Bug 489580 - when booted with P-state limit, limit can never be increased
Summary: when booted with P-state limit, limit can never be increased
Keywords:
Status: CLOSED WONTFIX
Alias: None
Product: Red Hat Enterprise Linux 4
Classification: Red Hat
Component: kernel
Version: 4.7
Hardware: All
OS: Linux
high
medium
Target Milestone: rc
: ---
Assignee: Prarit Bhargava
QA Contact: Red Hat Kernel QE team
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2009-03-10 18:51 UTC by Marc Milgram
Modified: 2018-11-14 18:29 UTC (History)
7 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of: 489566
Environment:
Last Closed: 2010-10-18 12:36:09 UTC
Target Upstream Version:
Embargoed:


Attachments (Terms of Use)
Possible fix (1.56 KB, patch)
2009-03-10 18:51 UTC, Marc Milgram
no flags Details | Diff

Comment 1 Marc Milgram 2009-03-12 13:47:57 UTC
Sanitized description:

Certain modern servers contain BIOS functionality to set a P-State limit. This
works by generating an ACPI_PROCESSOR_NOTIFY_PERFORMANCE event and re-reading
the APCI _PPC object.  _PPC indicates the smallest index (highest frequency)
that the OS is allowed to select.

In general, this works fine under EL5.3. But if the cpufreq driver
(acpi-cpufreq) initializes while the _PPC object is already active (_PPC > 0),
the max frequency limit is set such that it can't be changed to normal when a
new ACPI_PROCESSOR_NOTIFY_PERFORMANCE event arrives and _PPC is set back to 0.

Version-Release number of selected component (if applicable):
5.3

How reproducible:
Always

Steps to Reproduce:
1. reboot, or (rmmod acpi-cpufreq; modprobe acpi-cpufreq) while power
consumption is set to "minimum power".
2. change power setting to "max performance"

Actual results:
scaling_max_freq is still set to the frequency corresponding to "minimum power"

Expected results:
scaling_max_freq is set to maximum (=cpuinfo_max_freq)  

There is an upstream git commit that should fix this:

http://mirror.celinuxforum.org/gitstat/commit-detail.php?commit=22c970f3468a6766b362d57fa32ebb92cb8cd6db

Comment 2 RHEL Program Management 2009-03-12 18:56:50 UTC
Since RHEL 4.8 External Beta has begun, and this bugzilla remains 
unresolved, it has been rejected as it is not proposed as exception or 
blocker.

Comment 3 Prarit Bhargava 2009-03-16 15:39:18 UTC
Not a regression, too late for RHEL4.8.

P.

Comment 4 RHEL Program Management 2009-03-16 15:55:56 UTC
Development Management has reviewed and declined this request.  You may appeal
this decision by reopening this request.

Comment 5 Marc Milgram 2009-03-16 17:42:36 UTC
Please review this for inclusion in 4.9.

Comment 12 Prarit Bhargava 2010-10-18 12:35:07 UTC
CPU Frequency (cpufreq) is used by a large number of systems.  The issue
is that this patch enables new functionality in a final release of a
RHEL stream.

The code in the suggested patch was evaluated by RH engineering, and
requires a signficant test cycle for RHEL4.9.   One of RH's principles
in accepting a patch is the stability of the patch and since this patch
effects a large number of systems we cannot accept it for RHEL4.9.

In addition, the BZ is marked a blocker when in fact this is NOT a
regression.  It is likely the blocker flag was carried over from the
clone of the RHEL5 bug.


Note You need to log in before you can comment on or make changes to this bug.