Bug 113884 - IBM Thinkpad A20m - system hangs at shutdown and does not power off due possible ACPI interference
Summary: IBM Thinkpad A20m - system hangs at shutdown and does not power off due possi...
Status: RESOLVED FIXED
Alias: None
Product: SUSE LINUX 10.0
Classification: openSUSE
Component: Kernel (show other bugs)
Version: Beta 3
Hardware: PC All
: P5 - None : Normal
Target Milestone: ---
Assignee: Torsten Duwe
QA Contact: E-mail List
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2005-08-29 15:01 UTC by Adam Schuett
Modified: 2005-12-09 10:02 UTC (History)
5 users (show)

See Also:
Found By: Other
Services Priority:
Business Priority:
Blocker: ---
Marketing QA Status: ---
IT Deployment: ---


Attachments
beta3 hwinfo (186.87 KB, text/plain)
2005-08-31 20:36 UTC, Adam Schuett
Details
beta3 acpidmp (179.58 KB, text/plain)
2005-08-31 20:39 UTC, Adam Schuett
Details
beta4 hwinfo (186.35 KB, text/plain)
2005-09-01 20:17 UTC, Adam Schuett
Details
beta4 acpidmp (179.58 KB, text/plain)
2005-09-01 20:17 UTC, Adam Schuett
Details
/var/lib/boot.msg without any dsdt changes (31.33 KB, text/plain)
2005-09-04 13:06 UTC, Forgotten User v5inq2MBSE
Details
/var/lib/boot.msg with fixed dsdt changes (31.41 KB, text/plain)
2005-09-04 13:07 UTC, Forgotten User v5inq2MBSE
Details

Note You need to log in before you can comment on or make changes to this bug.
Description Adam Schuett 2005-08-29 15:01:24 UTC
When doing a shutdown of the system, it goes through normal procedure of ending
the gui and shutting down all processes.  When it gets to "the system will be
halted immediately", it hangs there for hours on end and never powers off the unit.

This also happens when trying a restart where it hangs giving the message
"please wait while your system restarts".  This problem also exists in trying to
do the Suspend-to-disk feature.

This exact hardware worked fine on 9.1, 9.2, and 9.3 and these problems did not
occur.
Comment 1 Hubert Mantel 2005-08-31 15:30:31 UTC
Andi, do you (or somebody else) have an idea? Could this behaviour be related to
ACPI?
Comment 2 Andreas Kleen 2005-08-31 15:35:22 UTC
Yes, it's probably related to ACPI (unless he turned ACPI off) 

However there is zero information about the hardware in the bug,
so it's useless in this form.  Full hwinfo and acpidmp output please.
Comment 3 Pavel Machek 2005-08-31 20:01:06 UTC
Forcing backtrace with magic key when it is hung would be helpfull.
Comment 4 Adam Schuett 2005-08-31 20:18:23 UTC
ACPI was never turned on as it never functioned with SUSE in the past.  I turned
it on in the BIOS to see if it helped, but did the same thing so it has been
turned back off.  The hardware is an IBM Thinkpad A20m.  If there are other
details that you need to help troubleshoot, please let me know how to get that
data for you, as I am rather new to linux besides day to day usage.
Comment 5 Andreas Kleen 2005-08-31 20:28:52 UTC
So you're likely using APM. It's  all hard to tell given how economic you are
with information. Can you please attach hwinfo?  

And do echo 1 > /proc/sys/kernel/sysrq and then press alt-sysrq-p when
the system hangs and write down or photograph the kernel output.
Comment 6 Adam Schuett 2005-08-31 20:36:17 UTC
Created attachment 48370 [details]
beta3 hwinfo
Comment 7 Adam Schuett 2005-08-31 20:39:55 UTC
Created attachment 48373 [details]
beta3 acpidmp
Comment 8 Andreas Kleen 2005-08-31 20:59:09 UTC
Ok you're actually using ACPI according to the boot log.
Comment 9 Adam Schuett 2005-09-01 20:17:25 UTC
Created attachment 48531 [details]
beta4 hwinfo
Comment 10 Adam Schuett 2005-09-01 20:17:59 UTC
Created attachment 48532 [details]
beta4 acpidmp
Comment 11 Pavel Machek 2005-09-02 08:06:45 UTC
Eric W. Biederman was recently doing changes in that area on mainline, and he's
quite willing to debug it...
Comment 12 Andreas Kleen 2005-09-02 09:16:19 UTC
Can you please run klogconsole -r0 -l8 and then tell us what the latest
message is before the hang? Do you see "acpi_power_off called" ? 
Comment 13 Adam Schuett 2005-09-02 17:32:36 UTC
(In reply to comment #12)
> Can you please run klogconsole -r0 -l8 and then tell us what the latest
> message is before the hang? Do you see "acpi_power_off called" ? 
> 

