powernow-k8: BIOS does not provide ACPI -PSS objects in a way that Linux understands

[Pyoverdine]

Hello,
In Jaunty freshly installed and updated.
Booting, I've got the message:
[Firmware bug]: powernow-k8: BIOS does not provide ACPI -PSS objetcs in a way that Linux understand....

If you want any file or information, please ask: i'll send you...
Thanks in advance.
Laurent

[Matt Zimmerman (mdz)]

I believe this message indicates a bug in your system BIOS (firmware), not in Ubuntu.

[Scott James Remnant (Canonical) (canonical-scott)]

This happens to affect me too, and I asked about it last week - it may be a Linux driver is missing a quirk kind of bug rather than a firmware bug proper.

I assume that it causes the reporter to have no CPU Frequency Scaling?

[Scott James Remnant (Canonical) (canonical-scott)]

Laurent: could you provide your dmesg output, lspci -vvnn output, /proc/cpuinfo, etc.

[Pyoverdine (pyoverdine)]

Hello,
dmesg output:

[Pyoverdine (pyoverdine)]

Hello
lspci -vvnn output

[Pyoverdine (pyoverdine)]

Hello,
I'm afraid, but /proc/cpuinfo is 0 bytes len...

[Scott James Remnant (Canonical) (canonical-scott)]

<mjg59> That normally means that the BIOS doesn't provide _PSS at all, which means that it either doesn't support frequency changing or that it's disabled in the BIOS

[Pyoverdine (pyoverdine)]

Hello,
If you want another file: said it!!! I'll (try to) send...
Laurent

[Pyoverdine (pyoverdine)]

Hello,
@Scott Jame Remnant: if you want, i can go to see in my bios to look at... But you've got to tell me what to look at for exactly... ;-)
Laurent

[Pyoverdine (pyoverdine)]

Hello,
While in Jaunty cpuinfo is 0 len, in Intrepid it's not: i'll send you (but it's not from Jaunty!!!)

[Scott James Remnant (Canonical) (canonical-scott)]

On Mon, 2009-04-20 at 15:15 +0000, Pyoverdine wrote:

> @Scott Jame Remnant: if you want, i can go to see in my bios to look at... But you've got to tell me what to look at for exactly... ;-)
>
I'm afraid we simply haven't seen every BIOS in existance.

You'd need to self-identify the relevant setting. Alternatively, you
could supply screenshots of your BIOS and we might be able to identify
something.

Scott
--
Scott James Remnant
<email address hidden>

[Leo Milano (lmilano)]

Hi. This bug seems a duplicate of an earlier submission:
https://bugs.launchpad.net/ubuntu/+source/acpi/+bug/363868

[Pyoverdine (pyoverdine)]

Hello,
@Scott James Remnant:
Well, here are the bios (but I do not find a way to take screen shots...)
And the Jaunty CPU Info.

@Leo Lilano:
It seems to be quite similar... (but I'm on 64 bytes Jaunty)

If you want more files, tell me...
Thanks,

Laurent

[Leo Milano (lmilano)]

Laurent: I think this should not depend on whether it's 64 or 32 bit, but rather on the motherboard BIOS. And we have different BIOS.

Scott: I think Matt right, the ACPI folks consider this a firmware bug, not a kernel issue. I reported upstream yesterday:
http://bugzilla.kernel.org/show_bug.cgi?id=13146

I think it would be helpful to have some more info, like the specs for the "correct" way of defining the _PSS. The kernel messages makes it appear as if Linux expected something special, but I am sure it must be a standard!

[Pyoverdine (pyoverdine)]

Hello,
@Leo Milano: I don't know. If you think that's the same, how can I do (or witch people can) mark it as duplicate of 363868 bug?
(I send my bios options Scott James Remnant, thinking that it may contain stupid errors: I'm not specialist).
If I can help, tell me...
Laurent

[Leo Milano (lmilano)]

Laurent: sorry, I meant to say the opposite, I think these are separate bugs in principle because they are different BIOS'es (I can;t open the attachents from here to confirm). Technically these are not duplicates, sorry for the confusion.

All: note that the tests I reported in the kernel's bugzilla were against the later kernel: 2.6.30rc2. This means these errors will also be there in the NEXT Ubuntu unless we do something!

