This is a regression from 6.11 to 6.12. The option VIDEO_INTEL_IPU6 for the Intel IPU6 driver depends on ARCH_HAS_DMA_OPS. ARCH_HAS_DMA_OPS can only be enabled implicitly be either enabling GART_IOMMU or XEN. I have an Intel-only system with CPU_SUP_INTEL enabled and CPU_SUP_AMD disabled. So GART_IOMMU is not available. Moreover, my system runs on a bare-metal PC and does not need guest OS support. So XEN is disabled, too. It seems to a a bug that I have either to select CPU_SUP_AMD and GART_IOMMU or XEN in order to have VIDEO_INTEL_IPU6 available. Expected behaviour: Being able to select VIDEO_INTEL_IPU6 without enabling AMD-support on an Intel-only system or XEN guest support on a bare-metal PC.
You really have to find the maintainer(s) of this piece of kernel and send your message to LKML and CC them otherwise this bug report has almost zero chance of being noticed and fixed.
The commit https://github.com/torvalds/linux/commit/de6c85bf918ea52d5c680f0d130b37ee2ff152d6 introduced the regression into 6.12.1 as a temporary, work-around for a faulty DMA re-mapping in the IPU6 driver. The commit https://github.com/torvalds/linux/commit/1d4a000289979cc7f2887c8407b1bfe2a0918354 fixed the underlying, true issue. The commit https://github.com/torvalds/linux/commit/c8e9120c2065868d97e9e94bceee777e5db08c3e fixed the dependency regression. Unfortunately, only the first commit has made it into 6.12.1 and the two remaining commits are scheduled for 6.13. Maybe it can be backported to 6.12?