Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 09 Mar 1999 11:16:28 -0500
From:      Dennis <dennis@etinc.com>
To:        "Randy A. Katz" <randyk@ccsales.com>, =?iso-8859-1?Q?=22Ronald_Wiplinger_=28=C3Q=A4=AF=AF=C7=29=22?= <ronald@trace.net.tw>, "freebsd-isp@FreeBSD.ORG" <freebsd-isp@FreeBSD.ORG>
Subject:   Re: Ethernet card problem
Message-ID:  <199903091607.LAA25321@etinc.com>
In-Reply-To: <3.0.5.32.19990307232217.05073d50@ccsales.com>
References:  <36E37494.14798FFC@trace.net.tw>

next in thread | previous in thread | raw e-mail | index | archive | help
At 11:22 PM 3/7/99 -0800, Randy A. Katz wrote:
>Hi Ronald,
>
>If the problem is with the Ethernet adapter then having two fxp's will be
>no problem. Assuming the fxp's are PCI based you will see fxp0 and fxp1 and
>will assign each an IP address and use them accordingly...I do this will a
>bunch of machines currently.
>
>If you are using the newer version of the driver you should investigate
>limiting the bandwith on the serial interface (eth0) as the later versions
>do not support limiting on devices other then the ETINC adapters...

This is not true at all. 90% of our customers using the ETBWMGR dont
have serial cards, so this certainly isn't the case.=20

Randy's analysis is likely due to some old problems...and also because
of the nature of his traffic. Typically, it is much more efficient to limit
on the exiting interface...since he runs web sites the exiting interface is
the T1/T3 card. For dual ethernet customers, you should set your limits on
the ethernet interface that is on the internet router side of the connection
So, for example if  you have fxp0 on the lan with your web servicers and=
 fxp1
on the lan with your T1 router, the limits for the web sites should be set=
 on
fxp1.

There was a bug that could cause overloads on "IN" limits that has been=
 fixed
in the most recent version.

The problem that Ronald is seeing is not related to the BWMGR, but to
a DMA problem on the card. "transmit underruns" (or underflows for it thats
what the driver author called it") are typically caused when the controller
cant
get data fast enough (typically bus masters not able to get the bus in time=
 to
fill the next data slot in a frame that is in the process of being
transmitted).=20

Im not sure what card ronald is using, but if it is an isa bus-master card=
 you
may just be overloading the bus. PCI should solve the problem

Dennis
>
>
>At 02:56 PM 3/8/99 +0800, Ronald Wiplinger (=C3Q=A4=AF=AF=C7) wrote:
>>It seems that I get a problem with my Ethernet card. I get following
>>message on the screen:
>>
>>
>>
>>de0: abnormal interrupt: transmit underflow (raising X threshold to
>>96|256)
>>de0: abnormal interrupt: transmit underflow (raising X threshold to
>>128|512)
>>de0: abnormal interrupt: transmit underflow (raising X threshold to
>>160|1024)
>>
>>
>>What does it mean, and how can I solve it?
>>
>>
>>This error comes up, since I have installed the bandwidth filter of
>>ETINC. The system crashs frequently after 2 hours of operation. I
>>believe it is related to the above message.
>>
>>
>>Suppose I have to change this Ethernet card (D-link) to a Intel Ethernet
>>Express 100. How can I have two fxp* cards in one machine and still know
>>which one is for which gateway ????
>>
>>
>>bye
>>
>>Ronald Wiplinger (=C3Q=A4=AF=AF=C7)  http://www.trace.net.tw  (phone=
 number =3D
>>e-mail)
>>e-mail: 0935869459@phonebook.com.tw or ronald@trace.net.tw
>>Attachment Converted: "d:\eudorapgp\attach\vcard353.vcf"
>>
>
>
>To Unsubscribe: send mail to majordomo@FreeBSD.org
>with "unsubscribe freebsd-isp" in the body of the message
>=20


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-isp" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199903091607.LAA25321>