Re: [PATCH v2 3/3] mtd: nand: sunxi: add randomizer support

From: Boris Brezillon
Date: Sat Jan 23 2016 - 03:18:55 EST


Hi Brian,

On Fri, 22 Jan 2016 18:57:13 -0800
Brian Norris <computersforpeace@xxxxxxxxx> wrote:

> All three look good, so pushed to l2-mtd.git/next. One comment below:

Thanks.

>
> On Wed, Dec 02, 2015 at 12:01:07PM +0100, Boris Brezillon wrote:
>
> ...
>
> > +static u16 sunxi_nfc_randomizer_state(struct mtd_info *mtd, int page, bool ecc)
> > +{
> > + const u16 *seeds = sunxi_nfc_randomizer_page_seeds;
> > + int mod = mtd->erasesize / mtd->writesize;
>
> Richard suggested you use the mtd.h helper here. Patch below.

Yep, I forgot to do this change.

>
> > +
> > + if (mod > ARRAY_SIZE(sunxi_nfc_randomizer_page_seeds))
> > + mod = ARRAY_SIZE(sunxi_nfc_randomizer_page_seeds);
> > +
> > + if (ecc) {
> > + if (mtd->ecc_step_size == 512)
> > + seeds = sunxi_nfc_randomizer_ecc512_seeds;
> > + else
> > + seeds = sunxi_nfc_randomizer_ecc1024_seeds;
> > + }
> > +
> > + return seeds[page % mod];
> > +}
>
> From: Brian Norris <computersforpeace@xxxxxxxxx>
> Date: Fri, 22 Jan 2016 18:54:02 -0800
> Subject: [PATCH] mtd: nand: sunxi: use mtd_div_by_ws() helper
>
> Suggested-by: Richard Weinberger <richard@xxxxxx>
> Signed-off-by: Brian Norris <computersforpeace@xxxxxxxxx>

Acked-by: Boris Brezillon <boris.brezillon@xxxxxxxxxxxxxxxxxx>

> ---
> drivers/mtd/nand/sunxi_nand.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/mtd/nand/sunxi_nand.c b/drivers/mtd/nand/sunxi_nand.c
> index 5f700719d5c2..b5ea6b312df0 100644
> --- a/drivers/mtd/nand/sunxi_nand.c
> +++ b/drivers/mtd/nand/sunxi_nand.c
> @@ -624,7 +624,7 @@ static u16 sunxi_nfc_randomizer_step(u16 state, int count)
> static u16 sunxi_nfc_randomizer_state(struct mtd_info *mtd, int page, bool ecc)
> {
> const u16 *seeds = sunxi_nfc_randomizer_page_seeds;
> - int mod = mtd->erasesize / mtd->writesize;
> + int mod = mtd_div_by_ws(mtd->erasesize, mtd);

Just a comment (which should not prevent you from applying this patch).
Isn't it a bit overkill to cast the erasesize to a 64 bit value, and
then do a do_div on it. Shouldn't happen often though, because
->writesize_shift should be != 0 in most (all?) cases.

Another related remark: with the MLC/paired pages stuff I'll have to
retrieve this information (number of write units per erase block) quite
often, so maybe we should have a field (and/or an helper) for that.

Thanks,

Boris


--
Boris Brezillon, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com