i2c: rk3x: fix divisor calculation for SCL frequency
diff mbox

Message ID 1409884333-3544-1-git-send-email-addy.ke@rock-chips.com
State New, archived
Headers show

Commit Message

addy ke Sept. 5, 2014, 2:32 a.m. UTC
I2C_CLKDIV register descripted in the previous version of
RK3x chip manual is incorrect. Plus 1 is required.

The correct formula:
- T(SCL_HIGH) = T(PCLK) * (CLKDIVH + 1) * 8
- T(SCL_LOW) = T(PCLK) * (CLKDIVL + 1) * 8
- (SCL Divsor) = 8 * ((CLKDIVL + 1) + (CLKDIVH + 1))
- SCL = PCLK / (CLK Divsor)

It will be updated to the latest version of chip manual.

Signed-off-by: Addy Ke <addy.ke@rock-chips.com>
---
 drivers/i2c/busses/i2c-rk3x.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Comments

Doug Anderson Sept. 5, 2014, 4:31 a.m. UTC | #1
Addy,

On Thu, Sep 4, 2014 at 7:32 PM, Addy Ke <addy.ke@rock-chips.com> wrote:
> I2C_CLKDIV register descripted in the previous version of
> RK3x chip manual is incorrect. Plus 1 is required.
>
> The correct formula:
> - T(SCL_HIGH) = T(PCLK) * (CLKDIVH + 1) * 8
> - T(SCL_LOW) = T(PCLK) * (CLKDIVL + 1) * 8
> - (SCL Divsor) = 8 * ((CLKDIVL + 1) + (CLKDIVH + 1))
> - SCL = PCLK / (CLK Divsor)

I'll trust that you tested this with a scope


> It will be updated to the latest version of chip manual.
>
> Signed-off-by: Addy Ke <addy.ke@rock-chips.com>
> ---
>  drivers/i2c/busses/i2c-rk3x.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/i2c/busses/i2c-rk3x.c b/drivers/i2c/busses/i2c-rk3x.c
> index e637c32..76b6604 100644
> --- a/drivers/i2c/busses/i2c-rk3x.c
> +++ b/drivers/i2c/busses/i2c-rk3x.c
> @@ -433,8 +433,8 @@ static void rk3x_i2c_set_scl_rate(struct rk3x_i2c *i2c, unsigned long scl_rate)
>         unsigned long i2c_rate = clk_get_rate(i2c->clk);
>         unsigned int div;
>
> -       /* SCL rate = (clk rate) / (8 * DIV) */
> -       div = DIV_ROUND_UP(i2c_rate, scl_rate * 8);
> +       /* SCL rate = (clk rate) / (8 * (DIV + 2)) */
> +       div = DIV_ROUND_UP(i2c_rate, scl_rate * 8) - 2;

Given the bug I just fixed in the Rockchip SPI driver, I was a little
worried about div becoming -1 (and thus being a really large positive
number since div is unsigned).

However, it seems that you get saved by the next statement:
  div = DIV_ROUND_UP(div, 2);

In the testing I did with the Linux macros, that magically transformed
a div of 0xFFFFFFFF (-1) to 0, so it's not technically a bug.  ...but
it's very non-obvious.  Can you do something a little cleaner?

