Using Lolo version 2.4.12

Connected controller with SOM to a non-managed ethernet switch (Netgear Model FS108). Create fatal loop by directly connecting two switch ports together with a patch cable. Power on controller and Lolo stops right after "Starting DHCP on sm0" instruction. 

Regards,

Brad

Unfortunately the development got lost in a office move many years ago, so I do not currently have it to test.

Today I was able to re-acquire a Logic i.MX31 Development kit and rerun the same tests on the development kit. The development kit had Lolo version 2.4.0, slightly earlier version than our production units (Lolo 2.4.12) although I had no problem in consistently reproducing the problem. 

1) Connected the development kit ethernet port to a powered non-managed switch (Netgear FS108).

2) Created a fatal loop on the ethernet switch by directly connecting two of the switches ports together with a patch cable.

3) Connected a serial connection to a PC running TeraTerm.

4) Powered on the developement kit and instantly exited the configuration script by pressing "Q".

5) At the Losh prompt entered "ifconfig sm0 dhcp".

6) As the echo "Starting DHCP on sm0.." is received, Lolo appears to freeze at this point and a high activity can be observed on the ethernet switch communication leds. 

7) The TeraTerm output below shows two tests that were ran. The first test is without the fatal loop connected. The first test responds as expected with a "error: ifconfig: unable to get ip address; still trying (sm0)" message.  For the second test, the fatal loop was connected prior to powering on the development kit. Once the losh command "ifconfig sm0 dhcp" was entered, the development kit remained at the "Starting DHCP on sm0 .." message for at least 5 minutes. My assumption is that the DHCP function sends out a broadcast message which gets trapped in a switch loop and causes a broadcast storm. This broadcast storm causes excessive interrupts that overload the CPU. Our main concern is not with Lolo but that this same condition happens during normal operation in the WinCE 5.0 operating system.

                         LogicLoader

 (c) Copyright 2002-2006, Logic Product Development, Inc.
 All Rights Reserved.
 Version 2.4.0-IMX31_10 0001
*****************************************************************

losh> ifconfig sm0 dhcp
Starting DHCP on sm0 ..........
error: ifconfig: unable to get ip address; still trying (sm0)
losh> }}
*****************************************************************

                         LogicLoader

 (c) Copyright 2002-2006, Logic Product Development, Inc.
 All Rights Reserved.
 Version 2.4.0-IMX31_10 0001
*****************************************************************

losh> ifconfig sm0 dhcp
Starting DHCP on sm0 ..

Gary,

I was able to verify that the i.MX31 SOM does become unresponsive due to the broadcast storm using the steps provided above while leaving the non-managed Netgear FS108 switch isolated from the network.  Broadcast storms was not consider as a test case during the development of Logic Loader or the WinCE Ethernet driver.   Both Logic Loader and the WinCE release have been considered maintenance releases for a while.  Only significant issues that affects many customers or updates required due to changes in hardware are being considered for future releases.  An recently released EOL notice shows the last time buy for all i.MX31 SOMs will be September 30, 2015.

Customers looking for improvements outside the scope of a maintenance release are encouraged to contract us for development assistance from our Product Development Services team.   Click here to submit your request.   Someone will contact you as soon as possible.

Regards,
Brad

Brad,

Thanks for your time and insight into this issue. Unfortunately this is not the resolution that we were looking for.  Our company has been producing a product with this SOM for many of years and we currently have a considerable installed base. We have become aware of this issue via problems that our customers have reported. The reliable operation of our machinery with utilizes this SOM is vital to our customers operation. The significance of this issue is certainly critical to those who have experienced it and the number affected could potentially be every application with an Ethernet connection. I certainly hope that your organization will consider the potential severity of this issue, and resolve this as a maintenance item.