Hi,

On Tue, Aug 31, 2021 at 08:49:04AM +0200, Thomas Marangoni wrote:
> > What kind of devices is it going to handle, specifically?
> The AXP209 is a power management chip for e.g. Allwinner A20 (ARM
> architecture), I want to focus on the axp20x-battery driver.
> 
> > > The device has multiple
> > > warnings like: temperature too high or low,
> > You should be able to use the existing thermal framework for this.
> > 
> > > over-voltage, under-voltage
> > But there's no generic way to communicate those AFAICS.
> > 
> > The question here is what user space is going to do about them, though.
> 
> Events I like to communicate to the user-space:
> 
> - Battery Temperature to high (as already mentioned, possible with thermal
> framework)
> - Battery Temperature to low (as already mentioned, possible with thermal
> framework)
> - AXP209Internal over-temperature (should also be possible with thermal
> framework)
> - Low Voltage Level 2 (shuts down the device)
> 
> - Low Voltage Level 1
> - Over Voltage
> - Under Voltage
> - Chargingcurrent shortage
> 
> In my case the application would prepare for a shutdown if the low voltage
> warning level 1 event is received. For the other events the user would get a
> warning with instruction how to handle them from the application.
> 
> > > etc. What is the recommended way to communicate such warnings to the
> > > user-space, besides dmesg, so programs can handle them? Or should I just
> > > create my own attribute for each warning and set it to a specific value
> > > from the kernel and let user-space applications reset them to
> > > acknowledge them? And if there is a recommended way can you name me a
> > > driver that is using it?
> > > 
> > > Thanks for your help!

power-supply drivers can expose the health property, which exposes
status like overvoltage or overheat (just one at a time though.
Drivers should report the most critical one).

In general userspace gets a uevent notification with all properties incl.
health when power_supply_changed() is called for the device. It
definetly should be called when health status changes.

You can find more information in Documentation/ABI/testing/sysfs-class-power
(look for "/health") and by searching the kernel tree for
POWER_SUPPLY_PROP_HEALTH.

-- Sebastian