Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.sys.apple2.programmer > #2782 > unrolled thread

Uthernet and half-duplex/full-duplex negotiation

Back to article view | Back to comp.sys.apple2.programmer

Contents

  Uthernet and half-duplex/full-duplex negotiation D Finnigan <dog_cow@macgui.com> - 2016-06-11 16:04 +0000
    Re: Uthernet and half-duplex/full-duplex negotiation Tom Greene <tom@cowgod.org> - 2016-06-17 08:24 -0700
      Re: Uthernet and half-duplex/full-duplex negotiation D Finnigan <dog_cow@macgui.com> - 2016-06-17 21:28 +0000

#2782 — Uthernet and half-duplex/full-duplex negotiation

Dear all,

I'm once again improving and testing the Marina IP stack for a June 2016
edition. I've set up two Apple II computers which are connected with a
straight-through Ethernet cable. One is running the original Uthernet, the
other has the new Uthernet II. I notice that the Uthernet II does not enter
into full-duplex mode, and further more, that it seems to receive its own
transmissions. Or at least there is some hang-up which causes Marina to get
stuck in the ARP Address Conflict Detection Probe loop until I pull the
cable and finally it settles into normal operation. This infinite loop does
not seem to happen on the Uthernet machine, but it ends up receiving the
flood of ARP probes.

This problem also does not happen at all when both machines are running an
Uthernet II. Both cards enter into 100 Mb/s, full duplex mode.

I've checked the manufacturer's documentation for both cards and it's not
very clear on how to change from full to half-duplex. Indeed, the W5100
manual includes basically no documentation on full vs half duplex.

Has anyone run into this problem and can provide a solution or something to
check for in my code?

Or maybe I just need to use an actual cross-over cable? I don't think I have
one with me at the moment, otherwise I would have tried it before posting
here.

-- 
]DF$
The Marina IP stack for Apple II--
http://marina.a2hq.com/

[toc] | [next] | [standalone]

#2792

On Saturday, June 11, 2016 at 12:04:21 PM UTC-4, D Finnigan wrote:
> Dear all,
> 
> I'm once again improving and testing the Marina IP stack for a June 2016
> edition. I've set up two Apple II computers which are connected with a
> straight-through Ethernet cable. One is running the original Uthernet, the
> other has the new Uthernet II. I notice that the Uthernet II does not enter
> into full-duplex mode, and further more, that it seems to receive its own
> transmissions. Or at least there is some hang-up which causes Marina to get
> stuck in the ARP Address Conflict Detection Probe loop until I pull the
> cable and finally it settles into normal operation. This infinite loop does
> not seem to happen on the Uthernet machine, but it ends up receiving the
> flood of ARP probes.
> 
> This problem also does not happen at all when both machines are running an
> Uthernet II. Both cards enter into 100 Mb/s, full duplex mode.
> 
> I've checked the manufacturer's documentation for both cards and it's not
> very clear on how to change from full to half-duplex. Indeed, the W5100
> manual includes basically no documentation on full vs half duplex.
> 
> Has anyone run into this problem and can provide a solution or something to
> check for in my code?
> 
> Or maybe I just need to use an actual cross-over cable? I don't think I have
> one with me at the moment, otherwise I would have tried it before posting
> here.
> 
> -- 
> ]DF$
> The Marina IP stack for Apple II--
> http://marina.a2hq.com/

Hmmm, the W5100 data sheet describes how speed and duplex can be selected by configuring a set of input pins. It doesn't appear that it is possible to change it in software. The Uthernet II manual doesn't seem to indicate there are any switches or jumpers on the card, so it is probably not configurable short of modifying the board. 

The CS8900A on the original Uthernet looks like it defaults to half duplex and can be configured for full duplex operation, but it does not auto-negotiate. Since the Uthernet II doesn't seem to be configurable, setting full-duplex on the Uthernet would not be helpful since it would be impossible to match the setting on the other end.

So, the end result should be that the link between the two cards always comes up as 10Mbit half-duplex, but that shouldn't be a problem in itself. 

Tom

[toc] | [prev] | [next] | [standalone]

#2794

Tom Greene wrote:
> 
> The CS8900A on the original Uthernet looks like it defaults to half duplex
> and can be configured for full duplex operation, but it does not
> auto-negotiate. 

The manufacturer's documentation leads me to believe that the CS8900A does
auto-negotiate.


> So, the end result should be that the link between the two cards always
> comes up as 10Mbit half-duplex, but that shouldn't be a problem in itself.
>

Exactly. I'm trying to figure out whether this is a software problem on my
end, or something to do with the hardware.

-- 
]DF$
The Marina IP stack for Apple II--
http://marina.a2hq.com/

[toc] | [prev] | [standalone]

Back to top | Article view | comp.sys.apple2.programmer

csiph-web