HP EliteBook 6930p: scaling_max_freq is less than highest frequencies in scaling_available_frequencies

[Brendan Donegan]

brendand@brendand-6930p:/sys/devices/system/cpu/cpu0/cpufreq$ cat scaling_max_freq
1520400
brendand@brendand-6930p:/sys/devices/system/cpu/cpu0/cpufreq$ cat scaling_available_frequencies
2534000 2533000 1600000 800000

[Colin Ian King (colin-king)]

There are some BIOS issues in your machine:

[ 0.728941] ACPI Exception: AE_AML_PACKAGE_LIMIT, Index (0x0000000000000006) is beyond end of object (20121018/exoparg2-418)
[ 0.728946] ACPI Error: Method parse/execution failed [\_TZ_.GETP] (Node ffff880139a31cf8), AE_AML_PACKAGE_LIMIT (20121018/psparse-537)
[ 0.728952] ACPI Error: Method parse/execution failed [\_TZ_.GFXZ._CRT] (Node ffff880139a31aa0), AE_AML_PACKAGE_LIMIT (20121018/psparse-537)
[ 0.729042] ACPI Exception: AE_AML_PACKAGE_LIMIT, Index (0x0000000000000005) is beyond end of object (20121018/exoparg2-418)
[ 0.729046] ACPI Error: Method parse/execution failed [\_TZ_.GETP] (Node ffff880139a31cf8), AE_AML_PACKAGE_LIMIT (20121018/psparse-537)
[ 0.729050] ACPI Error: Method parse/execution failed [\_TZ_.GFXZ._AC0] (Node ffff880139a31a78), AE_AML_PACKAGE_LIMIT (20121018/psparse-537)
[ 0.729274] ACPI Exception: AE_AML_PACKAGE_LIMIT, Index (0x0000000000000001) is beyond end of object (20121018/exoparg2-418)
[ 0.729280] ACPI Error: Method parse/execution failed [\_TZ_.PSWT] (Node ffff880139a31e38), AE_AML_PACKAGE_LIMIT (20121018/psparse-537)
[ 0.729285] ACPI Error: Method parse/execution failed [\_TZ_.GTTP] (Node ffff880139a31de8), AE_AML_PACKAGE_LIMIT (20121018/psparse-537)
[ 0.729290] ACPI Error: Method parse/execution failed [\_TZ_.GFXZ._TMP] (Node ffff880139a31ac8), AE_AML_PACKAGE_LIMIT (20121018/psparse-537)
[ 0.729378] ACPI Exception: AE_AML_PACKAGE_LIMIT, Index (0x0000000000000006) is beyond end of object (20121018/exoparg2-418)
[ 0.729382] ACPI Error: Method parse/execution failed [\_TZ_.GETP] (Node ffff880139a31cf8), AE_AML_PACKAGE_LIMIT (20121018/psparse-537)
[ 0.729387] ACPI Error: Method parse/execution failed [\_TZ_.GFXZ._CRT] (Node ffff880139a31aa0), AE_AML_PACKAGE_LIMIT (20121018/psparse-537)
[ 0.729474] ACPI Exception: AE_AML_PACKAGE_LIMIT, Index (0x0000000000000005) is beyond end of object (20121018/exoparg2-418)
[ 0.729477] ACPI Error: Method parse/execution failed [\_TZ_.GETP] (Node ffff880139a31cf8), AE_AML_PACKAGE_LIMIT (20121018/psparse-537)
[ 0.729482] ACPI Error: Method parse/execution failed [\_TZ_.GFXZ._AC0] (Node ffff880139a31a78), AE_AML_PACKAGE_LIMIT (20121018/psparse-537)
[ 0.729488] [Firmware Bug]: No valid trip found
[ 0.731928] ACPI Exception: AE_AML_PACKAGE_LIMIT, Index (0x0000000000000001) is beyond end of object (20121018/exoparg2-418)
[ 0.731934] ACPI Error: Method parse/execution failed [\_TZ_.PSWT] (Node ffff880139a31e38), AE_AML_PACKAGE_LIMIT (20121018/psparse-537)
[ 0.731939] ACPI Error: Method parse/execution failed [\_TZ_.GTTP] (Node ffff880139a31de8), AE_AML_PACKAGE_LIMIT (20121018/psparse-537)
[ 0.731944] ACPI Error: Method parse/execution failed [\_TZ_.DTSZ._TMP] (Node ffff880139a31fc8), AE_AML_PACKAGE_LIMIT (20121018/psparse-537)

For example, the _CRT (critical trip point) for GFXZ tries to index into \_TZ.TRIP in the GETP() method and ARG0 in this call references element 6 in a 2 element package, which results in a failure. So I guess a bunch of critical trip points are all invalid, which may mean your CPU is scaled down to the lowest speed.

