[Request] please add cpu temperature meter

[aszurnasirpal]

I think that one of the most important sensor that should be visualized in th system diagnostic tab should be CPU temp.

[MichaIng]

Good idea. I wonder how to implement best, I guess loading DietPi-Globals to run the G_OBTAIN_CPU_TEMP function in a shell is pretty overkill, though to give a precise temperature, different /sys tunables need to be read on every hardware. Probably there is a slim library available as Rust crate to do this with higher efficiency and same precision without invoking an external process.

Last but not least the logic could be cloned: https://github.com/MichaIng/DietPi/blob/733d81e/dietpi/func/dietpi-globals#L1692-L1722
But this means doubled code maintenance if something changes 🤔.

[ravenclaw900]

The crate that I'm currently using to read CPU, RAM, etc has a function for that: https://docs.rs/heim/0.1.0-rc.1/heim/sensors/fn.temperatures.html.

[MichaIng]

That is great. Hopefully it is accurate on the various SBCs we support.

[ravenclaw900]

I've been trying to work on this, but I've run into some problems (heim-rs/heim#354). Until that gets resolved, my basic changes are in the https://github.com/ravenclaw900/DietPi-Dashboard/tree/cpu-temp branch.

[BeekeeperNL]

isn't it something like this?....

root@DietPi:/# watch -n 2 sensors

Every 2.0s: sensors DietPi: Thu Feb 10 17:17:03 2022

amdgpu-pci-0008
Adapter: PCI adapter
vddgfx: 1.55 V
vddnb: 787.00 mV
edge: +18.0°C

k10temp-pci-00c3
Adapter: PCI adapter
temp1: +18.8°C (high = +70.0°C)
(crit = +100.0°C, hyst = +99.0°C)

fam15h_power-pci-00c4
Adapter: PCI adapter
power1: 9.05 mW (avg = 1.19 mW, interval = 0.01 s)
(crit = 6.00 W)

[ravenclaw900]

While that might work (though it doesn't on my RPi, does the sensors command need to be installed?), we try to use native Rust code instead of shell scripts, to keep the binary self-contained.

[MichaIng]

apt install lm-sensors and the (C) library is libsensors5.

[BeekeeperNL]

ah okay, i understand. sorry i was having trouble seeing why it was difficult to have the cpu temp seeing as it is posted each time i login via ssh. I understand now why this is not what you are trying to do, least not in this way. and yes, i used lm-sensors and no.. it's not a pi. but it should work just as well on a pi.

[MichaIng]

Via bash shell and sourcing /boot/dietpi/func/dietpi-globals, there is G_OBTAIN_CPU_TEMP available which does the job without additional dependencies/libraries. But loading a shell + this quite large script is probably too much overhead:

bash -c '. /boot/dietpi/func/dietpi-globals; G_OBTAIN_CPU_TEMP'

Otherwise the logic, looping though the kernel API until a valid integer was found, could be directly implemented: https://github.com/MichaIng/DietPi/blob/4eee867/dietpi/func/dietpi-globals#L1674-L1716

[MichaIng]

Implemented already for next release v0.6.1 with: #232