Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210213095724.3411058-1-daniel@0x0f.com> <20210215112409.1a755bf0@xps13>
 <CAFr9PXkh+attaCc6C2UxB=qvXksWriWOaaoEndy4k6SGE0QOHQ@mail.gmail.com>
 <20210215121653.4edd86c4@xps13> <CAFr9PXmVehtcm6FjBqi_hmEAj1rgtxMvarisjPmWhgjruVj++Q@mail.gmail.com>
 <20210322193213.18520b9a@xps13> <CAFr9PX=mL9UWcr-yWbYa5NXS-R=yHeGgM+hd_MKOa2XiBUAJOQ@mail.gmail.com>
 <20210323113233.3523d66b@xps13> <CAFr9PX=KMZuzp61Hq=2WdHyEzE=6J7HEPWZxPs7FEqiH-G8wFw@mail.gmail.com>
 <20210323150603.6b942a60@xps13>
In-Reply-To: <20210323150603.6b942a60@xps13>
From:   Daniel Palmer <daniel@0x0f.com>
Date:   Fri, 26 Mar 2021 23:09:28 +0900
Message-ID: <CAFr9PXntCmdrmg+i3BB1j-aY4VbE=2iqyLcoSr9cX2090jGiGQ@mail.gmail.com>
Subject: Re: [PATCH v2] mtd: spinand: add support for Foresee FS35ND01G-S1Y2
To:     Miquel Raynal <miquel.raynal@bootlin.com>
Cc:     linux-mtd@lists.infradead.org,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

Hi Miquel,

Sorry for the constant pestering on this..

On Tue, 23 Mar 2021 at 23:06, Miquel Raynal <miquel.raynal@bootlin.com> wrote:
> > # nandbiterrs -i /dev/mtd1
> > incremental biterrors test
> > Successfully corrected 0 bit errors per subpage
> > Inserted biterror @ 0/5
> > Read reported 4 corrected bit errors
> > ECC failure, invalid data despite read success
>
> This is not a valid behavior. There is something wrong with the way ECC
> status is read/retrieved. The read should indeed report 4 corrected bit
> errors, but then the data should be valid. Here it means that the
> introduced error appears corrected but in fact is not.
> We need to understand what status are available and write the
> appropriate vendor code.

I looked at the datasheet again and the encoding for ecc status seems
a bit funky.
The chip seems to have only two bits for ECC status with this encoding:
0 - Read success with 0-3 flipped bits.
1 - Read success with 4 flipped bits.
2 - Uncorrectable.
3 - Reserved.

This looks almost the same as the Winbond chips with 0 and 1 being successes
but with no totally error free status.

Anyhow, if 4 bits were corrected is returned for 1 then nandbiterrs
reports "ECC failure, invalid data despite read success".
If I return 3 bit errors for 0 and -EBADMSG for anything else
nandbiterrs doesn't complain anymore but is this right?:

# nandbiterrs -i /dev/mtd1
incremental biterrors test
Read reported 3 corrected bit errors
Successfully corrected 0 bit errors per subpage
biterror @ 0/5
Failed to recover 1 bitflips
Read error after 1 bit errors per page

When running nandbiterrs the status is always 0 before injecting the
error and then sometimes 1 and sometimes 2 after the error is
injected.

Thanks,

Daniel