On Tue, May 30, 2000, Johan Verrept <[EMAIL PROTECTED]> wrote:
> Johannes Erdfelt wrote:
>
> > > Not that I've had occasion yet to look at that code, but why
> > > shouldn't the PCI code just directly modprobe the right driver?
> > >
> > > After all, the OHCI and UHCI (and EHCI :-) controllers will
> > > have different PCI identifiers.
> >
> > This would require a large table of devices/classes and their drivers in
> > the kernel to know what to do.
> >
> > This is probably best left in userspace.
> >
> > The new PCI code has something like this, but the driver must register
> > itself first. It's the old chicken and the egg problem.
>
> I think it would be usefull not only to keep a list of chickens, but also a
> list of unmatched eggs, so you can compare chicken and egg later ;-).
>
> translation:
> keep a list of drivers AND (unmatched?) devices. If either is added, match
> them to the other list... this should solve the chicken and egg problem. The
> tables won't get too large (not that many devices (and thus drivers) in a
> system anyway...). New drivers should, off course, only be matched with
> unmatched devices...
You need to have a list of all possible devices and the appropriate
drivers. Remember, you don't know what will be plugged in.
> This would solve the problem with the ohci adapter in the cardbus interface
> too. (A ohci adapter is compiled in, thus in the driver database, and when the
> new device is inserted, the driver is initialised.)
>
> This can be done in both userspace of kernelspace (although I also prefer the
> first). There will be a need for a way to list unmatched devices though. (for
> pci, usb and cardbus, I think you can get this info in /proc ...). The main
> problem would be to create a general device and general driver model that can
> be matches on all this types of devices and preferably on every other (future)
> type of device too.
>
> I'll think about this for a while.
Umm, isn't this essentially what I proposed? Or am I missing something?
JE
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
