Re: [regression] suspend is flakey, "shutdown -h now" does not power off

Re: [regression] suspend is flakey, "shutdown -h now" does not power off

[Matthieu Dubuget]

Hello,

I'm sending this message again without attachments etc., because I
think (apologies if I'm wrong) that my first try did not succeed. The
original message was also sent to debian bug tracking system. You can
find it there, with the attachments (acpidump outputs, dmesg output)
and with some more (maybe less relevant) tests.

http://bugs.debian.org/648754

I'm experiencing some acpi-related problems with my laptop
([0.000000] DMI: Compal PBL21/PBL21, BIOS 1.71 04/19/2011)

With the most recent kernels I tested:

- "shutdown -h now" or "shutdown -h -P now":
     * produce an apparently normal shutdown log
     * then echoes "System halted"
     * but does never power off the computer

- sometime (rare & randomly), when I re-open the lid after the computer was
    manually turned off, it boots

- But most of the time, pm-suspend (or "suspend" from gnome-shell) is
    working (this fails sometime with a freeze, and fans at full speed),
    as expected

List of kernels I could test:

     - affected:

         * Debian 3.0.0-6 (which is very close to upstream stable v3.0.8)
         * Debian 3.1.0-1~experimental.1 (which is basically linus's v3.1)
         * Debian 2.6.32-39 (close to longterm v2.6.32.43)
         * Ubuntu 3.0.0-12.20-generic (3.0.4 ? started from the Ubuntu 
11.10 live CD)

     - not affected:

         * Ubuntu 2.6.32.41+drm33.18 (started from the Ubuntu 10.04.3 LTS
           live CD)

Attached files:

    - dmesg-not-working.txt: dmesg from an affected kernel
    - acpidump-not-working.txt: acpidump output
    - dmesg-working.txt: dmesg from an unaffected kernel
    - acpidump-working.txt: acpidump output obtain with a working kernel
    - acpidump-ubuntu-11-10.txt: obtained with 11.10 debian CD. Not 
working. Different dump, for the record.

Note:

       $ diff acpidump-not-working.txt acpidump-working.txt
       2805a2806,2813
  > RSDT @ 0xaf7fe0ac
  >   0000: 52 53 44 54 58 00 00 00 01 8a 49 4e 53 59 44 45 
RSDTX.....INSYDE
  >   0010: 48 52 20 43 52 42 20 20 01 00 00 00 20 20 20 20  HR CRB  ....
  >   0020: 13 00 00 01 00 c0 7f af 00 d0 7f af 00 b0 7f af 
................
  >   0030: 00 a0 7f af 00 90 7f af 00 f0 7e af 00 e0 7e af 
..........~...~.
  >   0040: 00 d0 7e af 00 b0 7e af 00 80 7e af 00 70 7e af 
..~...~...~..p~.
  >   0050: 00 60 7e af 00 50 7e af                          .`~..P~.
  >

Reference: some more tests are on debian's BTS: 
http://bugs.debian.org/648754

I would be happy to provide any help you would think necessary.

Salutations

Matthieu Dubuget

Re: [regression] suspend is flakey, "shutdown -h now" does not power off

[Matthieu Dubuget]

Hello,

I have some shutdown problems with my notebook, and it seems to be an
(old) regression, since it used to work.

(http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=648754#38)

I compared two acpidump outputs with a working and with a non working
kernel: it appears that the only difference is that there is no RSDT
table in the non working one.

Is it something expected, or could it be an explanation of my
problems?

How could I solve this?

Thanks in advance for any advice.

Salutations

Matthieu Dubuget

Re: [regression] suspend is flakey, "shutdown -h now" does not power off

[Rafael J. Wysocki]

On Saturday, April 27, 2013 04:29:26 PM Matthieu Dubuget wrote:
> Hello,
> 
> I have some shutdown problems with my notebook, and it seems to be an
> (old) regression, since it used to work.
> 
> (http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=648754#38)
> 
> I compared two acpidump outputs with a working and with a non working
> kernel: it appears that the only difference is that there is no RSDT
> table in the non working one.
> 
> Is it something expected, or could it be an explanation of my
> problems?
> 
> How could I solve this?
> 
> Thanks in advance for any advice.

What's the newest kernel you tried?

Rafael


-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.

Re: [regression] suspend is flakey, "shutdown -h now" does not power off

[Matthieu Dubuget]

Le 27/04/2013 16:42, Rafael J. Wysocki a écrit :
>
>  What's the newest kernel you tried?
>
>  Rafael
>
>

Hello Rafael,

It used to worked when I bought the computer. And I can check this
with a live CD:

* Ubuntu 2.6.32.41+drm33.18 (started from the Ubuntu 10.04.3 LTS live
   CD)

Since then, I tried:

* Debian 3.0.0-6 (which is very close to upstream stable v3.0.8)
* Debian 3.1.0-1~experimental.1 (which is basically linus's v3.1)
* Debian 2.6.32-39 (close to longterm v2.6.32.43)
* Ubuntu 3.0.0-12.20-generic (3.0.4 ? started from the Ubuntu live CD)

* And this morning, I downloaded, compiled and ran 3.8.10 from
   kernel.org without more success.

