Hi Joe, Thanks for your suggestions and your delicate test model! Follow your suggestion,I connect the system in a simple way,well, I have installed wireshark and change the MTU to 9000. But it is still no data comes. I forgot one thing to tell you that I have a model modified the PAPER's model for our use.This model has 4096+8=4104 bytes length per packet.Well this bof file can go send data through the network. Using the wireshark ,I can easily catch the packet send from roach board.Well for a packet of 8192+8=8200 bytes from whatever PAPER model or antother model I designed, There always been no data out from Roach v2 .It's for the same reason --overflow. As the biggest frame of 10gb block is 8K+512,so the size of the 8200 byetes packet is ok theoretically. Your ethernet_test model is very useful and delicate,I plan to modified a little to check whether it is caused by the frequency problem or the Tx buffer size. Thanks for your help! Have a good night/day! peter
At 2014-10-20 23:13:11, "Kujawski, Joseph" <jkujaw...@siena.edu> wrote: Peter, I am not sure if you are having the same problem that I was having, so I will give you some pointers on the debugging process and will be happy to help as you refine your debugging approach. 1) One of the most difficult issues I had initially was determining whether or not the ROACH board was even communicating. I found that simplifying the network to a direct connection between a computer and the Roach board helped me isolate issues related to the firmware and issues related the Ethernet framework. 2) If you are not already using one, you will need a network analyzer tool. I used 'wireshark' (https://www.wireshark.org/). This allowed me to determine if the ROACH board ever even broadcast packets. 2a) I found it useful to shut down most of the ethernet traffic into and out of the computer. This made it much easier to see the ROACH generated traffic. 3) While in theory, the ethernet protocol will support very large packet sizes, most infrastructure implementations actually limits the size to 1500 bytes of so. If you try to send larger packets over a backbone with these limitations, these packets will be ignored at the first switch. 4) The 10GBe block is a simplified version of the 10GBe protocol and *only* supports 10GBe. Specifically, it is not backwards compatible to the older 1GBe and 100MB versions of the interface. 4a) I had in infrastructure with a 1GBe switch that prevented data packets generated by my ROACH board from being forwarded to the receiving computer. 5) There are a number of setup issues on the 10GBe block that also need to be configured properly. I recommend that you develop a very simple model that only exercises the 10GBe port and have included my initial simplified Ethernet communications model for your use. Please try the following: A) Determine if your existing model sends anything to the network port by implementing 1) and 2) above. Report the results. B) If nothing is being sent to the port, then implement a simplified Ethernet port (or use the one I am sending you). Report the results. Good luck with everything. -Joe Kujawski -- ************************************** * Joe Kujawski * Siena College * Dept. of Physics and Astronomy, RB 113 * 515 Loudon Road * Loudonville, NY 12211-1462 * * Email: jkujaw...@siena.edu * Phone: 518-867-7509 <-- NEW NUMBER * Fax: 518-783-2986 **************************************