FreeBSD 15 on a MacBook Pro Mid-2012

Hardware Overview

What Works

1. Trackpad Fix — hw.usb.usbhid.enable=0

This is the most important fix. Understanding why it works will help you diagnose regressions.

The Problem

The Apple keyboard/trackpad is a single composite USB device (0x05ac:0x0252) with three HID interfaces:

USB Device 0x05ac:0x0252 (Wellspring5A ANSI)
├── Interface 0   HID Keyboard      (bInterfaceProtocol=0x01)
├── Interface 1   HID Touchpad      (bInterfaceProtocol=0x00)  <-- wsp needs this
└── Interface 2   HID Consumer Ctrl (bInterfaceProtocol=0x00)

FreeBSD has wsp, a driver written specifically for Apple trackpads — and 0x0252 is in its device table. But it never bound.

wsp is a uhub-level driver — it probes at the raw USB layer. usbhid is a bridge driver that by default claims every HID-class interface first, creating a hidbus for each one. Once usbhid claims Interface 1, the generic hms (basic mouse) driver attaches to the hidbus — and wsp never gets to probe at all. Result: basic mouse, no gestures, no /dev/wsp0.

The Fix

hw.usb.usbhid.enable=0

With usbhid disabled, the USB layer handles each interface directly:

With hw.usb.usbhid.enable=0 (working):
USB Device 0x05ac:0x0252
├── Interface 0 → ukbd1  → keyboard input ✓
├── Interface 1 → wsp0   → /dev/wsp0 → moused → /dev/sysmouse ✓
└── Interface 2 → (none) → brightness/backlight via ACPI/SMC still works ✓

Without it (broken default):
USB Device 0x05ac:0x0252
└── usbhid → hidbus
    ├── hkbd1  (keyboard) ✓
    ├── hms1   (trackpad = basic mouse only, no gestures) ✗
    └── hcons0 (consumer control / media keys) ✓

Verify It Worked

dmesg | grep wsp
# Expected: wsp0 on uhub5
#           wsp0: <Apple Inc. Apple Internal Keyboard / Trackpad> on usbus2

ls /dev/wsp0 # should exist

rc.conf — moused

devmatch_blacklist="ums.ko bcm5974.ko"
moused_enable="YES"
moused_port="/dev/wsp0"
moused_flags="-t auto"

X11 Touchpad Config — libinput

Section "InputClass"
    Identifier "Apple Trackpad"
    MatchProduct "Apple Inc. Apple Internal Keyboard / Trackpad"
    Driver "libinput"
    Option "Tapping"           "on"
    Option "TappingButtonMap"  "lrm"
    Option "NaturalScrolling"  "true"
    Option "ScrollMethod"      "twofinger"
    Option "ClickMethod"       "clickfinger"
    Option "DisableWhileTyping" "true"
EndSection

# Prevent libinput from trying to open /dev/wsp0 directly — moused owns it
# and libinput segfaults (null deref) instead of returning "device busy"
Section "InputClass"
    Identifier "Ignore wsp0"
    MatchDevicePath "/dev/wsp0"
    Option "Ignore" "true"
EndSection

TappingButtonMap lrm: 1-finger tap = left, 2-finger tap = right, 3-finger = middle. Physical 2-finger click = right click. The second block matters even though wsp isn't fighting usbhid for the device — without it, libinput crashes Xorg with SIGSEGV whenever it probes /dev/wsp0 while moused already holds it.

wsp Sensitivity Tuning

hw.usb.wsp.enable_single_tap_clicks=1
hw.usb.wsp.pressure_tap_threshold=25
hw.usb.wsp.pressure_untouch_threshold=20
hw.usb.wsp.z_factor=6    # scroll speed (lower = slower)
hw.usb.wsp.scale_factor=6  # cursor speed (higher = faster)

Recurring Issue — USB Disconnect After Days of Uptime

After several days of uptime the trackpad goes dead — even at the SDDM login screen, before any user session starts. If it happens mid-session, Xorg can crash outright (SIGABRT on device removal), which can crash-loop kwin/plasmashell and, if the saved KWin session has stale window entries, respawn a storm of Konsole windows on every crash-loop restart.

Root cause: the SMSC 2.0 hub (ugen2.3, vendor 0x0424 product 0x2513) that hosts the internal Apple devices on usbus2 enters pwr=SAVE (USB autosuspend) and eventually drops the trackpad (ugen2.9, vendor 0x05ac product 0x0252) entirely. Per usbconfig(8), power_save is the default for all USB hubs — nothing has to go wrong first, the hub starts every boot in this state.

