[PATCH v13 09/20] mtd: rawnand: Separate the ECC engine type and the ECC byte placement

[Miquel Raynal]

Hi Han,

Han Xu <xhnjupt at gmail.com> wrote on Fri, 16 Oct 2020 12:51:24 -0500:

> On Fri, Oct 16, 2020 at 12:14 PM Miquel Raynal
> <miquel.raynal at bootlin.com> wrote:
> >
> > Hi Han,
> >
> > Han Xu <xhnjupt at gmail.com> wrote on Fri, 16 Oct 2020 12:09:22 -0500:
> >  
> > > On Thu, Oct 15, 2020 at 2:25 AM Miquel Raynal <miquel.raynal at bootlin.com> wrote:  
> > > >
> > > > Hi Han,
> > > >
> > > > Han Xu <xhnjupt at gmail.com> wrote on Wed, 14 Oct 2020 22:09:56 -0500:
> > > >  
> > > > > On Wed, Oct 14, 2020 at 8:35 PM Fabio Estevam <festevam at gmail.com> wrote:  
> > > > > >
> > > > > > Hi Han Xu,
> > > > > >
> > > > > > On Wed, Oct 14, 2020 at 6:38 PM Han Xu <xhnjupt at gmail.com> wrote:
> > > > > >  
> > > > > > > Hi Miquel, the changes to the nested switch statement make the logic
> > > > > > > of this break wrong.  
> > > > > >
> > > > > > Care to send a fix, please?
> > > > > >
> > > > > > We have seen ECC NAND regressions on mx27 and LS2088 platforms.
> > > > > >
> > > > > > In which platform did you noticed the ECC issue?
> > > > > >
> > > > > > Thanks  
> > > > >
> > > > > Hi Fabio,
> > > > >
> > > > > CI test reported all LS platforms have the ECC regression. The error
> > > > > is "Driver must set ecc.strength when using hardware ECC"
> > > > >
> > > > > I haven't worked on any LS platforms so I tested on i.MX platforms
> > > > > with gpmi NAND controller,
> > > > > the code ran into NAND_ECC_ENGINE_TYPE_SOFT due to this issue. So
> > > > > these are two different issues, I can send a patch to fix the nested
> > > > > switch statement problem but won't fix the LS platform issue.  
> > > >
> > > >
> > > > I think the fact that the execution ran into NAND_ECC_ENGINE_TYPE_SOFT
> > > > has been fixed. However there is still the other issue that I am unable
> > > > to find just by reading the code :/  
> > >
> > > Could you please send me the link of the fix patch, I can test it on
> > > i.MX platforms.  
> >
> > The fix is in next since more than a week, if you still have the issue
> > then it's probably the same issue that sascha reported in another
> > thread and in this case debugging from your side is welcome.  
> 
> If I am correct, the patch "mtd: rawnand: Fix wrongful fallthrough
> NAND_ECC_SOFT" is the fix. But I didn't see it in the next branch,
> only the other two patches in the patch set applied.

Indeed, I squashed that patch with the faulty commit to try keeping a
clean/understandable history.


Thanks,
Miquèl