-Doug
addy ke Sept. 5, 2014, 10:17 a.m. UTC | #2
> Addy,
> 
> On Thu, Sep 4, 2014 at 7:32 PM, Addy Ke <addy.ke@rock-chips.com> wrote:
>> I2C_CLKDIV register descripted in the previous version of
>> RK3x chip manual is incorrect. Plus 1 is required.
>>
>> The correct formula:
>> - T(SCL_HIGH) = T(PCLK) * (CLKDIVH + 1) * 8
>> - T(SCL_LOW) = T(PCLK) * (CLKDIVL + 1) * 8
>> - (SCL Divsor) = 8 * ((CLKDIVL + 1) + (CLKDIVH + 1))
>> - SCL = PCLK / (CLK Divsor)
> 
> I'll trust that you tested this with a scope
> 
Yes ,I have tested on RK3188 and RK3288, and confirmed by oscilloscope.
> 
>> It will be updated to the latest version of chip manual.
>>
>> Signed-off-by: Addy Ke <addy.ke@rock-chips.com>
>> ---
>>  drivers/i2c/busses/i2c-rk3x.c | 4 ++--
>>  1 file changed, 2 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/i2c/busses/i2c-rk3x.c b/drivers/i2c/busses/i2c-rk3x.c
>> index e637c32..76b6604 100644
>> --- a/drivers/i2c/busses/i2c-rk3x.c
>> +++ b/drivers/i2c/busses/i2c-rk3x.c
>> @@ -433,8 +433,8 @@ static void rk3x_i2c_set_scl_rate(struct rk3x_i2c *i2c, unsigned long scl_rate)
>>         unsigned long i2c_rate = clk_get_rate(i2c->clk);
>>         unsigned int div;
>>
>> -       /* SCL rate = (clk rate) / (8 * DIV) */
>> -       div = DIV_ROUND_UP(i2c_rate, scl_rate * 8);
>> +       /* SCL rate = (clk rate) / (8 * (DIV + 2)) */
>> +       div = DIV_ROUND_UP(i2c_rate, scl_rate * 8) - 2;
> 
> Given the bug I just fixed in the Rockchip SPI driver, I was a little
> worried about div becoming -1 (and thus being a really large positive
> number since div is unsigned).
> 
> However, it seems that you get saved by the next statement:
>   div = DIV_ROUND_UP(div, 2);
> 
> In the testing I did with the Linux macros, that magically transformed
> a div of 0xFFFFFFFF (-1) to 0, so it's not technically a bug.  ...but
> it's very non-obvious.  Can you do something a little cleaner?

The following modifications is reasonable?

static void rk3x_i2c_set_scl_rate(struct rk3x_i2c *i2c, unsigned long scl_rate)
{
        unsigned long i2c_rate = clk_get_rate(i2c->clk);
        unsigned int div;

        /* set DIV = DIVH = DIVL
         * SCL rate = (clk rate) / (8 * (DIVH + 1 + DIVL + 1))
         *          = (clk rate) / (16 * (DIV + 1))
         */
        div = DIV_ROUND_UP(i2c_rate, scl_rate * 16) - 1;

        i2c_writel(i2c, (div << 16) | (div & 0xffff), REG_CLKDIV);
}
> 
> -Doug
> 
> 
>
Doug Anderson Sept. 5, 2014, 3:20 p.m. UTC | #3
Addy,

On Fri, Sep 5, 2014 at 3:17 AM, addy ke <addy.ke@rock-chips.com> wrote:

> The following modifications is reasonable?
>
> static void rk3x_i2c_set_scl_rate(struct rk3x_i2c *i2c, unsigned long scl_rate)
> {
>         unsigned long i2c_rate = clk_get_rate(i2c->clk);
>         unsigned int div;
>
>         /* set DIV = DIVH = DIVL
>          * SCL rate = (clk rate) / (8 * (DIVH + 1 + DIVL + 1))
>          *          = (clk rate) / (16 * (DIV + 1))
>          */
>         div = DIV_ROUND_UP(i2c_rate, scl_rate * 16) - 1;
>
>         i2c_writel(i2c, (div << 16) | (div & 0xffff), REG_CLKDIV);
> }

Yes, that looks much cleaner and is a nice solution, thanks!  Can you
send up a new patch version?

-Doug

Patch
diff mbox

diff --git a/drivers/i2c/busses/i2c-rk3x.c b/drivers/i2c/busses/i2c-rk3x.c
index e637c32..76b6604 100644
--- a/drivers/i2c/busses/i2c-rk3x.c
+++ b/drivers/i2c/busses/i2c-rk3x.c
@@ -433,8 +433,8 @@  static void rk3x_i2c_set_scl_rate(struct rk3x_i2c *i2c, unsigned long scl_rate)
 	unsigned long i2c_rate = clk_get_rate(i2c->clk);
 	unsigned int div;
 
-	/* SCL rate = (clk rate) / (8 * DIV) */
-	div = DIV_ROUND_UP(i2c_rate, scl_rate * 8);
+	/* SCL rate = (clk rate) / (8 * (DIV + 2)) */
+	div = DIV_ROUND_UP(i2c_rate, scl_rate * 8) - 2;
 
 	/* The lower and upper half of the CLKDIV reg describe the length of
 	 * SCL low & high periods. */