RE: [PATCH] Documentation: binding: Update endianness usage

-----Original Message-----
From: Scott Wood [mailto:oss@buserror.net]
Sent: Tuesday, December 05, 2017 8:16 AM
To: Prabhakar Kushwaha <redacted>; linux-
mtd@lists.infradead.org; devicetree@vger.kernel.org
Cc: dedekind1@gmail.com; computersforpeace@gmail.com
Subject: Re: [PATCH] Documentation: binding: Update endianness usage

On Mon, 2017-12-04 at 04:33 +0000, Prabhakar Kushwaha wrote:

quoted

quoted

-----Original Message-----
From: Scott Wood [mailto:oss@buserror.net]
Sent: Saturday, December 02, 2017 3:25 AM
To: Prabhakar Kushwaha <redacted>; linux-
mtd@lists.infradead.org; devicetree-discuss@lists.ozlabs.org
Cc: dedekind1@gmail.com; computersforpeace@gmail.com
Subject: Re: [PATCH] Documentation: binding: Update endianness usage

On Fri, 2017-12-01 at 08:42 +0000, Prabhakar Kushwaha wrote:

quoted

quoted

-----Original Message-----
From: Scott Wood [mailto:oss@buserror.net]
Sent: Friday, December 01, 2017 10:43 AM
To: Prabhakar Kushwaha <redacted>; linux-
mtd@lists.infradead.org; devicetree-discuss@lists.ozlabs.org
Cc: dedekind1@gmail.com; computersforpeace@gmail.com
Subject: Re: [PATCH] Documentation: binding: Update endianness usage

If big endian is the default, is this change really
necessary?  Particularly
since the big endian chips are older and thus have existing device
trees.

Earlier endianness information was only used for "how to"  access IFC-
NAND
register access.
Now this info  will also be used for defining swap requirement of NOR
flash.

Is this a difference between LS1021A and PPC-based chips?

Yes.
CONFIG_MTD_CFI_BE_BYTE_SWAP needs to be defined For LS1021A, LS1043A,
LS1046A

Only because you're running a little-endian kernel on those chips.  I still
don't see why the absence of a little-endian property isn't sufficient to
communicate that the hardware is big-endian given that that's the established
default.

I now see your patch to of_flash_probe... where is the non-IFC-specific
binding that says the *parent* of a CFI node should be looked at for this?
Where in general are endian properties kept in the parent of the node with
"reg"?  The right answer is to add endianness to mtd-physmap.txt.

Flashes are always littler endian. 
It is because of IFC controller behavior, endianness is required.  
So as per my understanding, this info should go in IFC binding. 

Please help me if I am not able to understand your view.

--prabhakar


N�����r��y���b�X��ǧv�^�)޺{.n�+���z��z��z)���w*jg���
�����ݢj.�۰\��M��gj��a����' ��ޢ�
���j:+v���w�j�m��������zZ+�����ݢj"��!�i