AMD/Xilinx Zynq UltraScale+ and Versal platform firmware interface
The communication layer between Linux and the on-chip platform management firmware found on AMD/Xilinx Zynq UltraScale+ MPSoC, Versal, and Versal NET adaptive SoCs and FPGAs. It lets the kernel ask the firmware to handle low-level tasks like power, clock, reset, and secure operations on these chips, which are widely used in embedded, industrial, aerospace, and data-center accelerator products.
recommendation
It should stay because this is the Linux side of the firmware interface used by AMD/Xilinx Zynq UltraScale+ MPSoC and Versal adaptive SoCs, which AMD still sells in 2025 with lifecycle commitments stretching to 2045. Upstream activity is healthy, with multiple AMD engineers adding new features such as secure IOCTL support through 2025 and 2026, and there is no alternative driver that talks to this firmware.
repository signals
sources
- lore.kernel.org
Upstream development is active in 2026; this patch adds new zynqmp firmware functionality rather than maintenance-only cleanup.
- lore.kernel.org
The directory continues to gain new secure read/write IOCTL support in 2025, indicating ongoing feature work.
- amd.com
AMD was still marketing Zynq UltraScale+ MPSoCs as current products, with long lifecycle language extending UltraScale+ adaptive SoCs through 2045.
- amd.com
AMD was still marketing Versal adaptive SoCs, including current product families relevant to the driver's compatible strings.
codex reasoning notes (technical)
Local shell inspection showed this is a real platform firmware driver, not helper-only code: Kconfig enables a firmware interface driver and zynqmp.c binds compatibles xlnx,zynqmp-firmware, xlnx,versal-firmware, and xlnx,versal-net-firmware. lore_file_timeline and lore_activity on drivers/firmware/xilinx/zynqmp.c found dense 2023-2026 patch traffic with feature additions from multiple AMD authors; cited lore URLs came from those MCP tool calls. A removal/deprecation subject scan via lore_regex timed out and a fallback lei query was blocked by local sandboxing, so removal evidence is absent rather than formally disproven. AMD product pages were obtained via web search on amd.com and show the underlying SoC families are still current, so this directory should be kept rather than deprecated; there is no natural replacement driver for the same firmware interface.