Bug 4454 - Mobile Intel Prescott with Enhanced Speedstep not recognized, acpi-cpufreq & speedstep-centrino don't recognize
Summary: Mobile Intel Prescott with Enhanced Speedstep not recognized, acpi-cpufreq & ...
Status: REJECTED DUPLICATE of bug 5483
Alias: None
Product: ACPI
Classification: Unclassified
Component: Power-Processor (show other bugs)
Hardware: i386 Linux
: P2 normal
Assignee: Venkatesh Pallipadi
URL:
Keywords:
Depends on:
Blocks:
 
Reported: 2005-04-06 17:05 UTC by Paul
Modified: 2005-10-29 08:49 UTC (History)
2 users (show)

See Also:
Kernel Version: kernel-smp#2.6.10-1.770_FC3 and also kernel 2.6.11.6
Subsystem:
Regression: ---
Bisected commit-id:


Attachments

Description Paul 2005-04-06 17:05:51 UTC
Distribution: Fedora Core 3
Hardware Environment: Mobile Pentium 4 538 (Prescott, 1mb L2, Enhanced Intel
Speed Step) Toshiba A75-S229 laptop
Software Environment: 
Problem Description: The Mobile Pentium 4 538 (3.2GHz) supports Enhanced Intel
Speed Step but that is not recognized by the ACPI.  I am not able to control
voltage changing.  It works under Windows, checked the voltage changing with a
program, so I know that my processor supports voltage changin.  I have included
an acpidmp, dmesg, and cpuinfo.  In the dmesg you will see that when I try and
modprobe acpi-cpufreq it gives me the error:
acpi-cpufreq: Transition failed

Also in dmesg, says it found a processor with E.I.S.T. but that it is
unsupported.  Your help is much appreciated.
Below are the links to the files, too big to post here.
http://plaza.ufl.edu/redboots/files/dmesg.txt
http://plaza.ufl.edu/redboots/files/cpuinfo.txt
http://plaza.ufl.edu/redboots/files/acpidmp.txt
Steps to reproduce: anytime I boot.
Comment 1 Luming Yu 2005-04-20 02:03:26 UTC
Please try kernel parmeter: acpi_serialize
Comment 2 Brian Gaynor 2005-04-26 10:09:04 UTC
Another data point on Fedora Core 3 (2.6.11-1.14_FC3smp) for this.

This is a Mobile Pentium 4 532 (Prescott, 1Mb L2, enhanced speed step, 3.06GHz)
on a Dell Inspiron 5160. Southbridge is 82801DB. Not recognized by the speedstep
driver (speedstep-ich.ko). I can get the p4-clockmod driver to work, but it
doesn't do voltage control. Tried the acpi_serialize kernel option mentioned but
it was no help. I am happy to provide any additional info.

Comment 3 Paul 2005-04-26 14:33:57 UTC
I tried acpi_serialize as a kernel parameter and it still didnt work.  Here is
the output from dmesg.  It still says acpi-cpufreq: Trasition failed

http://plaza.ufl.edu/redboots/files/dmesg2.txt
Comment 4 Andrew Morton 2005-05-25 22:04:54 UTC
Can someone please determine whether this problem is still
present in 2.6.12-rc5?

Thanks.
Comment 5 Brian Gaynor 2005-06-08 13:18:25 UTC
Andrew - still can't get it to work with 2.6.12-rc5. Modeprobe fails with "No
such device". Let me know if I can provide any more information. P.S. sorry for
the delay, been out of town.
Comment 6 Venkatesh Pallipadi 2005-07-27 19:05:51 UTC
Did you try both speedstep-centrino and acpi-cpufreq drivers? One of them should
work here..
Comment 7 Brian Gaynor 2005-08-18 13:38:20 UTC
Venkatesh - I can only get p4-clockmod to load. With p4-clockmod I can adjust
the processor speed but not the voltage. 
Comment 8 Venkatesh Pallipadi 2005-09-28 19:25:04 UTC
Sorry for the delayed reply...
So, it doesn't work with any kernel 2.6.11/2.6.12. For double check, please make
sure you have both acpi-cpufreq and speedstep-centrino configured and
SPEEDSTEP_CENTRINO_ACPI selected in your config.

