[ACPI-sppt] Loosing timer interrupts with ACPI

Please, CC me any reply to this message, since I'm not
subscribed to this list.


When periodically reading /proc/acpi/battery/BAT0/info (once
per second), running "adjtimex --compare=10000" indicates
slight slowing down of the system clock. It seems that
reading battery info may lead to loosing timer interrupts.

System: Compaq Presario 900Z notebook
        (Athlon XP 2000+, ATI IGP320M(U1) chipset,
         ALI southbridge)
Kernel: 2.4.21-ac1 (includes the latest ACPI patch)

First time the system clock drift was noticed when running
akpi applet for KDE: ntpdate time corrections were too
large. Then I've found that it happens when simply reading
battery info. It is well reproducable, the time drift is
between 20 to 50 ms per every 10 seconds. This is in both
2.4.20+acpi and 2.4.12-ac1.

To test this, run

adjtimex --compare=10000

in first xterm, and the following script on another xterm:

#!/bin/sh

for i in `find /proc/acpi/battery -type f`
do
    echo $i
    count=0
    while [ $count -lt 30 ]
    do
        cat $i > /dev/null 2>/dev/null
        sleep 1
        count=$[count+1]
    done
done

Serguei Miridonov.



-------------------------------------------------------
This SF.NET email is sponsored by: eBay
Great deals on office technology -- on eBay now! Click here:
http://adfarm.mediaplex.com/ad/ck/711-11697-6916-5