ext4 corruptions

ext4 corruptions

[Alexey Fisher]

I have really bad ext4 corruptions, i needed to reinstall 3 of my test
systems.
The symptoms:
is use karmic beta for this with ubuntu supplied kernel 2.6.31 ... after
testing 2.6.32-rc[1-3] and booting again to 2.6.31 fsck will find some
bad things and will try to fix it. After this action is the system
unusable.

Do any one had same problem? The is no sans to bisect it, after some
jumps between 2.6.31 and 2.6.32 the system will be broken.  Is it kernel
or fsck problem?

Re: ext4 corruptions

[Jiri Kosina]

[ adding some CCs so that this doesn't get lost in the lkml wilderness ]

On Fri, 9 Oct 2009, Alexey Fisher wrote:

> I have really bad ext4 corruptions, i needed to reinstall 3 of my test
> systems.
> The symptoms:
> is use karmic beta for this with ubuntu supplied kernel 2.6.31 ... after
> testing 2.6.32-rc[1-3] and booting again to 2.6.31 fsck will find some
> bad things and will try to fix it. After this action is the system
> unusable.

Does this happen also when you switch from vanilla 2.6.32-rcX to vanilla 
2.6.31, or only from vanilla 2.6.32-rcX to "Ubuntu 2.6.31"?

> Do any one had same problem? The is no sans to bisect it, after some
> jumps between 2.6.31 and 2.6.32 the system will be broken.  Is it kernel
> or fsck problem?

-- 
Jiri Kosina
SUSE Labs, Novell Inc.

Re: ext4 corruptions

[Eric Sandeen]

Jiri Kosina wrote:
> [ adding some CCs so that this doesn't get lost in the lkml wilderness ]
> 
> On Fri, 9 Oct 2009, Alexey Fisher wrote:
> 
>> I have really bad ext4 corruptions, i needed to reinstall 3 of my test
>> systems.
>> The symptoms:
>> is use karmic beta for this with ubuntu supplied kernel 2.6.31 ... after
>> testing 2.6.32-rc[1-3] and booting again to 2.6.31 fsck will find some
>> bad things and will try to fix it. After this action is the system
>> unusable.
> 
> Does this happen also when you switch from vanilla 2.6.32-rcX to vanilla 
> 2.6.31, or only from vanilla 2.6.32-rcX to "Ubuntu 2.6.31"?
> 
>> Do any one had same problem? The is no sans to bisect it, after some
>> jumps between 2.6.31 and 2.6.32 the system will be broken.  Is it kernel
>> or fsck problem?
> 

Not sure if this is the same as
http://bugzilla.kernel.org/show_bug.cgi?id=14354 but it's similarly
lacking in useful info ;)

Please provide the fsck output, any kernel messages, etc ...

-Eric

Re: ext4 corruptions

[Alexey Fisher]

Am Freitag, den 09.10.2009, 17:03 -0500 schrieb Eric Sandeen:
> Jiri Kosina wrote:
> > [ adding some CCs so that this doesn't get lost in the lkml wilderness ]
> > 
> > On Fri, 9 Oct 2009, Alexey Fisher wrote:
> > 
> >> I have really bad ext4 corruptions, i needed to reinstall 3 of my test
> >> systems.
> >> The symptoms:
> >> is use karmic beta for this with ubuntu supplied kernel 2.6.31 ... after
> >> testing 2.6.32-rc[1-3] and booting again to 2.6.31 fsck will find some
> >> bad things and will try to fix it. After this action is the system
> >> unusable.
> > 
> > Does this happen also when you switch from vanilla 2.6.32-rcX to vanilla 
> > 2.6.31, or only from vanilla 2.6.32-rcX to "Ubuntu 2.6.31"?
> > 
> >> Do any one had same problem? The is no sans to bisect it, after some
> >> jumps between 2.6.31 and 2.6.32 the system will be broken.  Is it kernel
> >> or fsck problem?
> > 
> 
> Not sure if this is the same as
> http://bugzilla.kernel.org/show_bug.cgi?id=14354 but it's similarly
> lacking in useful info ;)
> 
> Please provide the fsck output, any kernel messages, etc ...
> 
> -Eric

Thank you for the tip. These systems i need for testing graphic
regressions, so i'll try reproduce it on virtual system.

Alexey

Re: ext4 corruptions

[Theodore Tso]

On Fri, Oct 09, 2009 at 09:20:13PM +0200, Alexey Fisher wrote:
> I have really bad ext4 corruptions, i needed to reinstall 3 of my test
> systems.
> The symptoms:
> is use karmic beta for this with ubuntu supplied kernel 2.6.31 ... after
> testing 2.6.32-rc[1-3] and booting again to 2.6.31 fsck will find some
> bad things and will try to fix it. After this action is the system
> unusable.
> 
> Do any one had same problem? The is no sans to bisect it, after some
> jumps between 2.6.31 and 2.6.32 the system will be broken.  Is it kernel
> or fsck problem?

Can you reproduce this reliably?  What sort of things did fsck report?
Was it after a suspend/resume that failed?

I'm using a Ubuntu 9.04 with a 2.6.32-rc1+ mainline kernel, and it
works fine for me.  So we really need many more details what sort of
filesystem corruption (not just "bad things"), and more details about
what you did while testing a newer kernel, and what sort of things
fsck complained about.

						- Ted