EmbeddedRelated.com
Forums
Memfault Beyond the Launch

TCP sequence numbers.

Started by seec...@yahoo.com January 24, 2014
I'm using wireshark to try to understand what is happening with my rabbit module, and I have a question about the data wireshark is showing me. I tried the wireshark forum, but it seems to be down.

The info column displays the sequence number of the current packet, and the sequence number of the packet being acknowledged. I'm wondering how a packet can acknowledge a sequence number that hasn't been issued (sent) yet? At time stamp 3576.772512 sequence number 370 is acknowledged, and at time stamp 3576.81128 is when sequence number 370 is sent.

Steve
I don't think I'm interpreting the sequence numbers correctly, so my previous post is probably not valid.

What is happening though, is that about every 10-seconds the Rabbit is resetting it's connection with the client. The client then reconnects successfully, and normal communication continues for another 10-15 seconds, at which time the Rabbit abruptly resets the connection again. This continues for days until finally the client is not able to reconnect again. And the only way to be able to connect to the Rabbit is to move the Ethernet cable to another port on the switch.

The other observation I have is that there are a lot of Gratuitous ARPs being issued by the client. And according to the Wireshark wiki, that can be an indication of faulty cabling or hardware.

I'm suspecting the switch. We saw something like this a while back with a customer using a Dell switch. Over time the Rabbit would eventually just lose its connection. If we put a Netgear switch between the Dell and the Rabbit, the problem went away. We have never seen this phenomena with Netgear, or D-Link switches, which are what we use.

If someone can think of something I can look at or check to get the bottom of this, I'd appreciate it.

Steve

Steve
Hi Steve,

I had a similar problem last year.

Try to turn off automatic crossover ("auto MDI-X" or "auto-uplink") of the Rabbit ethernet port (you can try with the switch ports also, if there is such option at all).

Turn off auto negotiation too. Make sure to configure the right transmission parameters, such as speed, duplex mode, and flow control.

Wish you success with this annoying eth problem,
Alexander Hristov
I assume to mean to turn off auto-negotiation and auto-crossover in the switch, since I don't see that
option in the DC libraries.

Why do think these modes could be causing the problem? Are you thinking the switch is changing the
crossover configuration during communication, or losing the negotiated comm settings?

Steve

On Sunday, January 26, 2014 11:57 PM, "s...@tecosys.info" wrote:

 
>Hi Steve,
>I had a similar problem last year.
>
>Try to turn off automatic crossover ("auto MDI-X" or "auto-uplink") of the
Rabbit ethernet port (you can try with the switch ports also, if there
is such option at all).
>Turn off auto negotiation too. Make sure to configure the right transmission parameters, such as speed, duplex mode, and flow control.
>Wish you success with this annoying eth problem,
>Alexander Hristov
>
>
>
Hello,
You meant “Turning off automatic crossover “ the port of the Rabbit board?, if so, could you show me how. Thanks

Henry Huynh

From: r... [mailto:r...] On Behalf Of s...@tecosys.info
Sent: Monday, January 27, 2014 2:57 AM
To: r...
Subject: [rabbit-semi] RE: TCP sequence numbers.

Hi Steve,

I had a similar problem last year.

Try to turn off automatic crossover ("auto MDI-X" or "auto-uplink") of the Rabbit ethernet port (you can try with the switch ports also, if there is such option at all).

Turn off auto negotiation too. Make sure to configure the right transmission parameters, such as speed, duplex mode, and flow control.

Wish you success with this annoying eth problem,

Alexander Hristov
We are using the RCM3000 modules.  They use the Realtek Ethernet controller, which I thought was half-duplex. The customer turned off auto-negotiate and set the switch's port to 10mbit and tried both half duplex and full duplex. They report it didn't help.

I found out they are using a Cisco 3750x switch. I've asked if they could loan us one so we can try to duplicate what they are seeing.

In going through the Wireshark data, I don't see any abnormal packets. Just lots of Gratuitous ARPs from their system. Our device appears to be humming along just fine, and then abruptly resets the connection at random times. Very strange.

Steve

On Wednesday, January 29, 2014 7:13 AM, "Huynh, Henry" wrote:

 
>Hello,
>You meant “Turning off automatic crossover “ the port of the Rabbit board?, if so, could you show me how. Thanks

>Henry Huynh
>

>From:r... [mailto:r...] On Behalf Of s...@tecosys.info
>Sent: Monday, January 27, 2014 2:57 AM
>To: r...
>Subject: [rabbit-semi] RE: TCP sequence numbers.


>Hi Steve,

>I had a similar problem last year.
>Try to turn off automatic crossover ("auto MDI-X" or "auto-uplink") of the Rabbit ethernet port (you can try with the switch ports also, if there is such option at all).

>Turn off auto negotiation too. Make sure to configure the right transmission parameters, such as speed, duplex mode, and flow control.

>Wish you success with this annoying eth problem,
>Alexander Hristov


>

Memfault Beyond the Launch