ubifs error : read_znode: bad indexing node

[Artem Bityutskiy]

Hi,

On Mon, 2012-06-11 at 15:59 +0200, Matthieu CASTET wrote:
> we have products where rw ubifs partition are corrupted.
> They all failed with the "read_znode: bad indexing node at LEB 558:22056, error
> 6" or "read_znode: bad indexing node at LEB 446:62888, error 2".

So we read the indexing node, and its CRC is correct, but then we
validate it by checking that various fields are withing possible limits,
and we see that there is some garbage.

This smells more like memory corruption somewhere - we write garbage or
corrupted buffers to the media.

I guess it would be a good improvement to UBIFS if we validated when we
write, not only read, then we could catch issues earlier.

> The last ubifs stable commit we have is
> http://git.infradead.org/users/dedekind/ubifs-v2.6.32.git/commit/7eb3b6c0999bd2cbc37adb2bd0fb8127f98240ea

Does not look like something which could cause those errors. Is that
problem reproducible? Only on these devices or on other too?


> Do you know if the bug is known ?
> Google doesn't provide any info on this and looking at ubifs new commits don't
> provide explicit bug fix.

No, this is the first time I see such a problem.

> Are there any document that explain how znode works ?

Well, Adrian's PDF file may give some insight. znode is just an in-RAM
representation of an indexing node on the flash.

-- 
Best Regards,
Artem Bityutskiy
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part
URL: <http://lists.infradead.org/pipermail/linux-mtd/attachments/20120618/6ab9e229/attachment.sig>