[Leo Milano (lmilano)]

Wow, the release is out and there is no comment on the issue of the fglrx driver crash??? Is it too late to edit these notes?

[Leo Milano (lmilano)]

Arrg, please disregard my previous comment/email, I posted in the wrong bug report, I apologize !

[Pyoverdine (pyoverdine)]

Hello,
I'm sorry for asking stupid question, but, why the bug is in incomplete status??? Is there any information I've not send??
Regards

[Stefan Bader (smb)]

The bug is was incomplete as there is information missing. As the message indicates your bios provides information in a way that Linux cannot understand (or it provides nothing). So if this issue still exists:
1. Is there a newer BIOS for the motherboard / computer? If, yes, you should try this.
2. If no, run the following two commands and attach the resulting text files:
    - 'sudo acpidump -o acpidump.txt'
    - 'sudo dmidecode >dmidecode.txt'

[Pyoverdine (pyoverdine)]

There's no newer driver for my motherboard.

Here are the files:

It was just necessary to ask.

[Stefan Bader (smb)]

So what happens here is, that the ACPI BIOS does not show any Methods that would be required to enable power management and the PSB BIOS structure, which would be an alternative, is only valid on a single CPU system.

The big question: apart from those error messages, is there anything different? Did you have frequency scaling before and this is now gone? Something like that, or did just the error message catch your eye?

[Pyoverdine (pyoverdine)]

Hello,
Well, I don't know exactly but:
-reboot fail (it freeze on test page, testing USB. Reboot is ok with M$ XP)
-"Hiberner" and "Mettre en veille" crash (Sorry I don't know those translations: may be sleep and warm-boot? Those are options in Ubuntu Jaunty. Last option is stop the computer, before is reboot, before is "Hiberner" and before is "Mettre ne veille")
I do not know if that work before (I'm quite new user of Ubuntu) and when I tested Hiberner or Mettre en veille, I've got to unplug my computer, so I don't try much. I tested that to try to wake-up my computer on a precise time. But I don't know if there's any relation with those messages.

If I can help, tell me.
Laurent

[Stefan Bader (smb)]

If one is hibernate, the other likely is suspend (google translate thinks pause, so I guess I am close ;)) But I believe all this is not related to that module message. The module itself is used for cpu powersaving issues (frequency scaling). So the effects are that your CPU's run at full speed all the time. I once had another computer where I had to enable an option in the bios to get this).
The reboot issue might be solved with reboot=b or reboot=a either on the grub command line or in /boot/grub/menu.lst
The suspend/hibernate problem would be another case and will need more information (does the computer go into suspend or hibernate, does the fan stop and the machine seems off or does it have problems there?). But I wonder whether this would be better reported as a new bug with a better matching description. Like "[Jaunty] <laptop vendor and model> fails to suspend/hibernate".

[Pyoverdine (pyoverdine)]

So, I don't know if that's a problem.
But, yes the fan run et full speed all the time and I'll have to suspend the gnome power management (because it cause the computer to suspend and then to crash).
But very thank for you help.

[Stefan Bader (smb)]

Pyoverdine, that description sounds like it is a problem (though I must admit that I do not completely understand the description). That the fan is running during normal operation is very likely something that is related to the powernow module being unable to load. As the CPUs run at full speed all the time they generate more heat. Unfortunately this can only be solved by a correct BIOS (or in some cases the right BIOS settings). For this reason I would close this bug as won't fix. I would suggest, however, that you open a new bug for the suspend/resume problem. There is a wiki page that hopefully explains how to obtain more debugging information. You can find it at https://wiki.ubuntu.com/DebuggingKernelSuspendHibernateResume

Reading over your last comment again, it might be that this is related to CPU temperature getting too high. You could verify this by monitoring your cpu temperature. Unfortunately there is no standard way to do this. Maybe acpi -t reports it or the sensors command from lm-sensors. You would need to experiment a bit. Anyways, if that is the problem, then you should open another bug as well. But it is always better to have one bug for one specific problem. Thanks a lot.

[Pyoverdine (pyoverdine)]

Hello,
No problem. If I can give you other informations, tell me.
I report here for the message, and when you ask for the big question ;-) I'll try to find. But I don't know.
Thanks very much for your help.
Laurent

[Darko Dimitrovski (darkodimitrovski)]

Hello. I have the same problem ACPI_PSS bug is not letting me run live environment or installation of 9.10 or 9.04. I downloaded 8.04 LTS just to test and that runs fine.

