Re: [PATCH v3 4/7] thermal/drivers/tegra: Add driver for Tegra30 thermal sensor

[Thara Gopinath <thara.gopinath@linaro.org>]

On 6/15/21 3:32 PM, Dmitry Osipenko wrote:
> 15.06.2021 19:18, Daniel Lezcano пишет:
>> On 15/06/2021 15:01, Dmitry Osipenko wrote:
>>> 15.06.2021 13:26, Viresh Kumar пишет:
>>>> On 15-06-21, 12:03, Daniel Lezcano wrote:
>>>>>
>>>>> [Cc Viresh]
>>>>>
>>>>> On 29/05/2021 19:09, Dmitry Osipenko wrote:
>>>>>> All NVIDIA Tegra30 SoCs have a two-channel on-chip sensor unit which
>>>>>> monitors temperature and voltage of the SoC. Sensors control CPU frequency
>>>>>> throttling, which is activated by hardware once preprogrammed temperature
>>>>>> level is breached, they also send signal to Power Management controller to
>>>>>> perform emergency shutdown on a critical overheat of the SoC die. Add
>>>>>> driver for the Tegra30 TSENSOR module, exposing it as a thermal sensor
>>>>>> and a cooling device.
>>>>>
>>>>> IMO it does not make sense to expose the hardware throttling mechanism
>>>>> as a cooling device because it is not usable anywhere from the thermal
>>>>> framework.
>>>>>
>>>>> Moreover, that will collide with the thermal / cpufreq framework
>>>>> mitigation (hardware sets the frequency but the software thinks the freq
>>>>> is different), right ?
>>>
>>> H/w mitigation is additional and should be transparent to the software
>>> mitigation. The software mitigation is much more flexible, but it has
>>> latency. Software also could crash and hang.
>>>
>>> Hardware mitigation doesn't have latency and it will continue to work
>>> regardless of the software state.
>>
>> Yes, I agree. Both solutions have their pros and cons. However, I don't
>> think they can co-exist sanely.
>>
>>> The CCF driver is aware about the h/w cooling status [1], hence software
>>> sees the actual frequency.
>>>
>>> [1]
>>> https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit?id=344d5df34f5abd468267daa98f041abf90b2f660
>>
>> Ah interesting, thanks for the pointer.
>>
>> What I'm worried about is the consistency with cpufreq.
>>
>> Probably cpufreq_update_limits() should be called from the interrupt
>> handler.
> 
> IIUC, the cpufreq already should be prepared for the case where firmware
> may override frequency. Viresh, could you please clarify what are the
> possible implications of the frequency overriding?
> 
>>>> I am not even sure what the cooling device is doing here:
>>>>
>>>> tegra_tsensor_set_cur_state() is not implemented and it says hardware
>>>> changed it by itself. What is the benefit you are getting out of the
>>>> cooling device here ?
>>>
>>> It allows userspace to check whether hardware cooling is active via the
>>> cooling_device sysfs. Otherwise we don't have ability to check whether
>>> h/w cooling is active, I think it's a useful information. It's also
>>> interesting to see the cooling_device stats, showing how many times h/w
>>> mitigation was active.
>>
>> Actually the stats are for software mitigation. For the driver, create a
>> debugfs entry like what do the other drivers or a module parameter with
>> the stats.
> 
> Okay
> 
>>>>> The hardware limiter should let know the cpufreq framework about the
>>>>> frequency change.
>>>>>
>>>>> 	https://lkml.org/lkml/2021/6/8/1792
>>>>>
>>>>> May be post the sensor without the hw limiter for now and address that
>>>>> in a separate series ?
>>>>
>>>
>>> I wasn't aware about existence of the thermal pressure, thank you for
>>> pointing at it. At a quick glance it should be possible to benefit from
>>> the information about the additional pressure.
>>>
>>> Seems the current thermal pressure API assumes that there is only one
>>> user of the API. So it's impossible to aggregate the pressure from
>>> different sources, like software cpufreq pressure + h/w freq pressure.
>>> Correct? If yes, then please let me know yours thoughts about the best
>>> approach of supporting the aggregation.

Hi,

Thermal pressure is letting scheduler know that the max capacity 
available for a cpu to schedule tasks is reduced due to a thermal event.
So you cannot have a h/w thermal pressure and s/w thermal pressure. 
There is eventually only one capping applied at h/w level and the 
frequency corresponding to this capping should be used for thermal pressure.

Ideally you should not be having both s/w and h/w trying to throttle at 
the same time. Why is this a scenario and what prevents you from 
disabling s/w throttling when h/w throttling is enabled. Now if there 
has to a aggregation for whatever reason this should be done at the 
thermal driver level and passed to scheduler.

>>
>> That is a good question. IMO, first step would be to call
>> cpufreq_update_limits().
> 
> Right
> 
>> [ Cc Thara who implemented the thermal pressure ]
>>
>> May be Thara has an idea about how to aggregate both? There is another
>> series floating around with hardware limiter [1] and the same problematic.
>>
>>   [1] https://lkml.org/lkml/2021/6/8/1791
> 
> Thanks, it indeed looks similar.
> 
> I guess the common thermal pressure update code could be moved out into
> a new special cpufreq thermal QoS handler (policy->thermal_constraints),
> where handler will select the frequency constraint and set up the
> pressure accordingly. So there won't be any races in the code.
> 
It was a conscious decision to keep thermal pressure update out of qos 
max freq update because there are platforms that don't use the qos 
framework. For eg acpi uses cpufreq_update_policy.
But you are right. We have two platforms now applying h/w throttling and 
cpufreq_cooling applying s/w throttling. So it does make sense to have 
one api doing all the computation to update thermal pressure. I am not 
sure how exactly/where exactly this will reside.

So for starters, I think you should replicate the update of thermal 
pressure in your h/w driver when you know that h/w is 
throttling/throttled the frequency. You can refer to cpufreq_cooling.c 
to see how it is done.

Moving to a common api can be done as a separate patch series.

-- 
Warm Regards
Thara (She/Her/Hers)