In article <Pine.NEB.4.60.0502012028580.13437@kvik.lan.dinsen.net>,
Anders Dinsen  <anders@dinsen.net> wrote:
>I'm building a new server. Configuration is NetBSD 2.0 with raidframe set 
>up for RAID 1 on two identical disks. Box is ASUS P5P800 Intel based MB 
>with Celeron D CPU.
>
>dmesg below.
>
>It freezes for short periods from time to time. This started after I added 
>two 3C509 based netcards. The problem goes away when I remove one. It 
>looks like this when I ping one card from another host:
>
>64 bytes from 10.10.144.22: icmp_seq=6787 ttl=255 time=1.809 ms
>64 bytes from 10.10.144.22: icmp_seq=6788 ttl=255 time=1.784 ms
>64 bytes from 10.10.144.22: icmp_seq=6789 ttl=255 time=2606.026 ms
>64 bytes from 10.10.144.22: icmp_seq=6790 ttl=255 time=1606.939 ms
>64 bytes from 10.10.144.22: icmp_seq=6791 ttl=255 time=607.168 ms
>64 bytes from 10.10.144.22: icmp_seq=6792 ttl=255 time=1.809 ms
>64 bytes from 10.10.144.22: icmp_seq=6793 ttl=255 time=1.817 ms
>
>It's not only the network that freezes, the machine stops responding on 
>the console too.
>
>Breaking the kernel into the debugger when it's frozen reveals the 
>following backtrace:

You mean 3c9xx prolly. 3c509 is isa and it does not have mii. Anyway,
if you suspect it is mii related, why don't you try to turn autonegotiation
off and hard-code the link settings?

christos