Note some long-term difficult goals.

This commit is contained in:
jruoho 2011-06-23 16:56:07 +00:00
parent aade7fc010
commit 769cd829a3
1 changed files with 32 additions and 0 deletions

View File

@ -6,3 +6,35 @@ Important things to make ACPI useful on more machines:
has a more detailed analysis.
- sekiya, 21 December 2005
* The mismatch between the conventional and the ACPI device tree is the single
biggest architectural problem in the current stack. Various power management
features rely on the interaction between the firmware (ACPI) and the native
drivers. Examples include, but are not limited to, WoL, PCI power management,
and runtime power management (which involves GPEs on the ACPI side).
Devices should not attach to acpi(4), but should still use ACPI to reserve
PCI, ISA, and other resources. This includes also buses such as I2C where
resource conflicts and other bus errors are widely known to exist.
A satisfactory solution involves full or partial redesign of the x86
autoconfiguration process. The current ad-hoc solutions used for instance
in acpi_pci.c are inefficient and ugly.
* Related to previous, investigate how existing systems can interact with
the BIOS and ACPI. For instance, at the moment of writing this, IPMI is
already available via ACPI. In the future this may cause possible conflicts
between ipmi(4) and the firmware.
* Not directly related to ACPI, but implement S4 a.k.a. suspend-to-disk.
* Improve IA-64 ACPI support. Even if the architecture may be rare,
support for Itanium is a good way to ensure that the stack remains MI.
* Implement the ACPI requirements for CPU hot-plug.
* Support APEI (ACPI Platform Error Interface).
* Support ACPI 4.0 devices.
- jruoho, 23 June 2011