[Bug 2154140] Re: USBIO MTL-CVF ACPI binding missing for Dell Latitude 7450 IPU6 camera

** Tags added: kernel-daily-bug -- You received this bug notification because you are subscribed to linux in Ubuntu. Matching subscriptions: Bgg, Bmail, Nb https://bugs.launchpad.net/bugs/2154140 Title: USBIO MTL-CVF ACPI binding missing for Dell Latitude 7450 IPU6 camera Status in linux package in Ubuntu: New Bug description: On a Dell Latitude 7450, the internal IPU6 MIPI camera does not work with the Ubuntu `7.0.0-15-generic` USBIO module stack because the USBIO auxiliary GPIO and I2C devices are not associated with the platform's MTL-CVF ACPI nodes: - USBIO GPIO provider: `INTC10D1` / `_UID VGPO` - USBIO I2C adapters: `INTC10D2` / `_UID VIC0` and `INTC10D2` / `_UID VIC1` Without that binding, `int3472-discrete` cannot resolve the GPIO provider for the camera power resources and the OV02E10 sensor path does not probe correctly. ## Affected System - Hardware: Dell Latitude 7450 - OS: Ubuntu 26.04 LTS - Kernel tested: `7.0.0-15-generic` - USBIO bridge: `2ac1:20c9 SLS Lattice AI USB 2.0` - Sensor path: `OVTI02E1:00` / `ov02e10` - ACPI nodes observed: - `/sys/bus/acpi/devices/INTC10D1:00` - `/sys/bus/acpi/devices/INTC10D2:00` - `/sys/bus/acpi/devices/INTC10D2:01` ## Failure Mode Before the USBIO binding fix, the kernel logs contained: ```text intel-ipu6 0000:00:05.0: Found supported sensor OVTI02E1:00 intel-ipu6 0000:00:05.0: Connected 1 cameras int3472-discrete INT3472:01: cannot find GPIO chip INTC10D1:00, deferring int3472-discrete INT3472:0c: cannot find GPIO chip INTC10D1:00, deferring ``` After partial USBIO I2C work, the OV02E10 client existed but probe still failed without the complete binding/power path: ```text ov02e10 i2c-OVTI02E1:00: Error writing reg 0x00fd: -6 ov02e10 i2c-OVTI02E1:00: failed to find sensor: -6 ``` The media graph lacked an `ov02e10` entity and `/dev/video0` did not produce usable frames. ## Fix The attached patch is a reduced USBIO backport candidate for Ubuntu's `intel-usbio-dkms_0~git202510282139.ee221eca-0ubuntu1` source. It does the following: - adds `INTC10D1` to the USBIO GPIO ACPI HID table, - adds `INTC10D2` to the USBIO I2C ACPI HID table, - sets `gpio_chip.fwnode = dev_fwnode(dev)` for ACPI GPIO lookup, - changes `usbio_client_acpi_bind()` to search ACPI children under the USBIO auxiliary device's parent device chain, so matching remains scoped to this physical USBIO bridge rather than using a global HID lookup. The patch intentionally excludes local debug instrumentation and dead-end experiments. ## Test Case 1. Boot the affected kernel without the patch. 2. Check: ```bash sudo journalctl -b -k --no-pager -g 'INTC10D1|INTC10D2|INT3472|OVTI02E1|ov02e10|usbio|ipu' tools/v4l-utils/usr/bin/media-ctl -p timeout 20s gst-launch-1.0 -v v4l2src device=/dev/video0 num-buffers=30 ! fakesink ``` Expected broken result: - `int3472-discrete` cannot find GPIO chip `INTC10D1:00`, or - `ov02e10` fails probing with `failed to find sensor: -6`, and - no usable video stream from `/dev/video0`. 3. Build and boot with the attached USBIO patch. 4. Check the same commands. Expected fixed result: - USBIO GPIO chip is associated with `INTC10D1:00`, - `OVTI02E1:00` binds to `ov02e10`, - media graph contains an `ov02e10` entity linked to `Intel IPU6 CSI2 1`, - `/dev/video0` streams successfully. Validated locally after the reduced binding and related already-upstream power support were present: ```text ov02e10 16-0010 media entity present gst-device-monitor-1.0 Video/Source reports Intel MIPI Camera (V4L2) gst-launch-1.0 -v v4l2src device=/dev/video0 num-buffers=30 ! fakesink completes captured frames at 1280x960 show usable video once the physical shutter is open ``` ## Regression Potential Low to moderate. The patch adds new ACPI HIDs for MTL-CVF USBIO GPIO/I2C devices and adjusts the existing USBIO ACPI binding helper to search ACPI companions up the current USB device's parent chain. It deliberately avoids a global `acpi_dev_get_first_match_dev()` fallback, so systems with multiple USBIO bridges or multiple `INTC10D2` islands should not bind to an unrelated ACPI node. The GPIO fwnode assignment is required for ACPI GPIO lookup by consumers such as INT3472. To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/2154140/+subscriptions