Hello,
Thank you for your response.
No, I don't use flow control.
I have made progress on understanding and solving the problem.
My communication has a lot of parity error but also the reception of break error.
Here is the result of the ioctl command with the parameter "TIOCGICOUNT".

I think µP has a problem under these particular conditions.
Here's the solution I've implemented is one that works. When the number of errors reaches a threshold I do a software reset using the "UCR2_SRST" bit. In the "Part-of_imx.c" file the part of code added.
I have proof that it takes the µP out of this permanent FF sending state in the file "TheProof.txt"
do you think this is the right solution ?
Jean-françois