Arm Mali-D71/D32 and Arm China Linlon-D6 Komeda display processors
Display processing units that Arm licenses to SoC vendors for driving screens on Arm-based systems-on-chip. The Komeda family includes Arm's Mali-D71 and Mali-D32 DPUs, used in Arm reference platforms like Total Compute, and the Arm China Linlon-D6 found on newer boards such as the Radxa Orion O6.
recommendation
It should stay in the kernel because the hardware is still being shipped and enabled upstream. Patches landing in 2025 and early 2026 add support for the Arm China Linlon-D6 variant (tested on the Radxa Orion O6), and Arm's own Total Compute reference platforms continue to list Mali-D71 Linux display support, so this is active embedded-SoC code rather than legacy.
repository signals
sources
- lore.kernel.org
March 2026 upstream patch adds Arm China Linlon-D6 support to drm/komeda, showing ongoing enablement for new Komeda-compatible hardware.
- lore.kernel.org
May 2025 komeda maintenance patch ('Register sysfs groups through driver core') shows the driver is still receiving non-removal upkeep.
- spinics.net
The 2026 Linlon-D6 support series says Komeda support was tested on a Radxa Orion O6 board, indicating current real hardware bring-up rather than purely legacy maintenance.
- totalcompute.docs.arm.com
Arm's TC2 2023.10.04 release notes list 'Mali-D71 DPU and virtual encoder support for display on Linux', indicating continued deployment in modern Arm reference platforms.
codex reasoning notes (technical)
Local inspection with `sed`/`rg` of `komeda_drv.c` identified active compatibles `arm,mali-d71`, `arm,mali-d32`, and `armchina,linlon-d6`, so this is a real DRM platform driver. `lore_file_timeline` on `drivers/gpu/drm/arm/display/komeda/komeda_drv.c` produced the cited 2025 and 2026 lore URLs, showing fresh maintenance and new-hardware enablement rather than retirement. `web.search` found the Spinics cover letter documenting Radxa Orion O6 bring-up and Arm TC2 release notes documenting Mali-D71 Linux display support. Within the tool budget, no source-backed removal/deprecation discussion was found; the evidence points to a niche but still-live embedded/SoC display IP block, so keep rather than deprecate.