[PATCH v2] mtd: nand: sunxi: rely on nand_dt_init initialization

[PATCH v2] mtd: nand: sunxi: rely on nand_dt_init initialization

[Boris Brezillon]

nand_dt_init(), called from nand_scan_ident(), is already parsing the
generic MTD/NAND DT properties, and initializing the nand_chip struct
accordingly.
Rely on this initialization instead of manually parsing those properties.

Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
---
Changes since v1:
- drop unused strength and blk_size variables
---
 drivers/mtd/nand/sunxi_nand.c | 20 +++++---------------
 1 file changed, 5 insertions(+), 15 deletions(-)

diff --git a/drivers/mtd/nand/sunxi_nand.c b/drivers/mtd/nand/sunxi_nand.c
index 96dda70..92a2c58 100644
--- a/drivers/mtd/nand/sunxi_nand.c
+++ b/drivers/mtd/nand/sunxi_nand.c
@@ -1164,16 +1164,9 @@ static int sunxi_nand_ecc_init(struct mtd_info *mtd, struct nand_ecc_ctrl *ecc,
 			       struct device_node *np)
 {
 	struct nand_chip *nand = mtd->priv;
-	int strength;
-	int blk_size;
 	int ret;
 
-	blk_size = of_get_nand_ecc_step_size(np);
-	strength = of_get_nand_ecc_strength(np);
-	if (blk_size > 0 && strength > 0) {
-		ecc->size = blk_size;
-		ecc->strength = strength;
-	} else {
+	if (!ecc->size) {
 		ecc->size = nand->ecc_step_ds;
 		ecc->strength = nand->ecc_strength_ds;
 	}
@@ -1183,10 +1176,6 @@ static int sunxi_nand_ecc_init(struct mtd_info *mtd, struct nand_ecc_ctrl *ecc,
 
 	ecc->mode = NAND_ECC_HW;
 
-	ret = of_get_nand_ecc_mode(np);
-	if (ret >= 0)
-		ecc->mode = ret;
-
 	switch (ecc->mode) {
 	case NAND_ECC_SOFT_BCH:
 		break;
@@ -1312,15 +1301,13 @@ static int sunxi_nand_chip_init(struct device *dev, struct sunxi_nfc *nfc,
 	/* Default tR value specified in the ONFI spec (chapter 4.15.1) */
 	nand->chip_delay = 200;
 	nand->controller = &nfc->controller;
+	nand->dn = np;
 	nand->select_chip = sunxi_nfc_select_chip;
 	nand->cmd_ctrl = sunxi_nfc_cmd_ctrl;
 	nand->read_buf = sunxi_nfc_read_buf;
 	nand->write_buf = sunxi_nfc_write_buf;
 	nand->read_byte = sunxi_nfc_read_byte;
 
-	if (of_get_nand_on_flash_bbt(np))
-		nand->bbt_options |= NAND_BBT_USE_FLASH | NAND_BBT_NO_OOB;
-
 	mtd = &chip->mtd;
 	mtd->dev.parent = dev;
 	mtd->priv = nand;
@@ -1330,6 +1317,9 @@ static int sunxi_nand_chip_init(struct device *dev, struct sunxi_nfc *nfc,
 	if (ret)
 		return ret;
 
+	if (nand->bbt_options & NAND_BBT_USE_FLASH)
+		nand->bbt_options |= NAND_BBT_NO_OOB;
+
 	ret = sunxi_nand_chip_init_timings(chip, np);
 	if (ret) {
 		dev_err(dev, "could not configure chip timings: %d\n", ret);
-- 
1.9.1

[PATCH v2] mtd: nand: sunxi: rely on nand_dt_init initialization

[Brian Norris]

On Mon, Aug 24, 2015 at 12:10:31PM +0200, Boris Brezillon wrote:
> nand_dt_init(), called from nand_scan_ident(), is already parsing the
> generic MTD/NAND DT properties, and initializing the nand_chip struct
> accordingly.
> Rely on this initialization instead of manually parsing those properties.
> 
> Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
> ---
> Changes since v1:
> - drop unused strength and blk_size variables

Good catch, I was about to comment on v1 :)

