Re: acpicpu(4) strangeness

[Jukka Ruohonen <jukka.ruohonen%iki.fi@localhost>]

On Mon, Jan 31, 2011 at 04:04:36AM -0700, Sverre Froyen wrote:
> Is the bug in the ACPI table? If so, I seem to remember an option to load 
> patched ACPI tables from the root file system.

That is possible, as documented in acpi(4). However, there are obstacles:

        1. First of all, this is not sustainable. We could fix it for
           Dustin, but the bug would remain there for someone else to
           discover.

        2. On Intel systems, the "CPU tables" are usually not the same as
           the "main table". We have no easy way to patch these.

        3. The contents of Intel IA32_PERF_CTL and IA32_PERF_STATUS registers
           remain to be undocumented. In other words, patching the tables
           would require knowledge of what to put in there. Given that these
           registers are also model-specific, we would have the same
           problems as with est(4), which operates largely on educated
           guesses and can guarantee entirely reliable operation only with
           the early Pentium M CPUs that were meticulously tabulated.

- Jukka.