From: Matthew W. <wi...@de...> - 2003-03-24 12:19:55
|
On Mon, Mar 24, 2003 at 02:58:06PM +0900, Takayoshi Kochi wrote: > > So.. how to fix? Well, it depends. Right now we have a half-arsed belief > > that PCI root bridges can go away. Either we need to stop believing it > > or believe it properly. > > Yes, they can. OK. In that case the existing code was broken as there was no way for the acpiphp driver to be informed when the root bridge was removed. > > If we believe it, we need to have some mechanism for having two drivers > > for one device [1]. This doesn't need to be a generic mechanism; it's > > fairly clear that acpiphp is a `subdriver' of pci_root and will always > > be initialised after it. > > For 2.4, there is no acpi_bus_register_driver in the stock kernel > and we have luckily avoided the problem... Well, yeah. But I need the GPE Block support in the latest version of ACPI, so I have to upgrade my ACPI CA. > > I'm currently favouring an acpi_pci_{,un}register_driver() interface with > > add and remove callbacks, since it's the least amount of code change > > without disbelieving pci root bridge removal. > > I have no objection for going for it. OK. Will implement today. -- "It's not Hollywood. War is real, war is primarily not about defeat or victory, it is about death. I've seen thousands and thousands of dead bodies. Do you think I want to have an academic debate on this subject?" -- Robert Fisk |