On Tue, Aug 22, 2006 at 12:27:38PM +0200, Simon 'corecode' Schubert wrote: > On 21.08.2006, at 15:14, Joerg Sonnenberger wrote: > >>Well, I investigated doing this at some point, but it seemed quite > >>impossible at this time. We'd have to change the device handling, so > >>that a driver can hand out private devices for an open and not have > >>just one device per major/minor. > >Actually, no. It just should threat minor==0 or whatever special and > >replace the in-kernel file with an entry for the first free minor. That > >is possible already, through the mechanism for cloning in the tree > >sucks. > > could you show me a short snippet how this could be done? i know the > driver can find a free vchan, that's no problem, but it also needs to > return a different device than the one which was actually opened. this > was where i got stuck.
Check for lwp_dupfd. Joerg
