Michele De Candia (VT) wrote:
Scott Wood wrote:
Wolfgang Denk wrote:
Dear "Michele De Candia (VT)",
In message 4A37F7BF.2090101@valueteam.com you wrote:
this patch fixes a bug for 'nand erase' command: when bad blocks are present into erasing area, they were skipped but the erased size was updated anyway.
And what exactly is the bug in this behaviour?
I think that 'erase' should have the same behaviour of 'write' and 'read' commands: skip bad blocks until read/write size is reached. If you write a script that erases and then writes a NAND area and bad blocks are not skipped while erasing (as U-Boot actually does), the following 'write' is successfully done but ECC checks fail on next read on the same area.
I see - thanks for the explanation.
Hm... actually I think the write should fail in such a case...
Scott, what do you think?
I think the current behavior is reasonable. You're erasing a specific region of flash, not an amount needed to hold a certain amount of data.
From this point of view you're in the right; maybe this would be explained in 'README.nand' documentation or what do you think about add ing an option to 'nand erase' command to consider 'size' field as the effective blocks size to be erased and not as the area size?
Moreover, I think that if you want to erase a specific NAND area, the correct way to use 'nand erase' command would be:
'nand erase start end'
If you want to erase an area but you want to be sure that 'size' bytes were erased, you should use:
'nand erase off size'
In this case a bad block should be skipped and next block will be erased.
Obviously this behavior must be implemented.
What do you think about it?
While I can see the appeal of Michele's suggestion, I think it would be more error-prone as people trying to erase a region rather than just the size of data could erase too much.
It definitely should not be an error to erase a region that happens to contain a bad block. Bad blocks are expected and we need to work around them.
-Scott