>My original suggestion was to have a kernel API, but the persistent data should not be stored
>inside the kernel but on a file.
>
>I guess insmod/rmmod can do it too, to a point, if a format way is established
>by which insmod knowd where the parameters are stored (and makes it available
>to the module init) and the module lets rmmod know where the outgoing data
>is. Maybe just another well-known symbol like the init/cleanup.
>
>It is importanat that not only detection info is saved but also user settable
>modes (from ioctl).
This sounds awfully similar to the 'registry' of Microsoft Windows 95
and Windows NT. As far as I can tell, they have designed this quite
well, and for a similar purpose, so why not steal part or all of their
design? Their particular organization in the hierarchy looks a bit
weird to me, but that doesn't mean the basic idea isn't sound.
In particular, their hierarchy of access keys would fit quite well in
the wosisname filing system of Linux that contains all this kernel
information. (Sorry, the name escapes me for the moment.)
Most info on the registry can probably be found on the web. By request
I can do some digging there myself.
-- Kees van Reeuwijk Delft University of Technology, Faculty of Applied Physics, FI-CP email: reeuwijk@cp.tn.tudelft.nl URL: http://www.cp.tn.tudelft.nl/C.vanReeuwijk/home.html