Follow-up from bug 2483180 comment 45, split into its own ticket as requested.
The IR camera on the ThinkPad X1 Carbon Gen 14 (tested: 21V7CTO1WW) is an ST VD55G1 (mono, global shutter) behind ACPI HID TBE20A1, on the same IPU7 as the IMX471 (1 CSI-2 lane, port 2). Sensor identity was taken from the tuning-file metadata in Lenovo's Windows driver package for this HID and confirmed on the wire: the mainline vd55g1 driver's probe reads back model id 0x53354731.
With four small changes it streams end to end (Y10 804x704@30, working AE, concurrent with the IMX471):
1. vd55g1: ACPI match table with TBE20A1 (driver is fwnode-based; this is sufficient, as you noted)
2. ipu-bridge: IPU_SENSOR_CONFIG("TBE20A1", 1, 600000000) (600 MHz verified streaming)
3. int3472: power-enable GPIO mapped to con_id "vana" for TBE20A1 (same pattern as the IMX471 entries; without it probing fails on a dummy regulator)
4. ipu7-isys: add a V4L2_PIX_FMT_Y10 entry to the pixel format table — Y10_1X10 is already in csi2_supported_codes and the RAW10 mapping, but with no video-node format, mono sensors pass subdev negotiation and then fail STREAMON with EPIPE
Working patches, dmesg logs, and the capture recipe: https://github.com/jriff/x1c14-ir-vd55g1
I have the hardware (21V7CTO1WW) and am happy to test patches and provide Tested-by. Note for expectations: the IR emitter is not driven by anything yet, so indoor LED-lit scenes are near-black; daylight works fine.