DragonFly kernel List (threaded) for 2004-03
Re: Kernel Panic on boot
On Sat, Mar 20, 2004 at 01:19:15PM -0800, Matthew Dillon wrote:
> :There are three different "solutions" in order of increasing complexity:
> :- add a local hack wire the interrupt link for cardbus and add the necessary
> : chipset specific part for activation. I'm doing this right now.
> :- include the hooks for ACPI interrupt routing. This depends on a working
> : ACPI implementation instead and should work if you have to use ACPI on
> : FreeBSD 5 too.
> :- port the pcib interrupt code from NetBSD/OpenBSD which allows the kernel
> : to handle it w/ PCI BIOS at all.
> :The first point is some kind of scaled down version of the last point :)
> #3 sounds like it would be the most maintainable. I don't think we want
> to create a case where we have to completely depend on ACPI, if possible,
> and I would really love to be able to scrap PCI-BIOS.
I completely agree on this, esp. since the framework would be identical on
all PCI architectures, not only IA32. There are still some dependencies on
the PCI BIOS, since we need to get the PCIR table somewhere, that's the
mapping of interrupt pins to slots/devices.
> Could you point me at the NetBSD/OpenBSD sources involved? I can take a
> quick look and estimate the effort required.
It's in arch/i386/pci, pci_intr_fixup.c and chipset specific parts in the
same directory. I need to give the handling of PCI bridges some thinking
and start integrating that code and the IO code into our PCI tree as optional
component. But it has to wait a bit for my exams :)
> Matthew Dillon