I am worried about _PDC method error in dmesg AE_ALREADY_EXISTS. There were some
changes related to this in recent kernels. So, can you please test with 2.6.13
and check whether it works.

Thanks.
Comment 9 R Hill 2005-09-30 21:25:17 UTC
this is my dmesg output when booting with a kernel with ACPI_CPUFREQ enabled.  i
do override my DSDT with a fixed version, but I get the same _PDC error even if
i don't.  i also happened to get a kernel bug here.  this happens sometimes but
not always when i try to boot a kernel with ACPI_CPUFREQ enabled.

http://metawire.org/~dirtyepic/dmesg_acpi-cpufreq.log

(the ACPI error with PBST at the end is normal - or at least i've always gotten
it :P - it doesn't seem to do any harm.)

the laptop is a Toshiba Satellite A70 w/ a Mobile Intel Pentium 4 3.06GHz
Prescott chip.
Comment 10 R Hill 2005-09-30 21:26:20 UTC
erg.  forgot to mention.  this was using kernel 2.6.13-ck6.
Comment 11 Venkatesh Pallipadi 2005-10-14 10:03:09 UTC
Bob,

In which ACPICA version was the duplicate SSDT fix go in?

Does this look familiar
ACPI: Subsystem revision 20050408

    ACPI-0304: *** Error: Looking up [_PPC] in namespace, AE_ALREADY_EXISTS
    ACPI-1172: *** Error: Method execution failed [\_PR_.CPU0._PDC] (Node
dbf5a260), AE_ALREADY_EXISTS
acpi-cpufreq: CPU0 - ACPI performance management activated.
    ACPI-0304: *** Error: Looking up [_PPC] in namespace, AE_ALREADY_EXISTS
    ACPI-1172: *** Error: Method execution failed [\_PR_.CPU1._PDC] (Node
dbf5a1e0), AE_ALREADY_EXISTS
acpi-cpufreq: CPU1 - ACPI performance management activated.

    ACPI-0304: *** Error: Looking up [PBST] in namespace, AE_ALREADY_EXISTS
    ACPI-1172: *** Error: Method execution failed [\_SB_.PCI0.LPC0.BAT1._BST]
(Node dbf59140), AE_ALREADY_EXISTS


Comment 12 Venkatesh Pallipadi 2005-10-14 10:06:46 UTC
In the dmesg posted in comment #9, I see:

acpi-cpufreq: Transition failed
acpi-cpufreq: Transition failed

So, acpi-cpufreq driver is indeed getting loaded but failing.
Can you send me the cat output of all files under
/sys/devices/system/cpu/cpuX/cpufreq/
X=0,1

Thanks,
Venki
Comment 13 Venkatesh Pallipadi 2005-10-23 16:49:27 UTC
Can you try the patch in bug #5483 over 2.6.14-rc5 kernel with acpi-cpufreq 
driver enabled. Let me know whether that fixes the issue.

Thanks.
Comment 14 R Hill 2005-10-28 21:44:00 UTC
this is from 2.6.14-rc5-ck1 without the patch.

# CONFIG_X86_ACPI_CPUFREQ is not set
# CONFIG_X86_SPEEDSTEP_CENTRINO is not set
CONFIG_X86_P4_CLOCKMOD=y


root ~ #  for x in /sys/devices/system/cpu/cpu[0-1]/cpufreq/*; do echo -e "\n
$x" && cat $x; done

 /sys/devices/system/cpu/cpu0/cpufreq/affected_cpus
0 1

 /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq
3066590

 /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_max_freq
3066590

 /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_min_freq
383323

 /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies
383323 766647 1149971 1533295 1916618 2299942 2683266 3066590 

 /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_governors
conservative ondemand powersave userspace performance 

 /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq
3066590

 /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver
p4-clockmod

 /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
performance

 /sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq
3066590

 /sys/devices/system/cpu/cpu0/cpufreq/scaling_min_freq
383323

 /sys/devices/system/cpu/cpu1/cpufreq/affected_cpus
0 1

 /sys/devices/system/cpu/cpu1/cpufreq/cpuinfo_cur_freq
3066590

 /sys/devices/system/cpu/cpu1/cpufreq/cpuinfo_max_freq
3066590

 /sys/devices/system/cpu/cpu1/cpufreq/cpuinfo_min_freq
383323

 /sys/devices/system/cpu/cpu1/cpufreq/scaling_available_frequencies
383323 766647 1149971 1533295 1916618 2299942 2683266 3066590 