Fix, part 1 — boot-time rc.d script (does the real work)

#!/bin/sh
#
# PROVIDE: apple_usb_power
# REQUIRE: FILESYSTEMS
# KEYWORD: nojail

. /etc/rc.subr

name="apple_usb_power"
rcvar="apple_usb_power_enable"
start_cmd="apple_usb_power_start"

apple_usb_power_start()
{
	for dev in $(usbconfig list | cut -d: -f1); do
		desc=$(usbconfig -d "${dev}" dump_device_desc 2>/dev/null)
		if echo "${desc}" | grep -q "idVendor = 0x0424" && \
		   echo "${desc}" | grep -q "idProduct = 0x2513"; then
			usbconfig -d "${dev}" power_on
		fi
		if echo "${desc}" | grep -q "idVendor = 0x05ac" && \
		   echo "${desc}" | grep -q "idProduct = 0x0252"; then
			usbconfig -d "${dev}" power_on
		fi
	done
}

load_rc_config $name
run_rc_command "$1"

Matches devices by vendor:product ID rather than a hardcoded ugen name, so it survives renumbering across reboots. Enable and test:

sudo chmod 555 /usr/local/etc/rc.d/apple_usb_power
sudo sysrc apple_usb_power_enable="YES"
sudo service apple_usb_power start
sudo usbconfig list | grep -E "ugen2\.[39]"   # hub should now show pwr=ON

Fix, part 2 — devd rule (secondary, catches genuine runtime reconnects)

notify 100 {
    match "system" "USB";
    match "subsystem" "DEVICE";
    match "type" "ATTACH";
    match "vendor" "0x0424";
    match "product" "0x2513";
    action "usbconfig -d $device-name power_on";
};

notify 100 {
    match "system" "USB";
    match "subsystem" "DEVICE";
    match "type" "ATTACH";
    match "vendor" "0x05ac";
    match "product" "0x0252";
    action "usbconfig -d $device-name power_on";
};

If it disconnects again despite both fixes

sudo usbconfig list | grep -E "ugen2\.[39]"       # check current pwr= state
sudo service apple_usb_power start                # force power_on immediately
sudo service moused restart
sudo service sddm restart                          # if Xorg crashed

If power_on still doesn't hold across days, add hw.usb.no_suspend_wait=1 to /etc/sysctl.conf as a more aggressive suspend-prevention measure. If Xorg crashing triggers a KWin crash-loop, a bloated saved session (~/.config/session/kwin_saved*) can respawn dozens of Konsole windows on every restart — fix with pkill -9 -u joe konsole then clear/trim that session file.

2. Display / GPU — i915kms

Intel HD Graphics 4000 (Ivy Bridge) via the i915kms Linux KMS compatibility driver. GPU acceleration works for KDE Plasma 6.

i915kms_load="YES"

On boot you'll see the VT switch from efifb to drmfb in dmesg — this is normal. Xorg requires /dev/dri/card0 which this module creates; without it X fails with a fatal error at a black screen.

Major Version Upgrades — drm-kmod Breakage

i915kms is not part of base — it ships in the graphics/drm-66-kmod package, built against one specific kernel ABI (__FreeBSD_version / kern.osreldate). A freebsd-update version bump (e.g. 15.0 → 15.1) installs a new kernel but does not touch /boot/modules/*.ko — those come from pkg, not freebsd-update.

1 — In single-user mode, disable i915kms so the box boots

sysrc -f /boot/loader.conf -x i915kms_load
sysrc kld_list=""

Reboot to confirm the stale module was the cause — you'll be back to multi-user, but with no GPU driver: Xorg fails with open /dev/dri/card0: No such file or directory (expected, no black-screen mystery).

2 — Finish the freebsd-update upgrade

A panic on first reboot usually means you're mid-upgrade (kernel installed, userland not yet):

freebsd-version -k   # kernel version
freebsd-version -u   # userland version — if these differ, finish the upgrade:
sudo freebsd-update install
sudo pkg update -f
sudo freebsd-update install   # second pass, installs userland

3 — Get a kernel-matched drm-kmod build

pkg upgrade alone will likely report drm-kmod as "up to date" even when it isn't. drm-66-kmod is provided by two repos — the general FreeBSD-ports repo (lags behind, keeps an older build) and FreeBSD-ports-kmods (tracks the current kernel ABI). Plain pkg upgrade can silently pick the stale one.

sysctl kern.osreldate                              # e.g. 1501000
pkg info drm-66-kmod | grep FreeBSD_version         # must match osreldate above
pkg rquery -r FreeBSD-ports-kmods "%n %v" drm-66-kmod

If the installed FreeBSD_version doesn't match kern.osreldate, force it explicitly:

sudo pkg upgrade -r FreeBSD-ports-kmods drm-66-kmod drm-kmod

4 — Re-enable and test

No fresh reboot needed to test — kldload does the same attach as boot-time loading, and lets you watch dmesg live instead of guessing from a cold boot:

sudo kldload i915kms
dmesg | tail -20   # look for "[drm] Initialized i915 ..." and "Replacing driver efifb with drmfb"
ls /dev/dri/        # card0 should exist
sudo service sddm restart

Once confirmed working, make it permanent:

sudo sysrc -f /boot/loader.conf i915kms_load="YES"
sudo sysrc kld_list="i915kms"

3. Keyboard Fix — evdev for kbdmux

The keyboard works fine in TTY but feels laggy in X. The cause: libinput applies debounce/fuzz filtering to the kbdmux (keyboard multiplexer) device, making keys feel sluggish and unresponsive.

