ACPI problems (was: Re: Synaptics PS/2 driver and 2.6.0-test11)

ACPI problems (was: Re: Synaptics PS/2 driver and 2.6.0-test11)

[Lukas Hejtmanek]

Hello,

I've found that problem with synaptics is really related to some ACPI troubles.
I've suspected synaptics touchpad to do some interrupt bursts and it does.
Just single button press and relelase generates about 500 interrupts!!! I wonder
if it is driver related or really hardware related...

I've found that modem driver (slamr module for built in soft modem) suffers from
the same problem. Gnome battery stat applet freezes connection. 

However with 2.4.23 kernel is all OK. So it seems to be something broken with
ACPI. 

I remind that gnome battery stat applet causes many lost interrupts. (Interrupt
timeout in i8042.c)

-- 
Lukáš Hejtmánek

Re: ACPI problems (was: Re: Synaptics PS/2 driver and 2.6.0-test11)

[Peter Osterlund]

Lukas Hejtmanek <xhejtman@hell.sks3.muni.cz> writes:

> Hello,
> 
> I've found that problem with synaptics is really related to some ACPI troubles.
> I've suspected synaptics touchpad to do some interrupt bursts and it does.
> Just single button press and relelase generates about 500 interrupts!!! I wonder
> if it is driver related or really hardware related...

That's normal. The hardware generates 80 packets/second and it keeps
generating packets 1 second after you stop using the touchpad. Each
packet is 6 bytes and the i8042 controller generates one
interrupt/byte.

500 interrupts/s shouldn't be any problem to handle though.

-- 
Peter Osterlund - petero2@telia.com
http://w1.894.telia.com/~u89404340

Re: ACPI problems (was: Re: Synaptics PS/2 driver and 2.6.0-test11)

[Lukas Hejtmanek]

On Tue, Dec 30, 2003 at 01:37:11AM +0100, Peter Osterlund wrote:
> That's normal. The hardware generates 80 packets/second and it keeps
> generating packets 1 second after you stop using the touchpad. Each
> packet is 6 bytes and the i8042 controller generates one
> interrupt/byte.
> 
> 500 interrupts/s shouldn't be any problem to handle though.

Well it depends. If ACPI disables interrupts or it keeps some spin lock then
there is a high probability that some interrupts are lost. 

If I do (with no load nor i/o stress):

# time cat /proc/acpi/battery/BAT0/state
present:                 yes
capacity state:          ok
charging state:          unknown
present rate:            unknown
remaining capacity:      4000 mAh
present voltage:         14800 mV

real    0m1.118s
user    0m0.001s
sys     0m0.007s


However for me it looks odd that sound is not corrupted (and I believe 
that dma buffer does not hold 1 sec of sound data) while serio and soft modem
is.

-- 
Lukáš Hejtmánek