[PATCH BUGFIX] mtd: nand: mxc: fix obiwan error in mxc_nand_v[12]_ooblayout_free() functions

[Boris Brezillon]

On Fri, 16 Sep 2016 10:09:25 +0200
Lothar Waßmann <LW at KARO-electronics.de> wrote:

> Hi,
> 
> On Thu, 15 Sep 2016 18:06:05 +0200 Boris Brezillon wrote:
> > Hi Lothar,
> > 
> > On Fri,  9 Sep 2016 16:44:11 +0200
> > Lothar Waßmann <LW at KARO-electronics.de> wrote:
> >   
> > > commit a894cf6c5a82 ("mtd: nand: mxc: switch to mtd_ooblayout_ops")
> > > introduced a regression accessing the OOB area from the mxc_nand
> > > driver due to an Obiwan error in the mxc_nand_v[12]_ooblayout_free()
> > > functions. They report a bogus oobregion { 64, 7 } which leads to
> > > errors accessing bogus data when reading the oob area.
> > > 
> > > Prior to the commit the mtd-oobtest module could be run without any
> > > errors. With the offending commit, this test fails with results like:
> > > |Running mtd-oobtest
> > > |
> > > |=================================================
> > > |mtd_oobtest: MTD device: 5
> > > |mtd_oobtest: MTD device size 524288, eraseblock size 131072, page size 2048, count of eraseblocks 4, pages per eraseblock 64, OOB size 64
> > > |mtd_test: scanning for bad eraseblocks
> > > |mtd_test: scanned 4 eraseblocks, 0 are bad
> > > |mtd_oobtest: test 1 of 5
> > > |mtd_oobtest: writing OOBs of whole device
> > > |mtd_oobtest: written up to eraseblock 0
> > > |mtd_oobtest: written 4 eraseblocks
> > > |mtd_oobtest: verifying all eraseblocks
> > > |mtd_oobtest: error @addr[0x0:0x19] 0x9a -> 0x78 diff 0xe2
> > > |mtd_oobtest: error @addr[0x0:0x1a] 0xcc -> 0x0 diff 0xcc
> > > |mtd_oobtest: error @addr[0x0:0x1b] 0xe0 -> 0x85 diff 0x65
> > > |mtd_oobtest: error @addr[0x0:0x1c] 0x60 -> 0x62 diff 0x2
> > > |mtd_oobtest: error @addr[0x0:0x1d] 0x69 -> 0x45 diff 0x2c
> > > |mtd_oobtest: error @addr[0x0:0x1e] 0xcd -> 0xa0 diff 0x6d
> > > |mtd_oobtest: error @addr[0x0:0x1f] 0xf2 -> 0x60 diff 0x92
> > > |mtd_oobtest: error: verify failed at 0x0
> > > [...]
> > > 
> > > Signed-off-by: Lothar Waßmann <LW at KARO-electronics.de>
> > > ---
> > >  drivers/mtd/nand/mxc_nand.c | 4 ++--
> > >  1 file changed, 2 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/drivers/mtd/nand/mxc_nand.c b/drivers/mtd/nand/mxc_nand.c
> > > index 5173fad..fdee907 100644
> > > --- a/drivers/mtd/nand/mxc_nand.c
> > > +++ b/drivers/mtd/nand/mxc_nand.c
> > > @@ -893,7 +893,7 @@ static int mxc_v1_ooblayout_free(struct mtd_info *mtd, int section,
> > >  {
> > >  	struct nand_chip *nand_chip = mtd_to_nand(mtd);
> > >  
> > > -	if (section > nand_chip->ecc.steps)
> > > +	if (section >= nand_chip->ecc.steps)
> > >  		return -ERANGE;  
> > 
> > Hm, looking at the commit you're pointing to, it seems that this test
> > is correct (we have X + 1 free sections, where X is the number of ECC
> > steps).
> >   
> You are right. I didn't verify the v1 case (for which I have no HW here
> any more).
> I'll send a corrected patch.

No need to resend, I fixed it locally.

Thanks,

Boris