I'm new to both Ubuntu and bug reporting, should I just report here or open new report with my motherboard model?

In case you want to know, it is Gigabyte MA78G-DS3H, ver1.0. Factory BIOS was F3 and the error was present there. I updated to latest F8 and the error is still there.

The error appears with both Cool&Quiet enabled (default) or disabled. Also, the error is still there even if I edit (with TAB) the command line of the Try Ubuntu option adding at the end "acpi=off noacpi". I saw that alternative in another post but doesn't help for me.

So as far as it looks to me, I have two options: not install Ubuntu at all, or settle for 8.04. But I really want 9.10.

Any chance of this "bug" being solved? If you need more info just ask for but consider the following:
1. I'm new to Ubuntu so you would need to provide steps (not too detailed, I can find my way around)
2. I do not have a running 9.10 or 9.04 system due to the error above so don't know how would I provide logs/files from it. I have a running 8.04 but that doesn't give the error which is why it could be installed.

[Darko Dimitrovski (darkodimitrovski)]

Update: Kubuntu 9.10 x64 can install and run successfully. This ACPI_PSS error seem to be connected with BIOS which the Ubuntu team has no control over, but it's not only that.

Why would Kubuntu install and run and Ubuntu not? Does this help isolate the "problem"?

[Darko Dimitrovski (darkodimitrovski)]

Update: SUCCESS!!! Woohooo, as a beginner in Ubuntu I am quite proud. :)

Here is my understanding of this topic (please take it with reserve, I am only a beginner).

9.04 is generating the bug powernow-k8 acpi_pss and it clearly says so because it kicks you into busybox and the bug report is on the screen.

9.10 live cd or booting into an installation was restarting my PC just when it should show the login screen. I wrongly assumed that it is the same bug only that the busybox is not showing. To make further tests, I downloaded Kubuntu 9.10 and the live environment was running just fine. I installed it and it worked fine.

But I really wanted Ubuntu so I didn't settle for that. I asked myself, since the core files are the same, why would Kubuntu work and Ubuntu reboot the PC. Seems like a different problem than the BIOS, that would affect both. How about the video driver? I googled how to install a driver under recovery mode, did that, and it worked. Ubuntu is booting up, showing the login screen and working normally.

It seems the driver distributed with the package reboots some video cards/chips.

My suggestion for anyone having the acpi_pss bug or Ubuntu rebooting the PC:
1. Forget 9.04 and download 9.10. ACPI_PSS seems to be fixed there.
2. If trying the live environment of 9.10 reboots the PC, don't worry. Just select Install and go trough the whole process.
3. The first reboot after install if you select Ubuntu (normal mode) it will most probably again reboot the PC just before it shows the login screen.
4. Next time go into Ubuntu (recovery mode). When presented with the first options screen select "root with networking" to have internet access.
5. Install EnvyNG with:
sudo apt-get install envyng-core envyng-qt
6. You can run EnvyNG in text mode with:
envyng -t
7. Select to install ATI or Nvidia driver as needed (I have ATI maybe this issue doesn't exist on Nvidia).
8. After the install is complete it will offer to reboot, select the Ubuntu (normal mode) and you should be fine.

This worked for me and I am very happy now.

To the Ubuntu team:
Maybe this is not correct procedure but in case you want to investigate why 9.10 was rebooting my PC here is some info:
- motherboard Gigabyte MA78G-DS3H ver1.0 with AMD 780G northbridge, Radeon HD3200 integrated
- I wrote down the driver package that EnvyNG installed: 8.660-0ubuntu4
- With that driver it works fine, the driver inside the Ubuntu image seems to reboot with HD3200

Sorry for the long post, I hope I could help. Cheers. If someone needs any more info just ask.

[Tom Pringle (thomas-b-pringle)]

I see this problem on all 6 AMD machines I work with on a regular basis. CPU frequency scaling is disabled. Most of these machines have the log message

                                 [Firmware Bug]: powernow-k8: No compatible ACPI _PSS objects found.

