In current implementation dummy buswidth is set equal to address buswidth. In case of quad spi (mode 1-1-4), where address width is 1 the dummy bytes will be calculated to 1(8 dummy cycles) and dummy buswidth is set to 1. Due to this, the controller driver will introduce 8 dummy cycles on data line(D0) during read operation.

But since we are using 4 data lines in case of qspi, we need to change this dummy bus width to 4. This will make dummy bytes to 4 inplace of 1. This will be taken care in controller driver by dividing with dummy buswidth again as in below code, which makes dummy cycles to 8 as earlier.

dummy_cycles = op->dummy.nbytes * 8 / op->dummy.buswidth;

https://source.denx.de/u-boot/u-boot/-/blob/master/drivers/spi/zynqmp_gqspi....

So with this change dummy cycles will be on all data lines(D0-D3) and it is taken care for all the configurations(single, dual, quad and octal).

SPI experts, please advice if this change is good. What is the reason we are using dummy bus width as address bus width so far ?

I have tested this change on all xilinx platforms with single, quad and octal configurations. It works perfectly fine.

I would appreciate if anyone can test on your board and give feedback.

Signed-off-by: Ashok Reddy Soma ashok.reddy.soma@xilinx.com ---

drivers/mtd/spi/spi-nor-core.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/mtd/spi/spi-nor-core.c b/drivers/mtd/spi/spi-nor-core.c index a70fbda4bb..6849da9113 100644 --- a/drivers/mtd/spi/spi-nor-core.c +++ b/drivers/mtd/spi/spi-nor-core.c @@ -251,7 +251,7 @@ static void spi_nor_setup_op(const struct spi_nor *nor, op->addr.buswidth = spi_nor_get_protocol_addr_nbits(proto);

if (op->dummy.nbytes) - op->dummy.buswidth = spi_nor_get_protocol_addr_nbits(proto); + op->dummy.buswidth = spi_nor_get_protocol_data_nbits(proto);

if (op->data.nbytes) op->data.buswidth = spi_nor_get_protocol_data_nbits(proto);