Fix: force the evdev driver for the kbdmux device instead of libinput.

Section "InputClass"
    Identifier   "kbdmux evdev"
    MatchProduct "System keyboard multiplexer"
    Driver       "evdev"
EndSection

# KBDMUX(2) + HID_MOUSE(4) + HID_KBD(8) = 14
kern.evdev.rcpt_mask=14

4. Screen Brightness — Intel Backlight Device

Brightness is controlled via /dev/backlight/intel_backlight0, exposed by i915kms. The acpi_video sysctl nodes (hw.acpi.video.lcd0.*) exist but do not control the actual display on this hardware — use the backlight device directly.

Load acpi_video at boot (still needed for F-key keysym generation, even though the sysctl path doesn't work for brightness):

acpi_video_load="YES"

The video group owns the backlight device — no sudo needed:

pw groupmod video -m joe

Manual control

# Check current level (0–100)
backlight -f /dev/backlight/intel_backlight0

# Set level
backlight -f /dev/backlight/intel_backlight0 70

F1 / F2 hotkeys — let KDE Powerdevil handle them

The F1/F2 keys emit XF86MonBrightnessDown/Up. Let KDE Powerdevil handle these directly — don't grab them in xbindkeys. Powerdevil's built-in "Increase/Decrease Screen Brightness" shortcuts are bound to Monitor Brightness Up/Down by default and drive this backlight correctly (verified: KDE's org.kde.ScreenBrightness DBus service writes /dev/backlight/intel_backlight0).

~/bin/brightness.sh is kept for manual/CLI control:

#!/bin/sh
BACKLIGHT="/dev/backlight/intel_backlight0"
STEP=10
current=$(backlight -f "$BACKLIGHT" | awk '{print $2}')
case "$1" in
    up)   new=$(( current + STEP )); [ "$new" -gt 100 ] && new=100 ;;
    down) new=$(( current - STEP )); [ "$new" -lt  10 ] && new=10  ;;
esac
backlight -f "$BACKLIGHT" "$new"

5. Keyboard Backlight — asmc

The Apple SMC driver (asmc) is supported out of the box on FreeBSD 15 for MacBookPro9,2. No patching required.

asmc_load="YES"

# Keyboard backlight brightness — range 0–255
dev.asmc.0.light.control=100

Manual control

sysctl dev.asmc.0.light.control=150

~/bin/kbd-brightness.sh — F5 / F6 hotkeys

#!/bin/sh
STEP=25
current=$(sysctl -n dev.asmc.0.light.control)
case "$1" in
    up)   new=$(( current + STEP )); [ "$new" -gt 255 ] && new=255 ;;
    down) new=$(( current - STEP )); [ "$new" -lt   0 ] && new=0   ;;
esac
sysctl dev.asmc.0.light.control="$new" >/dev/null

Bound in ~/.xbindkeysrc to XF86KbdBrightnessDown / XF86KbdBrightnessUp.

6. Function Keys — xkbcomp + xbindkeys

The MacBook top-row keys send plain F1–F12 keycodes in X. They must be remapped to XF86 media keysyms. Use xkbcomp — not xmodmap. xmodmap wipes XKB level 5 (the Ctrl+Alt level) where XF86Switch_VT_N lives, which breaks Ctrl+Alt+Fn VT switching entirely.

~/.xkb_macbook.xkb

Generate a base keymap and edit the FK01–FK12 entries to put XF86 keysyms at levels 1–4 and preserve XF86Switch_VT_N at level 5:

# Generate base keymap from current X state
setxkbmap us
xkbcomp :0 ~/.xkb_macbook.xkb

# Edit FK01–FK12 entries to read, e.g.:
# key <FK01> { type="CTRL+ALT", symbols[Group1]= [
#   XF86MonBrightnessDown, XF86MonBrightnessDown,
#   XF86MonBrightnessDown, XF86MonBrightnessDown,
#   XF86Switch_VT_1 ] };
# ... and so on for FK02–FK12

# Apply immediately
xkbcomp ~/.xkb_macbook.xkb $DISPLAY

The CTRL+ALT key type maps: level 1 = unmodified, level 5 = Ctrl+Alt. The X server sees XF86Switch_VT_N on Ctrl+Alt+Fn and switches VT directly — no daemon needed.

Add to KDE autostart:

[Desktop Entry]
Type=Application
Name=xkb macbook keys
Exec=xkbcomp /home/joe/.xkb_macbook.xkb %DISPLAY
Hidden=false
NoDisplay=false
X-GNOME-Autostart-enabled=true

~/.xbindkeysrc

# NOTE: F1 / F2 screen brightness is deliberately NOT bound here —
# KDE Powerdevil owns XF86MonBrightnessDown/Up. See step 4.

