Subject: Re: increasing NMBCLUSTERS
To: Roberto <roberto@redix.it>
From: Simon Burge <simonb@wasabisystems.com>
List: tech-net
Date: 09/30/2005 01:50:12
"Roberto" wrote:

> 
> > On Wed, Sep 28, 2005 at 02:16:12PM -0600, Herb Peyerl wrote:
> >> [...]
> >> 714 mbufs in use:
> >>         146 mbufs allocated to data
> >>         567 mbufs allocated to packet headers
> >>         1 mbufs allocated to socket names and addresses
> >> 25619 calls to protocol drain routines
> >>
> >> [...]
> >> Memory resource pool statistics
> >> Name        Size Requests Fail Releases Pgreq Pgrel Npage Hiwat Minpg
> >> Maxpg Idle
> >> mclpl       2048   183151 22639  117629 40643  7882 32761 32761     4
> >> 32768    0
> >
> 
> Are there any manpages that can give me the meaning of this field?

Not really.  The best info is in /usr/include/sys/pool.h, look for
"struct pool".

> > Hum
> > 183151 - 117629 = 65522, so this really makes 32761 pages.
> > Looks like a leak somewhere, because it seems these clusters are not
> > attached to mbufs (there isn't 65522 mbufs allocated).
> >
> 
> Sorry but I do not understand what you figure out: why 65522 makes 32761
> pages? and why seems to you not attached to mbufs?
> 
> Maybe they are silly question but...

The size of the object is 2048 bytes, and each page is 4k (on most
architectures), so you can fit two per page.

About the mbufs - the top of the message says "146 mbufs allocated to
data" and "567 mbufs allocated to packet headers" - the sum of these
is nowhere near 65522.

Simon.
--
Simon Burge                            <simonb@wasabisystems.com>
NetBSD Support and Service:         http://www.wasabisystems.com/