> From: Matthias Drochner <drochner@zel459.zel.kfa-juelich.de>
>
> I was mildly annoyed by the fact that a normal kernel didn't
> get some devices attached right on a ThinkPad I got between
> my fingers. (The first serial line -the infrared one- used irq4
> but should use irq10, and the soundblaster compatible audio
> didn't find the right dma channels...)
>:::
 [ pnpbios callback subsystem ]
>:::
> What do you think - is this the way to go?

That's neat, but it's a little early.  To some extent, you are fixing an
existing problem by adding a new feature instead of revising the broken
existing approach.

There is really one problem that explains most of our isapnp lossage...

We quickly reset all the pnp devices' configurations.  This isn't right at
all, lots of useful information is lost that way, and the non-volatile
config information and resource allocation in the bios is just thrown away.
One of the reasons is because there might not be a pnp bios or even a bios
at all, but in that case the configurations will still be reset from startup,
so it's actually OK.

We also do this because we think we don't want the pnp devices to be attached
in the isa probe, but actually that's OK too, if we just check in the isapnp
probe for each device whether someone else configured it, and if we use
the bus_space I/O port extents to prevent reattachment.

So, all the information you want is in the config of the devices, but at
the moment, we toss it all.

I have two or three high-priority alpha projects stacked up, but in a month
or so I could make a stab at this...feel free to beat me to it. :-)

	Ross.Harvey@Computer.Org