[LEDE-DEV] Need fix / advice / idea

[Sergey Ryazanov]

On Wed, Mar 8, 2017 at 3:19 AM, Denis Periša <darkman.hr at gmail.com> wrote:
>>> Just curious, this bad sectors appeared during regular device
>>> operation or just after sysupgrade?
>
>
> Sysupgrade. I had to get to roof and replace device as I could no
> longer get firmware on it. Btw. I wrote that it would be nice to first
> check nand than write if possible.

Before sysupgrade starts checking NAND it should know target flash
type, but image does not contain this info now. Image name contains
tiny hint: "64m" is for chips with 512-bytes pages and "large" is for
chips with 2048-bytes pages. Your flash type could be revealed by:

# dmesg | grep "page size"

> sysupgrade told me that firmware
> was flashed successfully but on reboot no kernel was found.
>

Yep. sysupgrade uses nandwrite utility, which could only report data
write result but could not predict kernel reaction to such "data".

>>> I am asking because it could be "false positive" bad sectors. Couple
>>> of weeks ago I flash one rb411 with wrong image. I used image for
>>> flash chips with 512-bytes pages, while the board was equipped with
>>> flash chip with 2048-bytes pages. And after reboot kernel printed a
>>> lot of "bad sector" messages. Actually sectors was fine, but using of
>>> the wrong firmware image lead to corruption of OOB data of several
>>> pages.
>>>
>>> I solved this issue by flashing back the vendor's firmware with help
>>> of netinstall, this action had recover flash OOB. Then I flash the
>>> device again, but now with the correct LEDE image and it works! Looks
>>> like that the similar results could be reached with _nandwrite_
>>> utility, but I did not try it.
>
> You might be on to something there, I had same device and it was brand
> new out of box.. Flashed firmware once, flashed twice, lot more bad
> sectors, next time whole nand was bad. That time I experimented with
> yaffs1 and yaffs2, had lot of issues there. Returned it as RMA same
> day, got another device. I actually tried to find how I could from
> linux re-detect or re-check how to put it, those bad sectors but
> didn't find any info. This yours is very helpful info, I'll try that
> and return here, but question stays, maybe in future some NANDs will
> begin to fail. This could be helpful in future.
>

Even 2Mb kernel partition contains enough spare space to overcome rise
of few bad sectors. To get more bad sectors you should reflash your
board a couple of time per day for several years. So IMHO there are
nothing to worry about.

-- 
Sergey