[470] in SIPB_Linux_Development

home help back first fref pref prev next nref lref last post

[becker@super.org (Donald J. Becker): Re: AT-1700 timeouts]

daemon@ATHENA.MIT.EDU (Erik Nygren)
Thu Feb 24 17:09:04 1994

To: resnet-help@MIT.EDU, linux-dev@MIT.EDU, escher@MIT.EDU
Cc: linux-help@MIT.EDU
Date: Thu, 24 Feb 94 17:08:40 EST
From: Erik Nygren <nygren@MIT.EDU>


Here's an interesting reply from Donald Becker, the author of the Linux
AT1700 and 3C509 drivers.  He's responding to questions regarding the transmit
timeouts on the 3c509 and at1700 under Linux.  (The thing about
the Fujitsu chip is with regards to the at1700 hanging.)

Most importantly, he mentions a new at-1700 driver that just came out
in the latest kernel (pl15i).  I haven't looked at this yet, but
someone with an at1700 may want to try.  Tell me if you have any luck.

    --- Erik

------- Forwarded Message

Received: from MIT.MIT.EDU by po7.MIT.EDU (5.61/4.7) id AA28082; Thu, 24 Feb 94 16:47:21 EST
Received: from FOUNDATION.MIT.EDU by MIT.EDU with SMTP
	id AA23396; Thu, 24 Feb 94 16:47:11 EST
Received:  
	from super by foundation  
		 with smtp 
		
	(Linux Smail3.1.28.1 #14)  
	id m0pZntz-000045C; Thu, 24 Feb 94 16:47 EST
Received: from descartes.super.org by super (4.1/SMI-4.1)
	id AA07207; Thu, 24 Feb 94 16:47:02 EST
Date: Thu, 24 Feb 94 16:47:02 EST
From: becker@super.org (Donald J. Becker)
Message-Id: <9402242147.AA07207@super>
To: nygren@foundation.mit.edu
Subject: Re:  AT-1700 timeouts

>Maybe the problem is at a higher level in the netowrking code or in
>our physical network.  Do you have any idea what in the physical netowrk
>could cause problems like these on both cards?  What should be looked at
>here?

Both problems are the result of invalid data on the ethernet wire.  I have
no easy way of generating errors on my networks, so the error recovery code
is rarely tested.

>By the way, a soft reset causes the kernel not to detect the card at all.
>Only a hard reset gets it to work.

The reset problem is a bug in the Fujitsu ethernet chip -- you must power
cycle the device to fully reset it.

Both the AT1700 probe code and the error recovery has been updated in the
latest alpha kernels.  Let me know if the AT1700 error recovery works now --
I'm not confident the Fujitsu datasheet is correct about the procedure.

a
Donald Becker				       becker@super.org
 IDA Supercomputing Research Center
 17100 Science Drive, Bowie MD 20715		   301-805-7482

------- End of Forwarded Message


home help back first fref pref prev next nref lref last post