Jetpack 4.2.1 breaks USB modem (driver issue?)

I feel doubtful that new firmware is being uploaded to the device… on a working computer the same thing happens every time, where it starts up in flash mode, resets, and reboots in normal mode. I feel like if firmware were being uploaded, that would only happen once, and not every time you plug it in. I think that the device just starts up automatically in a mode for flashing new firmware, and restarts into normal mode if no action is taken (i.e. no attempt to upload firmware).

I’m going to try one more thing before just switching back to 3.3 for now. The modem is actually an mPCI-e form factor, and just uses the USB lanes in the mPCI-e slot. I have a USB mMPC-e adapter that I’m going to plug the card into, and plug that directly into a USB port on the TX2. Who knows, maybe its something related to the mPCI-e slot. Gotta make a wire though… I’m using an Elroy carrier.

If that doesn’t working, I’m switching back to Jetpack 3.3 so I can at least continue working on my project until I can get an answer from u-blox about why the modem wouldn’t reset into normal mode.

Will keep you posted.

Two significant breakthroughs!

  1. I figured out what makes the modem transition from flashing mode to normal mode. When the modem starts up in flashing mode, if cdc_acm binds to it, the modem will shut off and start back up in normal mode.

  2. I figured out that the issue is in the mPCI-e USB2.0 configuration. I made a USB cable to bypass the mPCI-e connection and plug into a different USB port, and the modem works exactly as expected in Jetpack 4.2.1.

So I now know this is not a driver issue related to the modem… there’s something wrong with the configuration/drivers related to the mPCI-e usb 2.0 lanes.

Looking at ConnectTech’s block diagram for the carrier board, the mPCI-e slot I’m using passes USB2.0 lanes directly to USB port 2 on the TX2.

So now the question becomes, who broke it between updates? Did NVIDIA change something related to USB port 2? Or did CTI change something with their BSP as it relates to mPCI-e or USB port 2?

I’ll check with CTI, but if anybody has any insight as to if JetPack 4.2.x made changes to USP port 2’s functionality, that would be super helpful.

EDIT: Opened up a new post here to troubleshoot USB2.0 Port 2. Jetson 4.2+ possibly breaks USB2.0 Port 2 - Jetson TX2 - NVIDIA Developer Forums

Different carrier boards use different device trees. The answer may be something as simple as a minor device tree edit. On the other hand, perhaps it was something else which is hard to detect, e.g., a reverse polarity on the D+/D-…don’t know, have not seen that carrier board before.

Someone from NVIDIA or CTI may have run into issues like this before and be able to comment about the mPCIe USB.

The issue should be clarified:
[url]Jetson 4.2+ possibly breaks USB2.0 Port 2 - Jetson TX2 - NVIDIA Developer Forums