These machines are configured with motherboards from various vendors using various AMD / NVIDIA chipsets. The common factor is the K8 / K10 cpu architecture and the Linux kernel (2.6.28 and .31 possibly earlier as well).

The problem is most acute on the recently released IBM thinkpad edge notebook which boots both Window7 and Ubuntu 10.04. In windows it appears to run at a much lower temperature. Based on this observation I assume frequency scaling is working in Windows so it is hard to understand how the lack of frequency scaling in Ubuntu could be a BIOS bug. Given the pervasiveness of the problem in Linux and the observation that Windows seems to scale CPU frequency, isn't this problem more likely to be a Linux kernel issue. Or does windows somehow workaround the BIOS bug. If so perhaps there is workaround that could be applied in the linux kernel.

more attachments to follow

[Tom Pringle (thomas-b-pringle)]

slight clarification. I neglected to explicitly list kernel 2.6.32 (10.04)

[David Gaarenstroom (david-gaarenstroom)]

@Pyoverdine:
You should look in your BIOS at the CPU setting for something called "Cool 'n' Quit", "Dynamic Clock" or "Frequency Scaling" that should be "Enabled". Do not use "Auto", because that may only work on Windows. If you you cannot find such an option, you may have an option "AM2 Boost" instead, please disable that (only if there is no C 'n' Q option)...

"Your BIOS does not provide ACPI _PSS objects in a way that Linux understands" almost always means that Cool 'n' Quiet has been disabled in the BIOS. This kernel message also isn't really an error, it just tells that the kernel will not use CPUFREQ powernow-k8 support because Cool 'n Quiet has been disabled, so it's doing exactly what is requested.

Please verify that C 'n' Q has indeed been disabled in the BIOS, and then determine what you want. If you don't want Cool 'n' Quiet you can ignore this message (powernow-k8 has been built-in in the kernel so you cannot prevent this message from being displayed).

[Tom Pringle (thomas-b-pringle)]

I checked one of the machines where this problem shows up and "Cool and Quiet" was set to Auto. However there was no option to outright enable it. The 2 options in the BIOS i looked at were Auto and Disabled.

[David Gaarenstroom (david-gaarenstroom)]

@Tom Pringle: even though you get the same kernel warning (symptom), for you this is a different situation, because you are using completely different hardware (cause). That is also why I think similar bug reports for different hardware should not be merged.

AFAIK, there can be three main causes for this warning:
1) Your BIOS is not configured to support C 'n' Q
2) Your BIOS is incomplete
3) You are using a different (newer) CPU than the ones your motherboard supports

For Lenovo I know they offers _incomplete_ ACPI information, you will have to override your ACPI data by using a custom DSDT. You should be able to find some howto's on the Internet, I don't have the full details...

But either way, these bug reports do not apply to any software in Ubuntu, except that you cannot use a customized powernow-k8 driver because this module is built into the kernel image...

[Tom Pringle (thomas-b-pringle)]

I've already "fixed" this problem on my Lenovo laptop by building in a
custom DSDT. As I understand it, this means I have to merge the table with
each new kernel change and rebulid. I guess that will be need less often
once 10.04 is released.

Note that i see this problem on all 6 AMD platforms I work with.

I've also filed this bug up stream with no positive outcome.

If Windows can handle the situation I'm not sure how we can conclude the
BIOS is the problem?

On Tue, Apr 13, 2010 at 3:30 AM, David Gaarenstroom <
<email address hidden>> wrote:

> @Tom Pringle: even though you get the same kernel warning (symptom), for
> you this is a different situation, because you are using completely
> different hardware (cause). That is also why I think similar bug reports
> for different hardware should not be merged
>
> AFAIK, there can be three main causes for this warning:
> 1) Your BIOS is not configured to support C 'n' Q
> 2) Your BIOS is incomplete
> 3) You are using a different (newer) CPU than the ones your motherboard
> supports
>
> For Lenovo I know they offers _incomplete_ ACPI information, you will
> have to override your ACPI data by using a custom DSDT. You should be
> able to find some howto's on the Internet, I don't have the full
> details...
>
> But either way, these bug reports do not apply to any software in
> Ubuntu, except that you cannot use a customized powernow-k8 driver
> because this module is built into the kernel image...
>
> --
> powernow-k8: BIOS does not provide ACPI -PSS objects in a way that Linux
> understands
> https://bugs.launchpad.net/bugs/364156
> You received this bug notification because you are a direct subscriber
> of the bug.
>
> Status in The Linux Kernel: In Progress
> Status in “linux” package in Ubuntu: Won't Fix
>
> Bug description:
> Hello,
> In Jaunty freshly installed and updated.
> Booting, I've got the message:
> [Firmware bug]: powernow-k8: BIOS does not provide ACPI -PSS objetcs in a
> way that Linux understand....
>
> If you want any file or information, please ask: i'll send you...
> Thanks in advance.
> Laurent
>
> To unsubscribe from this bug, go to:
> https://bugs.launchpad.net/linux/+bug/364156/+subscribe
>

