Hello,

I've got a machine running NetBSD-i386 3.1_STABLE. The machine has the
bad habit of completely locking up.

Today I've active a cronjob which logs the result of "vmstat -m" to a
file every five minutes. Here's the result so far:

20070615-131000:In use 107673K, total allocated 203060K; utilization 53.0%
20070615-131500:In use 107676K, total allocated 203060K; utilization 53.0%
20070615-132000:In use 107695K, total allocated 203060K; utilization 53.0%
20070615-132500:In use 107728K, total allocated 203060K; utilization 53.1%
20070615-133000:In use 107707K, total allocated 203064K; utilization 53.0%
20070615-133500:In use 107758K, total allocated 203064K; utilization 53.1%
20070615-134000:In use 107782K, total allocated 203064K; utilization 53.1%
20070615-134500:In use 107781K, total allocated 203064K; utilization 53.1%
20070615-135000:In use 107810K, total allocated 203064K; utilization 53.1%
20070615-135500:In use 107823K, total allocated 203064K; utilization 53.1%
20070615-140000:In use 107848K, total allocated 203064K; utilization 53.1%
20070615-140500:In use 107865K, total allocated 203064K; utilization 53.1%
20070615-141000:In use 107878K, total allocated 203064K; utilization 53.1%
20070615-141500:In use 107921K, total allocated 203064K; utilization 53.1%
20070615-142001:In use 108144K, total allocated 203064K; utilization 53.3%
20070615-142500:In use 108185K, total allocated 203068K; utilization 53.3%

That looks to me like the kernel is leaking memory somewhere.
Is that assumption correct?

	Kind regards

P.S. The kernel produces a lot of errors like this one:

arplookup: unable to enter address for xxx.xxx.xxx.xxx@yy:yy:yy:yy:yy:yy on ex0 (host is not on local network)

     This is just a side effect of the network setup. Can that code path
     trigger a memory leak?

-- 
Matthias Scheler                                  http://zhadum.org.uk/