Well the reason that I am assuming kernel memory is because the memory
usage brings the buffers and cache down to about 200k each and the swap
only gets up to about 5 Megs ( on a 32 Meg system )
>
>
>I am going to try and update to 2.1.42 and see if this helps things.
I'm not ready to make the jump yet, I have been VERY happy with the 2.0.x
performance thus far, with this little exception.
>
>Please let me know if you come up with a solution!
An Update:
I did a <shift> scroll lock after the system had been up for several
hours. I noticed that the "Network Buffers in use" was growing
constantly. At the time they were at about 32,000 buffers ( This seemed
odd to me. ) At the suggestion of Alexey Kuznetsov I recompiled the
kernel with "RT_CACHE_DEBUG" in route.h set to "2". and adjusted the
"ARP_MAXSIZE" in arp.c back to 256 ( to keep the debug traffic from
arp_force_expire() messages down. )
The result: this seems to have fixed the problem. My Network buffers in
use does not grow constantly. It seems to hover around 1. The only thing
that I can think of is that there is a timing issue with the locking and
unlocking of a resource somewhere that the extra printk()s from the
debugging is fixing.
I know this is not a fix... but it seems to work for now. I will begin to
take out debugging code and see when the problem resurfaces.
If any one has helpful comments on where in the source I can look or some
debugging tricks to try please send them to me.
Thanks for all of your help thus far,
~Sean
-- Sean@ibasys.net @work http://www.ibasys.net Sean@fredcom.com @play http://www.fredcom.comChief Linux Advocate and HP9000 Systems Admin Guy
My 2 year olds answer to everything: "No!" You can learn a lot from a 2 year old....
May the farce be with you...