[PATCH v13 17/20] mtd: rawnand: Use the ECC framework OOB layouts

[Janusz Krzysztofik]

Hi Miquel,

On Tuesday, September 29, 2020 10:15:01 A.M. CEST Miquel Raynal wrote:
> Hi Janusz,
> 
> Janusz Krzysztofik <jmkrzyszt at gmail.com> wrote on Sun, 20 Sep 2020
> 14:21:44 +0200:
> 
> > On Thursday, August 27, 2020 10:52:05 A.M. CEST Miquel Raynal wrote:
> > > No need to have our own in the raw NAND core.  
> > 
> > Hi Miquel,
> > 
> > This patch introduced a regression for me, at least on ams-delta which 
sets 
> > nand_chip->ecc.engine_type to NAND_ECC_ENGINE_TYPE_SOFT and 
> > nand_chip.ecc->algo to NAND_ECC_ALGO_HAMMING.  No related kernel messages 
> > could be observed, but nanddump fails with "pread: Numerical result out of 
> > range" for me.
> 
> Thanks for the report and sorry for the delay!
> 
> > The ECC framework OOB layouts refer to nand_device->ecc.ctx.total which no 
> > driver nor ECC engine initializes so far, while the old layouts used to 
refer 
> > to nand_chip->ecc.total, still initialized by nand_scan_tail().  Is there 
any 
> > work in progress which will address the issue (initialization of 
> > nand_device->ecc.ctx.total at least)?
> 
> Indeed, there are more changes coming and the update of ecc.ctx.total
> was done in an upcoming patch. It makes sense to update this entry when
> switching to the "generic" layouts, so I moved the line in this patch.
> 
> I just updated nand/next, please tell me if it fixed your setup?

I won't be able to test it before next week, however looking at the updated 
patch I think the issue should be solved since a similar line did help when I 
added it at the end of ams-delta's .probe().

Thanks,
Janusz

> 
> Thanks,
> Miquèl
>