hi,

> There's clearly some sort of memory leak here.  I'm not sure why only
> Kent and I are seeing it but I don't think this bodes well that 
> something like this can be extant for 6+ months...
> 
> About a week ago, I bumped kern.somaxkva=64MB and the machine ran
> fine for a week. Just now it got clogged up again.  I couldn't bring it
> back alive by bumping kern.somaxkva higher so I had to reboot it. Having
> to reboot a NetBSD machine once a week seems a little sub-optimal.

if one sends some data by loan and then keeps the socket,
the mbuf will stay in the socket's local so_pendfree queue 
up to forever.  i think this is a problem.
although i doubt if it eats 64MB...

YAMAMOTO Takashi