Corruption of in-memory data detected

[Eric Sandeen]

Thomas Gutzler wrote:
> Eric Sandeen wrote:
>> Thomas Gutzler wrote:
...

>>> What can I do to help getting this fixed?
>>>
>> can you make an xfs_metadump of the filesystem in question, and then try
>> an xfs_repair?  Capture/save the repair output.  If repair finds errors,
>> then perhaps the bug is triggered by bad error checking on a corrupted
>> image, and we might reproduce it w/ the metadump image.
> 
> I tried...
> 
> root at io:~# xfs_metadump /dev/sda xfs_metadump_sda
> *** glibc detected *** xfs_db: double free or corruption (out):
> 0x00000000017b8000 ***

...

> Aborted

:(  what version of xfsprogs?

> and tried again:
> root at io:~# xfs_metadump -w /dev/sda xfs_metadump_sda
> root at io:~# ll xfs_metadump_sda
> -rw-r--r-- 1 root root 588407296 2009-03-11 19:09 xfs_metadump_sda
> (where do I upload this to?)

You can bzip2 it and probably shrink it pretty well (it's sparse).  See
how big that is, and we can find a place for it.

> xfs_repair fixed "bad names" of 4 inodes (see attached log file)
> 
> another xfs_metadump /dev/sda xfs_metadump_sda_2 (without -w):
> -rw-r--r--  1 root root 588261888 2009-03-11 19:23 xfs_metadump_sda_2
> 
>> It'd be nice if ubuntu had debug kernel variants (Fedora does this, I
>> dunno about ubuntu) - if you are hitting any kind of memory corruption
>> then a kernel with debug checks enabled might catch it sooner.
> 
> I haven't seen any - probably have to build my own debug kernel. Oh joy.
> 
> Is the metadump of any use?

it might be, let's see how well it shrinks.

-Eric

> Tom
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> xfs mailing list
> xfs at oss.sgi.com
> http://oss.sgi.com/mailman/listinfo/xfs