[U-Boot] [PATCHv2 5/8] drivers: mtd: spi: Modify read/write command for sfl256s flash.

[Sourav Poddar <sourav.poddar@ti.com>]

+ jagan,

On Tuesday 23 July 2013 02:29 PM, Sourav Poddar wrote:
> Reading using the already supported read command is causing regression
> after 4k bytes, as a result doing a page by page read. Its happening, because
> ti qpsi controller CS will get de asserted after 4096 bytes.
>
> Signed-off-by: Sourav Poddar<sourav.poddar@ti.com>
> ---
>   drivers/mtd/spi/spi_flash.c |   32 +++++++++++++++++++++++++++++++-
>   1 files changed, 31 insertions(+), 1 deletions(-)
>
> diff --git a/drivers/mtd/spi/spi_flash.c b/drivers/mtd/spi/spi_flash.c
> index 6a6fe37..5f8db7b 100644
> --- a/drivers/mtd/spi/spi_flash.c
> +++ b/drivers/mtd/spi/spi_flash.c
> @@ -303,6 +303,36 @@ int spi_flash_cmd_read_fast(struct spi_flash *flash, u32 offset,
>   		else
>   			read_len = remain_len;
>
> +#ifdef CONFIG_TI_QSPI
> +		unsigned long page_addr, byte_addr, page_size;
> +		size_t chunk_len, actual;
> +		int ret = 0;
> +
> +		page_size = flash->page_size;
> +		page_addr = offset / page_size;
> +		byte_addr = offset % page_size;
> +
> +		for (actual = 0; actual<  read_len; actual += chunk_len) {
> +			chunk_len = min(read_len - actual, page_size - byte_addr);
> +
> +			cmd[1] = page_addr>>  8;
> +			cmd[2] = page_addr;
> +			cmd[3] = byte_addr;
> +
> +			ret = spi_flash_read_common(flash, cmd, sizeof(cmd),
> +					data + actual, chunk_len);
> +			if (ret<  0) {
> +				debug("SF: read failed");
> +				break;
> +			}
> +
> +			byte_addr += chunk_len;
> +			if (byte_addr == page_size) {
> +				page_addr++;
> +				byte_addr = 0;
> +			}
> +		}
> +#else
>   		spi_flash_addr(offset, cmd);
>
>   		ret = spi_flash_read_common(flash, cmd, sizeof(cmd),
> @@ -311,7 +341,7 @@ int spi_flash_cmd_read_fast(struct spi_flash *flash, u32 offset,
>   			debug("SF: read failed\n");
>   			break;
>   		}
> -
> +#endif
>   		offset += read_len;
>   		len -= read_len;
>   		data += read_len;
Elaborating a bit more on this,
There is a constrain on our hardware, which goes like this..

As soon as the words transfered is 4096 bytes, the CS gets deasserted 
automatically.
As a result of this bottleneck, I am not able to use the current use 
read api in mtd framework.
This requires me to send the read command every time in range upto 4096 
bytes only.

To overcome this, I have updated the mtd read based on TI_QSPI config as 
done above.

[Jagan]:
Do you have any suggestion of dealing this in a better way?
I don't see a way to get around this apart from updating mtd read 
depending on TI_QSPI config.