Jayjwa, Aside from the kref_init issue, another difference between 2.6.27.x and 2.6.28.x is that the /dev/536ep0 device does NOT get deleted on device rmmod with the 536 driver source code you loaded. Please reload the latest version from http://vouters.dyndns.org:8080/Intel/ For the other character oriented drivers, one of the problems comes in if TTY_DRIVER_DYNAMIC_DEV is set in the tty struct flags AND the last argument to tty_register_device is set to NULL. It is better to comment out the TTY_DRIVER_DYNAMIC_DEV flag as well as the associated tty_register_device call. At module __exit time and even if the TTY_DRIVER_DYNAMIC_DEV flag is unset, one MUST call tty_unregister_device. In conclusion, tty handling in 2.6.28.x does not seem to match the documentation. Best regards to you, - http://vouters.dyndns.org:8080/ Philippe Vouters (Fontainebleau/France) Le dimanche 08 février 2009 à 11:35 -0500, jayjwa a écrit : > > > > > Seems OK, thanks again. > > There's massive differences between the codebase I'd had and the above. What's > odd is that they both work up to 2.6.27.x while only yours above works beyond, > even with the kref_init changes added. There must be other differences between > the 2.6.27.x and 2.6.28.x > > As a side note, I noticed the 2.6.27.x is still continuing on with 2.6.27.15 > while 2.6.28.x is at 2.6.28.4. Both were add on Feb. 6th according to the date > on the kernel.org ftp server. Maybe people are having troubles with 2.6.28.x > beyond the kref_init changes... > > >
