Re: flash-kernel failure during apt-get upgrade. Now what?

[Forest <forestcode@ixio.org>]

On Sun, 26 Feb 2017 10:45:03 +0100, Uwe Kleine-König wrote:

>>> Without having checked every byte, this looks okish. Is this NAND or NOR
>>> memory?
>> 
>> Looks to me like it's NOR:

>I'd something like this:
>
>dd if=/dev/zero of=zero bs=10240 count=1
>flashcp -v zero /dev/mtd1
>memtool md -s /dev/mtdblock1 0+0x2800 | grep -vE '(0{8} ){4}'

Only 10KiB?  Okay.  Unfortunately, flashcp once again said "File does not
seem to match flash data", and the grep output shows nonzero bytes at every
single offset.

>flash_erase /dev/mtd1 0 0
>memtool md -s /dev/mtdblock1 0+0x2800 | grep -vE '(f{8} ){4}'

Again, grep shows non-FF bytes at every single offset.

I just noticed something, though:  In both tests above, the hex dumps aren't
full of 0x00 or 0xFF, but they aren't full of garbage, either.  For example:

00000000: 56190527 9aac8fc2 4e5a9258 28a41f00  '..V....X.ZN...(
00000010: 00800000 00800000 db880208 00020205  ................
00000020: 6e72656b 33206c65 2e36312e 2d342d30  kernel 3.16.0-4-
00000030: 6b72696b 646f6f77 00000000 00000000  kirkwood........

That looks to me like the start of a kernel image.  Could that be the old
kernel image, completely intact?  Perhaps the problem here is not
corruption, but some kind of read-only mode?  What do you think?