From sjohn@alamak.bchem.washington.edu  Mon Dec 11 21:46:35 1995
Received: from alamak.bchem.washington.edu (alamak.bchem.washington.edu [128.95.12.209])
          by freefall.freebsd.org (8.7.3/8.7.3) with SMTP id VAA08777
          for <FreeBSD-gnats-submit@freebsd.org>; Mon, 11 Dec 1995 21:46:35 -0800 (PST)
Received: (from sjohn@localhost) by alamak.bchem.washington.edu (8.6.12/8.6.12) id VAA26263; Mon, 11 Dec 1995 21:48:48 -0800
Message-Id: <199512120548.VAA26263@alamak.bchem.washington.edu>
Date: Mon, 11 Dec 1995 21:48:48 -0800
From: sjohn@alamak.bchem.washington.edu
Reply-To: sjohn@alamak.bchem.washington.edu
To: FreeBSD-gnats-submit@freebsd.org
Subject: Ethernet, DEC chipset, mb_map error, grabs memory, then stops working.
X-Send-Pr-Version: 3.2

>Number:         890
>Category:       kern
>Synopsis:       Kingston card de0 driver grabs memory error mb_map full
>Confidential:   no
>Severity:       serious
>Priority:       high
>Responsible:    freebsd-bugs
>State:          closed
>Quarter:
>Keywords:
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Mon Dec 11 21:50:02 PST 1995
>Closed-Date:    Mon Dec 11 23:09:40 PST 1995
>Last-Modified:  Mon Dec 11 23:12:15 PST 1995
>Originator:     John D. Saario
>Release:        FreeBSD 2.1-STABLE i386
>Organization:
>Environment:

Kingston Ethernet Card....
FreeBSD 2.1 on a P133 with 64MB of memory 77MB swap space. Adaptec SCSI 
controler with Conner 1.0GB drive. High volume web server running Apache!

>Description:

I bought a Kingston card with the DEC DC21040 chipset since it said this
chipset was supported. I used the de0 driver and it worked fine. At first
I forgot to turn off PlugNPlay and the network stopped working after 6
hours with a mb_map full messsge to the console. Then I noticed there
seemed to be about twice as much Active memory used as with the 3C503 card
I used before and also the Cache memory as displayed by top was 20M.... I
rebooted and everything worked. Later I disabled plug and play, it didnt'
crash but is still using up all that memory and slowly grabbing more. It
didn't seem to do much at first but after an hour of heavy use it seemed
to shoot up in a few minutes when I was not watching top. Then it
stabilzed for awhile. 

>How-To-Repeat:

Running a high volume web server with a Web Chat page....many hits a second.

>Fix:

I don't know but I may try a SMC card?

John
>Release-Note:
>Audit-Trail:
State-Changed-From-To: open->closed 
State-Changed-By: davidg 
State-Changed-When: Mon Dec 11 23:09:40 PST 1995 
State-Changed-Why:  
The system caches all file data. That's what "cached" pages are. 
This is not a bug and has nothing to do with the network card. 
The mb_map full is caused by not configuring the kernel correctly 
for WWW/HTTP loads. 
Add: 
options "NMBCLUSTERS=2048" 
to your kernel config file to fix this problem. 
>Unformatted:
