[Yaffs-archive] Re: yaffs: serious error in yaffs guts

[Charles Manning]

Hi Christian

It is easiest for me if someone using the Linux MTD with NANDECC can test 
this. Otherewise I will have to write a software emulator.

Could you roll this together with Thomas' recent patchs for NANDECC, with 
your proposed ammendments, and I can apply them as a consistent patch set. 

Thanx

-- Charles


On Fri, 28 Feb 2003 06:27, Christian Gan wrote:
> Also,
>
> In yaffs_guts there is no real need to check eccres1 and eccres2 you should
> only need to check the return value of nandmtd_ReadChunkFromNand.  The two
> extra ints from the NAND MTD are only useful when you are performing
> multiple page reads when calling nand_read_ecc, since they will tell you
> which of the pages you requested has failed.  But YAFFs only a reads single
> page at a time, looking at the return value of nand_read_ecc is enough to
> know if the write was successful or if an ecc error has occurred for that
> page.  This value is checked in nandmtd_ReadChunkFromNand and it passes on
> a YAFFS_OK or YAFFS_FAIL (for any ecc errors) appropriately.
>
> Christian
>
> > -----Original Message-----
> > From: Nick Bane [mailto:nick@cecomputing.co.uk]
> > Sent: Thursday, February 27, 2003 10:31 AM
> > To: yaffs list
> > Subject: yaffs: serious error in yaffs guts
> >
> >
> > Charles
> >
> > A couple of bugs. One serious, one minor.
> >
> > in yaffs_ReadChunkFromNAND when dev->useNANDECC is true, it is
> > assumed that
> > readChunkFromNAND fills in the two ecc result integers on the end
> > of the oob
> > data.
> > These are then tested to see if there has been and ecc read
> > fail/correction
> > by the mtd layer.
> >
> > The problem with this is that readChunkFromNAND never actually passes the
> > ecc correction data back but leaves it in a local variable
> > tmpSpare and only
> > copies YAFFS_BYTES_PER_SPARE from it.
> >
> > The serious consequence of this is that all chunk reads are then
> > flagged as
> > failures and their blocks are marked as needing retiring once the data is
> > erased so yaffs nand becomes effectively write-once. The really serious
> > consequence of this is that one cannot now distinguish between
> > yaffs_retired_blocks and nand_factory_bad_blocks so one has to
> > speculatively
> > erase the entire device losing the factory erase data (which seems to be
> > zero on devices I have seen). Shouldn't we use a value other than 0 for
> > yaffs to mark the block as bad (with more than two zero bits) so
> > they can be
> > distinguished?
> >
> > The other bug is in testing the ecc correction results in
> > yaffs_ReadChunkFromNAND.
> > if(nspare.eccres2 || nspare.eccres2)
> > Clearly one of them should be eccres1
> >
> > Nick
> >
> > -------------------------------------------
> > Nick Bane
> > Cambridge, UK.
> > +44(0)1954 719270
> >
> >
> >
> >
> > ------------------------------------------------------------------
> > ---------------------
> > This mailing list is hosted by Toby Churchill open software
>
> (www.toby-churchill.org).
> If mailing list membership is no longer wanted you can remove yourself from
> the list by
> sending an email to yaffs-request@toby-churchill.org with the text
> "unsubscribe"
> (without the quotes) as the subject.
>
>
> ---------------------------------------------------------------------------
>------------ This mailing list is hosted by Toby Churchill open software
> (www.toby-churchill.org). If mailing list membership is no longer wanted
> you can remove yourself from the list by sending an email to
> yaffs-request@toby-churchill.org with the text "unsubscribe" (without the
> quotes) as the subject.


---------------------------------------------------------------------------------------
This mailing list is hosted by Toby Churchill open software (www.toby-churchill.org).
If mailing list membership is no longer wanted you can remove yourself from the list by 
sending an email to yaffs-request@toby-churchill.org with the text "unsubscribe" 
(without the quotes) as the subject.