Medium Errors when running cpufreq test in fwts

[Jeff Lane ]

While running the cpufreq test on a server, we encountered the following (Here's the full log)

Results generated by fwts: Version V13.09.01 (2013-09-17 07:41:02).

Some of this work - Copyright (c) 1999 - 2010, Intel Corp. All rights reserved.
Some of this work - Copyright (c) 2010 - 2013, Canonical.

This test run on 05/11/13 at 16:05:51 on host Linux ubuntuc1n1 3.11.0-12-generic
#19-Ubuntu SMP Wed Oct 9 16:20:46 UTC 2013 x86_64.

Command: "fwts -q --stdout-summary -r /root/.checkbox/scaling_test.log cpufreq".
Running tests: cpufreq.

cpufreq: CPU frequency scaling tests.
--------------------------------------------------------------------------------
Test 1 of 1: CPU P-State Checks.
For each processor in the system, this test steps through the various frequency
states (P-states) that the BIOS advertises for the processor. For each processor
/frequency combination, a quick performance value is measured. The test then
validates that:
  1. Each processor has the same number of frequency states.
  2. Higher advertised frequencies have a higher performance.
  3. No duplicate frequency values are reported by the BIOS.
  4. BIOS doing Sw_All P-state coordination across cores.
  5. BIOS doing Sw_Any P-state coordination across cores.

CPU 0: 14 CPU frequency steps supported.
 Frequency | Relative Speed | Bogo loops
-----------+----------------+-----------
  2.45 GHz | 100.0 % | 155783
  2.45 GHz | 92.3 % | 143795
  2.35 GHz | 92.3 % | 143798
  2.25 GHz | 92.3 % | 143796
  2.15 GHz | 92.3 % | 143797
  2.05 GHz | 92.3 % | 143794
  1.95 GHz | 92.3 % | 143815
  1.85 GHz | 92.3 % | 143710
  1.75 GHz | 92.3 % | 143789
  1.65 GHz | 92.3 % | 143795
  1500 MHz | 92.3 % | 143798
  1400 MHz | 92.3 % | 143799
  1300 MHz | 92.3 % | 143796
  1200 MHz | 92.3 % | 143797

FAILED [MEDIUM] CPUFreqSlowerOnCPU: Test 1, Supposedly higher frequency 1300 MHz
is slower (143796 bogo loops) than frequency 1300 MHz (143797 bogo loops) on CPU
0.
FAILED [MEDIUM] CPUFreqSlowerOnCPU: Test 1, Supposedly higher frequency 1500 MHz
is slower (143798 bogo loops) than frequency 1500 MHz (143799 bogo loops) on CPU
0.
FAILED [MEDIUM] CPUFreqSlowerOnCPU: Test 1, Supposedly higher frequency 1.65 GHz
is slower (143795 bogo loops) than frequency 1.65 GHz (143798 bogo loops) on CPU
0.
FAILED [MEDIUM] CPUFreqSlowerOnCPU: Test 1, Supposedly higher frequency 1.75 GHz
is slower (143789 bogo loops) than frequency 1.75 GHz (143795 bogo loops) on CPU
0.
FAILED [MEDIUM] CPUFreqSlowerOnCPU: Test 1, Supposedly higher frequency 1.85 GHz
is slower (143710 bogo loops) than frequency 1.85 GHz (143789 bogo loops) on CPU
0.
FAILED [MEDIUM] CPUFreqSlowerOnCPU: Test 1, Supposedly higher frequency 2.05 GHz
is slower (143794 bogo loops) than frequency 2.05 GHz (143815 bogo loops) on CPU
0.
FAILED [MEDIUM] CPUFreqSlowerOnCPU: Test 1, Supposedly higher frequency 2.25 GHz
is slower (143796 bogo loops) than frequency 2.25 GHz (143797 bogo loops) on CPU
0.
FAILED [MEDIUM] CPUFreqSlowerOnCPU: Test 1, Supposedly higher frequency 2.45 GHz
is slower (143795 bogo loops) than frequency 2.45 GHz (143798 bogo loops) on CPU
0.
Frequency scaling not supported.

