[U-Boot] [PATCH v3 8/9] tegra: i2c: Enable new CONFIG_SYS_I2C framework

[Simon Glass]

Hi,

On Fri, Aug 2, 2013 at 9:55 PM, Heiko Schocher <hs at denx.de> wrote:

> Hello Stephen,
>
> Am 02.08.2013 23:43, schrieb Stephen Warren:
>
>  On 08/01/2013 10:40 PM, Heiko Schocher wrote:
>>
>>> Am 01.08.2013 22:32, schrieb Stephen Warren:
>>>
>> ...
>>
>>> Given how long this discussion is going on, can we please just revert
>>>> the commit so that the code works for everyone who's trying to use it,
>>>> then fix the problem later?
>>>>
>>>
>>> Yes, but not reverting the hole commit, please just remove the
>>> i2c_init_board() call in i2c_init_all() and test it, and send a
>>> new patch, thanks!
>>>
>>> diff --git a/drivers/i2c/i2c_core.c b/drivers/i2c/i2c_core.c
>>> index d1072e8..5f10eb3 100644
>>> --- a/drivers/i2c/i2c_core.c
>>> +++ b/drivers/i2c/i2c_core.c
>>> @@ -246,7 +246,6 @@ void i2c_init_board(void)
>>>    */
>>>   void i2c_init_all(void)
>>>   {
>>> -       i2c_init_board();
>>>          i2c_set_bus_num(CONFIG_SYS_**SPD_BUS_NUM);
>>>          return;
>>>   }
>>>
>>
>> That change doesn't work. Instead of hanging immediately after printing
>> "I2C:", it hangs after printing:
>>
>> I2C:   Caller requested bad clock: periph=-49, parent=2
>>
>> I guess now various data is simply uninitialized since DT parsing hasn't
>> been run at all?
>>
>
> I do not know this ... and what happen if you make init_func_i2c()
> weak and default just dummy?
>
>
>  This is exactly why I suggest a full revert of the problematic patch. If
>> we do that, it should immediately allow all Tegra boards to actually
>> work again. Right now, nobody can use or test u-boot.git/master is
>>
>
> nobody = only tegra boards ... I am working fine with current mainline
> on other arm targets ...
>
>
>  completely blocked by this issue. This is dangerous, since it could
>> easily allow all manner of other problems to appear in other areas
>> without anyone being able to notice. It's my opinion we should
>> immediately unblock them by a revert, and then later work out how to
>> resolve the issues.
>>
>> For the record, "git revert 1f2ba722ac06393d6abe6d4734824d**3b98ea9108"
>> has one simple conflict in README, and certainly allows at least Beaver
>> to work without issue.
>>
>
> Ah, you mean only the switch from the tegra i2c driver to the new
> framework ... Hmm.. Ok, but I would prefer the way fixing the
> current problem ... as I think, nobody want to do this work again ...
>
> But it is alberts descicion on the end, and if nobody has currently
> time for searching the real reason, the best way.


Yes I think we should fix the problem properly rather than just revert. I
will take a look.

Regards,
Simon