Re: [PATCH v2] mtd/nand : workaround for Freescale FCM to support large-page Nand chip

[Artem Bityutskiy <dedekind1@gmail.com>]

On Mon, 2011-08-15 at 11:11 -0500, Scott Wood wrote:
> On 08/15/2011 10:59 AM, Artem Bityutskiy wrote:
> > On Tue, 2011-07-12 at 12:48 +0800, b35362@freescale.com wrote:
> >> +	/*
> >> +	 * Hack for supporting the flash chip whose writesize is
> >> +	 * larger than 2K bytes.
> >> +	 */
> >> +	if (mtd->writesize > 2048) {
> >> +		elbc_fcm_ctrl->subpage_shift = ffs(mtd->writesize >> 11) - 1;
> >> +		elbc_fcm_ctrl->subpage_mask =
> >> +			(1 << elbc_fcm_ctrl->subpage_shift) - 1;
> >> +		/*
> >> +		 * Rewrite mtd->writesize, mtd->oobsize, chip->page_shift
> >> +		 * and chip->pagemask.
> >> +		 */
> >> +		mtd->writesize = 2048;
> >> +		mtd->oobsize = 64;
> >> +		chip->page_shift = ffs(mtd->writesize) - 1;
> >> +		chip->pagemask = (chip->chipsize >> chip->page_shift) - 1;
> >> +	}
> > 
> > So basically if the flash has 4KiB NAND pages, you are considering it as
> > a flash with 2KiB NAND pages. But surely this will work only if the
> > underlying flash has NOP 2 at least. Or even, if you consider that JFFS2
> > and YAFFS want to write to OOB, you need NOP 4 (2 ECC writes and 2
> > writes from YAFFS/JFFS2) ? So this won't work for NOP1 flashes?
> 
> Right.  The set of chips that work with this controller is still larger
> with this than without this.
> 
> It looks like NOP1 tends to be MLC -- you probably wouldn't want to use
> MLC with this controller anyway as it only does 1-bit ECC.
> 
> > Isn't it an ugly hack?
> 
> Less ugly than some other approaches that were considered. :-)
> 
> But yes, it's a hack (even says so in the comment).  The other option is
> "it doesn't work".

Could there be at least a fat comment that NANDs with 4KiB pages have to
be at least NOP4? And probably NANDs with 8KiB pages and larger should
simply be rejected?

-- 
Best Regards,
Artem Bityutskiy