Re: Polling for battery stauts and lost keypresses

[Pavel Machek <pavel@ucw.cz>]

On Wed 16-08-06 09:31:48, Giuseppe Bilotta wrote:
> On Mon, 14 Aug 2006 16:17:01 -0400, Dmitry Torokhov wrote:
> 
> > On 8/14/06, Giuseppe Bilotta <bilotta78@hotpop.com> wrote:
> >> On Mon, 14 Aug 2006 12:06:06 -0400, Dmitry Torokhov wrote:
> >>
> >>> On many laptops (including mine) polling battery takes a loooong time
> >>> and is done in SMI mode in BIOS causing lost keypresses, jerky mouse
> >>> etc. It is pretty common problem. I think I have my ACPI client
> >>> refreshing every 3 minutes.
> >>
> >> BTW, polling battery status takes a lot on a Dell Inspiron 8200 too,
> >> and all keypresses and mouse movements (and I think even network
> >> IRQs?) are totally *dead* while polling.
> >>
> >> However, The Other OS(tm) *seems* to do it right enough to have no
> >> noticeable keypress losses, even when updating the battery status. Is
> >> it using different system calls, or what?
> >>
> > 
> > I am not sure, but there are many things that may affect it:
> > 
> > 1. Battry attributes are divided into 2 groups - static (i think they
> > go into /proc/acpi/battery/<name>/info and dynamic
> > (/proc/acpi/batetry/state). Static attributes take really long time to
> > pull and they do not change so it may wery well be they are polled one
> > at startup. Dynamic attributes are cheaper to poll and even then OS
> > may cache access or limit rate.
> 
> Well, this would explain why Linux freezes while polling only if Linux
> polls for the slow, static ones just as much as it does for the
> dynamic ones ...

I guess patch caching battery/*/state would be welcome.
-- 
Thanks for all the (sleeping) penguins.