UBIFS Recovery Issue: 'grab_empty_leb: could not find an empty LEB'

[Brent Taylor]

On Mon, Sep 17, 2012 at 8:34 AM, Brent Taylor <motobud at gmail.com> wrote:
> On Fri, Aug 31, 2012 at 8:23 AM, Brent Taylor <motobud at gmail.com> wrote:
>>
>> I'm currently running the test again, and will know the results on wed, sept 5.  If I do see a failure, I can send you an image, the problem is that the image will be either 256MB, 1GB, or 15GB depending on which partition fails.  Will either of the larger sizes be an issue for you?  What would be the best way to get the image to you?
>>
>> Thanks for your time.
>> Bud
>>
>> On Aug 31, 2012 7:39 AM, "Artem Bityutskiy" <dedekind1 at gmail.com> wrote:
>>>
>>> On Mon, 2012-07-16 at 13:22 -0500, Brent Taylor wrote:
>>> > I'm developing on custom hardware using an at91sam9g45 processor with
>>> > 64 MB of RAM and 64 MB of NOR flash running Linux 3.0.13.
>>> >
>>> > I'm doing some power-cycle testing and after several hours, UBIFS
>>> > failed to mount the root partition and printed the message
>>> > "grab_empty_leb: could not find an empty LEB".  The system boots and
>>> > starts a script that will power cycle the board anytime between 10 and
>>> > 20 seconds after the script has started.  Just before a power cycle
>>> > occurs, syslog and our main process are the only processes that could
>>> > be writing to disk.
>>> >
>>> I was trying to do power cut emulation testing with mtdram, and I see
>>> there are some issues which do not seem to exist on NAND. But they are
>>> different to yours. I remember I saw these -ENOSPC issues in the past,
>>> but they were very difficult to reproduce.
>>>
>>> I will try to investigate why power-cut emulation testing fails on
>>> mtdram, but I have limited amount of time and I do not know when I'll be
>>> able to do this.
>>>
>>> --
>>> Best Regards,
>>> Artem Bityutskiy
>
>
> Artem,
>    I was finally able to capture the issue again with linux version
> 3.5.1, and I now have a 64MB mtd image for you to look at.  The issue
> occurred in the root filesystem this time.  Let me know what's the
> best way to get this image to you.  The image was retrieved from a NOR
> flash chip instead of a NAND flash chip, but I've seen the error on
> both which leads me to believe it's not a flash driver issue and it's
> something in the UBIFS subsystem.  If there is more data you need from
> me to help track this down, just let me know.
>
> Thanks for your time,
> Bud

Ping,  I just didn't want to get lost in the e-mail shuffle.