Salutations

Matt




--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [regression] suspend is flakey, "shutdown -h now" does not power off

[Rafael J. Wysocki]

On Saturday, April 27, 2013 04:43:46 PM Matthieu Dubuget wrote:
> 
> Le 27/04/2013 16:42, Rafael J. Wysocki a écrit :
> >
> >  What's the newest kernel you tried?
> >
> >  Rafael
> >
> >
> 
> Hello Rafael,
> 
> It used to worked when I bought the computer. And I can check this
> with a live CD:
> 
> * Ubuntu 2.6.32.41+drm33.18 (started from the Ubuntu 10.04.3 LTS live
>    CD)
> 
> Since then, I tried:
> 
> * Debian 3.0.0-6 (which is very close to upstream stable v3.0.8)
> * Debian 3.1.0-1~experimental.1 (which is basically linus's v3.1)
> * Debian 2.6.32-39 (close to longterm v2.6.32.43)
> * Ubuntu 3.0.0-12.20-generic (3.0.4 ? started from the Ubuntu live CD)
> 
> * And this morning, I downloaded, compiled and ran 3.8.10 from
>    kernel.org without more success.

I see.

Did it ever worked with a kernel that's not from Ubuntu?

Rafael


-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [regression] suspend is flakey, "shutdown -h now" does not power off

[Matthieu Dubuget]

Le 27/04/2013 17:09, Rafael J. Wysocki a écrit :
> Did it ever worked with a kernel that's not from Ubuntu?

I do not know… You suggest that I shourd find a « pure »
kernel that is working, around 2.6.30 version?

I'll come back when/if I will have found one.

Salutations

Matt
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [regression] suspend is flakey, "shutdown -h now" does not power off

[Rafael J. Wysocki]

On Saturday, April 27, 2013 05:16:37 PM Matthieu Dubuget wrote:
> Le 27/04/2013 17:09, Rafael J. Wysocki a écrit :
> > Did it ever worked with a kernel that's not from Ubuntu?
> 
> I do not know… You suggest that I shourd find a « pure »
> kernel that is working, around 2.6.30 version?

Yes.  Basically, I'd like to know if the fact that it worked was due to an
Ubuntu patch or something we did in the mainline that changed.

Also, it would be good to know what the failing hardware is.

> I'll come back when/if I will have found one.

Thanks!

Rafael


-- 
I speak only for myself.
Rafael J. Wysocki, Intel Open Source Technology Center.
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [regression] suspend is flakey, "shutdown -h now" does not power off

[Aaron Lu]

Filed a bug report for you here:
https://bugzilla.kernel.org/show_bug.cgi?id=57231

Thanks,
Aaron

On 04/27/2013 10:29 PM, Matthieu Dubuget wrote:
> Hello,
> 
> I have some shutdown problems with my notebook, and it seems to be an
> (old) regression, since it used to work.
> 
> (http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=648754#38)
> 
> I compared two acpidump outputs with a working and with a non working
> kernel: it appears that the only difference is that there is no RSDT
> table in the non working one.
> 
> Is it something expected, or could it be an explanation of my
> problems?
> 
> How could I solve this?
> 
> Thanks in advance for any advice.
> 
> Salutations
> 
> Matthieu Dubuget
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

Re: [regression] suspend is flakey, "shutdown -h now" does not power off

[Matthieu Dubuget]

Le 28/04/2013 03:27, Aaron Lu a écrit :
> Filed a bug report for you here:
> https://bugzilla.kernel.org/show_bug.cgi?id=57231
>
> Thanks,
> Aaron
>

I am to thank you. I'll try to gather more details to narrow the problem 
and add them in the bug report.

Is there any common known distrib that uses the kernel untouched I could 
use to narrow the revision
at which the problem was introduced?

Thanks

Matt
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [regression] suspend is flakey, "shutdown -h now" does not power off

[Matthieu Dubuget]

It seems that the problem reported is not a regression in acpi, but a
bug in the amd64 version.  For all the live systems I tested, each
time I used a 386/686 version it worked fine, each time I tried a
amd64 version it failed.

Hoping this will help narrowing the bug!

 From now on, I'm not sure about what I could do to understand the
problem?

Maybe a kernel recompilation with some debugging options added? But
which?

Salutations, and thanks for your help

Matt