Subject: Re: Fix for KAME PF_KEY DUMP-truncation marshalling bug
To: Jason Thorpe <thorpej@wasabisystems.com>
From: Jonathan Stone <jonathan@dsg.stanford.edu>
List: tech-net
Date: 05/28/2004 18:01:50
In message <1E19C422-B105-11D8-9AE5-000A957650EC@wasabisystems.com>Jason Thorpe
>There is no limit on little mbufs, per se.  I.e. the pool has no high 
>water mark.
>
>How about we just pick some value?  Say, 512K?  More?  Less?

I dunno; The numbers I personally run at are exceessive for most
people's taste. (5 Mbytes?)  Given the very low packing density we
get, I'd probably go for roughly 3x to 4x kern.sbmax, whatever it is
at boot.

How about this: we aim for a low default, around 768K, and I'll add
two more sysctl knobs: one to record the peak data enqueued on a
registered socket, and a second to count drops due to exceeding the
registered-message limit.

Is it worth clipping the default limit further on very
low-memory machines?