Subject: is this bad (UDP stats)?
To: None <tech-kern@netbsd.org>
From: Erik E. Fair <fair@clock.org>
List: tech-kern
Date: 10/25/1998 09:15:46
% uptime
 9:09AM  up 20:21, 1 user, load averages: 0.11, 0.12, 0.14
% netstat -i -n
Name  Mtu   Network       Address              Ipkts Ierrs    Opkts Oerrs  Coll
le0   1500  <Link>        08:00:20:04:d3:67    43614    54    19248     0    13
le0   1500  198.68.110    198.68.110.24        43614    54    19248     0    13
le1   1500  <Link>        08:00:20:04:d3:67   472814     0   431891     0  2389
le1   1500  192.168.110   192.168.110.24      472814     0   431891     0  2389
lo0   32976 <Link>                               264     0      264     0     0
lo0   32976 127           127.0.0.1              264     0      264     0     0
% netstat -s
[...]
udp:
        449098 datagrams received
        0 with incomplete header
        0 with bad data length field
        0 with bad checksum
        47 dropped due to no socket
        2363 broadcast/multicast datagrams dropped due to no socket
        0 dropped due to full socket buffers
        446688 delivered
        443201 PCB hash misses
        449136 datagrams output

This is from a SPARC Classic that I rebooted yesterday with a -current
kernel from two days ago. I am just about to reboot it again with today's
kernel, but thought I'd have a look/see at the system stats while I was at
it. This is from "netstat -i" and I'm wondering about the number of UDP
packets received, versus the number of PCB hash misses - 98.6% missed.

This machine has /usr/src NFS mounted read-only for builds against a SPARC
LX running NetBSD 1.3 on a private Ethernet. It also speaks rwhod and NTP.

What's up with all those PCB hash misses?

	curious,

	Erik <fair@clock.org>