CRC on Read-Only partition

CRC on Read-Only partition

[Artem B. Bityuckiy]

But if I formulate the question as:
we have device with RO partition. It is Read-Only always.

Does NAND technology imply that this partition may become bad, and it is 
normal?

The blocks from this partition are not ever erased. They are once writte 
when the device was shipped, checked.

--
Best Regards,
Artem B. Bityuckiy,
St.-Petersburg, Russia.

Re: CRC on Read-Only partition

[David Woodhouse]

On Thu, 2005-01-13 at 12:25 +0000, Artem B. Bityuckiy wrote:
> The blocks from this partition are not ever erased. They are once writte 
> when the device was shipped, checked.

I don't care. You show me formal proof that there cannot ever be an
error, or we check the CRC anyway :)

-- 
dwmw2

Re: CRC on Read-Only partition

[Artem B. Bityuckiy]

> I don't care. You show me formal proof that there cannot ever be an
> error, or we check the CRC anyway :)
I'm sorry, I mistiteled the mail. No doubt, CRC checking is needed to 
detect problems. They may come from different sources, not only NAND 
technology.

But I formulated question as: "does NAND technology imply that RO 
partition may become bad, and it is normal?".

The question is about NAND technology.

Ok, thin about Joern's phrase:
> What happens, if crucial data in flash gets corrupted and is unusable?
> Say, /sbin/init.

Now imagine we have this data on RO partition. Other non-sensitive data on 
RW partition. So, does this mean that only non-sensitive data may be 
corrupted? Can we keep safe our RO data this way? I only mean the 
corruptions due to *NAND technology*

--
Best Regards,
Artem B. Bityuckiy,
St.-Petersburg, Russia.

Re: CRC on Read-Only partition

[Jared Hulbert]

My understanding is that NAND technology is prone to read error even
if you aren't programming or eraseing it.   Such errors can occur
randomly.  By that I mean the data reads fine and then its bad and
then it reads fine...

Re: CRC on Read-Only partition

[Artem B. Bityuckiy]

On Tue, 18 Jan 2005, Jared Hulbert wrote:

> My understanding is that NAND technology is prone to read error even
> if you aren't programming or eraseing it.   Such errors can occur
> randomly.  By that I mean the data reads fine and then its bad and
> then it reads fine...
> 
Hmm. Thanks. So we just need to re-read several times.

--
Best Regards,
Artem B. Bityuckiy,
St.-Petersburg, Russia.

Re: CRC on Read-Only partition

[Artem B. Bityuckiy]

On Wed, 19 Jan 2005, Artem B. Bityuckiy wrote:

> On Tue, 18 Jan 2005, Jared Hulbert wrote:
> 
> > My understanding is that NAND technology is prone to read error even
> > if you aren't programming or eraseing it.   Such errors can occur
> > randomly.  By that I mean the data reads fine and then its bad and
> > then it reads fine...
> > 
> Hmm. Thanks. So we just need to re-read several times.
> 
Ok guys, I looked through documentation and articles and have not 
found anything which may approve this.

The question is: if we've read page and encounter ECC error, is it 
possible that if we re-read it several times we will have no errors? This 
means does NAND technology assume random occasional bit flips which go 
away after we re-read page?

I do not mean solar rays, radiation, etc. I only mean NAND technology 
itself.

Would be good to have some URL.

P.S. :
Toshiba's guide stands bit flipping is possible, but they stay until block 
is erased. Re-read as many times as yiu want - you will still have this 
bit error. But this doesn't mean block is bad - erase the block and be 
happy.


> --
> Best Regards,
> Artem B. Bityuckiy,
> St.-Petersburg, Russia.
> 
> ______________________________________________________
> Linux MTD discussion mailing list
> http://lists.infradead.org/mailman/listinfo/linux-mtd/
> 

--
Best Regards,
Artem B. Bityuckiy,
St.-Petersburg, Russia.

Re: CRC on Read-Only partition

[Jared Hulbert]

> The question is: if we've read page and encounter ECC error, is it
> possible that if we re-read it several times we will have no errors? This
> means does NAND technology assume random occasional bit flips which go
> away after we re-read page?

Whether or not the error goes away was not the point.  My point is
that with NAND you can get errors even with RO data.  That's all.
 
> P.S. :
> Toshiba's guide stands bit flipping is possible, but they stay until block
> is erased. Re-read as many times as yiu want - you will still have this
> bit error. But this doesn't mean block is bad - erase the block and be
> happy.

I think the real question is whether the likelyhood of errors getting
past the NAND's ECC is > or < than the users tolerance level.  I don't
believe that the answer should be terribly different for RO or RW
data.  Yet this depends on temperature, frequency of read, and yes
even solar flares :)  Therefore the answer is different for a set top
box with an expected useful life of 2 years and a car.