We are attempting to update the firmwares on the IGX Developer’s Kit in order to flash the IGX OS onto it. We have been in contact with the Nvidia solutions team for the past few months but unfortunately have not been able to resolve this issue.
Description
Attempting to flash the IGX OS (iGPU and dGPU)
We attempted to flash the iGPU OS following the instructions in the user manual using the webUI following the firmware updates. After choosing the iGPU option from the GNU GRUB menu, on port 2200 the execution is seen. However, after waiting for 30-40 minutes and rebooting the IGX it opens a screen with(DP connected to the dGPU, have tried connecting to iGPU DP but nothing was observed on the monitor)
Ubuntu 22.04.4 LTS localhost tty1
Localhost login
What should be the next steps, as the guide does not show what screens or messages should be seen following the flash.
Have also attempted to flash the dGPU OS from the GNU GRUB menu following a similar process and the exact same screen is observed.
Can we still attempt to use the dGPU vbios tool if so how do we ssh into the system(i.e. Are the sudo commands intended to be ran via ssh into port 2200?)
Note: We were able to verify the version of the BMC and the SMCU using the commands found in the older guide before it got updated to web UI method. However, there was no method that indicated how to check the version of the QSPI flash.
Furthermore, we have attempted to try the QSPI firmware via the webUI but no indication of whether it was successful or not is shown. After clicking start update and waiting for 20-30 minutes there is no indication of if it was successful or failed. The solution teams mentioned it should only take approximately 6 minutes for it to finish updating. We have rebooted the system after waiting for 20-30 minutes but it falls back to the same spot. Feel free to ask for any other information or commands you would like us to run to help you.
Thank you for the help in advance!
Thanks for following up, I am attaching logs from the iGPU IGX OS ISO installation below as well as dGPU. Is this the correct information you are looking for?
Yes, the IGX Dev kit I have comes with the discrete RTX A6000 log_dgpu_update.txt (326.2 KB) log_igpu.txt (91.7 KB)
Could you share the result of cat /etc/os-release in your BMC console?
Please also refer to the following steps to update IGX SW through WebUI + Virtual Media
[WebUI]
Operations -> Virtual media -> Add file -> Select “.iso” -> Start
[Tegra Console]
$ lsusb
Bus 001 Device 007: ID 1d6b:0104 Linux Foundation Multifunction Composite Gadget
Reboot device (Operations -> Server power operations -> Reboot server -> Reboot)
Wait UEFI starts -> Press “ESC” to enter menu -> Boot Manager -> Select “UEFI OpenBMC Virtual Media Device” -> ENTER
You can run the following command from your host to access the serial console to check the update logs.
We followed this procedure: Wait UEFI starts → Press “ESC” to enter menu → Boot Manager → Select “UEFI OpenBMC Virtual Media Device” → ENTER
After waiting for ~30minutes the execution being shown on the port 2200 terminal window stopped but the screen remained blank. We rebooted the server and were led back to the same screen as highlighted in the original post. (DP is connected to the dGPU port, iGPU DP showed no signal)
For this previous attempt we installed the iGPU stack. Although in the past we have attempted both first iGPU then the dGPU following a similar procedure. In both cases it falls back to the screen mentioned in the original post.
Apart from the execution that was constantly being generated until after ~25 minutes when it stopped. No further message was shown. When we rebooted the system it simply goes to the screen mentioned above after the Nvidia splash screen is shown. When we ssh into port 2200 it simply shows the same window as what is shown on the monitor connected to the Nvidia IGX server.
If you installed it with iGPU stack, then you should connect DP port next to USB port.
If you installed it with dGPU stack, then your should connect the DP port on your external GPU card.
There should be many serial console logs output from port 2200 during flash the board.
Please check if the flashing process finished successfully.
