Re: Fwd: EXT3 File System Corruption 2.6.34

[Eric Sandeen <sandeen@sandeen.net>]

Jeffrey Merkey wrote:

> OK.  I will set this up.  You may want to make this option the default
> in the build scripts.  here is a corrupted file.  

It was default, but Linus changed it a while back.

> This was a .gif
> image file I saved THEN AFTER SAVING THE FILE I pulled the power to
> the machine and during recovery the file was FUCKED.  

I assume your application did not sync the data, and buffered data
loss is expected on a power loss.

> At any rate,
> this does not happen with 2.6.28.

that I can't explain for sure.... different timing perhaps.

> I dumped the file with xdump a util I use internally for my own use so
> you could see the file contents as text and I could post it here.
> This was an image file but look what ended up in it -- directory
> blocks and such.  Take a look:

As I said, stale blocks exposed due to data=writeback.  Known behavior,
unfortunately the default for ext3.  If you find similar problems
when mounted data=ordered, it's a more interesting report.

-Eric

>           0  1  2  3  4  5  6  7  8  9  A  B  C  D  E  F
> 00000000  6C 73 0A 63 64 20 2E 2E 0A 63 6C 73 0A 6C 73 0A  ls.cd ...cls.ls.
> 00000010  63 64 20 6C 69 6E 75 78 2D 32 2E 36 2E 33 34 2D  cd linux-2.6.34-
> 00000020  6D 64 62 2F 0A 63 6C 73 0A 6C 73 0A 63 64 20 2E  mdb/.cls.ls.cd .
> 00000030  2E 0A 63 6C 73 0A 6C 73 0A 63 64 20 6C 69 6E 75  ..cls.ls.cd linu

<giant snip>