 /sys/devices/system/cpu/cpu1/cpufreq/scaling_available_governors
conservative ondemand powersave userspace performance 

 /sys/devices/system/cpu/cpu1/cpufreq/scaling_cur_freq
3066590

 /sys/devices/system/cpu/cpu1/cpufreq/scaling_driver
p4-clockmod

 /sys/devices/system/cpu/cpu1/cpufreq/scaling_governor
performance

 /sys/devices/system/cpu/cpu1/cpufreq/scaling_max_freq
3066590

 /sys/devices/system/cpu/cpu1/cpufreq/scaling_min_freq
383323


i'll build 2.6.14 with the patch and try those three drivers.  is it enough to
build them as modules or should they be compiled into the kernel?
Comment 15 R Hill 2005-10-28 22:49:03 UTC
cool.  it's loading the acpi-cpufreq module without a problem now.  i haven't
tested it but it looks good.  at least there's no errors in dmesg and it doesn't
freeze up during boot like it did.  thanks. :D

dmesg:
acpi-cpufreq: CPU0 - ACPI performance management activated.
acpi-cpufreq: CPU1 - ACPI performance management activated.

here's the cpufreq /sys info:
dirtyepic ~ $  cat 2.6.14-ck1-acpi_cpufreq.log 

 /sys/devices/system/cpu/cpu0/cpufreq/affected_cpus
0

 /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_max_freq
3067000

 /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_min_freq
1867000

 /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_frequencies
3067000 2667000 2400000 2133000 1867000 

 /sys/devices/system/cpu/cpu0/cpufreq/scaling_available_governors
conservative ondemand powersave userspace performance 

 /sys/devices/system/cpu/cpu0/cpufreq/scaling_cur_freq
3067000

 /sys/devices/system/cpu/cpu0/cpufreq/scaling_driver
acpi-cpufreq

 /sys/devices/system/cpu/cpu0/cpufreq/scaling_governor
userspace

 /sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq
3067000

 /sys/devices/system/cpu/cpu0/cpufreq/scaling_min_freq
1867000

 /sys/devices/system/cpu/cpu0/cpufreq/scaling_setspeed
3067000

 /sys/devices/system/cpu/cpu1/cpufreq/affected_cpus
1

 /sys/devices/system/cpu/cpu1/cpufreq/cpuinfo_max_freq
3067000

 /sys/devices/system/cpu/cpu1/cpufreq/cpuinfo_min_freq
1867000

 /sys/devices/system/cpu/cpu1/cpufreq/scaling_available_frequencies
3067000 2667000 2400000 2133000 1867000 

 /sys/devices/system/cpu/cpu1/cpufreq/scaling_available_governors
conservative ondemand powersave userspace performance 

 /sys/devices/system/cpu/cpu1/cpufreq/scaling_cur_freq
3067000

 /sys/devices/system/cpu/cpu1/cpufreq/scaling_driver
acpi-cpufreq

 /sys/devices/system/cpu/cpu1/cpufreq/scaling_governor
userspace

 /sys/devices/system/cpu/cpu1/cpufreq/scaling_max_freq
3067000

 /sys/devices/system/cpu/cpu1/cpufreq/scaling_min_freq
1867000

 /sys/devices/system/cpu/cpu1/cpufreq/scaling_setspeed
3067000
Comment 16 Venkatesh Pallipadi 2005-10-29 08:49:21 UTC
Thats great news!

I would like you to stop using p4-clockmod for power saving reasons. Though it
runs at lower frequency, it won't save much power. It is mostly useful under
Thermal (cpu too hot) conditions.

acpi-cpufreq and speedstep-centrino are the drivers that you should be using.
Depending on the BIOS configuration and the actual processor, one of those
drivers should work and you shoul dbe able to change the freq.

As the patch in bug #5483 seems to be doing the trick here, I will mark this bug
as a duplicate of that one.

Thanks.

*** This bug has been marked as a duplicate of 5483 ***

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