0.5.10 - new sysfs/uevent battery interface - a couple of issues

[Ash Milsted]

On 28/10/2007, Ash Milsted <thatistosayiseenem at gawab.com> wrote:
> Hi,
> I ended up trying HAL 0.5.10 with a 2.6.24-rc1 kernel and have
> discovered a couple of issues with battery monitoring relating to the
> new kernel interface.
>
> 1. If the procfs interface is enabled (CONFIG_ACPI_PROCFS) as well as
> the new sysfs interface, HAL registers my battery twice (two show up
> in g-p-m). I guess this is a general issue for kernels so configured.
> Duplicate detection based on the name (i.e. BAT1)?
>
> 2.  My Toshiba Satellite S1110 has a mildly broken DSDT such that
> battery (dis)charge uevents do not occur (there does not appear to be
> a BIOS update available). Assuming this is not a very common issue it
> would seem reasonable to create a quirk that causes HAL to fall back
> to polling the sysfs files for changes as it appears to do with
> procfs. Alternatively HAL could start polling if uevents do not appear
> where they are expected e.g. when we are supposed to be (dis)charging.
> Without such a fix the battery monitoring on this laptop regresses wrt
> using the deprecated proc interface, and hacking the DSDT is, of
> course, not within everyone's reach.
>
> Will gladly supply further information.
>
> Cheers,
> Ash
>

It just occured to me that, regarding the second issue, it is probably
significant that uevents for ac plug/unplugs *are* sent by the kernel.
If HAL already has a polling fallback, the presence of these events
(and a single battery 'changed' event sent shortly after bootup - no
more after that one) might be throwing it off.