> ---
>  drivers/mtd/nand/sunxi_nand.c | 20 +++++---------------
>  1 file changed, 5 insertions(+), 15 deletions(-)
> 
> diff --git a/drivers/mtd/nand/sunxi_nand.c b/drivers/mtd/nand/sunxi_nand.c
> index 96dda70..92a2c58 100644
> --- a/drivers/mtd/nand/sunxi_nand.c
> +++ b/drivers/mtd/nand/sunxi_nand.c
> @@ -1164,16 +1164,9 @@ static int sunxi_nand_ecc_init(struct mtd_info *mtd, struct nand_ecc_ctrl *ecc,
>  			       struct device_node *np)
>  {
>  	struct nand_chip *nand = mtd->priv;
> -	int strength;
> -	int blk_size;
>  	int ret;
>  
> -	blk_size = of_get_nand_ecc_step_size(np);
> -	strength = of_get_nand_ecc_strength(np);
> -	if (blk_size > 0 && strength > 0) {
> -		ecc->size = blk_size;
> -		ecc->strength = strength;
> -	} else {
> +	if (!ecc->size) {
>  		ecc->size = nand->ecc_step_ds;
>  		ecc->strength = nand->ecc_strength_ds;
>  	}
> @@ -1183,10 +1176,6 @@ static int sunxi_nand_ecc_init(struct mtd_info *mtd, struct nand_ecc_ctrl *ecc,
>  
>  	ecc->mode = NAND_ECC_HW;
>  
> -	ret = of_get_nand_ecc_mode(np);
> -	if (ret >= 0)
> -		ecc->mode = ret;
> -

I think you've missed something here. You're now *only* allowing HW ECC!
You now want to make this a default case.

There's also a subtle change here that affects more than just here: you
don't have a good way to tell the difference between "no ECC mode
provide" and NAND_ECC_NONE. Perhaps we can modify the nand_ecc_modes_t
enum to include a "uninitialized" value? I'm not sure if that'd work
best as 0, -1, or something else, in case there are people making
assumptions elsewhere...

>  	switch (ecc->mode) {
>  	case NAND_ECC_SOFT_BCH:
>  		break;
> @@ -1312,15 +1301,13 @@ static int sunxi_nand_chip_init(struct device *dev, struct sunxi_nfc *nfc,
>  	/* Default tR value specified in the ONFI spec (chapter 4.15.1) */
>  	nand->chip_delay = 200;
>  	nand->controller = &nfc->controller;
> +	nand->dn = np;
>  	nand->select_chip = sunxi_nfc_select_chip;
>  	nand->cmd_ctrl = sunxi_nfc_cmd_ctrl;
>  	nand->read_buf = sunxi_nfc_read_buf;
>  	nand->write_buf = sunxi_nfc_write_buf;
>  	nand->read_byte = sunxi_nfc_read_byte;
>  
> -	if (of_get_nand_on_flash_bbt(np))
> -		nand->bbt_options |= NAND_BBT_USE_FLASH | NAND_BBT_NO_OOB;
> -
>  	mtd = &chip->mtd;
>  	mtd->dev.parent = dev;
>  	mtd->priv = nand;
> @@ -1330,6 +1317,9 @@ static int sunxi_nand_chip_init(struct device *dev, struct sunxi_nfc *nfc,
>  	if (ret)
>  		return ret;
>  
> +	if (nand->bbt_options & NAND_BBT_USE_FLASH)
> +		nand->bbt_options |= NAND_BBT_NO_OOB;
> +
>  	ret = sunxi_nand_chip_init_timings(chip, np);
>  	if (ret) {
>  		dev_err(dev, "could not configure chip timings: %d\n", ret);

Brian

[PATCH v2] mtd: nand: sunxi: rely on nand_dt_init initialization

[Boris Brezillon]

Hi Brian,

On Tue, 1 Sep 2015 18:16:16 -0700
Brian Norris <computersforpeace@gmail.com> wrote:

> On Mon, Aug 24, 2015 at 12:10:31PM +0200, Boris Brezillon wrote:
> > nand_dt_init(), called from nand_scan_ident(), is already parsing the
> > generic MTD/NAND DT properties, and initializing the nand_chip struct
> > accordingly.
> > Rely on this initialization instead of manually parsing those properties.
> > 
> > Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
> > ---
> > Changes since v1:
> > - drop unused strength and blk_size variables
> 
> Good catch, I was about to comment on v1 :)
> 
> > ---
> >  drivers/mtd/nand/sunxi_nand.c | 20 +++++---------------
> >  1 file changed, 5 insertions(+), 15 deletions(-)
> > 
> > diff --git a/drivers/mtd/nand/sunxi_nand.c b/drivers/mtd/nand/sunxi_nand.c
> > index 96dda70..92a2c58 100644
> > --- a/drivers/mtd/nand/sunxi_nand.c
> > +++ b/drivers/mtd/nand/sunxi_nand.c
> > @@ -1164,16 +1164,9 @@ static int sunxi_nand_ecc_init(struct mtd_info *mtd, struct nand_ecc_ctrl *ecc,
> >  			       struct device_node *np)
> >  {
> >  	struct nand_chip *nand = mtd->priv;
> > -	int strength;
> > -	int blk_size;
> >  	int ret;
> >  
> > -	blk_size = of_get_nand_ecc_step_size(np);
> > -	strength = of_get_nand_ecc_strength(np);
> > -	if (blk_size > 0 && strength > 0) {
> > -		ecc->size = blk_size;
> > -		ecc->strength = strength;
> > -	} else {
> > +	if (!ecc->size) {
> >  		ecc->size = nand->ecc_step_ds;
> >  		ecc->strength = nand->ecc_strength_ds;
> >  	}
> > @@ -1183,10 +1176,6 @@ static int sunxi_nand_ecc_init(struct mtd_info *mtd, struct nand_ecc_ctrl *ecc,
> >  
> >  	ecc->mode = NAND_ECC_HW;
> >  
> > -	ret = of_get_nand_ecc_mode(np);
> > -	if (ret >= 0)
> > -		ecc->mode = ret;
> > -
> 
> I think you've missed something here. You're now *only* allowing HW ECC!
> You now want to make this a default case.

Indeed, I missed that point.

> 
> There's also a subtle change here that affects more than just here: you
> don't have a good way to tell the difference between "no ECC mode
> provide" and NAND_ECC_NONE. Perhaps we can modify the nand_ecc_modes_t
> enum to include a "uninitialized" value? I'm not sure if that'd work
> best as 0, -1, or something else, in case there are people making
> assumptions elsewhere...

It really depends on what we want to do. I see at least two solutions
for this problem:
1/ keep the nand_dt_init() implementation as is and ask
   nand_scan_ident() callers to pre-set their default value in ecc.mode.
   This should work since ecc.mode is only overridden if ecc_mode is >= 0.
2/ modify the nand_dt_init function to assign the ecc_mode even if it
   is negative, or create a new value in the enum (NAND_ECC_UNSPECIFIED
   = -1) and assign it to ecc.mode when ecc_mode is negative.

1# does not require any change, but if we choose this solution, we
should make it clear that NAND controller drivers should set their
default value before calling nand_scan_ident().
And if go for 2#, that's not like we'll have to patch a lot of
drivers (AFAICT, the only user in 4.2 is the brcmnand driver) ;-).

Best Regards,

Boris


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

[PATCH v2] mtd: nand: sunxi: rely on nand_dt_init initialization

[Brian Norris]

On Wed, Sep 02, 2015 at 10:21:08AM +0200, Boris Brezillon wrote:
> On Tue, 1 Sep 2015 18:16:16 -0700
> Brian Norris <computersforpeace@gmail.com> wrote:
> > There's also a subtle change here that affects more than just here: you
> > don't have a good way to tell the difference between "no ECC mode
> > provide" and NAND_ECC_NONE. Perhaps we can modify the nand_ecc_modes_t
> > enum to include a "uninitialized" value? I'm not sure if that'd work
> > best as 0, -1, or something else, in case there are people making
> > assumptions elsewhere...
> 
> It really depends on what we want to do. I see at least two solutions
> for this problem:
> 1/ keep the nand_dt_init() implementation as is and ask
>    nand_scan_ident() callers to pre-set their default value in ecc.mode.
>    This should work since ecc.mode is only overridden if ecc_mode is >= 0.
> 2/ modify the nand_dt_init function to assign the ecc_mode even if it
>    is negative, or create a new value in the enum (NAND_ECC_UNSPECIFIED
>    = -1) and assign it to ecc.mode when ecc_mode is negative.
> 
> 1# does not require any change, but if we choose this solution, we
> should make it clear that NAND controller drivers should set their
> default value before calling nand_scan_ident().

I'm OK with your choice of #1. Your v3 patch looks fine.

> And if go for 2#, that's not like we'll have to patch a lot of
> drivers (AFAICT, the only user in 4.2 is the brcmnand driver) ;-).

:) Well, I thought there were others in the pipeline. I've at least
reviewed another addition somewhat recently, but it wasn't quite ready.

Brian