I have done several shutdowns, and all hang at the same point.  The last few
lines are as follows:

md: stopping all md devices
md: md0 switched to read-only mode
shutdown: hda

Since updating to beta4, I did a lot of shutdowns (some in verbose, some in
silent mode) as well as some reboots in the same type.  Those results are as
follows:

Shutdowns in Verbose = 10 attempts, 4 succesfully powered off the laptop
Shutdowns in Silent = 10 attempts, 3 succesfully powered off the laptop
Reboots in Verbose = 5 attempts, 5 successul
Reboots in Silent = 5 attempts, 5 successul

Comment 14 Forgotten User v5inq2MBSE 2005-09-04 13:01:41 UTC
Hello,  
I've installed 10beta4 today on my ASUS M6842NW (first on tested on the 
notebook), I do have the same problem (did not do so many shutdowns so I can't 
send a statistic). I've tried BIOS Versions 0211A and 0214A (the latest with 
fixed battery entries in DSDT) with included and with fixed DSDT (using initrd 
recompiled with the included iasl). I also don't get a battery entrie 
in /proc/acpi. 
I've attached the two /var/log/boot.msg files with BIOS 0214A, maybe this will 
help you. 
 
I also had no ACPI problems (in this direction, Bug 61106 and 46254 - old suse 
bugzilla - are something else) with SUSE 9.1, 9.2 and 9.3 on this notebook. 
Comment 15 Forgotten User v5inq2MBSE 2005-09-04 13:06:12 UTC
Created attachment 48726 [details]
/var/lib/boot.msg without any dsdt changes
Comment 16 Forgotten User v5inq2MBSE 2005-09-04 13:07:31 UTC
Created attachment 48727 [details]
/var/lib/boot.msg with fixed dsdt changes
Comment 17 Adam Schuett 2005-09-15 17:37:28 UTC
(In reply to comment #14)

I looked through the files you attached and saw the differences between them.  I
checked my /var/lib and was unable to find anything similar on my machine.  Is
there something I need to do to create them on my machine

I am now on version 10 RC1 and still have the same issues.
Comment 18 Forgotten User v5inq2MBSE 2005-09-15 19:22:37 UTC
Since RC1 here shotdown works fine now. The battery state is also displayed 
correctly. The files are in /var/log/, sorry my fault. 
Comment 19 Thomas Renninger 2005-10-25 16:37:17 UTC
Could you try this kernel please:
ftp.suse.com/pub/people/trenn/10_0_kernel_reboot_fix/kernel-default-2.6.13-2.i586.rpm
If it works, please close this one, the patch will be included in the next YOU update kernel.
Comment 20 Adam Schuett 2005-10-26 14:51:35 UTC
I tried installing the kernel patch both through yast, as well as on the terminal window, which gave the following result:

ComputerName:/home/adams/downloads # rpm -Uvh kernel-default-2.6.13-2.i586.rpm 
Preparing...                ########################################### [100%]
        package kernel-default-2.6.13-2 is already installed
ComputerName:/home/adams/downloads # 

I am running 10.0 OSS, and the problems still occur
Comment 21 Thomas Renninger 2005-10-27 07:21:43 UTC
I didn't upgrade the version...
Please try with --force.

> I am running 10.0 OSS, and the problems still occur
->With the *new* kernel installed?
Comment 22 Adam Schuett 2005-10-27 13:30:32 UTC
I have used the force command for the kernel, and after installing I did an rpm -qi kernel-default to prove it was in fact installed.  I have done several shutdowns and it never powers off.

Restarting the system has been working fine since RC1.

Can I get the kernel source for the kernel package I just installed?  My VPN client will not work without having the kernel source.
Comment 23 Thomas Renninger 2005-10-28 07:44:33 UTC
Better override it with the original 10.0 kernel again if it doesn't help.

IBM Thinkpad A20m -> I already had another one with this problem (#118997).
See comment #15, be careful this bug report contains two different problems.

