Re: [PATCH v4 REPOST 1/5] of: Add descriptions of thermtrip properties to Tegra PMC bindings

[Alexandre Courbot <acourbot@nvidia.com>]

On 11/12/2014 10:07 PM, Mikko Perttunen wrote:
> On 11/12/2014 02:29 PM, Thierry Reding wrote:
>> On Wed, Nov 12, 2014 at 02:07:51PM +0200, Mikko Perttunen wrote:
>>> On 11/11/2014 08:37 AM, Alexandre Courbot wrote:
>>>> On 11/10/2014 10:12 PM, Mikko Perttunen wrote:
>>>>> From: Mikko Perttunen <mperttunen@nvidia.com>
>>>>>
>>>>> Hardware-triggered thermal reset requires configuring the I2C
>>>>> reset procedure. This configuration is read from the device tree,
>>>>> so document the relevant properties in the binding documentation.
>>>>>
>>>>> Signed-off-by: Mikko Perttunen <mperttunen@nvidia.com>
>>>>> Reviewed-by: Wei Ni <wni@nvidia.com>
>>>>> Tested-by: Wei Ni <wni@nvidia.com>
>>>>> ---
>>>>>   .../bindings/arm/tegra/nvidia,tegra20-pmc.txt      | 24
>>>>> ++++++++++++++++++++++
>>>>>   1 file changed, 24 insertions(+)
>>>>>
>>>>> diff --git
>>>>> a/Documentation/devicetree/bindings/arm/tegra/nvidia,tegra20-pmc.txt
>>>>> b/Documentation/devicetree/bindings/arm/tegra/nvidia,tegra20-pmc.txt
>>>>> index 68ac65f..dc13fb0 100644
>>>>> ---
>>>>> a/Documentation/devicetree/bindings/arm/tegra/nvidia,tegra20-pmc.txt
>>>>> +++
>>>>> b/Documentation/devicetree/bindings/arm/tegra/nvidia,tegra20-pmc.txt
>>>>> @@ -47,6 +47,21 @@ Required properties when nvidia,suspend-mode=<0>:
>>>>>     sleep mode, the warm boot code will restore some PLLs, clocks and
>>>>> then
>>>>>     bring up CPU0 for resuming the system.
>>>>>
>>>>> +Hardware-triggered thermal reset:
>>>>> +On Tegra30, Tegra114 and Tegra124, if the 'i2c-thermtrip' subnode
>>>>> exists,
>>>>> +hardware-triggered thermal reset will be enabled.
>>>>> +
>>>>> +Required properties for hardware-triggered thermal reset (inside
>>>>> 'i2c-thermtrip'):
>>>>> +- nvidia,i2c-bus : Phandle to I2C bus containing the PMU
>>>>> +- nvidia,bus-addr : Bus address of the PMU on the I2C bus
>>>>> +- nvidia,reg-addr : I2C register address to write poweroff command to
>>>>> +- nvidia,reg-data : Poweroff command to write to PMU
>>>>
>>>> This binding is taking two different routes to provide values to the
>>>> driver:
>>>>
>>>> 1) It uses a phandle for i2c-bus (which must then be provided by
>>>> another
>>>> binding implemented in the two following patches)
>>>>
>>>> 2) It uses direct values for bus-addr, reg-addr and reg-data.
>>>>
>>>> Do we need to use both approaches? bus-addr could just as well be
>>>> obtained through a phandle to the i2c device and reading its reg
>>>> property. From this phandle you could also go back up to the bus,
>>>> making
>>>> the i2c-bus property unnecessary. reg-addr and reg-data cannot be
>>>> specified that way, obviously.
>>>
>>> This was in fact how I used to implement this, but Stephen or Thierry
>>> pointed out that the reg property actually might not contain the correct
>>> address (I think because the PMIC could have multiple addresses, and
>>> the one
>>> in DT might not be the one that accepts the reset command).
>>>
>>> The workaround for that was to either add this integer property for
>>> bus-addr
>>> or add a new PMIC API for querying. I went for this as it is much
>>> simpler.
>>>
>>>>
>>>> Actually I think I'd prefer to see i2c-bus become an integer property
>>>> instead of a phandle, because at the end of the day it is a value field
>>>> of a particular register and the reference is only used to retrieve
>>>> that
>>>> value. It is not like we are actually going to call functions on the
>>>> bus
>>>> instance or change its state. And for the single purpose of retrieving
>>>> that value, this binding requires the addition of a new property on the
>>>> bus node that will probably never be used for something else.
>>>
>>> And this was how I used to implement this even earlier, but Thierry
>>> objected
>>> to that since it was duplicating information :)
>>
>> If I remember correctly what I was asking for was to derive as much as
>> possible from simply a phandle. That is, what I was after is a phandle
>> to the PMU and ideally a way for the PMU to pass back information about
>> the register and value for the power off command.
>>
>> Given the lack of a PMU abstraction and how this is probably very Tegra
>> specific I was okay with leaving reg-addr and reg-data in the DT. But if
>> we can't derive even the slave address from a phandle along with the I2C
>> bus master, then I think there remains little point in doing it this way
>> at all.
>>
>> If we're going to duplicate three properties, adding a fourth isn't
>> going to make it much worse.
>>
>> Thierry
>>
>
> Yeah, I guess that's sensible. I'll change the phandle to an integer if
> that's preferred.

As far as I'm concerned, definitely - at the end of the day these values 
are really here just to be written into a register, and not because we 
plan to do fancy things with the PMU. So to me it is really not an issue 
if the DT bindings reflect that fact (not to mention doing otherwise 
would be uselessly cumbersome).