Re: [PATCH 4/5] ARM: dts: sam9x60: add rtt

[Claudiu.Beznea]

On 14.04.2020 15:47, Alexandre Belloni wrote:
> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
> 
> On 14/04/2020 12:13:46+0000, Claudiu.Beznea@xxxxxxxxxxxxx wrote:
>>
>>
>> On 14.04.2020 14:16, Alexandre Belloni wrote:
>>> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
>>>
>>> On 14/04/2020 08:42:08+0000, Claudiu.Beznea@xxxxxxxxxxxxx wrote:
>>>>> Why would one use the RTT while the RTC is far superior?
>>>>
>>>> I didn't enabled this for a particular use case, but: couldn't this be used
>>>> by some user that wants to generate multiple alarms? from multiple RTCs?
>>>>
>>>
>>> I very much doubt that as Linux is able to properly multiplex alarms and
>>> basically, the only one we are interested in is actually wakeup.
>>
>> I think you can use the wakealarm sysfs exported file to prepare an alarm
>> and take user space actions based on that without being suspended.
>>
>>>
>>>> Moreover, this IP's counter has the possibility of being clocked at 1Hz.
>>>> Couldn't this minimize the power consumption while being in a power saving
>>>> mode?
>>>>
>>>
>>> And that 1Hz clock is coming from the RTC so using the RTC is
>>> definitively consuming less power.
>>
>> Datasheet specifies this: "Configuring the RTPRES field value to 0x8000
>> (default value) corresponds to feeding the real-time counter with a
>>
>> 1Hz signal (if the slow clock is 32.768 kHz)."
>>
>> So, it is not the RTC, it is the slow clock divided by 32768.
> 
> This is not what you described previously,

I said this way: "this *IP's counter* has the possibility of being clocked at 1Hz"

> using RTPRES means running
> the RTT at 32kHz. This is exactly what happens with the RTC but you get
> the added clock calibration circuitry that is probably not drawing to
> much power but the added consumption of the configurable prescaler
> versus the static prescaler of the RTC is probably similar.
> 
> Using RTC1HZ would be driving the RTT at 1Hz.
> 
>>> But this is very unlikely to happen because this would be limited to a
>>> single board device tree instead of impact every sam9x60 based boards.
>>
>> Very unlikely but a having a patch with diff like this:
>>
>> +&gpbr {
>> +     status = "okay";
>> +};
>> +
>> +&rtt {
>> +     atmel,rtt-rtc-time-reg = <&gpbr 0x0>;
>> +     status = "okay";
>> +};
>> +
>>
>> and reverting it may affect the other users of gpbr in sam9x60ek.dts.
>>
> 
> Again, this affects only sam9x60ek.dts instead of possibly multiple DTs
> that may be out of tree. So the risk of doing that is null.

Anyway... I'll merge it although I don't consider is the right way.

> 
> --
> Alexandre Belloni, Bootlin
> Embedded Linux and Kernel engineering
> https://bootlin.com
>