### System details

Lenovo ThinkPad X9-15p Gen 1 (21VVCTO1WW), BIOS N4QET31W 1.10 (03/12/2026) — Intel Core Ultra X7 358H (Panther Lake), Arc B390 (`xe`, Mesa 26.2.2); IPU7 `8086:b05d`; Omarchy 4.0.3-1; kernel `7.2.3-arch1-3`; `intel-ipu7-camera 1.0.5-1`; `v4l2-relayd 0.2.0-2.2`; `libcamera 0.7.2-4`; `imx471-dkms-git r28.a56bd1b-1` (AUR).

The fitted sensor is a **Sony IMX471** (ACPI `TBE20A0`), not an OV08X40.

### What's wrong?

`install/hardware/intel/ipu7-camera.sh` installs the package when ACPI HID `OVTI08F4` is present:

```bash
if grep -q "OVTI08F4" /sys/bus/acpi/devices/*/hid 2>/dev/null; then
  omarchy-pkg-add intel-ipu7-camera
fi
```

This board's ACPI lists four camera HIDs while only one sensor is fitted — `OVTI08F4`, `OVTI01AF`, `TBE20A0`, `TBE20A1` — so the check passes even though the sensor that actually binds is the IMX471 behind `TBE20A0`.

The package then cannot drive it: `/etc/camera/ipu75xa/pipe_scheduler_profiles.json` defines graphIds **100000, 100002 and 100200**, but the IMX471 pipeline requests **100005**.

This is separate from the jsoncpp breakage in #10837 — that has to be worked around first to reach this failure at all — and separate from #12178, where the needed plugin isn't built. Here `ipu75xa` is the right plugin and it loads fine.

### Evidence

The kernel side is healthy; the sensor binds:

```
intel-ipu7 0000:00:05.0: firmware cpd file: intel/ipu/ipu7ptl_fw.bin
intel-ipu7 0000:00:05.0: Found supported sensor TBE20A0:00
intel-ipu7 0000:00:05.0: Connected 1 cameras
intel_ipu7_isys.isys intel_ipu7.isys.40: bind imx471 0-001a nlanes is 4 port is 0
intel_ipu7_isys.isys intel_ipu7.isys.40: All sensor registration completed.
```

With #10837 worked around so the HAL loads, `icamerasrc` advertises exactly one camera, correctly the IMX471:

```
$ gst-inspect-1.0 icamerasrc | grep -A2 device-name
   (0): imx471-uf
```

The stock `/etc/v4l2-relayd.d/ipu7.conf` asks for `device-name=ov08x40-uf`, which is not in that enum, and the relay logs only:

```
gst_element_set_state: assertion 'GST_IS_ELEMENT (element)' failed
```

Setting `device-name=imx471-uf` clears that. The HAL then loads the correct IMX471 data:

```
CamHAL[INF] CameraParserInvoker: parseSensors: I will Load config file: sensors/imx471-uf.json
CamHAL[INF] CameraSensorsParser: resolveCsiPortAndI2CBus: I2CBus:0-001a <=> CSI Port:0
CamHAL[INF] AiqInitData: aiqb file name IMX471_BBG803N3_PTL.aiqb
```

and then fails on the missing profile:

```
CamHAL[ERR] SchedPolicy: setConfig: no config for graphId 100005
CamHAL[ERR] Scheduler: configurate 100005 error
CamHAL[ERR] CameraDevice: @configure Failed to configure H-Scheduler
CamHAL[ERR] CameraHal: failed to config streams.
streaming task paused, reason not-negotiated (-4)
```

I tried adding a 100005 profile using both executor groupings that already appear in that file (`video_lb=[lbff]` with `video_bb=[bbps,post_1,post_2]`, and `video_lb=[lbff,bbps]` with `video_bb=[post_1,post_2]`). Both clear the scheduler error and then fail on graph topology:

```
CamHAL[ERR] GraphConfig: getPSysContextId: Can't find node, stream 60001, outerNode ctxId 3
CamHAL[ERR] PipeManager: Failed to bind output ports
CamHAL[ERR] ProcessingUnit: @configure configure psys dag failed:-22
```

The package does ship the rest of the IMX471 support — `IMX471_BBG803N3_PTL.aiqb`, `gcss/IMX471_BBG803N3.IPU75XA.bin`, `sensors/imx471-uf.json`, and `imx471-uf-0` in `libcamhal_configs.json` — so the missing scheduler profile looks like an oversight.

