[gentoo-user] Kernel Panic fglrx

[gentoo-user] Kernel Panic fglrx

[4k3nd0]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

hi guys,

somehow my fglrx is unstable. Got sometime a crashing fglrx that
pulling down the hole system. Someone maybe a idea to fix this?

http://pastebin.com/izFnWD7N

I'm using the 11.9 ati-drivers, i got a ATI Readon HD6990 M

uname -a
Linux Slaxy 3.0.7-gentoo #1 SMP Sat Oct 22 20:44:41 CEST 2011 x86_64
Intel(R) Core(TM) i7-2820QM CPU @ 2.30GHz GenuineIntel GNU/Linux

Using xorg-server 1.10.4-r1

Greeting's from Germany
4k3nd0
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.17 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iQEcBAEBAgAGBQJOpXfQAAoJEARqOhiQNM7Zm9cH/0cs8Pkb8YIGKloBVMQhdFxq
+bzFWxbv6aLTqph5uxUkcTZoWYiAZw55TVImaB3RIVbd248RHGrCJMwWywOnNHCY
iJ2IVmUrXMFPfwTBSi1L9+2eMEZJqr8jcHCKEUUkJvKpgMcg4koTcFLWu/1RjMdw
JecoMUwYZAb+0NC8kN+CQpytn6KlXdoMKYhprQ8QOj2HfCe3ZF5R3Shz62Nj99NB
cAx87cX8YLYynPbhrmj2Fg44Q5ItsO4gAw4Z+zAC2Ikih+3jnSzFhwCPk1TRhmcn
KffLGxun9eAKaJqKhcWXYWH9k3qYqN16x0/3cESAHFqY5FETOD3PRTu1shS3S0k=
=ChkA
-----END PGP SIGNATURE-----

Re: [gentoo-user] Kernel Panic fglrx

[Michael Schreckenbauer]

Hi,

Am Montag, 24. Oktober 2011, 16:36:00 schrieb 4k3nd0:
> hi guys,
> 
> somehow my fglrx is unstable. Got sometime a crashing fglrx that
> pulling down the hole system. Someone maybe a idea to fix this?
> 
> http://pastebin.com/izFnWD7N

[fglrx] ASIC hang happened

is a message indicating a deadlock in a compute-kernel (shader, GPGPU), which 
in almost all cases is a userspace bug or the timeout in fglrx is too low for 
the kernel in question.
I'm no expert in tweaking fglrx, but afaict there's nothing you can do except 
wait for an update of the driver.
You could also try the opensource-drivers (I think r600 in your case), cause 
those do their own heuristics to determine hanging or deadlocked compute-
kernels.

Best,
Michael

Re: [gentoo-user] Kernel Panic fglrx

[4k3nd0]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 10/24/11 17:31, Michael Schreckenbauer wrote:
> Hi,
> 
> Am Montag, 24. Oktober 2011, 16:36:00 schrieb 4k3nd0:
>> hi guys,
>> 
>> somehow my fglrx is unstable. Got sometime a crashing fglrx that 
>> pulling down the hole system. Someone maybe a idea to fix this?
>> 
>> http://pastebin.com/izFnWD7N
> 
> [fglrx] ASIC hang happened
> 
> is a message indicating a deadlock in a compute-kernel (shader,
> GPGPU), which in almost all cases is a userspace bug or the timeout
> in fglrx is too low for the kernel in question. I'm no expert in
> tweaking fglrx, but afaict there's nothing you can do except wait
> for an update of the driver. You could also try the
> opensource-drivers (I think r600 in your case), cause those do
> their own heuristics to determine hanging or deadlocked compute- 
> kernels.
> 
> Best, Michael
> 
> 
> 
Thank you. Do you think using a newer version of xserver would help?
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.17 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iQEcBAEBAgAGBQJOpY5BAAoJEARqOhiQNM7ZwysH/RDWXwBdqLHgKL2R5tSgcOal
tL3gYduHOB0vJdITrbm3hvAdq6axvOVW3UXlI9dskLvD/OlA2uH9U70Lbh5GjdZH
+bqp8+B8gKBmm0yNIT9/PWPAI+yxSpf1JPXLwV08R5B9F1pRL4sXMhLzcjNbywZM
V9V9tBaE/vVaAwvipN8kn9sK7vCorLkA9LcOF0IzNBJDSJIqxUZR9x99O2FA51re
57MyoQPyXxX0zlqe8mHHggLKCZnKYNmI5XNngX++2GL6vUVeerxKqJcPQJZ2vX+V
xScQY1jFOqt1QPsVjqmhlTSbND7tcTl5yY6inrdS7ar1FhzpJjHZ8l4hJVS4ayU=
=P9Xf
-----END PGP SIGNATURE-----