> ACPI was never turned on as it never functioned with SUSE in the past.
-> Are you sure? On above mentioned bug, it came out that ACPI modules cause the machine to not shutdown correctly. To verify could you try to:
 - rm /var/lib/acpi/laptop_modules
 - touch /var/lib/acpi/laptop_modules
 - chkconfig powersaved off
Can you shutdown after rebooting? (This only makes sense with ACPI on - if a /proc/acpi directory exists it's turned on).
Comment 24 Adam Schuett 2005-10-28 15:07:46 UTC
(In reply to comment #23)
I ran the following commands on my IBM a20p

>  - rm /var/lib/acpi/laptop_modules
>  - touch /var/lib/acpi/laptop_modules
>  - chkconfig powersaved off

I performed a reboot, and then did a full shutdown and it shutdown and powered itself off.  

HOWEVER, it now gives an error every time I boot into gui stating "The powersave daemon is not running.  Starting it will improve performance: /usr/sbin/rcpowersaved start"

I opened a console and typed in 'rcpowersaved status' and it stated unused.  I did an 'rcpowersaved start' and then a full shutdown and it worked again.

The error still appears afterward, but because it was not set to start in RunLevelEditor.  I set it in RunLevelEditor to start powersaved in 3&5 and then did a reboot.  It does not shutdown, but does not give me an error.  The only way shutting the laptop off works now is to go into a console and type 'rcpowersaved stop', then 'rcpowersaved start', otherwise the machine never shuts off.

Comment 25 Thomas Renninger 2005-10-29 13:43:46 UTC
Puhhh, too much info here mixed up...:
To avoid the nasty message you have to disable the kpowersave applet in the kde kicker (the power cable on the down-right). You can start it later by simply writing kpowersave into an X-shell). Another nice example why we need a button to disable the message, Danny (have a quick look at comment #24).

Let me tell you how far I came on Holgi's machine:
  1) The problem seem to come from the cpufreq or acpi modules as not starting
     the powersave daemon helps.
  2) Simply loading ACPI modules is not enough to trigger the problem
  3) ec_burst=0/1 does not help

Hmmm, maybe it is the speedstep-ich or speedstep-smi modules (that should manage cpufreq on this machine, but do not work for what reasons ever...). Maybe some bad data is written into the wrong register...
To verify you could still start the powersave daemon (chkconfig powersaved on).
But modifiy CPUFREQD_MODULE="off" in /etc/sysconfig/powersave/cpufreq before (and reboot).
Does it help?

If it does not help you could set:
ACPI_MODULES="NONE" in /etc/sysconfig/powersave/common
If this does help, you could slowly add one module after the other until you found the bad one...
Yes that takes some time..., but I fear there is no other way to find it, booting into init 3 should be enough (add a simple 3 to the boot options). Going through all services in the Yast->System->Runlevel editor and disable all unneeded services also helps a lot for faster booting.
Comment 26 Thomas Renninger 2005-11-04 08:59:06 UTC
*** Bug 118997 has been marked as a duplicate of this bug. ***
Comment 27 Adam Schuett 2005-11-17 20:41:37 UTC
By adding acpi=off in the bootloader for SUSE, the problem is gone and it powers off without fail.

Because of a past fix which was a new kernel, I have no kernel source for my machine.  Because of this, I am unable to install and AS400 (IBM5250) or VPN clients.  Is there a place where I can get the kernel source, or is there a way to revert back to a former kernel in which I have the kernel source for?
Comment 28 Thomas Renninger 2005-11-18 08:17:16 UTC
This is the latest kotd (10.0 kernel with latest fixes):
ftp.suse.com/pub/people/mantel/kotd/10.0-i386/SL100_BRANCH/kernel-default.i586.rpm

and here are the sources of it:
ftp.suse.com/pub/people/mantel/kotd/10.0-i386/SL100_BRANCH/kernel-source.rpm
Comment 29 Adam Schuett 2005-11-18 16:52:38 UTC
The kernel-source file you referred to does not exist.  Can you please point me to the correct kernel-source?


(In reply to comment #28)
> This is the latest kotd (10.0 kernel with latest fixes):
> ftp.suse.com/pub/people/mantel/kotd/10.0-i386/SL100_BRANCH/kernel-default.i586.rpm
> 
> and here are the sources of it:
> ftp.suse.com/pub/people/mantel/kotd/10.0-i386/SL100_BRANCH/kernel-source.rpm
> 

Comment 30 Thomas Renninger 2005-12-09 10:02:15 UTC
Sorry for not repling, I was on holidays..., a YOU update kernel should be available now.