svn commit: r297954 - in head/sys: boot/efi/loader/arch/amd64 boot/i386/libi386 x86/acpica

[Bruce Evans]

On Thu, 14 Apr 2016, Warner Losh wrote:

> Log:
>  Deprecate using hints.acpi.0.rsdp to communicate the RSDP to the
>  system. This uses the hints mechnanism. This mostly works today
>  because when there's no static hints (the default), this value can be
>  fetched from the hint. When there is a static hints file, the hint
>  passed from the boot loader to the kernel is ignored, but for the BIOS
>  case we're able to find it anyway. However, with UEFI, the fallback
>  doesn't work, so we get a panic instead.

The new static_hints.disabled option allows overriding static hints.  The
override global, so hints generally have to be duplicated in both static
hints and /boot/device.hints, like they are or should be in the example
source files.

>  Switch to acpi.rsdp and use TUNABLE_ULONG_FETCH instead. Continue to
>  generate the old values to allow for transitions. In addition, fall
>  back to the old method if the new method isn't present.

This only works if there is no static env (the default).  The new
static_env.disabled option allows overriding static env.

I used to have this problem with all 50-100 hints and tunables that I
use.  static hints and env killed everything in /boot config files,
and also killed all additions and changes in loader.  This rarely
mattered since I rarely use loader, and I keep the /boot files in sync
with static hints and env, but it required a kernel recompile or editing
the kernel binary to change anything.  Editing the kernel binary is
actually fairly easy.  It needs an editor that doesn't clobber binary
data, but the editor doesn't need to dispay or have search capability
for binary data since the hints and env string are text.

> ...
> Modified: head/sys/x86/acpica/OsdEnvironment.c
> ==============================================================================
> --- head/sys/x86/acpica/OsdEnvironment.c	Thu Apr 14 04:54:47 2016	(r297953)
> +++ head/sys/x86/acpica/OsdEnvironment.c	Thu Apr 14 04:59:51 2016	(r297954)
> ...
> @@ -59,6 +60,16 @@ acpi_get_root_from_loader(void)
> {
> 	long acpi_root;
>
> +	if (TUNABLE_ULONG_FETCH("acpi.rsdp", &acpi_root))
> +		return (acpi_root);
> +
> +	/*
> +	 * The hints mechanism is unreliable (it fails if anybody ever
> +	 * compiled in hints to the kernel). It has been replaced
> +	 * by the tunable method, but is used here as a fallback to
> +	 * retain maximum compatibility between old loaders and new
> +	 * kernels. It can be removed after 11.0R.
> +	 */
> 	if (resource_long_value("acpi", 0, "rsdp", &acpi_root) == 0)
> 		return (acpi_root);

Having 2 mechanisms stupid.  It basically gives more than 2 ways of
foot shooting by using 2 spellings of the same variable (both easy to
forget or misspell in interactive use) and possibly not supporting all
combinations.  Despite its verboseness, the above doesn't detect
inconsistent settings or document the precedence for inconsistent
settings anywhere uses can see it.  It documents planned breakage for
1 of the mechanisms.

Bruce