UBI assert failed in ubi_wl_init

Timo Ketola timo at exertus.fi
Thu Mar 22 06:57:57 PDT 2018


On 21.03.2018 00:02, Richard Weinberger wrote:
> Timo,
> 
> On Fri, Mar 16, 2018 at 2:34 PM, Timo Ketola <timo at exertus.fi> wrote:
>> On 18.02.2018 22:31, Richard Weinberger wrote:
>>> Okay. That's a bit odd. Maybe my analysis is wrong. Can you try the
>>> following, replay the image to your NAND and attach again.
>>> Then you should get the same UBI assert a second time.
>>
>> Of course I can but what would be the difference to simple rebooting? I
>> get the assertion failure on every boot from the second one onwards.
>> Before I commit to that one, let's look what I found recently.
>>
>> I already tried the replay but bumped into ECC errors. Due to that and
>> other things I switched the kernel to Freescale IMX 4.9.74 with Boundary
>> fixes. I reduced the size of the user partition to 1800M and got rid of
>> the 'could not get any free erase block' and 'Unable to write new
>> fastmap' errors. The assertion failures from the second boot onwards
>> remained.
>>
>> I wondered why I'm not seeing the same issue with my smaller 64M system
>> mtd. Then I reduced the user partition to 48M and the mtd to 64M and the
>> issue vanished. Then I kept the partition at 48M but restored the mtd
>> size to 1800M and the issue returned. If I make the mtd size 0x4e680000
>> or smaller there is no issue. If I make it 0x4e6a0000 or larger the
>> issue is there. Every time I erased the NAND and burned a fresh image.
>>
>> What do you want me to do next?
> 
> Did you verify the image or not? :)
> I asked you do to so because the image you sent me does not make sense
> and I'm not sure whether I can trust it.
> Hence I asked for a double-check.

Ok, sorry.

Here is a dump after the second boot:

https://drive.google.com/open?id=1oa2lV_OB_tC-SX_c4jylnMXK6x1xhj2o

After I dumped it I erased the whole mtd, wrote the dump back there,
dumped another dump and verified that it was identical with the first
one. Then I rebooted and observed that the issue was still there just as
before.

--

Timo



More information about the linux-mtd mailing list