Re: [PATCH 2/4][v4] driver: mtd: update struct map_info's swap as per map requirement

[Boris Brezillon <boris.brezillon@bootlin.com>]

On Thu, 15 Feb 2018 17:20:00 +0530
Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com> wrote:

> It is not necessary for all device's maps to be CFI_HOST_ENDIAN.
> Maps device can be Big endian or little endian.
> 
> Currently it is being taken care using CONFIG_MTD_CFI_LE_BYTE_SWAP or
> CONFIG_MTD_CFI_BE_BYTE_SWAP i.e. compile time.
> 
> Now update struct map_info's swap field based on device characteristics
> defined in device tree.
> 
> Signed-off-by: Jagdish Gediya <jagdish.gediya@nxp.com>
> Signed-off-by: Prabhakar Kushwaha <prabhakar.kushwaha@nxp.com>
> ---
> Changes for v2: Incorporated review comments from Boris Brezillon
> http://patchwork.ozlabs.org/patch/758381/
> Changes for v3: None
> Changes for v4: Updated code as per new defined binding of physmap.txt
> 
>  drivers/mtd/maps/physmap_of_core.c | 8 ++++++++
>  1 file changed, 8 insertions(+)
> 
> diff --git a/drivers/mtd/maps/physmap_of_core.c b/drivers/mtd/maps/physmap_of_core.c
> index b1bd4faecfb2..6eee160dfb3d 100644
> --- a/drivers/mtd/maps/physmap_of_core.c
> +++ b/drivers/mtd/maps/physmap_of_core.c
> @@ -20,6 +20,7 @@
>  #include <linux/mtd/map.h>
>  #include <linux/mtd/partitions.h>
>  #include <linux/mtd/concat.h>
> +#include <linux/mtd/cfi_endian.h>
>  #include <linux/of.h>
>  #include <linux/of_address.h>
>  #include <linux/of_platform.h>
> @@ -233,6 +234,13 @@ static int of_flash_probe(struct platform_device *dev)
>  		info->list[i].map.bankwidth = be32_to_cpup(width);
>  		info->list[i].map.device_node = dp;
>  
> +		if (of_property_read_bool(dp, "big-endian"))
> +			info->list[i].map.swap = CFI_BIG_ENDIAN;
> +		else if (of_property_read_bool(dp, "little-endian"))
> +			info->list[i].map.swap = CFI_LITTLE_ENDIAN;
> +		else
> +			info->list[i].map.swap = CFI_HOST_ENDIAN;
> +

Are you sure we want the default to be CFI_HOST_ENDIAN when the
property is not defined? I think we should leave ->swap to 0 and let
the CFI logic use the CFI_DEFAULT_ENDIAN value. What you're doing here
you're likely to break platforms that are not explicitly specifying
endianness in their DT and are instead relying on the Kconfig choice.

I think that was one of Scott's comments on your previous version.

>  		err = of_flash_probe_gemini(dev, dp, &info->list[i].map);
>  		if (err)
>  			goto err_out;



-- 
Boris Brezillon, Bootlin (formerly Free Electrons)
Embedded Linux and Kernel engineering
https://bootlin.com