I was upgrading from 10.2.0.5 to 126.96.36.199
and I faced the problem repeatedly
during running of catupgrd.sql which I ran manually
it was hanging
I figured out that the problem was the flashback recovery area which was set in my instance to 2 GB
during the startup upgrade you cannot release the space of flashback recovery area which is used for online archivelogs by executing RMAN commands
I resolved this problem by setting the flashback recovery area to 10GB
The process of catupgrd.sql took 6 GB of space
and once this was done I shutdown the instance and started up the instance again normally and now I could connect with RMAN and backup the archivelogs delete input - so the space used by the flashback recovery area was released.
subesquently I ran the utlrp.sql script and this also consumed a lot of flashback recovery area space which I could release using RMAN commands - I had to regularly monitor the space usage and periodically run Rman commands to backup the archive logs and release the space.