[U-Boot] U-boot nand bug, read.part should fail

[Scott Wood]

On 02/07/2013 04:13:55 PM, Harvey Chapman wrote:
> [ I started this conversation off-list before I joined the list. ]
> 
> The idea is to add .part as a valid command suffix to nand read/write  
> so it would match nand erase.part. The code to implement it makes  
> "nand read.part" act identically to "nand read".
> 
> On Feb 7, 2013, at 4:59 PM, Scott Wood <scottwood at freescale.com>  
> wrote:
> 
> >> >> In fact, I think erase should be modified to deprecate  
> erase.part and make erase work like read does now.
> >> >
> >> > Erase used to work like read does.  I deliberately changed it so  
> that typos (e.g. "nand erase $partition $fliesize") don't end up  
> erasing your entire partition or chip.
> >> Ah, then maybe we should add .part to nand read for consistency?  
> I'm ok either way.
> >
> > That would get messy because it would be orthogonal to other  
> suffixes.  Reading too much is not as harmful as
> 
> Nothing would change other than do_nand() would treat "nand read" and  
> "nand read.part" identically.

The only reason to add .part/.chip is if the unsuffixed versions no  
longer operate on entire partitions/chips.

> > erasing too much.  Writing too much can be bad, though.  Perhaps we  
> should just eliminate the ability to do reads/writes without explicit  
> size (it already has problems with the size needing adjustment due to  
> bad blocks).
> 
> I liked that I didn't have to specify the size.

It's fine until you get a bad block in the partition, and you end up  
accessing the first block of the next partition (or getting "Attempt to  
read/write outside the flash area" if it's the last partition).

Of course, fixing partition/chip accesses to account for this when  
determining size would be even better. :-)

-Scott