>>>>> "Tim" == Tim Kelly <hockey@dialectronics.com> writes:

    Tim> Indeed, the fix turns out to be two lines.
    Tim> [...]
    Tim> [Wed Oct 13 01:31:17 basis pci]$ cvs diff -bu pci_machdep.c
    Tim> Index: pci_machdep.c
    Tim> ===================================================================
    Tim> RCS file: /cvsroot/src/sys/arch/macppc/pci/pci_machdep.c,v
    Tim> retrieving revision 1.24.2.1
    Tim> diff -b -u -r1.24.2.1 pci_machdep.c
    Tim> --- pci_machdep.c       9 Apr 2004 02:31:12 -0000       1.24.2.1
    Tim> +++ pci_machdep.c       12 Oct 2004 23:31:30 -0000
    Tim> @@ -363,6 +363,8 @@
    Tim>                      if (len != 4*ilen)
    Tim>                               continue;
    Tim>                      len = find_node_intr(node, &iaddr.phys_hi,irqs);
    Tim> +                    if (len == -1)
    Tim> +                       len=OF_getprop(node,"AAPL,interrupts",irqs,4);
    Tim>                      }
    Tim>                      if (len > 0) {
    Tim>                         intr = pci_conf_read(pc, tag,PCI_INTERRUPT_REG);

In addition for the good news: I did some (stress) testing of the
'new' ex0 dev in my beige g3, even with the *csum stuff turned on: Not
a tug of a problem so far, it's working absolutely flawlessly for me!

So, I have no objections against committing this patch. ;-)

Thanks for the great support!

Markus.