On Tue, May 01, 2001 at 05:07:08PM -0700, Chris G. Demetriou wrote:
	> It would seem to be a mistake to do something which:
	> 
	> 	* involved significant code size increase, or
	> 	* involved significant data size increase
	> 
	> to accomodate something which happens almost never.

	The thing is.. the table is small-- how many pci devices are in the
	system, anyways?  and you -can't- do hotplug right without this kind of
	functionality.  e.g. you could use an exceptionall small linked list for
	this purpose, e.g. <20 bytes per device.

Shouldn't that information be available in the device tree?

	It also assumes that rescanning things after beginning to use them is
	safe-- which might not necessarily be a safe assumption with broken
	hardware.

Huh?  If you get a hotplug event, don't you need to reprobe the bus
to find out what the new devices are?

Eduardo