[PATCH linux dev-4.7] drivers/fsi: Add more time between clocking and data sampling

[Brad Bishop]

Hi Chris

> On Mar 15, 2017, at 5:25 PM, Christopher Bostic <cbostic at linux.vnet.ibm.com> wrote:
> 
> Under high work loads the ARM core can occasionaly hit invalid
> timings between the FSI clock and data lines which leads to
> data crc errors.  Add further delay between clock switch and
> data line changes.
> 
> Signed-off-by: Christopher Bostic <cbostic at linux.vnet.ibm.com>
> ---
> drivers/fsi/fsi-master-gpio.c | 8 ++++----
> 1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/fsi/fsi-master-gpio.c b/drivers/fsi/fsi-master-gpio.c
> index a8976d5..a93453f 100644
> --- a/drivers/fsi/fsi-master-gpio.c
> +++ b/drivers/fsi/fsi-master-gpio.c
> @@ -13,7 +13,7 @@
> 
> #include "fsi-master.h"
> 
> -#define	FSI_GPIO_STD_DLY	1	/* Standard pin delay in nS */
> +#define	FSI_GPIO_STD_DLY	1	/* Standard pin delay in uS */

How did you arrive at 1000x ?  Does this mean we spend 1000x spinning
the core when sending data out?

I wonder if 100x or even 10x might still get rid of enough of the jitter?

> #define	FSI_ECHO_DELAY_CLOCKS	16	/* Number clocks for echo delay */
> #define	FSI_PRE_BREAK_CLOCKS	50	/* Number clocks to prep for break */
> #define	FSI_BREAK_CLOCKS	256	/* Number of clocks to issue break */
> @@ -86,9 +86,9 @@ static void clock_toggle(struct fsi_master_gpio *master, int count)
> 	int i;
> 
> 	for (i = 0; i < count; i++) {
> -		ndelay(FSI_GPIO_STD_DLY);
> +		udelay(FSI_GPIO_STD_DLY);
> 		gpiod_set_value(master->gpio_clk, 0);
> -		ndelay(FSI_GPIO_STD_DLY);
> +		udelay(FSI_GPIO_STD_DLY);
> 		gpiod_set_value(master->gpio_clk, 1);
> 	}
> }
> @@ -97,7 +97,7 @@ static int sda_in(struct fsi_master_gpio *master)
> {
> 	int in;
> 
> -	ndelay(FSI_GPIO_STD_DLY);
> +	udelay(FSI_GPIO_STD_DLY);
> 	in = gpiod_get_value(master->gpio_data);
> 	return in ? 1 : 0;
> }
> -- 
> 1.8.2.2

thx - brad