unclean shutdown of usb hdd destroyed xfs partially

[Marko Weber|8000]

Am 2014-09-14 00:09, schrieb Dave Chinner:
> On Sat, Sep 13, 2014 at 04:36:59PM +0200, Marko Weber|8000 wrote:
>> 
>> an output of  xfs_repair -v -L /dev/sde1
> ...
>> 
>>  # xfs_repair -v -L /dev/sde1
> 
> What version?

first it was 3.1.10
later like posted 3.2.1

> 
>> Phase 1 - Superblock finden und überprüfen...
>>         - Berichts-Prozess in Abständen von 15 Minutes
>>         - Block-Zwischenspeichergröße ist auf 1487792 Einträge gesetzt
>> Phase 2 - ein internes Protokoll benutzen
>>         - Null-Protokoll...
>> zero_log: head block 40 tail block 40
>>         - freier Speicher und Inode-Karten des Dateisystems werden
>> gescannt...
>> bad magic numberbad magic numberbad magic number
> ....
>> falscher on-disk-Superblock 12 - falsche Magische Nummer
>> Metadata corruption detected at block
>> 0x20bf8e50/0x1000primäre/sekundärer Superblock-12-Konflikt -
>> AG-Superblock-Geometrie-Info hat einen Konflikt mit der
>> Dateisystem-Geometrie
>> flasche magische # 0x0 für agf 12
>> falsche Version # 0 für agf 12
>> 
>> Metadata corruption detected at block 0x417f1c90/0x1000
>> falscher on-disk-Superblock 6 - falsche Magische Nummer
>> primäre/sekundärer Superblock-6-Konflikt -
>> AG-Superblock-Geometrie-Info hat einen Konflikt mit der
>> Dateisystem-Geometrie
>> falscher on-disk-Superblock 3 - falsche Magische Nummer
>> ungenutzten Anteil des »sekundär«-Superblocks nullen (AG #6)
>> Metadata corruption detected at block 0x57542610/0x1000
>> falsche Sequenz # 0 für agf 12
>> Metadata corruption detected at block
>> 0x7813b450/0x1000Speicherzugriffsfehler
> 
> I don't read german(?) but that looks like many AG header block
> have been overwritten with zeros (0 magic number, 0 sequence #, 0
> length, etc) and so even if we can repair the filesystem, I'd
> suggest that you need to verify that the data in every file in the
> filesystem is correct.
> 
> Is that as far as xfs_repair got? If so, it would have appeared to
> crash.  Can you run the lastest version inside gdb to get a stack
> trace when it dies? Or, alternatively, provide a metadump for one of
> us to look at more closely?

yes, this is as far xfs_repair got it.
what is ment 'run latest version in gdb?' gnudebugger? how do i do that? 
console example needed.
and how do i do the metadump?

thank you

marko
> 
> Cheers,
> 
> Dave.