Re: [PATCH] [RFC] ondemand governor: dynamic cpufreq scaling withdifferent CPUs

From: Venki Pallipadi
Date: Wed Aug 10 2011 - 12:59:39 EST


On Mon, Aug 1, 2011 at 7:51 AM, Len Brown <lenb@xxxxxxxxxx> wrote:
> Re: software vs hardware coordination
>
> Speaking for Intel, everything we make, from hand-sets to supercomputers
> supports hardware coorination.
>
> Think about it -- if we did not, then we would not be able
> to support a legacy OS that doesn't know how to do SW coordination.
>
> HW coordiation is the simplest thing for the SW because
> SW doesn't have to know anything at all about topology constraints.
> It is also the simplest thing for the BIOS, which doesn't have
> to correctly describe topology.
>
> Yes, there is an MSR bit that the BIOS can flip to disable
> HW coordination.  90% of the time that bit is set it is a BIOS bug.
> eg. the bit is set, but the BIOS fails to properly set up the
> ACPI tables to properly enable SW coordination, resulting
> in a system with really strange performance issues.
>
> Indeed, I proposed before, and I will propose again that
> the underlying ACPI _PSD support behind SW coordination
> be completely removed from Linux.
>
> Venki,
> I recall that you mentioned one platform that depends on it --
> and I'd  like to understand that issue and address it with a quirk
> rather than expose all systems to BIOS bugs due to bogus _PSD support.
>

Agree. Addressing this as a quirk would be fine. The processor that
needs it is E74xx series, IIRC.

Thanks,
Venki

> Winston,
> Your system is running in SW coordination mode?
> What do you see if you run with the patch below?
> I'd be intersted to see the output from adpidump for your system.
>
> thanks,
> -Len Brown
> Intel Open Source Technology Center
>
> ---
> diff --git a/include/acpi/pdc_intel.h b/include/acpi/pdc_intel.h
> index 552637b..d081595 100644
> --- a/include/acpi/pdc_intel.h
> +++ b/include/acpi/pdc_intel.h
> @@ -22,7 +22,6 @@
>
>  #define ACPI_PDC_EST_CAPABILITY_SWSMP  (ACPI_PDC_SMP_C1PT | \
>                                         ACPI_PDC_C_C1_HALT | \
> -                                        ACPI_PDC_SMP_P_SWCOORD | \
>                                         ACPI_PDC_SMP_P_HWCOORD | \
>                                         ACPI_PDC_P_FFH)
>
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/