Re: [PATCH] ACPI / Battery: Return -ENODATA for unknown values in get_property()

[Sitsofe Wheeler <sitsofe@yahoo.com>]

On Sun, Oct 17, 2010 at 03:19:53AM -0200, Henrique de Moraes Holschuh wrote:
> On Sun, 17 Oct 2010, Rafael J. Wysocki wrote:
> > In fact, the driver is supposed to return -ENODATA in that case, which will
> > result in the read from /sys/class/power_supply/BAT0/current_now
> > fail (I guess upower should be able to cope with that).
>
> ENODATA?  Shouldn't it be ENXIO?  There is no non-blocking data stream
> involved in a sysfs attribute.

Using ENODATA and ENXIO appears to solve the problem (upower reports a
rate of 0.0). However when plugging the battery in after previously only
being on AC power none of the /sys/class/power_supply/BAT0/uevent:*
files are created so upower never realises a battery has been plugged
in. A further issue with ENXIO is is the following repeatedly appears in
dmesg:

power_supply BAT0: driver failed to report `current_now' property

> Of course, the RIGHT thing would be to not expose in sysfs attributes
> that are unsupported by the firmware/hardware in the first place, but
> that's easier said than done.

My understanding is that this very hard to do because you can't tell if
the problem was transient (battery settling) or permanent (feature not
supported).

-- 
Sitsofe | http://sucs.org/~sits/