[Tom Pringle (thomas-b-pringle)]

I've already "fixed" this problem on my Lenovo laptop by building in a custom DSDT. As I understand it, this means I have to merge the table with each new kernel change and rebulid. I guess that will be need less often once 10.04 is released.

Note that i see this problem on all 6 AMD platforms I work with.

I've also filed this bug up stream with no positive outcome.

If Windows can handle the situation I'm not sure how we can conclude the BIOS is the problem?

[David Gaarenstroom (david-gaarenstroom)]

AFAIK, If you placed your DSDT.aml in /etc/initramfs-tools/ a kernel change does not require any additional action... Only if you update your BIOS you should generate a new DSDT based on the new BIOS.

On all 4 AMD platforms I can test on, I've never seen this problem. You should look at each platform individually. I noticed only one dmidecode.txt, so I wasn't aware of additional platforms.

Maybe you can post me the details on the other platforms to my email address, it's probably of no use to keep discussing this here, but maybe I can be of any assistance to you...

[Tom Pringle (thomas-b-pringle)]

Hi David,

I had followed some instructions on the web which required building the DSDT
into kernel. I'll try the approach you suggest here this evening.

I'll try to get you more details on some of the other machines where I
observe this problem on the weekend.

Thanks,
Tom

On Tue, Apr 13, 2010 at 8:29 AM, David Gaarenstroom <
<email address hidden>> wrote:

> AFAIK, If you placed your DSDT.aml in /etc/initramfs-tools/ a kernel
> change does not require any additional action... Only if you update your
> BIOS you should generate a new DSDT based on the new BIOS.
>
> On all 4 AMD platforms I can test on, I've never seen this problem. You
> should look at each platform individually. I noticed only one
> dmidecode.txt, so I wasn't aware of additional platforms.
>
> Maybe you can post me the details on the other platforms to my email
> address, it's probably of no use to keep discussing this here, but maybe
> I can be of any assistance to you...
>
> --
> powernow-k8: BIOS does not provide ACPI -PSS objects in a way that Linux
> understands
> https://bugs.launchpad.net/bugs/364156
> You received this bug notification because you are a direct subscriber
> of the bug.
>
> Status in The Linux Kernel: In Progress
> Status in “linux” package in Ubuntu: Won't Fix
>
> Bug description:
> Hello,
> In Jaunty freshly installed and updated.
> Booting, I've got the message:
> [Firmware bug]: powernow-k8: BIOS does not provide ACPI -PSS objetcs in a
> way that Linux understand....
>
> If you want any file or information, please ask: i'll send you...
> Thanks in advance.
> Laurent
>
> To unsubscribe from this bug, go to:
> https://bugs.launchpad.net/linux/+bug/364156/+subscribe
>

[Andy S (andys-bristolwireless)]

Just had my CPU pull a thermal overload shutdown,

Ubuntu Lucid upgraded 31.5.10
Gigabyte GA-MA785*** + AMD PhenomII 550BE (quadded) - no issues with this setup with various Debian based distros until today.

1st indication of problem when scaling app reverted to performance instead of ondemand, then post fglrx manual install using sgfxi - desktop reported scaling problem.

Shutdown pc just in time - cpu temp too high to reboot for ~15mins!!

kern.log:>>
powernow-k8: Found 1 AMD Engineering Sample processors (4 cpu cores) (version 2.20.00)
May 31 15:35:28 P***** kernel: [ 0.941662] [Firmware Bug]: powernow-k8: No compatible ACPI _PSS objects found.
May 31 15:35:28 P***** kernel: [ 0.941664] [Firmware Bug]: powernow-k8: Try again with latest BIOS.

