Re: imx27: Unsupported ECC algorithm

From: Han Xu <xhnjupt@gmail.com>
To: Miquel Raynal <miquel.raynal@bootlin.com>
Cc: linux-mtd <linux-mtd@lists.infradead.org>,
	Sascha Hauer <s.hauer@pengutronix.de>,
	Fabio Estevam <festevam@gmail.com>,
	Sascha Hauer <kernel@pengutronix.de>,
	Lucas Stach <l.stach@pengutronix.de>
Subject: Re: imx27: Unsupported ECC algorithm
Date: Thu, 15 Oct 2020 22:04:14 -0500	[thread overview]
Message-ID: <CA+EcR200pOT8F=xS+Fh_gyJMF9ZGxFU6Y+RDRg6XyntyR2yGdA@mail.gmail.com> (raw)
In-Reply-To: <20201014152659.6ef6a2fc@xps13>

On Wed, Oct 14, 2020 at 8:32 AM Miquel Raynal <miquel.raynal@bootlin.com> wrote:
>
> Hi Fabio, Sascha, Lucas,
>
> Fabio Estevam <festevam@gmail.com> wrote on Wed, 14 Oct 2020 09:23:06
> -0300:
>
> > Hi Lucas,
> >
> > On Tue, Oct 13, 2020 at 2:14 PM Lucas Stach <l.stach@pengutronix.de> wrote:
> >
> > > Just a heads up: our lab was broken for some time due to a local issue.
> > > It's now back to working and the backlog of jobs is being executed on
> > > the boards right now, so results should show up in the near furture.
> >
> > Thanks. kernelci results are showing up again.
> >
> > Hi Miquel,
> >
> > The ECC issues are still present on imx27-phytec-phycard-s-rdk:
> > https://storage.kernelci.org/next/master/next-20201013/arm/imx_v4_v5_defconfig/gcc-8/lab-pengutronix/baseline-imx27-phytec-phycard-s-rdk.html
> >
> > BTW, ECC errors are also reported on a Freescale Layerscape 2088A RDB Board:
> > https://www.spinics.net/lists/iommu/msg49772.html
> >
>
> I supposed the failing division is this one:
>
>         /*
>          * Set the number of read / write steps for one page depending on ECC
>          * mode.
>          */
>         ecc->steps = mtd->writesize / ecc->size;
>         if (ecc->steps * ecc->size != mtd->writesize) {
>                 WARN(1, "Invalid ECC parameters\n");
>                 ret = -EINVAL;
>                 goto err_nand_manuf_cleanup;
>         }
>
> Which produces the warning right after. This would mean that
>
>         ecc->size == 0
>
> But this entry is set before calling nand_scan() in mxcnd_probe() to
> 512, and that does not make sense.
>
> I guess we should be able to reproduce this error with any board using
> the mxc NAND driver with hardware correction (now also called "on
> host"). Do you have such a board available to at least verify that the
> beginning of my logic is right?

Hi Miquel,

I debugged on one LS1043A platform and found in rawnand_dt_init(), the
chip->ecc.strength was set to nand->ecc.user_conf.strength, which may
not have been touched before. Same as chip->ecc.size.

>
> Thanks,
> Miquèl
>
> ______________________________________________________
> Linux MTD discussion mailing list
> http://lists.infradead.org/mailman/listinfo/linux-mtd/

-- 
Sincerely,

Han XU

______________________________________________________
Linux MTD discussion mailing list
http://lists.infradead.org/mailman/listinfo/linux-mtd/