Repeatable r8169 kernel oops testcase

Repeatable r8169 kernel oops testcase

[Dale Ogilvie]

[-- Attachment #1: Type: text/plain, Size: 794 bytes --]

Hello Francois,

after having some issues with networking on Fedora 10 I was given your
name to provide some data to. See this fedora forum
post :http://forums.fedoraforum.org/showthread.php?t=205609

I hope you'll be able to get to the bottom of this failure. I have
attached the information suggested in the above post, plus some further
information. In a nutshell, on my hardware the network is unusable after
bootup, until the user crashes the r8169 module by timing out a hostname
lookup. I have included the repeatable crash testcase in the attached
file.

I believe that simply removing the module and modprobing it back in
again after bootup also delivers a working network without the oops.

Let me know if you need any further information, or want me to test
further.

Regards

Dale


[-- Attachment #2: r8169_oops.tar.gz --]
[-- Type: application/x-compressed-tar, Size: 16053 bytes --]

Re: Repeatable r8169 kernel oops testcase

[Jay Cliburn]

On Sun, 30 Nov 2008 23:29:14 +1300
Dale Ogilvie <k1qeelx02@sneakemail.com> wrote:

> after having some issues with networking on Fedora 10 I was given your
> name to provide some data to. See this fedora forum
> post :http://forums.fedoraforum.org/showthread.php?t=205609

This line from your dmesg caught my eye.

Clocksource tsc unstable (delta = -249993826 ns)

Does booting with clocksource=hpet change anything?  What about the
others in 
/sys/devices/system/clocksource/clocksource0/available_clocksource

Jay

Re: Repeatable r8169 kernel oops testcase

[Francois Romieu]

Jay Cliburn <jcliburn@gmail.com> :
[...]
> Does booting with clocksource=hpet change anything?  What about the
> others in 
> /sys/devices/system/clocksource/clocksource0/available_clocksource

Once this is done, can your send your .config and tell if simply
issuing an 'ethtool -s eth0 autoneg off' then 'ethtool -s eth0 autoneg on' 
makes a difference ?

-- 
Ueimor

Re: Repeatable r8169 kernel oops testcase

[Dale Ogilvie]

[-- Attachment #1: Type: text/plain, Size: 952 bytes --]

[root@gordon ~]#
cat /sys/devices/system/clocksource/clocksource0/available_clocksource
acpi_pm jiffies tsc

Since I don't have hpet in my available_clocksources I tried notsc and
clocksource=acpi_pm. I also tried the two ethtool commands below.

None of these changes made any difference. Networking is dead until the
r8169 module is unloaded and reloaded. Also, while the network is up but
"dead", mouse input is erratic. Moving the mouse is not smooth at all.
Something is severely wrong while the r8169 module is in this state.

See attached .config.


> Jay Cliburn <jcliburn@gmail.com> :
> [...]
> > Does booting with clocksource=hpet change anything?  What about the
> > others in 
> > /sys/devices/system/clocksource/clocksource0/available_clocksource
> 
> Once this is done, can your send your .config and tell if simply
> issuing an 'ethtool -s eth0 autoneg off' then 'ethtool -s eth0 autoneg on' 
> makes a difference ?
> 
> -- 
> Ueimor
> 

[-- Attachment #2: config.gz --]
[-- Type: application/x-gzip, Size: 23379 bytes --]

Re: Repeatable r8169 kernel oops testcase

[Francois Romieu]

Dale Ogilvie <k1qeelx02@sneakemail.com> :
[...]
> None of these changes made any difference. Networking is dead until the
> r8169 module is unloaded and reloaded. Also, while the network is up but
> "dead", mouse input is erratic. Moving the mouse is not smooth at all.
> Something is severely wrong while the r8169 module is in this state.

Is it possible for you to log 'mii-tool -v -v eth0' output:
- before the watchdog fires
- after it fires and before the module removal
- after the module removal / insertion, when everything seems right

-- 
Ueimor