev/null || echo "-"; }
is_new_lg4ff_active() {
# The DKMS module registers itself as hid-logitech-new even though the
# file is named hid-logitech; /proc/modules shows the registered name.
grep -q '^hid_logitech_new ' /proc/modules
}
wheel_event_dev() {
readlink -f /dev/input/by-id/usb-Logitech_G29_Driving_Force_Racing_Wheel-event-joystick 2>/dev/null || true
}
# Unbind and rebind the wheel so the freshly installed module takes over
# without a reboot. hid core rebinds the device on modprobe by itself.
reload_module() {
log "reloading $MODULE"
sudo modprobe -r "$MODULE"
sudo modprobe "$MODULE"
sleep 1
}
regen_initramfs() {
# Only the images that actually carry the module need rebuilding.
local img rebuilt=0
for img in /boot/initramfs-*.img; do
[[ -f $img ]] || continue
if lsinitcpio -l "$img" 2>/dev/null | grep -q 'hid-logitech'; then
rebuilt=1
fi
done
if ((rebuilt)); then
log "initramfs bundles hid-logitech: regenerating all presets"
sudo mkinitcpio -P
else
log "initramfs does not bundle hid-logitech: nothing to regenerate"
fi
}
show_status() {
local dev caps
log "package: $(pacman -Q "$AUR_PKG" 2>/dev/null || echo 'not installed')"
# `dkms status <name>` is a prefix filter that prints everything on a miss.
log "dkms: $(dkms status 2>/dev/null | grep "^$DKMS_NAME/" || echo 'none')"
log "module file: $(module_path)"
if is_new_lg4ff_active; then
log "loaded now: hid-logitech-new (new-lg4ff)"
elif grep -q '^hid_logitech ' /proc/modules; then
log "loaded now: hid_logitech (in-tree)"
else
log "loaded now: none"
fi
local d
for d in /sys/bus/hid/devices/*"$VID_PID"*; do
[[ -e $d ]] || continue
# USB interface 1 is unbound by design (hid-lg rejects it); only
# interface 0 carries the wheel, so "none" there is normal.
if [[ -e $d/driver ]]; then
log "$(basename "$d") driver: $(basename "$(readlink -f "$d/driver")")"
else
log "$(basename "$d") driver: none"
fi
done
dev="$(wheel_event_dev)"
if [[ -n $dev ]]; then
caps="$(cat "/sys/class/input/$(basename "$dev")/device/capabilities/ff")"
# In-tree lg4ff: "300040000 0" (constant, gain, autocenter only).
log "evdev $dev ff caps: $caps"
local sysdev
sysdev="$(readlink -f "/sys/class/input/$(basename "$dev")/device/device")"
local knob
for knob in range gain autocenter spring_level damper_level friction_level; do
[[ -f $sysdev/$knob ]] && log " $knob=$(cat "$sysdev/$knob")"
done
else
log "wheel: not plugged in"
fi
# An `&&` list as the last statement would make the function return 1
# when the shifter fix is absent, and set -e would abort on it.
if [[ -x $SHIFTER_FIX ]]; then
"$SHIFTER_FIX" --status | sed 's/^/ shifter: /'
fi
}
do_install() {
install_packages
if pacman -Q "$AUR_PKG" &>/dev/null; then
log "$AUR_PKG already installed"
else
command -v yay >/dev/null || die "yay not found; install it first"
log "installing $AUR_PKG from AUR (builds against $(uname -r))"
yay -S --needed --noconfirm "$AUR_PKG"
fi
dkms status | grep "^$DKMS_NAME/" | grep -q "$(uname -r).*installed" ||
die "dkms did not install $DKMS_NAME for $(uname -r): $(dkms status | grep "^$DKMS_NAME/")"
[[ "$(module_path)" == */updates/dkms/* ]] ||
die "modprobe still resolves $MODULE to $(module_path)"
regen_initramfs
if [[ -n "$(wheel_event_dev)" ]]; then
reload_module
is_new_lg4ff_active || die "reload done but hid_logitech_new is not loaded"
else
log "wheel not plugged in: module swap takes effect on next plug/boot"
fi
show_status
log "done. Feel check: fftest $(wheel_event_dev)"
}
do_revert() {
if pacman -Q "$AUR_PKG" &>/dev/null; then
# The dkms pacman hook removes the module from updates/dkms/.
sudo pacman -Rns --noconfirm "$AUR_PKG"
else
log "$AUR_PKG not installed"
fi
sudo depmod -a
[[ "$(module_path)" != */updates/dkms/* ]] ||
die "in-tree module not restored: $(module_path)"
regen_initramfs
if [[ -n "$(wheel_event_dev)" ]]; then
reload_module
! is_new_lg4ff_active || die "reload done but hid_logitech_new still loaded"
fi
show_status
}
main() {
case "$MODE" in
install) do_install ;;
revert) do_revert ;;
status) show_status ;;
esac
}
while [[ $# -gt 0 ]]; do
case $1 in
--status) MODE="status" ;;
--revert) MODE="revert" ;;
-h | --help) usage ;;
*)
echo "Unknown option: $1" >&2
exit 1
;;
esac
shift
done
main