Devicetree only booting exposes ACPI

[alarixnia]

Just installed RPi3_UEFI_Firmware_v1.32.zip to a SD card and attempted to boot NetBSD. I was told that the ACPI support for the RPI3 doesn't use the standard ARM interrupt controller, so needs to be disabled (the kernel panics when booting in ACPI mode). Problematically, the NetBSD bootloader still detects ACPI when devicetree-only mode is selected through EDK2's device configuration, which leads to it attempting ACPI booting anyway and immediately panicking.

The operating system shouldn't be able to detect the presence of ACPI when EDK2 is set to use only devicetrees, right?

[jlinton]

See if its still doing it with recent releases, I think the default was a dual acpi/DT mode, so you may have to switch it to "DT" in the advanced firmware menu.
edit: I think I misunderstood what you were saying here. Because it appears the behavior has changed slightly, I just upgraded my pi3 and it was doing an odd dance with the old firmware where linux would see the ACPI tables but default to DT after displaying the discovered ACPI tables. The current firmware must be in DT mode, or acpi=off must be passed to the kernel because it appears that linux discovers the entire set of ACPI tables rather than aborting early.

If you can retest that would be helpful because I think we may be providing a stubbed ACPI entry, which should be ignored, but maybe doing it that way isn't sufficient on netbsd to have it conclude the ACPI tables are disabled.