Hi Sergei,

   CCing DaveM (as you should have done from the start)...

Thank you. I appreciate your help.

   Did you also build mdio-bitbang.c as a module?

Yes. Sure.

For the in-kernal driver, not being able to rmmod the 'ravb' one sounds
logical. :-)

root@rcar-gen3:~# lsmod|grep ravb
ravb                   40960  1
mdio_bitbang           16384  1 ravb
root@rcar-gen3:~# modprobe -r ravb
modprobe: FATAL: Module ravb is in use.
root@rcar-gen3:~# modprobe -r mdio_bitbang
modprobe: FATAL: Module mdio_bitbang is in use.

quoted

Fixed to execute mdio_init() at open and free_mdio() at close, thereby rmmod is

   Call ravb_mdio_init() at open and free_mdio_bitbang() at close.

OK. 
Include the exact function name in the commit log, not the abbreviated function name.

  Dave, would you tolerate the forward declarations here instead (to avoid the function moves, to be later
done in the net-next tree)?

Wait for Dave's reply for a while.
(If Dave's reply is slow, I will only correct Sergei's issue and post it)

Thanks & Best Regards,
Yuusuke Ashizuka [off-list ref]