Re: ext3: Freeing blocks not in datazone

[Eric Sandeen <sandeen@redhat.com>]

Jan Kara wrote:
>> Mark Lord wrote:
>>> Eric Sandeen wrote:
>>>> Mark Lord wrote:
>>>>> One of my test systems here, running 2.6.25-rc8-git*,
>>>>> just now crapped out with this in the tail end of dmesg:
>>>>>
>>>>> [   20.927780] EXT3-fs error (device sda1): ext3_free_blocks: Freeing blocks not in datazone - block
>>>>> [   20.928756] Aborting journal on device sda1.
>>>> I've got a similar report from the F9 installer, also on 2.6.25-rc8* ...
>>>>  trying to reproduce locally.
>>> ..
>>>
>>> Well, if it's of any help, here it was on a 2.4GHz Intel QuadCore,
>>> 32-bit kernel/userspace, 4GB RAM, PAE-enabled.
>> Any chance you had done a resize2fs, online or offline, before this?
>> Just a guess based on the installer thing I'm looking at...
>   Hmm, Eric, how exactly did the corruption looked like? Are you running
> SLUB allocator? I'm just wondering whether it doesn't have something in
> common with the memory corruption as discussed in the thread starting at
> http://lkml.org/lkml/2008/4/19/85

It turns out that that problem (sorry, should have followed up) was due
to the installer not copying the last bit of a filesystem image onto the
device, and the fs was then trying to use whatever it found on the
un-copied portion of the disk as metadata.

So not an ext3 bug in that case.

-Eric