Re: [PATCH] mtd/nand: don't use {read,write}_buf for 8-bit transfers

[Huang Shijie <b32955@freescale.com>]

于 2013年02月28日 18:48, Uwe Kleine-König 写道:
> On Thu, Feb 28, 2013 at 10:47:43AM +0800, Huang Shijie wrote:
>> 于 2013年02月27日 23:10, Uwe Kleine-König 写道:
>>> According to the Open NAND Flash Interface Specification (ONFI) Revision
>>> 3.1 "Parameters are always transferred on the lower 8-bits of the data
>>> bus." for the Get Features and Set Features commands.
>>>
>> yes. the set/get features should works in 8-bit.
>>
>> I have never met a 16-bit onfi nand yet. :)
>>
>>> So using read_buf and write_buf is wrong for 16-bit wide nand chips as
>>> they use I/O[15:0]. The Get Features command is easily fixed using 4
>>> times the read_byte callback. For Set Features error out as there is no
>> yes. for get features, it's easy to fix it.
>>> write_byte callback.
>> Most of the time, the nand controller will overwrite the write_buf hook...
>> I also think we need a write_byte callback.
> a default implementation could be something like that:
>
> 	static void nand_write_byte(struct mtd_info *mtd, uint8_t byte)
> 	{
> 		struct nand_chip *chip = mtd->priv;
>
> 		if (chip->options&  NAND_BUSWIDTH_16)
> 			chip->write_buf(mtd, (uint8_t[]){ byte, 0 }, 2);
> 		else
> 			chip->write_buf(mtd,&byte, 1);
> 	}
>
> (Is this the correct order in the array? Or might that depend on
> endianess?)
>
> Does this look right?
>
IMHO, the nand_write_byte() should not call the chip->write_buf() again. 
Since the ->write_buf() could
be the nand_write_buf16(). it makes a little mess.

In my opinion, the default nand_write_byte() hook could use the 
nand_write_buf() to  write just one byte;
and the nand controller can overwrite the nand_write_byte() hook if it 
could.
Of course, it's just a suggest.

thanks
Huang Shijie



> Alternatively something could be done with chip->cmd_ctrl (but it seems
> not all drivers implement this, e.g. mxc_nand doesn't).
>
> Best regards
> Uwe
>