Re: [PATCH v3 48/52] mtd: nand: vf610: switch to mtd_ooblayout_ops

[Boris Brezillon]

Hi Stephan,

On Fri, 26 Feb 2016 11:43:46 -0800
Stefan Agner <stefan@xxxxxxxx> wrote:

> Hi Boris,
> 
> On 2016-02-25 16:57, Boris Brezillon wrote:
> > Implementing the mtd_ooblayout_ops interface is the new way of exposing
> > ECC/OOB layout to MTD users.
> > 
> > Signed-off-by: Boris Brezillon <boris.brezillon@xxxxxxxxxxxxxxxxxx>
> > ---
> >  drivers/mtd/nand/vf610_nfc.c | 34 ++++------------------------------
> >  1 file changed, 4 insertions(+), 30 deletions(-)
> > 
> > diff --git a/drivers/mtd/nand/vf610_nfc.c b/drivers/mtd/nand/vf610_nfc.c
> > index 293feb1..da34de1 100644
> > --- a/drivers/mtd/nand/vf610_nfc.c
> > +++ b/drivers/mtd/nand/vf610_nfc.c
> > @@ -175,34 +175,6 @@ static inline struct vf610_nfc *mtd_to_nfc(struct
> > mtd_info *mtd)
> >  	return container_of(mtd_to_nand(mtd), struct vf610_nfc, chip);
> >  }
> >  
> > -static struct nand_ecclayout vf610_nfc_ecc45 = {
> > -	.eccbytes = 45,
> > -	.eccpos = {19, 20, 21, 22, 23,
> > -		   24, 25, 26, 27, 28, 29, 30, 31,
> > -		   32, 33, 34, 35, 36, 37, 38, 39,
> > -		   40, 41, 42, 43, 44, 45, 46, 47,
> > -		   48, 49, 50, 51, 52, 53, 54, 55,
> > -		   56, 57, 58, 59, 60, 61, 62, 63},
> > -	.oobfree = {
> > -		{.offset = 2,
> > -		 .length = 17} }
> > -};
> > -
> > -static struct nand_ecclayout vf610_nfc_ecc60 = {
> > -	.eccbytes = 60,
> > -	.eccpos = { 4,  5,  6,  7,  8,  9, 10, 11,
> > -		   12, 13, 14, 15, 16, 17, 18, 19,
> > -		   20, 21, 22, 23, 24, 25, 26, 27,
> > -		   28, 29, 30, 31, 32, 33, 34, 35,
> > -		   36, 37, 38, 39, 40, 41, 42, 43,
> > -		   44, 45, 46, 47, 48, 49, 50, 51,
> > -		   52, 53, 54, 55, 56, 57, 58, 59,
> > -		   60, 61, 62, 63 },
> > -	.oobfree = {
> > -		{.offset = 2,
> > -		 .length = 2} }
> > -};
> > -
> >  static inline u32 vf610_nfc_read(struct vf610_nfc *nfc, uint reg)
> >  {
> >  	return readl(nfc->regs + reg);
> > @@ -781,14 +753,16 @@ static int vf610_nfc_probe(struct platform_device *pdev)
> >  		if (mtd->oobsize > 64)
> >  			mtd->oobsize = 64;
> >  
> > +		/*
> > +		 * mtd->ecclayout is not specified here because we're using the
> > +		 * default large page ECC layout defined in NAND core.
> > +		 */
> >  		if (chip->ecc.strength == 32) {
> >  			nfc->ecc_mode = ECC_60_BYTE;
> >  			chip->ecc.bytes = 60;
> 
> 
> When I add a printk in nand_ooblayout_free_lp, I get this values:
> free offset 2 lenth 62
> 
> This seems wrong to me. It seems that nand_ooblayout_free_lp needs steps
> to be set to calculate the free bytes correctly. With chip->ecc.steps =
> 1 the calculated values look good to me.

Actually, ecc->steps is calculated in nand_scan_tail(), but it happens
a bit too late (after mtd_ooblayout_count_freebytes() is called). I
reordered things in nand_scan_tail() to address that.

> 
> I quickly tried to run mtd_oobtest.ko test, however that faild:
> Kernel BUG at 80399764 [verbose debug info unavailable]
> Internal error: Oops - BUG: 0 [#1] ARM
> Modules linked in: mtd_oobtest(+)
> CPU: 0 PID: 676 Comm: insmod Tainted: G        W      
> 4.5.0-rc1-00097-g4c60bde-dirty #320
> Hardware name: Freescale Vybrid VF5xx/VF6xx (Device Tree)
> task: 8db27180 ti: 8ce74000 task.ti: 8ce74000
> PC is at nand_fill_oob+0x48/0xac
> LR is at mtd_ooblayout_free+0x5c/0x74
> 
> mtd_ooblayout_set_databytes seems to return -ERANGE...
> 
> Do you know what is going on here?

A bug in the mtd_ooblayout_set/get_bytes() implementation. Just fixed
it and pushed the changes on this branch [1].
The bugfix patches are here [2], [3].

Sorry for the inconvenience (I made some changes I thought were
trivial and didn't retest the whole thing, which is bad :-/), and thanks
for testing.

Boris

[1]https://github.com/bbrezillon/linux-0day/commits/nand/ecclayout
[2]https://github.com/bbrezillon/linux-0day/commit/070393f6269fef4ba1929b5ab74b19466d3b5862
[3]https://github.com/bbrezillon/linux-0day/commit/51fcc44a6c98cd2dbbd22279ca4cd393528ea42a

-- 
Boris Brezillon, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com