Root cause confirmed, and a request to resync intel/vision-drivers (fix already available upstream)
Following up on the original report with the confirmed root cause and a specific request for the OEM kernel. == Root cause == The failure originates in the Intel CVS driver (`intel_cvs`, shipped in- tree as `ubuntu/vision/intel_cvs.ko`), not in the ov02c10 sensor driver itself. On this platform the OV02C10 (ACPI `OVTI02C1`) is a Connected Camera (LCHS=1) that is powered through the CVS path, so the sensor depends on a successful CVS probe. `cvs_common_probe()` issues the `SET_HOST_IDENTIFIER` command unconditionally. Protocol 1.0 devices do not support it and return -EIO, so the CVS probe fails. The sensor probe then fails downstream with -ENXIO, which is the symptom in the original report. == Fix is already available upstream == intel/vision-drivers resolves this by issuing `SET_HOST_IDENTIFIER` conditionally (only when the device advertises magic-number / host- identifier support). This landed as PR #38 and is included in the upstream release tag `26WW19.4_NVL` (2026-05-07). It addresses the same Lunar Lake protocol 1.0 probe failure as PR #35. I have validated this fix on the affected hardware. == Current OEM kernels still ship the unpatched snapshot == The in-tree `ubuntu/vision/intel_cvs.ko` in `linux-oem-6.17` is still the pre-fix snapshot on both `6.17.0-1023-oem` and `6.17.0-1025-oem`: - `cvs_common_probe` is 0x4f7 bytes (the unconditional `SET_HOST_IDENTIFIER` call) versus 0x519 bytes once the conditional fix is applied. - The in-tree module `srcversion` (`59B5F70...`) is unchanged across 1023 -> 1025, i.e. the bundled snapshot has not been refreshed. Note: string-based detection is unreliable here (the `host_identifier` / `magic_num_support` strings are identical in the patched and unpatched builds). The function-size / disassembly check above is the reliable discriminator. == Request == Please resync the bundled intel/vision-drivers snapshot in `linux- oem-6.17` (`ubuntu/vision/`) to `26WW19.4_NVL` or later, so the fix ships in-tree. This would let affected users drop the local DKMS workaround they currently need. == Workaround currently in use == A DKMS build of intel/vision-drivers carrying the PR #35 patch makes the camera fully functional (1920x1080 capture through the IPU7 stack via the Intel Camera HAL / icamerasrc). == Affected hardware == Dell XPS 13 9350 (2024), Intel Core Ultra 7 258V (Lunar Lake), OVTI02C1 / OV02C10, IPU7. Same hardware as bug #2147610 (Confirmed). -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2142382 Title: ov02c10 camera sensor fails to probe on Dell Lunar Lake with IPU7 (ENXIO) To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux-signed-oem-6.17/+bug/2142382/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
