[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH] xen: fix interrupt routing

On Wed, 2011-05-18 at 18:53 +0100, Stefano Stabellini wrote:
> From: Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx>
> Match the routing informations built by seabios:
> - remove i440fx_write_config_xen
> we don't need to intercept pci config writes to i440FX;
> - introduce piix3_write_config_xen
> we do need to intercept pci config write to the PCI-ISA bridge to update
> the PCI link routing;
> - remove xen_pci_slot_get_pirq
> we are now using the same link routing as seabios and qemu so we don't
> need a diffirent get_pirq function;
> - fix xen_piix3_set_irq
> we always inject one of the 4 pci intx, so we can use
> xc_hvm_set_isa_irq_level to inject the interrupt. Use pci_irqs as
> initialized by seabios to map a pirq into an ISA irq. This has the
> benefit of removing all the calls to xc_hvm_set_pci_intx_level that
> doesn't work correctly anymore because from the same device number and
> intx Xen calculates a different PCI link compared to Qemu and Seabios.

Why does Xen calculate a different PCI link? What are the other
downsides of this disparity?

For example, in the hypervisor I see that hvm_pci_intx_link is used in
the passthrough code as well as in the HVMOP_set_pci_intx_level, is that
safe/sane in conjunction with this change?

IOW -- are we hiding other problems for later by doing an end run around
the interface instead of addressing the disconnect directly? It seems
like it would be a relatively contained change on the SeaBIOS end to
allow it to use the routing which Xen expects.


Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.