# Keyboard backlight — F5 / F6
"$HOME/bin/kbd-brightness.sh down"
  XF86KbdBrightnessDown

"$HOME/bin/kbd-brightness.sh up"
  XF86KbdBrightnessUp

# Media keys — F7 / F8 / F9
"playerctl previous"
  XF86AudioPrev

"playerctl play-pause"
  XF86AudioPlay

"playerctl next"
  XF86AudioNext

# F3 — Konsole
"konsole"
  XF86Launch1

# F4 — Dolphin
"dolphin"
  XF86LaunchA

# Right-click workaround (Ctrl + click)
"xdotool click 3"
  control + b:1

7. Volume Keys & KDE OSD

The Cirrus Logic CS4206 is detected automatically by snd_hda. Volume key control (F10/F11/F12) and the KDE OSD popup are handled through PulseAudio and the plasma6-plasma-pa Plasma applet.

Install plasma-pa

sudo pkg install plasma6-plasma-pa

Log out and back in. The volume icon will appear in the system tray and the OSD will show on F10/F11/F12.

How it works

xkbcomp maps F10/F11/F12 to XF86AudioMute / XF86AudioLowerVolume / XF86AudioRaiseVolume at XKB level 1. KDE's audioshortcutsservice (a kded module) registers these as global shortcuts via kglobalaccel, routes changes through PulseAudio, and plasma-pa displays the OSD.

F10/F11/F12
  → xkbcomp (XF86Audio* keysyms at XKB level 1)
  → kglobalaccel → audioshortcutsservice (kded)
  → PulseAudio → plasma-pa applet
  → volume OSD popup ✓

Manual volume via mixer: mixer vol.volume=+5% (relative) · mixer vol.volume=0.75 (absolute) · mixer vol.mute=toggle

8. Audio

The Cirrus Logic CS4206 is detected automatically — no configuration needed beyond the default FreeBSD install.

cat /dev/sndstat   # list devices and their state
mixer              # check/set volume

9. WiFi — BCM4331 Not Supported + USB Dongle

Setup

wlans_rtwn0="wlan0"
ifconfig_wlan0="WPA DHCP"

network={
    ssid="YourNetworkName"
    psk="YourPassword"
}

Verify

dmesg | grep rtwn
# rtwn0: <Realtek RTL8821AU...>
# rtwn0: MAC/BB RTL8821AU, RF 6052 1T1R

ifconfig wlan0  # shows IP, SSID, channel

10. Fans, Heat & CPU C-States

The machine runs noticeably warm at idle — cores at 58–62°C with the desktop doing nothing and the single exhaust fan at its floor speed. Not dangerous (no thermal throttling has ever occurred; Tjmax is 105°C), but a hot chassis and less headroom before the fan has to ramp under load.

Root cause — CPU never leaves C1 on AC power

FreeBSD's default on AC power leaves cx_lowest at C1, the shallowest idle state. /etc/defaults/rc.conf ships performance_cx_lowest="NONE" ("don't touch it"), so the kernel default of C1 stands and the cores never enter deep sleep — most of the CPU package stays powered around the clock. The battery-power default (economy_cx_lowest="Cmax") is already correct; this is an AC-only problem, which is why the machine ran hot while plugged in.

sysctl dev.cpu.0.cx_supported dev.cpu.0.cx_lowest dev.cpu.0.cx_usage
# dev.cpu.0.cx_supported: C1/1/1 C2/2/148 C3/3/198   <- hardware supports C1-C3
# dev.cpu.0.cx_lowest: C1                            <- but only C1 is allowed
# dev.cpu.0.cx_usage: 100.00% 0.00% 0.00%            <- so it's all C1, all the time

The fix

Allow the deepest supported C-state. Test live first — instantly reversible with =C1, resets on reboot:

sudo sysctl hw.acpi.cpu.cx_lowest=Cmax

Then persist it:

sudo sysrc performance_cx_lowest="Cmax"
sudo sysrc economy_cx_lowest="Cmax"

Measured result

Idle core temps dropped from 58–62°C to 48–51°C (~10°C) within about two minutes, with idle residency settling around 27% C1 / 8% C2 / 65% C3. powerd frequency scaling (1200–2500 MHz) kept working normally throughout.

Fan reference (asmc)

The single exhaust fan is controlled automatically by the Apple SMC via asmc(4). Useful sysctls:

sysctl dev.asmc.0.fan.0        # speed, targetspeed, minspeed/maxspeed (3000/6200), manual
sysctl dev.asmc.0.temp         # all SMC temperature sensors (CPU, GPU, battery, palm rest…)
sysctl dev.cpu.0.temperature   # per-core temps via coretemp(4)

dev.asmc.0.fan.0.minspeed=3000 is pinned in /etc/sysctl.conf (stock floor is lower). SMC auto control has proven fine once C-states are fixed; a manual fan curve (fan.0.manual=1 + fan.0.targetspeed) exists as a lever but hasn't been needed.

11. Lid Close — Screen Blank

S3 sleep is broken on this hardware. The solution is to blank the screen on lid close using a UPower DBus watcher, with KDE Powerdevil's lid action disabled entirely.

Why the obvious approaches don't work

devd rules: ACPI lid events do not reach /dev/devctl on this machine — closing the lid produces nothing on devctl. A devd rule matching system="ACPI" subsystem="Lid" is inert here.

