Correction: Earlier I used initials “CTS/DTS”, it’s actually “CTS/RTS” (instructions are still correct, but it stands for “Clear To Send” and “Request To Send”).
Keep in mind that the 9-pin D-sub connector has serial console bound to it, and I am assuming you want this for serial console. If you’ve mixed other use in with this you’d have to first disable serial console. However, I’m continuing on the the idea that you are intending serial console, the above settings should work.
If the serial connection settings at both sides of the wire do not match you could experience issues similar to what you have. The fact that you get data transfer in one direction says the 115200 8N1 is probably correct on both sides of the wire. My thoughts on flow control are because I use CTS/RTS where possible, and in fact I’ve used this with a couple of JTK1s successfully for a very long time using every L4T version. This is a form of hardware flow control and adds reliability up through much faster speeds than 115200…without this 115200 is probably the limit before starting to worry about quirks.
My thoughts on CTS/RTS versus software flow control are because I have experienced exactly the issue you have (on different hardware running L4T) when CTS/RTS was attempted, but had the problem go away under software flow control. Can you look at the 9-pin D-sub connector side and check if the pins are all in place and not bent or missing?