spi: dw: Fix off-by-one for checking fifo depth

The fifo depth could be from 2 to 256 from HW spec, so fix off-by-one for checking fifo depth.
Without this patch, fifo is 258 after for loop iteration for the "no-match" case. So below line does not catch the "no-match" case.

        dws->fifo_len = (fifo == 257) ? 0 : fifo;

Signed-off-by Axel Lin <[email protected]>

---
Hi,
I don't have this hardware handy, so please test it.

BTW, I'm wondering if it make sense to set dws->fifo_len to 0. Isn't it suppose to be 2 ~ 256?

Thanks.

 drivers/spi/spi-dw.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

drivers/spi/spi-dw.c +1 -1

index d0d5542..74b27479 100644
@@ -621,7 +621,7 @@ static void spi_hw_init(struct dw_spi *dws)
 	if (!dws->fifo_len) {
 		u32 fifo;
 
-		for (fifo = 2; fifo <= 257; fifo++) {
+		for (fifo = 2; fifo <= 256; fifo++) {
 			dw_writew(dws, DW_SPI_TXFLTR, fifo);
 			if (fifo != dw_readw(dws, DW_SPI_TXFLTR))
 				break;

Quoted 5 lines

The fifo depth could be from 2 to 256 from HW spec, so fix off-by-one for checking fifo depth.
Without this patch, fifo is 258 after for loop iteration for the "no-match" case. So below line does not catch the "no-match" case.

        dws->fifo_len = (fifo == 257) ? 0 : fifo;

Seems reasonable, but never tried because in case of Medfield device we have fifo size defined.

I would try this in January.

2014-12-19 19:46 GMT+08:00 Andy Shevchenko <andriy.shevchenko-VuQAYsv1563Yd54FQh9/[email protected]>:

Quoted 10 lines · Axel Lin

Quoted 5 lines

The fifo depth could be from 2 to 256 from HW spec, so fix off-by-one for checking fifo depth.
Without this patch, fifo is 258 after for loop iteration for the "no-match" case. So below line does not catch the "no-match" case.

        dws->fifo_len = (fifo == 257) ? 0 : fifo;

Seems reasonable, but never tried because in case of Medfield device we have fifo size defined.

Quoted 1 line

I would try this in January.

hi Andy,

I check the code again and I think what current code does is:
It tries to find the highest valid fifo depth so it checks the value it wrote to DW_SPI_TXFLTR. I think the valid value range is 2 ~ 256.
So it will break out when writing 257 to DW_SPI_TXFLTR.

There is an off-by-one in dws->fifo_len setting because it assumes the latest register write fails so the latest valid value is fifo - 1.

So I think you can set dws->fifo_len to 0 to test current behavior first.

I guess below change should work, please test this instead my previous patch.

drivers/spi/spi-dw.c +1 -1

index d0d5542..1a0f266 100644
@@ -621,13 +621,13 @@ static void spi_hw_init(struct dw_spi *dws)
  if (!dws->fifo_len) {
  u32 fifo;
 
- for (fifo = 2; fifo <= 257; fifo++) {
+ for (fifo = 2; fifo <= 256; fifo++) {