Re: [PATCH] ARM: dts: imx6ull: fix ubi mount failed on MYS-6ULX-IOT board

[dillon min]

Hi Sascha,

Thanks for reviewing.

On Tue, Mar 9, 2021 at 8:18 PM Sascha Hauer <s.hauer@xxxxxxxxxxxxxx> wrote:
>
> On Tue, Mar 09, 2021 at 02:15:19PM +0800, dillon.minfei@xxxxxxxxx wrote:
> > From: dillon min <dillon.minfei@xxxxxxxxx>
> >
> > This patch intend to fix ubi filesystem mount failed on MYS-6ULX-IOT board,
> > from Micron MT29F2G08ABAEAWP's datasheets, we need to choose 4-bit ECC.
> >
> > Table 18: Error Management Details
> >
> > Description                                   Requirement
> >
> > Minimum number of valid blocks (NVB) per LUN  2008
> > Total available blocks per LUN                        2048
> > First spare area location                     x8: byte 2048 x16: word 1024
> > Bad-block mark                                        x8: 00h x16: 0000h
> > Minimum required ECC                          4-bit ECC per 528 bytes
> > Minimum ECC with internal ECC enabled         4-bit ECC per 516 bytes (user data) + 8
> >                                               bytes (parity data)
> > Minimum required ECC for block 0 if PROGRAM/
> > ERASE cycles are less than 1000                       1-bit ECC per 528 bytes
>
> 4-bit ECC is the minimum this chip requires. There's nothing wrong with
> choosing a better ECC like the GPMI driver does by default.
>
Yes, indeed, the mt29f2g08's minimum ecc is 4-bit, you can use 8-bits ecc.
but there is a dependency between new kernel gpmi-nand with the old
mfg-kernel's , which means
if the old nand ecc layout is 4-bits, you should use ecc 4-bit in the
new kernel (by fsl,use-minimum-ecc),
else use 8-bits.

For my case, the ubifs filesystem was created by ecc 4-bits, without
reflash filesystem or change
gpmi-nand.c , so the better way is to use dts configure.

use legacy_set_geometry() :
[    6.713115] GF length              : 13
[    6.713115] ECC Strength           : 8
[    6.713115] Page Size in Bytes     : 2110
[    6.713115] Metadata Size in Bytes : 10
[    6.713115] ECC Chunk Size in Bytes: 512
[    6.713115] ECC Chunk Count        : 4
[    6.713115] Payload Size in Bytes  : 2048
[    6.713115] Auxiliary Size in Bytes: 16
[    6.713115] Auxiliary Status Offset: 12
[    6.713115] Block Mark Byte Offset : 1999
[    6.713115] Block Mark Bit Offset  : 0

use set_geometry_by_ecc_info()
[    6.701262] GF length              : 13
[    6.701262] ECC Strength           : 4
[    6.701262] Page Size in Bytes     : 2084
[    6.701262] Metadata Size in Bytes : 10
[    6.701262] ECC Chunk Size in Bytes: 512
[    6.701262] ECC Chunk Count        : 4
[    6.701262] Payload Size in Bytes  : 2048
[    6.701262] Auxiliary Size in Bytes: 16
[    6.701262] Auxiliary Status Offset: 12
[    6.701262] Block Mark Byte Offset : 2018
[    6.701262] Block Mark Bit Offset  : 4

here are some background tips (fsl,use-minimum-ecc) for your refer:

The bug was submitted in regression test,
[1] https://linux-mtd.infradead.narkive.com/hfRFPn3P/gpmi-mtd-ecc-regression

First patch to fix it
[2] https://stable.vger.kernel.narkive.com/mupPGUPj/patch-v2-mtd-gpmi-fix-the-ecc-regression

Introduce devicetree to configure by user(more flexibility)
[3] https://linux-mtd.infradead.narkive.com/tqQZ75Iz/patch-v2-mtd-gpmi-add-a-new-dt-property-to-use-the-datasheet-s-minimum-required-ecc



> It looks like you are papering over some other problem.
>
> Sascha
>
> --
> Pengutronix e.K.                           |                             |
> Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
> 31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
> Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |