arm64: Force Morello kernels to use the FDT

[markjdb]

No description provided.

[jrtc27]

Older firmware doesn’t provide an FDT, and being EDK2 won’t load one. We install one as part of bsdinstall, and enforce it in loader.conf, but that doesn’t help the installer itself. On such systems we still have to boot with ACPI, and the lack of GPU drivers isn’t a problem because it can use efifb for the VT.

Also, QEMU uses ACPI. It maybe can also provide an FDT, but we want to use ACPI there, both because it’s the fancier thing and because it tests ACPICA in CI.

[jrtc27]

I think what we want is for GENERIC-MORELLO to just reverse the order back to what it used to be.

[markjdb]

I think what we want is for GENERIC-MORELLO to just reverse the order back to what it used to be.

I need to test in QEMU, but I believe this change has exactly that effect: if there's an FDT, we'll use it, otherwise we'll fall back to the default, which is ACPI.

[bsdjhb]

The code in question does fall back to auto-guessing and will find ACPI if there is no FDT table:

	env = kern_getenv("kern.cfg.order");
	if (env != NULL) {
		order = env;
		while (order != NULL) {
			if (has_acpi &&
			    strncmp(order, "acpi", 4) == 0 &&
			    (order[4] == ',' || order[4] == '\0')) {
				arm64_bus_method = ARM64_BUS_ACPI;
				break;
			}
			if (has_fdt &&
			    strncmp(order, "fdt", 3) == 0 &&
			    (order[3] == ',' || order[3] == '\0')) {
				arm64_bus_method = ARM64_BUS_FDT;
				break;
			}
			order = strchr(order, ',');
			if (order != NULL)
				order++;	/* Skip comma */
		}
		freeenv(env);

		/* If we set the bus method it is valid */
		if (arm64_bus_method != ARM64_BUS_NONE)
			return (true);
	}
	/* If no order or an invalid order was set use the default */
	if (arm64_bus_method == ARM64_BUS_NONE) {
		if (has_acpi)
			arm64_bus_method = ARM64_BUS_ACPI;
		else if (has_fdt)
			arm64_bus_method = ARM64_BUS_FDT;
	}

[bsdjhb]

Though the CI failures seem to imply that the code didn't work the way I thought it would.

[bsdjhb]

I do wonder if the static environment is overriding the environment from the loader (in which case the kernel won't "see" the ACPI tables)? I thought Kyle fixed it so that we merged the static and dynamic environments, but perhaps not?

[jrtc27]

	/*
	 * If no option was set the default is valid, otherwise we are
	 * setting one to get cninit() working, then calling panic to tell
	 * the user about the invalid bus setup.
	 */
	return (env == NULL);

So it'll fall back on picking the right default, but since acpi isn't listed in the env it'll panic as soon as it's used that to configure the console to tell the user that it couldn't find the requested method.

[jrtc27]

(That is, like I said, it needs to be a reversal of the order, not requesting only fdt)