Hi,

Le mercredi 31 mai 2017 à 19:32 +0200, Pavel Machek a écrit :
> The status reported directly by the battery controller is not always
> > > > reliable and should be corrected based on the current draw information.
> > > > 
> > > > This implements such a correction with a dedicated function, called
> > > > when retrieving the supply status.
> > > > @@ -1182,6 +1196,8 @@ static int bq27xxx_battery_status(struct
> > > > bq27xxx_device_info *di,
> > > >  		else
> > > >  			status = POWER_SUPPLY_STATUS_DISCHARGING;
> > > >  	} else {
> > > > +		curr = (int)((s16)curr) * 1000;
> > > 
> > > Umm.
> 
> As in "two casts in one expression -- too ugly to live".

Oh, I had skipped that comment, sorry about that. Yeah I understand your
concern. However, this line was mostly inspired by another part of the code,
below the following comment: /* Other gauges return signed value */

I think we should fix the first occurence first and then used the fixed syntax
in v2 of this patch. What do you think?

> > > > @@ -1190,6 +1206,18 @@ static int bq27xxx_battery_status(struct
> > > > bq27xxx_device_info *di,
> > > >  			status = POWER_SUPPLY_STATUS_CHARGING;
> > > >  	}
> > > >  
> > > > +
> > > > +	if (curr == 0 && status != POWER_SUPPLY_STATUS_NOT_CHARGING)
> > > > +		status = POWER_SUPPLY_STATUS_FULL;
> > > > +
> > > > +	if (status == POWER_SUPPLY_STATUS_FULL) {
> > > > +		/* Drawing or providing current when full */
> > > > +		if (curr > 0)
> > > > +			status = POWER_SUPPLY_STATUS_CHARGING;
> > > > +		else if (curr < 0)
> > > > +			status = POWER_SUPPLY_STATUS_DISCHARGING;
> > > > +	}
> > > 
> > > Are you sure this works? On N900, we normally see small currents to/from
> > > "full" battery.
> > 
> > In my case, this works perfectly and I am quite surprised of what you're
> > describing. Is it the case when the battery has a PSU connected?
> 
> "PSU"? This is cellphone. It has USB connection and charges from that.
> 
> It has been charging for long while now, and current_now fluctuates
> between 20706 and -2856. USB has limitted current, so I guess "draw
> current from battery if we need more than USB can provide" is quite common.

Ah right, I had forgotten about the USB current limitation thing. In this case,
I guess the battery is never actually full and IMO, it should be reported as
such.

> pavel@n900:~$ cat /sys/class/power_supply/bq27200-0/current_now
> 5355
> pavel@n900:~$ cat /sys/class/power_supply/bq27200-0/current_now
> 5355
> pavel@n900:~$ cat /sys/class/power_supply/bq27200-0/current_now
> -4105
> pavel@n900:~$ cat /sys/class/power_supply/bq27200-0/current_now
> -4105
> pavel@n900:~$ cat /sys/class/power_supply/bq27200-0/current_now
> -7675
> pavel@n900:~$ cat /sys/class/power_supply/bq27200-0/current_now
> -5712
> pavel@n900:~$ #screen on
> pavel@n900:~$ cat /sys/class/power_supply/bq27200-0/current_now
> 4641
> pavel@n900:~$ cat /sys/class/power_supply/bq27200-0/current_now
> 4641
> pavel@n900:~$ cat /sys/class/power_supply/bq27200-0/current_now
> 37842
> pavel@n900:~$ cat /sys/class/power_supply/bq27200-0/current_now
> 16600
> pavel@n900:~$
> 
> > I guess I would consider this a hardware issue (leak currents) and we could
> > definitely set some range (in device-tree) to distinguish between full +
> > leak
> > currents and bad reporting from the fuel gauge. That would work well in my
> > case
> > too.
> 
> I'd pass to userspace what the controller reports. Yes, I seldom see
> "STATUS_FULL" but that may be a problem we need to track down.

The controller is known, from my experience, to not be reliable in that regard,
so I don't think it makes sense to pass a state that doesn't reflect the actual
state of charging just because the chip tells us so.

Worst case, we could also have a dt property to enable that kind of fixup
workaround and let every device maintainer decide whether it is relevant for
their device.

What do you think?

-- 
Paul Kocialkowski, developer of free digital technology and hardware support

Website: https://www.paulk.fr/
Coding blog: https://code.paulk.fr/
Git repositories: https://git.paulk.fr/ https://git.code.paulk.fr/