KDE Powerdevil lid actions: All broken or wrong. The Plasma 6 LidAction values mapped by trial and error:

Fix — Part 1: Disable KDE lid action

[AC][SuspendAndShutdown]
AutoSuspendAction=0
LidAction=0

[Battery][SuspendAndShutdown]
LidAction=0

Fix — Part 2: UPower DBus watcher

UPower already tracks lid state (KDE uses it internally). This script taps that existing signal:

#!/bin/sh
export DISPLAY=:0

dbus-monitor --system \
    "type='signal',interface='org.freedesktop.DBus.Properties',member='PropertiesChanged',path='/org/freedesktop/UPower'" |
awk '
/string "LidIsClosed"/ { wait=1; next }
wait && /boolean true/  { system("xset dpms force off"); wait=0 }
wait && /boolean false/ { wait=0 }
'

chmod +x /usr/local/bin/lid-screen.sh

[Desktop Entry]
Type=Application
Name=Lid Screen Blank
Exec=/usr/local/bin/lid-screen.sh
Restart=always
X-KDE-autostart-phase=1

Result

• Close lid → screen goes dark immediately
• Open lid → DPMS wakes on first mouse movement or keypress
• No input grab, no lock screen, no partial suspend — everything is clickable after open

12. KDE Plasma 6 Desktop

Minimal KDE Plasma 6 install with SDDM display manager, on top of a stock X.Org base. Wayland does not work on this hardware — use the X11 session.

Install X.Org

pkg install xorg

The full x11/xorg metaport (server, drivers, libraries, fonts, client utilities) — the Handbook-recommended route. x11/xorg-minimal is a lighter alternative but needs fonts and xf86-input-libinput added by hand. Requires the i915kms driver from step 2 to be loaded first, or X fails with "no screens found". No xorg.conf is needed — the only local X config is the trackpad drop-ins in /usr/local/etc/X11/xorg.conf.d/ (step 1).

Enable D-Bus

sysrc dbus_enable="YES"
service dbus start

Plasma uses D-Bus for its message bus — hard requirement. (The Handbook's /proc-mount requirement applies to GNOME/MATE/Cinnamon/LXQt, not Plasma; this box mounts procfs only for OpenJDK.)

Install Plasma 6 (minimal)

The Handbook route is the plasma6-plasma metaport (all of Plasma) or kde (Plasma plus the full KDE application suite); this machine uses a hand-picked minimal set instead:

pkg install plasma6-kwin-x11 plasma6-plasma-workspace plasma6-plasma-desktop \
  plasma6-systemsettings plasma6-kscreenlocker plasma6-polkit-kde-agent-1 \
  plasma6-xdg-desktop-portal-kde plasma6-plasma-activities \
  plasma6-plasma-activities-stats plasma6-breeze plasma6-libkscreen \
  plasma6-kscreen plasma6-powerdevil plasma6-ksystemstats \
  plasma6-libksysguard plasma6-kde-cli-tools plasma6-kdecoration \
  plasma6-kactivitymanagerd plasma6-kglobalacceld plasma6-milou \
  plasma6-plasma-pa sddm konsole

KDE Apps

pkg install dolphin ark kate gwenview okular plasma6-spectacle networkmgr

Enable SDDM

sysrc lightdm_enable="NO"
sysrc sddm_enable="YES"

Why Wayland Specifically Fails: kwin_wayland SIGSEGV

Beyond the missing Qt Wayland platform plugin, kwin_wayland also hits a real crash: a Qt 6.11.1 bug where QKdeTheme::createKdeTheme() SIGSEGVs in handleThemeChanged() because KDE_SESSION_VERSION is set in the environment (injected by SDDM for KDE sessions) before a required global is initialized.

Fix: replace the real binary with a wrapper that strips the offending env vars before exec'ing it:

sudo mv /usr/local/bin/kwin_wayland /usr/local/bin/kwin_wayland.bin
sudo tee /usr/local/bin/kwin_wayland <<'EOF'
#!/bin/sh
exec env -u KDE_SESSION_VERSION -u KDE_FULL_SESSION /usr/local/bin/kwin_wayland.bin "$@"
EOF
sudo chmod 755 /usr/local/bin/kwin_wayland

XLibre — Alternative X11 Server

XLibre is a fork of X.Org Server (split off June 2025) that's since picked up active CVE backporting, code cleanup, and a few features X.Org itself doesn't ship by default — notably TearFree enabled out of the box and atomic modesetting support. It's packaged for FreeBSD (x11/xlibre, x11-servers/xlibre-server) and GhostBSD has adopted it as the default X server for MATE/Xfce. Confirmed working on this exact hardware (Intel HD Graphics 4000 IVB, i915kms, Apple internal trackpad) on 2026-07-18: same modesetting + glamor + libinput stack as stock X.Org, correctly picks up the existing xorg.conf.d quirks (SWcursor, the /dev/wsp0 libinput-segfault ignore rule from step 1), and the trackpad works identically. You now have two solid X11 choices on this machine.

Switch to XLibre

pkg install xlibre-server xlibre-xf86-input-libinput xlibre-xf86-video-intel \
  xlibre-xf86-input-evdev

xlibre-xf86-input-evdev wasn't part of the original 2026-07-18 switch — it replaced stock xf86-input-evdev two days later, needed because 10-keyboard.conf (see the Keyboard Fix section) hard-requires the evdev driver name and stock xf86-input-evdev can't coexist with the rest of the XLibre stack.

This removes xorg-server, plasma6-plasma-desktop, and the stock driver packages as a side effect of the conflict above. To get Plasma running on top of XLibre again, its dependency on stock xf86-input-libinput/xorg-server has to be bypassed — the input driver is something only the X server loads, XLibre already ships its own equivalent (xlibre-xf86-input-libinput), so Plasma's own binaries (kwin_x11, plasmashell) never actually need the stock package present:

# fetch plasma6-plasma-desktop + full dependency tree as local files
mkdir -p /tmp/plasma-fetch
pkg fetch -d -y -o /tmp/plasma-fetch plasma6-plasma-desktop

# drop the two packages that would drag stock X.Org back in
rm /tmp/plasma-fetch/All/Hashed/xorg-server-*.pkg
rm /tmp/plasma-fetch/All/Hashed/xf86-input-libinput-*.pkg

# install everything else, accepting that one now-missing dependency
find /tmp/plasma-fetch -name '*.pkg' -print0 | xargs -0 pkg add -M

Keeping Plasma through future upgrades

The bypass above is not a one-time fix. The plasma6-plasma-desktop installed via pkg add -M has no recorded xf86-input-libinput dependency, but the repository's copy still does — so the next time the port is revbumped, a routine pkg upgrade tries to upgrade it, hits the xorg-server conflict again, and its second solver pass resolves it by removing plasma6-plasma-desktop altogether. This happened on 2026-08-13 with the 6.6.6 → 6.6.6_1 revbump.

The tell is in the second plan pkg prints after Cannot solve problem using SAT solver, trying another plan:

Installed packages to be REMOVED:
        plasma6-plasma-desktop: 6.6.6

To make pkg refuse instead of silently dropping it:

pkg lock -y plasma6-plasma-desktop

A locked package cannot be removed by the solver — pkg delete on it exits 7 with locked or vital and may not be removed — so the conflict surfaces as an error you have to deal with rather than as a package that quietly vanishes. The cost is that Plasma revbumps no longer apply automatically: when you want one, pkg unlock plasma6-plasma-desktop, redo the fetch / strip / pkg add -M sequence above for the new version, then lock it again.

Verify which server is running

xdpyinfo | grep vendor        # "vendor string: XLibre"
grep -m1 "X Server" /var/log/Xorg.0.log  # "XLibre X Server 1.25.1.9" — note the legacy "1." prefix

Revert to stock X.Org + Plasma

pkg install xorg-server plasma6-plasma-desktop xorg-drivers \
  xf86-video-intel xf86-input-libinput xf86-input-mouse \
  xf86-input-keyboard xf86-video-scfb xf86-video-vesa

Same removal-as-side-effect behavior in reverse: pkg will remove the xlibre-* packages to resolve the conflict and restore the stock stack exactly.

Known bug — modesetting page-flip SIGBUS (25.1.x)

The XLibre 25.1.x line (25.1.8 when this was diagnosed; still 25.1.9, the current FreeBSD port version, as of 2026-08-13) crashes the X server to a bus error during normal desktop use — anywhere from every few minutes to a few times a session — dumping you back to SDDM. The Xorg log ends with a backtrace in modesetting_drv.so and Caught signal 10 (Bus error) (the kernel logs it as signal 6 because XLibre catches SIGBUS and then abort()s). The fault is in the modesetting driver's page-flip / TearFree present path (ms_present_check_flip), which XLibre exercises hard because it enables both PageFlip and TearFree by default where stock X.Org leaves them off.

The real fix is on the XLibre 25.2.x branch (commit 2e31b2fdc adds a geometry guard to ms_present_check_flip, plus null-deref guards) and was never backported to the 25.1.x release line — so the June-2026 25.1.8 tag still lacks a February-2026 fix. The 25.1.9 changelog was checked on 2026-08-13 and contains no modesetting or present-path changes either, so upgrading within 25.1.x does not help. Until the port reaches 25.2.x, disable page flips in the modesetting Device section (/usr/local/etc/X11/xorg.conf.d/15-modesetting-swcursor.conf on this box):

Section "Device"
    Identifier  "Intel HD 4000"
    Driver      "modesetting"
    Option      "SWcursor" "true"
    Option      "PageFlip" "false"
EndSection

This fully removes the crashing path and is reversible by deleting the one line. The cost is flip-based vsync for fullscreen video/games; the composited desktop is unaffected. Option "TearFree" "false" is a lighter alternative but doesn't close the fullscreen-app flip path as completely. Remove the workaround once x11-servers/xlibre-server updates to 25.2.x.