/var/apt/history.log:>>

Start-Date: 2010-05-31 09:59:40
Upgrade: libgtk2.0-common (2.20.0-0ubuntu4, 2.20.1-0ubuntu1), libc-bin (2.11.1-0ubuntu7, 2.11.1-0ubuntu7.1), gwibber-service (2.30.0.1-0ubuntu1, 2.30.0.1-0ubuntu3), libgail18 (2.20.0-0ubuntu4, 2.20.1-0ubuntu1), xsane (0.996-2ubuntu1.1, 0.996-2ubuntu2), libgail-common (2.20.0-0ubuntu4, 2.20.1-0ubuntu1), gnome-screensaver (2.30.0-0ubuntu1, 2.30.0-0ubuntu2), python-gtksourceview2 (2.10.0-0ubuntu1, 2.10.1-0ubuntu1), gtk2-engines-pixbuf (2.20.0-0ubuntu4, 2.20.1-0ubuntu1), libevview2 (2.30.0-0ubuntu1, 2.30.1-0ubuntu3), libc-dev-bin (2.11.1-0ubuntu7, 2.11.1-0ubuntu7.1), python-apt (0.7.94.2ubuntu6, 0.7.94.2ubuntu6.1), gir1.0-gtk-2.0 (2.20.0-0ubuntu4, 2.20.1-0ubuntu1), jockey-common (0.5.8-0ubuntu8, 0.5.8-0ubuntu8.1), libc6-i686 (2.11.1-0ubuntu7, 2.11.1-0ubuntu7.1), libgtk2.0-bin (2.20.0-0ubuntu4, 2.20.1-0ubuntu1), evince (2.30.0-0ubuntu1, 2.30.1-0ubuntu3), libevdocument2 (2.30.0-0ubuntu1, 2.30.1-0ubuntu3), xsane-common (0.996-2ubuntu1.1, 0.996-2ubuntu2), libc6-dev (2.11.1-0ubuntu7, 2.11.1-0ubuntu7.1), librsvg2-2 (2.26.2-0ubuntu2, 2.26.3-0ubuntu1), jockey-gtk (0.5.8-0ubuntu8, 0.5.8-0ubuntu8.1), jockey-kde (0.5.8-0ubuntu8, 0.5.8-0ubuntu8.1), gwibber (2.30.0.1-0ubuntu1, 2.30.0.1-0ubuntu3), libc6 (2.11.1-0ubuntu7, 2.11.1-0ubuntu7.1), librsvg2-common (2.26.2-0ubuntu2, 2.26.3-0ubuntu1), libgtk2.0-0 (2.20.0-0ubuntu4, 2.20.1-0ubuntu1)

End-Date: 2010-05-31 10:01:26

sgfxi script simply d/l Ati driver - compiles and installed - only pulled in missing libgl1-mesa-glx from repo...

can't see any other errors in logs.

More info if needed..hoping full shutdown + restart will now repeat this incident and it was just a random spurious inexplicable incident!!

[Andy S (andys-bristolwireless)]

Going to check 'cool and quiet' hasn't disabled itself in bios as suggested here:

http://ubuntuforums.org/archive/index.php/t-1477743.html

[Andy S (andys-bristolwireless)]

Yep - was one of those 'random inexplicable incidents'! (just not imaginary!)

[LAZA (laza74)]

Got the same problem here on my second machine:

Asrock ALiveXFire-eSATA2
AMD Athlon 64 X2 Dual Core Processor 4400+

Ubuntu Lucid 10.04.2

So if it helps i also can post some reports...

[Joachim Schwender (jschwender)]

Got the same problem on all my sun v20z with upgradung from 8.04 to 11.04.
powernow used to work with 8.04, with 11.04 it does not, the fans are running full speed. Kernel option noacpi did not change anything. The BIOS has no option to set something related to ACPI and no updates are available. acpidump on both versions are identical. powernowd is no more available, cpufreq complains "No cpufreq supprt". For me, this is a showstopper, i probably have to replace all hardware. This is very disappointing: ubuntu folks and kernel folks keep blaming the BIOS - but Windows can do it on the same hardware ......