Re: pci_intr_map(9) oddity

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

On Wed, Jan 26, 2011 at 10:29:31AM -0600, David Young wrote:
> (I'm not subscribed to port-xen, please CC me on replies.)
> 
> In my local tree, I made const the pci_attach_args argument to
> pci_intr_map(9).  Xen seems to be the lone port that writes to
> the pci_attach_args, any ideas why?  Before any PCI device is
> probed/attached, pci_probe_device() initialized pa_intrline.

I think this is a leftover from Xen-2. For xen2, a
HYPERVISOR_physdev_op(PHYSDEVOP_PCI_INITIALISE_DEVICE) was done in the
pci_intr_map() function, which, among other things, did the
interrupt routing and wrote the result to PCI_INTERRUPT_REG (or maybe it
did fake the PCI_INTERRUPT_REG, as all this did go through hypercalls).

I think it's not needed any more with Xen-3.

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--