Re: ubifs error at boot "bad node type"

[hujianyang <hujianyang@huawei.com>]

On 2014/7/16 16:29, Artem Bityutskiy wrote:
> On Tue, 2014-07-08 at 12:10 +0200, Richard Genoud wrote:
>> ubiattach is ok:
>> # ubiattach -m 3
>> [   50.164062] UBI: default fastmap pool size: 95
>> [   50.164062] UBI: default fastmap WL pool size: 25
> 
> I do not know why this happened, I never saw reports like this before.
> Most probably this is because of fastmap. I did not hear any report that
> it was extensively verified WRT power cuts, and my theory is that there
> is a bug in fastmap which causes this, and this may be related to power
> cuts. I do not have proves, but suggest you to dig in this direction.
> E.g., setup power-cut testing.
> 

My colleague who use UBIFS with kernel v2.6.34(WR4?) hit
an error like this:

UBIFS error (pid 0): ubifs_read_node: bad node type (255 but expected 0)
UBIFS error (pid 0): ubifs_read_node: bad node at LEB 112:67824, LEB mapping status 1
UBIFS error (pid 0): ubifs_iget: failed to read inode 1, error -22

He said he was running UBIFS for 3 years and would hit this
error about one time per year.

I googled these error messages and found some same reports
in 2010~2012. So I think this problem is fixed by early
fixes, I asked him to backport this patch set

[PATCH 00/13] UBI reliability improvements

but I'm sure of that. After this error report, I was
worrying about this bug still left.