What does the following output?

 for...

[Colin Ian King (colin-king)]

You may like to try out the following kernel parameters

thermal.crt=-1

thermal.off=1

and see what that does.

[Brendan Donegan (brendan-donegan)]

/sys/class/thermal/thermal_zone0:
105000
critical
95000
passive
/sys/class/thermal/thermal_zone1:
112000
critical
60000
passive
/sys/class/thermal/thermal_zone2:
112000
critical
98000
passive
/sys/class/thermal/thermal_zone3:
90000
critical
85000
passive
/sys/class/thermal/thermal_zone4:
107000
critical
98000
active

[Colin Ian King (colin-king)]

OK, now I am confused. That's not what I expected.

[Colin Ian King (colin-king)]

Can you try the following:

Edit /etc/default/grub (you need to do this using root privilege) and change

GRUB_CMDLINE_LINUX_DEFAULT="quiet splash"

to

GRUB_CMDLINE_LINUX_DEFAULT="quiet splash processor.ignore_ppc=1"

and then run:

sudo update-grub

and reboot. Maybe this will work around the issue.

[Colin Ian King (colin-king)]

Brendand, did the ignore_ppc workaround help?

[Brendan Donegan (brendan-donegan)]

Maybe - I noticed that just after booting scaling_max_freq was 2534000, but it dropped down soon after. I can provide the same logs as before again and you can check them to make sure that a.) I enabled ignore_ppc properly and b.) it had the expected effect. Sound ok?

[Colin Ian King (colin-king)]

Either load the machine and check to see the scaling frequencies by hand, or run:

sudo fwts cpufreq

[Brendan Donegan (brendan-donegan)]

I got this now, with ignore_ppc: http://paste.ubuntu.com/6513870/

[Colin Ian King (colin-king)]

Since pastebin data can get reaped after a while, the info that's useful is as follows:

CPU 0: 4 CPU frequency steps supported.
 Frequency | Relative Speed | Bogo loops
-----------+----------------+-----------
  2.58 GHz | 100.0 % | 74603
  2.58 GHz | 93.6 % | 69857
  1.65 GHz | 91.7 % | 68432
   800 MHz | 76.3 % | 56956

WARNING: Test 1, Frequency 1600000 not achievable; _PSS limit of 1520400 in
effect?
CPU 1: 4 CPU frequency steps supported.
 Frequency | Relative Speed | Bogo loops
-----------+----------------+-----------
  2.58 GHz | 95.3 % | 73874
  2.58 GHz | 100.0 % | 77558
  1.65 GHz | 97.9 % | 75927
   800 MHz | 98.5 % | 76409

FAILED [MEDIUM] CPUFreqSlowerOnCPU: Test 1, Supposedly higher frequency 1.65 GHz
is slower (75927 bogo loops) than frequency 1.65 GHz (76409 bogo loops) on CPU
1.
WARNING: Test 1, Frequency 1600000 not achievable; _PSS limit of 1520400 in
effect?
FAILED [MEDIUM] CPUFreqSlowerOnCPU: Test 1, Supposedly higher frequency 2.58 GHz
is slower (73874 bogo loops) than frequency 2.58 GHz (77558 bogo loops) on CPU
1.
FAILED [MEDIUM] CPUFreqCPUsSetToSW_ANY: Test 1, Processors are set to SW_ANY.
FAILED [MEDIUM] CPUFreqSW_ANY: Test 1, Firmware not implementing hardware
coordination cleanly. Firmware using SW_ANY instead?.

[Colin Ian King (colin-king)]

So it seems some CPU scaling works on CPU0, but not on CPU1. What are the values of

/sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies

and

/sys/devices/system/cpu/cpu1/cpufreq/scaling_available_frequencies

with the ignore_ppc=1 setting?

[Colin Ian King (colin-king)]

@Brendan, I will close this bug in a short while as it has not been updated for nearly 3 months

[Brendan Donegan (brendan-donegan)]

Hi Colin,

Close it now, this laptop is deceased - unless you think someone else with the same model is going to come along and provide the info.

Thanks,