### Steps to reproduce

1. Install Omarchy on a ThinkPad X9-15p Gen 1, or any board whose ACPI lists `OVTI08F4` but which fits an IMX471 (`TBE20A0`).
2. Open anything that uses the camera. `/dev/video50` exists but only ever yields a flat black frame.

### Ask

Either ship the IMX471 scheduler profile, or gate installation on the sensor that actually binds rather than on ACPI HID presence.

### Related

- #10837, #11358 — jsoncpp `.26`→`.27` soname break; has to be worked around before reaching this.
- #12178, #7697 — same class: the package is installed by detection onto hardware it cannot serve.

### Workaround

Drive the sensor with libcamera's `simple` pipeline handler and software ISP instead of the package's userspace pipeline. Everything required:

1. **Sensor driver:** install `imx471-dkms-git` (AUR). The IMX471 is not in the mainline kernel and `intel-ipu7-camera` only ships ov08x40. This package also installs the libcamera tuning file `/usr/share/libcamera/ipa/simple/imx471.yaml`, without which the simple IPA falls back to `uncalibrated.yaml`.
2. **Userspace:** install `libcamera` and `pipewire-libcamera` (plus `libcamera-tools` if you want `cam -l` for diagnostics).
3. **Keep `intel-ipu7-camera` installed.** It owns the `ipu7-drivers` DKMS source providing `intel_ipu7`, `intel_ipu7_isys`, `intel_ipu7_psys`, `ipu_bridge` and `ipu_acpi`. Kernel `7.2.3-arch1-3` has in-tree IPU6 but no IPU7, so removing it leaves no IPU7 driver at all.
4. **Stop its userspace pipeline:** mask `intel-ipu7-camera.service` (it is `WantedBy=graphical.target` and starts the rest) and `[email protected]`.
5. **Drop v4l2loopback:** override `/usr/lib/modules-load.d/v4l2-relayd.conf` with an empty `/etc/modules-load.d/v4l2-relayd.conf`, and unload the module, so the unused `/dev/video50` stops appearing as a camera.
6. **Enable the libcamera monitor.** `intel-ipu7-camera` ships `/usr/share/wireplumber/wireplumber.conf.d/disable-libcamera.conf` setting `monitor.libcamera = disabled`; override it with `monitor.libcamera = required` in `~/.config/wireplumber/wireplumber.conf.d/`.
7. **Keep WirePlumber's V4L2 monitor disabled** (`monitor.v4l2 = disabled`). It stalls enumerating the 32 raw `Intel IPU7 ISYS Capture` nodes, which leaves the audio nodes without ports and kills the microphone.
8. **Grant the session access to the ISYS nodes.** `71-ipu7-hide-isys.rules` from the package sets `MODE="0600" GROUP="root"` on them, which blocks libcamera running as a normal user. A later rule has to restore access — on this machine `/etc/udev/rules.d/99-imx471-camera.rules` does it with `SUBSYSTEM=="video4linux", KERNEL=="video*", GROUP="video", MODE="0660"`, and the user is in the `video` group. Its `TAG+="uaccess"` has no effect at that point, since `73-seat-late.rules` has already run, so the nodes get no ACL and stay out of application camera pickers.
9. **Fix the colour matrix** in `/usr/share/libcamera/ipa/simple/imx471.yaml`. Its rows sum to 1.15 / 0.50 / 1.30, and the deficient green row gives a strong magenta cast; renormalise each row to 1.0.
10. **For browsers**, Chromium needs the `PipeWireCamera` feature flag. Omarchy already sets it in its default `chrome-flags.conf`.

Result: 1920x1088 at ~57 fps through `pipewire-libcamera`, with the debayer running on the iGPU via EGL.