Changing the shell image after OpenNIC shell configuration on the U45N

[FilippoCarloni]

Hi everyone,

I am using the OpenNIC shell on the U45N board, but I am encountering an issue when changing the shell back to the golden image once the OpenNIC one is configured.

I configured the shell via PCI using the Xbflash2 tool with the default AXI Quad SPI address 0x50000 since I do not have the JTAG cable.

However, when trying to revert to the golden MCS configuration with the Xbflash2 option --revert-to-golden, using the same AXI Quad SPI address gives me a segmentation fault error.

Is the SPI address in the OpenNIC shell the same as the golden MCS image or should I use a different address?

Thank you for the help!

[cneely-amd]

Hi @FilippoCarloni ,

I'm going to respond in slightly round about way, mentioning some background and then points you towards a couple of spots in the shell design.

Step by step instructions for instantiating the Card Management IP and QSPI flash programming are posted online here, and these instructions are what I followed when adding both CMS and QSPI to the shell. However because of the OpenNIC's exisiting address map, these IP were instantiated at other addresses in the shell (described further below):
https://adaptivesupport.amd.com/s/article/Alveo-Updating-Vivado-images-via-PCIe-with-xbflash?language=en_US

The system config part of OpenNIC shell contains a top level address decoder and an AXI crossbar IP instance. Please see the comments section at the top of this file and also within for some address map details.
https://github.com/Xilinx/open-nic-shell/blob/main/src/system_config/system_config_address_map.sv

However, there is a second place where the addresses are specified and that is within parameters for the AXI crossbar instance called system_config_axi_crossbar, for example:

I'm mentioning these three parts in case you decide to make any edits based on your requirements.

Best regards,
--Chris

[FilippoCarloni]

Hi @cneely-amd,

Thank you for the answer; those addresses are exactly what I was looking for!

However, I tried to reset the FPGA to the golden image using the same guide (https://adaptivesupport.amd.com/s/article/Alveo-Updating-Vivado-images-via-PCIe-with-xbflash?language=en_US) but I had two different situations.

By using sudo xbflash2 program --spi --revert-to-golden -d 65:00.0 --bar-offset 0x340000, I got a "segmentation fault" error; while using xbflash2 program --spi --image ../u45n_mfg.mcs --bar-offset 0x340000 -d 65:00.0 the execution stucks at this point without moving ahead:

flashing via QSPI controller located at 0x340000 on BAR0 INFO: ***Found 298 ELA Records Preparing flash chip 0

The 65:00.0 is the PCI device (my case has both 65:00.0 and 65:00.1, but the procedure output is the same), while the 0x340000 should be the QSPI address. I followed the base example on the Readme to reproduce the U45N shell generation with two CMAC port.

Can you help me?
Let me know if you need further information,
Thank you for your time!
Best regards,

Filippo Carloni