Re: [PATCH net] net: dsa: microchip: fix KSZ9477 set_ageing_time function

[Paolo Abeni <pabeni@redhat.com>]

On Tue, 2024-05-28 at 14:36 -0700, Tristram.Ha@microchip.com wrote:
> From: Tristram Ha <tristram.ha@microchip.com>
> 
> The aging count is not a simple 11-bit value but comprises a 3-bit
> multiplier and an 8-bit second count.  The code tries to find a set of
> values with result close to the specifying value.
> 
> Note LAN937X has similar operation but provides an option to use
> millisecond instead of second so there will be a separate fix in the
> future.
> 
> Fixes: 2c119d9982b1 ("net: dsa: microchip: add the support for set_ageing_time")
> Signed-off-by: Tristram Ha <tristram.ha@microchip.com>
> ---
>  drivers/net/dsa/microchip/ksz9477.c     | 64 +++++++++++++++++++++----
>  drivers/net/dsa/microchip/ksz9477_reg.h |  1 -
>  2 files changed, 54 insertions(+), 11 deletions(-)
> 
> diff --git a/drivers/net/dsa/microchip/ksz9477.c b/drivers/net/dsa/microchip/ksz9477.c
> index f8ad7833f5d9..1af11aee3119 100644
> --- a/drivers/net/dsa/microchip/ksz9477.c
> +++ b/drivers/net/dsa/microchip/ksz9477.c
> @@ -1099,26 +1099,70 @@ void ksz9477_get_caps(struct ksz_device *dev, int port,
>  int ksz9477_set_ageing_time(struct ksz_device *dev, unsigned int msecs)
>  {
>  	u32 secs = msecs / 1000;
> +	u8 first, last, mult, i;
> +	int min, ret;
> +	int diff[8];
>  	u8 value;
>  	u8 data;
> -	int ret;
>  
> -	value = FIELD_GET(SW_AGE_PERIOD_7_0_M, secs);
> +	/* The aging timer comprises a 3-bit multiplier and an 8-bit second
> +	 * value.  Either of them cannot be zero.  The maximum timer is then
> +	 * 7 * 255 = 1785.
> +	 */
> +	if (!secs)
> +		secs = 1;
>  
> -	ret = ksz_write8(dev, REG_SW_LUE_CTRL_3, value);
> +	ret = ksz_read8(dev, REG_SW_LUE_CTRL_0, &value);
>  	if (ret < 0)
>  		return ret;
>  
> -	data = FIELD_GET(SW_AGE_PERIOD_10_8_M, secs);
> +	/* Check whether there is need to update the multiplier. */
> +	mult = FIELD_GET(SW_AGE_CNT_M, value);
> +	if (mult > 0) {
> +		/* Try to use the same multiplier already in the register. */
> +		min = secs / mult;
> +		if (min <= 0xff && min * mult == secs)
> +			return ksz_write8(dev, REG_SW_LUE_CTRL_3, min);
> +	}
>  
> -	ret = ksz_read8(dev, REG_SW_LUE_CTRL_0, &value);
> -	if (ret < 0)
> -		return ret;
> +	/* Return error if too large. */
> +	if (secs > 7 * 0xff)
> +		return -EINVAL;
> +
> +	/* Find out which combination of multiplier * value results in a timer
> +	 * value close to the specified timer value.
> +	 */
> +	first = (secs + 0xfe) / 0xff;
> +	for (i = first; i <= 7; i++) {
> +		min = secs / i;
> +		diff[i] = secs - i * min;
> +		if (!diff[i]) {
> +			i++;
> +			break;
> +		}
> +	}
> +
> +	last = i;
> +	min = 0xff;
> +	for (i = last - 1; i >= first; i--) {
> +		if (diff[i] < min) {
> +			data = i;
> +			min = diff[i];
> +		}
> +		if (!min)
> +			break;
> +	}

Is the additional accuracy worthy the added complexity WRT:

	mult = DIV_ROUND_UP(secs, 0xff);

?

Paolo