freenode
Kernel & Low-Level

Dell's ARM XPS 16 forces laptop drivers out of the x86 tree

Firmware on the RTX Spark machine answers Dell's SMBIOS calls with zeroes, so battery and thermal drivers would otherwise report fake state.

NVIDIA engineer Brian Mak has asked the Linux kernel to stop treating Dell's laptop platform drivers as x86-only, so they can run on the Dell XPS 16 Creator Edition, an ARM64 machine built around NVIDIA's RTX Spark. The laptop's firmware exposes the same ACPI-WMI and SMBIOS interfaces Dell has long used on x86, but the drivers sit behind an x86 platform-device gate and cannot be enabled on ARM64 today.

The proposed move would put the Dell drivers in their own platform directory and let the ones that do not depend on x86-only helpers build for ARM64. A few stay behind: the UART backlight driver needs x86 ACPI code that never creates its serial device elsewhere, and the SMM path, remote BIOS update, and the older systems-management driver remain x86-only. Mak also wants the laptop driver's hard requirement on a PS/2 (i8042) controller relaxed, arguing that a stub already returns an error when that hardware is absent.

Two firmware problems on the XPS 16 (model DX16263) are why a plain move is not enough. The WMI SMBIOS driver looks for Dell's miscellaneous BIOS-flags table at a fixed handle that this machine numbers differently, so the driver fails to attach. Separately, the proprietary SMBIOS calling interface answers every request with an all-zero buffer, including calls that should fail. Drivers that trust those answers report bogus battery charge modes and thresholds, register microphone and mute LEDs that do nothing, and the PC thermal driver aborts while probing platform profiles. Mak's workaround keeps the shared SMBIOS layer loaded, because other Dell WMI code depends on it, but skips the laptop and thermal drivers on this model via a DMI quirk. If the familiar BIOS-flags handle is present it is still used; otherwise the driver accepts the table only when exactly one exists.

Longtime Dell driver reviewer Pali Rohár asked whether Dell knows about the all-zero interface and whether firmware will fix it. Mak said he will report the bug and update the list when he hears back. Rohár called that report important: firmware authors often do not learn about bugs that Linux driver work uncovers, and a fix in later models keeps the kernel's list of broken machines from growing. He also pushed back on the i8042 change. On many Vostro and Inspiron laptops that controller drives the touchpad LED and wireless-kill events, with no detection and no error checking, so he wants those paths compiled out on non-x86 rather than treated as optional everywhere.