13. Webcam

The Apple FaceTime HD camera uses the webcamd userspace driver.

pkg install webcamd
sysrc webcamd_enable="YES"
pw groupmod webcamd -m joe
service webcamd start

Packages Installed

Final Configuration Files

kern.geom.label.disk_ident.enable="0"
kern.geom.label.gptid.enable="0"
zfs_load="YES"
i915kms_load="YES"
wsp_load="YES"
asmc_load="YES"
acpi_video_load="YES"
coretemp_load="YES"
hw.usb.usbhid.enable=0
kern.hz=100
vfs.zfs.arc_max="4294967296"

hostname="zippy"
ifconfig_bge0="DHCP"
ifconfig_bge0_ipv6="inet6 accept_rtadv"
sshd_enable="YES"
powerd_enable="YES"
powerd_flags="-b adaptive -a adaptive -i 25 -r 85 -N"
performance_cx_lowest="Cmax"
economy_cx_lowest="Cmax"
moused_nondefault_enable="NO"
dumpdev="NO"
zfs_enable="YES"
dbus_enable="YES"
lightdm_enable="NO"
devmatch_blacklist="ums.ko bcm5974.ko"
moused_enable="YES"
moused_port="/dev/wsp0"
moused_flags="-t auto"
webcamd_enable="YES"
sddm_enable="YES"
apple_usb_power_enable="YES"
wlans_rtwn0="wlan0"
ifconfig_wlan0="WPA DHCP"
cupsd_enable="YES"
avahi_daemon_enable="NO"

vfs.zfs.vdev.min_auto_ashift=12
kern.evdev.rcpt_mask=14       # KBDMUX(2)+HID_MOUSE(4)+HID_KBD(8) — do NOT use 12
dev.asmc.0.light.control=100
hw.usb.wsp.enable_single_tap_clicks=1
hw.usb.wsp.pressure_tap_threshold=25
hw.usb.wsp.pressure_untouch_threshold=20
hw.usb.wsp.z_factor=6
hw.usb.wsp.scale_factor=6
hw.pci.do_power_nodriver=3
dev.asmc.0.fan.0.minspeed=3000
kern.coredump=0

Section "InputClass"
    Identifier   "kbdmux evdev"
    MatchProduct "System keyboard multiplexer"
    Driver       "evdev"
EndSection

Section "InputClass"
    Identifier "Apple Trackpad"
    MatchProduct "Apple Inc. Apple Internal Keyboard / Trackpad"
    Driver "libinput"
    Option "Tapping"           "on"
    Option "TappingButtonMap"  "lrm"
    Option "NaturalScrolling"  "true"
    Option "ScrollMethod"      "twofinger"
    Option "ClickMethod"       "clickfinger"
    Option "DisableWhileTyping" "true"
EndSection

Section "InputClass"
    Identifier "Ignore wsp0"
    MatchDevicePath "/dev/wsp0"
    Option "Ignore" "true"
EndSection

See USB Disconnect fix for /usr/local/etc/rc.d/apple_usb_power and /usr/local/etc/devd/apple-usb-power.conf — omitted here for space.

KDE Autostart (~/.config/autostart/)

xmodmap.desktop          → xkbcomp /home/joe/.xkb_macbook.xkb %DISPLAY
xbindkeys.desktop        → xbindkeys
kbd-backlight-daemon.desktop → /home/joe/bin/kbd-backlight-daemon.sh
lid-screen.desktop       → /usr/local/bin/lid-screen.sh

Scripts (~/bin/)

brightness.sh       → manual/CLI screen brightness via /dev/backlight/intel_backlight0
                      (NOT bound to F1/F2 — KDE Powerdevil owns those, see step 4)
kbd-brightness.sh   → F5/F6 keyboard backlight via dev.asmc.0.light.control

F10/F11/F12 volume is handled entirely by KDE's audioshortcutsservice via kglobalaccel — no shell script involved. Do not bind these in xbindkeys.

How to Revert

Revert trackpad to basic mouse only

1. Remove hw.usb.usbhid.enable=0 from /boot/loader.conf
2. Reboot

The trackpad falls back to hms (basic mouse, no gestures). /dev/wsp0 won't exist — moused fails silently — but the trackpad still works as a basic mouse through X11 evdev. hcons returns, restoring full HID media key events.

Revert X11 touchpad config to synaptics

Replace /usr/local/etc/X11/xorg.conf.d/20-apple-touchpad.conf with:

Section "InputClass"
    Identifier "Apple Touchpad synaptics"
    MatchProduct "Apple Inc. Apple Internal Keyboard / Trackpad Touchpad"
    Driver "synaptics"
    Option "Protocol"       "event"
    Option "TapButton1"     "1"
    Option "TapButton2"     "3"
    Option "RTCornerButton" "3"
    Option "VertEdgeScroll" "on"
    Option "FingerLow"      "1"
    Option "FingerHigh"     "5"
    Option "MinPressure"    "0"
    Option "MaxPressure"    "255"
EndSection

Troubleshooting

/dev/wsp0 doesn't exist after reboot

dmesg | grep wsp
sudo usbconfig list | grep -i trackpad
kldstat | grep wsp

Verify hw.usb.usbhid.enable=0 and wsp_load="YES" are in /boot/loader.conf.

