On 04/28/2013 10:28:14 PM, Kuo-Jung Su wrote:
2013/4/27 Scott Wood <scottw...@freescale.com>:
> Shouldn't .eccpos depend on HWECC?
>
>
Actually it means nothing here, the ECC function is designed to be
some kind of
a blackbox to users (i.e softwares), the actual position of ECC codes
are not documented
in the datasheet, I have to dig it out from RTL which I do not have
permission to access.
Maybe get rid of .eccpos altogether, and supply your own (much simpler)
hwecc read/write_page functions -- similar to what fsl_elbc_nand, which
also has transparent (but not hidden) ECC, does? This would also let
you get rid of hwctl/calculate/correct.
> Is it really all in one step, regardless of page size?
>
Theoretical, NO.
It uses R-S or BCH for Forwarding Error Correction, which is selected
at RTL level.
Which means it actually 522 bytes per step at RS and 528 bytes per
step at BCH.
However it's all blackbox to software, the FTNANDC021 simply supports
only
page read, so looks like everything is done at 1 step.
It would matter if you could detect correctable errors, but since you
can't, it doesn't.
-Scott
_______________________________________________
U-Boot mailing list
U-Boot@lists.denx.de
http://lists.denx.de/mailman/listinfo/u-boot
