this bug is now documented in the RHEL5.3 release notes. please consult the following link to view the most recent build of this document:

http://documentation-stage.bne.redhat.com/docs/en-US/Red_Hat_Enterprise_Linux/5.3/html-single/Release_Notes/

this document is synced daily. as such, any recent additions are expected to show up within 24 hours or so.

Release note added. If any revisions are required, please set the 
"requires_release_notes" flag to "?" and edit the "Release Notes" field accordingly.
All revisions will be proofread by the Engineering Content Services team.

New Contents:
cpuspeed has been updated to version 1.2.1-5. With this update, the cpuspeed init script now loads the speedstep-centrino module if all other module loads fail. In addition, a user-space bug that prevented the Powernow-k8 module from loading is now fixed.

on AMD x86_64 arch machine hp-xw9400-01.rhts.bos.redhat.com:

[root@hp-xw9400-01 cpufreq]# lsmod | grep powernow
powernow_k8            58073  0 
freq_table             40889  2 powernow_k8,cpufreq_ondemand

I can use "modprobe -r powernow_k8" to remove the module, while on AMD i686
arch machine hp-xw9400-02.rhts.bos.redhat.com:

[root@hp-xw9400-02 function]# lsmod | grep powernow
powernow_k8            23009  2

I can not remove by "modprobe -r powernow_k8"
[root@hp-xw9400-02 function]# modprobe -r powernow_k8
FATAL: Module powernow_k8 is in use.

This should be not a big problem, just want to know whether need to improve or
change it, to make them look the same.

tested on hp-xw4550-01.rhts.bos.redhat.com bare metal kernel is OK with test case cpuspeed/function and test case in https://bugzilla.redhat.com/show_bug.cgi?id=451119#c1

cpuspeed is not supported under xen kernel, see kernel function powernowk8_init()

............
        if (preregister_acpi_perf == 1 && cpu_family == CPU_OPTERON) {
                char * dmi_data = dmi_get_system_info(DMI_BIOS_VENDOR);
                if (dmi_data && !strncmp(dmi_data, "Hewlett-Packard", 15)) {
#ifdef CONFIG_XEN
                        /* Disable cpufreq for HP AMD Opteron systems */
                        printk("%s: This BIOS is %s .... disabling cpufreq "
                               "support\n", __FUNCTION__, dmi_data);
                        return -EPERM;
..............

tested on sun-x4600m2-01.rhts.bos.redhat.com bare metal kernel is OK.
but on xen kernel, output of "service cpuspeed status" is:

[root@sun-x4600m2-01 ~]# service cpuspeed status
Frequency scaling enabled using performance governor

ondemand driver is not loaded by default:

[root@sun-x4600m2-01 cpufreq]# cat scaling_available_governors 
userspace performance 

after load  the ondemand driver and echo ondemand > scaling_governor,
then "service cpuspeed status" becomes to:

[root@sun-x4600m2-01 cpufreq]# service cpuspeed status
Frequency scaling enabled using ondemand governor

I don't know whether xen kernel on AMD machine is intended to use performance governor, but I found that on several intel cpu machine, all xen kernel use ondemand governor by default.

so put the status to ON_DEV.

From the initscript:

    # Frequency scaling on AMD rev F CPUs under Xen can result in
    # timekeeping problems for fully virtualized guests, so we disable
    # it by default.
    if [ -d /proc/xen ] && [ "$cpu_vendor" == AuthenticAMD ] \
       && [ "$cpu_family" -le 15 ]; then
      default_governor=performance
    fi

In other words, things are behaving as designed.

Is there a reason why the release notes flag was set back to ?

this note is currently in the 5.3 release notes...

An advisory has been issued which should help the problem
described in this bug report. This report is therefore being
closed with a resolution of ERRATA. For more information
on therefore solution and/or where to find the updated files,
please follow the link below. You may reopen this bug report
if the solution does not work for you.

http://rhn.redhat.com/errata/RHEA-2009-0098.html