Trackpad moves but no tap or scroll

xinput list  # Apple trackpad should appear as a pointer device
grep -i "libinput\|trackpad\|apple" /var/log/Xorg.0.log

Cursor / scroll speed adjustment

sysctl hw.usb.wsp.scale_factor=8   # cursor (default 6, higher = faster)
sysctl hw.usb.wsp.z_factor=4       # scroll (default 6, lower = slower)

Add to /etc/sysctl.conf to persist.

Brightness keys not working

kldstat | grep acpi_video
ls -la /dev/backlight/intel_backlight0

Ensure acpi_video_load="YES" and i915kms_load="YES" are in /boot/loader.conf. Ensure your user is in the video group: id | grep video.

Keyboard backlight not responding

kldstat | grep asmc
sysctl dev.asmc.0.light.control=200

Ensure asmc_load="YES" is in /boot/loader.conf.

Ctrl+Alt+Fn VT switching not working

Almost always caused by xmodmap wiping the XKB keymap. Verify the correct keymap is applied:

xkbcomp :0 /tmp/check.xkb && grep -A3 "key <FK01>" /tmp/check.xkb

You should see XF86Switch_VT_1 at position 5. If you see NoSymbol there, xmodmap has run and wiped level 5. Reapply:

xkbcomp ~/.xkb_macbook.xkb $DISPLAY

Volume OSD not appearing (F10/F11/F12)

pkg info plasma6-plasma-pa
ps aux | grep pulse
grep -i "XF86Audio" ~/.xbindkeysrc  # must return nothing

If any XF86Audio* bindings appear in xbindkeys, remove them — then reload:

killall xbindkeys && xbindkeys

After opening lid — desktop visible but nothing clickable

KDE Powerdevil's lid action grabbed X input without a visible UI. Check:

grep LidAction ~/.config/powerdevilrc

Should be LidAction=0. If it's 64, that's the culprit. Fix: set LidAction=0 in all [SuspendAndShutdown] sections and use the UPower DBus script instead. If stuck right now: type your login password and press Enter — kscreenlocker may be running invisibly and that may dismiss it.

WiFi — harmless dmesg noise

ieee80211_sta_join: BSS ...: 2GHz channel, VHT info; ignoring — router is advertising 802.11ac on 2.4GHz (invalid). Driver ignores it, connects normally. Not a bug.

Machine idles hot (~60°C) with the fan at floor speed

sysctl dev.cpu.0.cx_usage  # 100.00% 0.00% 0.00% = stuck in C1, the AC-power default

Fix with sudo sysctl hw.acpi.cpu.cx_lowest=Cmax and persist via performance_cx_lowest="Cmax" in rc.conf. Expect ~10°C cooler idle. See Fans, Heat & C-States.

X server crashes to SDDM every few minutes (bus error in modesetting)

XLibre 25.1.x's modesetting page-flip bug. The Xorg log ends with a modesetting_drv.so backtrace and Caught signal 10 (Bus error). Disable page flips with Option "PageFlip" "false" in the modesetting Device section — see Known bug — modesetting page-flip SIGBUS. The ProcXkbGetKbdByName childbuf size mismatch warnings in the log are unrelated and harmless.

Screen brightness resets after the display dims on idle

The screen dims on idle and moving the mouse restores it to the wrong level, not the one you set. Something is writing the backlight outside KDE (an xbindkeys→brightness.sh binding, or a direct backlight(8) write), so Powerdevil re-asserts its own stored value on undim. Fix: hand the brightness keys to KDE — remove the XF86MonBrightnessDown/Up bindings from ~/.xbindkeysrc. See Screen Brightness → F1/F2 hotkeys. Log out and back in afterwards so KDE grabs the freed keys — don't use the blockGlobalShortcuts toggle on its own, it kills xbindkeys' other keys (F3–F6) until xbindkeys is restarted.

Kernel panics on boot right after a freebsd-update version bump

This is drm-kmod ABI breakage — see Major Version Upgrades for the full recovery procedure. Short version: boot single-user, sysrc -f /boot/loader.conf -x i915kms_load and sysrc kld_list="" to get back to multi-user, finish the freebsd-update, then pkg upgrade -r FreeBSD-ports-kmods drm-66-kmod before re-enabling i915kms.

Trackpad dead again after days of uptime, or dead at the SDDM login screen

See Recurring Issue — USB Disconnect. Quick check: sudo usbconfig list | grep -E "ugen2\.[39]" — if the hub shows pwr=SAVE, run sudo service apple_usb_power start to force it back on immediately.

Wayland session (kwin_wayland) SIGSEGVs again after a KDE package upgrade

pkg upgrade of plasma6-kwin overwrites /usr/local/bin/kwin_wayland with the raw binary, silently dropping the wrapper. See KDE Plasma 6 → Why Wayland Specifically Fails for the reapply command. Not urgent day-to-day since this machine runs Session=plasmax11 (X11).