From: Dave Jones <davej@redhat.com>
Subject: [RHEL5] Fwd: Re: [PATCH] allow the highest frequency if bios think so.
Date: Tue, 21 Nov 2006 20:19:12 -0500
Bugzilla: 218106
Message-Id: <20061122011912.GC533@redhat.com>
Changelog: allow the highest frequency if bios think so.


This is a pretty fundamental 'must have' for RHEL5.

It's amazing this went unnoticed for so long, as it means that
certain systems won't be able to scale their CPU speed to
the maximum frequency.

This has been queued for 2.6.19 (and will likely turn up in
a future 2.6.18.x too)

processor_perflib.c::acpi_processor_ppc_notifier() checks if the value
returned by the processor's _PPC method is 0 and return failed if so.
This is wrong since 0 indicate that the bios think the processor can go
to the highest frequency.  This patch for example fix the HP NX 6125 to
allow its highest frequency to be available.

		Dave

-- 
http://www.codemonkey.org.uk

User-Agent: Mutt/1.5.9i
To: "Pallipadi, Venkatesh" <venkatesh.pallipadi@intel.com>
Cc: davej@redhat.com, cpufreq@lists.linux.org.uk, linux-acpi@vger.kernel.org
Message-ID: <20061121163514.GM4945@poupinou.org>

On Tue, Nov 21, 2006 at 07:08:57AM -0800, Pallipadi, Venkatesh wrote:
> Hmmm Great find.....
> A perfect example of the bug that was quietly hiding
> under our noses all these years....

Indeed.  I'm actually still surprised why I haven't found it in the
past.

> 
> >>  	if (ppc > pr->performance->state_count)
> >
> >BTW I'm wondering if this check should be :
> >	if (ppc >= pr->performance->state_count)
> >>  		goto out;
> >
> 
> Yes. I think this should change to >= as well.
> Otherwise indexing states with this ppc can go out of bounds.
> 

Updated.

Correct bound checking from the value returned from _PPC method.

Signed-off-by: Bruno Ducrot <ducrot@poupinou.org>

--- linux-2.6.19-rc6/drivers/acpi/processor_perflib.c	2006/11/21 10:18:39	1.1
+++ linux-2.6.19-rc6/drivers/acpi/processor_perflib.c	2006/11/21 16:28:04
@@ -83,10 +83,8 @@ static int acpi_processor_ppc_notifier(s
 		goto out;
 
 	ppc = (unsigned int)pr->performance_platform_limit;
-	if (!ppc)
-		goto out;
 
-	if (ppc > pr->performance->state_count)
+	if (ppc >= pr->performance->state_count)
 		goto out;
 
 	cpufreq_verify_within_limits(policy, 0,

-- 
Bruno Ducrot

--  Which is worse:  ignorance or apathy?
--  Don't know.  Don't care.