"Ronald G. Minnich" <rminnich(a)lanl.gov> writes:
On Tue, 10 Dec 2002, Kevin Hester wrote:
If folks think we can trust the link value fields to show interrupt sharing,
then all is well. I'm happy to make a new function that given an arbitrary
pirq table will make default IRQ assignments.
Well, it does seem to work. This is another area I've only looked at a
bit, to fix IRQ setup in Plan 9 and such.
I've been looking at the Linux arch/i386/kernel/pci-irq.c file and it
seems to be able to manage using PIRQ tables to set up IRQs, even handling
the weirdnesses of some chipsets. Have you checked this out? It's
educational. This is the file I was hoping to start with for IRQ
assignment in linuxbios. If there is a problem with my idea let me know.
Eric I don't get your comment about the "kernel PCI code not assigning
IRQs" in light of all the kernel code that does this. I think I
misunderstood you, can you explain.
There is an pci irq scratch register, that reports the BIOS's irq
assignments. The kernel does not and should not populate that
register in pci configuration space. But LinuxBIOS needs to populate
that register, so OS's that don't know about the pirq table can find
the irq assignments.