[U-Boot] [PATCH 3/3] mxc_i2c: remove setting speed at each start

[Troy Kisky]

On 5/5/2012 2:36 PM, Marek Vasut wrote:
> Dear Troy Kisky,
>
>> On 5/5/2012 6:08 AM, Marek Vasut wrote:
>>> Dear Troy Kisky,
>>>
>>>> On 4/24/2012 8:33 PM, Troy Kisky wrote:
>>>>> Other then being very weird, this code was also wrong.
>>>>> For example, say I set speed to 100K. I'll read back the speed
>>>>> as 85937. But the speed is really 85937.5, so we I reset
>>>>> the speed to 85937, I'll get 73660.7. After a couple of transactions
>>>>> my speed is now exactly 68750 so it will remain there.
>>>>>
>>>>> Signed-off-by: Troy Kisky<troy.kisky at boundarydevices.com>
>>>>> ---
>>>>>
>>>>>     drivers/i2c/mxc_i2c.c |    6 ------
>>>>>     1 files changed, 0 insertions(+), 6 deletions(-)
>>>>>
>>>>> diff --git a/drivers/i2c/mxc_i2c.c b/drivers/i2c/mxc_i2c.c
>>>>> index 416ffee..fc68062 100644
>>>>> --- a/drivers/i2c/mxc_i2c.c
>>>>> +++ b/drivers/i2c/mxc_i2c.c
>>>>> @@ -231,12 +231,6 @@ int i2c_imx_start(void)
>>>>>
>>>>>     	struct mxc_i2c_regs *i2c_regs = (struct mxc_i2c_regs *)I2C_BASE;
>>>>>     	unsigned int temp = 0;
>>>>>     	int result;
>>>>>
>>>>> -	int speed = i2c_get_bus_speed();
>>>>> -	u8 clk_idx = i2c_imx_get_clk(speed);
>>>>> -	u8 idx = i2c_clk_div[clk_idx][1];
>>>>> -
>>>>> -	/* Store divider value */
>>>>> -	writeb(idx,&i2c_regs->ifdr);
>>>>>
>>>>>     	/* Enable I2C controller */
>>>>>     	writeb(0,&i2c_regs->i2sr);
>>>> Marek would you care to ack/nak this? It is deleting code that you
>>>> added.
>>> Ok, who will set the controller speed if you remove this?
>> i2c_init is the only function that writes the ifdr register after this
>> patch.
> And i2c_init() is called on every boot. Correct?
>
>> This is fine because this register is not affected by a software reset.
> I take it you verified this or that you're sure here :)
>
I haven't looked at every reference manual, but if some oddball chip 
needs it reinitialized after
a software reset, then the fix should go into the i2c_reset function, 
not i2c_imx_start.
And this patch would not be introducing a regression for the oddball 
chip anyway.

Troy