[PATCH v1 2/2] spi: stm32: split large transfers based on word size instead of bytes

[Alain Volmat]

Hi Leonard,

thanks for your patch.  I agree with this patch and tested it ok
as well on my side.

On Fri, Mar 10, 2023 at 10:20:53AM +0100, Leonard Göhrs wrote:
> The TSIZE register in CR2, to which the number of words to transfer
> is written, is only 16 Bit. This limits transfers to 65535 SPI
> _words_ at a time. The existing code uses spi_split_transfers_maxsize
> to limit transfers to 65535 _bytes_ at a time.
> 
> This breaks large transfers with bits_per_word > 8, as they are
> split inside of a word boundary by the odd size limit.
> 
> Split transfers based on the number of words instead.
> This has the added benefit of not artificially limiting the maximum
> length of bpw > 8 transfers to half or a quarter of the actual limit.
> 
> The combination of very large transfers and bits_per_word = 16 is triggered
> e.g. by MIPI DBI displays when updating large parts of the screen.
> 
> Signed-off-by: Leonard Göhrs <l.goehrs at pengutronix.de>
> ---
>  drivers/spi/spi-stm32.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/spi/spi-stm32.c b/drivers/spi/spi-stm32.c
> index def09cf0dc14..d2e16f16fae6 100644
> --- a/drivers/spi/spi-stm32.c
> +++ b/drivers/spi/spi-stm32.c
> @@ -984,9 +984,9 @@ static int stm32_spi_prepare_msg(struct spi_master *master,
>  	if (spi->cfg->set_number_of_data) {
>  		int ret;
>  
> -		ret = spi_split_transfers_maxsize(master, msg,
> -						  STM32H7_SPI_TSIZE_MAX,
> -						  GFP_KERNEL | GFP_DMA);
> +		ret = spi_split_transfers_maxwords(master, msg,
> +						   STM32H7_SPI_TSIZE_MAX,
> +						   GFP_KERNEL | GFP_DMA);
>  		if (ret)
>  			return ret;
>  	}

Acked-by: Alain Volmat <alain.volmat at foss.st.com>
> -- 
> 2.30.2
>