================================================================================
0 passed, 8 failed, 0 warnings, 0 aborted, 0 skipped, 0 info only.
================================================================================

0 passed, 8 failed, 0 warnings, 0 aborted, 0 skipped, 0 info only.

Test Failure Summary
================================================================================

Critical failures: NONE

High failures: NONE

Medium failures: 8
 cpufreq: Supposedly higher frequency 1300 MHz is slower (143796 bogo loops) than frequency 1300 MHz (143797 bogo loops) on CPU 0.
 cpufreq: Supposedly higher frequency 1500 MHz is slower (143798 bogo loops) than frequency 1500 MHz (143799 bogo loops) on CPU 0.
 cpufreq: Supposedly higher frequency 1.65 GHz is slower (143795 bogo loops) than frequency 1.65 GHz (143798 bogo loops) on CPU 0.
 cpufreq: Supposedly higher frequency 1.75 GHz is slower (143789 bogo loops) than frequency 1.75 GHz (143795 bogo loops) on CPU 0.
 cpufreq: Supposedly higher frequency 1.85 GHz is slower (143710 bogo loops) than frequency 1.85 GHz (143789 bogo loops) on CPU 0.
 cpufreq: Supposedly higher frequency 2.05 GHz is slower (143794 bogo loops) than frequency 2.05 GHz (143815 bogo loops) on CPU 0.
 cpufreq: Supposedly higher frequency 2.25 GHz is slower (143796 bogo loops) than frequency 2.25 GHz (143797 bogo loops) on CPU 0.
 cpufreq: Supposedly higher frequency 2.45 GHz is slower (143795 bogo loops) than frequency 2.45 GHz (143798 bogo loops) on CPU 0.

Low failures: NONE

Other failures: NONE

Test |Pass |Fail |Abort|Warn |Skip |Info |
---------------+-----+-----+-----+-----+-----+-----+
cpufreq | | 8| | | | |
---------------+-----+-----+-----+-----+-----+-----+
Total: | 0| 8| 0| 0| 0| 0|
---------------+-----+-----+-----+-----+-----+-----+

This results in a few questions...

First, what exactly is causing the failures to trigger?
Second, why does it say "Frequency scaling not supported." at the end the run messages? It appears that the cpu is being scaled up and down to various speeds, so that output is confusing and we need some clarification.

Can you please take a look and give us some insite for the two questions above?

Finally, is this a failure we should be concerned about?

[Jeff Lane  (bladernr)]

another question:

we noticed that the frequency scaling test is ONLY run on CPU0, but this is an 8 core system, why is it not run against all cores/sockets?

[Brad Figg (brad-figg)]

This bug is missing log files that will aid in diagnosing the problem. From a terminal window please run:

apport-collect 1248621

and then change the status of the bug to 'Confirmed'.

If, due to the nature of the issue you have encountered, you are unable to run this command, please add a comment stating that fact and change the bug status to 'Confirmed'.

This change has been made by an automated script, maintained by the Ubuntu Kernel Team.

[Colin Ian King (colin-king)]

So, the ACPI _PSS object returns the number of supported CPU frequencies and this test sets the CPU(s) to each frequency and checks the performance of each frequency. The amount of compute possible "bogo cycles" in a fixed amount of time should scale accordingly. In this machine's case, it has turboboost as the top frequency and all other frequencies don't seem to scale (they are all set to 93.3% of the maximum turboboost compute rate), which appears erroneous to fwts (and to me too!)

Hence the test says "Frequency scaling not supported.", because it does not appear to do so because it fails so many scaling settings.

Because the test discovered that CPU frequency scaling looks broken or not supported it does not need to waste time to exercise the other 7 CPUs since they all will respond the same way.

[Colin Ian King (colin-king)]

It would be interesting to see what the firmware is doing, so a dump of the ACPI tables would be useful:

sudo fwts --dump

and attach just the acpidump.log to the bug report

[Narinder Gupta (narindergupta)]

attaching the acpidump.log

[Jeff Lane  (bladernr)]

I'm going to mark this as invalid. Narinder discovered the actual issue appears to be that HP had installed an incorrect processor on this machine. After they swapped out the proc for the correct one, the issue went away.

[Colin Ian King (colin-king)]

That's useful to know, seems like fwts did discover an issue then ;-)