[[
UPDATE: see-also: https://firmwaresecurity.com/2016/01/22/who-created-the-acpi-aspt-spec/
]]
Colin King of Canonical has added a new ACPI test to the FirmWare Test Suit (FWTS). The new test is for ASPT (ACPI System Performance Tuning). The problem is that ASPT is an undocumented ACPI table. As Colin says:
This table is not well described anywhere, however it is a frequently used table on AMD machines and the format is relatively simple set of 4 32 bit addresses. This table has been discussed on the ACPICA devel mailing list:
https://lists.acpica.org/pipermail/devel/2015-November/000850.html
and this description matches the various acpi dumps of this table on AMD machines that I have access too. I believe the table refers to an AMD performance monitoring feature.
Here are some scary comments from the new test code to clarify the problem:
/* ASPT Table (reverse engineered, table is common on AMD machines) */
/* Without a specification to work with there is very little we can do to validate this apart from the
implest sanity check */
/* ACPI ASPT: determined by reverse engineering */
For more information see the fwts-devel list or the fwts source code:
https://lists.ubuntu.com/archives/fwts-devel/2015-December/007107.html
http://www.uefi.org/acpi
IMO, the UEFI Forum, who has recently taken over ownership of ACPI, should be working with this vendor to provide a proper public spec, if the table is being used in modern hardware.
Firmware Security 