Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 31 Aug 2005 17:29:08 +1000
From:      Peter Jeremy <PeterJeremy@optushome.com.au>
To:        hackers@freebsd.org
Subject:   Determining where mbufs are in use
Message-ID:  <20050831072908.GA73852@cirb503493.alcatel.com.au>

next in thread | raw e-mail | index | archive | help
I think I have a system that is leaking mbufs and I'd like some ideas
on how to track down what mbufs are in use.

Background: The system is running 4.9p1 and has been up for about 8
months.  I haven't noticed any problems before.  This afternoon, I
opened an xterm, did an slogin to another system and ran "ipfw l" on
that system.  The output froze about halfway through and I noticed
that my system was barely responding.  A check of the console showed
that the system was out of mbuf clusters (1024).

I killed a couple of ssh sessions and the in-use mbufs and clusters
dropped to 440 but that still strikes me as excessive.  My laptop
shows 12 clusters in use with a maximum of 74 after a month of
uptime.

The system in question is my main workstation at work.  It has dual
heads with a virtual WM and I tend to have lots of xterms open (maybe
50-60, both local and ssh).  It's an NFS client but not server.
netstat showed only a couple of dozen TCP sockets, though there are
a lot of Unix domain sockets (for X) - none had any data queued.

Is there any easy way to work out where all the mbufs have gone?

-- 
Peter Jeremy



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20050831072908.GA73852>