Re: [gentoo-user] Kernel Panic fglrx

[Michael Schreckenbauer]

Am Montag, 24. Oktober 2011, 18:11:45 schrieb 4k3nd0:
> On 10/24/11 17:31, Michael Schreckenbauer wrote:
> > Hi,
> > 
> > Am Montag, 24. Oktober 2011, 16:36:00 schrieb 4k3nd0:
> >> hi guys,
> >> 
> >> somehow my fglrx is unstable. Got sometime a crashing fglrx that
> >> pulling down the hole system. Someone maybe a idea to fix this?
> >> 
> >> http://pastebin.com/izFnWD7N
> > 
> > [fglrx] ASIC hang happened
> > 
> > is a message indicating a deadlock in a compute-kernel (shader,
> > GPGPU), which in almost all cases is a userspace bug or the timeout
> > in fglrx is too low for the kernel in question. I'm no expert in
> > tweaking fglrx, but afaict there's nothing you can do except wait
> > for an update of the driver. You could also try the
> > opensource-drivers (I think r600 in your case), cause those do
> > their own heuristics to determine hanging or deadlocked compute-
> > kernels.
> > 
> > Best, Michael
> 
> Thank you. Do you think using a newer version of xserver would help?

Always worth a try imo. But I wouldn't expect too much from doing this. The 
driver thinks, that a compute-kernel hangs or is deadlocked. If this 
assumption is correct, the userspace app needs a fix otherwise the driver is 
faulty.
Is this problem related to running a special app or DE? Or does it happen 
"always"?

Best,
Michael

Re: [gentoo-user] Kernel Panic fglrx

[4k3nd0]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 10/24/11 20:27, Michael Schreckenbauer wrote:
> Am Montag, 24. Oktober 2011, 18:11:45 schrieb 4k3nd0:
>> On 10/24/11 17:31, Michael Schreckenbauer wrote:
>>> Hi,
>>> 
>>> Am Montag, 24. Oktober 2011, 16:36:00 schrieb 4k3nd0:
>>>> hi guys,
>>>> 
>>>> somehow my fglrx is unstable. Got sometime a crashing fglrx
>>>> that pulling down the hole system. Someone maybe a idea to
>>>> fix this?
>>>> 
>>>> http://pastebin.com/izFnWD7N
>>> 
>>> [fglrx] ASIC hang happened
>>> 
>>> is a message indicating a deadlock in a compute-kernel
>>> (shader, GPGPU), which in almost all cases is a userspace bug
>>> or the timeout in fglrx is too low for the kernel in question.
>>> I'm no expert in tweaking fglrx, but afaict there's nothing you
>>> can do except wait for an update of the driver. You could also
>>> try the opensource-drivers (I think r600 in your case), cause
>>> those do their own heuristics to determine hanging or
>>> deadlocked compute- kernels.
>>> 
>>> Best, Michael
>> 
>> Thank you. Do you think using a newer version of xserver would
>> help?
> 
> Always worth a try imo. But I wouldn't expect too much from doing
> this. The driver thinks, that a compute-kernel hangs or is
> deadlocked. If this assumption is correct, the userspace app needs
> a fix otherwise the driver is faulty. Is this problem related to
> running a special app or DE? Or does it happen "always"?
> 
> Best, Michael
> 
> 
First i suspect flash player, but it's randomly. There is no special
reason i can detect.

I will try using the other Kernel, maybe it will help.


