Bug? PCI link device _SRS

[John Baldwin]

Note that in the _PRT, the interrupt associated with a link device object has 
a source index indicating which resource holds the IRQ.  When we do _SRS to 
route an interrupt, we try to set all the resources of the device to a single 
IRQ resource.  Shouldn't we probably instead be modifying the single resource 
object crsres points to in crsbuf and resetting all the resources when we 
call _SRS?

Also, the handling of BUS_CONFIG_INTR() is giving me a heartache.  We should 
defer that until bus_alloc_resource() (I have a hack to do this now) as it 
causes the link devices currently to configure an interrupt for PIC mode that 
will _never_ be used if we are using the APIC.  However, it can do things 
like screw up the SCI if the SCI isn't using level/low.  Grr.

-- 
John Baldwin <jhb at FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve"  =  http://www.FreeBSD.org