Re: Timer issues

[Stefan Krüger <skrueger@xxxxxxxxxxxxxxxx>]

Matthew Dillon wrote:
> :
> :Matthew Dillon wrote:
> :> :
> :> :
> :> :Hey Matt (and the src gang),
> :> :
> :> :I am having some serious timer issues.  My clock is hosed, which wouldn't be a
> :> :problem if it wern't for the tme of day.  The clock is so far off that cvs is
> :> :hosed and it totally screws up my projects.
> :> :
> :> :I can fool around with my test box, so any patches I can try easily enough.
> :> :--
> :> :jm
> :> 
> :>     Do a boot -v and post your dmesg output.  What do you mean by 'hosed' ?
> :I have that problem, too
> :
> :It happens when I run some kernel/regression tests in single-user mode,
> :I had them running for a few hours and the clock was off nearly an hour
> :
> :strange...
> 
>     In Tomaz's case the video BIOS plays a big role in the time getting
>     off because he isn't running ACPI so his system is trying to use 
>     the second 8254 timer for timekeeping instead of the ACPI timer.
>     Video BIOS calls seem to mess up that 8254 timer on a number of BIOSes.
> 
>     If you are using any vesa console modes, a mouse in console mode,
>     splash screen, etc, try turning it all off and just use a plain 
>     vanilla console.
> 
>     Does the problem occur if you are not using the console?  e.g. if
>     you are using an X session or ssh into the box remotely to run the
>     tests?
I found out how to reproduce this easily :-)

first compile src/test/stress/eattls.c and then run the executable 10
times (could be less, but I did it w/ 10)

#!/bin/sh

. /eattls &
. /eattls &
. /eattls &
. /eattls &
. /eattls &
. /eattls &
. /eattls &
. /eattls &
. /eattls &
. /eattls &

my clock was out of sync after a few secs

hope this helps

SK