-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.17 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iQEcBAEBAgAGBQJOpbjxAAoJEARqOhiQNM7ZnvoH/0tj+06hg/8Si0A0usMDZAYn
3L6vyGjmYmOMW8eTzw7MTE3QFbw8rFrF9KeONzasZU83Z12saA/f+uRZkExcmrM1
kjYCDyzqlOqPz/u76X3HYP8DKIdFxI/mOws/xKf/kdOioI1TOpas5Lm18VkzfWiF
JYQvJ7KjUIo/MfqMoTtWjUMn6lDfdQ9fs9DvQCQQuJtPZXp/N4cZb3WFHhXAl8mw
LnoWi3vBh7EZv3W4pisaxKVti2cGmUFjzBjarQL89EB67coya6Iy7ZIDWoPHIFC0
WroGKaIFPwbMqWiSZxqSgwWAycpgs6Z9OuiQaz/BysYBQ7cwm6MosBG5stAbU3c=
=W9dW
-----END PGP SIGNATURE-----

Re: [gentoo-user] Kernel Panic fglrx

[4k3nd0]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 10/24/2011 08:27 PM, Michael Schreckenbauer wrote:
> Am Montag, 24. Oktober 2011, 18:11:45 schrieb 4k3nd0:
>> On 10/24/11 17:31, Michael Schreckenbauer wrote:
>>> Hi,
>>> 
>>> Am Montag, 24. Oktober 2011, 16:36:00 schrieb 4k3nd0:
>>>> hi guys,
>>>> 
>>>> somehow my fglrx is unstable. Got sometime a crashing fglrx
>>>> that pulling down the hole system. Someone maybe a idea to
>>>> fix this?
>>>> 
>>>> http://pastebin.com/izFnWD7N
>>> 
>>> [fglrx] ASIC hang happened
>>> 
>>> is a message indicating a deadlock in a compute-kernel
>>> (shader, GPGPU), which in almost all cases is a userspace bug
>>> or the timeout in fglrx is too low for the kernel in question.
>>> I'm no expert in tweaking fglrx, but afaict there's nothing you
>>> can do except wait for an update of the driver. You could also
>>> try the opensource-drivers (I think r600 in your case), cause
>>> those do their own heuristics to determine hanging or
>>> deadlocked compute- kernels.
>>> 
>>> Best, Michael
>> 
>> Thank you. Do you think using a newer version of xserver would
>> help?
> 
> Always worth a try imo. But I wouldn't expect too much from doing
> this. The driver thinks, that a compute-kernel hangs or is
> deadlocked. If this assumption is correct, the userspace app needs
> a fix otherwise the driver is faulty. Is this problem related to
> running a special app or DE? Or does it happen "always"?
> 
> Best, Michael
> 
> 
I tried also updating to a newer Version of xorg-xserver als well
using the 3.0.0.-generic-12 .

Neither did it.

The problem seems to be gone with a Linux Kernel 3.1.0
(no-gentoo-patch). I using now the lastet ATI-Driver from the Webpage,
but can't tell how stable it is. I'll see it in around a week.

But thanks for the help!

Greeting's from Germany
4k3nd0

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.17 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iQEcBAEBAgAGBQJOp8/IAAoJEARqOhiQNM7ZlvgIALswN1Rh7HxuBvIIM9+VmSI7
bGZ7ezAvX0Su0NfnH1npwrGwFdxfa6q8UVdCdyy7PzEZ54oYfaOUJOnn9FtIGMjs
HGtS4Kek4zvxa3yRYH2yoarjTUiyLhTzk5TdbwcfgKQ2RzSPhNB+W2CeEakNPJRv
uyIKLJysslwbdgx30hsbldBEV5MbU1tqII8WEk4pzFjE5Se9OgXgo2TQQjOYVx/V
PB9w2oTUZ3TIvEV3J+IzbiTPhH2FbeJmRCbiM11JrSGcYPPZC4EzDgunUvnrwtdx
0c8fHZG04INnk4PdfHxCm12HJ+nRPZzTEUe8fe+GQZ1+H3apyIqLWabwTKzAQI0=
=cdC0
-----END PGP SIGNATURE-----

Re: [gentoo-user] Kernel Panic fglrx

[Volker Armin Hemmann]

