Re: [PATCH] mtd: brcmnand: set initial ECC params based on info from HW

[Brian Norris <computersforpeace@gmail.com>]

On Mon, Feb 01, 2016 at 12:33:06PM -0800, Florian Fainelli wrote:
> On 01/02/16 11:02, Brian Norris wrote:
> > On Mon, Feb 01, 2016 at 10:09:28AM -0800, Florian Fainelli wrote:
> >> The Marvell PXA3XX NAND controller has a "marvell,nand-keep-config"
> >> which seems to be in par with what Rafal is after here, maybe it would
> >> be worth standardizing/mimicing that kind of property for the brcmnand
> >> driver?
> > 
> > Maybe. Honestly, it's not that clear to me what the pxa3xx-nand binding
> > means. But if we make that clear, that could be a possibility.
> 
> I would argue that if you were down the road of looking at the
> bootloader defaults to figure out why things did not work, it's not that
> much of a stretch to put the correct information in Device Tree, leading
> to predictable results, so ...

I wasn't suggesting trying to reintroduce any complex
bootloader-matching logic. I think either we should stick with:

(a) all info (ECC step size + strength) goes in DT (as-is) or
(b) we provide a DT option to say "keep the ECC settings configured by
    the bootloader"

If we go with (a), I'd still propose my original suggestion, if that
helps Rafal's use case:

> >>> What if we just print out the hardware defaults when we bail out due to
> >>> missing ECC DT properties? Then developers can choose to set up their DT
> >>> with these properties, if those are actually proven correct. Would that
> >>> save you the hours of setup you mention?

Regards,
Brian