From owner-freebsd-questions  Tue May 23 13:16:05 1995
Return-Path: questions-owner
Received: (from majordom@localhost)
          by freefall.cdrom.com (8.6.10/8.6.6) id NAA11576
          for questions-outgoing; Tue, 23 May 1995 13:16:05 -0700
Received: from easynet.com (easy1.easynet.com [199.2.26.10])
          by freefall.cdrom.com (8.6.10/8.6.6) with SMTP id NAA11568
          for <freebsd-questions@freebsd.org>; Tue, 23 May 1995 13:16:02 -0700
Received: by easynet.com (Smail3.1.28.1 #7)
	id m0sE0N3-000rdJC; Tue, 23 May 95 13:16 WET DST
Message-Id: <m0sE0N3-000rdJC@easynet.com>
From: brian@mediacity.com (Brian Litzinger)
Subject: troubles with mbuf's
To: freebsd-questions@FreeBSD.org
Date: Tue, 23 May 1995 13:16:04 -0700 (PDT)
X-Mailer: ELM [version 2.4 PL23]
Content-Type: text
Content-Length: 1042      
Sender: questions-owner@FreeBSD.org
Precedence: bulk

I have a device driver which works correctly on FreeBSD 2.0R, but on
-current it leaks MBUF CLUSTERS, while it doesn't leak MBUFs.

Or perhaps it  leaks MBUF clusters on 2.0R too, but I can't
tell cuz netstat -m on 2.0R doesn't report MBUF clusters.

What happens on my -current system is that the MBUF Clusters field
fills up (reports 256/256 clusters in use, 512K in use by network).
However, vmstat -m, doesn't reflect this.  And the mbuf pool seems
in a reasonable state.  I.E. other drivers relying on just mbufs
work fine.

However, the ed driver stops working because in -current it fails if
it can't get a cluster, while in 2.0R it falls back to using
individual mbufs.

Looking at the mbuf allocation code, I don't immediately see where the
accounting for mbuf clusters is done.

Also, I get the message 'mb_map full' on the console each time the
mbuf cluster field fills up.

Lastly I can beat the device driver to death for days under 2.0R and
the mbuf pool remains stable.

Any ideas?

Thanks,

Brian Litzinger
brian@easynet.com