Am Mittwoch 26 Oktober 2011, 11:15:52 schrieb 4k3nd0:
> On 10/24/2011 08:27 PM, Michael Schreckenbauer wrote:
> > Am Montag, 24. Oktober 2011, 18:11:45 schrieb 4k3nd0:
> >> On 10/24/11 17:31, Michael Schreckenbauer wrote:
> >>> Hi,
> >>> 
> >>> Am Montag, 24. Oktober 2011, 16:36:00 schrieb 4k3nd0:
> >>>> hi guys,
> >>>> 
> >>>> somehow my fglrx is unstable. Got sometime a crashing fglrx
> >>>> that pulling down the hole system. Someone maybe a idea to
> >>>> fix this?
> >>>> 
> >>>> http://pastebin.com/izFnWD7N
> >>> 
> >>> [fglrx] ASIC hang happened
> >>> 
> >>> is a message indicating a deadlock in a compute-kernel
> >>> (shader, GPGPU), which in almost all cases is a userspace bug
> >>> or the timeout in fglrx is too low for the kernel in question.
> >>> I'm no expert in tweaking fglrx, but afaict there's nothing you
> >>> can do except wait for an update of the driver. You could also
> >>> try the opensource-drivers (I think r600 in your case), cause
> >>> those do their own heuristics to determine hanging or
> >>> deadlocked compute- kernels.
> >>> 
> >>> Best, Michael
> >> 
> >> Thank you. Do you think using a newer version of xserver would
> >> help?
> > 
> > Always worth a try imo. But I wouldn't expect too much from doing
> > this. The driver thinks, that a compute-kernel hangs or is
> > deadlocked. If this assumption is correct, the userspace app needs
> > a fix otherwise the driver is faulty. Is this problem related to
> > running a special app or DE? Or does it happen "always"?
> > 
> > Best, Michael
> 
> I tried also updating to a newer Version of xorg-xserver als well
> using the 3.0.0.-generic-12 .

which is not a vanilla kernel.

vanilla-sources 3.0.7 is

-- 
#163933

Re: [gentoo-user] Kernel Panic fglrx

[Volker Armin Hemmann]

Am Montag 24 Oktober 2011, 16:36:00 schrieb 4k3nd0:
> hi guys,
> 
> somehow my fglrx is unstable. Got sometime a crashing fglrx that
> pulling down the hole system. Someone maybe a idea to fix this?
> 
> http://pastebin.com/izFnWD7N
> 
> I'm using the 11.9 ati-drivers, i got a ATI Readon HD6990 M
> 
> uname -a
> Linux Slaxy 3.0.7-gentoo #1 SMP Sat Oct 22 20:44:41 CEST 2011 x86_64
> Intel(R) Core(TM) i7-2820QM CPU @ 2.30GHz GenuineIntel GNU/Linux
> 
> Using xorg-server 1.10.4-r1
> 
> Greeting's from Germany
> 4k3nd0

well, the first thing is always: don't use patched kernels.

gentoo kernels are patched.

Use vanilla. See if the problems persist.

fglrx has been solid for me for the last half of a year. With uptimes of two 
weeks and more, reboots always caused by kernel updates.

-- 
#163933

Re: [gentoo-user] Kernel Panic fglrx

[Michael Schreckenbauer]

Am Montag, 24. Oktober 2011, 18:49:36 schrieb Volker Armin Hemmann:
> Am Montag 24 Oktober 2011, 16:36:00 schrieb 4k3nd0:
> > hi guys,
> > 
> > somehow my fglrx is unstable. Got sometime a crashing fglrx that
> > pulling down the hole system. Someone maybe a idea to fix this?
> > 
> > http://pastebin.com/izFnWD7N
> > 
> > I'm using the 11.9 ati-drivers, i got a ATI Readon HD6990 M
> > 
> > uname -a
> > Linux Slaxy 3.0.7-gentoo #1 SMP Sat Oct 22 20:44:41 CEST 2011 x86_64
> > Intel(R) Core(TM) i7-2820QM CPU @ 2.30GHz GenuineIntel GNU/Linux
> > 
> > Using xorg-server 1.10.4-r1
> > 
> > Greeting's from Germany
> > 4k3nd0
> 
> well, the first thing is always: don't use patched kernels.
> 
> gentoo kernels are patched.
> 
> Use vanilla. See if the problems persist.

Good advice in general. But afaik the deadlock detection happens completely in 
the driver, so I am not convinced other parts of the kernel could cause this. 
I might be wrong about this though.

> fglrx has been solid for me for the last half of a year. With uptimes of two
> weeks and more, reboots always caused by kernel updates.

Best,
Michael